UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Epistemic progress in biology : a case study Ogden, Athena Dawn 2002

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2002-73224X.pdf [ 12.83MB ]
Metadata
JSON: 831-1.0090448.json
JSON-LD: 831-1.0090448-ld.json
RDF/XML (Pretty): 831-1.0090448-rdf.xml
RDF/JSON: 831-1.0090448-rdf.json
Turtle: 831-1.0090448-turtle.txt
N-Triples: 831-1.0090448-rdf-ntriples.txt
Original Record: 831-1.0090448-source.json
Full Text
831-1.0090448-fulltext.txt
Citation
831-1.0090448.ris

Full Text

EPISTEMIC PROGRESS IN BIOLOGY: A CASE STUDY by ATHENA D A W N OGDEN B.A. (First Class Honours), 1993 U N I V E R S I T Y O F B R I T I S H C O L U M B I A A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF GRADUATE STUDIES Department of Philosophy We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A January 2002 © Athena Dawn Ogden, 2002 In p resen t ing th is thesis in partial fu l f i lment of t h e requ i rements fo r an advanced degree at the Univers i ty of Brit ish C o l u m b i a , I agree that t h e Library shall make it f reely available f o r re ference and s tudy. I fu r ther agree that permiss ion f o r extensive c o p y i n g o f this thesis f o r scholar ly pu rposes may be g ran ted by the head o f my d e p a r t m e n t or by his o r her representat ives. It is u n d e r s t o o d that c o p y i n g o r pub l i ca t i on o f this thesis for f inancial gain shall n o t be a l l o w e d w i t h o u t m y w r i t t e n permiss ion . D e p a r t m e n t The Univers i ty o f Brit ish C o l u m b i a Vancouver , Canada DE-6 (2/88) ABSTRACT The aim of this dissertation is to explore the nature of scientific progress and to broaden existing theories of what constitutes progress in science. I do this by means of a close analysis of the main post-Kuhnian philosophical accounts of scientific progress, namely those put forward by Imre Lakatos, Larry Laudan and Philip Kitcher. I test these three accounts by reconstructing a series of scientific episodes in evolutionary ecology in terms of each account and then assessing the degree to which each account incorporates what is progressive. The episodes I have selected concern the resource competition research of Dolph Schluter on Galapagos finches and related work leading up to it. After distinguishing between macroscopic and microscopic levels in science, I attend carefully to the microscopic level of each episode as it relates to epistemic progress. This investigation demonstrates that some important aspects of scientific progress have been overlooked. I conclude that there are three main ways in which the philosophies of science surveyed do not adequately represent instances of scientific progress. First, the accumulation of factual knowledge is not well accommodated. Second, the role of evidence and argument in scientific theories is not adequately captured. Third, the fine-grained level at which much important epistemic progress in science occurs is often not accounted for. These criticisms relate to a more general tendency of contemporary philosophical accounts to emphasize the macroscopic level of entire research programmes and traditions while failing to attend to the microscopic level of progress inherent in a detailed case study. I end by offering a positive account of scientific progress in light of these criticisms. ii TABLE OF CONTENTS Abstract ii Table of Contents iii List of Figures vi Acknowledgements vii Dedication viii Chapter I Introduction 1 1.1 The Problem of Progress 2 1.2 Methodology 4 1.3 The Kuhnian Backdrop 7 1.4 The Philosophers 11 1.5 The Case Study 13 Chapter II Resource Competition in the Galapagos Finches 17 2.1 Background and Precursors to the Resource Competition Case 18 2.1.1 The Neo-Darwinian Evolutionary Synthesis 18 2.1.2 Speciation and the Genesis of Isolation Mechanisms 20 2.1.3 Character Displacement and Competitive Exclusion 21 2.1.4 Adaptive Radiation and Adaptive Landscapes 22 2.2 Overview of the Galapagos Finch Competition Case 27 2.3 David Lack 30 2.4 Robert Bowman 34 2.5 Abbott, Abbott and Grant 37 2.6 Stochastic Challenges to the Competition Hypothesis 43 2.6.1 Strong, Szyska and Simberloff 44 2.6.2 Conner and Simberloff 46 2.7 Schluter and Grant 1984 48 Chapter III Lakatos on Scientific Progress 63 3.1 Lakatos's Methodology of Research Programmes 63 3.2 Series of Theories 68 3.3 Scientific Progress and Progressive Problemshifts 70 3.4 What Does and Does Progress 77 3.5 Before and After Establishment of a Research Programme 79 3.5.1 Prior to a Research Programme 80 3.5.2 How Research Programmes Begin 82 3.6 Size and Scope of Established Research Programmes 85 iii Chapter IV Evaluating Lakatos's Account 89 4.1 Applicability of Lakatos's Methodology to the Resource Competition Case 90 4.2 Overview of the Case in Lakatosian Terms 93 4.3 Research Programmes and Progress in Lack's Work 94 4.3.1 The Larger Evolutionary Context 94 4.3.2 Lack's Programme and Hard Core Commitments 97 4.3.3 Character Displacement as a Kind of Adaptation 102 4.3.4 Other Lakatosian Elements in Lack's Work 103 4.3.5 Predictions and Progress in Lack 105 4.3.6 Lack's Programme: Empirical Progress for Neo-Darwinism 110 4.4 Competing Research Programmes 111 4.4.1 Bowman 113 4.5 Abbott, Abbott and Grant 117 4.6 Competing Programmes, Round Two: Three Stochastic Theorists 123 4.6.1 Strong, Szyska and Simberloff 124 4.6.2 Conner and Simberloff 129 4.6.3 Summary of Lakatos applied to the Stochastic Theorists 130 4.7 Schluter and Grant 132 4.7.1 Expected Densities 134 4.7.2 Five-Way Test 137 Chapter V Laudan on Scientific Progress 140 5.1 Introduction 140 5.2 Empirical Problems 142 5.3 Conceptual Problems 145 5.4 Weighting Problems 146 5.5 Problems, Theories and Progress 149 5.6 Research Traditions 154 5.7 Evaluating Research Traditions for their Progress 160 Chapter VI Evaluating Laudan's Account 164 6.1 Evolutionary Research Traditions 165 6.2 The Resource Competition Case: Tradition or Theory? 167 6.3 Problem-Solving Effectiveness in the Galapagos Finch Case 172 6.4 Lack 174 6.4.1 Number of Problems Solved by Lack 175 6.4.2 Weight of Problems Solved by Lack 177 6.4.3 Conceptual Problems and Anomalies in Lack 182 6.4.4 Problem-Solving Effectiveness in Lack 183 6.5 Competitors 184 6.6 Bowman 184 6.7 Abbott, Abbott and Grant 186 6.8 The Stochastic Challenge 191 iv 6.9 Schluter and Grant 195 6.10 Accumulation of Facts 197 Chapter VII Kitcher on Scientific Progress 199 7.1 Consensus Practice 200 7.2 Preliminaries 202 7.3 Conceptual Progress 203 7.4 Explanatory Progress 204 7.5 Erotetic Progress and Significant Questions 207 7.6 Other Kinds of Progress 211 7.7 Progress, Cumulativity and Significance 215 Chapter VIII Evaluating Kitcher's Account 224 8.1 Conceptual Progress in the Galapagos Finch Competition Case 225 8.2 Schematization and Explanatory Progress in the Finch Case 226 8.2.1 The Darwinian Schema 227 8.2.2 Schematizing Darwin and Lack on Competition 229 8.2.3 The Neo-Darwinian Schema 235 8.2.4 Schematizing Resource Competition and Floristic Adaptation 237 8.2.5 Schematizing Optimality and Adaptive Radiation 241 8.3 Lack's Contribution to Progress in Accepted Statements 244 8.4 Bowman's Contributions to Progress 247 8.5 The Contribution of Abbott, Abbott, and Grant 249 8.6 Progress applied to the Stochastic Theorists 250 8.7 Schluter and Grant 251 8.8 Significance and Cumulativity 260 Chapter IX Conclusions 263 9.1 Comparison of the Three Accounts 263 9.1.1 Laudan 264 9.1.2 Lakatos 267 9.1.3 Kitcher 268 9.2 Progress not Accommodated by these Three Philosophers 271 9.2.1 Accumulation of Factual Knowledge 272 9.2.2 The Role of Evidence and Argument 274 9.2.3 Fine-Grained Detail of the Finch Competition Controversy 276 9.3 Kuhnian Postscript 280 9.4 A Positive Account of Scientific Progress 282 Endnotes 287 Bibliography 295 v LIST OF FIGURES Figure Page 1 Simpsonian Adaptive Landscape 24 2 The Galapagos Ground Finches 29 3 Graph of Mean Seed Hardness by Log Beak Depth 52 4 Graph of Finch Biomass by Seed Biomass 53 5 Graph of Expected Population Densities 57 vi ACKNOWLEDGEMENTS I am very grateful to the members of my supervisory committee for all their hard work-Andrew Irvine, Alan Richardson, Paul Bartha and Stephen Straker. Thanks are also due to my university examiner, Sally Otto, whose criticisms increased the clarity of this dissertation. In particular I wish to thank my supervisor, Andrew Irvine, who always came through when it mattered most and whose gentle persistence has been a source of continued encouragement. For his patience and his tolerance I will always be grateful. Thank you to the scientists who consented to having me interview them-Dolph Schluter, John Pinel, Dennis Chitty, David Randall and Donald Wilkie. Thank you most of all to my friends for their support and patience. Thanks to Jack Webb, Elizabeth Negrave, Christine Adkins and Patrick Carrier for encouragement and support. Thank you to Alex Boston for cheery notes of encouragement and distraction. Thank you to Brad Murray for financial and moral support and intellectual stimulation during my undergraduate years. Thanks are also due to Mark Battersby for his encouragement over the years. Thank you especially to Kate Talmage, without whose emotional support, encouragement and advice this dissertation would not have been completed. A heartfelt thank you also to the Ogden family, who helped Mom and Dad in innumerable ways when I could not be there, especially Fred Ogden and Shirley Ogden Naso. Thank you to the UBC Faculty of Graduate Studies and to the Social Sciences and Research Council of Canada for financial support. Finally, a special thank you to J.S. Bach for the Goldberg aria, L. van Beethoven for the symphonies and piano sonatas, J. Brahms for the piano concertos, and S. Prokofiev (and Evgeny Kissin) for the Piano Sonata No. 6. vii DEDICATION To Kate Talmage and Jack Webb-for supererogatory acts of friendship. And in memory of Stanley Victor Ogden May 1940 - February 2002. viii Chapter I Introduction Which historiographical research programme is superior may be tested by seeing how successfully they explain scientific progress. (Lakatos and Zahar 1978, 192) The aim of this dissertation is to explore the nature of scientific progress in the context of an extended case study, thereby broadening existing theories of what constitutes progress in science. I do this by means of a close analysis of the views of three main contemporary philosophers of science, Imre Lakatos, Larry Laudan and Philip Kitcher, all of whom explicitly discuss scientific progress. I do so against a background of specific scientific episodes taken from recent work in evolutionary ecology. More specifically, a test of these philosophers' views on progress is provided by reviewing the research on the relative importance of resource adaptation, resource competition and stochastic effects in the evolutionary radiation of the Galapagos ground finches. The first main component of this dissertation is a chapter (Chapter II) detailing a series of scientific episodes beginning with David Lack's (1947) study of the ecology and biogeography of the Galapagos finches and ending with Dolph Schluter's computer model of the distribution of Galapagos ground finches given a number of observed facts about them. That chapter also contains additional background material designed to serve two 1 purposes: first, it will introduce the reader to some of the relevant biological background, and second, it will establish a larger context within the field of evolutionary biology in which these episodes occur. The next component consists of a detailed explication of the views of Imre Lakatos, Larry Laudan and Philip Kitcher. This explication occurs in Chapters III, V and VII. The central feature of this dissertation is then the evaluation of these views in light of the Chapter II case study. In this way it is possible to assess the ability of each account to accommodate the progressive aspects of the case at issue. When each philosophical account is examined in detail in the context of the Galapagos finch resource competition controversy, internal tensions often emerge. Finally, in Chapter IX, there is a comparison of these different philosophical frameworks and their ability to accommodate scientific progress. This is most instructive where they differ in their emphases and especially where they differ in the elements of the case that they are most able to accommodate. Also in the concluding chapter is a critical section discussing those kinds of progress that they are not capable of accommodating. I end by offering a positive account of scientific progress in light of the discussion up to this point. 1.1 The Problem of Progress This dissertation deals exclusively with epistemic progress. It is not about progress in terms of creating a world more hospitable to human life or desires. Similarly, it is not 2 about improvements in technological efficiency. It regards improvements in explanation and knowledge of facts about the natural world. Also at issue are the comparative abilities of three philosophical accounts of such progress to capture the fine-grained progress inherent in the Galapagos finch competition controversy. I attempt both to illuminate each philosopher's account with respect to his views regarding progress and to assess each in the context of this detailed case study. Each of the philosophers considered in this dissertation operates in a post-Kuhnian, and hence historically-sensitive, philosophical context. All three of their accounts purport to take seriously and attempt to incorporate actual scientific advancements. However, most theories of scientific change ... have not yet been extensively or systematically tested against the empirical record. The historical examples liberally scattered throughout the writings of these theorists are illustrative rather than probative and none is developed in sufficient detail to determine whether the analysis fits the case in hand. (Laudan et al. 1988, 5) The methodology here employed involves very close attention to the fine-grained detail of actual science. This kind of attention to scientific detail is intended to help fill a gap in the literature. Accordingly, I am here concerned to determine whether the accounts on offer live up to their aspirations to be relevant and applicable to real cases. The present 3 assessment can provide only a partial test, for it concerns a series of episodes in evolutionary ecology that may exhibit idiosyncrasies not shared by other sciences. While this limitation restricts the generality of the result, it is nevertheless the case that if epistemic progress occurs in the finch competition dispute, it should be accommodated by philosophical accounts purporting to provide an account of progress. 1.2 Methodology The methodology of this dissertation is empirical in the following sense. I reconstruct a series of episodes in an extended case study in terms of each of the philosophies of science being considered so as to determine the degree to which each philosophical account satisfactorily incorporates as much of what is significant to progress as possible. This procedure demonstrates how each philosopher's view is to be cashed out in the case of a real scientific episode. The procedure allows me both to detail progressive aspects demonstrated by the scientific episodes under scrutiny, particularly in relation to the philosophers of science examined, and to determine which of these aspects are problematic for the conceptual system of each philosopher. The methodology primarily involves a reflective application of the philosopher's views to the microstructure of the Galapagos finch competition controversy. Specifically, the source data used for the purpose of assessing each theory's ability to accommodate progress is past research into the Galapagos finch radiation. For simplicity, I am taking the episodes detailed in the case as factual, and will assume that concerns regarding the 4 theory-ladenness of observation do not apply to this data set. Although this is primarily a simplifying assumption, it is justifiable for another reason as well. A certain amount of theory-ladenness of one kind or another seems to be an irreducible factor in perception of any kind (reviewed in Gillies 1993, 132-149). I suggest therefore that there are more and less insidious kinds of theory-ladenness. Taking as the relevant data set the episodes comprising the extended case study, there are, for simplicity, three levels at which they might be infected by theory. There is first the perceptual level, the level at which we cannot perceive anything at all without experiencing it through our perceptual filters. Next there is the level at which the point is usually made: in science, even the simplest observation statements have the theoretical interpretation of them built right in. These two levels of theory-ladenness are accepted by most philosophers of science, and I assume they are at work here. However, the scientific episodes detailed in the case study might also be infected by meta-level theory. For present purposes, the worst instance of this kind of theory-ladennes would be for the data collected here to be tainted by prior notions about the nature of scientific progress. Accordingly, in explicating the Galapagos finch case in Chapter II, I attempt to present the scientific episodes without having in mind any account of scientific progress whatsoever. To the extent that the case might be considered theory-laden, I believe that at least it is contaminated very little by any prior notions of the nature of progress. In other words, the data may be theory laden at the level that all scientific observation is probably theory-laden, but I want to suggest that these data are contaminated minimally by prior 5 notions of what makes for progress in science. I will have more to say below about the selection of the case study itself. In the analytic chapters (Chapters IV, VI and VII), the reconstruction procedure involves provisionally accepting each philosopher's system and entering into his account as far as possible. To "extensively or systematically" test these accounts of scientific change "against the empirical record" as urged by Laudan et al. would seem necessarily to require an attention to the fine-grained detail of the scientific episodes under scrutiny. While entering into each philosopher's account in this way, I also take note of ambiguities of interpretation of the philosophical account in the process. The application of the three philosophers' accounts to the case at hand is also reflective in that the recasting of each episode in terms of each philosopher's historiography is interspersed with pro and con considerations regarding the best way to accomplish this recasting. At the same time, I adduce criticisms on the basis of what I perceive to be the best way of applying the given philosopher's position to the data set. As a result, this procedure is simultaneously a hermeneutical approach and a philosophico-critical approach. Thus, the methodology here employed is an attempt to incorporate the most valuable elements of both approaches. Laudan et al. (1988) suggest that the most appropriate way to test different philosophical historiographies as applied to scientific change is to reduce their commitments to a single common language. One of the reasons for doing this is that "most of the case studies to date attempt to apply some theory in toto to a particular episode, so that points of agreement and disagreement between the theory and the episode 6 are difficult to disentangle" (Laudan et al. 1988, 6). While this is at least partly true, another way to address the same difficulty would be to apply the rival historiographies to the same case study and to determine which best accommodates various aspects of the case. These authors come close to making this suggestion when they say that "thus far, virtually none of the 'tests' has attempted to assess the relative adequacy of rival theories of science, so that we simply do not know whether apparent confirmations of one theory might not also support theories that differ in very important respects" (ibid.). This latter difficulty can be addressed by doing a comparison of rival historiographies as they suggest here, but not necessarily in the way that they suggest. To the extent that Laudan et al. are committed to understanding philosophies of science as empirical theories of the practice of science, I am in agreement with them. However, I disagree with them in their employment of the hypothetic-deductive method in testing these alternative historiographies (cf. Richardson 1992). Instead, I am in favour of a more holistic or hermeneutical approach to comparison of alternative philosophies of science with one another in the presence of data consisting of episodes in the history of science. 1.3 The Kuhnian Backdrop As already suggested, the three philosophers whose views of scientific progress will be canvassed in this dissertation work in a post-Kuhnian context.1 This has implications for what their views are and for what they take the solved and unsolved problems of 7 scientific progress to be. Although I do not here assess what these philosophers owe to Kuhn, nor what they accept him as having demonstrated, it is worth bearing in mind that a case might easily be made for connections of these kinds. A potential criticism on the basis of what these philosophers assume as their post-Kuhnian mandate will be taken up in the concluding chapter. Accordingly, it is worth making explicit here some of Kuhn's views on scientific progress. Kuhn's historically-informed account of the development of science divides it into three main phases: pre-paradigm science, normal science and crisis (revolutionary) science. Pre-paradigm science does not play much of a role in his account; Kuhn is mainly interested in sciences that have reached the stage of commitment to a single or occasionally to two main paradigms, with periodic crisis interludes leading to the adoption of a new paradigm. In his "Postscript" (1970), Kuhn admits to two main uses of the term "paradigm" in The Structure of Scientific Revolutions: disciplinary matrix and exemplar. Exemplars are the "concrete problem-solutions that students encounter from the start of their scientific education, whether in laboratories, on examinations, or at the ends of chapters in science texts" (187). These are the examples that are shared among the practitioners of a scientific discipline, are agreed by that community to be exemplary results in its scientific domain, and are learned as part of the enculturation of scientists into their scientific subdisciplines. A "disciplinary matrix" is to be understood as "the constellation of group commitments," including, but not restricted to, the "formal or the readily formalizable components" used by the relevant community, shared metaphysical 8 commitments, values concerning the proper ways of doing science and what counts as an adequate experimental result, and, finally, exemplars as already described (181-85). At the broadest level, Kuhn attributes progress in science to the special kind of communities in which scientists participate, and to their acting in accordance with the values that they collectively endorse. However, progress is differently manifested in normal (paradigm-defined) science verses extraordinary science (during which the switch is made from former to new paradigm). In spite of its commitment to progress occurring in science over a series of paradigm shifts, the Kuhnian view must not be confused with the view of some prior philosophies of science that endorse convergent scientific realism or cumulativity. Although Kuhn warns not "to see [the whole history of] scientific development as a process of accretion" (3), he is, however, committed to some degree of cumulativity. He says, for instance, "though new paradigms seldom or never possess all the capabilities of their predecessors, they usually preserve a great deal of the most concrete parts of past achievement" (Kuhn 1970, 169). So it is clear that he is not committed to complete loss of the achievements of earlier paradigms after a scientific revolution has occurred.4 Given that Kuhn denies total accretion or cumulativity, this naturally enough entails that there must be explanatory "losses as well as gains in scientific revolutions" (167), although some of these losses have later been regained (148-9). Whereas extraordinary (or crisis) science is progressive but only partly cumulative, normal science appears to be both progressive and cumulative.5 This is due to the nature of normal science as a puzzle-solving endeavour. During normal science, 9 scientists set achievable goals relative to the current paradigm, specify what would count as solutions to them, and (usually) solve them (Kuhn 1970, Chapter IV). The three ways in which normal science has "success" are by "extending the knowledge of those facts that the paradigm displays as particularly revealing, by increasing the extent of the match between those facts and the paradigm's predictions, and by further articulation of the paradigm itself (24). These points correspond to the "three classes of problems" to be met with in normal science: "determination of significant fact, matching of fact with theory, and articulation of theory" (34). Presumably, although he does not say so explicitly, Kuhn would advocate progress corresponding to each of these classes of problems as it is clear that it is the puzzle-solving nature of normal science that makes it so successful: In its normal state, then, a scientific community is an immensely efficient instrument for solving the problems or puzzles that its paradigms define. Furthermore, the result of solving those problems must inevitably be progress. (Kuhn 1970, 166) So the solving of problems is what allows for progress, and the broad kinds of problems to be solved in normal science are these: determining the relevant facts, showing that the paradigm's predictions accord with those facts and further theoretical articulation. 10 1.4 The Philosophers In this dissertation I examine the accounts of scientific progress put forth by Imre Lakatos, Larry Laudan and Philip Kitcher. All of these philosophers examine epistemic progress. Laudan and Kitcher term it "cognitive progress" (Laudan 1977, 7; Kitcher 1993, 92), while it is clear that Lakatos's "theoretical" and "empirical" progress are both epistemic in nature. Writing after the historical turn in the philosophy of science, their views are both historically sensitive and characteristically philosophical, in that they concern themselves with issues such as rationality, epistemic progress and the demarcation between science and non-science. Lakatos is an intellectual heir of Popper, and Laudan an heir of Kuhn, so between them they continue two major philosophical views of scientific progress that have been dominant in the second half of the Twentieth century. Kitcher, in contrast, draws from all of the philosophers just mentioned, as well as others, but particularly from Kuhn. He also appends his own novel solutions to the problems of progress. Given that Kitcher succeeds these other philosophers, it might be expected that he eludes some of the difficulties encountered by the earlier views of Lakatos and Laudan. In particular, Kitcher explicitly attempts to accommodate the fine grain of scientific episodes in a way that Lakatos and Laudan do not. Together these three philosophers constitute the main post-Kuhnian philosophical accounts of progress. Because Kuhn's views are incorporated to some extent in the views of each of these three, and because they have more explicit and more extensively worked-out accounts of progress than Kuhn, Kuhn's account is omitted from the analyses in the upcoming chapters. However, whether these three post-Kuhnian philosophers take themselves to be primarily concerned with what Kuhn would call revolutionary science or with his normal science, or with some category that might cut across Kuhn's definitions, it is evident in the cases of Laudan and Kitcher at least that their intentions are to account for scientific progress to be found at any of these levels. As to progress within normal science, Laudan says, "I have claimed that any research tradition [his paradigm analogue] which can exemplify this process [i.e., problem solution] through time is a progressive one" (Laudan 1977, 124). As to the question of progress across paradigm shifts, Laudan first of all rejects the rigid distinction between Kuhnian normal and revolutionary science (Laudan 1977, 134). He suggests in any case that problem solving, his vehicle of scientific progress, can also illuminate "the nature of scientific revolutions" (121). Further, he claims that within-research tradition problem solving occurs under two circumstances that Kuhn associates with revolution (134). What is clear is that Laudan sees a role for problem-solving, and hence progress, in both of what Kuhn would consider normal and crisis science. Kitcher notes that his Darwinian example spans episodes that Kuhn would consider both normal and revolutionary (Kitcher 1993, 110). Moreover, he traces "something like cumulative progress" throughout this time frame (ibid.), so it is clear that he believes that at least some of the same kinds of progress can be demonstrated for both normal science and revolutionary science. Consistent with this, he also rejects what he takes to be Kuhn's commitment to "significant epistemological differences between the 12 course of science within a [paradigm] and the inter[paradigm] transitions" (Kitcher 1993, 87 footnote). The case of Lakatos is somewhat different. He refers to the continuity of series of theories that he says make up a research programme as "reminiscent of Kuhnian 'normal science'" (Lakatos 1970, 132). On the other hand, Lakatos's stated purpose does not accommodate Kuhnian normal science further than this. He says, The history of science has been and should be a history of competing research programmes (or, if you wish, 'paradigms'), but it has not been and must not become a succession ofperiods of normal science: the sooner the competition starts, the better for progress. (Lakatos 1970, 155) Within Lakatos's methodology, it makes sense to refer to progress at whatever level it is cogent to describe research programmes in competition with one another. As we will see in Chapters III and IV, it can be argued that Lakatos's methodology can be extended to apply at least sometimes to the relatively small grain of scientific change. His own example of Bohr's neutrino programme provides an instance of this. 1.5 The Case Study The case was initially chosen with few or no presuppositions regarding its progressiveness. In its selection I was guided only by the intrinsic interest inherent in 13 Schluter's work and by the knowledge that it is highly regarded by other scientists. For the latter reason, I assumed that upon further analysis the finch case would likely prove progressive, but I was not committed to selecting a case study on the basis of any prior notions of progress. I did not assume that any particular account of progress would apply to the competition case, nor did I assume that any particular kinds of progress would be evident. The methodology is an empirical-exploratory one in that I began with as few presuppositions regarding scientific progress as possible in an attempt to come to a fresh understanding of the nature of epistemic progress in science. Lakatos's philosophy of science takes as part of its subject matter the continuity of theories over time. So to provide a fair test of his views, it became apparent that the chosen case should be one with theoretical continuity over time. The science comprising the background context of Schluter and Grant's 1984 paper lent itself to this project. Thus the history of the relative importance of interspecific competition in the Galapagos finches beginning with David Lack (1947) became the relevant case study. There is no reason to assume that, in itself, the commitment to detailing a series of related episodes imports any presuppositions regarding progress. Lack's work, as we shall see, counts as a significant contribution both to knowledge of the ecology and biogeography of the Galapagos finches and as an important example of some of Darwin's theories in action. The next researcher, Bowman, adds extensively to the information regarding the Galapagos finches while disagreeing with Lack on some fundamental processes. Following him, Abbott, Abbott and Grant begin to determine the relative import of Lack's and Bowman's conflicting views. There follows 14 research on the importance of stochastic effects contradicting both Lack's and Bowman's results, but particularly the former. Finally, the relative import of the foregoing processes is partially assessed on the basis of Schluter's important mathematical reconceptualization of the interrelations between the various elements postulated by the earlier researchers as important to the distribution and morphology of the Galapagos ground finches. This result is regarded as significant by scientists working in the relevant fields as we shall see in the next chapter. That some of the scientific episodes detailed in Chapter II are admired by scientists working in the field provides guidance only in our quest for scientific progress, since it might be the case that what philosophers consider progress is not what scientists consider progress. If this were the case, philosophers of science would need to come up with principled reasons for such a disagreement. More importantly, an episode in science may be admired by scientists for any number of reasons, none of them, perhaps, relating to what philosophers would understand to be progress in any philosophically relevant sense. So we see that a case's being admired by scientists is neither a necessary nor a sufficient condition for its containing progress as understood by philosophers. As will be evident from the foregoing, a backward-looking approach to choosing the episodes and their emphases was employed. Schluter and Grant (1984), the paper in which Schluter's mathematization of adaptive landscapes is first introduced, looks at the relative importance of competition verses resource adaptation simpliciter. As a result, I consider only those aspects of earlier scientists' work that are relevant to the concepts of adaptation and competition. So part of the answer to a question of what was selected for 15 inclusion in the case study is that later concerns dictated which elements of earlier work are relevant. This might be criticized as exactly the kind of ahistorical science that historically-sensitive philosophers warn against. It selects for inclusion in the scientific narrative only those elements that will become relevant later, as though science had a sort of teleological structure. I want to suggest that this procedure is necessary in order to get a case study of manageable size, particularly one that deals in multiple scientific episodes. More importantly, I believe that it is possible in principle to follow theoretical threads without significantly altering their character.6 This is possible so long as one does not draw unjustified conclusions from seeing these theoretical threads thus stripped of some of their attendant context. It is always to be kept in mind that these focal theories are surrounded by other theories and data, even where the latter two are left out of the retelling due to space constraints. Finally, to serve as an accurate and sympathetic test of the three accounts of progress, the chosen case should be one in which some progress has occurred. Fortunately this turned out to be the case, as will emerge in Chapters IV, VI and VII. 16 Chapter II Resource Competition in the Galapagos Finches In this chapter I will describe the relevant research and scientific background concerning resource competition among Galapagos finches that culminated in Schluter and Grant (1984). Schluter and Grant's two main results, as well as research leading up to this work, provide the data that I will use to evaluate the three philosophical accounts of progress I wish to assess. I begin with a section detailing the relevant conceptual background for the series of episodes in the case study. First, I briefly detail the Darwinian and neo-Darwinian context that acts as a backdrop for all that follows. This background figures importantly in all of the analysis chapters. Similarly, the section about speciation and isolation mechanisms provides an understanding of where competition, as detailed in the following section, fits into the recent history of evolutionary biology. The concepts of adaptive radiation and adaptive landscapes, as detailed in the following subsection, also figure prominently in Schluter and Grant (1984). David Lack is the first member of the roster of players in the case study. His research identifies the initial problem and examines in greater detail than anyone prior to him the adaptive radiation in the Galapagos finches. Bowman then provides a largely contradictory point of view. Abbott et al., the next group of researchers in our series of scientific episodes, is committed to a conjunction of the accounts of these two scientists. 17 Subsequently some stochastic challenges are raised to the commitments of all of these workers. Finally we come to Schluter and Grant, whose (1984) paper has resolved the conflict, at least for the present. 2.1 Background and Precursors to the Resource Competition Case 2.1.1 The Neo-Darwinian Evolutionary Synthesis In situating Schluter's work within evolutionary theory more generally, I rely heavily upon the second edition of Futuyma (1998). This text is considered by many evolutionary biologists to be the most comprehensive single source; indeed, journal articles in evolutionary biology frequently quote Futuyma for basic definitions. In what Futuyma says is a convincing manner (Futuyma 1998, 21), Mayr argues for there being "a whole set of more or less independent theories" in The Origin of Species (Mayr 1982,426). By "independent" he seems to mean only that each can be viewed in isolation, not that they are in fact separate. The theories that he abstracts are (1) evolution as such, (2) evolution by common descent, (3) gradualness of evolution, (4) natural selection, y and (5) populational speciation. Briefly, (1) is just the thesis of evolution as opposed to stasis (Mayr 1982, 506). Futuyma understands thesis (5) in terms of change, not speciation: it is the theory that "evolution occurs by changes in the proportions of individuals within a population that differ in one or more hereditary characteristics • (Futuyma 1998, 22). The other theories are self-explanatory. Of the various theoretical commitments that can be 18 extracted from Darwin's The Origin of Species (1859), the "two major theses" are descent from a common ancestor and natural selection (Futuyma 1998, 21). There was a major reassessment of evolutionary biology during the 1930s and 1940s. This resulted in the so-called "new synthesis" or "modern synthesis." Futuyma lists twenty points that "In the aftermath of the Evolutionary Synthesis ... were widely accepted and became the orthodox modern theory of evolution," although some of these points have subsequently been challenged (Futuyma 1998, 26-7). Mayr is more concise about the commitments of the neo-Darwinian synthesis: a meeting of the minds came quite suddenly and completely in a period ... from 1936 to 1947... . [Bjiologists ... accepted two major conclusions: (1) that evolution is gradual, being explicatory in terms of small genetic changes and recombination and in terms of the ordering of this genetic variation by natural selection; and (2) that by introducing the population concept, by considering species as reproductively isolated aggregates of populations, and by analyzing the effect of ecological factors (niche occupation, competition, adaptive radiation) on diversity and on the origin of higher taxa, one can explain all evolutionary phenomena in a manner that is consistent both with the known genetic mechanisms and with the observational evidence of the naturalists. (Mayr 1982, 567) 19 These can be boiled down to the two most fundamental concepts in the synthesis, viz., natural selection8 and genetics. Simply put, genetically transmitted mutations provide the variation upon which natural selection acts (Futuyma 1998, 24; Hale 1995, 593). 2.1.2 Speciation and the Genesis of Isolating Mechanisms In 1942, Ernst Mayr defined the so-called "biological species concept" as follows: "Species are groups of actually or potentially interbreeding populations, which are reproductively isolated from other such groups" (Mayr 1942, 120). Although it is not the only definition of species, Mayr's biological species concept is now widely used by evolutionary biologists. Mayr hypothesizes that species arise in allopatry; that is, that speciation occurs between initially related populations when they become geographically separated. Allopatric speciation also entails that there is no gene flow between the two populations for enough time that they become reproductively isolated even if they were subsequently to occur in the same environment. The essential element of speciation in allopatry is that it occurs during geographic separation. All evolutionary biologists now accept that allopatric speciation has occurred (Futuyma 1998, 498). The other main theoretical kind of speciation is sympatric speciation, in which one population splits into two species while its members spatially coexist. More generally, the state of sympatry (as opposed to sympatric speciation) exists when two differentiable populations exist in the same locale. Sympatric speciation is highly controversial (Futuyma 1998, 499). After an original population has been separated into two subpopulations by some 20 geographic barrier, during which time they are reproductively isolated from each other, and then subsequently come into contact with each other again, individuals of one sub-population may or may not be able to mate and produce fertile or viable offspring with members of the other sub-population. This depends upon how different the two populations have become due to genetic drift and adaptation to their different environments. There are a variety of genetic changes that might have occurred during the time the two sub-populations had been separated and that would result in isolating mechanisms-i.e., barriers to gene flow between the populations arising specifically from various aspects of the organisms' biology (Futuyma 1998, 457). 2.1.3 Character Displacement and Competitive Exclusion Reproductive character displacement is "a pattern whereby characters that reduce mating between populations differ more where the two taxa are sympatric than where they are allopatric ..." (Futuyma 1998, 491). Character displacement is in evidence wherever "sympatric populations of two species differ more than allopatric populations in their use of food or other resources, and the difference may be reflected in their morphology" (Futuyma 1998, 259-260). This was a pattern first labeled with the term "character displacement" by Brown and Wilson in 1956 (ibid., 554), although Darwin himself recognized the role of competition in causing species to diverge (Weiner 1994, 141). "The competitive exclusion principle holds that two or more competing species that use exactly the same resources cannot coexist indefinitely" (Futuyma 1998, 75). One 21 possible result is that one of the two species will be driven to extinction (competitive exclusion).9 Another is character displacement. When biologists find a pattern of apparent character displacement the implication is that, in the past, there must have been competition between the two species for limited resources.10 If either species had not been present, the other species would have utilized these resources. However, because of the presence of another species that utilized the same resource, either species might be forced (via natural selection) to diverge genetically in its disposition to use these resources. There are various results. One species might utilize the resource, while the other does not, or does so considerably less, or the relevant morphological traits can diverge. So when a pattern of apparent character displacement is detected, this suggests that coevolution between competing species has occurred (Futuyma 1998, 555); i.e., it suggests that the existence of past competition for resources between the two species has forced them to "coevolve" in the sense of selection for adaptations to exploit different resources. Perhaps nowhere else has this pattern been so convincingly demonstrated as in the beak characteristics of the finches of the Galapagos Islands (Sulloway 1982, 1), to which I will return. 2.1.4 Adaptive Radiation and Adaptive Landscapes Evolutionary radiation is divergent evolution of numerous related lineages within a relatively short time;" accordingly, adaptive radiation is evolutionary radiation that has been driven by the adaptation of organisms to differing environments (Futuyma 1998,117). Radiation, in other words, refers to the divergence of an ancestral stock into more than one 22 species fairly quickly. Non-adaptive evolutionary radiation in turn constitutes genetic drift. Building on Sewall Wright's 1932 "adaptive topography, which is a surface of mean fitness plotted as a function of allele frequencies" (Hartl and Clark 1989, 166), George Gaylord Simpson introduced a heuristic tool to help explain adaptation and that has been much used in evolutionary biology ever since: adaptive landscapes. These three-dimensional landscapes are "pictorial representation[s] that graphically portray selection, structure [i.e., phenotype] and adaptation" (Simpson 1944, 89). A three-dimensional landscape can be represented in two dimensions much like a contour map, with so-called "adaptive peaks" indicating phenotypic traits of maximal fitness for a particular environment. The "adaptive valleys" represent attributes of the phenotype that are not well adapted to a given environment. Selection then pushes organisms' traits up the slope toward the adaptive peaks-toward local maxima in fitness for each phenotypic trait. Furthermore, this landscape is not thought of as static, but shifts as environments change over time (Simpson 1944, 89-92). Simpson diagrams organisms' characteristics tracking the relevant adaptive peaks through time (Simpson 1944, 204). In particular, he diagrams the shift in horse evolution from a browser adaptive peak to a grazer peak; here tooth shape is the morphological trait that adapts to the peaks (Simpson 1944, 92) (see figure l).1 1 Simpson's more metaphorical adaptive landscapes (in which two different environments are represented on one graph) can be employed here to make more intelligible both adaptive radiations and character displacement due to coevolution. Essentially, competition changes the adaptive landscape so that what would have been an adaptive peak 23 T A X O N O M Y Egwrne ^?rf^ Ifyracoftsriinae Fig. 1. Simpsonian adaptive landscape showing the movement of a lineage from one adaptive peak, through a divergence of the ancestral population into two, so that each descendant population colonizes its own adaptive peak. Represented is an equid (horse) phenotype moving from a browsing adaptive peak (B) to a grazing adaptive peak (G). Reproduced from G.G. Simpson, Tempo and Mode in Evolution (1944), 92. 24 for an organism, had its competitor not shared the same environment, is no longer, and so it is forced to "colonize" another adaptive peak-i.e., to utilize a different resource and to change, via natural selection, in order to do so. Thus at least one of its phenotypic characters has been "displaced", forced toward a different adaptive peak, by the presence of a competitor for the same resources. Furthermore, there tends to be a set minimum distancebetween these peaks with fitness valleys between them due to the intensity of competition that occurs when the same trait is most similar in competing individuals. Adaptive radiations, as we have seen, were no more than evolutionary radiations that were caused by adaptations to different environments. Differing environments can be represented metaphorically by different adaptive peaks. In one environment, a phenotypic trait, such as a particular leg length, will be more fit. For example, we could represent leg length in this environment by an adaptive surface, with the most fit leg length at the top. So an environment in which leg length, x, is maximally adaptive can be represented by one adaptive peak in this (Simpsonian) landscape, while an environment in which leg length, y, is maximally adaptive is represented by a different peak in the same landscape. An adaptive radiation (due to character displacement) in the leg length of these two species-assuming that leg length were positively correlated with ability to exploit the resource-would then track the first species' leg length as it ascends one peak, and the other species' leg length as it ascends the other. One can also think of the adaptive forces that are at work on a species as competition that changes the landscape. If one peak is already "colonized" by one species, that peak is no longer an adaptive peak as far as the other species is concerned. The reason is that the 25 second cannot improve its fitness by "climbing" this peak since that resource is already exploited at close to capacity by the first species. As alluded to earlier, the mere existence of evolutionary radiation is insufficient evidence to conclude that the radiation is adaptive. Accordingly, to assert that a group of species exhibits "adaptive radiation" is to take a stand about the processes that produced the members of that group. Schluter defines adaptive radiation as the evolution of ecological and phenotypic diversity within a rapidly multiplying lineage. It occurs when a single ancestor diverges into a host of species that use a variety of environments and that differ in traits used to exploit those environments. (Schluter 2000, 2) Ecological diversity here just means diversity in the distribution of organisms. Schluter presents these conditions as those that are necessary to be fulfilled in order for a radiation to be adaptive. He adds that this does not take into account whether divergent selection was the cause of the radiation; this needs to be assessed separately, and is made explicit in the third possibility, directly below: Three main causes of adaptive radiation have been proposed: (1) "phenotypic differentiation between populations and species caused directly by differences in the environments they inhabit and the resources they consume" (2) "divergence in phenotype resulting from resource competition" 26 (3) "ecological speciation, whereby new species arise in adaptive radiation by the same processes that drive differentiation of phenotypes, namely divergent natural selection stemming from environment and resource competition" (Schluter 2000, 5). The importance of (1) lies in the resources and environments that the organisms adapt to; the stress here is on differential adaptation to these. In (2) the emphasis is on inter-specific competition. Finally (3) is opposed to "many non-ecological processes [that] also lead to speciation" (Schluter 2000, 5) such as genetic drift; here the emphasis is on speciation. 2.2 Overview of the Galapagos Finch Resource Competition Case It is plausible to understand Lack (1947) as initiating the current case, at least as I am delineating it. The case then terminates in Schluter's mathematical modeling and testing of the Galapagos finch data with respect to adaptive landscapes. It is worth pausing here to present a short (non-comprehensive) summary of the case study. Lack provides the first hypothesis with respect to the biogeography of Darwin's finches, namely that reproductive isolation began in allopatry but was completed in sympatry due to competition between incipient species for food resources resulting in exaggeration of their differences (Grant 1981, 654, 656). Bowman, by contrast, argues that the particular morphologies of the plants (and hence food availability) of each island were sufficient by themselves to explain the adaptive radiation of the finches; competition leading to exaggeration of species' differences was not necessary to explain those 27 differences because adaptation to differing environments as provided by particular plant morphologies would be sufficient by itself to explain the observed differences between the finches (Grant 1981, 656). Neither of these authors tested their hypotheses (ibid). In 1973, Abbott et al. (including Grant) initiated a field study of the genus Geospiza, the granivorous ground finches (see figure 2), on eight islands of the archipelago. The study was designed to decide between the two hypotheses or to reconcile them (Grant 1981, 656). The researchers found that Bowman's hypothesis was supported as the principle mechanism leading to the biogeography of the finch species; however, there was an irregularity suggesting that diversity of available food was not the only element influencing the finch radiation (Grant 1981, 657). Consequently, Abbott et al. argued that the observed speciation of the grain-eating Galapagos finches was due both to floral distribution on the islands (Bowman's so-called "floristic hypothesis") and to competition, with emphasis on the former. Subsequently others challenged the necessity of invoking competition at all. Simberloff, Conner, Strong and Szyska in various papers "showed that morphological and distributional patterns of the finches do not differ significantly from patterns in hypothetical communities produced by randomly combining species" (Schluter and Grant 1984,175). Finally, Schluter argued that competition was necessary in addition to floristic diversity to give an account of finch speciation in the Galapagos. He did this by producing, "for each island ... a curve giving the expected population density of solitary finch species as a function of its mean phenotype" (Schluter and Grant 1984, 176), and then demonstrating that this curve could be explained only by recourse both to competition and adaptation to food resources. 28 Fjfr 2 ; T h / e n G ^ l d P a g o s ground finches. (1) Geospizafortis. (2) G. magnirostris. (3) G T U O , r 7 f u l l 8 l n o s a - ( 5) G- conirostris. (6) G. sccmdens. Reproduced from J. Weiner / he Beak of the Finch (1994), 41. 29 To reiterate, Abbott, Grant and Schluter all argue that the granivorous Galapagos finches evolved by adapting to the local food sources (as provided by the plants present) in combination with intra-species competition. Bowman argued that competition need not be invoked because diversity of plant life was sufficient to account for all evolved differences in the finches. Lack was the only one to espouse competition, but not adaptation to local environmental factors, as contributing to the morphology of the finches. Simberloff and Strong (primarily) argued against the competition hypothesis by suggesting that chance could account for the elements in the finch distribution that seemed anomalous on the hypothesis that adaptation to food resources alone led to the morphology we see in today's Galapagos finches. Most of the work of the Grants has focused on the ground finches, genus Geospiza. This genus forms one branch of the radiation of finches on the Galapagos islands according to both the phylogeny Lack originally proposed in 1947, and Peter Grant's later adoption of it (Grant 1986, 11). 2.3 D a v i d L a c k In his explanation for the biogeography of the finches of the Galapagos Archipelago (the so-called "Darwin's finches"), David Lack (1947) emphasizes the effects of initial geographic barriers followed by competition for food resources. He holds that allopatric speciation in geographic isolation was followed by subsequent contact of the incipient species, that 30 resulted in larger morphological differences due to competition for resources. This model was not novel to Lack, although he presented it in the most detail. What Grant (1981) calls the "Darwin-Stresemarrn-Lack model" is the four-step allopatric model of speciation in the Galapagos finches. It is encapsulated in the following passage. I consider that the adaptive radiation of Darwin's finches can have come about only through the repeated differentiation of geographical forms, that later met and became established in the same region, that this in turn led to subdivision of the food supply and habitats, and then to an increased restriction in ecology and specialization in structure of each form. (Lack 1947, 149) The four steps of the allopatric model as applied to the finches are as follows: (1) "a single over-water colonization of the Galapagos archipelago" by one initial finch species; (2) dispersal to other islands so that allopatric populations of the same initial species are formed; (3) "[secondary contact between original and derived populations," during which there is interbreeding between sufficiently similar populations and between less similar populations, where there is enhancement of isolating mechanisms, and character displacement; and (4) a repetition of the other steps (Grant 1981, 654-5). Grant reports that this four-step model is "remarkably faithful to Darwin's general reasoning (1859)" (Grant 1981, 654); however, it appears that the allopatric model of speciation was first emphasized 31 by Mayr (1963) (Futuyma 1998, 482), even though it was also present in rudimentary form in Darwin. It should be noted that responses to competition are, of course, a kind of adaptation-they are an adaptation to the existence of competing species in the organisms' environment, that Darwin himself recognized as a factor in natural selection (1859, quoted in Lack 1947, 115). So we see that the present case study has its recent beginning with Lack, who combined questions regarding the origin of the Galapagos finch fauna with the notion, due to Darwin (1859) and Gause (1934), of resource competition leading to diversification and speciation in the sense of morphological gaps between species. Lack agreed with Darwin that "physical and climatic conditions are very similar on the various Galapagos islands, so there is no reason to think that the differences between island forms are correlated with differences in the physical conditions to which they are subjected" (Lack 1947, 117). At the same time, in a qualified way, Lack postulates adaptation as a mechanism involved in the finch radiation. Although Lack notes that "beak differences between most of the subgenera of Darwin's finches are clearly adapted to differences in feeding methods, the same does not seem to hold for the beak differences between closely related species"-that includes the granivorous ground finches and two species of insectivorous tree-finch (Lack 1947, 60-61). Accordingly, he had to come up with another explanation for the morphological variation in those finches. In the case of the three granivorous ground finches, G. magnirostris, G.fortis and G. fuliginosa (61), that are found "together in the same habitat on the same Galapagos islands" (62), and in the case of the two sympatric 32 species of insectivorous tree-finch, resource competition (specifically competitive exclusion) was the mechanism that had led to their morphological differences (62-4). Lack (1947) also postulates genetic drift, or the "Sewall Wright effect," as governing some of the morphological characteristics of the finches, but only in the cases of G. fortis on the island of Daphne, G. fuliginosa on Crossman, and G. conirostris on Tower (now Genovesa) (122-4). To summarize briefly, Lack is committed to adaptation to food resources as a factor influencing the morphology (particularly with regard to body size and beaks) of most of the Galapagos finches. However, for a couple of species he postulates that genetic drift effects may have been operative. For other species, and in particular three sympatrically-occurring ground finches, he suggests that the distribution that he saw was due to competitive effects. Lack is recognized as demonstrating that "when certain species of seed-eating finches co-occurred on the same island, they were more divergent in beak depth than when they occurred alone; species with very similar beaks rarely co-occurred" (Givnish 1997, 4). Lack's (1947) book has been described as "pivotal," and it "probably did more to bring adaptive radiation-and ecological thinking generally-into the Modern Synthesis than any other single work" (Givnish 1997, 4). This was in spite of the fact that Lack did not explicitly attempt to generalize his (1947) results beyond the Galapagos finches. Although he did not do this, it appears that others extrapolated his results beyond their application to the finches. Lack's book "supplies abundant evidence for considering these birds ... as a classic paradigm of evolution and adaptive radiation in action" (Sulloway 1982, 45). Even so, Lack's thesis of the adaptive character of the radiation seen in the finches on the 33 Galapagos Archipelago was controversial and has since been criticized in a number of ways. Lack demonstrated that the distribution of finches on the Galapagos islands was consistent with competitive exclusion and character displacement, taking this as evidence that these processes had formed the fauna. However, because the events shaping the present finch distribution had occurred in the past, they were undemonstrable (Weiner 1994, 145). Therefore, critics could adduce a number of other possibilities to account for the pattern of species colonization seen in the archipelago. In other words, Lack had not given sufficient evidence that the biogeography he witnessed in the finch fauna of the Galapagos must have arisen due to competition (and the resulting selection) between species upon secondary contact. Lack demonstrates a guarded acceptance of the applicability of the heritability of phenotypic traits: "It is now generally agreed that in animals the differences between geographical races of the same species are hereditary, though in Darwin's finches, as in most other birds, experimental proof of this statement is lacking" (Lack 1947, 117). This can be taken as evidence that Lack accepted the synthetic evolutionary theory that had congealed in the 1930s and 1940s (Futuyma 1998, 24). 2.4 Robert Bowman Robert Bowman carried out an extensive survey of the Galapagos finches, including extremely detailed measurements of their skulls and facial musculature; he also observed the food preferences of live finches on the island of Indefatigable (now Santa Cruz). 34 Contra Lack, Bowman suggests that it is unnecessary to postulate a role for competition in the divergence of the finch species, because the islands are sufficiently different in their respective floras (and hence the food supply provided by that flora) for one to postulate adaptation to different environments as the primary cause of the variety of the Galapagos finch morphologies. He states that examples would suggest that interisland variations in size and shape of bill within species of the Geospizinae are correlated with differences in their feeding habits. Presumably, if more behavioral information were available, most and possibly all of the intra- and interisland variations in bill structure of the Geospizinae could be related to feeding. (Bowman 1961, 268) In a six-page section, Bowman addresses many of the passages in Lack in which the latter argues for the occurrence of competition. Bowman says, It is implied in Lack's remarks that there is a "competition" between G. fortis and G. magnirostris for certain foods, and that the intensity of the competition is related to the size of the populations of the competing species.... I believe that the conditions described may be interpreted in a different way, namely, that bill variation and population density are above all related to the availability and diversity of the food. (Bowman 1961, 273) 35 Bowman explicitly contrasts his own "floristics hypothesis" (as Abbott et al. 1977 dub it) with Lack's competition hypothesis. Further, Bowman suggests that "since there is no direct evidence that competition is occurring at the present time, I see no logical reason to assume that it must have occurred in the past" (Bowman 1961, 275). He also argues for the colonization of the islands being due to finch species surviving when they reach an island with food resources to which their beaks are more-or-less suited, modified by adaptation to local resources. What is clear throughout is that Bowman strongly believes that competition is not needed to explain the finch distribution on the Galapagos. Bowman therefore emphasizes (1) the adaptations of the finches to aspects of the plant life of the islands and in particular to the food that the plants provided. In addition, he suggests that both (2) the predators present there14 and (3) the "genetic constitution of the ancestral colonists from the mainland" are important forces shaping evolutionary development (Bowman 1961, 292). He adds that "the failure of certain food specialists to have evolved, even with suitable niches available, is probably due to the genetic inability of the geospizines to evolve them, and not necessarily to the presence of 'ecological equivalents' from the mainland" (ibid.). Here, too, he is implicitly arguing against Lack's competition hypothesis, given that Lack holds that the "absence of other land birds has had a most important influence on the evolution of Darwin's finches, since it has allowed them to evolve in directions which otherwise would have been closed to them" (Lack 1947, 113). This is a reference to the competitive exclusion hypothesis. Bowman also argues against Lack's contention (mainly in Lack's 1945 paper) that drift was responsible "for differences in island forms of the same species" (Bowman 1961, 36 269). Once again, Bowman's emphasis is on adaptation to local, floristic, conditions to account for the morphology of the finches. 2.5 Abbott, Abbott and Grant Abbott et al. (1977) state that prior to their own work, "The [finch] radiation has been inferred almost entirely from morphological evidence and from the distribution of the species among the islands" (Abbott et al. 1977, 151). Neither Lack's nor Bowman's main hypotheses were directly tested by their authors, although both of them adduce theoretical reasons and circumstantial evidence to defend their hypotheses. In 1973, Abbott, Abbott and Grant set out to conduct field studies in the Galapagos in order to gather data which would either decide between, or reconcile, the two hypotheses (Grant 1981, 656). Peter and Rosemary Grant and their collaborators have been conducting field studies there ever since-banding, measuring, observing food preferences, and tracing the genealogies of every bird on some of the islands. Their work "is one of the most intensive and valuable animal studies ever conducted in the wild; zoologist and evolutionists already regard it as a classic. It is the best and most detailed demonstration to date of the power of Darwin's process" (Weiner 1994, 9). Furthermore, their observations and measurements have been so thorough and painstaking that they and their collaborators have managed to predict evolutionary changes resulting from selection events (Greenwood 1993, 699), something that critics of evolutionary theory have often claimed was impossible-and until recently it was. One of Grant's most significant contributions has been the masses of data collected about this 37 group-the most extensive natural history study ever conducted. Although Peter Grant is a major player in studies of the natural history and radiation of the Galapagos Geospizae, his main contributions for present purposes were in the collaborative work of Abbott et al. The Galapagos finch field study which began in 1973 culminated in Abbott et al. (1977). In the introduction to this paper, they present the conflict between Lack and Bowman as primarily addressing the question of competition. They say, "Our ignorance of the feeding ecology of these finches has allowed a conflict of views to persist on the importance of interspecific competition in the adaptive radiation" (Abbott et al. 1977, 152). It is noteworthy that they put the case in this way, as they could equally have presented the conflict of views between these previous writers as over whether the floristic diversity of the islands primarily determined the finch distribution. Abbott et al. admit to exaggerating the contrast between Lack and Bowman in their article: "The Lack and Bowman theories are largely nonoverlapping but complement each other to a small extent.15 Yet... we have deliberately contrasted them where possible..." (Abbott et al. 1977, 153). Indeed, they implicitly accept the hypothetico-deductive method: "The competition theory of Lack and the floristics theory of Bowman were investigated by testing their explicit hypotheses and by testing predictions we have made from their remaining arguments" (Abbott et al. 1977, 153). Furthermore, they employ a loose form of crucial experiment where they decide between the implications of the two theories by teasing out the observable implications of the theories' corollaries and then determining which consequence obtains: they "have allowed [their] data to 'choose' between" the two hypotheses (Abbott et al. 1977, 153). 38 The Abbott et al. paper notes widespread agreement in the literature of the time with Lack's competition hypothesis and slightly more limited, though also widespread disagreement, with Bowman's floristic emphasis (Abbott et al. 1977, 153). Abbott et al. have collected data and analyzed it with respect to the two hypotheses with exemplary thoroughness. It would be excessive to restate all of their evidence and arguments here, so I will instead mention only a few important arguments for each hypothesis. Abbott et al. made a direct test of the competition hypothesis by looking for correlations between "the diversity of seeds and fruits ... and ... the absolute abundance of all other sympatric Geospiza species" but found none (Abbott et al. 1977, 158). Then they tested Bowman's floristic hypothesis directly by "seeking a relation between the diversity of seeds and fruits eaten ... by each of 21 Geospiza populations and the diversity of seeds and fruits ... available at each site" and found a "significant positive correlation" supporting Bowman's hypothesis (Abbott et al. 1977, 158-9). Although even the early results provide ample evidence supporting Bowman's hypothesis (Grant 1981, 656), there are "several regularities in the distribution and morphology of the finches that are predicted by a hypothesis of interspecific competition for food" (Grant 1981, 657). One of the most striking examples of this was that "study of the distributions of beak depths in specimens from eight island sites reveals a regular pattern of spacing compatible with a hypothesis of interspecific competition for food" (ibid.). In other words, although there is a positive correlation between beak size and the diet that is provided by the plant life on those islands, "there is not a one-to-one relationship between peaks in the frequency distributions of beak sizes of sympatric species and peaks in the 39 frequency distributions of seed and fruit size and hardness" (ibid.). Furthermore, "Since the species not occurring together are very similar in morphology and ecology ... there is a strong implication that one of the missing species has been present but has gone extinct for reasons of competitive inferiority" (Abbott et al. 1977, 176). "[B]y this indirect argument we infer that some morphological and ecological character displacement probably occurred . . . although we cannot say in which particular cases" (Abbott et al. 1977, 176). In their 1977 paper, there are two sub-sections that are particularly relevant to both the relative importance of Lack's and Bowman's main hypotheses, and to subsequent researchers: On the competition hypothesis, closely related species wi l l persist on the same island i f their diets differ "sufficiently." Differences in beak depth are thus seen as adaptations for taking foods of different size and hardness, hence sympatric finch species should exhibit more pronounced differences in beak depth than allopatric species. (Abbott et al., 1977,163-4) These are both implications of Lack's competition hypothesis. In contrast, "The floristic hypothesis states that interspecific differences in beak depth are caused by populations adapting to relatively constant interisland differences in diversity and availability of food" (Abbott etal. 1977,164). " A testable prediction of the competition hypothesis is that the more different in beak depth two sympatric Geospiza species are, the less they should overlap in diet" 40 (Abbott et al. 1977, 164). They tested this by evaluating overlap in observed food preferences of sympatric pairs of species against the ratio of the beak depths of the same two species in each case (ibid.). The resulting correlations clearly show that a large difference in beak morphology is associated with a large difference in diet. This is in accordance with Lack's reasoning that morphological differences between sympatric congeners16 are adaptive in the sense that they allow them to use different resources and hence permit their continued coexistence. (Abbott et al. 1977, 164) Furthermore, "A corollary of the competition hypothesis is that islands with only two breeding sympatric Geospiza species should hold two such species of very dissimilar beak depth," and this is indeed what is seen (ibid.). On the other hand, Bowman's hypothesis suggests that "overlap between sympatric pairs... . would tend to be highest when the lowest variety of foods is available" (164). This is testable by "relating overlap in the seed and fruit types eaten by each sympatric pair of Geospiza species to the diversity of seeds and fruits available" (ibid.). Here the result was that there was no significant correlation, thus counting against Bowman's hypothesis in this instance. When competition is assumed in the Galapagos Geospizae, what should be seen is that the beak depth of one species will influence the beak depth of sympatric congeners (164). This consequence of competition can be tested in one of two ways. First, in a 41 comparison of the (mean) shallowest and next-shallowest (mean) beak pairs at a variety of sites, the competition hypothesis would predict that there would be a demonstrable influence, that could be measured by correlations (164-5). This was not found, so in this test, Lack's hypothesis was not confirmed. Second, the same relationship should hold between the deepest mean beaked species and the next deepest. Again, a significant correlation was not found (165). A prediction from the "floristic hypothesis" regarding the same data sets "is that inter-island differences in beak depth should relate to interisland differences in food supply" (165). Two comparisons were made. First, the availability of plant species producing small, soft seeds was compared with the existence of small beaks at the same sites. These small seeds should set a mean lower limit for the size of the beaks of finches in the vicinity. This was not found (ibid.). Similarly, the second assumption was that the "abundance of large, hard seeds and fruits should set an upper limit to mean beak depth of the largest-beaked breeding species... . The expectation is realized" (165). Here Bowman's hypothesis is supported. One way in which the foregoing detailed studies are relevant to our purposes is in demonstrating the kind and variety of tests that were made. A second importance of these details is to demonstrate the variability of the support that was and was not given by the data to the two competing hypotheses. In the final analysis, this group proposed a role for both the differences in flora across the islands and interspecific competition in shaping the speciation and distribution of studied finch species on the islands (Abbott et al. 1977, 175). However, they add that "Allocating relative importance to these two classes of factors is difficult" for a number of 42 reasons; "First, they interact" (Abbott et al. 1977, 175). As far as the finch colonization and speciation in the archipelago goes, they accept Lack's two-part allopatric model; they say, "The allopatric model (Lack 1947) seems the most likely descriptor of the speciation process" (Abbott et al. 1977, 175). From their data and analysis they conclude that "Interisland variation in vegetation favoured the initial steps of differentiation. Competitive interactions among species influenced later stages by determining which ecological types could coexist on an island with a given array of foods" (Abbott et al. 1977, 151). In the end, the Abbotts and Grant came up with (strongly) suggestive patterns of competitive exclusion and character displacement in the evolutionary history of the finches; however, "Peculiarities of the local food supply by themselves may have led to the same phenomenon, in which case character displacement and/or differential colonization have been mimicked by noncompetitive processes" (Grant 1986, 332). So their results, although they were consistent with processes caused by competition, were still non-definitive and left open the possibility of dissent. 2.6 Stochastic Challenges to the Competition Hypothesis Schluter and Grant (1984) represent the relevant controversy between the competition hypothesis of Lack and the "floristic" hypotheses of Bowman (with additional considerations given by Abbott et al.). They also describe later arguments against both of these hypotheses (Schluter and Grant 1984,175). Of these, they are most interested in the arguments against competition. They take the (most) significant arguments against 43 competition to be those made by Strong et al. (1979), Simberloff (1978), and Conner and Simberloff (1978) (Schluter and Grant 1984, 175). Because Simberloff (1978) does not so much as mention Lack or Bowman, I will only touch on his work here before detailing the positions held by the others. Simberloff addresses the claim that "if competitive interactions did not exist most published biogeographic distributions would be similar to those now observed" (Simberloff 1978, 715). His is an example of "the more traditional methods of comparing populations in sympatry with those in allopatry ... [in] island communities" (Schluter and Grant 1984, 176). He examines (and argues against) competitive exclusion rather than character displacement, and concludes that "a model of colonization ... which is purely stochastic and rests only on properties of individual species comes close to accord with some plant and insect data and can in any event be used as a baseline test to see if other phenomena must be assumed" (Simberloff 1978, 724). 2.6.1 Strong, Szyska and Simberloff The two commitments of Abbott et al. that Strong et al. take themselves to be challenging are the existence of character displacement in the Galapagos ground finches (Strong et al. 1979, 900), and "that there is greater character displacement on islands with fewer Geospiza species" (901). They note that "Abbott et al. (1977) do not compare actual communities to null communities which have not been shaped by competition caused by the present coexistence" (Strong et al. 1979, 900). In order for character displacement to be 44 demonstrated with respect to beak length, two species existing together must show a ratio of beak sizes greater than would be expected by a chance selection of species. Again, the theory behind character displacement is that competition has forced two sympatric or adjacent species to be more unlike each other in some trait than they would be i f character displacement due to competition were not in operation. Therefore a reasonable assumption is that the average of beak length ratios of sympatric species chosen at random from a pool of all the finch species of the Galapagos archipelago should be significantly smaller than the actual ratios of species in which character displacement has occurred. There are a few subtly different ways of analyzing the resultant ratios to determine whether the Abbott et al. competition and character displacement hypothesis is supported, but basically what Strong et al. found was that the number of observed ratios which are higher than would be expected by chance "does not lead to rejection of the null hypothesis," i.e., that the ratios observed are due to chance alone (Strong et al. 1979, 901). This leaves open the possibility that the ratios of those sympatric species which are in fact higher than the ratio expected by chance may be thus due to character displacement; however, the trend in the data is such that character displacement is ruled out as a general phenomenon governing all of the finch species studied. On the other hand, the researchers found "some evidence for character convergence" that they note is "consistent with Bowman's (1961) interpretation of intraspecification variation in Galapagos finches" (Strong et al. 1979, 907). In another section of their paper, they present their own position as an alternative to Abbott et al. (1977), whose position they represent as being that "food supply and interspecific competition have jointly determined the patterns of evolution and species 45 diversity of the Galapagos finches" (Strong et al. 1979, 910). Instead, they suggest that "only a portion of the evolution and ecology of the finches has been determined by these factors" (ibid.). Their point in putting forth and testing their null hypothesis is not to suggest that random factors are the only causes (if non-deterministic processes can be considered causes) operant in shaping the Galapagos and other avifauna they consider. Instead, they suggest that "it is reasonable that apparent randomness be disproved before more structured or ordered versions of ecological nature are accepted as true" and furthermore that it is likely to be one of the many complex interactive factors influencing the development of any particular fauna (ibid.). They conclude that "Our results do not disprove interspecific competition. Rather, they show by use of null hypotheses that community-wide character displacement, which is indirect evidence of competition, is not as common in nature as believed" (Strong et al. 1979, 910). 2.6.2 Conner and Simberloff Conner and Simberloff present "Two null hypotheses concerning the determination of floral and avifaunal compositional similarities among the Galapagos Islands" (Conner and Simberloff 1978, 219), and do a pairwise comparison of which species one would expect by chance, compared to which species are actually present on adjacent islands. Their first null hypothesis assumes that all species have equiprobable persistence and dispersal abilities; the second does not assume this and so is, they claim, more realistic. Both hypotheses are geared toward exploring "compositional similarities" of colonizing species (ibid.). 46 The emphasis of the research question that Conner and Simberloff examine is slightly different than that of either Lack's or Bowman's: Conner and Simberloff ask why there are similarities in the species present on one Galapagos island compared to the next; Bowman and Lack are looking for the causes not only of the replication of species across islands, but also why some islands are missing some species. Furthermore, because Conner and Simberloff compare pairs of species only on adjacent islands, this means that character displacement is not one of the things they tested via their methods.17 Distribution of species on adjacent islands does, however, relate to competitive exclusion (although Abbott et al. do not test for this) and especially to adaptation to local flora. Significantly, their analysis includes all bird species, not just the Geospizae, as well as the flora recorded on the islands. Null Hypothesis II was marginally better for the plant species and genera studied (Conner and Simberloff 1978, 242). So we see that "a substantial proportion of the number of species shared between 2 islands can be viewed as resulting from stochastic processes of persistence and dispersal" (244). However, according to their application of the first null hypothesis, there is a statistically significant difference between the observed distribution of bird species and one expected by purely stochastic processes 31.4% of the time (244). For the second null hypothesis, it is 69.5% (ibid.). In other words, if we were to take their results at face value, the more realistic model yields the conclusion that something other than stochastic processes account for the species distribution on adjacent islands almost 70% of the time, for the totality of bird species in the archipelago. These researchers explain the discrepant bird results with reference to their initial assumptions. At any rate, if we do take this result as indicative, it shows that about 30% of 47 the distribution of bird species on neighbouring islands is due to stochastic processes. So for species whose distribution was (putatively) determined merely stochastically, neither competition nor adaptation to local plants would have been required to explain their distribution. Obviously the stochastic hypothesis is an alternative explanation to both Lack's and Bowman's hypotheses. For present purposes, I need say no more about the stochastic null hypotheses employed here than that they are estimates, of how often, by chance alone, any pair of species that are selected at random from a common pool of species, would be found on two islands. The second null model takes into account that some species have different abilities with respect to dispersal and persistence. The authors conclude, "Although both hypotheses are found to be inadequate models of compositional similarity in the Galapagos, the results suggest that a substantial proportion of compositional similarity can be considered a consequence of stochastic processes of dispersal and persistence, and that compositional similarity ... [does not arise] in a fashion determined by the physical environment or species interactions" (Conner and Simberloff 1978, 219). 2.7 Schluter and Grant 1984 Most of the work of the Grants has focused on the ground finches, genus Geospiza. This genus forms one branch of the radiation of finches on the Galapagos islands according to both the phylogeny Lack originally proposed in 1947 and Peter Grant's later adaptation of it (Grant 1986, 11). Schluter's early work primarily involves these granivorous ground finches. 48 The main advantage of studying the ecology of this subset of the ground finches is that they feed primarily on seeds-especially in the dry season when selection is most highly operative because there is less food available. Furthermore, the different sizes and hardnesses of the seeds are both something that can be measured and that can demonstrably be correlated with the beak sizes of the finches that feed on them. By the time Schluter joined the Galapagos finch field study for his doctoral research, Grant and his collaborators had isolated a number of lines of circumstantial evidence in support of the thesis that Lack had proposed-that the degree of difference in morphology of finches, particularly on the same islands, could not be adequately explained without invoking interspecific competition. The most significant irregularity is that there is a relatively smooth gradation of seed sizes, but there is a clumping of beak sizes (depths) in sympatric species. If it were only the seed sizes that influence the beak sizes, then the beaks should show a smooth gradation of sizes as well, given that it is the size of the beak that is the most salient adaptation to seed size. This pattern suggests but does not prove that there is competition between the finch species, since a purely random distribution of species would result in the same pattern (Grant 1981, 658). Schluter and Grant note that the disagreements between the floristic, competition and stochastic hypotheses "have been difficult to settle because data to evaluate alternative explanations are usually unavailable, and because the means to do so are often equivocal" (Schluter and Grant 1984, 176). They add, "Our contribution here is the development of a procedure which overcomes many of these difficulties" (ibid.). In their 1984 paper, there are two main results. They first graph the "population density of a solitary finch species as a 49 function of the mean size of its beak" that would be expected on the basis of available seed food on each of 15 Galapagos islands (Schluter and Grant 1984,182). Then they develop five predictive models based on only 12 of those data sets. Their method allows researchers to decide between a number of hypotheses that could account for the distribution of finches in the Galapagos: they set up computer simulations of each hypothesis, and then compared the results to the actual distribution of finches on the islands. The five models they tested "differ in the extent to which the evolution and/or assembly of species is directed by food supply, through its effects on population density, and by interspecific competition for food" (Schluter and Grant 1984,176). In the first model, "Food supply limits the range of feasible morphologies, without influencing the probability of persistence or evolution within this range. Interspecific competition is absent" (Schluter and Grant 1984, 177). At the other end of the spectrum, the fifth model accommodates adaptation to available resources while competition between species for limited resources determines the species on various islands and their relative abundance. In essence, the five-way test is a computer simulation, based on the available data, of the various hypotheses that have been presented to explain the distribution of the finches on those islands, restricted to analysis of the generalist seed-eating finches. I turn now to their first result. Their "procedure estimates the abundance of a colonizing finch species on islands in the absence of interspecific competition" (Schluter and Grant 1984, 176). The abundance and size of available seeds on 15 islands were determined by random sampling, and observations were made of which seeds were consumed by finches of which beak sizes. Seed size is correlated with seed hardness, but 50 the latter is assumed to be more relevant to which seeds a finch can manage. Generally, these finches consume seeds that the length and depth of their beaks make them best adapted to, even though it is possible for them to manage seed sizes outside the range of their preferred type. It might be wondered why finches with bigger beaks, and therefore the ability to crack even the largest seeds, do not consume all the seed sizes from smallest to the largest that they are capable of cracking. Although finches with larger beaks (that are also larger birds) are capable of eating the smallest seeds, they rarely do because those seeds are "relatively less profitable for large finches than small finches, and hence are more likely to be ignored"18 (Schluter and Grant 1984,179). Only the species of finch that, particularly during the dry season specialize in eating a variety of seeds, were utilized in the study: G. fuliginosa, G. fortis, and G. magnirostris. G. difficilis was included in the analysis on islands where it, too, was primarily a generalist19 granivore. There are two relationships that will be of particular importance in what follows. First, there is a linear relationship between the log of seed hardness and the log of beak depth of finches' preferred seeds (see figure 3). Schluter determined the availability of seeds, converted that to biomass, and determined empirically how much each finch species 20 biomass could be supported per unit biomass of each species' preferred seeds. The result, and second important relationship, is a species-specific linear relationship between finch biomass and the total biomass of all of the seeds consumed by a particular species (see figure 4). These two 51 SEED BIOMASS (mg/m2) Fig. 3. "Mean hardness (kgf) of hardest seed preferred regressed against mean beak depth of adult male finches. Linear regression uses log-transformed data; Y= 3.OX- 5.3. Dashed line adds 1 SD of the residuals to 7." Symbols: G. fuliginosa (•); G.fortis (•); G. magnirostris (o); G. difficilis (A).;G. conirostris (filled diamonds);"hardness of seeds not consumed by these populations" (stars). Reproduced from Schluter and Grant, American Naturalist, 123 (2) (1984), 180. 5 2 LU LU CO O o 1.0-0 . 0 1 1.8 2 . 2 2 . 6 3 . 0 L O G B E A K D E P T H (mm) Fig. 4."Number of finches encountered per hour in census walks, converted to biomass, plotted against the biomass density of preferred seeds." Symbols as in Fig. 3. Reproduced from Schluter and Grant, American Naturalist, 123(2) (1984), 181. 53 relationships allow the construction of curves describing expected population density of a solitary finch species as a function of the mean size of its beak... . For a given beak depth the preferred size-hardness range of seeds is determined using the [previously established] relation... . Those seeds on the particular island that fall between these limits are identified and their biomass is summed. Seed biomass is converted to finch numbers using the [discovered finch biomass x seed biomass relation] and the known mass of individual finches of a given mean beak size... . This procedure is repeated for the entire feasible range of beak depths, for all islands. (Schluter and Grant 1984, 182) This method generated different curves for different islands in the study, when expected population density (on the y-axis) is graphed against log beak depth (on the x-axis). In summary, and more simply, "The result, for each island, is a curve giving the expected population density of a solitary finch species as a function of its mean phenotype" in terms of beak depth (Schluter and Grant 1984, 176). For example, on the island of Daphne, a log beak depth of 2.2 mm is the optimal beak depth given the food availability on that island: a maximum density of finches would survive there with that beak measurement. Initially one might wonder whether the resultant graphs merely expressed what was already known, but in different terms. Says Grant, 54 The problem [was] to devise a method of analysis that predicts the properties of finch communities from a knowledge of food supplies without merely reassembling the known relationships to predict what is known already. The solution to that problem was developed by Dolph Schluter. It is to estimate and use a third variable: expected population density, as a function of beak depth. (Grant 1986, 332) This passage indicates something of the importance of the result, while demonstrating that it fulfils the following requirement: "The same prediction may also help us to see if the theory is correct through a matching of prediction with observation" (Grant 1986, 331). Furthermore, in spite of the 1984 article's being published under both names, Grant here gives credit to Schluter for coming up with the relevant mathematization of the data. What was interesting about the resultant "landscapes was that they were hilly with lots of peaks and valleys, and the second interesting thing about them was that they differed between islands" (Schluter 1999)-in position, breadth, and number of peaks (Schluter and Grant 1984, 185). What these graphs demonstrated, given the availability of seed types and amounts on the island, was what would be the ideal beak size(s) to deal with that food supply. "The most important factor responsible for these [expected] patterns is the gaps in the frequency distributions of seed characteristics on an island, and the difference among islands in the position of these gaps" (Grant 1986, 333-335). Thus, by comparison of the expected to the observed pattern, it was once again demonstrated that the main factor influencing the biogeography of the finches was adaptation to food resources. 55 What might have been expected instead of these polymodal (i.e., having peaks and valleys) curves was a normal curve (Schluter and Grant 1984, 182). The beak depth corresponding to the peak of the normal curve would have represented the one best beak depth for that island, given the food supply. What they found instead was that many of the islands studied had more than one ideal beak depth. Furthermore, in most cases, these peaks in expected population density occurred over the mean values of the beaks of the species that were actually observed on the islands (see figure 5). Grant suggests that each of these 15 curves "is analogous to a complex, environmentally determined, adaptive landscape" (Grant 1986, 335), and Schluter suggests that these landscapes serve to estimate adaptive landscapes "from the ecological mechanisms upwards" (Schluter 2000, 111). The 15 computer-generated graphs already detailed were to give the expected population density on each island for every reasonable (log) beak depth. The next step, their second major result, was then to examine the predictions of the various models representing how the islands might have been colonized against the actual distribution of species found. This was done for only 12 of the islands. These computer simulations would, for each island with its respective flora and other factors as detailed below, generate which phenotype would be expected. So rather than specifying a species, the simulation would specify which (mean) sized beaks should be borne by the birds on that island. Then the expected mean beak sizes could be compared to the mean beaks of the species actually found on those islands. Model I generated the species (beak sizes) that should be found on the islands by "a process of random colonization, with all phenotypes being equally likely to colonize, within 56 Fig. 5. "Expected population density of a solitary granivorous finch species as a function of beak depth on fifteen islands.... Mean log beak depths of male ground finches on the islands are indicated by the positions of symbols. Symbols as in Fig. 3. Reproduced from Schluter and Grant, American Naturalist, 123(2) (1984), 184. 57 limits set by the food supply" only, i.e., wherever there is at least some grain food (Grant 1986, 336). "By model I all phenotypes are equally likely to successfully colonize or evolve where expected population density is positive; competition and variation in food supply play no role"21 (Schluter and Grant 1984, 185). Therefore, "all forms [would be]equiprobable, [and] the mean phenotype of a species should be distributed as a uniform random variable" within the limits of beak size and in the range of positive population densities as given in the calculated graphs (ibid.). This yields the prediction that on each 22 island, any viable beak size is equiprobable. Thus, one can compare the actual distribution of species (each with its respective mean beak depth) against the probability of those species having occurred randomly in the distribution seen there. The result is that it is extremely improbable that the actual species observed to be present on their islands were randomly assembled: the species are more different from each other with respect to beak depth than would be likely to happen randomly. Another way of conceptualizing this is to say that there were bigger, fairly regular gaps between mean beak depth in the species actually present than there should have been according to chance. This had been argued in earlier papers of Grant and his collaborators (and was alluded to above in the background to Schluter's work), but the 1984 joint paper demonstrates this result via computer simulation. Model II has more initial plausibility in that actual food abundance is taken into account. It simulated colonization of virtual islands by random selections of species of the same number as have actually colonized the respective real islands, but while randomly assigning mean beak depth to each species within the limits of beak depth actually observed on those islands (Grant 1986, 338). Here the success of those species would be influenced 58 by the expected population density, that was in turn a function of how much food was available (Schluter and Grant 1984,185). The results still differed compared to the actual presence of species, in that the gaps between existing species were bigger than this model predicted. Model III assigned species randomly (with respect to beak depth) to islands, but this time success was entirely a function of whether the relevant beak depth occurred close to a peak in expected population density. This model was meant to allow for the possibility that a finch species that did not have the optimal beak size for the food resources might colonize the island, and subsequently evolve toward one of the peaks in expected population density (Grant 1986, 338). The model accomplished this by only allowing, in the simulation, beak sizes associated with peaks. A peak is defined as a local maximum in expected population density greater than the density at points up to 0.10 log beak depth units away on either side of the maximum, (ibid.) As noted earlier, there is a fairly high correspondence between peaks expected on the basis of these calculations and the beak depths of species actually present. What is not seen in the actual species distribution, however, is more than one species occurring on the same peak, or on peaks very close to each other-both states of affairs that are predicted by model III. Thus the first three models, which did not include interspecific competition, were unable to account completely for the distance between the mean beak sizes of the ground 59 finches actually found on the 12 islands studied. This alone suggests that too few relevant factors were being taken into account to produce an accurate result. In short, the alternative models proposed to account for the pattern of finch species on the islands-purely random distribution within the viable beak sizes, distribution based on food availability alone, and a combination of random colonization with the ability of finches to evolve toward available food sources-were insufficient to predict the actual pattern of colonization. However, with the addition of competition into the calculations, the observed distribution is modeled very closely (models IV and V). One inconsistency between the first three models and the observed finch distribution is that the former predict that more than one finch species may occupy the same adaptive peak, but this is never seen in the real Galapagos ground finches. The assumption, going back to David Lack (1947), has been that only competition could account for this phenomenon. What is remarkable is that the Schluter and Grant (1984) paper finally confirms this. The simulation based on the fourth model is "for colonization with competitive exclusion, but no evolution", in other words, competitive exclusion, but not character displacement. Three species with any feasible beak depths are allowed to colonize a given island. The conditional expected population density of each is then computed as its expected density excluding all foods lying within the preferred range ... of any of the other species. The process identifies those species (one or more) whose conditional density is zero; the one with the 60 lowest expected density is then eliminated. Conditional densities for the remaining species are recomputed, and serial deletion (extinction) of species continues until all remaining species have positive conditional density. A new colonist with a different beak depth is then added and the entire process is repeated until no additional species can successfully invade. The order of colonization does not matter, for the procedure is entirely deterministic and gives a single result for each island. (Grant 1986, 339) The results compared to the actual finch distribution were highly positively correlated (correlation coefficient (r) = 0.90, with N= 23, i.e., the number of species examined on the • 23 12 islands), compared to what would be maximally expected by chance (r = 0.26). The fifth model differs from the fourth mainly in allowing for the "coevolution of forms driven by interspecific competition" (Schluter and Grant 1984, 190), i.e., character displacement.24 What this means is that "morphological adjustments in response to competitor species" were taken into account in the calculations (ibid., 177). The results here 25 were slightly better (r = 0.92, N= 32) (ibid., 190). "Thus, in conclusion, incorporating competitive effects into the models yields a close correspondence between actual and predicted properties of finch communities" (Grant 1986, 339). So in the end, the most complex model, the one long-postulated by the Grant group, was shown to be the best fit to the available data. Even so, because "there is some overlap between the beak sizes of different species on the Galapagos," it is "not possible to predict the identity of the species that would be lying under the peak" (Schluter 1999). So what was 61 predicted was the mean beak sizes of the species present, but nothing as specific as the species themselves, where there was an overlap in beak size between species. The significance of this work, according to Futuyma, is that it demonstrates "that in at least some instances, (1) the distributions of species are affected by those of competing species, and (2) the numbers and distributions of species in a locality are predictable from ecological information" (Futuyma 1998, 219). Additionally, Taper and Case suggest in their review that the only hard evidence for character displacement's occurrence is in the work of Schluter and Grant (1984) on Geospiza fortis and G. fuliginosa on the Galapagos island of Daphne (Taper and Case 1992, 77). Of course, these two species on a single island constitute only a subset of Schluter and Grant's (1984) results. Recall that character displacement was only one of the two kinds of competition postulated to influence the Galapagos finch radiation. The five-way test revealed that Model IV (that included competitive exclusion but no character displacement) accounted for most of the discrepant beak depths of sympatric species. 62 Chapter I I I Lakatos on Scientific Progress Imre Lakatos advocates what he calls a "methodology of scientific research programmes." Each research programme is made up of a hard core of unfalsifiable commitments together with an often-extensive network of auxiliary hypotheses surrounding the hard core and protecting it from refutation. The programme also includes methodological suggestions generated by so-called positive and negative heuristics. In what follows I describe Lakatos's account and tease out some of its implications. After laying out the main elements of his account, I examine what he considers to be the structure of scientific theories. Then follows a section on what does and does not progress. This is followed by an attempt to extract Lakatos's views on how research programmes begin. Finally, I address the size and scope of what, in Lakatos's views, are genuine research programmes. This last section is particularly relevant for the analytic chapter that follows, for it tells us how to understand the Galapagos finch competition case in terms of research programmes. 3.1 Lakatos's Methodology of Research Programmes What Lakatos primarily means by "methodology" when he speaks about the "methodology of research programmes" is implied by his reconstruction of various historiographies of science and his referring to them as "methodologies." He says, "The methodology of 63 scientific research programmes constitutes, like any other methodology, a historiographical research programme" (Lakatos 1971, 101-2). This sort of "methodology" is not to be confused with his usage of the same term to refer to the positive and negative heuristics within his model of research programmes. What Lakatos refers to as "heuristic" is methodological in the sense of "giv[ing] advice about what methods to employ to achieve some end" (Hacking 1981, 131). For Lakatos, the basic unit of advancement for scientific knowledge must not be "an isolated theory or conjunction of theories" (Lakatos 1971, 99). Instead it is the research programme itself. There are four components within each research programme: a hard core, a protective belt, and the positive and negative heuristics. The hard core is a small body of theories or hypotheses. It is designated as non-falsifiable and has associated with it the negative heuristic. As we will see below, the hard core, in conjunction with auxiliary hypotheses, can to be used to make predictions. The hard core by itself cannot, however, for that would make it falsifiable. It is not possible for the research programme's advocates to give up the hard core without abandoning the programme altogether (Larvor 1998, 51). The negative heuristic designates the hard core as the irrefutable portion of the programme (Lakatos 1970, 134) and dictates to scientists "what paths of research to avoid" (Lakatos 1970, 132). Moreover, at least one kind of research that it directs the programme's advocates to avoid is work on theories that are inconsistent with the programme's hard core (Lakatos 1970, 133). All this is not to say that the hard core is never given up: "if and when the programme ceases to anticipate novel 64 facts, its hard core might have to be abandoned" (Lakatos 1970, 134); however, this would constitute a giving up of the whole programme, not just a revision of the hard core. The protective belt is comprised of (auxiliary) theories or hypotheses as well, but these are falsifiable, and have associated with them the positive heuristic. The protective belt often also contains models. Lakatos appears always to understand these as physical models. An example he gives is Bohr's series of models of the hydrogen atom, starting with the first model "based on a fixed proton-nucleus with an electron in a circular orbit" (Lakatos 1970, 146). Henceforth, for simplicity, I will follow Lakatos in referring predominately to the auxiliary hypotheses in the protective belt, but we are to keep in mind that models also count as a type of auxiliary hypothesis. Compared to the smaller hard core, the protective belt makes up the majority of the research programme. It is not entirely clear whether the positive heuristic is to be thought of as part of the protective belt, although it appears that it is not. The role of the positive heuristic is to give advice to scientists on how to proceed with the research programme, which it does, in part, by offering "a set of problem-solving techniques" (Lakatos and Zahar 1978, 179). Specifically, among its many roles, the positive heuristic "[1] defines problems, [2] outlines the construction of a belt of auxiliary hypotheses, [3] foresees anomalies and [4] turns them victoriously into examples, all [5] according to a preconceived plan" (Lakatos 1971, 99). I will now touch on each of these functions. When Lakatos gives the positive heuristic the role of "defm[ing] problems", I take this to mean (in part) that it specifies the domain to which a theory will apply. For example, "Bohr's problem was ... to explain the paradoxical stability of the Rutherford atom" 65 (Lakatos 1970, 147). We can take a Lakatosian "problem" as similar to a Kuhnian "puzzle," but only insofar as it is a phenomenon that is accepted as requiring a solution. One significant difference would be that for Kuhnian puzzles, the scientist knows the sort of answers that will count as solutions to puzzles, whereas Lakatos does not have this element explicitly built into his notion of what problems are. The positive heuristic is charged with the function of encouraging scientists to come up with auxiliary hypotheses that, when falsified, do not take down the hypotheses of the hard core with them. One example of the positive heuristic in action has "it encouraging] work on auxiliary hypotheses which might have saved [the research programme] from apparent counterevidence" (Lakatos 1970, 133), which is an instance of "forsee[ing] anomalies" (hence [3]). Furthermore, the positive heuristic "consists of a partially articulated set of suggestions or hints on how to change, develop, the 'refutable variants' of the research-programme" (Lakatos 1970, 135). In this way, the positive heuristic plans the outline of the protective belt ([2] from above). Again, these are specified in advance of the establishment of the research programme proper. Lakatos's example of [4], the positive heuristic turning an anomaly into an example consistent with a research programme's hard core, is the discovery of Neptune (the positive example), from the perturbation of the orbit of Uranus (the anomaly) (Lakatos 1978b, 4). Here scientists, in accordance with the dictates of the positive heuristic, took the supposed counterexample, not as evidence of the failure of the hard core commitments of Newton's programme, but instead as a reason to look for a way to explain the anomaly that would leave Newtonian mechanics intact. Accordingly, the perturbation in Uranus's orbit could be 66 explained without abandoning Newtonian commitments by postulating another planet whose gravitation impinged upon Uranus. I turn now to some of the ways that the positive heuristic constitutes a "preconceived plan" ([5]). Part of the way in which the positive heuristic generates the protective belt is in its function as a strategy for producing models, which Lakatos says are "bound to be replaced during the further development of the programme, and one even knows, more or less, how" (Lakatos 1970, 136). Additionally, the positive heuristic provides other forms of positive guidance to the scientists engaged in the research programme (Lakatos 1970, 132). For instance, the positive heuristic provides a "long-term research policy" indicating the order in which scientists in a research programme are to tackle anomalies (Lakatos 1970, 135). From the role of the positive heuristic in generating auxiliary hypotheses, one should not get the idea that the programme has to wait for its protective belt. Indeed, the scaffolding of the auxiliary hypotheses is laid out in advance to a significant degree: "in the positive heuristic of a powerful programme there is, right at the start, a general outline of how to build the protective belts" (Lakatos 1970, 175). I will return below to how research programmes begin. Finally, it is possible to "appraise research programmes, even after their 'elimination,' for their heuristic power: how many new facts did they produce, how great was 'their capacity to explain their refutations in the course of their growth'?" (Lakatos 1970, 137). Heuristic power differs from heuristic progress (to which I will return) in emphasizing the degree to which theories are in accordance with the heuristic as set out at the beginning of the programme rather than the theoretical progressiveness of a programme. 67 From this quick overview we can see that the positive heuristic is doing a lot of work in Lakatos's research programme and, hence, in his historiography as well. However, Lakatos nowhere explicitly mentions his heuristics as giving suggestions regarding experimental methodology. Although his heuristics are arguably consistent with laying out suggestions for the kinds of experiments that may and may not be used in proceeding with the programme, it is a shortcoming of his model that it is mute on this issue. 3.2 Series of Theories A research programme develops by having its belt of auxiliary hypotheses "modified, increased, [and] complicated", while its hard core remains unchanged (Lakatos and Zahar 1978, 178-9). At the same time, within a programme, theories, each of which is a conjunction of the hard core's theoretical commitments and auxiliary hypotheses (Zahar 26 1976, 215), are modified and expanded over time. Because the entities referred to here are individual theories, this may be a source of confusion since Lakatos says that the "basic unit of appraisal" cannot be individual theories, or conjunctions thereof (Lakatos 1971, 99), or individual hypotheses or conjunctions thereof (Lakatos and Zahar 1978, 178), but must be 27 whole research programmes. Individual theories looked at in isolation from the research programmes in which they are embedded are not, for Lakatos, rightly understood as the building blocks constituting science. We are thus to understand the individual theories that Zahar here refers to-because they are inextricably bound up with a research programme (Lakatos's "basic unit")-as the sorts of entities that do admit of progress on Lakatos's 68 account. An equivalent formulation of talk about individual theories changing over time is to assert that series of theories, each theory of which is a slight modification upon its predecessor, are conjunctions of the hard core and some auxiliary hypotheses, with changes occurring within the auxiliary hypotheses in each successive permutation. Zahar provides a succinct statement of the relationship between the (positive) heuristic and the series of theories involved in a progressive shift: "Applications of the heuristic to specific problems (which may or may not be set by 'refutations' or anomalies) generate a sequence of theories" (Zahar 1978,215-6). Because I will be relying for some of my arguments in the next chapter on an interpretation of Lakatos in which he is committed to the inability of the hard core to make predictions, I will review the argument for this assumption here. In a passage in which he argues against the tenability of dogmatic falsificationism, Lakatos says that "exacrty the most admired scientific theories simply fail to forbid any observable state of affairs" (Lakatos 1970, 100). Again, "it is exactly the most important, 'mature' theories in the history of science which are prima facie undisprovable" (Lakatos 1970, 102). That these theories do not forbid any observations is equivalent to saying that the relevant theories are non-falsifiable. I also take it that Lakatos here means to refer to the most centrally-held theories, the hard core of a research programme. From this commitment it is possible to derive that the theories of the hard core by itself cannot predict novel facts, for if they could, then these would render the hard core falsifiable. However, a theory is made up of a hard core plus an auxiliary hypothesis and we have seen that it is the hard core in conjunction with at least one auxiliary hypothesis that 69 makes predictions. Series of theories progress. Progress is made only by the introduction of novel predictions. Hard cores, as we have just seen, should not be able to make predictions 28 by themselves, because that would make them falsifiable . We are left with auxiliary hypotheses as the things that either predict or constitute predictions-in conjunction with the hard core. The hard core, though, cannot by itself predict. In addition to this argument, Lakatos gives an example of "an imaginary case of planetary misbehaviour" (Lakatos 1970, 100). There every auxiliary hypothesis is simultaneously a prediction. 3.3 Scientific Progress and Progressive Problemshifts Lakatos takes it for granted that science grows (Hacking 1981, 129), although he sometimes appears to make a distinction between growth and progress, at least insofar as "growth" is usually a non-technical term in his usage. The success of a research programme is defined as its machinery leading to a progressive problemshift, whereas a degenerating problemshift is the result of an unsuccessful programme (Lakatos 1970, 133). First, then, it is worth considering what Lakatos means by "problemshift." He says that the "developing series of theories" making up his research programme is "a special kind of'problemshift'" (Lakatos and Zahar 1978, 179). "Problemshifts" are referred to throughout Lakatos's elucidation of his research programmes and those of Popper, while the most thorough definition he seems to give is that problemshifts are "series of propositions which are the result of progressive or ad hoc modifications" (Lakatos 1978a, 221). So in making a problemshift, a research programme adds something new to, or modifies some aspects of, 70 the propositions that comprise the auxiliary hypotheses. At the same time, the series of propositions constituting the problemshift is what makes for a series of research programmes. "Problemshift" seems to be no more than an alternative way to refer to the 29 difference between one theory and its successor in a series of theories. Progress in general "is measured by the degree to which a problemshift is progressive, by the degree to which the series of theories leads us to the discovery of novel facts" (Lakatos 1970, 118). Lakatos stresses in a number of places the prediction of novel facts (by theory) as a necessary component of progress within (a series of) research programmes. He is particularly concerned to have the research programme's theoretical segments predict facts rather than having theory accommodate additional empirical data after its discovery, an emphasis that seems to reflect an attempt to avoid having the addition of post hoc hypotheses to a programme inadvertently count as progressive. For instance, he says, "It is always easy for a scientist to deal with a given anomaly by making suitable adjustments to his programme... . Such manoeuvres are ad hoc, and the programme is degenerating, unless they not only explain the given facts they were intended to explain but also predict some new fact as well" (Lakatos and Zahar 1978, 179). Here it is suggested most clearly that what is motivating Lakatos in his having theoretically progressive shifts lead "to new unexpected predictions" (ibid.) is his attempt to disallow the possibility that ad hoc additions of theory will inadvertently count as progressive. There is one emendation to be noted here. Lakatos, in his later work, incorporates a suggestion of Zahar: so long as the scientist who proposes the new theory does not set out to explain an existing fact, it can count as a "novel" fact even though it is already known (Lakatos and Zahar 1978, 185). 71 There are two kinds of progress in Lakatos's model: theoretical and empirical. There is a hierarchy between the two, and theoretical progress is primary: a series of theories is theoretically progressive (or 'constitutes a theoretically progressive problemshift') if each new theory has some excess empirical content over its predecessor, that is, if it predicts some novel, hitherto unexpected fact. Let us say that a theoretically progressive series of theories is also empirically progressive ... if some of this excess empirical content is also corroborated, that is, if each new theory leads us to the actual discovery of some new fact. (Lakatos 1970, 118) Notice that a programme's progress is first assessed in terms of theory (or a series of theories), and it is the new prediction of facts by theory that gets the scientific process, as Lakatos understands it, going in the first place. Parasitic upon theoretical progress is empirical progress, that is made when the predictions of theory are found to obtain: Finally, let us call a problemshift progressive if it is both theoretically and empirically progressive and degenerating if it is not. We 'accept' problemshifts as 'scientific' only if they are at least theoretically progressive; if they are not, we 'reject them as 'pseudoscientific' (Lakatos 1970, 118) 72 I understand Lakatos in this passage as implicitly committed to the position that if a theory is only empirically progressive, then it is degenerating. Furthermore, a purely empirical progressive problemshift would not count as scientific. As additional evidence that he holds this view, Lakatos equates the "scientific ... form" of a theory with its "content-increasing" characteristic, i.e., its having empirical consequences (Lakatos 1970, 139). Apparently, therefore, the gathering of additional data about the world in the absence of a theory that predicts and perhaps explains it, is non-scientific and non-progressive. It is also possible to observe that theoretical progress and empirical progress are both necessary but not individually sufficient conditions for Lakatos's progress. Also, A research programme is ... stagnating if its theoretical growth lags behind its empirical growth, that is, as long as it gives only post hoc explanations either of chance discoveries or of facts anticipated by, and discovered in, a rival programme (^degenerating problemshift). (Lakatos 1971, 100) So we see that even in the case in which a research programme has theories about the relevant domain, so long as the discovery of new data outstrips the predictive theory invented by its research programme, the programme cannot be understood to be progressing. The following example is worth examining because it illuminates how Lakatos understands a progressive problemshift versus a merely theoretical progressive shift. 73 Lakatos refers to it as a "contrived micro-example of a progressive Newtonian 30 problemshift" (Lakatos 1970,133-4), and says that "the example constitutes a consistently progressive theoretical shift" (Lakatos 1970, 134). The example is remarkable in that it demonstrates some of the elements of Lakatos's model pushed to their extremes. It illustrates "an imaginary case of planetary misbehaviour" (Lakatos 1970, 100) in which the hard core of Newton's programme is never abandoned, but increasingly inventive auxiliary hypotheses are added to the protective belt; when these are tested and found not to be borne out, the researcher blithely continues inventing more auxiliary hypotheses having novel predictions. However, even though Lakatos calls this example a theoretically progressive shift, he considers it only partially progressive when assessed at the time his hypothetical Newtonian came up with the novel predictions. Again, "a problemshift [is] progressive if it is both theoretically and empirically progressive and [is] degenerating if it is not" (Lakatos 31 1970, 118). To be progressive on all of Lakatos's measures of progress, a case would have to demonstrate both empirical and theoretical progress. It will demonstrate empirical progress later, but it does not at this time when Lakatos considers it only theoretically 32 progressive. Furthermore, in a footnote discussing the kinds of ad hoc auxiliary hypotheses that Lakatos recognizes, he includes "those which do have such excess [empirical] content [over their predecessors] but none of it is corroborated (Lakatos 1971, 125; footnote 36). This way of defining what it means to be ad hoc is equivalent to the "progressive theoretical shift" just outlined, and so he would consider his Newtonian thought experiment to be ad hoc. Nevertheless, this contrived Newtonian example does constitute a scientific episode: "We 'accept problemshifts as 'scientific' only if they are at 74 least theoretically progressive; if they are not, we 'reject them as 'pseudoscientific'" (Lakatos 1970, 118). Notice that Lakatos takes from Popper the importance of having falsifiable predictions at least here in his protective belt, although clearly not in the hard core; as with Popper, without these falsifiable predictions, the research programme in which these theories are embedded, is not to be considered scientific. Another difference characterizing theoretical versus empirical progress is how soon each type of progress may be verified: the latter cannot be verified immediately, and is often verified only much later; in contrast, the existence of theoretical progress is "verified" by the research programme's having novel predictions of some empirical fact (Lakatos 1970, 134). These novel predictions are, of course, obvious as soon as they are made, whereas whether they obtain will usually not be evident until later. Accordingly, Lakatos "require[s] that each step of a research programme be consistently content-increasing: that each step constitutefs] a consistently progressive theoretical problemshift" whereas it is sufficient for the 33 empirical shift to be only "intermittently progressive" (Lakatos 1970, 134). Lakatos also refers to another kind of "progress" that should be mentioned for the purposes of clarification-heuristic progress. His first mention of heuristic progress is in Lakatos and Zahar (1978). In a singly-authoured postscript to that paper (Lakatos 1978c, 189, editor's footnote), Lakatos includes "heuristic progress" among the "three standard criteria for appraising research programmes" (Lakatos 1978d, 189). However, what Lakatos appears to mean by "heuristic progress" is counter-intuitive. It is not to be understood as progress within the positive heuristic itself over time. Heuristic progress is not a kind of progress that is separate from, and to be contrasted with, what he has called theoretical and 75 empirical progress. He says that his "methodology also contains a notion of heuristic progress: the successive modifications of the protective belt must be in the spirit of the heuristic. Scientists rightly dislike artificial ad hoc devices for countering anomalies" (Lakatos and Zahar 1978, 179). What Lakatos is trying to get at in this talk of so-called "heuristic progress" seems to be the following: for a research programme to be thoroughly progressive, it must not be ad hoc. In this context, Lakatos adopts from Popper three kinds of ad hoc auxiliary hypotheses: those which have no excess empirical content over their predecessor ('ad hoc\'), those which do have such excess content but none of it is corroborated (W hoci) and finally those which are not ad hoc in these two senses but do not form an integral part of the positive heuristic ('ad hoc^'). (Lakatos 1971, 125; footnote 36) Antithetical to the first two kinds of ad hoc hypotheses are theoretical and empirical progress respectively. The final sense of "ad hoc" represents another kind of degeneration (anti-progress), so for a research programme to be non-ad hoc3 might seem to make it progressive in some third sense of progressiveness, hence "heuristic progress". What I want to suggest is that although Lakatos considers "heuristic progress" to be an important characteristic of a research programme, the goal of emphasizing this characteristic could be met more parsimoniously by emphasizing that "successive modifications of the protective belt must be in the spirit of the heuristic" (Lakatos and Zahar 76 1978, 179). He makes this claim elsewhere, as we have seen. What he wants to block in a programme is that it "anticipates novel facts but does so in a patched-up development rather than by a coherent, pre-planned positive heuristic" (Lakatos 1971, 125; footnote 36). So he means no more by "heuristic progress" than the degree to which a programme's theories are in accordance with the positive heuristic as set out at the programme's beginning. This interpretation is consistent with the requirement for "more heuristic unity" that Lakatos offers when expanding upon his use of the term "heuristic progress" (Lakatos 1978d, 189). Finally, supersession of one research programme over another is based on comparisons of the progress of each. "If a research programme progressively explains more than a rival, it 'supersedes' it, and the rival can be eliminated" (Lakatos 1971,100). Similarly for theories within a research programme: "a theory can only be eliminated ... by one which has excess empirical content over its predecessors, some of which is subsequently confirmed" (Lakatos 1971, 100), which was how he defined progress. 3.4 What Does and Does Not Progress Since Lakatos has stipulated that there are only two true types of progress (theoretical and empirical), it is possible to say in summary that he does not take any of the other elements of his methodology of research programmes to be, in themselves, progressive. The hard core and the positive and negative heuristics are not amenable to progress, although the latter two can change under certain conditions. Neither is progress going on in the protective belt, since the (series of) theories that are the units that undergo progress for 77 Lakatos have the elements of the protective belt as one of their subordinate parts. Again, theories were the smallest unit undergoing progress, and the auxiliary hypotheses of the protective belt are subunits of it. The protective belt and the heuristics have important roles to play in progress, but none of them, in itself, progresses. Other aspects of science that do not progress for Lakatos include experimental method and our conceptual understanding of the world. With respect to the latter, however, Lakatos holds that what he calls "creative shifts" or "meaning-shifts" within the protective belt are sometimes instrumental in making progressive shifts. This is stated most clearly where Lakatos says that one particular "'progressive' development, incidentally, hinged on a 'creative shift'" (Lakatos 1970, 167). So these creative shifts, or innovations, are admitted to be implicated in progress even though they do not, in themselves, progress. Furthermore, to reiterate, the addition of more factual information about the world in the absence of theory that predicted it does not count as progressive. It is worth summarizing also what Lakatos allows as changeable. I will argue subsequently that he has too narrow a conception of progress. However, a defender of Lakatos would rightly argue that for many instances in which I see epistemic progress but Lakatos does not, the elements of his programme that do not constitute progress in themselves nevertheless contribute crucially to the progress that he does admit. This is a serious criticism of my position, especially given that a Lakatosian might further argue that if we are not to define every good thing in science as progress, there needs to be some principled demarcation between progressive elements and non-progressive elements that contribute to progress. Change is a necessary (although not sufficient) condition for 78 progress, and so it is important, in giving Lakatos his due, to consider in which elements of research programmes he is committed to recognizing change. So in summary, once a programme has been established, Lakatos allows for change within all the elements of his research programme except the hard core. As we have already seen, individual theories and programmes both admit of change (as will sub-programmes). Of course, the most significant changes occur in the protective belt of (auxiliary) hypotheses. This is primarily where (in conjunction with the immutable hard core) predictions of novel empirical consequences happen. The positive heuristic, although it is set out at the beginning "according to a preconceived plan" (Lakatos 1971, 99) can undergo some change. Finally, as we shall see below, conceptual changes are also possible. 3.5 Before and After Establishment of a Research Programme In this section I will discuss the characteristics of research programmes before and after their establishment, and indicate some of the ways that they may come into being. These details are important so as not to misrepresent Lakatos's intentions when rewriting scientific episodes in his terms. First I will talk about the preparatory stages of the research programme. This will lead into how the research programme proper is initiated. Then I will detail some of the most significant characteristics of an established research programme. It is important to add that these characteristics of research programmes and pre-programmes apply to a science only once it is to be considered a "mature science" (cf. Lakatos 1970, 175). 79 3.5.1 Prior to a Research Programme It is important to look at how research programmes begin and what they look like before and after their initiation in order thoroughly to understand their methodology. Accordingly, I will explicate what I understand to be Lakatos's position on the characteristics of science before and after the initiation of a research programme proper. This will be based more on some tantalizing hints that Lakatos gives than on any concentrated explication on his part. If, however, one does not recognize the passages in which Lakatos is referring to pre-programme development for what they are, he might be understood as contradicting himself. One suggestion is that Lakatos considers a whole research programme to be instituted at a particular point in time. This occurs where he says, "The best opening gambit ... is a research programme" (Lakatos 1971, 99-100). Since the most central and defining characteristic of the research programme is its hard core, I take Lakatos here to be committed to a hard core's being established at a particular time. In his discussion of the genesis of the Copernican research programme, Lakatos suggests that "The geocentric hypothesis 'hardened' into a real hard core assumption only with the development of an elaborate Aristotelian terrestrial physics" (Lakatos and Zahar 1978, 181). Here with the "hardening" or solidification of the research programme, Lakatos commits himself to its being "real." This is another tantalizing suggestion that he has in mind a time prior to the existence of the research programme proper. It is at this time that the hard core's original components are undergoing mutation before they become the final immutable hard core. 80 Also consistent with an interpretation committing Lakatos to both the research programme's and the hard core's having a determinable beginning in time is the following comment: "The actual hard core of a programme does not actually emerge fully [formed]... .It develops slowly, by a long, preliminary process of trial and error" (Lakatos 1970, 133 footnote). Here Lakatos hints at a process that is finished before the proper beginning of the research programme. Further, this footnote provides evidence that Lakatosian research programmes, properly understood, have a beginning, before which there is a process of bringing the elements of the research programme together and deciding what they will look like. This is a process of gradual "evolution," and ends as soon as the research programme is established. Upon this establishment, the preliminary process of hard core formation has finished and the positive heuristic is established. The positive heuristic's role in the research programme is carried out "according to a preconceived plan" (Lakatos 1971, 99), that suggests that it is in place at the beginning of the programme and changes relatively little thereafter. Indeed, the positive heuristic of what Lakatos calls "old quantum theory" "was planned right at the start" (Lakatos 1970, 146). We should not, however, take this to mean that "all developments in the programme were foreseen and planned when the positive heuristic was first sketched;" they were not, for example, in Bohr's research programme (Lakatos 1970, 153). Furthermore, Lakatos's reference to a "coherent, pre-planned positive heuristic" (Lakatos 1971, 125; footnote 36) is evidence for his wanting the positive heuristic to be there "at the start" and for it not to change (much) thereafter. "Pre-planned" also of course implies that there is a time prior to the instantiation of the programme during which this planning of the positive heuristic occurred. 81 3.5.2 How Research Programmes Begin Understanding Lakatos's account of how research programmes begin is an important component in our understanding of his scientific progress. As we have seen, progress occurs in only some sub-components of research programmes and not in others. Therefore, knowing where the various Lakatosian boundaries lie is important in understanding actual cases. These distinctions are also important for helping us to recognize and delineate one research programme from another. Lakatos clearly finds it unproblematic that research programmes find their initial impetus in problems or anomalies confronting other research programmes, and that their hard cores constitute attempted solutions to puzzles given by our theorizing about the physical world. One instance of this is demonstrated in Bohr's light emission programme in which "The background problem was the riddle of how Rutherford atoms ... can remain stable" (Lakatos 1970, 141).34 Indeed, there are a variety of ways that one research programme can be related to another and usually these ways bear directly upon how research programmes begin. One programme can be an outgrowth of another, or even an inconsistent addition to another, as we will see below. The first of these two kinds of relationship between an earlier and a later programme can be thought of as a sort of branching. Lakatos accepts that historically one research programme can split into more than one, and that each of these can in turn split, and furthermore, that each of these new programmes operates independently of the "parent" programme. An instance of one layer of this sort of branching is demonstrated by Lakatos's 82 and Zahar's Copernicus example. They say that Ptolemy's and Copernicus's research programmes branched off from the Pythagorean-Platonic programme whose basic principle was that since heavenly bodies are perfect, all astronomical phenomena should be saved by a combination of as few uniform circular motions ... as possible. This principle remained the cornerstone of the heuristic of both programmes. (Lakatos and Zahar 1978, 180) An interesting twist in this "proto-programme" (i.e., the Pythagorean-Platonic programme) is that "the heuristic ... was primary, the 'hard-core' secondary" (Lakatos and Zahar 1978, 180-1). The positive heuristic here is to be understood as the suggestion that "astronomical phenomena should be saved by a combination of as few uniform circular motions ... as possible." It appears furthermore that what they mean here by the hard core is "where the centre of the universe lies" (180). From the positive heuristic of the original programme, that Lakatos here calls its "basic principle," came the heuristics of the two new programmes. Incidentally, this heuristic must be seen as a positive one, since at the time there was no hard core for a negative heuristic to defend. Lakatos adds, as a footnote, "The demarcation between 'hard core' and 'heuristics' is frequently a matter of convention as can be seen from the arguments proposed by Popper and Watkins ..." (Lakatos and Zahar 1978, 181 footnote). Further, he appears to intend that this observation apply to his methodology of research programmes as well. Larvor also takes this "matter of convention" as an admission 83 applying to Lakatos's own methodology (Larvor 1998, 103). Assuming it is the case that he means for this conventional demarcation to apply to his own programme as well, that allows for a fair amount of flexibility in the assignment of aspects of a scientific episode. As we will see in the next chapter, this flexibility has troubling consequences when Lakatos's method is applied to at least some cases. Another main way a new research programme may be instituted is by being "grafted on" to an earlier one: "some of the most important research programmes in the history of science were grafted on to older programmes with which they were blatantly inconsistent" (Lakatos 1970, 142). In addition to this pattern being characteristic of significant research programmes, it also appears to be quite frequent: "For instance, Copernican astronomy was 'grafted' on to Aristotelian physics, Bohr's programme to Maxwell's" (Lakatos 1970, 142), and Newton's programme was grafted on to "Cartesian push-mechanics" (Lakatos 1970, 145). A significant aspect about this grafting is the interesting observation that newer programmes can initially contradict older programmes to which they are appended. Finally, another way that a new research programme may be related to an older one is by having an earlier research programme's hard core subsumed in its own hard core. An example of this comes where the Copernican "hard core was incorporated in the ... Newtonian research programme" (Lakatos and Zahar 1978, 184). One assumes that part of impetus for the new programme must have then come from the older. Kepler's and Galileo's research programmes provide another example in that "They took off from the point where the steam ran out of the Copernican programme" (Lakatos and Zahar 1978, 188), while 84 keeping Copernicus's hard core (189). Furthermore, both abandoned the old heuristic (189), while Kepler, at least, added a new one (188). To summarize, Lakatos hints at a few different ways that a research programme can begin, and there are two main sorts of relation between a new programme and an older programme-acceptance of some of the tenets, and rejection of them. 3.6 Size and Scope of Established Research Programmes In rewriting specific scientific episodes in terms of Lakatos's research programmes, it is necessary to situate them properly in order to be true to Lakatos's methodology. In the following I will show that Lakatosian research programmes can be of different scopes or sizes. Why considerations of grain are important for present purposes, particularly Lakatos's apparent allowance for differing scales or scopes of research programmes, is that it means that case studies of manageable size, such as that provided by the Galapagos finch competition controversy, will be amenable to Lakatosian treatment; it is these considerations that make it possible to assess Lakatos's ability to account for a research programme's progressive elements. The first indication that Lakatos takes different sizes or scopes of research programme to count equally as proper research programmes comes when he suggests that "Even science as a whole can be regarded as a huge research programme" (Lakatos 1970, 132). It is clear therefore that he is committed to the possibility of his methodology being applied to units larger than those he ordinarily accepts as proper research programmes. 85 Here the research programme must be larger both in terms of size and scope. First, science 35 as a whole must have a larger list of theoretical commitments than each of the various scientific fields. At the same time, the whole of science, understood in this way, will have a larger scope insofar as its subject matter will be more extensive. That differences of size or scope are allowable is further illustrated in the following example, which demonstrates both a supra-research programme and a sub-programme nested within it. Lakatos appears to consider both to be proper research programmes in spite of the differences in size and scope. In 1933-4 [Fermi] reinterpreted the beta-emission problem in the framework of the research programme of the new quantum theory. Thus he initiated a small new research programme of the neutrino (which later grew into the programme of weak interactions). (Lakatos 1970, 170) In this passage Lakatos demonstrates that he considers different sizes and scopes of research programme to count equally as proper research programmes. Here, for instance, a programme with a commitment to neutrinos has a smaller scope than a research programme centred on whatever he considers the hard core of the "new quantum theory" to have been at that time. At the same time, a nested research programme having a smaller scope in this way simultaneously has a larger hard core in that it includes the hard core commitments of the programme of which it is a sub-part, as well as its own new theoretical commitments. Consistent with this interpretation of the passage (i.e., in terms of it being possible to 86 consider research programmes of different sizes and scopes as proper research programmes), Lakatos makes further references to Heisenberg's "new big programme" (ibid., 172), underscoring here its size, and contrasting it in the same sentence with the "neutrino programme which promised to fill a sensitive gap "-in the bigger programme, presumably. If we can read the passage in this way, it shows that the neutrino programme was to be thought of as both subsumed by the larger programme, and as its own partly-autonomous programme. The Fermi passage quoted above also demonstrates an important element of flexibility: one research programme can sometimes be a component of another (larger) research programme. Consistent with this interpretation is the following: because Lakatos considers Fermi here to have initiated the neutrino research programme by means of "reinterpret[ing] the beta-emission problem in the framework of... the new quantum theory", the only interpretation seems to be that Lakatos allows for the neutrino programme to be a research sub-programme of the larger new quantum research programme. In short, Lakatos is apparently amenable to research sub-programmes. At the same time, Lakatos calls the neutrino research programme "Fermi's daring application of Heisenberg's new big programme to the nucleus" (1970, 172). What is noteworthy here is that a sub-programme gets to count as an application of the larger programme. This is interesting at the very least because, in demonstrating that a sub-programme, for Lakatos, might concurrently be an application of the larger programme in which it is embedded, it shows that a single element of a scientific episode (described in different ways) might be representable by more than one of the elements in Lakatos's methodology of research programmes. 87 To reiterate a point about the structure of research programmes and how they begin, it appears that there can be either a "grafted on" relation of the smaller research programme to the larger, or the larger research programme can have an imbedded sub-programme that is nevertheless considered to be a proper research programme in its own right. 88 Chapter I V Evaluating Lakatos's Account In this chapter, I first specify some criteria given by Lakatos and Hacking for applying Lakatos's methodology to specific cases. Then I give a brief overview of the Galapagos finch competition controversy in Lakatos's terms. Next there is an extensive analysis of the Lack case. This includes discussion of the larger evolutionary context, a sorting out of how to characterize competition and adaptation, the Lakatosian elements of Lack's programme and, finally, the theoretical and empirical progress provided by Lack's work. Interestingly, Lakatos's methodology renders Lack's work not progressive in itself, although it will count as empirical progress toward the larger neo-Darwinian research programme. A Lakatosian understanding of Lack's research programme does, however, have Lack's small programme make progress when it is taken up and extended by other researchers. The programmes of the latter researchers, Abbott, Abbott, Schluter and Grant will be detailed as well. Competing programmes, in the form of Bowman's programme and that of the stochastic theorists also enter the fray. The question of whether the stochastic theorists are best seen as forming part of a single programme is addressed. A considerable amount of space is devoted to assessing whether each of the sets of researchers have individual research programmes. This is essential, given that Lakatos requires progress to be capable of evaluation only against the background of a research programme. 89 4.1 Applicability of Lakatos's Methodology to the Resource Competition Case It is important first to demonstrate that Lakatos should, for consistency, count the Galapagos finch competition controversy as amenable to his methodology of scientific research programmes. "To test [the] supposition" that "knowledge grows by the triumph of progressive programmes over degenerating ones ... we select an example which must prima facie illustrate something that scientists have found out" (Hacking 1981, 139-140). In our present case, scientists have found out many facts about the ecology of Galapagos finches; much of this data, as we shall see, will not count toward progress for Lakatos, but presumably it would nevertheless count toward Hacking's "something ... found out" criterion. Surely at least one discovery of our finch competition case should count as something scientists have found out, even by fairly strict criteria: that competition plays a role in the diversification of the Galapagos finches. This finding, in turn, can be broken down into at least two aspects: Lack's contributions and the contributions of Schluter and Grant (1984). We have seen in Chapter II that Lack's 1947 book was understood to be a landmark in bringing ecological thinking into the evolutionary synthesis (Givnish 1977, Sulloway 1982) and in particular, in suggesting the importance of competition (Abbott et al. 1977). Presumably, then, competition's applying to the finches should count both as something that a scientist has discovered and perhaps even as progress according to Lakatos's criteria. (It may count as a discovery without being progress, but if it were progress, some elements of that progress would have to include its counting as a bona fide discovery.) 90 Hacking has a stronger condition as well for cases that are candidates for Lakatosian treatment: "the example should be currently admired by scientists or people who think about the appropriate branch of knowledge" (Hacking 1981, 140). This condition would seem to be met in Lack's case for reasons just mentioned, and the Schluter and Grant (1984) results appear to satisfy this condition as well. Schluter has been inducted into the Royal Society (of the U.K. and Canada) for his research on evolution, and in his own estimation his most important work to date has been the 1984 paper (Schluter 1999; 2001). In their review of character displacement, Taper and Case suggest that the only hard evidence for this phenomenon occurs in work on Geospiza fortis and G. fuliginosa on Daphne, one of the Galapagos islands, and they attribute this result to the work of Lack (1947) and Schluter and Grant (1984) (Taper and Case 1992, 77). Surely if it is considered at least by some as the only actual demonstration of this phenomenon-a phenomenon first argued for by Darwin-then it must count as an important scientific discovery. As we have seen in Chapter II, Futuyma takes their result to apply reliably more broadly within the Galapagos archipelago than just to Daphne (Futuyma 1998,219).36 As an additional condition governing the proper application of Lakatos to specific episodes from science, Hacking says that "Having chosen an example we should read all the texts we can lay hands on, covering a complete epoch spanned by the research programme, and the entire array of practitioners". (Hacking 1981, 140). Furthermore, acting in accordance with the internal history (as Lakatos understands the term) which is associated with Lakatos's methodology, means that "Within what we read we must select the class of sentences that express what the workers of the day were trying to find out, and 91 how they were trying to find it out" (Hacking 1981, 140). It would be impossibly onerous for a research programme of the size and scope of neo-Darwinian evolution to meet these conditions. However, so long as it is acceptable to apply the Lakatosian machinery to a subset of a larger research programme-as I argue below-the finch competition case is amenable to this sort of treatment. Lakatos himself says that "The historian who accepts this [i.e., Lakatos's own] methodology as a guide will look in history for rival research programmes, for progressive and degenerating problem-shifts" (Lakatos 1971, 102). The existence of rival research programmes, then, can be taken as one of the most important conditions that a case must meet in order for it to be a candidate for a progressive shift in Lakatos's sense. In the Galapagos finch competition dispute there are conflicting research programmes, so long as we can assume that the conflicting positions are not too small in grain and scope to count for Lakatos as (rival) research programmes; the validity of this assumption will be addressed below. Specifically, there are two or three mini-research programmes in the Galapagos finch competition controversy: the competition programme, as instigated by Lack; the floristic programme as endorsed primarily by Bowman, and the stochastic programme endorsed by a few different research groups. The latter may or may not constitute its own programme, as we shall see. Again according to the above quote, we must also find in our case both "progressive and degenerating problem-shifts." I will demonstrate below that these can be discovered in the relevant case and translated into Lakatosian constructs. 92 4.2 Overview of the Case in Lakatosian Terms Here, then, are the components of our scientific episode in broad brush strokes. The hard core of Lack's research programme is comprised of neo-Darwinian hard core commitments (viz. adaptation to resources), plus competition and the allopatric model of speciation in the Galapagos. Lack's programme considered as its own separate programme will not be seen to make progress until it is taken up by other workers. However, when understood as an auxiliary hypothesis of neo-Darwinism, it will provide empirical progress for the larger research programme in which it is embedded. Bowman's hard core includes the neo-Darwinian hard core, plus the negation of competition, at least in the case of the Galapagos finches. His core also includes the presence of a commitment to finch adaptation to food resources available on the Galapagos islands as processes influencing the species distribution. His work will be seen to count as progressive within the neo-Darwinian programme in a qualified way. Abbott, Abbott and Grant take both Lack's and Bowman's mini-programmes to apply to the Galapagos finch case. They accept all the significant elements of both hard cores except for Bowman's commitment to non-competition. Accordingly, we can see Abbott et al. as operating in both programmes, or as committed to a new programme that amalgamates the cores of both. They make progress that counts toward all three of Lack's and Bowman's and neo-Darwinism's programmes. When we come to the stochastic challenge, we see that in the case of Strong et al., at least, their commitment can be seen more as a positive heuristic. They are willing to 93 entertain, at least for the sake of argument, either of Lack's or Bowman's hard cores, and on their basis, make progress of a sort for each of these programmes. While all three of the stochastic groups touched on here are broadly committed to a stochastic hypothesis of some sort, they cannot really be understood as part of a congealed research programme. Finally, Schluter and Grant bring some innovative new predictions to bear, thus contributing to the progress of both Bowman's and Lack's small original programmes. In what immediately follows I focus heavily on Lack's work. This is because there are issues of interpretation that arise in the context of Lack's programme that are relevant to the programmes that follow him. Furthermore, Lack can be seen as instigating a research programme that others subsequently adopt and defend. Additionally, in the rest of the chapter, I devote a substantial amount of space to determining whether various parts of the Galapagos finch competition controversy are research programmes or not. This is essential because of Lakatos's commitment to progress's occurring only against the background of a research programme. 4.3 Research Programmes and Progress in Lack's Work 4.3.1 The Larger Evolutionary Context It is valuable to digress momentarily to fi l l in a bit of background here. John Losee considers the principle of natural selection to be a Lakatosian hard core principle (Losee 1993, 229), although Lakatos himself nowhere mentions natural selection or, for that matter, 94 any biological cases at all. Robert J. Richards also suggests that Lakatos would consider natural selection to be a "core principle" of "Darwin's conceptual system" (1981, 65). I agree that natural selection must form at least part of what Lakatos would consider to be the hard core of an evolutionary programme. Certainly Darwinian evolutionary theory and the neo-Darwinian evolutionary synthesis have selection at their cores, as well as presumably a few other core commitments most of which will not concern us here. Lack's work post-dated the neo-Darwinian evolutionary synthesis. Chapter II traces the specific elements of evolution that end up being relevant to the most recent work detailed there (i.e., Schluter and Grant [1984]), and briefly details the neo-Darwinian evolutionary synthesis. As we have seen there, Futuyma lists twenty points that might be understood as the foundational commitments of neo-Darwinian theory (Futuyma 1998, 26-7). Each of these points is a candidate for inclusion in the hard core of the synthetic theory. I will not argue here for or against considering each of these points to be theoretical commitments of a Lakatosian hard core of modern evolutionary theory, although it appears to be consistent with Lakatos's methodology that there could be this many hard core commitments. The two most fundamental concepts in the synthesis are natural selection and genetics (Futuyma 1998, 24; Mayr 1982, 567; Hale 1995, 593). Accordingly, I will take as hard core commitments of modern evolutionary theory at least these two elements: the occurrence of natural selection and the genetic transmission of phenotypic traits. For my purposes, it is natural selection that will get the most attention. Adaptation is a consequence of natural selection, and it counts equally as a hard core commitment. It was stressed both by Darwin and by the neo-Darwinians. 95 As we have seen in Chapter II, we can understand "two major theses" of Darwin's The Origin of Species (1859) as natural selection and descent from a common ancestor (Futuyma 1998, 21). As we have also seen, Darwin himself recognized the role of competition in causing species' characteristics to diverge (Weiner 1994, 141; Darwin 1859, quoted in Lack 1947, 115). First I want to assess what Lakatosian role competition might play in Darwin's account, since Darwin did recognize its operation in species divergence. Mayr argues convincingly (Futuyma 1998, 21) for the existence of "a whole set of more or less independent theories" in The Origin of Species (Mayr 1982, 426). Each of the theoretical commitments of the five theories that Mayr isolates might best be understood as a hard core commitment of Darwin's programme. Mayr presents character divergence caused by resource competition as part of the solution to a criticism facing one of Darwin's five isolable theories: gradualness of evolution (Mayr 1982, 509). The criticism was that the gradual changes that Darwin was committed to would seem to imply that there would be a continuous gradation of morphology among taxonomic groupings. However, this is not what we see. Darwin "explained the gaps between genera and still higher taxa by the dual processes of character divergence and extinction" (Mayr 1982, 509). Thus character divergence caused by competition for similar resources is more appropriately understood as an auxiliary hypothesis: if Mayr is right, it is clearly being used here as a defense of gradualism, one of Darwin's hard core commitments. Defending the hard core is one of the main functions of auxiliary hypotheses (Lakatos 1970,133). It is also the function of the heuristics to defend the hard core, but they do so by "giv[ing] advice about what methods to employ to achieve some end" (Hacking 1981, 131). Clearly, resource competition is not an 96 aspect of advice on how to protect the hard core. It is a hypothesis about what relationships obtain in the world, and so is most appropriately understood, not as a heuristic but as one of the theoretical components of the research programme. So we see that there are good reasons for understanding competition to be an auxiliary hypothesis of Darwin's programme. Consistent with its being an auxiliary hypothesis for Darwin's programme is the observation that the role of competition in adaptive radiations was not emphasized by evolutionary biologists until after Lack (Abbott et al, 1977, 153)-when the dominant evolutionary research programme was neo-Darwinism. Indeed, the term "character displacement" was not coined until the work of Brown and Wilson in 1956 (Futuyma 1998, 554). For all of the foregoing reasons, it is appropriate to view competition within Darwin's programme as an auxiliary hypothesis of his programme. 4.3.2 Lack's Programme and Hard Core Commitments We have seen that resource competition is best understood as an auxiliary hypothesis relative to Darwin's research programme. If we are to take Lack's work as constituting a programme of its own, however, competition seems to occupy a different role in Lack's programme. Here I examine the arguments for and against considering Lack's work as its own programme, and competition its hard core. The main reasons Lakatos might not want to count Lack's work as a proper programme are its small size and scope and its seeming embededness within the larger neo-97 Darwinian programme. More on this state of affairs momentarily. However, all of these elements are present in Lakatos's own Fermi example, which he takes to be a proper research programme in spite of these characteristics. Lakatos says both that the "beta-emission problem" could be "reinterpreted ... in the framework of the research programme of the new quantum theory" and that this was simultaneously to initiate Fermi's "small new research programme of the neutrino" (Lakatos 1970, 170). Here the point is that a small programme, one which is at the same time understood as nested within a larger programme, nevertheless has sufficient scope to be considered a research programme in its own right. At the same time, it is operating both as an independent research programme and as part of the larger research programme in which it is embedded. Furthermore, the continuing role of the neutrino in quantum physics seems to be of approximately the same magnitude as the 38 continuing role of competition in evolutionary biology . So the scale and scope of Lack's resource competition seems to be similar to Lakatos's Fermi example. I conclude that it is consistent with Lakatos's commitments to understand Lack's work as constituting its own research programme. The next question is what role competition plays in Lack's programme. For all of the reasons just given, it would seem that we can understand Lack's commitment to competition, as compared to the larger neo-Darwinian programme, in a similar light as Fermi's commitment to the neutrino within quantum physics. Lakatos does not make explicit that Fermi's commitment to the neutrino is the hard core of his new programme, but this is the likeliest interpretation. Lack's commitment to competition will therefore count as a hard core commitment of his small research programme. Further, we have seen that one of 98 the characterizing features of Lack's (1947) work was its commitment to competition's role in the Galapagos finch evolutionary radiation. This commitment seems best understood as a 39 hard core commitment. However, because the role of competition in Darwin's programme is as an auxiliary hypothesis, it is worth assessing whether it would better be understood as an auxiliary hypothesis within Lack's sub-programme as well. The main function of auxiliary hypotheses is to provide a refutable protective belt surrounding the irrefutable hard core. This role does not seem to capture the significance of competition in Lack's programme. Those who came after Lack saw competition as a major commitment of his work. This by itself does not guarantee that it was intended as such by him, although it provides circumstantial evidence for that conclusion. Another reason for regarding competition as a hard core commitment for Lack is that he was still defending competition in 1971 in his last publication on the finch colonization of the islands (Grant 1986, 294). If it had not been a hard core commitment when he wrote the (1947) book, at least it seems to have hardened into part of his hard core in the meantime. For simplicity, I will take it that one of the hard core commitments of Lack's programme was competition, and that by the time his 1947 book was published, the programme with its hard core had been properly established. Thus we can take Lack's hard core commitments to be those of the neo-Darwinian programme plus resource competition. It seems plausible, then, that Fermi's commitment to the neutrino would count for Lakatos as a hard core commitment of Fermi's small programme. What is less clear is how Fermi's commitment to the neutrino was to be understood relative to the larger quantum mechanics programme. It is a hypothesis that certainly seems to function as an auxiliary 99 one, in the standard sense of the word, in the quantum mechanical programme. It is also hard to imagine what other element of Lakatos's machinery it would count as, since it does not seem to be sufficiently central to quantum mechanics to count as a hard core commitment at the time it was introduced. Furthermore, it is a hypothetical commitment rather than a suggestion for how to proceed. At any rate, it could be understood as providing part of a protective buffer around the hard core, which is one of the functions of auxiliary hypotheses. At any rate, a commitment to resource competition and a commitment to the neutrino will fulfill parallel roles relative to their respective research programmes and sub-programmes. I conclude that competition functions as an auxiliary hypothesis in the larger neo-Darwinian programme. The findings of the next section present additional reasons to understand competition in this way. It is, however, not a wholly satisfactory state of affairs that one element (here competition) functions simultaneously as a hard core in a sub-programme of the larger programme and as an auxiliary hypothesis in the larger programme. Whatever Lakatos's considered view would be on the status of the neutrino in Fermi's sub-programme and in quantum physics, it is unlikely that he would reject a demonstrably parallel role for resource competition in Lack's and the neo-Darwinian programmes. Let us review. It appears that Lack's commitment to competition's being a major factor in the ground finch radiation functions as a sub-programme of the neo-Darwinian programme. It can be viewed simultaneously as the hard core of its own programme and as acting somewhat independently of the neo-Darwinian programme in which it is embedded. We have seen that resource competition before Lack counted as an auxiliary hypothesis for 100 the Darwinian programme. Resource competition would also be understood as an auxiliary hypothesis of the neo-Darwinian programme prior to Lack's work, at least. I have not yet mentioned Lack's four-stage allopatric model of speciation of the Galapagos finches. Grant (1981) calls it the "Darwin-Stresemann-Lack model" (Grant 1981, 654), which underscores for us that it was not entirely new to Lack. Indeed, Mayr, whose name Grant does not include here, would most appropriately be understood as having allopatric speciation as one of the hard core commitments of his own neo-Darwinian sub-programme, given that the allopatric model of speciation was most emphasized by Mayr (1963) (Futuyma 1998,482). Thereafter, Stresemann and Lack co-opted allopatric speciation (mainly from Mayr) to explain part of the Galapagos finch radiation, and Lack added the competition emphasis. Even to the extent that allopatric speciation was inchoate in Darwin and presumably a hard core proper of Mayr's programme, Stresemann would have been the first to apply it to Darwin's finches,40 and so it is appropriately understood as either the hard core of a new research programme at that time, or as a research programme that was starting to harden. Accordingly, we might view this four-stage speciation model as also one of Lack's hard core commitments. Although it seems most in keeping with Lakatos's methodology to understand Lack's hard core as containing both competition and the four-step model of speciation, it is the former that has mainly been emphasized in the workers carrying on with his programme; it is for this reason that the allopatric model was not emphasized earlier. Additionally, the allopatric model will have a minimal role to play in the analysis that follows. 101 4.3.3 Character Displacement as a Kind of Adaptation There is still a tension to be resolved regarding competition in the research programmes of Darwin and Lack. When competition is a factor in a divergence, it leads to a difference in character states via one or both of two processes: competitive exclusion and character displacement. Character displacement caused by resource competition in particular is a kind of adaptation (Schluter 2000, 69). If we are to understand competition as a hard core for Lack's programme, we are led to a seeming inconsistency since adaptation ought to be understood by Lakatos as part of the hard core commitments of Darwin, but the above analysis on the basis of Mayr (1982) suggests that character divergence caused by competition would more appropriately be understood as an auxiliary hypothesis of Darwin's theory. The resolution of this seeming inconsistency is quite simple. Although " [ compet ing species are simply part of the environment," I suggest we take to heart the usefulness of "distinguish[ing] the impact of interactions between species of the radiating lineage itself from that of external factors," e.g., aspects of the environment excluding congeneric competitors (Schluter 2000, 69). So even though competition between organisms for resources is one of the kinds of selection that organisms may be subject to, we can distinguish between kinds of adaptation for the purposes of hard core commitments. So one of the hard core commitments of neo-Darwinism is natural selection for adaptation to resources. Selection in the presence of competing organisms does not get enshrined in the 102 hard core, but counts as an auxiliary hypothesis both for Darwin's programme, as we have seen, and for the neo-Darwinian programme. 4.3.4 Other Lakatosian Elements in Lack's Work I turn now to the other components of Lakatos's model as applied to Lack. Lakatos does not specify the negative heuristic in even his most detailed illustrative example, that of Bohr, although he does talk about it in the case of Newton. Anyway, we can take Lakatos's minimum negative heuristic to be an exhortation to defend the hard core. In the case of Lack, we can see this as an exhortation to defend the most strongly-held tenets of neo-Darwinian evolutionary theory and the competition hypothesis. The evidence upon which Lack's hard core commitment is largely based does not count as any of the elements of Lakatos's methodology of research programmes. We might understand the positive heuristic of the programme as an implicit exhortation to take Galapagos finches as the subject matter of the programme. In the last chapter, we saw that there are five roles specified by the positive heuristic. Of these five, only two are fulfilled by the suggestion to study the finches: defining problems and giving the researcher(s) a modicum of guidance in how to proceed. It appears initially that Lack's programme has few components that would be understood as auxiliary hypotheses prior to other researchers taking it up. This is something that requires closer examination. To clarify, although auxiliary hypotheses were assessed 103 above with respect to whether Lack's competition hypothesis was one, here I am assuming that competition does form part of Lack's hard core. One sort of auxiliary hypothesis one might expect to find, given that the function of auxiliary hypotheses is to deflect falsifiability from the hard core, is counter-counter-arguments. That is, arguments that address actual or possible counter-arguments against the hard core. In his book, Lack almost never resorts to this second level of justification in any of the passages in which he suggests that competition helps to explain the finch biogeography. Instead, he seems most often to take evidence of the patterns of morphological difference as sufficient justification. There is one exception, however, to the generalization that Lack does not address possible counter-arguments to competition's purported workings. Lack offers that the "meeting of two forms in the same region to form new species must, when both persist, result in subdivision of the food or habitat, and so to increased specialization" (Lack 1947, 162). This is just character displacement again. He adds that this pattern is not seen in "land birds of single oceanic islands" (ibid.). Here Lack might be seen as addressing a potential counter-argument that would underscore the rarity of the patterns he has observed. The reason these other cases do not apply is, he claims, "because these provide no opportunities for the formation of new species in geographical isolation" (ibid.). Furthermore, "[ijnstances are rare even on other archipelagos, probably because most archipelagos are too accessible to fresh colonization by more efficient birds from outside areas" (ibid.). The reasons Lack gives here could be understood as auxiliary hypotheses on Lakatos's account insofar as they 104 deflect possible counter-arguments (i.e., the rarity41 of other such examples) from Lack's hard core commitments. So we see that aside from a hard core, Lack's programme has only an implicit negative heuristic, and a positive heuristic that does not give much guidance.42 In other words, Lack has very little in the way of auxiliary hypotheses. 4.3.5 Predictions and Progress in Lack The instigation of a new programme by itself does not count for Lakatos as progress. It is often after a programme successfully makes novel predictions of phenomena that it can be understood as progressing. Further, predictions are usually made by the auxiliary hypotheses. There are no new predictions explicitly appended to the defensive arguments mentioned in the last section. However, if admitting implicit predictions is consistent with Lakatos's commitments, we might understand Lack's work as containing some of these. That is, it might be implicit in Lack's conclusions that if ecologists were to discover an archipelago having similar conditions to these of the Galapagos Archipelago, then similar results would obtain. Presumably it would count as a novel prediction for Lakatos that adaptive radiations of species that had not previously been detailed would exhibit similar patterns of resource partitioning on the basis of their differential morphologies. This is similar to Bohr's predictions of the "wavelengths of hydrogen's line emission spectrum" (Lakatos 1970, 147), although more approximate. The predicted wavelength series would 1 0 5 be comprised of quantitative predictions, whereas the sort of implicit prediction I am postulating on the basis of Lack's work would be more qualitative. However, given that the predictions are not made explicit, presumably Lakatos would have to view the auxiliary hypotheses that bear the supposed implicit predictions as ad hoc auxiliary hypotheses. This is because Lakatos considers auxiliary hypotheses to be ad hoc when even they "have no excess empirical content over their predecessor" (Lakatos 1971, 125; footnote 36). Excess empirical content comes in the form of predictions of novel empirical phenomena. If these are not made explicit, presumably they do not count toward having "excess empirical content." In short, these auxiliary hypotheses with their possibly implicit predictions would not count toward progress within Lack's programme. Lack does briefly compare his results to those seen in other groups: "The manner of evolution of Darwin's finches is peculiar in some details, but is considered to be fundamentally typical of that which has occurred in many other organisms" (Lack 1947, 162). Lack's reference to the evolution of other taxonomic groups is consistent with understanding Lack's programme as an attempt to understand competition as an explanatory principle in at least some adaptive radiations, but not confined solely to the finch case. At any rate, noting these other cases in which there are some similarities with his own case will not count as progressive for Lakatos, since Lakatos counts only obtainments of novel predictions, and clearly here Lack refers to established cases. Further, they are not cases that support competition, but adaptation more generally. That is, I take the peculiarities mentioned in this passage to be the workings of competition, whereas the way in which they are similar to other groups is in their adaptation. 106 In addition to the considerations just mentioned, it is worth also assessing whether there are instances in which Lack's programme might be seen as making predictions when considered only in relation to the finches. Lack takes the evidence provided by all of the Galapagos finches and bases his conclusions upon that, rather than delaying data collection for some of the finch species in order to see if they, too, follow the same pattern. There is a trivial sense in which one could "postdict" that two species X and Ton an island would demonstrate a pattern of resource competition, but this is dissatisfying given that the evidence from X and 7 was part of the pool of data used to build the original hypothesis. Furthermore, that is exactly the kind of situation that Lakatos seems to be attempting to rule out by his commitment to novel predictions and, in particular, to his making theoretical progress dominant over empirical. Zahar puts Lakatos's commitment succinctly: "facts provide little or no evidential support for the theory, [when] the theory was specifically designed to deal with the facts" (Zahar 1976, 218). Lakatos was amenable to the emendation proposed by Zahar redefining the notion of novel fact, which is what progressive programmes are supposed to predict. "A fact will be considered novel with respect to a given hypothesis if it did not belong to the problem-situation which governed the construction of the hypothesis" (Zahar 1976, 218). This dispenses with the requirement that the prediction of a novel fact must occur prior to the discovery of the same fact. As just noted, however, the only relevant facts in the Lack case were those used to derive the theory supporting any novel predictions that might be made. Consequently, they would not count as "novel" in Zahar's amended sense, either. 107 For Lack, as probably for many scientists, there is a reciprocal relation between evidence or experiment and the hypotheses that are extracted from them. In the case of Lack, the relevant hypothesis is the hard core itself. It appears that Lack did not set out in advance to defend his competition hypothesis; indeed his 1945 paper argued "that the intra and interspecific morphological variation among islands was ... merely the result of genetic drift and founder effect" rather than competition (Abbott et al. 1977, 153).43 Two years later, he writes, "My views have now completely changed" (Lack 1947, 62). So it appears that Lack did not set out with his hard core firmly in place prior to gathering the data that he describes in 1947. Therefore, to the extent that the data corroborates his hard core, it cannot be seen as something that was predicted on the basis of a firmed up hard core. Accordingly, this type of evidentiary corroboration does not count toward progress for the reasons given in the last paragraph. Before concluding that Lack's programme does not make the requisite kind of predictions that Lakatos would take as necessary for its counting as progressive, there is yet one more consideration to assess. Perhaps it makes sense to suggest that during the course of his study, Lack made predictions on the basis of his commitment to competition, which were then borne out when he made the requisite observations. This process would most likely not be reflected in the finished book, and is an instance in which obeying Lakatos's dictum to obtain all the written information by the various players could have an effect on the outcome of the understanding of the Galapagos finch competition controversy in Lakatos's terms. However, I take Lakatos's requirement for novel predictions to be of a 1 0 8 broader scope than just their relevance in the initial reciprocal interaction between evidence and hypotheses as seen in Lack. Evidence for this suggestion comes from the Bohr case again. The first model in Bohr's progressive programme predicted "the wavelengths of hydrogen's line emission spectrum" (Lakatos 1970, 149). This prediction is roughly of the same scope and magnitude as a (hypothetical) prediction that the next Galapagos finch discovered will have a pattern suggesting character displacement. More relevant to present purposes, though, Lakatos notes that some of these wavelengths were already known, but that the model predicted others in addition to the known ones. Furthermore, it was not Bohr himself who "corroborated its novel content" (ibid.), unlike our present case. Later workers such as Abbott et al. (1977), and Schluter and Grant (1984) give additional reasons for agreeing with Lack's original conclusions, but these were not a matter of discovering new finch species that fit the predicted pattern. So if these are to count as corroborating the novel content of Lack's programme, this would have to be understood in a more subtle way than it seems Lakatos can manage. In the context of Abbott et al., I will examine the extent to which these other evidential considerations might count as the obtainment of Lack's novel predictions. In summary, I want to suggest that in none of the ways one might understand Lack as making predictions does he do so in a way that Lakatos would count as leading to progress. This is not to assert that Lack's work is unimportant, and it does not dispute that the work is taken by scientists to have provided good evidence and arguments for the existence of the processes of competition in the Galapagos finch fauna. It is merely to 109 suggest that Lakatos does not have the machinery to accommodate the elements of progress made by Lack. So we have seen that there is good reason to suppose that Lack's programme, considered in isolation, would not count as progressive for Lakatos. 4.3.6 Lack's Programme: Empirical Progress for Neo-Darwinism In understanding Lack's programme as an individual programme in isolation from others, i.e., before his work was taken up and carried further by other researchers, it is best not to understand it as predicting new phenomena. However, as we have seen, Lack's programme can also be understood as a sub-programme nested within the larger neo-Darwinian programme. Here I will be concerned with ascertaining whether understanding Lack's small programme in this way might make elements of it progressive within the larger programme. For a programme to be progressive, it needs to be both theoretically and empirically progressive. If we can take Darwinism to be predictive at least in suggesting that examples exhibiting evolution by natural selection and adaptation will be discovered, then it is theoretically progressive to that extent. Neo-Darwinian theory will make similar predictions, but in addition, those predictions will include a genetic component. Once convincing new cases exhibiting natural selection have been discovered, then Darwinism's empirical progress has caught up to its theoretical progress and the whole programme becomes progressive. This is where cases such as Lack's become relevant. Lack's (1947) work is primarily evidentiary. First, adaptation to resource competitors is a kind of adaptation and Lack has shown it to obtain. Second, Lack's evidence suggests competition 110 between congeners for the same resources as an element of their evolutionary radiation. If we can understand Darwin's auxiliary competition hypothesis as predictive, then Lack's evidence corroborates it. In both of these ways, Lack's sub-programme contributes to the empirical progress of Darwinian and neo-Darwinian evolutionary theory. So Lack cannot be seen to have made progress relative to his own programme, although relative to Darwin's, he can be seen to make empirical progress only, which is a necessary, but not a sufficient, condition for progress proper. If, however, we understand Lack's programme as being extended and continued by those who followed him and accepted resource competition as a hard core commitment, then we might be able to detect progress in the subsequent development of his programme. 4.4 Competing Research Programmes We saw earlier that one of the conditions that must be found in any case that is amenable to Lakatosian treatment is that there must be rival research programmes. Here I will be dealing with the floristic programme as endorsed primarily by Bowman, and the stochastic programme endorsed by a few different research groups. There is first of all a line of reasoning based on some of Lakatos's commitments that makes both of these rivals to Lack's programme. I will turn later to considerations that suggest otherwise. Lakatos suggests that "behind any alleged single battle between theory and experiment, there is a hidden war of attrition between two research programmes" (Lakatos 1971,102). This is how the methodology of research programmes would accommodate the same incident that a falsificationist would take as a "crucial negative experiment" (ibid.). One of the two main foci of Schluter and Grant (1984) is their five-way test. I suggest that this might be understood along falsificationist lines as a five-way crucial experiment. I will return to Schluter and Grant's research programme below. For present purposes, what is relevant is that if their result were to be viewed as a crucial experiment by falsificationists, Lakatos would describe it as a "hidden war of attrition" between research programmes. Here three of them: resource competition, adaptation to floristic resources, and stochastic colonization of the Galapagos. Basically, Schluter and Grant assembled the various published positions on the radiation of the finch fauna of the Galapagos, and tested them using computer models against the data they had on a subset of that fauna, the Geospizae. The first two computer models both "involve random colonization of islands without evolution of beak depths." The second takes into account the "expected [population] density associated with their beak sizes" (Grant 1986, 338). In other words, adaptation to the available food resources is factored in. The third model takes into account the ability of finch beaks to evolve to local adaptive peaks; here if a finch species colonizes an island but has a sub-optimal beak for the food resources present there, its beak will adapt to the food source. The fourth and fifth model competition: competitive exclusion in the fourth, and finally competitive exclusion plus character displacement in the fifth. Clearly it is Bowman's floristic hypothesis and the stochastic hypothesis that are being compared to the competition hypothesis. This argues for their counting as rival research programmes on Lakatos's account, given his commitment quoted at the beginning of this paragraph. 112 4.4.1 B o w m a n Bowman certainly challenges the importance of competition to the radiation of the finches, and it is clear that his hard core must be centred on a commitment to the finches' adaptation to the food resources provided by the flora of the island. For the same reasons that Lack's programme counted as a proper research programme on Lakatos's account, this one does as well. In his summary of the most important factors influencing the distribution and morphology of the Galapagos finches, Bowman suggests three: feeding adaptations, the influence of predators, and "[t]he genetic constitution of the ancestral colonists" (Bowman 1961, 292). All three of these factors may count as hard core commitments in Bowman's research programme, but the adaptation to food provided by the available flora is the main focus of his book and is most relevant to our purposes, so I will ignore the others. With respect to the third of his conditions, Bowman argues that "the failure of certain food specialists to have evolved ... [is not due] to the presence of 'ecological equivalents' from the mainland" (ibid.). This is explicitly a questioning of the role of competition (here from non-congeners) in leading to the observed morphologies. A more specific response to Lack's competition hypothesis is seen in a six-page section Bowman devotes to refuting it. What is clear throughout that section is that Bowman does not consider competition a required part of the explanation for the Galapagos finch distribution. The most that he will grant the competition hypothesis is the following: 1 1 3 Anatomical differences between closely related species on any one island are best thought of as biological adjustments that have been evolved when the forms were in isolation. These adjustments could have prevented "competition" from occurring between the forms when subsequently they came together on the same island. (Bowman 1961, 296) In other words, Bowman is committed to there being no character displacement in the finches upon secondary contact. The adaptations they demonstrate can be wholly accounted for before the incipient (or true) species met again in sympatry. Bowman acknowledges that there are sufficient differences in these sympatric species so that they do not in fact compete, but he denies that this is a result of any process occurring after they came into secondary contact. Detailing all of this is to demonstrate that Bowman's position with regard to the Galapagos finches is not that competition may have been one of the factors influencing the radiation we see. He is prepared actually to commit himself to the negation of Lack's hard core commitment to the role of competition. This negation counts as part of Bowman's hard core. I suggested earlier that one hallmark of at least some kinds of auxiliary hypothesis is provided to us by counter-arguments to arguments that threaten the relevant hard core commitment. In the case of Bowman, competition is the counter-argument to his commitment to floristic adaptation, and his way of countering competition in turn is by reinterpretation in terms of floristic adaptation of the data that Lack gives as evidence of 114 competition. In other words, here the hallmark draws our attention back to the hard core commitment of Bowman rather than to an auxiliary hypothesis of his programme. Bowman has gathered reams of data supporting the hypothesis of adaptation to food resources made available by local flora on the Galapagos islands. As with Lack's (1947) work, Bowman's evidence and his hard core commitment reciprocally support each other, rather than there being auxiliary hypotheses that both make predictions and provide a protective belt around the hard core. However, unlike in the case of Lack, Bowman's hard core commitment (i.e., to adaptation) seems to have been made before the data collection began. Therefore we might try to understand Bowman's hard core as making nebulous qualitative predictions regarding the fit of morphology of the finches to their floristic environments. As we have seen in Chapter III, the hard core cannot make predictions by itself, since then it would be falsifiable. I will therefore take it that when we locate predictions, they will be conjunctions of at least one hard core commitment and at least one auxiliary hypothesis. So although one might think that there would be predictions implicit in Bowman's searching for evidence to fit his hard core hypothesis, Lakatos's machinery does not seem easily able to accommodate this state of affairs. It does little good to suggest that all Lakatos would have to do to address this difficulty would be to suggest that here an auxiliary hypothesis is being conjoined implicitly to the hard core commitment. By Lakatosian fiat hard core commitments by themselves are not testable. Both Lack's and Bowman's research programmes can be seen as branching from the neo-Darwinian programme. Lakatos allows for branching as a possible genesis of 115 programmes, as demonstrated by his Pythagorean-Platonic-Ptolemaic-Copernican example in which the research programmes of Ptolemy and Copernicus began by enshrining the positive heuristic of the Pythagorean-Platonic research programme among the positive heuristics of their own programmes (Lakatos and Zahar 1978). This is then somewhat different than the Bowman and Lack cases in which new hard cores were appended to the hard core of the modern evolutionary research programme. Both Lack's and Bowman's programmes can be understood as having branched in different directions from the same neo-Darwinian hard core commitments. By way of summary of the application of Lakatosian machinery to Bowman, Bowman cannot be seen to make progress, relative to his own programme, in spite of his bringing to bear extensive new discoveries about the morphology and ecology of the Galapagos finches. When Bowman's programme is considered as an auxiliary hypothesis of the neo-Darwinian, however, it counts as giving evidential support to that programme. Specifically, it includes empirical progress, evidence catching up to the predictions of adaptation that were made in Darwin's time. This is, of course, only if we are allowed to understand the Darwinian (and the neo-Darwinian) programme as (implicitly) predicting that cases exemplifying adaptation will be found. This is a more qualitative sort of prediction than any Lakatos gives, so for this reason, at least, it is not certain that Lakatos would be happy to have his methodology applied so as to have neo-Darwinism count as progressive in this way. What one can question here is that qualitative, and perhaps only implicit, predictions would count for Lakatos as novel predictions capable of yielding theoretical progress. At most, then, Bowman's work would count as empirical progress only. 116 Again, empirical progress by itself does not count as progress. Neo-Darwinism counts as progressive only when it makes novel predictions (theoretical progress) and then later shows that they obtain, as with Bowman's work. 4.5 Abbott, Abbott and Grant For simplicity, I will lump the Abbott et al. paper and most of Grant's work into one research programme; this is justified because the Abbott et al. paper marks the beginning of Grant's researches into the Galapagos finches. The hard core commitments relevant to our purposes are the same in both; all of these researchers accept both Bowman's floristic hypothesis and Lack's competition hypothesis. I have demonstrated that Bowman was committed to the negation of competition in the case of the Galapagos finches. It therefore may initially appear that the Abbott and Grant research programme was built on inconsistent foundations. One might assume that this would not be allowed; in fact, Lakatos does allow it, at least in the case of the "grafting on" of a new research programme onto an older (Lakatos 1970, 142). However, this is not the case here anyway. Here we have a combination of the hard core commitments of two recent programmes combined into the hard core of either a new programme, or of an extension of both. It seems from the foregoing that Lakatos might not be against their conjunction in this way,44 even if they had been strictly contradictory. Anyway, we can take Bowman's commitment to the negation of competition as a hard core commitment separable from his commitment to adaptation to food resources. Then the combination of the other hard core commitments of these two 117 programmes is not contradictory. Abbott et al. have clearly abandoned Bowman's commitment to the negation of resource competition. Next let us determine whether the Abbotts and Grant have made progress in their research programme(s). This can be construed in either of two ways: have they made progress in their own, new programme, and have they made progress that could be attributed to either of Bowman or Lack programme? For simplicity, I will assume without arguing for it that progress for the former also counts as progress toward each of the latter where it extends the original programmes (i.e., Lack's and Bowman's), and extends them by extending their respective hard cores. (In Bowman's case, the hard core referred to here does not include the negation of competition.) Once again, the work of the Abbotts and Grant can be seen as part of the penumbra of the neo-Darwinian programme. Abbott et al. (1977) and Grant (1986) (which is primarily a review of his work to that point) continue adding to the sum of data on every aspect of the behaviour and ecology of the finches, a process that once again, in itself, does not count as progressive for Lakatos. As before, it counts as fulfilling a prediction of neo-Darwinism, and so counts as empirical progress. We saw previously that where a research programme makes some prediction, this prediction must be attributable to its auxiliary hypotheses (in conjunction with elements of its hard core). Abbott et al. "allow their data to 'choose' between" the hypotheses of Bowman and Lack (Abbott et al. 1977, 153). This is a loose form of crucial experiment. Everywhere they do this they tease out some consequence of each of the two hard core hypotheses and examine these consequences against the background of the Galapagos 118 finches. Each of these consequences of the hard core hypothesis is itself a hypothesis, but one that is not identical to the hard core. These consequences, then, count as auxiliary hypotheses within each of the two programmes, and as candidates for generating novel predictions on the basis of one or the other of the two hard cores. Clearly they count as auxiliary hypotheses unless, of course, the case is written up as though the predictions had been made in advance, when they were in fact not. If the latter, evidence found to corroborate predictions would not count as novel phenomena in Lakatos's original sense. They would not count as novel in Zahar's appended sense, either, since the evidence would then have been part of "the problem-situation which governed the construction of the hypothesis" (Zahar 1976, 218). Here it is frustrating that these results would not count toward the progressiveness of the case unless the predictions were in fact made before the relevant data were acquired. In attempting to rule out ad hoc auxiliary hypotheses from counting toward progress, Lakatos's commitments mean that he might eliminate some of what I would consider to be real progress. Let us assume for the sake of argument that these predictions were made before the evidence was either collected or collated into a form that would allow researchers to assess predictions, keeping in mind that if they were not, they would not count as progress for Lakatos. Some of these crucial-experiment-style analyses support Bowman's hypothesis, some support Lack's, and some support both. Without looking in detail at all of the many crucial experiments they set up, one can assert that both Bowman's and Lack's programmes have been progressively expanded-and so Abbott et al. have made progress for each of the two research programmes that they take over and extend. Lakatos still counts it as progress 119 where the prediction of an auxiliary hypothesis makes false predictions as well as true ones, for example as in the Bohr case where "[n]ot all the novel content of Bohr's first model M\ was corroborated" (Lakatos 1970, 147). Presumably, he also would not hold it against the progressiveness of a programme that some of the auxiliary hypotheses that it generates give false predictions, so long as some such predictions can be corroborated. Let us look at one of the examples from Abbott et al. to demonstrate both their procedure and how it counts as being progressive for Lakatos. "A corollary of the competition hypothesis is that islands with only two breeding sympatric Geospiza species should hold only two such species of very dissimilar beak depth- This is indeed correct" (Abbott et al. 1977, 164). Here we have an auxiliary hypothesis within Lack's programme (although not one that Lack himself came up with) which is simultaneously a prediction about what should obtain. Although Lack himself was aware of these gaps in beak depth of the species, there is no direct evidence that this data went into the development of his original hard core competition hypothesis. If it did, then the above prediction would have been part of "the problem-situation which governed the construction of the hypothesis" (Zahar 1976, 218), and hence would be ad hoc on Lakatos's view. If we can assume that it did not, then this result constitutes one instance of progress in Lack's extended programme. With respect to the same data set, Bowman's hypothesis would suggest the following. The alternative floristic hypothesis is that overlap between sympatric pairs is related to the diversity of food available. Overlap would tend to be highest 120 when the lowest variety of foods is available. This was tested directly by relating overlap in the seed and fruit types eaten by each sympatric pair of Geospiza species to diversity of seeds and fruits available.... No significant correlation was found... . (Abbott et al. 1977, 164) That overlap "would tend to be highest when the lowest variety of foods is available" is a consequence of Bowman's hypothesis and not identical to it. So it can be seen as auxiliary to Bowman's hard core commitment and, hence, as counting both as an auxiliary hypothesis and as a prediction. In this case, the prediction is not borne out by the data. I presume that, in addition, this prediction was made in advance of the correlation's being measured and does not count as an ad hoc hypothesis. There are also cases in which Abbott et al. set up the comparison between the two programmes by directly "testing their explicit hypotheses"-in Lakatos's terms, Bowman's and Lack's hard cores (Abbott et al. 1977, 153). An example of this sort of analysis occurs where they found "a relation between the diversity of seeds and fruits eaten ... by each of 21 Geospiza populations and the diversity of seeds and fruits" available (Abbott et al. 1977, 158). A direct test such as this in which they do not have to come up with consequences of Bowman's and Lack's hard core commitments (thus introducing the possibility of error) would seem to be at least as valuable a comparison, but for Lakatos it is not. So long as we understand the commitments of Lack and Bowman as hard core elements of their own small research programmes, to test them directly will not be possible. Qua hard core elements, 121 they are by definition non-falsifiable and so cannot have falsifiable consequences. Of course, as we have seen, both could count as auxiliary hypotheses of Darwin's larger programme, in which case they could generate novel predictions for that programme and so be seen as candidates for part of the progress of neo-Darwinism. That they count as progressive when viewed as one kind of element within a research programme but not when viewed as a different element of another programme is not a satisfying state of affairs. One might even say that it constitutes a reductio ad absurdum of the original assumption that it is possible within Lakatos's methodology to understand part of a big programme's auxiliary hypothesis as the hard core of a small programme. While I do not want to be committed to this constituting a true reductio, the problem does appear to underscore an inflexibility within Lakatos's methodology of scientific research programmes in two ways. Most obviously, it does not account well for programmes that can be seen simultaneously as different programmes even according to Lakatos's own commitments. Second, it appears to demonstrate that Lakatos's programme does not easily accommodate the smaller-grained research programmes and their progress. Abbott et al. also make much clearer, compared to Lack's treatment, the concept of character displacement. They do this in part, but not wholly, by coming up with auxiliary hypotheses. It would seem that this kind of increase in conceptual clarity is something Lakatos's system is blind to. He makes allowances for a time before a hard core is adopted. During this time, conceptual modifications may well be made to the hard core in particular, and this may account for some of the examples of increase in conceptual clarity seen in research programmes. However, in the Galapagos finch resource competition case, it would 122 seem that conceptual clarity with regard to character displacement occurred after the instigation of the programme. 4.6 Competing Programmes, Round Two: The Stochastic Theorists I turn now to the stochastic challenge to the two main commitments we have been examining. Earlier we saw that we might understand the stochastic challenge to competition both as a rival to Lack's programme and as its own research programme. This suggestion was on the basis of a comment of Lakatos's regarding what falsificationists would likely consider a crucial experiment. Let us now examine the supposed stochastic programme in a bit more depth. The first thing to notice is that if it counts as a programme, it is not an integrated one. In Chapter II, I detailed the two most relevant of the three groups which use stochastic models to account for the observed finch radiation. I will skip much of Simberloff s (1978) analysis, since it does not directly mention Bowman, Lack or Abbott et al. Simberloff does, however, argue against the competition hypothesis on the basis of stochastic factors, and this would count as an attack on Lack's hard core. Although Lack is not mentioned in Simberloff s paper, his work was most responsible for bringing competition into the modern synthesis (Givnish 1997,4) so, by proxy, an attack on competition in general is an attack on Lack's hard core commitment. Simberloff s first hard core commitment is that it is not the case that "diffuse competition plays a large role in shaping island distribution patterns" (Simberloff 1978, 715). This commitment is very 123 similar to Bowman's. Simberloff concludes that "a model of colonization ... which is purely stochastic and rests only on properties of individual species comes close to accord with some plant and insect data and can in any event be used as a baseline test to see if other phenomena must be assumed" (Simberloff 1978, 724). So it appears that another commitment (perhaps also part of his hard core) is a heuristic one: namely the recommendation to rule out the influence of stochastic factors before making conclusions about other processes influencing biogeography and, in particular about competition's role. I will say more shortly about a similar heuristic. 4.6.1 Strong, Szyska and Simberloff Strong et al. make a prediction on the basis of what should be seen if character displacement is as common in the Geospizae as Abbott et al. suggest. They first calculate ratios of the beak lengths of sympatric species selected by chance.45 Their prediction on the basis of character displacement is then that the ratios seen between actual species in situ should be higher than if those species were arrayed by chance alone. Furthermore, those ratios should be consistently higher than those predicted by chance if character displacement is a widespread phenomenon in this group. Here we see quite clearly an auxiliary hypothesis being appended to Lack's programme for the sake of testing its prediction. The prediction on the basis of this auxiliary hypothesis, however, is not corroborated 4 6 As we have seen earlier, Lakatos will count as progressive auxiliary hypotheses that have predictions that are not corroborated, so long as some of their predictions are. This seems initially not to be the 124 case here (i.e., the prediction is disconfirmed), so the addition of this auxiliary hypothesis will not count as progress within Lack's programme. There may be another way to understand this prediction, however. If one looks at each pair of ratios-the one expected by chance and the actual ratio-as a separate prediction based partially on Lack's hard core, then some of the predictions are corroborated. However, fewer are corroborated than are falsified by means of this procedure. This in itself does not count as a non-progressive state of affairs for Lakatos, since he requires for empirical progress only that "some of this excess content [i.e., the prediction] is also corroborated" (Lakatos 1970, 118). Also, it depends upon exactly what we are taking Lack's hard core commitment to be. If we were to understand Lack as committed to the proposition that at least some of the morphology of the finches was determined by character displacement47 then even finding only some ratios that are higher than would be expected by chance would count as corroboration of his hypothesis. However, Lack seems to be committed to competition's (and character displacement's) being a common, rather than an occasional, process governing radiation. In that case, testing the character displacement hypothesis in the finches as Strong et al. have done here, viz., by determining whether it constitutes a general trend in the actual finch distribution, is a test of Lack's actual commitment. Thus there is one prediction made by the respective auxiliary hypothesis and it is not found to obtain. Therefore, in the final analysis there is no progress made by this auxiliary hypothesis for Lack. It seems counter-intuitive that the same result counts as progress if Lack had one commitment, but as non-progress if his commitment was slightly different. However, this 125 will be one of the consequences of making progress contingent upon the scientists' programmes as originally conceived, rather than upon the results of their researches compared to what was known before. If we are to take their paper as centred on a research programme, it is difficult to determine what, exactly, is the hard core of the Strong et al. group's work. From one perspective, they do not seem to have a hard core of their own, so much as they are attempting to assess the hard cores of other programmes. The question guiding their research is "What sort of species-set would lead to the conclusion that character displacement had not occurred?" (Strong et al. 1979, 898). This is a Popperian sort of question, asking what would count as falsifying the competition hypothesis. On the basis of their stochastic modeling, they suggest that "only a portion of the evolution and ecology of the finches has been determined by" the factors isolated by Abbott et al., i.e. "food supply and interspecific competition" (Strong et al. 1979, 910). From this one might gather that their hard core is a negation of Abbott et al.'s hard core; that is, not a simple negation of competition or adaptation to food availability or both, but a negation of the predominance of both of these as explaining the Galapagos finch radiation. Further, Strong et al. suggest that "it is reasonable that apparent randomness be disproved before more structured or ordered versions of ecological nature are accepted as true" (910). Also, among their final conclusions, they say "that apparent randomness would account for a substantial proportion of variation in many real ecological communities, were null hypotheses employed that assumed no structure at the outset" (911). These two passages suggest a commitment primarily to ruling out randomness as the explanation for 126 species distribution in ecological communities before making stronger claims for the processes that shaped those communities. In fact, their central commitment seems to be more of a positive heuristic than a hard core per se; it is a recommendation for the right way to proceed in determining the causes of evolutionary radiations. This, i.e., having as a central commitment a heuristic rather than a hard core, is a state of affairs that is allowed for Lakatos, for example in what he calls the Pythagorean-Platonic "proto-programme'' (Lakatos and Zahar 1978, 180). However, the Pythagorean-Platonic proto-programme did have a hard core. It was secondary rather than primary (ibid.), whereas it is perhaps unclear what we should view as the hard core of the Strong et al. programme-if it counts as such. It is appropriate to add that Strong et al.'s conclusions are intended to apply more broadly than merely to the Galapagos finches. This is evidenced in the last two Strong et al. quotes as well as in their methodology: the Galapagos finches are only one of the groups they examine in their (1979) paper. With respect to Lakatos's "hidden war of attrition," in the presence of perceived crucial experiments, interestingly, it may not be a war of the one programme with an actual rival, but with a perceived rival. Let us take the five-way test as a crucial experiment, or as close enough to one so as to be subject to Lakatos's "hidden war of attrition" view. It is clear that Schluter and Grant (1984) consider there to be rivals to the competition hypothesis. Even so, it is less clear that they are arguing against their rivals' actual hard cores' obtaining than against what they perceive to be the hard core(s) of their rivals, or some generalized amalgamation thereof. This point is especially acute in the stochastic case, since there are a 127 few different research programmes with slightly different actual hard core commitments even where they can be understood as having hard cores. At any rate, the most important question is whether Strong et al. have made progress. They have. In their three cases, the Galapagos finches and two other bird groups, the novel predictions of the stochastic theorists have been confirmed. For Lakatos's account, even if Strong et al.'s having a hard core is in doubt, it is peculiar that they are able to make progress. Lakatos might take this as evidence that they do in fact have a hard core, or even that they borrow the hard cores of other programmes in order to conjoin them with their own auxiliary hypotheses in order to test the viability of those cores. It is also interesting that this case raises the possibility that if scientists do not have their own hard core, they can nevertheless challenge other research programmes. This does not seem easily accommodated by Lakatos. One way of reconciling these non-standard Lakatosian threads is to understand the stochastic challenge as a nebulous beginning or incipient research programme that has not yet settled upon a unified hard core. It might be the case that what I am here calling incipient or beginning programmes are fairly common in science and that they contribute substantially to the progress of science. If so, it is an inadequacy of Lakatos's methodology that he does not make more provision for them. Our case does not give us enough evidence to decide this. However, here is an instance in which Lakatos's methodology does not make adequate provisions for the fine-grained structure of science. 128 4.6.2 Conner and Simberloff Conner and Simberloff (1978) conduct a pairwise comparison of the species that would have turned up on adjacent Galapagos islands by chance versus what we in fact see. They do this for all plant and bird species. They say, "The techniques of... Abbott et al. (1977) implicitly assume that compositional similarity results from deterministic processes, but our results show that stochastic factors may contribute more than previously expected" (Conner and Simberloff 1978, 245). This is the only direct conflict they note with any of the commitments of the scientists we have been examining. A deterministic commitment here referred to might be understood as an implicit hard core commitment of the Abbott group, although it is not one we have been examining, and there is no pressing need to examine it here either. Although the results of Conner and Simberloff and their commitment to the predominance of stochastic determinants of the species compositions of adjacent Galapagos islands are broadly relevant to the work of both Abbott et al. and Bowman, it appears that their results do not conflict directly with the hard core elements that we have been examining of those researchers. What is clear about the work of Conner and Simberloff is that they are committed to stochastic processes being the major determinant of the biogeography of the Galapagos Archipelago, at least with respect to the plant and bird life found there. They do not rule out other factors altogether. The predictions made on the basis of this commitment can be understood as progressive for the same reasons that those of Strong et al. can be understood as progressive. They have made progress by generating 129 predictions on the basis of auxiliary hypotheses appended to their hard core commitments; these hypotheses have generated data showing "that a substantial proportion of the number of species shared between 2 islands can be viewed as resulting from stochastic process of persistence and dispersal" (244). If there is a hard core that they are committed to, it would be that "compositional similarities among the Galapagos Islands" result (predominately) from "stochastic dispersal 48 and persistence" (Conner and Simberloff 1979, 219). This suggestion forms the basis of both of the null hypotheses they test, although "both hypotheses are found to be inadequate models of the compositional similarity of the Galapagos" (ibid.). Nevertheless, their results "suggest that a substantial proportion of compositional similarity can be considered a consequence of stochastic dispersal and persistence" (ibid.). As noted in Chapter II, however, they do not test character displacement, so their hard core, if such it is, might contain the commitment to the negation of competitive exclusion, but it would not explicitly contain the negation of character displacement. On the other hand, a commitment to stochastic composition of adjacent islands implies that intra-island populations would have resulted through stochastic means as well. 4.6.3 Summary of Lakatos applied to the Stochastic Theorists We are left still with the question of whether the work of the stochastic theorists taken together counts as an isolable research programme, an incipient programme, a small sub-programme, or their own larger programme that would be in competition with neo-130 Darwinism itself. As we saw earlier, adaptation counted as a hard core commitment of Darwinism. Therefore a commitment to the biogeography of a particular region as being determined primarily or exclusively by stochastic processes would count as negating Darwin's hard core. It is less clear that it would also count as negating any hard core commitment of neo-Darwinian evolutionary theory. Futuyma counts among the "major tenets" of the evolutionary synthesis that "the change in genotype proportions within a population can occur by either of two principle processes [including] random fluctuations in proportions (random genetic drift)" (Futuyma 1998, 26; bold face text in the original). As mentioned earlier, it is likely that the tenets that Futuyma mentions here would count for Lakatos as hard core commitments, a proposal that is sufficiently plausible that I do not feel the need to argue for it. Genetic drift the way it is described here is not exactly the same suggestion as the stochastic theorists we have been examining. What this passage does indicate, however, is that some of the most centrally-held commitments of neo-Darwinian theory are consistent with stochastic processes. Random factors in gene frequencies (and in the phenotypic expressions of gene frequencies), then, were grafted on to a programme (namely the Darwinian programme) that they were inconsistent with. This is in keeping with Lakatos's commitments (Lakatos 1970, 142). Interestingly, even though stochastic elements have been grafted into an amalgamated evolutionary theory, they are here being separated out for the sake of tracing the relative importance of these factors in particular faunas. In other words, the stochastic theorists might be understood as having a sub-programme within the larger neo-Darwinian programme, similar to Lack's small programme. 131 Nevertheless, it is difficult to conceptualize all stochastic theorists as members of the same research programme. They utilize different approaches, different heuristics and different hard cores. Conner and Simberloff can be seen as having a hard core, but it is less clear that Strong et al. can be understood as having one. 4.7 Schluter and Grant Schluter and Grant begin their (1984) paper with a brief review of several conflicting views regarding the evolutionary radiation of the Galapagos finches. Prominent among the players in the controversy are Lack (1947), Bowman (1961), and the stochastic theorists, as well as some others. They suggest that the arguments between these differing views "have been difficult to settle primarily because data to evaluate alternative explanations are usually unavailable, and because the means to do so are often equivocal" (Schluter and Grant 1984, 176). Their claim is that their "contribution here is the development of a procedure that overcomes many of these difficulties" (ibid.). Their initial commitments are, of course, firmly those of the neo-Darwinian programme. Grant at least had long been an advocate of competition's playing a role in the radiation of the finches (Grant 1981, 660), although his graduate student Dolph Schluter began the project "quite dizzy with the idea that everyone-including Grant-had overestimated the importance of competition" (Weiner 1994,149). Grant seems to have been committed, ever since his initial work with the Abbotts, to a combination of Lack's competition hypothesis and Bowman's adaptation to food resources hypothesis, and 132 Schluter seems to have come around to the two hypotheses by the time he had spent a few years in the Galapagos collecting finch data (Weiner 1994, 151). So we can understand Schluter and Grant as already committed to these hypotheses serving as hard core elements even before they used them as the basis, in combination with auxiliary hypotheses, for the predictions detailed in their (1984) paper. (As we shall see shortly, there is some question regarding the nature of the auxiliary hypotheses regarding the result of the expected densities.) The two main results of the Schluter and Grant 1984 paper are the estimation of adaptive landscapes on the basis of environment, and the five-way test of the various proposed causes for the finch distribution on the Galapagos. To be able to view these results as constituting progress for the relevant research programme(s), as always, they must be demonstrated to contain corroborated novel predictions. The expected densities graph (i.e., the former of these results) primarily corroborates Bowman's floristic hypothesis. The five-way test provides support mainly for Lack's competition hypothesis, although it also indicates that adaptation to floristically-available resources is a contributing factor. First let us determine which research programme(s) Schluter and Grant contribute to in this (1984) paper. I suggested earlier (Section 4.5) that the work of Abbott, Abbott and Grant might coherently be understood as contributing simultaneously to their own new programme and to each of Bowman's and Lack's programmes. On this way of construing the interrelations between the various research programmes, Abbott et al. could make a contribution to their own amalgamated programme without contributing to both of the others. To make progress for Lack's programme, for example, they would have to make a 133 corroborated prediction on the basis of a conjunction of Lack's competition hypothesis with at least one auxiliary hypothesis. This could be accomplished, and would count as progress within their own programme, without also counting as progress within Bowman's programme. Similarly, Schluter and Grant (1984) can be understood as working within Abbott et al.'s research programme, and as simultaneously extending the programmes of Lack and Bowman, where their predictions accord with both of the latter's hard core commitments. Schluter and Grant (1984) do not introduce any new hard core commitments, and so are not properly understood as forging their own new research programme. They do, however, make important contributions to those they can be understood as working within. 4.7.1 Expected Densities To determine whether Schluter and Grant have made Lakatosian progress for any of these programmes, we need again to look for successful. We also need to assess whether or not each prediction candidate counts as ad hoc. In making the calculations required to generate the expected density graphs, Schluter not only used original data collected for this purpose, he also "took information ... coming from a variety of studies, including my own, that relate beak size to ability to handle seeds of different... sizes, and that relates population size to the abundance of the seeds" (Schluter 1999). In other words, the latter determination was of the amount of finch biomass that could be supported by the amount of each species' preferred seeds (fig. 4) available on each of fifteen islands. Furthermore, Galapagos ground finches have an observed tendency 134 to consume seeds of sizes and hardnesses to which their beaks are best adapted. This relationship can be operationalized by graphing (log) seed hardness against (log) beak depth (see fig. 3). As the final result of all these computations, Schluter and Grant graph "expected population density of a solitary finch species as a function of the mean size of its beak" (fig. 5) (Schluter and Grant 1984,182). If all of the relevant factors have been taken into account, the peaks in the graph correspond to the theoretical mean (log) beak depth that is associated with the maximum number of finches supportable by the available food supply on that island. In other words, the kind and number of seeds available on a particular island determines the density of finches of given beak sizes that would be optimal. The hard core commitment being accepted here is clearly Bowman's adaptation to food made available by the plant life on the islands, here seed-bearing plants. That commitment is being taken for granted by this procedure which asks, in effect, what are the fittest beak depths given the available food. The fifteen graphs themselves are not the predictions; they are the result of very clever mathematical rearrangement of the available data in the context of the hard core commitment. One way of stating the relevant prediction is this: if the ground finches are optimally adapted to their food supplies, then the mean beak sizes of the species we observe on those fifteen islands should correspond (more or less closely) to the peaks in expected density as a function of mean beak size. It is less clear what the auxiliary hypotheses are, however. The prediction itself looks like a bare consequence of mathematical reasoning applied to reams of available data. There is not an auxiliary hypothesis here so much as there are direct consequences of Bowman's hard core applied to the observable data and derived by mathematical reasoning. 135 At the same time, none of these reasonings and data seem to provide much protection as elements of a protective belt, either. What might seem to count as an auxiliary hypothesis here, then, is the suggestion, accepted for the sake of the calculations, that optimal adaptation to resources obtains. However, that is little more than a statement of adaptation and, if this is right, then optimality of adaptation is as falsifiable as adaptation itself; if the former is falsifiable, the falsifiability would seem equally to apply to adaptation-Bowman's hard core-itself. The next question concerns whether the prediction was novel. The first possible reason why Lakatos might not see it as such is that, by the very nature of the prediction, it takes known data and relationships as its input. We have seen in the last chapter that Lakatos does not confine himself to a definition of novelty such that predictions are "novel" only when they predate the data that they "predict." However, as Zahar points out with respect to his emendation of Lakatos's novel prediction, an amendment that Lakatos accepts, "facts provide little or no evidential support for the theory, [when] the theory was specifically designed to deal with the facts" (Zahar 1976, 218). What we do not see in the present prediction is theory being used to accommodate the facts, here the mean beak depths of ground finch species actually found. What is not factored into the calculations resulting in the graphs is the mean beaks of the actual finches to be found on the islands. The available seed food is taken into account, and the observed seed preferences of finches with various beak sizes, but not which mean beak depths actually occur on the islands. That data is overlayed on the graphs afterward for comparison of prediction to the mean beak depths of the finches that are found. 1 3 6 On the criterion just quoted, it appears that the prediction does count as "novel," and we can take the just-mentioned considerations as motivating Lakatos and Zahar in having a novelty requirement for predictions. Therefore it seems that the prediction should count as sufficiently novel for Lakatos such that it counts as theoretical progress. Similarly, should this prediction obtain, this should as empirical progress. Indeed, the prediction is often borne out, as can be seen easily by inspecting Figure 5. As in the Strong et al. case, it is a bit unclear how to delineate the prediction; it might count as one prediction or many, or as a multi-modal prediction. However, what is most important for our purposes is that the prediction or predictions are often corroborated. So, to reiterate, the first main result of Schluter and Grant (1984) constitutes progress for Bowman's small research programme. 4.7.2 Five-Way Test We saw earlier that the five-way test might best be understood as a five-way crucial experiment. Basically, the procedure is to take the main hypotheses about the processes determining the radiation of the ground finches in the Galapagos and use the data to predict, on the basis of each hypothesis, the mean beak depth found on the twelve islands for which there was suitable data. Again, which actual finch species (with their corresponding mean beak depths) are found there is a data set that is not used in the computations. The raw data used were as before, leading to the expected densities (fig. 5), but the results for only twelve islands were utilized. Model I assumed only stochastic colonization where there were seeds for these finches to feed upon (i.e., where the expected densities were positive). This first 137 model thus evaluates what might be understood as some stochastic theorist's hard core commitment. Model II assumed random colonization again, but in conjunction with food abundance and, as such, tests a minimal version of Bowman's hypothesis; that is, persistence of a particular finch species at any given location is dependent upon the kind and abundance of seed food found there.49 Model III is just model II with the addition of the ability of the species to evolve in response to local food sources. This is more in keeping with Bowman's commitments given that he demonstrates at length the correlation between diets and finch skull morphology. It is not to be assumed that the high degree of the finches' beaks' fittedness is due purely to chance, particularly as mainland forms of the finches have different morphology. As a result, we can assume that Bowman believes that the finches have evolved in situ. The predictions made on the bases of all of these models so far do not accord well with the observed mean beak sizes on the studied islands. It is not until model IV that the predictions correlate closely to the observed data. Model IV is the first model to take competition into account, although it tests only competitive exclusion of the two kinds of competition. Model V takes character displacement into account as well, and the prediction on this basis accords slightly better with what is observed. We might understand these five models as auxiliary hypotheses conjoined to the hard core commitments of the various players in the Galapagos finch competition controversy. Then, on the basis of new auxiliary hypotheses, the predictions of the research programme centred on Lack's hard core of competition are confirmed empirically. Thus Lack's programme has been shown to surge progressively ahead of its rivals. To the extent that the Abbott et al. programme embraces Lack's hard core as one of its own hard core 138 commitments, it has progressed as well. Lakatos is not, however, committed to this state of affairs being the last word on the relative progressiveness of the small research programmes given here. 139 Chapter V Laudan on Scientific Progress In this chapter I will summarize Larry Laudan's problem-solving account of progress as it is presented in his book Progress and its Problems (1977). Specifically, I will describe Laudan's two main kinds of problem, empirical and conceptual. These form the bases of the kinds of progress that Laudan admits, given that Laudan makes progress parasitic upon problem solution. The first section details Laudan's somewhat vague system of weighting the two kinds of progress that arise from these two kinds of problem. Next, I detail the relationships between problems, the theories that solve them, and the way that this, in conjunction with problem weighting, constitutes progress. Theories are one of the main elements that constitute Laudan's research traditions, the others being a research tradition's methodology and ontology. Finally, I detail the way that Laudan evaluates a research tradition for progress. This section is also relevant for progress in terms of individual theories. One of the more significant results of this chapter is that Laudan does not appear to require that progress be assessed only against the background of a research tradition. 5.1 Introduction The sort of progress Laudan exclusively concerns himself with in his (1977) book is what he calls "cognitive progress" or "progress with respect to the intellectual aspirations of 140 science" (Laudan 1977, 7). Because Laudan believes that problem solving is the central goal of science, and hence that it is constitutive of the most general characterization of science, he makes problem solution the unifying element in his philosophy of science. Furthermore, for Laudan, problems are intimately involved with progress. Therefore, an account of his view of progress must be preceded by a close examination of what he means by problem solving, including the kinds of problem solving he admits. It is first useful to give a general characterization of a problem. Laudan calls the abstract example which follows "a very crude model of scientific evolution" (Laudan 1977, 67); however, it seems consistent with the rest of his discussion in demonstrating the kind of thing a problem needing solution can be. He says, "Imagine some domain in which we notice a certain puzzling phenomenon, p. The phenomenon p constitutes an unsolved problem for the scientist who wishes to develop a theory T\, specifically with a view toward resolving p" (Laudan 1977, 67). For Laudan the initial (empirical) problem is presented by some characteristic ("puzzling phenomenon") of nature in the domain of interest. The puzzling property of some item in the domain of interest is basic and sets the stage for a problem; it does not constitute a problem proper itself. On Laudan's view, progress occurs by solving problems. In turn, theories are the entities that provide solutions to problems. The "complex of functions" that Laudan refers to when he calls theories solutions to problems is "to resolve ambiguity, to reduce irregularity to uniformity, to show that what happens is somehow intelligible and predictable" (Laudan 1977, 13). From this we see the multifarity of his notion of problem solution. A theory which "provides acceptable answers to interesting questions" is roughly equivalent to its 141 "providing] satisfactory solutions to important problems" (13). The stress is upon the solving of important problems and similarly, the answering of interesting questions, rather than upon numbers of solutions, an emphasis to which I will return. Further, the adequacy of a solution and the significance of a problem are more important than whether the solutions are "'true,' 'corroborated,' well-confirmed'" (14). Laudan also makes reference to experimental methodologies, that I will later detail, but they are not the sort of entities that solve problems. 5.2 Empirical Problems There are two main kinds of problem for Laudan: empirical and conceptual. Although he separates them for the sake of simplicity, he admits that a continuum exists between these two endpoints; it is for simplicity that he explicates the extreme ends of the spectrum (Laudan 1977,48). Empirical problems are those that we often unreflectively associate with science. Laudan calls them "first order problems" and they concern the subject matter of scientific theories-"anything about the natural world which strikes us as odd, or otherwise in need of explanation" (15). He calls them empirical problems, in spite of the possibility of the theory-ladenness of observation. He does so because "we treat empirical problems as if they were problems about the world" (ibid.). So in order to determine "the adequacy of solutions to empirical problems" we study "the objects in the domain" to which the problems refer (15). 142 Laudan is careful to stress the empirical nature of these problems. He does not mean to equate extra-theoretical facts (which he calls "states of affairs") with "empirical problems"; for example, he says that empirical problems are not to be understood as what is "directly given by the world as veridical bits of unambiguous data" (15). In other words, we are to think of problems as a construct of language, rather than as the entities themselves (what he calls "facts") in the domain we are examining. Part of the significance of this distinction for Laudan is that a problem can be solved without the solution's being true. Alternatively, there can be problems that do not in fact correspond to what is in the world, but which scientists only believe to do so: "all that is required is that it be thought to be an actual state of affairs by some agent" (16). Falling under his category of empirical problems are problems not yet solved by any theory, solved problems, and anomalous problems, namely "those empirical problems which a particular theory has not solved, but which one or more of its competitors have" (Laudan 1977, 17). This is a novel definition of the term "anomaly," in that anomalies are not defined, as Kuhn has them, only as those observations that are logically contradictory with the theory in question; in fact, problems that logically contradict the theory are not considered anomalies on this account unless a rival theory has solved them. Anomalous problems count as "evidence against a theory, and unsolved problems simply indicate lines for future theoretical inquiry" (Laudan 1977, 18). The significance of anomalies for Laudan is that they constitute "an empirical situation which, while perhaps not offering definitive grounds for abandoning a theory, does raise rational doubts about the credentials of a theory" (28). 143 Of course the most important of these three kinds of empirical problem for the sake of progress is the solved problem: "one of the hallmarks of scientific progress is the transformation of anomalous and unsolved empirical problems into solved ones" (16). For Laudan, "a theory may solve a problem so long as it entails even an approximate statement of the problem" (23)-or an approximate solution to the problem, presumably, so long as this does not entail a comparison to extra-theoretic reality. This commitment is not terribly contentious. Laudan merely wants to recognize that many results in science are approximations rather than exact matches to the predictions made by theories. However, the commitment that it leads him to is more contentious: it entails that two different theories from different times can both solve a problem even though their solutions "are formally inconsistent" (24). So calling something a solution to a problem should not be understood as pronouncing it correct. This is spelled out more explicitly where Laudan claims that "one need not, and scientists generally do not, consider matters of truth and falsity when determining whether a theory does or does not solve a problem" (24). For example, "Lavoisier's theory of oxidation, whatever its truth status, solved the problem of why iron is heavier after being heated than before" (24-5). Another corollary of this understanding of problem solution is that solutions to problems need not be permanent to count as solutions (25). 144 5.3 Conceptual Problems "[Conceptual problems are higher order questions about the well-foundedness of the conceptual structures (e.g., theories) which have been devised to answer the first order questions" (Laudan 1977, 48). They are "characteristics of theories and have no existence independent of the theories which exhibit them" (ibid.). Conceptual problems are of two types: conceptual difficulties internal to a particular theory; and external conceptual problems, in which some element of a theory is in conflict with another, well-established theory (49). From examples, it appears that these sort of inter-theory tensions are usually to be understood as occurring between theories in different domains (50-54). There are two kinds of internal conceptual problem: logical inconsistency within a theory (research tradition), and "conceptual ambiguity and circularity within the theory" (49). Both are "important in the process of theory appraisal" (50). Laudan allows that "the ambiguity of concepts is a matter of degree rather than kind. Some degree of ambiguity is probably ineliminable in any except the most vigorously axiomatized theories" (ibid.). Because of this ineliminability of ambiguity, as well as its being sometimes advantageous, Laudan is not concerned to eliminate all conceptual ambiguity associated with theories. However, "systematic and chronic ambiguity or circularity ... [is] highly disadvantageous" (49). One important way of addressing internal inconsistencies, is "[t]he increase of the conceptual clarity of a theory through careful clarifications and specifications of meaning" (Laudan 1977, 50). Furthermore, it is "one of the most important ways in which science progresses" (ibid.). 145 External conceptual problems have been more important historically, according to Laudan, than internal problems (50). They can be generated even by conflicts with extra-scientific problems. Among these are conflicts with accepted religious traditions (46, 63) and with notions about the freedom of human action (102). Among external conceptual problems, there are three types: logical inconsistency, joint implausibility and mere compatibility (Laudan 1977, 51-3). Logical inconsistency between two theories needs no further explication. Joint implausibility occurs "when the acceptance of either [theory] makes it less plausible that the other is acceptable," where they are not logically incompatible (52). Mere compatibility is not always a problem, but it counts as one where scientists expect that the second theory should reinforce the first, but instead it turns out only to be compatible with the first (53)-that is, the first theory implies nothing about the second (54). 5.4 Weighting Problems Because the weighting of problems will be integral to Laudan's assessment of progress, I will here touch on some considerations he gives regarding relative weights of problems. Laudan's weighting criteria are to be taken as guidelines rather than as exhausting the ways that problems might be weighted rationally (Laudan 1977, 32). Furthermore, he is interested only in "cognitively rational weighting" rather than analysis based on other dimensions along which problems might matter to scientists and society (ibid.). 146 Let us turn first to empirical problems. In "a domain in which no adequate, systematic theories have yet been developed, almost all empirical problems are on a par" (33) . However, once there are a number of theories in a given domain, they can be differentially weighted. Laudan lists eight weighting considerations (33-8). Here I will detail only those that will be relevant to the Galapagos finch competition controversy. Once a problem has been solved by some theory in a domain, there is a problem inflation by solution effect: other theories in the domain must either solve the problem as well or suggest a principled reason why they should not have to solve it. One might think of this kind of problem weighting as the most minimal one. It is the kind of "weighting" that gets a problem to count as a problem at all. By way of defining problem inflation by anomaly solution Laudan says, "If a problem has proved anomalous for, or resisted solution by, certain theories in the domain, then any theory which can transform that anomalous problem into a solved one will have strong arguments in its favor" (33-4). One of Laudan's examples of this is the Michaelson-Morley results, which counted as anomalous empirical problems for "earlier aether theories" (34) . So Laudan allows that a problem discovered at a time can become a problem for earlier theories in a domain. Another way in which a theory "may endow certain empirical problems with greater significance than others" is problem inflation by archetype construction (34). For Laudan, an archetypal empirical situation is one to which "other processes in the domain must be reduced;" (34) thus, problems associated with it are more acute. For example, if one solves a problem by demonstrating it to be an archetypal case of some underlying phenomenon, 147 this is a form of problem solution that is then to be weighted particularly highly (35). What Laudan seems to be getting at here, without using the term, is paradigmatic instantiations of processes that figure importantly among a discipline's commitments. This interpretation is lent weight by a subsequent passage: "certain problems can be given prominence by the emergence of a new theory that gives them special importance" (36); however, here the emphasis seems to be on the new theory rather than the problem that is solved by it. With his problem weighting by generality (35) Laudan suggests that a problem that is more general is more important than a problem that has as its subject matter a smaller domain. Presumably, then, a problem can accrue greater weight if scientists show its solution to be an instance of a more general phenomenon. As for conceptual problems, it is important to note that Laudan considers them generally to be more serious than empirical anomalies (64). First, "[o]ther things being equal, the greater the tension between two theories, the weightier the problem will be" (65). Second, where there is a tension between two theories, the seriousness of that tension for the first theory is proportional to the degree of problem-solving ability of the other theory. Where its problem-solving effectiveness has been most certainly demonstrated, its incompatibility with the target theory will be most heavily weighted (ibid.). Third, where there are two competing theories "which exhibit the same conceptual problem(s), then those problems count no more against one than against the other and become relatively insignificant in the context of comparative theory appraisal" (65). Finally, a tension or inconsistency with an older problem is more significant than a newly-discovered problem 148 because there is hope when a problem is first located that "with very minor modifications in the theory, we can bring it into line" (ibid.). It is not until a later chapter that Laudan begins to talk about research traditions; he admits to using the term "theory" earlier for research traditions (Laudan 1977, 69). That makes it important to ascertain whether in his section on conceptual problems he is referring to what he later means by "theory" or what he later means by "research tradition." As we saw at the beginning of the section on conceptual problems, the examples of theories that Laudan gives are theories from different domains. Moreover, the tensions between theories that Laudan alludes to in his section on weighting conceptual problems put theories from different research traditions into a state of tension. We can assume that when Laudan refers to tensions between theories, he generally means theories from one research tradition that are in conflict with theories from a different tradition. 5.5 Problems, Theories and Progress Next we come to the relationship Laudan perceives between problems and progress. Although we will see momentarily that, for Laudan, assessing progress is an irreducibly comparative matter, it is useful to begin with progress's smallest sub-unit. This is the single empirical problem solved by a theory: 'Ti has solved its initial empirical problem, p, and to that extent, we can say that 'progress' has been made" (Laudan 1977, 67). This does not count yet as true progress. To introduce a more robust notion of progress, one that takes the 149 weights and significance of problems into account, requires alluding to "problem-solving effectiveness." Laudan states that, the overall problem-solving effectiveness of a theory is determined by assessing the number and importance of the empirical problems which the theory solves and deducting therefrom the number and importance of the anomalies and conceptual problems which the theory generates (Laudan 1977, 68) From there it is possible to move to "a rudimentary notion of scientific progress": "progress can occur if and only if the succession of scientific theories in any domain shows an increasing degree of problem solving effectiveness" (Laudan 1977, 68). Already the temporal component of progress is in evidence. Furthermore, there are "many ways in which such progress can occur" (ibid.). Laudan gives three of these here although it is not clear whether he intends these to be exhaustive of the kinds of progress there are. First, there is the replacement of one theory by another which is "an expansion of the domain of solved empirical problems" alone, with all of the other aspects of the two theories remaining the same. Another kind of progress occurs when a modification of a theory "eliminates some troublesome anomalies or which resolves some conceptual problems" (ibid.). It is worth underscoring here that these three factors, viz., increase in empirical problems and minimizing of anomalies and conceptual problems, form the cornerstones of Laudan's 150 account of progress. The third way that progress can be made, the way it most often is, is "as a result of all the relevant variables shifting subtly" (68). We learn later in his (1977) book that Laudan considers the section (66-9) regarding the problem-solving model of progress as discussing "how to evaluate the problem-solving effectiveness of individual theories" (107). For the purposes of assessing our resource competition cases, it is useful to have guidelines appropriate to the smaller grain of science we see there. Laudan says, Localizing the notion of progress to specific situations rather than to large stretches of time, we can say that any time we modify a theory or replace it by another theory, that change is progressive if and only if the later version is a more effective problem solver (in the sense just defined) than its predecessor. (Laudan 1977, 68) In turn, "the sense just defined" was that "progress can occur if and only if the succession of scientific theories in any domain shows an increasing degree of problem solving effectiveness" (ibid.). Next we discover that progress for Laudan is defined at all times as comparative superiority in problem solving (Laudan 1977, 120). He says, All evaluations of research traditions and theories must be made within a comparative context. What matters is not, in some absolute sense, how 151 effective or progressive a tradition or theory is, but, rather, how its effectiveness or progressiveness compares with its competitors. (Laudan 1977, 120) So progress is to be assessed comparatively: the progressiveness of theories or research traditions is to be assessed by comparing, on balance, the weighted difference of important solved problems minus important anomalies created by each. What is not being compared is any theory or research tradition to extra-theoretic reality (Laudan 1977, 125-6). What motivates this rejection of verisimilitude and other attempts to define progress as movement toward some extra-theoretic reality is Laudan's position that all such attempts have so far foundered: "no one has been able even to say what it would mean to be 'closer to the truth,' let alone to offer criteria for determining how we could assess such proximity" (126).50 He nevertheless rightly claims that his account is not anti-realist insofar as realism is consistent with his account of progress as problem solving: for all we know, his problem-solving theories may be true or approaching truth (ibid.), although he does not assess this. By these means he has neatly sidestepped the problem of how to compare theory to extra-theoretic reality in that his account of progress does not depend upon realism's tenability; he is able to define progress and rationality (in the case of empirical problems) in terms of progressively more adequate theories explaining the phenomena. His defining the phenomena (empirical problems) as non-equivalent to "states of affairs" is part of his care in attempting to have an account of progress that does not compare the substance of theories to the extra-theoretic reality they are aimed at. 152 I take from Hacking (1983) the useful (although somewhat implicit) distinction between subsumption and accumulation. A theory subsumes its predecessor (in Laudan's terms) when it solves all of the problems of its predecessor and more (Laudan 1977, 147; Hacking 1983, 67-8). Laudan calls this state of affairs the "cumulative conception of progress" (1977, 147), and he rejects its necessity as a condition of progress. Nevertheless, Laudan accepts that common threads filter through science across time: "it is basically the shared empirical problems which establish the important connections between successive research traditions" (140). The result is that "[fjhere is much continuity in an evolving research tradition. From one stage to the next, there is a preservation of most of the crucial assumptions of the research traditions" (98). So this is the way in which a research tradition is "cumulative," but not by subsuming everything that went before. Research traditions will be detailed more thoroughly below. We can also assess Laudan's commitment to accumulation, Hacking's "heapings up of knowledge" (Hacking 1983, 55). Significantly, Laudan defines progress in terms of the scope, rather than the numbers, of solved problems. He says that the following constitute the "core assumptions" of a problem-solving model of progress: (I) the solved problem-empirical or conceptual-w the basic unit of scientific progress; and (2) the aim of science is to maximize the scope of solved empirical problems, while minimizing the scope of anomalous and conceptual problems. (Laudan 1977, 66) 153 Although he does not define what he means by "scope" and this appears to be its only mention, the next paragraph following this quote might provide a clue. Laudan says, "The more numerous and weightier the problems are which a theory can adequately solve, the better it is" (ibid.). It seems, then, that what he intends by scope is just the weighted number of (adequately) solved problems. In the next chapter, I will have more to say about Laudan's notion of adequacy. I think Laudan is being careful in this quote to avoid being committed to the idea that an increase merely in the number of solved problems constitutes progress. Indeed, he nowhere asserts that an increase in numbers of solved problems by itself counts as progress. Indeed, numbers of solved and unsolved problems alone do not, according to Laudan, allow us adequately to evaluate the relative progress of two theories. We must also take into consideration the significance both of the various problems and of different proposed solutions. In other words, there is a weighting both of solutions and of problems, both solved and unsolved, which must be taken into account in the determination of which is the most progressive theory. 5.6 Research Traditions A research tradition is not to be reduced to the sum of its individual theories. Its theories only partially constitute and exemplify it (Laudan 1977, 78). Rather, it is "associated with a series of specific theories, each of which is designed to particularize the ontology of the research tradition and to illustrate, or satisfy, its methodology" (81). An important 154 difference between theories and research traditions is that the former are usually "empirically testable for they will entail (in conjunction with other specific theories) some precise predictions," whereas the latter "are neither explanatory, nor predictive, nor directly testable" (81-2). Apart from the theories that partly constitute it, the research tradition can be recognized by some of its other characteristics. It is the "metaphysical and methodological commitments which, as an ensemble, individuate the research tradition and distinguish it from others" (79). The research tradition's ontological and methodological commitments have a few different roles. The methodological components of the research tradition specify the "legitimate methods of inquiry open to a researcher within that tradition," including "experimental techniques, modes of theoretical testing and evaluation" (ibid.). Furthermore, the research tradition changes its makeup over time, unlike specific, individual theories (ibid.). It also "provides a set of guidelines for the development of specific theories. Part of those guidelines constitute an ontology which specifies, in a general way, the types of fundamental entities which exist in the domain" (79). The individual theories of a research tradition then reduce empirical problems in the relevant domain to the entities to which the research tradition is ontologically committed (ibid.). "Put simplistically, a research tradition is thus a set of ontological and methodological 'do's' and 'don'ts'" (80), but as we have seen, theories are part of what constitutes research traditions as well. There are passages, such as this one, in which Laudan seems to abandon his commitment to theories being part of what constitutes the traditions in which they are embedded but, even so, I will assume that he is committed to this. 1 5 5 There are four main "modes of interaction" between research traditions and the theories that partially constitute them: (1) a problem determining role (86), (2) a constraining role, (3) a heuristic role (89) and (4) a justificatory role. Each of these deserves a brief explication, except for the second among these, that will be least important for our purposes. First, "research traditions do not offer solutions to specific problems. Even so, "they are nevertheless not "outside of the problem-solving process. To the contrary, the whole function of a research tradition is to provide us with the crucial tools we need for solving problems, both empirical and conceptual" (82). The most important factor here is that the research tradition in part determines the problems within the domain to which it applies. Laudan says that "[e]ven before specific theories are formulated within a tradition, and continuously thereafter, a research tradition will often strongly influence (although it does not fully determine) the range and the weighting of the empirical problems with which its component theories must grapple" (86). It also influences what can count as "conceptual problems that the theories in that tradition can generate" (ibid.). It carries out these influences in a few ways. It determines the domain to which the theories under its umbrella will apply (86), and excludes some situations from counting as properly part of the domain (87). Its methodology usually also specifies "certain experimental techniques which alone are the legitimate investigational modes for determining what are the data to be explained" (87). Conceptual problems that arise due to a conflict between theories and the research tradition of which they are a part are the commonest kind of conceptual problems (88). 156 Due to the research tradition's role in specifying "certain types of entities and certain methods for investigating the properties of those entities," they can provide a heuristic function in generating new theories (89-90). Laudan again stresses here that theories cannot be deduced from the research traditions in which they are embedded. However, the ontology of the research tradition suggests certain kinds of theoretical solutions that might be used to explore the problems in its domain. It does this in part by providing scientists with theoretical entities that they can then apply to new problems. For instance, Descartes's general research tradition asserted "that the only properties which bodies can have are those of size, shape, position, and motion" (Laudan 1977, 91). When he developed his subsequent theory of light and colours, he was constrained by these ontological commitments. The most significant justificatory role of the research tradition consists in allowing scientists not to have to start afresh with each new theory. They can take some things for granted and do not have to justify to their fellow researchers in that domain, at least, their initial assumptions (93). They are then freed up to "pursue specific problems of interest" (ibid.). I turn now to what Laudan calls the evolution of research traditions. Laudan's "important and substantive changes ... occur[ing] within an on-going research tradition" will be more relevant to the Galapagos finch competition case than "how research traditions are displaced by other ones" (96). It is worth underscoring that these changes do not in themselves constitute progress, but changes to the research tradition's component theories are prerequisites for progress since increased problem-solving effectiveness will, of course, require changes to existing theories, and the addition of new theories. "The most obvious 157 way a research tradition changes is by a modification of some of its subordinate, specific theories" (96). These theories are not core commitments, and it is a matter of tinkering, not of making a complete overhaul of the existing theories. These modifications can be effected by "[sjlight alterations in previous theories, modifications of boundary conditions, revisions of constants of proportionality, minor refinements of terminology, expansion of the classificatory network or a theory to encompass newly discovered processes or entities" (96). Another way of modifying the set of the subordinate theories is by adding new theories, such as when researchers "discover that there is, within the framework of the tradition, a more effective theory for dealing with some of the phenomena in the domain than they had realized previously" (96). Importantly, the overall research tradition is only being tweaked, not overhauled. The research tradition can also be modified by means of "a change of some of its most basic core elements" (96). This is unlike Kuhnian and Lakatosian treatments of research tradition analogues (i.e., paradigms and research programmes, respectively) in which a whole new tradition is generated when there are any changes made to its most fundamental commitments (Laudan 1977, 97). Interestingly, Laudan is still (like Lakatos) committed to the idea that "certain elements of a research tradition are sacrosanct and thus cannot be rejected without repudiation of the tradition itself (99). The way he then squares this suggestion with his commitment to changeability of the core commitments is "to insist that the set of elements falling in this (unrejectable) class changes through time" (ibid.). As evidence of the Tightness of making changeable core elements one of the characteristics of his research tradition, Laudan notes "that there is scarcely any interesting set of doctrines 158 which characterizes any one of these research traditions throughout the whole of its history" (97). Significantly, he includes Darwinism as among "the great research traditions" for which this is the case (ibid.). In expanding Laudan's account of progress in terms of individual theories to research traditions, it is necessary to make explicit what are for Laudan the relationships between progress and what he labels "problem-solving effectiveness." For example, "the problem-solving effectiveness of a theory depends on the balance it strikes between its solved problems and its unresolved problems" (Laudan 1977, 67). The choosing of a theory that has improved problem-solving effectiveness over a predecessor explaining some aspect of the same domain counts as progress (68), and, incidentally, as rationality (109). Laudan says that with respect to "the progressiveness of a research tradition... . our chief concern is to determine whether the research tradition has, in the course of time, increased or decreased the problem-solving effectiveness of its components" (107). Incidentally, it is also worth drawing attention here to Laudan's commitment to the temporality of progressiveness. In summary, "[a] research tradition is a set of assumptions: assumptions about the basic kinds of entities in the world, assumptions about how those entities interact, assumptions about the proper methods to use for constructing and testing theories about those entities" (Laudan 1977, 97). 159 5.7 Evaluating Research Traditions for their Progress Laudan clearly prefers to assess the progressiveness of whole research traditions as we have seen, and as we shall see in more detail later in this section. However, there are at least two hints that he is not (or should not be) committed to progress's evaluation being possible only against the background of a research tradition. The first of these is that in his evaluation of progress, he mentions only problems, and not the hallmarks of his research traditions-methodology and ontology. We have already seen a number of passages in which Laudan insists that progress is to be determined by assessing whether an increase in problem-solving effectiveness with respect to empirical problems has been made and, at the same time, whether conceptual problems and anomalies have been overcome by the relevant theory or theories. These are the only entities he mentions in any of his statements of what progress consists in. Second, Laudan argues against the distinction made by Kuhn and Lakatos between "the 'early' and 'advanced' stages of scientific activity" (150). He titles the section in which he argues this as "In Defense of'Immature' Science" (ibid.), which suggests that he is in favour of even immature science progressing; however, he does not explicitly defend immature science nor give an account of it in his terminology and machinery. Still, he argues against "the Kuhn-Lakatos thesis that they [i.e., 'mature' sciences] would be intrinsically more progressive and more scientific than 'immature' ones" (151), so presumably we can assume that Laudan is committed to the assessability of progress in units smaller than whole Kuhnian paradigms, Lakatosian research programmes, or his own 160 analogue-research traditions. To make it clear that these entities play a sufficiently analogous role in the work of these three philosophers of science, it is important to compare their various commitments. Laudan's "working definition of a research tradition" is that it is "a set of general assumptions about the entities and processes in a domain of study, and about the appropriate methods to be usedfor investigating the problems and constructing the theories in that domain" (81). For comparison, Kuhn's paradigm in the sense of "disciplinary matrix" is to be understood as "the constellation of group commitments" including, but not restricted to, the "formal or the readily formalizable components" used by the relevant community, shared metaphysical commitments, values concerning the proper ways of doing science and what is to count as an adequate experimental result and, finally, paradigms in the sense of exemplars (Kuhn 1970, 181-5). Lakatos's programme consists of a hard core of irrefutable theoretical commitments, a negative heuristic that defines and defends the hard core, a positive heuristic that "defines problems, outlines the construction of a belt of auxiliary hypotheses, foresees anomalies and turns them victoriously into examples, all according to a preconceived plan" (Lakatos 1971, 99). It also has a collection of (auxiliary) hypotheses or theories (just mentioned), regarding the domain of interest, that comprise the bulk of the programme (cf. Chapter II). It should now be clear that these three are broadly analogous entities. Although in the next chapter I will not be assessing the relative merits of large research traditions, an understanding of how Laudan evaluates whole traditions is relevant to the evaluation of their sub-components, that will concern us. There are two chief ways of 161 appraising the relative merits of research traditions. The first is synchronic and requires us to assess the "(momentary) adequacy of a research tradition... . how effective the latest theories within the research tradition are at solving problems" (106). In turn, this requires a determination of the problem-solving effectiveness of these current theories (ibid.). This is detailed in Laudan (66-9), and above at most length in the section on problems, theories and progress. With the sum of the effectiveness of constituent problems in hand, "we need only combine those appraisals to find the adequacy of the broader research tradition" (107). The significance of this suggestion is that the adequacy of a research tradition is equivalent to the sum of the problem-solving effectiveness of its constituent current theories. No assessment of a research tradition's ontology or methodology is therefore accepted into this measure, i.e., adequacy. This is unsurprising, given Laudan's commitment to the supremacy of problem solution in science, but deserves mention. Assessing progress, which is "diachronic and developmental" (Laudan 1977, 106) in the context of research traditions also requires a return to the notion of problem-solving effectiveness. There are two particularly important kinds of problem-solving effectiveness to be concerned with in this connection, viz., general progress and the rate of progress (107). General progress is to be assessed by comparing the adequacy of a research tradition's oldest theories with that of its most recent theories. In contrast, the rate of progress is determined by "changes in the momentary adequacy of the research tradition during any specified time span" (ibid.). Later Laudan claims to "outline machinery for objective, rational comparisons between competing scientific theories and research traditions" (142); but most of this 162 discussion centres on a defense of the comparison of research traditions in spite of the potential objection that observation cannot be theory-neutral. Laudan does, however, defend assessments of progress made internally to a research tradition as "an approximate determination of the effectiveness of a research tradition" (145). This is to be done by assessing whether the research tradition has solved the problems which it set for itself; we ask whether, in the process, it generated any empirical anomalies or conceptual problems. We ask whether, in the course of time, it has managed to expand its domain of explained problems and to minimize the number and importance of its remaining conceptual problems and anomalies. (Laudan 1977, 146) Once this is done "we should be able to construct something like a progressive ranking of all research traditions at a given time" (ibid.), given only these internal assessments. Both earlier measures of progress in terms of generality and rate are thus internally assessable. Because Laudan treats this internal assessment as only an approximation of relative progress, it is not strictly a contradiction of his earlier assertion that all progress is irreducibly comparative. 163 Chapter VI Evaluating Laudan's Account I begin this chapter by setting out the evolutionary context for the Galapagos finch competition controversy using Laudan's vocabulary. I then discuss the proper scale in which to understand the finch case. There is next a section that details some considerations preliminary to applying Laudan's problem-solving account of progress to the finch case. A peculiar aspect of Laudan's account is here detailed: any problem solution, no matter how strange, will count as a solution for Laudan. This state of affairs gives us few ways to adjudicate the relative merits of such solutions. Next I detail how Laudan's account is to be cashed out in Lack's research. Bowman and Lack can be understood as solving the same problems, so in the section on Bowman I detail a strange result of applying Laudan to them: neither counts as having more problem-solving effectiveness. With Abbott et al. there are the case's first anomalies (in Laudan's sense) generated for Bowman and Lack. This allows us for the first time to compare the theories of these two theorists within Laudan's account. However, because the commitments of Lack and Bowman count as part of a disjunctive account, I question whether such a comparison accords with Laudan's purposes in comparative assessment of theories. In the section on the stochastic theorists, there is a reprise of the question of whether all of the scientific episodes in the Galapagos finch competition controversy count as small theories within a larger research tradition, or whether the stochastic challenge can be understood as a budding research tradition in itself. 164 Laudan is shown not to be able to accommodate the progressive results of Schluter and Grant. The final section details Laudan's commitments in relation to the accumulation of additional factual knowledge. 6.1 Evolutionary Research Traditions In situating the episodes detailed in the Galapagos finch competition controversy, it is necessary first to understand Darwinism and neo-Darwinism within Laudan's approach. For Laudan, Darwinism counts as a "classic research tradition" (Laudan 1977, 78) and as one of "the great research traditions in the history of scientific thought" (97). He does not mention the neo-Darwinian evolutionary synthesis. This leaves us to determine, from his other commitments, how he would most likely characterize it. As to whether the neo-Darwinian synthesis counts as its own autonomous research tradition or as an extension of, and so part of, Darwinism, there are two main considerations to be taken from Laudan-unfortunately pulling in opposite directions. First, there is Laudan's suggestion that the core commitments of a single research tradition are amenable to change over time (99). Second, there is the suggestion that separate research traditions can be amalgamated (105). Laudan believes that Darwinism's "core assumptions ... were explicit even from [its] inception" (75), so we see that he is committed to the Darwinian tradition's having core commitments. Further, with respect to the core elements of research traditions in general, he wants "to insist that the set of elements falling in this (unrejectable) class changes over time" without that entailing the introduction of a new tradition (99). Here he also poses a 165 question that is germane to our present concern: "if a research tradition can undergo certain deep-level transformations and still remain in some sense the 'same' tradition, how do we distinguish change within a research tradition from the replacement of one research tradition by another?" (ibid.). Given that a single research tradition can morph over time with respect to its most central commitments, one might wonder whether neo-Darwinian evolutionary theory counts as a mutated version of the Darwinian tradition. However, viewing neo-Darwinism as an integrated tradition seems the likeliest way to characterize the evolutionary synthesis, particularly as even its practitioners understand it as an integration primarily of selection and mutation, the central commitments of Darwinism and population genetics (Futuyma 1998, 24).51 In his section regarding the amalgamation of two whole research traditions (103-5), Laudan suggests two ways this can happen. The first way individual research traditions might be integrated is by one being "grafted onto another, without any serious modifications of the presuppositions of either" (103-4). The other way involves "the repudiation of some of the fundamental elements of each of the traditions being combined. In the latter kinds of cases, the new research tradition, if successful, requires the abandonment of its predecessors" (104). Because Laudan also describes this as "old ingredients ... combined" (ibid.), we should understand this process as some of the elements of the earlier tradition becoming part of the new and thus replacing both original traditions. It would appear that the neo-Darwinian synthetic theory could count as an example of either of these ways of amalgamating traditions. The definitive resolution of the question of whether to understand neo-Darwinism as a mutated Darwinian tradition, or an amalgam of Darwinism and other traditions is not 166 terribly important. A gesture toward a solution is valuable here insofar as it provides a backdrop for the smaller-grained episodes of the Galapagos competition controversy that will play themselves out under the auspices of the neo-Darwinian research tradition. So let us take the still-reigning evolutionary tradition to be neo-Darwinism; it was also accepted in the time of Lack's (1947) book, and by Lack himself, as we have seen in Chapter II. 6.2 The Resource Competition Case: Tradition or Theory? Given that different considerations apply in assessing the comparative progress of whole research traditions versus theories, it serves us well to determine which aspects of the finch competition dispute can be understood in which way. I want to suggest that Laudan's commitments make it the case that most of the resource competition versus food adaptation aspects of our case are best understood as theories, problems and problem solutions, all subsumed within the neo-Darwinian research tradition. They are not to be understood as (small) individual research traditions in themselves. To assess which is the most appropriate way to understand the episodes in the Galapagos finch competition controversy, it is necessary to begin by reviewing the differences between the two most obvious levels: the research tradition and the theories which partly constitute it. In a succinct characterization of the differences between the two, Laudan says, 167 A research tradition ... specifies a general ontology for nature, and a general method for solving natural problems within a given natural domain. A theory, on the other hand, articulates a very specific ontology and a number of specific and testable laws about nature (Laudan 1977, 84). The former characteristics certainly seem broadly to apply to neo-Darwinian evolution and the latter to the details of the resource competition case. One could understand the general methodological suggestion of neo-Darwinian theory as an injunction to use (mainly) genetics and natural selection to solve puzzling questions (i.e., problems) about the distribution and morphology of organisms. Constraining the domain to organisms slides into the general ontological commitment of neo-Darwinism. On the other hand, "specific ontology" is borne out in our case study by the Galapagos finches and, subsequent to Lack (1947), primarily the ground finches. The main "specific and testable laws about nature" in the resource competition case can be understood as follows. (1) The Galapagos finch radiation has been governed primarily by speciation in allopatry followed by adaptation to food resources and, especially in the ground finches, the effects of resource competition (Lack). (2) The Galapagos finch radiation has been governed by adaptation to food resources (and predation), but not at all by resource competition (Bowman). (3) The Galapagos (ground) finch distribution is due mainly (or 52 exclusively) to historical contingency (stochastic theorists). (4) The Galapagos ground finch radiation has been governed by both adaptation to food resources and resource 53 competition (Abbott et al., Schluter and Grant) . 168 It should be added that these candidates for "specific and testable" laws regarding the narrower domain of interest, although specific, were not tested until Abbott et al., and even then these proposed "laws" were not tested in a standard, experimental sense. Prior to Abbott et al., and sometimes afterward, explanatory statements about the Galapagos finch distribution were supported purely by observational evidence rather than by the confirmation of predictions. Whether Laudan would then consider these observations "testable laws about nature" is not entirely clear. However, they were eventually tested, although not experimentally, so presumably even when they were first formulated they might be counted as testable. Of course it would not be possible to test any of these suggestions about the historical finch radiation experimentally, given that it has already occurred. Anyway, Laudan is only committed to theories being testable in the usual sense. This is evident when he says "The individual theories ... will generally be empirically testable for they will entail (in conjunction with other specific theories) some precise predictions about how objects in the domain behave" (81). His position that they should "generally be empirically testable" may leave room for theoretical statements that are not testable, but amenable to confirmation or disconfirmation on the basis of observational evidence. Anyway, whole research traditions, in contrast to individual theories, "are neither explanatory, nor predictive, nor directly testable" (81-2). Another suggestion of Laudan's which may give us purchase on the appropriate way to understand his commitments within the Galapagos finch competition controversy is this: "the research tradition functions heuristically to suggest an initial theory for some domain" (91). We saw in Chapter II that Lack's (1947) suggestion that competition applied to the 169 Galapagos finch biogeography was novel to him. However, it was Darwin who first suggested resource competition as a factor in evolving lineages more generally. So we should understand resource competition as first an element of the Darwinian tradition. It is not yet clear, however, whether it can be understood as functioning heuristically so as to suggest competition in the specific case of the finches. Presumably to do so, resource competition should assume a methodological role in the Darwinian tradition. It would not form part of the general ontology of that tradition, for as we have seen, the Darwinian general ontology specifies that its focal domain is the set of organisms. However, another entity within a research tradition that resource competition might count as is one of its theories. I suspect that characterizing resource competition either as a theory within the Darwinian tradition, or as an aspect of its methodology would be consistent with Laudan's commitments. As an aspect of the tradition's methodology, resource competition would be part of the proscriptive aspects of the research tradition. That is, the tradition would suggest that one of the ways to account for the distribution and morphology of organisms is by means of resource competition. Darwin (1859) says that "the nature of the other inhabitants with which each has to compete, is at least as important [as the physical conditions of a country], and generally a far more important element of success" (quoted in Lack 1947, 115). This extract occurs in a passage in which Darwin refers to the organisms of the Galapagos. So competition is at the same time both a solution to a relatively specific problem and a more abstract member among Darwin's theoretical commitments. Then is resource competition between species empirically testable or not? Laudan seems not to 170 make explicit allowances for this sort of ambiguity between theory and research tradition, which can be seen as a deficiency in his view. There is a simple solution, however. It is to make room among Laudan's commitments for two broad classes of theory. One kind of theory is more general, and this class of theories can include even those that are initially suggested by observation, but their characterizing feature is that they are general. These can count as part of the methodology of research traditions. The other class of theories is more consistent with the way that Laudan specifies theories (quoted above) as applying to a narrow ontology, and as suggesting laws about nature that are directly testable (Laudan 1977, 84). If we take the passage from Darwin above to be indicative of Darwin's accepting resource competition as part of the metaphysical furniture of the world, then it can count as part of the ontological commitment of his tradition. Darwin's suggestion of competition, then, might have functioned in some heuristic way so as to suggest to Lack that competition could apply to the finches. It was Gause's hypothesis that was the proximal cause of Lack's acceptance of competition as applied to the finches (Lack 1947,62), so to be accurate, we would have to see Gause as an intermediary in the heuristic functioning of Darwin's tradition in ultimately leading to Lack's initial theory. Anyway, Lack was the first explicitly to apply resource competition to the Galapagos finches,54 and he would have been influenced in some way (and indirectly through Gause) by the existence of resource competition among Darwin's commitments. All this is to suggest that perhaps Laudan's condition of a research tradition's "function[ing] heuristically to suggest an initial theory" (91) may be met in the case of Lack 171 in relation to Darwin's tradition. There is one last wrinkle here: Lack was working after the shift to neo-Darwinism and his acceptance of the synthesis is evident in 1947.1 have been referring to the Darwinian tradition thus far with respect to the current question. However, as demonstrated in a quote in Section 2.2.1, the synthesis accepts competition as well (Mayr 1982, 567). It is possible to assess in more detail my claim that (1) - (4) count as theories within Laudan's framework, rather than as (small) research traditions in themselves, but I will take it that I have already made a plausible case for that conclusion. As detailed later, however, although (3) is plausibly understood as a theory, as I have it here, it might also (perhaps simultaneously) be understood as an independent incipient tradition. So far as research traditions in evolutionary biology go, it seems, then, that there is no current tradition below the level of the very large neo-Darwinian tradition.55 This may create a problem for Laudan in that for him progress is usually to be assessed by comparing the relative progress of whole traditions. As we have seen in the last chapter, progress can also be assessed within a particular tradition, but he would prefer to assess the relative progress of traditions. However, neo-Darwinian evolutionary theory seems not to have any viable rival research traditions, unless Creationism counts as one. 6.3 Problem-Solving Effectiveness in the Galapagos Finch Case Let us begin with Laudan's broadest notion of a problem. A scientist notices "a certain puzzling phenomenon, p. The phenomenon p constitutes an unsolved problem for the 172 scientist who wishes to develop a theory T\, specifically with a view toward resolvingp" (Laudan 1977, 67). In our case, the puzzling phenomenon is the distribution of finches across the Galapagos Archipelago: what caused them to assume the distribution that we see today? Again, for Laudan, theories are what attempt to solve problems, and I have already detailed above the relevant theories that have been put forth to resolve the finch distribution problem. As we have seen, there are two ways of evaluating progress within a research tradition: synchronic and diachronic-developmental (Laudan 1977, 106). The former assesses "the (momentary) adequacy of a research tradition" (ibid.), which is to be assessed, in turn, by summing up the problem-solving effectiveness of all of the current theories within a research tradition. This would be an onerous task in a tradition of the size and scope of neo-Darwinism. However, given that determining this measure is by means of a simple sum, looking at the effectiveness of a subset of theories will count toward that project. The diachronic-developmental form of evaluation evaluates progress and is a temporal measure (107). It has two submeasures: the general progress of a research tradition and its rate of progress, both of which are reliant upon momentary adequacy (ibid.). Given that all of these evaluative measures of a research tradition, and hence of the sub-components of a research tradition, reduce to momentary adequacy, I am justified in beginning by assessing adequacy. With respect to the weight of empirical problems, it is not until after there is a first theory that solves a particular problem that we can begin to talk about the importance of that problem (Laudan 1977, 33). 173 It seems initially a disadvantage of Laudan's account that he gives no guidelines concerning the substance of what would count as a solution to a problem; any bizarre "solution" counts as a solution for Laudan.56 This will be a technical problem for his account when there is only one solution to a problem. Simply put, he does not compare theories to the extra-theoretical reality they attempt to describe, but instead has a weighting system for theoretical problem solution that is irreducibly comparative. This unsatisfactory state of affairs might appear to be resolved as soon as there is more than one solution, for then we might expect the better to prevail. 6.4 Lack Lack's theoretical solution to the problem posed by the finch species distribution in the Galapagos Archipelago was his four-stage allopatric model. Our analysis is simplified by the fact that Laudan does not require that a solution be correct, although presumably Lack's is at least approximately so. Accordingly, it is not necessary to review additional evidence from our case of his model's correctness. Lack supported both adaptation and resource competition as mechanisms determining the finch fauna, with the emphasis on competition (Abbott et al. 1977,153). Competition mainly enters the allopatric model in the third step, "[sjecondary contact between original and derived populations" (Grant 1981, 654-5). We can see competition and adaptation as solutions to sub-problems within the larger problem of the radiation of the Galapagos finches. It is these smaller problems that I will mainly be 174 focusing on in the series of episodes presented by the Galapagos finch competition controversy. To find the problem-solving effectiveness (and adequacy) of a theory, we need to determine the number and weight of empirical problems solved, and subtract the number and weight of both conceptual problems and anomalous empirical problems (Laudan 1977, 68). I turn now to this task in the case of Lack. 6.4.1 Number of Problems Solved by Lack It is sometimes difficult to know how to delineate problems, and Laudan gives little or no guidance on this, seeming to take it as evident. For example, I have already alluded to the distribution of Galapagos finches as a puzzling phenomenon for which Lack sought a solution. Does this count as one problem for Laudan, or does each species on each island count as a separate sub-problem of this larger problem? The answer to this question is not clear. An initial suggestion about how to proceed could be to delineate the number of broad abstract classes of solutions, and for each of these assume that there is one problem. That is, if we take, for example, character displacement as one broad solution, then the class of problems to which it is a solution counts as one (kind of) problem. In other words, every time there are gaps in the smooth gradation between sympatric organisms that appear not to be solved by adaptation or other neo-Darwinian solutions, these taken together would count as a single problem. 175 To illustrate what would not count as an individual problem on this view, let us take a particularly small-grained problem, such as the one I will detail below, of why three sympatric species of Geospizae have gaps in beak morphology. On the understanding of problems presented directly above, this would not count as a problem unto itself because it is one of many that has the same solution: character displacement caused by resource competition.57 Another consequence of this "abstract classes as problems" way of viewing what problems are is that it might result in the following conclusion. Even the finch distribution problem is just an instance of a larger problem, and so not regarded as a problem proper in itself. This is because it might be characterizable more abstractly as an adaptive radiation; then adaptive radiations more generally could be understood as the relevant problem. So we see that there must be a hierarchy of problems ranging from the very small in scope to problems about the whole of a domain covered by a research tradition. It is the comparative evaluation of problems that matters for Laudan. So while it is instructive to a point to determine the size of problems that are understood by Laudan as proper problems, when comparing theories or research traditions, together with their ability to handle problems, what will count as problems will be more obvious. Presumably, the smallest problems that will be relevant to both theories (or traditions) will be those for which rival theories (or traditions) have different solutions. There is no need to look at any sub-problems of the problems at that level. In this context, I will take it that each micro-problem requiring solution (even if it is the same solution as in many other instances of evolutionary radiations) could count for 176 Laudan as an individual problem so long as some other theory or research tradition has a separate solution for it. To address every micro-problem within Lack's 1947 book would be tedious. This suggests that an assessment of a whole research tradition in this much detail would be a daunting task in practice. Almost certainly Laudan did not have this fine-grained scale in mind. However, if these micro-problems are to count as genuine problems for Laudan, there seems little option but to understand him as committed to the necessity of that degree of detail in analyzing progress. For all of the above reasons, I will not assess how many micro-problems were solved by Lack. Below I will compare some micro-problems solved by Lack versus some solved by Bowman, but this will constitute only a sub-set of all of the problems solved by each in their respective ways. I will not detail (and weight) every one of their solved problems. Lack solved the broad problem of the biogeography of the finches by means of his four-stage allopatric model. Counting this as a solved problem does not commit us, under Laudan's account, to how well he solved it or whether the "solution" was close to being true. 6.4.2 Weight of Problems Solved by Lack It makes sense in examining Lack's work here to confine the analysis to problems posed by the ground finches and, in particular, to a problem that is subsequently addressed by both Bowman and Abbott et al.: morphological gaps in sympatric Geospizae. More specifically, the phenomenon that Lack notes is the gaps in clustered beak depths between three 177 sympatric Geospizae on a number of islands where they occur. This is to detail a subset of Lack's (1947) work that is particularly relevant to the competition question. It is one of the more obvious problems (attended by his solution) that he lays out in his text. Moreover, Lack takes it as such an important problem that he talks about it at length in two different places: in his Chapter VI and Chapter VIII where the whole chapter is devoted to the question of food adaptation differences in sympatric congeners. We saw in Chapter II that Lack was particularly interested in explaining beak differences in closely-related species because these differences did not seem to be adapted to differential feeding (60). The clustering and gaps between beak depths of sympatric species of Geospizae (G. magnirostris, fortis and fuliginosa) is especially apparent on four islands (Lack 1947, 82). In 1945, because Lack "could not perceive any habitat variation among islands that would account for the observed morphological variation among the 58 finches," he espoused genetic drift and the founder effect (Abbott et al. 153). By 1947, as we have seen, he accepted genetic drift as a possibility in only two populations, and he emphasized competition as accounting for the morphological discrepancies not explained, in his opinion, by adaptation simpliciter. "The significance of these marked beak differences between species otherwise similar has excited speculation from all who have discussed Darwin's finches," he says (Lack 1947, 61). So the problem had been noted before and had at least been speculated about, probably "solved" in accordance with Laudan's minimal requirements regarding what counts as problem solution. I mention this since Laudan requires, before the weighting process can begin, that there be "one or more theories in the 178 domain" after which there can be "criteria for increasing the importance of certain empirical problems" (33). I turn now more directly to weight assessment of the empirical problem posed by the differences in sympatric ground finches. Although there was a question raised earlier about the imprecise account given of the nature of theories which counted as solving problems, Laudan does refer to "adequate" and "viable" theories although he does not appear to have any more detailed account of how they would get to have those characteristics (33). So for the sake of argument, let us assume that a more robust account of adequate theories consistent with Laudan's other commitments could be developed. "Once we have one or more theories in the domain ... we immediately have certain criteria for increasing the importance of certain empirical problems" (ibid.). As we have just seen, the different beak characteristics with respect to the sympatric ground finches formed a problem that was already solved by Lack in 1945, at least, if not by others who had speculated about the phenomenon. The 1945 solution is not important for our purposes here, except insofar as it provides us with an earlier theory solving the relevant problem so that we are able to apply Laudan's weighting criteria. Laudan says that if "a particular problem has been solved by any viable theory in the domain, then that problem acquires considerable significance;" this problem then accrues importance or weight because any other theory in the domain will then also have to solve it or give principled reasons why it need not (33). This constitutes Laudan's category of problem weighting by solution. The example Laudan gives in this section is of Galileo solving the problem of free fall; "every subsequent theory of mechanics was under strong 179 constraints" to provide as good a solution (ibid.). By "theory" in the latter quote it is pretty clear Laudan means "research tradition."59 So we see that the example he gives is not of the same scope as that of the three sympatric ground finches, which might be considered a micro-problem within neo-Darwinian theory. Furthermore, there are no (credible) research traditions competing with neo-Darwinism on the same scale, so what Laudan says here regarding other theories (traditions) having to account for the same phenomenon will not apply to competing traditions at the same grand level, for there are none.60 Anyway, Laudan here refers to problem inflation by solution. Problems, as we saw earlier, appear to exist at all levels of generality and scope, so it should not count against the current analysis that Laudan's own example of a problem is of a quite different scope than that detailed here. So Lack has solved a problem in the domain that has already had at least one solution, which makes the problem currently under consideration at least minimally weighted. Given that a problem can become an anomaly for earlier theories once it has been solved by later theories, Lack's competition solution applied to the sympatric ground finches will count as an anomaly for adaptation which (putatively) did not solve it. As such, it becomes more acute for advocates of the adaptation hypothesis to solve this problem, something Bowman will take up, as we shall see. At the same time, this raises an interesting difficulty for Laudan. The neo-Darwinian tradition is a disjunction of theories, some of them contradictory. The disjunction of course includes selection and competition. So one disjunct within neo-Darwinism can, in principle at least, solve a problem that is not solved by another disjunct, and thus present the second disjunct with an anomaly that it needs to solve. However, this does not make selection a 180 poorer solution in general, even though Laudan's commitments force us to say that this particular problem is an anomaly for adaptation. Usually adaptation does solve problems involving the characteristics, and to some extent, the distribution of species found in a region. So in the present example, we see that competition will have a problem weighted in its favour compared to adaptation, but that is just one more weighted problem to be taken into account in assessing the numbers and weights of solved problems. What I want to suggest is that where Laudan's weighting (and numbering) system will give differential weights to these two theories, that does not translate simply into what we see by inspection. That is, we see that adaptation is a good theory for addressing some problems, and competition is a good theory for addressing other problems, but this is not captured by Laudan's analysis. All we get at the end of the day is (at best, if it were possible to put exact weights on problems and their solutions) some kind of measure that takes into account only the weighted sum of problems solved. In the case of a tradition made up of a disjunction of theories, some of those theories will be good for some kinds of solution, and some kinds will be good for others. In practice, this tension may not be a big problem for adaptation versus competition, since competition is a kind of adaptation, but it could present problems, for example, when the ability of genetic drift to solve problems is compared to the ability of adaptation to solve other problems. In summary, the three ground finches present a weighty (and hence important) problem on Laudan's account. 181 6.4.3 Conceptual Problems and Anomalies in Lack External conceptual problems exist as tensions between the focal theory and some other theory and, according to Laudan, these have been historically the more significant class of conceptual problems (Laudan 1977, 50). Of course, it is difficult to imagine theories from non-biological domains that would be in conflict with resource competition61. Presumably any that were would be in conflict with neo-Darwinism more generally as well. As such, if there are external conceptual problems associated with Lack's work, they will be conceptual problems that count equally for at least some other theories contained within the neo-Darwinian tradition. The two kinds of internal conceptual problem are logical inconsistency within a research tradition, and "conceptual ambiguity and circularity" (49). Actually, in these passages, Laudan makes reference to theories rather than research traditions, but because he uses theories as stand-in entities for research traditions early in his book, it is plausible to assume that he means to be referring to research traditions here as well. So internal conceptual problems are to be understood as internal to a whole research tradition. Lack's competition hypothesis is just one small theoretical element within neo-Darwinism. The neo-Darwinian tradition seems to explain the phenomena in its domain as being due to one or more of a number of possibilities. If on one occasion, one kind of explanation applies, and a different kind of explanation applies at a different time, that does not seem to make it internally inconsistent in the way that Laudan has in mind in outlining this kind of internal conceptual problem. Therefore, it is hard to see how resource competition's 182 supposedly applying to a particular case would be seen as contradicting other aspects of neo-Darwinism. As to ambiguity, it does not seem particularly ambiguous, and Laudan allows a modicum of ambiguity in any case (50). There are no anomalies yet in Lack, although the very fact that Laudan allows that anomalies for a theory can show up later means that the non-existence of anomalies in 1947 does not predict that there will never be any. Indeed, some will appear, as we shall see below. 6.4.4 Problem-Solving Effectiveness in Lack We are now in a better position to talk about adequacy or problem-solving effectiveness. Specifically, Laudan says, the overall problem-solving effectiveness of a theory is determined by assessing the number and importance of the empirical problems which the theory solves and deducting therefrom the number and importance of the anomalies and conceptual problems which the theory generates (Laudan 1977, 68) Since by "the (momentary) adequacy of a research tradition" Laudan means the problem-solving effectiveness of the most recent theories of that research tradition (106), Lack's 183 competition solution would contribute to the adequacy of the neo-Darwinian tradition assessed in 1947. 6.5 Competitors Laudan gives very little in the way of guidance as to when we are to understand a scientific conflict as a conflict between research traditions, and when we should view it as a squabble internal to one research tradition. However, as we saw in the last chapter, for Laudan, progress can be assessed whether or not the theories that lead to it are understood within the context of a research tradition. In the Galapagos finch competition case, it is not always clear whether Laudan would understand a particular scientific conflict as subordinate to a research tradition, or its own (perhaps budding) research tradition. Furthermore, the scale of the finch resource competition case is such that I will not, in most cases be able to compare the relative progressiveness of alternative research traditions. 6.6 Bowman According to Laudan, we cannot assess progress before there are competitors: "All evaluations of research traditions and theories must be made within a comparative context" (Laudan 1977,120). Again, the theories of Bowman and Lack count as part of the same research tradition. We can, however, compare them as competitors of a sort. Laudan asserts, "What matters is not, in some absolute sense, how effective or progressive a tradition or 184 theory is, but, rather, how its effectiveness or progressiveness compares with its competitors" (120). This passage implies that both theories and research traditions can have competitors. So it should be allowable within Laudan's machinery to assess progress between competing theories, both of which are members of the same research tradition. Bowman and Lack clearly have different theories. Indeed, Bowman pretty clearly sees Lack's competition hypothesis as a theory competing with his own, as evidenced by his arguing against it at length and suggesting that adaptation to food is a complete explanation of the aspects of finch morphology and distribution that Lack attributes to competition. However, a peculiar characteristic of Laudan's account arises when we attempt to compare Lack's problem solutions to Bowman's. We need to do this for the purposes of assessing their relative problem-solving effectiveness as a precursor to determining the progress that they afford their (same) research tradition. Bowman and Lack give alternative explanations of the very same phenomenon in every case in which Lack suggests that competition has occurred. (I am comparing here only those cases, for they are not in competition where Lack, too, postulates adaptation to food resources.) Bowman and Lack should each solve equally many problems of equal weights, given that they are the same problems. Both theories count for Laudan as perfectly fine solutions. The only way to assess them comparatively is by means of "assessing the number and importance of the empirical problems which the theory solves and deducting therefrom the number and importance of the anomalies and conceptual problems which the theory generates" (Laudan 1977, 68). We cannot even say that one presents an anomaly for the other, since they both claim to account for same the data, and Laudan's definition of anomaly is "those empirical problems 185 which a particular theory has not solved, but which one or more of its competitors have" (Laudan 1977, 17). Finally, we are left with possible conceptual problems that each theory generates. Since the theories are different (although the problems are not), it is possible in principle that they could have different conceptual problems associated with them. However, the same considerations apply with respect to conceptual problems as did for Lack (see Section 6.4.3). In other words, we cannot compare Lack's and Bowman's theories with respect to conceptual progress either. This leaves us with no way comparatively to assess them. Alternatively one might then conclude that the two theories have equal 62 problem-solving effectiveness Either way, this problem demonstrates, at very the least, that Laudan's account is incomplete. The progress inherent in the work of Lack and Bowman cannot be captured by Laudan. 6.7 Abbott, Abbott and Grant Abbott et al. generate the first anomalies for Lack and Bowman. They attempt to tease out consequences of Lack's and Bowman's central hypotheses, and then test those consequences. The results of this procedure vary. In some cases, Bowman's floristic theory is supported and, in other cases, it is Lack's resource competition theory that is. Where one theory accounts for the relevant data, but the other does not, this counts as an anomaly for the latter. For the first time, therefore, it is possible (at least in principle) to apply Laudan's comparative criteria to the two theories and have different weightings. Since everything else is equal, as before (viz., number and weight of problems solved and conceptual problems), 186 whichever theory produces the most and weightiest anomalies will have less problem-solving effectiveness. Much of Abbott et al. (1977) utilizes a crucial experimenta style of argumentation. That is, the researchers attempt to derive testable predictions from each of Lack's and Bowman's theories, and then determine which prediction obtains in the Galapagos for the relevant data sets. Let us take the same example from Abbott et al. that was used in Chapter IV. I will have more to say in Chapter IX about the relative merits of the two philosophers in accommodating this example. The competition hypothesis implies that "islands with only two breeding sympatric Geospiza species should hold only two such species of very dissimilar beak depth," and this is what is in fact seen (Abbott et al. 1977, 164). Here we see the presentation of a problem which Lack's theory is demonstrated to solve. By comparison, Bowman's hypothesis would suggest that "overlap between sympatric pairs ... would tend to be highest when the lowest variety of foods is available" (Abbott et al. 1977, 164). The way this was assessed was "by relating overlap in the seed and fruit types eaten by each sympatric pair of Geospiza species to diversity of seeds and fruits available," and the result was that "[n]o significant correlation was found" (ibid.). To determine whether this constitutes an anomaly for Bowman, we first need to ascertain whether the problems being compared here are indeed the same problem. The dissimilar beak depths predicted by Lack where there are two sympatric species could be understood as the negation of a situation in which two sympatric species demonstrated convergence of beak depths in the presence of a low variety of foods. Conversely, finding no correlation between overlap in beak depths between species is equivalent to finding that 187 the negation of the state of affairs predicted by Lack's theory does not obtain. In this convoluted way, the two theories can be seen to be about the same phenomena in at least some cases. Therefore I take it that Lack's theory can account for the same phenomenon that Bowman's cannot. In other words, Lack's confirmation instance here counts as an anomaly for Bowman. It would be tedious to assess, in each crucial experiment set up by Abbott et al., which of Lack's or Bowman's theory accounts for the specific micro-phenomenon which is anomalous and then to sum these up for the purposes of comparing them. What is evident here is that this procedure would be possible. I would also here like to gesture at the way in which progress might be assessed in this micro-example. Again, because of considerations to be put forth shortly, a detailed assessment of the relative progress of all of Bowman's and Lack's theories is moot, although confining the analysis to a few of these is coherent. Laudan suggests that "any time we modify a theory or replace it by another theory, that change is progressive if and only if the later version is a more effective problem solver... than its predecessor" (Laudan 1977, 68). Here we do not have a replacement of either of Bowman's or Lack's theories, but we can assess whether those theories count as being modified. They have been modified only insofar as some of their implications have been determined and described in the context of a particular fauna. This does not count as a modification of their core hypotheses, but might be seen as an expansion of their commitments insofar as the consequences of their theories have been made explicit. If this were to count as a sort of modification of theory for Laudan, 1 8 8 it would allow the respective theories here to be understood as making progress, and the one with the least weight multiplied by the number of anomalies to be more progressive. An additional consideration should be raised here. Using Laudan's machinery comparatively to assess these two theories in this way is not entirely in line with Laudan's intentions for comparative progress. Abbott et al.'s purpose in their study was to amass "data that would allow a rational choice between Lack's competition hypothesis and Bowman's floristic, noncompetition hypothesis, or a reconciliation of the two views" (Grant 1981, 656). In their final analysis, they conclude that "both floristic diversity and interspecific competition have influenced the morphology and ecology of Geospiza finches" (Abbott et al. 1977, 175). They found both processes to be operative, and the two theories are therefore not strictly in competition, except perhaps in a few individual cases in which one or the other seems to solve a problem that the other does not. For these reasons, although it is possible to compare the relative progress of these two theories in the way outlined here, the ultimate reason for doing so may be lacking. That is, presumably the main reason for comparative assessments of the progress of theories is to help us ascertain something about the relative merits of (competing) theories in adding to our knowledge and understanding. Here Abbott et al. conclude that both of these theories do so. Accordingly, it seems odd to take an anomaly for one to be indicative of general superiority of progressiveness of the other. Where an anomaly for one may indicate superiority of the other is with respect to individual micro-problems that one case can handle but that the other cannot. At least according to Abbott et al., the following case would have to be considered to present Bowman with an anomaly. There are "several regularities in the distribution and 189 morphology of the finches that are predicted by a hypothesis of interspecific competition for food," but which are not amenable to Bowman's solution (Grant 1981, 657). For instance, there is a positive correlation between beak size and the diet which is provided by the plant life on eight islands, but "there is not a one-to-one relationship between peaks in the frequency distributions of beak sizes of sympatric species and peaks in the frequency distributions of seed and fruit size and hardness" (ibid.). In other words, there are gaps in the depth of beaks of sympatric species that are not consistent with the hypothesis that only adaptation to food resources mattered in the history of these birds. This leads to the "implication that one of the missing species has been present but has gone extinct for reasons of competitive inferiority" and suggests that character displacement has occurred (Abbott etal. 1977, 176). So, relative to this narrow sub-problem within the larger problem of the distribution and morphology of the finches in the Galapagos, Abbott et al. seem to have isolated an anomaly for Bowman's solution by demonstrating that Lack's solution does solve this problem. With respect to the larger problem of the finch biodiversity, as we have seen, each of these solutions solves parts of this larger problem, and so they are not rightly compared for their comparative progress. Laudan does, however, make allowances for assessing progress internal to a research tradition, that is, without comparing it to competing research traditions, which he says is "an approximate determination of the effectiveness of a research tradition" (Laudan 1977,145). This is to be accomplished by determining "whether a research tradition has solved the problems which it set for itself while generating a minimum of new anomalies and conceptual problems, and reducing existing ones (146). 190 Presumably then, it would be possible to assess individual theories internally as well for an approximation of their progressiveness, and so one would not have to compare them to other theories. This might be a way of getting at the approximate progressiveness of each of Bowman's and Lack's problem-solutions, but it would not count for Laudan as a true assessment of their progressiveness since he is committed to the irreducibly comparative nature of progress. 6.8 The Stochastic Challenge We can understand the stochastic theorists as solving the broad problem of the distribution of species in the Galapagos. Simberloff (1978) uses some plant and insect data as the basis of his stochastic solution. It is more a matter of his using that data to make his case than a matter of the data constituting the problem itself. Conner and Simberloff (1978) examine all of the bird and plant species of the islands in their analysis. Finally, Strong et al. (1979) directly address the problem of the biogeography of the Galapagos ground finches (as well as two other bird groups). The results of Conner and Simberloff would count equally as an anomaly for Lack's competitive exclusion (the procedure did not test character displacement) and for Bowman's adaptation to food resources theory-if they truly count as an anomaly. For the second (more realistic) null hypothesis of Conner and Simberloff, there is a significant difference between the observed distribution of bird species and one expected by purely stochastic processes 69.5% of the time (Conner and Simberloff 1978, 244). They conclude that "a substantial 191 proportion of the number of species shared between two islands can be viewed as resulting from stochastic processes of persistence and dispersal" (ibid.). There is a question about whether these results truly count as an anomaly for Lack's and Bowman's theories as solutions to the finch distribution problem, since it tests all of the bird species. Also, their model shows that the distribution that is in fact seen should occur by chance only (close to) 30% of the time. That leaves room for either Bowman or Lack to be right in specific instances. Accordingly, we do not have enough information to assess whether Conner and Simberloff s results constitute anomalies for the fine-grained problems to be encountered in Lack's and Bowman's problem-solution. This raises an important issue in comparing traditions (for, as we will shortly see, the stochastic challenge might be understood as its own tradition). Furthermore, in assessing the relative merits of traditions, unless one falls back upon Laudan's approximate measure of internal progress, it may not be possible to compare two traditions at the fine-grained level in their ability to accommodate the same problems. This is for the simple reason that when the two traditions are examined in sufficient detail, they may be seen not to be addressing exactly the same micro-problems. The two commitments of Abbott et al. that Strong et al. take themselves to be challenging are the existence of character displacement in the Galapagos ground finches (Strong et al. 1979, 900), and "that there is greater character displacement on islands with fewer Geospiza species" (901). The result will count as an anomaly for Lack only if his account does not have a solution to exactly the same problem. Lack's solution to finch distribution in general implies that two species existing together would show a ratio of beak sizes greater than would be expected by chance, i.e., demonstrating character displacement. 192 The average of beak length ratios of sympatric species chosen at random from a pool of all the finch species of the Galapagos archipelago should be smaller than the actual ratios of species in which character displacement has occurred. Strong et al. found that the beak size ratios observed are consistent with chance, and so there is no need to postulate competitive forces as a general phenomenon influencing the finch distribution (Strong et al. 1979, 901). This leaves open the possibility that the ratios of those sympatric species that are in fact higher than the ratio expected by chance may be so due to character displacement. Once again we see that if we were to attend to a sufficiently fine-grained account of the problem of the Galapagos finch distribution, it is unclear whether the stochastic results would provide an anomaly for Lack's solution. This result does seem to constitute an anomaly for Lack at the general level of distribution of finches in the archipelago. If we take his solution to the distribution to be something like the proposal that competition is usually at work, Strong et al. have produced data to show that this does not seem to be the case. In other words, Strong et al.'s results, that two species existing together do not show a ratio of beak sizes greater than would be expected by chance, is a problem that has to be accommodated somehow by Lack's competition solution; otherwise, it constitutes an anomaly for Lack and his account will be less progressive, when it is ultimately assessed over time. The researchers also find a trend in the same data "consistent with Bowman's (1961) interpretation of intraspecification variation in Galapagos finches" (Strong et al. 1979, 907). The anomaly for Lack is not also an anomaly for Bowman in this particular instance, and so Lack's solution would fall slightly behind on this basis if an assessment of progress were to be made. 193 I turn now to the question of whether the stochastic challenge constitutes its own budding research tradition. Although I think it would be wrong to present initial stochastic colonization as a contradiction of any of the neo-Darwinian core commitments listed by Futuyma (1998, 26-7), it certainly presents a difference of emphasis. At any rate, the neo-Darwinian tradition appears to take it as axiomatic that populations will change over time either in response to genetic drift, selection, recombination, or combinations of these. However, one might understand the (or a) stochastic challenge in one of two ways. First, it could be suggesting that the species arrived initially in the Galapagos Archipelago in a random manner. Second, this randomness is still primarily responsible for the distribution we see today, as opposed to selection's being primarily responsible for that biogeography. If we do attribute this second, stronger commitment to at least some of the stochastic theorists, then their commitment does appear to contradict the core commitments of the modern evolutionary research tradition. Presumably the existence of at least one contradictory core commitment is a sufficient condition to render it (the core of) a conflicting research tradition. Therefore it appears that a strict understanding of stochastic researchers' commitments results in viewing their work as part of its own tradition. Now obviously, it is only an incipient tradition, and not a full-blown one in the sense of the neo-Darwinian evolutionary tradition. Nevertheless, having what can be understood as a rival tradition gives us the opportunity to assess the relative progress of it and neo-Darwinism. I do not think I need to defend the progressiveness of neo-Darwinism. Its scope must be very wide indeed and, in comparison, the stochastic tradition is at best a 194 promising upstart. I therefore take it for granted that no matter how progressive it turns out to be, its general progress would not count as greater than that of neo-Darwinism. 6.9 Schluter and Grant I begin with expected landscapes. As before, the problem requiring solution is the finch distribution, here specifically the distribution of the Geospizae. "Our initial premise is that food characteristics are potentially the most important determinants of finch morphology on Galapagos islands" (Schluter and Grant 1984, 176). This is clearly Bowman's floristic solution to the finch distribution problem. Schluter takes this hypothesis for granted and uses it as the basis of his computer model of expected densities. If this is a sound original assumption, and given all of the available data combined with some innovative data manipulations, it should be possible to "predict" the optimal beak depth of the ground finches that are likely to be found on the islands. More accurately, the calculations produce expected population densities given the food availability, food preferences as assessed by beak depths of finches found on the islands and the biomass of finches supportable by the amount of seeds present. Combined with the assumption of optimality, expected population densities calculated in this way should predict the mean beak depths of the species to be found on those islands. So, given the initial assumption that Bowman's hypothesis is correct in its essentials, the resulting graphs are a good test of Bowman's solution to the finch distribution problem. 195 We saw before that "progress can occur if and only if the succession of scientific theories in any domain shows an increasing degree of problem solving effectiveness" (Laudan 1977, 68). So, in order to show that the expected densities result demonstrates progress, it has to be demonstrated that its problem-solving effectiveness is superior to previous theories in the domain. We have seen that this result counts as solving the general problem of distribution of the finches, and doing so using Bowman's hypothesis (or problem-solution). Curiously, although the result is demonstrated with much more definitiveness of evidence and argument, the number of problems it solves is unchanged. Let us take it that each finch species present that corresponds to a peak in the expected densities graph counts as a separate solved problem. Now, although presumably the number of problems Bowman's adaptation to resources solution has been demonstrated individually to solve has increased, the number of problems that have been solved by Bowman's floristic hypothesis should not be understood to have increased. This is because Bowman would always have been committed to adaptation's being applicable to those same species, even though he might not have mentioned every case individually. He was committed to floristic adaptation being what governed all of the finch radiation. An increase in problem-solving effectiveness is to include an increase in "the number and importance of the empirical problems which the theory solves and deducting therefrom the number and importance of the anomalies and conceptual problems which the theory generates " (Laudan 1977, 68). Once again, the problems are the same as before, and so we should not expect their importance to have changed, nor for the anomalies and conceptual problems they generate to have increased, since they are the same problems as 196 always. Might it be the case that because the way that the problems have been demonstrated to be solved is now different, that they are at least candidates for a change in the conceptual problems they generate? Even if so, it is difficult to see how any conceptual problems would be generated by assuming that the calculations of expected densities are a viable way to proceed. I turn now to the results of the five-way test. If we take the gaps in mean beak depths to be inexplicable on the adaptation hypothesis (as suggested by Abbott et al.), this will count as an anomaly for the floristic solution to the problem, and a very weighty problem for the competition hypothesis to solve. However, competition has already been claimed to solve the problem. What the five-way test does is to give additional reasons to believe that the competition problem is adequately solved for the Galapagos finches. Once again, Laudan's account is hampered by his insufficient guidelines as to what counts as an adequate or better solution. The solution is the same, but Schluter and Grant justify it better than before. We might say that they have given better argument and evidence for competition as a solution to the Galapagos finch radiation problem. Laudan's account does not make this improved justification of the solution progressive. 6.10 Accumulation of Facts We have seen that by Laudan's criteria, no progress in the Galapagos finch controversy occurred before Abbott et al. However, that is not to say that nothing new was discovered from the beginnings of this casestudy until then. We have seen in Chapter II that both Lack 197 and Bowman contributed new facts, that were not previously known about the Galapagos finches. I want to suggest that these facts are not best understood as solutions to any specific problems. Perhaps they might be understood as solutions to the general problem of lack of knowledge of all of the facts of the world, but this is clearly not the way Laudan understands a problem needing a solution; he has something much more specific in mind. I want to suggest that increases in knowledge regarding the facts in the world count as progress, but Laudan's account does not make room for this kind of progress. There is evident in the episodes detailed in the Galapagos finch competition controversy a first step of gathering information about the world before that data can even suggest the very problems that Laudan takes as central to his account. It should not be the case that the only value of that initial data is to provide the scientist with problems to solve, but this is the only function it can have for Laudan, given his stated commitments. 198 Chapter VII Kitcher on Scientific Progress In this chapter, I set out the account of progress detailed in Philip Kitcher's The Advancement of Science (1993). I begin by laying out his notion of consensus practice, which serves something of the same function as Kuhn's paradigm (disciplinary matrix) (Kuhn 1971). Next there is some preliminary material regarding the constraints within which Kitcher is operating for present purposes. Thereafter I lay out the main kinds of progress that Kitcher details: conceptual, explanatory and erotetic. Of these three, explanatory progress is perhaps the most significant for present purposes in that it introduces the schemata around which Chapter VIII is structured. The section on erotetic progress is important primarily because of the introduction of the notion of significant questions. In turn, significant questions will be especially relevant to progress in the statements scientists accept, which is one of the most important kinds of progress in the Chapter VIII analysis. This kind of progress is detailed in the next section on other kinds of progress. Finally there is a section on progress, cumulativity and significance. In this section I will be teasing out some of the implications for progress of Kitcher's ideas on significance and evaluating the question of whether an accumulation of phenomena counts as progressive for Kitcher, given his other commitments. 199 7.1 Consensus Practice Kitcher says, "In conceiving of science as progressive we envisage it as a sequence of consensus practices that get better and better with time" (Kitcher 1993, 90); furthermore, "If we are to understand the progress of science, we need to be able to articulate the relations among successive consensus practices" (87). Therefore the first order of business is to explicate what Kitcher means by consensus practice. Says Kitcher, "The consensus practice of a community at a given time is ... represented by (i) the core consensus ... (ii) the acknowledgements of authority . . . (iii) . . . subcommunities ... (iv) a virtual consensus ..." (Kitcher 1993, 88). I will touch on each of these in turn. What Kitcher is calling "the core consensus" is perhaps the most important element of those mentioned; it is constituted by "the elements of individual practice common to the individual practices of all members of the community" (88). In turn, individual practices are made up of seven components: (1) the language used by the scientists, (2) the questions they take to be the significant questions in their discipline, (3) the "statements (pictures, diagrams) they accept about the subject matter of the field," (4) the schemata underlying what are, in their view, the explanatory texts of the field, 200 (5) exemplars of credible informants in their discipline and the criteria of credibility, (6) "the paradigms of experimentation and observation, together with the instruments and tools which the scientists takes to be reliable, as well as [their]criteria for experimentation, observation, and reliability of instruments," and (7) "Exemplars of good and faulty scientific reasoning, coupled with the criteria for assessing proposed statements (the scientist's 'methodology')." (74) Component (ii) of the consensus practice of a scientific community was the communally accepted "acknowledgements of authority." This means that certain scientists' views within the discipline are accepted as authoritative on issues of importance in the discipline. As a corollary of this, (iii) the community is divided into accepted subdisciplines that are widely recognized as authoritative on specific issues-a sort of "division of authoritative labour." This leads to (iv), which Kitcher calls a "virtual consensus." This is a kind of consensus that is in one sense not actual, since no individual in the community is aware of all of the accepted elements of the consensus, although these could, in principle, be listed. Kitcher admits that his notion of consensus practice shares similarities with Kuhn's notion of paradigm, but he says that his own variant does not agree with what he calls Kuhn's commitment to "significant epistemological differences between the course of 201 science within a segment and the intersegment transitions" (Kitcher 1993, 87 footnote). Indeed, he argues explicitly against "Kuhn loss," and for cumulativity in science (116, 173-177). I will say more about this in the cumulativity section below. 7.2 Preliminaries As a simplification, Kitcher ignores, during most of his analysis of progress, that scientific fields split and merge. As mentioned above, progress is to be assessed by comparing successive consensus practices within the same "lineage" and evaluating the changes between them. This is to be done by comparing them pairwise (Kitcher 1993, 90). There are two important points to mention here. First, in comparing them pairwise, it is important that the last member of a sequence of practices be more progressive than the first. This is more important than that the latter of every pair has demonstrated progress; he calls the former state of affairs "weakly progressive" (91). Second, there are several different dimensions upon which the progress of consensus can be compared. Unlike Laudan, Kitcher does not weight the different dimensions. Furthermore, his account of progress is "keyed to investigating the fine structure" of science (123) and hence to the fine-grained detail of the changes that occur within consensus practices. Kitcher contrasts practical ("the relief of man's estate") with cognitive or epistemic progress, and says that he is concerned here only with the latter (92). He is a scientific realist and holds that "part of scientific progress consists in accepting statements that are both significant and true" (117). What he counts as true, though, is what is given to us by 202 our best current science, and he uses this standard of truth in evaluating past science (90). This will trouble not only the anti-realist (whom Kitcher addresses at length in his Chapter V), but cautious realists who suppose that some of what scientists now believe is likely to 63 be false . Truth, however, is not the only thing that matters; what we want, he suggests, is significant truth (94), which I will say more about later. The basic idea of significant truth is simple enough: some truths are not very interesting. Kitcher also appeals to the legitimate goals of science in his explications of individual kinds of progress as well (e.g. 104, 111). There are seven kinds of progress that Kitcher mentions explicitly: (1) conceptual, (2) explanatory, (3) erotetic, (4) experimental-instrumental, (5) progress in the statements scientists accept, (6) methodological and (7) organizational progress. The first three are the main ones, and he details them at some length. It is uncertain whether he considers the others sufficiently important to warrant their own categories, but it is important to recognize that he accepts all of these other four types. He also alludes to progress toward greater verisimilitude (Kitcher 1993, 117; 120-124), although he does not specifically give a name to this sort of progress, unless he intends "verisimilitude" itself to incorporate the notion of progress. Presumably he would also be willing to count as progress any beneficial changes along any of the dimensions he recognizes for consensus practice. 7.3 Conceptual Progress As an overview of what he means by conceptual progress, Kitcher states that conceptual progress "is made when we adjust the boundaries of our categories to conform to kinds and 203 when we are able to provide more adequate specifications of our referents" (Kitcher 1993, 95-6). So, in Kitcher's own example, our current use of the term "planet" has undergone progress since the pre-Copernican use of that term. The extension of the term has expanded to include more members of the kind planet, while our conceptual understanding of what a planet is has become more sophisticated. It is the latter process that Kitcher is particularly concerned to flesh out. In addition, what he calls the "ideal situation" is the case in which a scientist obeys all of the following three maxims: "Refer to that to which others refer" (conformity), "Refer to natural kinds" (naturalism) and "Refer to that which you can specify" (clarity) (104). One of the important goals of science according to Kitcher is to generate a language in which these three maxims are not at odds with one another, a language with reference terms such that it is possible to obey all three maxims simultaneously. Accordingly, conceptual progress is to be "assessed in terms of proximity to the ideal state" (ibid.). More specifically, progress ("Improvements") is supposed to "come about by abandoning modes of reference that are not in accord with one of the maxims or by adding modes of reference that would be in accord with both clarity and naturalism" (105). 7.4 Explanatory Progress "Explanatory progress consists in improving our view of the dependencies of phenomena" where some phenomena are considered by scientists to be explanatorily prior, and other phenomena are to be explained by their dependence upon (or reduction to) those prior phenomena (Kitcher 1993,105). More specifically, explanatory progress is made when 204 there are improvements in the explanatory schemata of a subsequent consensus practice over an earlier one (106). In general, Kitcher's schemata take the form of a standardized argument, headed by a question that is answered by the schema's conclusion. Each of the premises of the schema is a schematic sentence that replaces some of the nonlogical expressions with variables (82). There is also a set of filling instructions, usually left implicit, for what sort of entities are to replace the variables. Of particular note, some of the premises can "hand-wave" at input from future research, and disjunctive premises and disjunctions among premises are also acceptable (e.g. Kitcher 1993, 45). For the realist, a commitment to there being improvements in an explanatory schema presupposes that there is "an objective order of dependency in nature" that a later schema better captures64 (106). It is worth mentioning that Kitcherian schematization presupposes that philosophers of science are able coherently to extract a schema from both the consensus practice of a whole scientific discipline and from the work of a scientist or scientists who are part of a subset of that discipline. Kitcher's justification for his SIMPLE INDIVIDUAL SELECTION schema is the claim that "it underlies the simplest illustrations Darwin provides of natural selection" (28). In other words, he extracts from instantiations, and his view is that the schema itself is implicit in the instantiations. Additionally, as here, it is often instantiations that lead to improved explanatory schemata in the first place, so there is a reciprocal relationship between instantiations and the schemata that they instantiate, and each can help generate the other. There are four processes that Kitcher explicitly counts as explanatory progress (Kitcher 1993,109-110). First and second there are "the introduction of correct schemata" 205 (109) and "the elimination of incorrect schemata" (110). "Correct" is to be understood as identifying "a class of dependent phenomena and specifying] some of the entities and properties on which those phenomena depend" (111)65. Notice that a schema can be considered correct even if it does not tell the whole story of dependencies. Next there is what he calls an increase in the completeness of schemata, or "the generalization of schemata, rendering them able to deal correctly with a broader class of instances" (110). A more complete schema "identifies a more inclusive set of relevant entities and properties" in the relevant domain of the external world than its less complete counterpart. Alternatively it "is correct for a more inclusive class of dependent phenomena" (111). This latter is the sense in which increased completeness is a form of greater generalization of a schema. One might notice, although Kitcher does not say so, that both of these facets of increased completeness are increases in the class of things to which a schema can refer; this is analogous to the increased reference potential of terms, that was one of the two ways in which conceptual progress was understood to happen. Here, of course, the reference is not to kinds, but to dependencies. The other aspect of improvement of a conceptual term was increased clarity in its intension. Presumably the analogue here, viz., increasingly clear specifications of the nature of the dependencies, would count as explanatory progress as well. The fourth and final kind of explanatory progress is "explanatory extension, when the picture of dependencies is embedded within some larger scheme" (110). To be an extension of a schema, the improved schema must have "a schematic premise ... derived from" the prior schema (111). Furthermore, to count as a correct extension of the prior 206 schema, "the properties attributed to the entities in instances of the conclusion [in fact] depend on the entities and properties referred to in the corresponding instances of the premises" (111). His main example of "extending] the explanations" of a prior schema is that of NEO-DARWINIAN SELECTION extending the schema that Kitcher attributes to Darwin: SIMPLE INDIVIDUAL SELECTION (47). The "precise sense[s]" in which the former extends the explanation of the latter is that it (1) "preserves the main structure of Darwin's selectionist patterns (while [2] enlarging the scope to treat a broader family of questions) and simultaneously [3] derives statements that had previously been taken as premises" (ibid.). It is noteworthy that for all this talk of improvements of schemata, instantiations of schemata are not to be understood as explanatory progress. As to whether some other kind of Kitcherian progress might be involved in finding new instantiations, I address this question below. 7.5 Erotetic Progress and Significant Questions Erotetic progress is progress associated with the asking of better or more significant questions. For Kitcher, issues surrounding explanatory schemata which generate significance, and these issues are of two main kinds (Kitcher 1993, 112)66. First, there are issues generated by attempts to demonstrate instantiations of explanatory schemata, i.e., to demonstrate that the schemata apply to phenomena in nature. Kitcher terms questions regarding this first class of issues Kitcher terms "application questions." Finding instantiations of a schema is important and significant, especially when the schema is new 207 and there have not been (many) paradigmatic instantiations yet detailed. Thereafter they decrease in significance "because the record of success in instantiating the schema gives everyone confidence that they could ... [be instantiated], and the task of grinding out the details looks like hack work" (113). Questions surrounding the further instantiations of that particular schema are then significant only when they "involve special difficulties of producing instantiations" and such questions derive their significance from their potential to lead to improvements and extensions of the existing schemata (ibid.). The way that schemata generate important new instantiation questions can be seen in the following: New questions thus arise as the new schema makes it possible to inquire which kinds of instantiations are most important to the emergence of various major features of living things and to pose precisely problems about the strength of selection or the tempo of evolution. In their turn, these questions inspire new types of observational fieldwork and experimental activity: it becomes important to find ways of measuring fitness values, mutation rates, genetic variation, and so forth. (Kitcher 1993,48) Here it is the NEO-DARWINIAN SELECTION schema that leads to these new questions being asked. There are also prior conditions that have to be met before it is possible to instantiate the schemata; thus, presuppositional questions constitute the second kind of significant 208 issues surrounding schemata. "Presuppositional questions arise when instantiations of some accepted schema presuppose the truth of some controversial claim" (Kitcher 1993, 113). Here there is already a coherent instantiation, but some of its presuppositions are questionable. This, of course, generates research to resolve the apparent anomalies. Kitcher adds in a footnote that his presuppositional questions bear some resemblance both to Kuhnian notions and to Laudan's conceptual problems. Incidentally, it appears that for Kitcher, the questions that schemata are answers to (i.e., the questions heading the schemata) are not also to be counted as significant questions, nor therefore as questions amenable to erotetic progress. In summary, there are three types of intrinsically significant question, those whose answers would (1) instantiate an accepted schema (with the caveat that this occurs before there are a lot of paradigmatic cases of such instantiations), (2) hold out the promise of instantiating a schema in a problematic case, or (3) hold out the promise of helping to solve some problematic presupposition of the schema's being instantiated (Kitcher 1993, 114). There are also significant questions that are derivative from intrinsically significant questions: "the explanatory schemata of a practice generate significant primary questions, which spawn derivative questions: answering an appropriate sequence of the latter is seen as a way to address the former" (119). Instrumental significance is just significance that is derived via a chain leading back to a genuinely significant question in a different field. Erotetic progress occurs when "we have an erotetically well-grounded consensus practice in which we pose genuinely significant questions that were not previously asked" (Kitcher 1993,114). The significance of erotetic well-groundedness is just that the 209 questions that are being asked by the practitioners in the consensus practice are in fact significant relative to the explanatory schemata enshrined in the consensus practice: the practitioners do not only believe that the questions are significant; they are also correct to think so. There are at least four kinds of erotetic progress, two of which piggyback on the two kinds of progress already mentioned. When improvements are made to a referring term, thus constituting conceptual progress, these can sometimes be understood as leading to improved questions, thus constituting erotetic progress as well (114). Kitcher's vague example of this is, "Priestley's questions about the role of dephlogisticated air in various reactions are better formulated as questions about oxygen" (ibid.). Improvements in explanatory schemata generate new questions (Kuhn would call them new puzzles that need solving), particularly, as we have seen, questions regarding the presuppositions inherent in instantiated schemata. Indeed, it is intuitively obvious that a new or expanded schema would raise questions that were not previously at the forefront of the consensus practice. Erotetic progress is also made when scientists ask questions that are more amenable to being answered, in particular by substituting more precise questions for ones that are more vague (114-5). Finally, scientists make erotetic.progress when they break down existing (significant) questions into subquestions (115). Again, this is relative to an existing schema; for every premise in a schema, questions will arise as to how to recognize when something instantiates it. This will generate derivatively significant questions, questions that will need to be answered in order to recognize things of the kinds instantiating the variables 210 within the premises of the schema; whole subfields can be generated in attempts to answer these kinds of practical difficulties. 7.6 Other Kinds of Progress Experimental or instrumental progress67 is progress made when new instruments or experimental techniques allow scientists better to answer significant questions than their old instruments or techniques (Kitcher 1993, 117). He says also that this kind of progress "rests on the idea that the mixture of instruments and techniques now employed yields in each case a value that is at least as close to the actual value as that previously generated" with rare or insignificant exceptions to this generalization (124). Kitcher says rather little about this kind of progress, but he does have another tantalizing passage that refers to it tangentially: "The account of progress I am offering is thus keyed to investigating the fine structure of what [eg.] Fresnel and Brewster wrote and said (and what they did-for we can compare the instruments and experimental techniques they employed) . . ." (123). In the passage immediately preceding this quote, Kitcher explicates fine structure with respect to sets of accepted statements, and at a smaller level, the assessment of whether individual statements were significant and true (ibid.). Kitcher is explicitly sympathetic to verisimilitude, and makes use of it in his brief elucidation of this kind of progress: there is progress made in experimentation when experimental techniques and instruments allow scientists to acquire values that are closer to the true values (123). Then he seems to contradict himself by suggesting that "The claim 211 that experimental practice is progressive rests on the idea that the mixture of instruments and techniques now employed yields in each case a value that is at least as close to the actual value as that previously generated" (124). In the case in which the value generated by new instruments and experimental techniques is only as close to the actual value as that given by previous technique, it would seem that no progress has occurred. Presumably we are to understand this passage as assuming that there are a number of results given by the body of new experimental technique, and this body of technique is progressive over the previous one just in case it yields some results closer to the true ones, and in cases where they are no closer, they are at least as close as those yielded by the previous body of techniques. He adds that "if there are exceptions, that these are either rare, insignificant, or both" (124). "Making this conception of progress [i.e., instrumental-experimental progress] precise requires us to look more carefully at progress in the set of accepted statements" (Kitcher 1993,117). This latter kind of progress, i.e., with respect to the set of accepted statements, is made when scientists "[1] eliminate falsehood in favour of truth, [2] abandon the insignificant, [3] add significant truths,... [4] reconceptualize already accepted truths ... [and] [5] replace statements that are further from the truth with those that are closer to the truth" (117). From there Kitcher detours through elucidations of these five ways in which the statements accepted by a consensus practice can improve, including a whole section on verisimilitude, and ends up at the eventual definition of experimental-instrumental progress quoted in the last paragraph (i.e., 123). However, "progress with respect to the set of accepted statements" (117) is certainly an important category of progress in its own right, 212 and not merely a fleshing out of the background of instrumental and experimental progress. Therefore, I detail it here. The first way of making progress in the set of statements that scientists accept (i.e., [1] above) can be unpacked, according to Kitcher, as their "accepting statements that are both significant and true" (117). Let us recognize that "A significant statement is a potential answer to a significant question" (118). The first three sorts of progress in the set of accepted statements then bear a family resemblance to one another, as all are concerned with significance. Kitcher suggests that rather than exceptionless generalizations, science strives primarily for statements that are true answers to significant questions, and many of these will be generalizations with restrictions in scope; these true statements will turn out to be exceptionless generalizations with varying degrees of frequency depending on the field in question (118). While these first three subtypes of progress in accepted statements recall some of the other kinds of progress so far enumerated, they are importantly different. As compared to explanatory progress, they are improvements in statements rather than schemata. As compared to conceptual progress, they are improvements not solely in referring terms, but again in whole statements. Looking at it in this way, it would seem that progress in the set of statements could be seen as having a scope intermediate between conceptual and explanatory progress. Reconceptualizing already accepted truths, or "using improved language to reformulate antecedently enunciated significant truths" (120) (point [4] above), may sometimes overlap with conceptual progress, but again, the emphasis is on the statements accepted by scientists; statements encompass not only referring terms, but also relations between referring terms. Additionally, the third point above-adding 213 significant true statements-bears on a cumulativity of facts, to which I will return below. Here it is worth mentioning that this gets to count as progressive through its being considered a subtype of progress in the set of accepted statements. The last category of improvement in the accepted statements within a consensus practice is the replacement of statements that are not as close to the truth with ones that are closer. This is an explicit nod to verisimilitude, which Kitcher then goes into in more detail. He bases his notion of verisimilitude on Popper's: "In general, if there is a class A whose members have property B except under rare conditions C, D, ... , then the generalization 'All A's are B' will be relatively close to the truth (exceptions will be infrequent)" (121). Thereafter, there is movement closer to the truth when there is a restriction of the generalization excluding one of the exceptions while at the same time "adding a true generalization about the properties of those A's that are subject to the exceptional condition" (ibid.). Finally, there are two sorts of overarching progress Kitcher touches on briefly. He says "We make progress in our methodological principles by formulating strategies that give us greater chances of making conceptual progress, explanatory progress, or progress in the statements we accept" (Kitcher 1993, 120). It is not clear whether he believes that methodological progress would also apply to instrumental-experimental progress if scientists were to formulate strategies relevant to experimental improvement. It should be said that he mentions methodological progress only in passing, so although he does make room for it, it is not emphasized. 214 Kitcher raises the notion of organizational progress in the context of a discussion about the splitting and merging of fields. A reorganization of fields and subfields can be counted progressive so long as such a reorganization is to be "conceived as improvement of the accepted relations among the sciences" (124). Part of the consensus practice of any particular field of science concerns the "relationship of the phenomena ... [it studies] to phenomena studied by other fields" (124). The specification of these interrelationships between fields (where they are explicit) would also include a division of authoritative labour, here between fields (125). Kitcher allows that organizational progress can be involved in the three main types of progress that he details (125), but it is not clear whether he means it to apply to the other kinds as well. Although Kitcher also makes explicit reference to cumulative progress, it is not a category of progress in the sense that these others are, even though it applies to some of those categories. I turn now to an evaluation of how cumulativity comes into Kitcher's account. 7.7 Progress, Cumulativity and Significance There is implicit in Hacking (1983) a distinction between accumulation and subsumption. He refers to "heapings up of knowledge" (Hacking 1983, 55) and elsewhere to, what appears to be the same thing for him, accumulation of phenomena (56). He also recognizes accumulation of "Manipulative and technological skills" and "styles of scientific reasoning" (56). On the other hand, he refers to Nagelian subsumption in which "the new theory T 215 ought to be able to explain the phenomena that T explains, and it should also make whatever true predictions are made by T' (Hacking 1983, 67-8). I find this a valuable distinction between types of cumulativity-although there may be other distinctions that one could also draw. In one instance, Kitcher explicitly denies a commitment to an accumulative conception of progress: "We cannot think of evolutionary biology, even at the level of phenomena, as making progress through accumulation" (Kitcher 1993, 51). However, he seems to want to have it both ways, as later in the same paragraph he says I suggest we recognize that, while some descriptions of phenomena are revised or discarded, many become stable parts of consensus practice. The seeming growth in our understanding ... is ... partly captured by the presence in later practices of an increasing number of stable reports of phenomena, with the rate of increase greatly outstripping the rate of revision. (51) It seems that Kitcher is trying here to get a kind of accumulativity while claiming not to be in favour of it. He might be trying to make a very subtle point here about a non-trivial form of accumulation in that the "increasing number of stable reports of phenomena" is a general phenomena to be recognized, and constitutes growth (or perhaps progress), while we must be agnostic about the progressiveness of the accumulation of specific instances because they might later be jettisoned from consensus practice. Whether Kitcher is to be understood in 2 1 6 these passages as being against a progressive accumulativity, or as backhandedly endorsing it, he has other commitments that conflict with it. There is a tension between an accumulative understanding of cumulativity and Kitcher's account of significance. On the one hand, some of the other things he says seem to be evidence for a commitment to accumulativity. When Kitcher admits that the Darwinian case does not exhibit "simple accumulation," he nevertheless suggests that it demonstrates "a growing corpus of stable phenomena" (56) which appears to indicate a commitment to accumulation in the sense derived from Hacking. Also, as we have already seen, one of the subtypes of progress in terms of the set of statements accepted by scientists is "add[ing] significant truths" (117). Another of the ways in which progress could be made in the set of accepted statements is by "eliminat[ing] falsehood in favour of truth" (ibid). These considerations taken by themselves suggest that Kitcher might be amenable to progress being constituted by accumulation of true statements so long as they are significant as well. On the other hand is Kitcher's account of significance. As we have seen, Kitcher makes significance parasitic upon issues involved in the instantiation of schemata. Those issues were of two types, and the one that concerns us here is the first type: application issues or questions. Kitcher suggests that "In the early stages after a new schema has been introduced into consensus practice, almost all instantiations of it are important" (112), but "once a field has established a set of paradigm answers to application questions, further instantiations of its schemata are no longer on a par" and the questions that arise from these further instantiations are not significant except for "those that seem to involve special difficulties of producing instantiations" (113). Furthermore, some questions (such as the 217 latter) are intrinsically significant and some are only instrumentally significant (ibid.). Instrumentally significant questions are questions that derive their significance via a chain of derivative questions meant ultimately to help answer an intrinsically significant question (119)—in the case of application questions, by means of an instantiation (112). What all of this means is that when it comes to getting knowledge of new facts, more is not better unless the new fact is instrumental (via however long a chain) in answering an intrinsically significant fact. This is where Kitcher's account of significance creates a tension with an accumulative conception of progress: he is committed to accumulation's progressiveness only in those cases in which the accumulated statements describing phenomena are significant. However, a full account of progress in my view must include accumulation of facts not previously known, and not just those that Kitcher would define as significant. I will return to this criticism in Chapter VII; here it will suffice to lay the groundwork for it. The reason for Kitcher's commitment to the progressiveness of an accumulation of only significant truths is not hard to find. He agrees with previous philosophers that it is easy to generate new truths, but many of them are boring (94). Evidently he wants to rule out the accumulation of trivial truths as counting toward the progress of a scientific field. Here we encounter echoes of Lakatos. A n additional reason Kitcher does not want to have the accumulation of statements count as progress is that they may later drop out of consensus practice (51). At any rate, these commitments then rule out an accumulative sort 68 of cumulativity in cases in which the relevant newly-discovered phenomena are not significant according to Kitcher's definitions of significance. 218 It appears, furthermore, that there is no room in Kitcher's view for an accumulation of instantiations to count as progressive, although the textual evidence for this is mixed. In Kitcher's list of kinds of explanatory progress, his third is that "One schema is more complete than another just in case the former identifies a more inclusive set of relevant entities and properties ... "(111). This might initially seem to be evidence of a commitment to an accumulative conception regarding instantiations, but the more inclusive set he refers to here regards classes of instantiation types, not an accumulation of individual instantiating phenomena per se. In the context of evolutionary schemata, Kitcher refers to "a large class of statements reporting evolutionary phenomena"-some of which "are elevated to the status of paradigms for the theoretical study of evolution" and others of which act as examples in evolutionary subdisciplines of various evolutionary processes (50). Although Kitcher does not say so explicitly, these various phenomena must be understood as instantiations of schemata, since the examples he provides are the kinds of exemplars that do instantiate the various evolutionary schemata he earlier details. This is where he suggests that "We cannot think of evolutionary biology, even at the level of phenomena, as making progress through accumulation" (51). Here "the level of phenomena" can be understood as the level of instantiations given of the various schemata. So it appears that an accumulation of instantiations is not to be counted as progress, at least in evolutionary biology (although perhaps in other disciplines). Of course, for the realist, the instantiation itself is in the world and is discovered. However, instantiations are to be represented by statements. Since instantiations can be represented by statements, and he takes the accumulation of significant statements to count 2 1 9 as progress, one can ask whether Kitcher takes accumulation of instantiations to be progressive under no circumstances whatsoever, or whether accumulating significant instantiations will count as progressive for him. For my part, first I want to establish that instantiations can count for Kitcher as significant statements. He says, "significant statements ... [are] answers to significant questions" (Kitcher 1993, 118). Furthermore, as we have also seen, an instantiation of a schema is one of the kinds of answers to (intrinsically) significant questions (114). Accordingly, instantiations of schemata will count as significant statements, or perhaps more accurately, as conjunctions of significant statements. More simply, one can refer to "significant instantiations;" the considerations that bear on the accumulation of significant statements as being progressive should also apply to significant instantiations. So there is a partial contradiction: a growing corpus of instantiations is not to be counted as progress, but an accumulation of significant instantiations is. At least part of the reason that Kitcher does not want to have the accumulation of (at least some) statements or instantiations count as progress is that they may later be dropped from consensus practice (51). Of course, at any particular time, one cannot predict which statements will eventually be dropped from consensus practice, and this fact might be part of what is motivating his caution regarding any endorsement of an accumulation of statements or instantiations. Additionally, as quoted earlier, he claims that "A significant statement is a potential answer to a significant question" (118). If one can take Kitcher at his word, here, one assumes that a potential answer to even a significant question is even more likely to drop out of subsequent consensus practice. The thing to note is that although a 220 statement will count as an answer to a significant question only potentially, and is thus presumably more likely to drop out of consensus practice, it nevertheless gets to count as a significant statement. Little wonder, then, that Kitcher is cagey about admitting the accumulation of even significant statements into progress. At any rate, Kitcher's argument against (at least) evolutionary biology's "making progress through accumulation" in the phenomena it instantiates, is that statements of some of these phenomena will not be retained by future consensus practice, even if current consensus practice endorses them. If we can assume that this is a generalized rather than a localized worry for Kitcher, to be consistent he might have to abandon any commitment to accumulation of statements or instantiations as progress. In any case, he would have to abandon one of the two conflicting commitments, so either he would have to endorse a form of accumulativity that included statements, including (significant) instantiations, which might later be lost from consensus practice, or he would have to give up accumulativity's counting as progress. In summary, then, there is some question about whether an accumulation of significant instantiations will count for Kitcher as progress or not. Perhaps this observation is best expressed by identifying it as yet another tension between significance and an accumulative conception of cumulativity as progress. It is, however, incoherent to suggest that getting true answers to significant questions counts as progress while an accumulation of such answers does not. Therefore, in the following chapter, I will assume for the sake of argument that Kitcher would endorse the accumulation of significant answers as progress. 221 This accumulation includes significant instantiations, which are represented by conjunctions of significant statements. It is nevertheless important to detail the kind of cumulativity that Kitcher does accept, since cumulativity is an important concomitant of progress. He is committed to there being "something like cumulative progress" in his Darwinian case study, and he takes his explication of conceptual, and especially explanatory progress, to exhibit cumulativity (Kitcher 1993, 110). Kitcher also suggests that "tracing the details of the relations among the schemata employed at successive stages enables us to make precise the idea that there is a cumulative process of extending, correcting, and articulating a basic picture of the order of some natural phenomena" (109). Although he does not here say explicitly that this cumulativity is a form of progress, it is reasonable to assume that he holds this view, given both the surrounding context and his commitment to explanatory progress's exhibiting cumulativity. Furthermore, when he introduces the issue of what he calls "Kuhn loss," "that major shifts in science often involve the abandonment of explanatory insights that later developments in the field will reestablish," he suggests that even were Kuhn loss inevitable, the transition from one theory to the next would involve erotetic progress because it would (apparently necessarily) involve the replacement of less tractable with more tractable questions (116). So Kitcher is committed to cumulativity being manifest in (at least) his three main categories of progress. Kitcher later returns to the issue of "Kuhn loss" and argues against those who suggest that all scientific progress is accompanied by some kind of explanatory loss (173-7). In making this argument, Kitcher uses Kuhn's own example of a loss of explanatory power 222 in the theoretical shift between Aristotle and Newton. Assuming that this argument works, Kitcher still has not disproved explanatory loss in the general case. However, what is relevant to present purposes is that he hereby demonstrates a commitment to cumulativity and progress occurring across what Kuhn would call a paradigm shift. 223 Chapter VIII Evaluating Kitcher's Account In what follows, I first explain that Kitcher's category of conceptual progress is not relevant to the case study under consideration. Thereafter, there is an extended section on the schematization of Darwinian and neo-Darwinian theory. This section simultaneously deals with explanatory progress, which is parasitic upon schemata. There I show that the work of none of the scientists in the Galapagos finch competition controversy counts as explanatorily progressive. The discussion of schemata forms the backbone of the rest of the chapter, as schemata generate directly or indirectly the other kinds of epistemic progress that Kitcher allows. Next Lack is evaluated both in terms of the progress his work affords and in terms of progress in the statements that scientists accept. There I argue that Lack's work contributes toward making competition part of the consensus practice of evolutionary biologists, and advances true significant statements. Thus there is progress in accepted statements. The progressive contributions of the interim researchers-Bowman, Abbott et al. and the stochastic theorists-is then assessed. Bowman's, Abbott et al.'s and the stochastic theorists' contributions turn out not to count as significant statements, and so not progressive on Kitcher's account. The contributions of Schluter and Grant (1984) are then detailed, but Kitcher's machinery can accommodate only one small element of the progress inherent in that case. Finally, there is a section regarding Kitcher's inability to accommodate the cumulative progress evident in the finch competition case. 224 8.1 Conceptual Progress in the Galapagos Finch Competition Case Kitcher's first major component of progress is what he calls conceptual progress. None of the main elements of the current case study will end up counting as conceptual progress. Conceptual progress consists in scientists better obeying three maxims. They must refer to (1) what others refer to, (2) natural kinds, and (3) what can be specified (Kitcher 1993, 104). Ordinarily natural kind terms are nouns of a special sort that are intended to designate a natural grouping or class of things in the world. The examples that Kitcher details fit with this understanding of kinds: planet, oxygen and phlogiston (96, 97-100). However, Kitcher includes "homology" included among the "terms for which faulty modes of reference have been improved" (96). Homology is defined as "the condition of being homologous" (Hale et al. 1995, 324), which makes it a noun and its mention a referring term. Even so, it is not the kind of thing that one ordinarily thinks of as a kind, denoting as it does sets of traits of different species of organism which are similar due to common evolutionary origins (as opposed to being due to convergence caused by chance or byadaptation to similar environments). It seems therefore something of a relational property. So at least three options arise in interpreting Kitcher on this point. We can assume (1) that his considered opinion would be not to have "homology" among his examples of referring terms, or (2) that he considers "homology" to be a kind, or (3) that, since homology is not a kind, not only kind terms properly understood count toward conceptual progress, his frequent references to them notwithstanding. 225 I suspect that (1) is likeliest and (2) is unlikely. I would like to accept (3), as it would give Kitcher's conceptual progress more scope. However, attending to Kitcher's more schematized definition of conceptual progress reveals that kind terms are central. I therefore take it that on a strict interpretation of Kitcher's conceptual progress, it does not apply to the present case because the terms in the Galapagos finch competition controversy which undergo conceptual improvement are not classical kind terms. 8.2 Schematization and Explanatory Progress in the Finch Case In what follows I will assume that it is coherent to imagine that we can extract schemata from the work of scientists-schemata that they would recognize as schematizing their theoretical commitments. This point is debatable, but it will be valuable to assume it and to see how far it might take us in relation to the Galapagos finch competition controversy. I will sketch out in rough outline what extensions of Kitcher's evolution schemata would look like as applied to the present case. Indeed, extensions of schemata will be the only kind of explanatory progress (of the four kinds of explanatory progress Kitcher mentions) that the resource competition case exhibits. My position is that during the course of the history from Lack (1947) to Schluter and Grant (1984), evolutionary biology has made Kitcherian explanatory progress, but it is unclear whether it can be attributed to any of the players in the case study and, furthermore, the genuinely progressive contributions that they make cannot be captured by Kitcher's explanatory progress, even for the one kind of explanatory 226 progress that the whole of evolutionary biology has made during the same time frame as these scientists were working. 8.2.1 The Darwinian Schema Since I will be extending schemata that Kitcher himself provides, I will begin with what Kitcher says about these and their context. In the second chapter of his (1993) book, Kitcher lays out key points in the history of Darwinism and neo-Darwinism, thus providing "an extended illustrative example, on which subsequent discussion will be able to draw" (Kitcher 1993, 9). He calls the schema of neo-Darwinian selection, derived from the synthesis of population genetics and Darwinian natural selection in the 1940s, an "explanatory extension ... of Darwin's own selectionist patterns" (Kitcher 1993, 51). Furthermore, he claims that the SIMPLE INDIVIDUAL SELECTION schema can be embedded within the NEO-DARWINIAN SELECTION schema (46). The way that Kitcher claims that this embedding works is that "NEO-DARWINIAN SELECTION preserves the main structure of Darwin's selectionist patterns (while enlarging the scope to treat a broader family of questions) and simultaneously derives statements that had previously been taken as premises" (47). I will assume for now that the NEO-DARWINIAN SELECTION schema is an accurate representation of that subset of the consensus practice of modern evolutionary biologists, and I will use it as the basic background schema in what follows. Lack's (1947) book came after the neo-Darwinian evolutionary synthesis, so I will assume that we can count Kitcher's NEO-DARWINIAN SELECTION schema as part of Lack's 227 (Kitcherian) practice. Incidentally, with the exception of members of the subfield of population genetics, evolutionists do not bring population genetics into their analyses each time they analyze one of the phenomena they study. The neo-Darwinian schema that Kitcher gives is implicit but probably does not ever, in its entirety, enter the consciousness of most field evolutionists. Partly for this reason, and in order to assess Lack's contribution properly, we first need to refer to S I M P L E I N D I V I D U A L S E L E C T I O N . The schema is presented as an answer to the question, "Why do (virtually) all members of [group] G have [phenotypic trait] PT (28). Here are two of the seven original premises most relevant to present concerns: (2) Analysis of the ecological conditions and the physiological effects on their bearers ofP, Pi, ... , Pn Showing (3) Organisms with P had higher expected reproductive success than organisms with Pt(\>i> nf. (Kitcher 1993, 28) As alluded to before, S I M P L E I N D I V I D U A L S E L E C T I O N is embedded systematically within N E O - D A R W I N I A N S E L E C T I O N . Furthermore, premise (3) of the neo-Darwinian schema is essentially identical to premise (2) of S I M P L E I N D I V I D U A L S E L E C T I O N , as we shall see shortly. Although Kitcher has (2) only as detailed above, it is likely that it could be fleshed out further and still be consistent with Darwin's arguments in The Origin of Species. Of the S I M P L E I N D I V I D U A L S E L E C T I O N schema, Kitcher says, "I suggest that this elementary pattern 228 underlies the simplest illustrations that Darwin provides of natural selection" (28), so probably Kitcher is here committing himself only to the minimal schema both assumed by Darwin and accepted into the consensus practice of those who followed him, rather than suggesting that (2) could not be fleshed out further as a commitment of Darwin's. Nevertheless, Darwin did of course include adaptation to resources in his argument and, to a lesser extent as we saw in Chapter II, competition between organisms for resources as part of the environment to which species had to adapt (1859, quoted in Lack 1947, 115)— although this reference is to character displacement rather than competitive exclusion. 8.2.2 Schematizing Darwin and Lack on Competition I propose, then, a minimal Darwinian ecological extension, viz.: 70 (2') An analysis of the ecological conditions , including (a) adaptation to resources, or (b) competition between organisms utilizing the same resources, or (c) other ecological factors, and the effects of these ecological conditions on their bearers of Pi,... , Pn. Adaptation is rightly understood as part of the ecological conditions. Notice that although fitness, i.e., "higher expected reproductive success," occurs in (3), it is not sufficient by itself to account for adaptation. Adaptation is fitness in response to and in comparison with 229 environment; the degree to which the organisms' phenotype is compatible with their environment is what makes for their differential survival and reproductive success. Accordingly, adaptation should be detailed in the ecological premise. Lack is committed to (2'), in addition to the operation of genetic drift, although as we have seen, in 1947 he proposes the influence of the latter only in three finch species in the archipelago (Lack 1947,123-124), those whose populations were small enough to allow genetic drift to occur. Drift, however, would be subsumed by another part of the schema and, indeed, cannot easily be represented before the neo-Darwinian synthesis gave us a schema that takes population genetics into account. In Lack's view, the major process responsible for the differentiation of the finches is (b), competition between congeners. He is also committed to (a), adaptation to resources, but he thinks that it is not instantiated by the Galapagos granivorous ground finches71 "While the beak differences between most of the subgenera of Darwin's finches are clearly adapted to differences in feeding methods, the same does not seem to hold for the beak differences between closely related species" such as Geospiza magnirostris, G.fortis, and G. fuliginosa (Lack 1947, 60-61). So Lack was committed in his practice to typically Darwinian ecological conditions as leading to differential reproductive success, plus genetic drift, a theory that was due initially to Sewall Wright in the 1930s (Futuyma 1998, 297). So far, Lack cannot be seen to have made progress over his predecessors as far as schemata are concerned, given that the schemata he would have accepted were already implicit in the work of other scientists. The subpremises of (2') can be understood as implicit in Darwin. So, in order to see Lack as having made Kitcherian explanatory 230 progress, one would have to demonstrate that Lack was committed to a more fine-grained ecological extension than (2')-which Darwin himself would have accepted-although the importance of (2')(b) was probably not accepted into the consensus practice of ecologists and theoretical evolutionists until after Lack. Let us pause to examine the nature of schemata, instantiations and extensions, since these will be the basis of some of my claims below. The first thing to note is that Kitcher is vague about how these differ and how we would delineate the boundaries between them. I argued in the last chapter that Kitcher's example of SIMPLE INDIVIDUAL SELECTION demonstrates (although Kitcher does not make it explicit) that there is a reciprocal relationship between instances and the schema that it is possible to abstract from them. He says that this schema, in relation to Darwin's work, "underlies the simplest illustrations Darwin provides of natural selection" (Kitcher 1993, 28). Further, Kitcher defines schemata as capturing "an objective order of dependency in nature" (Kitcher 1993, 106). We saw also that increased generality of a schema counted as one of the ways that it could progress: it "is correct for a more inclusive class of dependent phenomena" (111). For all of these reasons, we can understand it as being a kind of abstraction from nature. An explanatory extension occurs "when the picture of dependencies is embedded within some larger scheme" (Kitcher 1993,110). Again, these dependencies are the objective dependencies that we find in nature and which scientists attempt to represent abstractly and more generally by means of schemata. The extension of an original schema, although perhaps being more precise, is still, at least ideally, a replicable generalization that will apply to multiple cases in a lawlike manner. Kitcher seems to be attempting, in 231 principle, at least, to make a fairly concise distinction between an extension of a schema and an instantiation of one. One passage in which it is relatively clear what Kitcher means by instantiation is the following: within evolutionary biology the "explanatory enterprise" of "explaining the prevalent traits among groups of organisms (or more generally, accounting for distributions of traits)" counts as instantiating "some selectionist pattern" (Kitcher 1993, 48 footnote). Here we see that an instantiation is an instance from nature of a more general pattern or schema. I suggested above that the extension Lack can be understood as making to SIMPLE INDIVIDUAL SELECTION (by extending the ecological premise) is no more than Darwin himself would have admitted. A contribution to explanatory progress would be to introduce a new schema or a refined extension of one. What we see in Lack, however, is not a more refined extension of the ecological premise, but rather many observational considerations lending more credibility to the proposition that competition shaped the finch radiation. In Kitcher's terms, these taken together, would count as an instantiation. They would even seem to count as a paradigmatic instantiation (cf. Kitcher 1993, 50). Sulloway says that Lack's 1947 book "supplies abundant evidence for considering these birds ... as a classic paradigm of evolution and adaptive radiation in action" (Sulloway 1982, 45). More specifically, as we have seen in Chapter II, there was widespread agreement with Lack's emphasis on the interspecific competition hypothesis (Abbott et al., 1977,153). If Kitcher were to count it as progress when a scientist's evidence and arguments lead to an instantiation's being accepted into consensus practice, Lack's evidence and argumentation 232 would be progress indeed. However, there is no evidence that Kitcher does take this as progress, even in the case of paradigmatic instantiations. So far, I have considered Lack only in terms of his commitment to competition. Perhaps his allopatric model is amenable to characterization as Kitcherian explanatory progress. O f course, this is contingent upon its being true or approximately true in order for Kitcher to count it as progress. Abbott et al. assume that it is the best model of the speciation process on the islands (Abbott et al. 1977, 175). Grant, who refers to it as the Darwin-Stresemann-Lack model (Grant 1981, 654), suggests that Lack made important contributions to the allopatric model o f finch colonization of the Galapagos: "Stresemann's contribution in particular was eclipsed by Lack's far more comprehensive treatment" (ibid.). So we see that according to our best current science it is true or approximately so (Kitcher 1993, 90). A s we have seen in Chapter VII, the purpose of schemata is to capture the objective "dependencies of phenomena," and this is when explanatory progress is made (Kitcher 1993, 104-5). Furthermore, one of the examples Kitcher gives of dependencies is that "the characteristics of contemporary organisms are objectively dependent on the evolutionary histories of those organisms" (Kitcher 1993,106). So it might seem initially plausible that Kitcher would be amenable to the construction of a schema that answered the question, "How was the Galapagos Archipelago colonized by ancestral finches such that we see the distribution of species that we see today?" O f course, such a question is too specific, more like part of an instantiation than a schema. A more appropriate schematic question would 72 be, "How was archipelago X colonized by taxonomic grouping Y such that we see the 233 distribution of species that we see today?" As we have seen in Chapter II, Lack presents a four-step suggestion (i.e., the allopatric model) as to how the Galapagos was colonized by ancestral finches, followed by their adaptive radiation. These insights would be too specific and non-replicable to be of use in their greatest detail as a schema of the kind Kitcher endorses. In short, such a schema would cross the line into instantiation. The idea of schemata is to capture not just objective dependencies alone, but ones that are generalizable to other cases-here other cases of the evolutionary radiation of organisms. Alternatively, Lack's four-step allopatric model of finch speciation could be incorporated in some way as a disjunct into premise (3) of the NEO-DARWINIAN SELECTION schema, with variables where appropriate, but this too would likely cross the line from part of a schema to an instantiation because of the very specificity of it. Ernst Mayr provided evidence for allopatric speciation on the basis of geographic distribution in 1942 (Futuyma 1998, 484). Lack's application of it to the Galapagos finches would count for Kitcher at best as an instantiation of the schema that Mayr had already come up with (and perhaps as instantiating part of other schemata as well). In short, the advances that Lack had made in terms of his model of the Galapagos finch origins would not count as explanatory progress for Kitcher; they would count as an instantiation, or count toward an instantiation. As we have seen, Kitcher suggests that "once a field has established a set of paradigm answers to application questions, further instantiations of its schemata are no longer on a par" (113). Mayr had already presented evidence (instantiations) for a hypothetical allopatric speciation model (assuming for the moment that Kitcher would be amenable to such a model at all). Presumably, though, those instantiations have not been 234 incorporated into the consensus practice the way that Lack's example of "Darwin's finches" 73 has been , so we can conclude that Lack's is one of the paradigmatic instantiations of allopatric speciation and, presumably, adaptive radiation and competition as well. It is worth also noting that this appears to be so, even though competition, for example, had not been conclusively proved in the finches, as evidenced by subsequent challenges to the competition hypothesis by Bowman and the stochastic theorists. 8.2.3 The Neo-Darwinian Schema Although the part of the N E O - D A R W I N I A N S E L E C T I O N schema, as Kitcher gives it, which is most relevant to present concerns is premise (3), it is necessary to lay out here more of the schema to indicate something of the context of the overall schema and how (3) fits into it. The schema itself is presented as an answer to the question, "Why is the distribution of properties [i.e., phenotypic traits] P\,... , Pk in relative frequencies r\,... , rk (S n••= 1) found in group G ? " (46). (1) Among the ancestors of G there was a group of contemporaneous organisms, G o , such that (i) there was variation at n loci among members of G o ; (ii) at the /th locus there were m\ alternative alleles present in the G o - G sequence. (2) In the environment74 common to all organisms in the G o - G sequence, organisms with the allelic combination anau ... aniani have probability s\... \ j of having trait Pj, ... {specification of the gene-environment-phenotype relations for all 235 the allelic combinations}/3. (3) Analysis of the ecological conditions and the effects on their bearers of Pt, ..., Pk Showing (4) The fitness of the allelic combination anau ... aniani is wi... 1 [etc.] ... {continued through all allelic combinations} (Kitcher 1993, 46-7). The rest of the schema deals with the mathematical details of the allelic composition of the population as it transverses the generations, ending up at the present group G. Where fitness is subsequently mentioned, the reference is to the ascriptions of fitness which were already derived at premise (4). A few other notes about this schema may help to clarify this claim. Here Vs" represent single alleles, and "anau ... anian" represents the first combination of alleles. So in the case of the beaks of finches, that expression would represent one combination of alleles responsible for a particular beak depth. The w's stand in for the relevant or fitnesses that accrue to particular allelic combinations. Premise (2) basically provides a conversion "dictionary" between the allelic combinations and the phenotypes that are produced from them under various environmental conditions. Accordingly, it is possible to refer to phenotypes in premise (3), and they will be converted into allelic combinations by premise (2) for the purposes of premise (4). It is the job of premise (3) to assign fitness values on the basis of ecological conditions to the various phenotypes. Of course, more than one genotype 236 might result in the same phenotype, in which case different phenotypes would have the same fitness assigned to them. Notice that (3) is a "handwaving" premise (hence its italicization). One presumes that Kitcher presents it this way in order to await further insights from ecology, or to allow for different ecological conditions to apply in different cases. 8.2.4 Schematizing Resource Competition and Floristic Adaptation For present purposes, the most relevant contribution that Lack, Bowman, Abbott et al., Grant and Schluter have made to understanding the Galapagos ground finch radiation has been their researches into the relative importance of competition and adaptation to food resources (the floristic hypothesis) in that particular radiation. Incorporating that contribution into Kitcher's NEO-DARWINIAN SELECTION schema gives us the following as a revised premise (3). The following is a minimal ecological extension of (3) that would incorporate the work of Lack, Bowman, Abbott et al., Grant and Schluter: (3') An analysis of the ecological conditions, including 76 77 (a) adaptation to resources, or (b) competition between organisms utilizing the same resources, or (c) other ecological factors, and the effects of these ecological conditions on their bearers of Pi, ..., Pk. 237 Notice that (a) mentions only "resources" and not food resources as would be the commitment of those supporting Bowman's floristic hypothesis. This is done in order to keep the schema as general as possible while yet taking into account that adaptation to resources (here food) has been debated extensively in the Galapagos finch competition controversy. I do not wish to call into question the ability of Kitcherian schemata to capture the mathematics that would be inherent in a fully-detailed account of how the ecological condition has been extended in consensus practice. Since I am not challenging this aspect of Kitcher's account, it is not necessary to render these subpremises in full detail in what follows. Furthermore, none of my conclusions hinges on the exact rendering of these subpremises. The reader can take these to be stand-ins for the actual, more mathematically rigorized subpremises that would apply. (3") An analysis of the ecological conditions and the effects on their bearers of Pi, ... , Pk, is to include (a) adaptation to resources, such that bearers of some of Pi, ... , Pk will have varying levels of fitness (the associated value of w), or (b) competition between organisms utilizing the same resources, such that, in addition to (a), bearers of some of P\,..., Pk, where those are less than 78 some minimum value m compared to bearers of other P\, ... ,Pk, will have associated w's with lower values, or 238 (c) other ecological factors, and the effects of these ecological conditions on their bearers of Pi, ... , Pk. Here premise (a) represents the minimum adaptation to environment conditions. Premise (b) represents the competitive exclusion case in which the competitor is driven to extinction at that particular site. With reference to (a), eventually the bearers of phenotypes that have 79 lower fitness should decrease substantially, perhaps to extinction. This trend will be covered by the iterative part of the schema (not included above). There might be a more elegant version of the schematization of the Lack-Bowman-Abbott-Grant-Schluter ecological insights, but this one will serve for present purposes. In addition, Kitcher himself is amenable to different versions of schematizations of scientists' views (Kitcher 1993, 28f; 175). I claim that both the competition and floristic hypotheses are best understood within Kitcher's machinery as helping to refine and deepen Kitcher's N E O - D A R W I N I A N S E L E C T I O N schema by fleshing out (3), the ecological aspects. Therefore I will now review the evidence that Kitcher would view it as a progressive advance to revise the original schema by substitution of (3') or (3") for (3). Kitcher is not averse to disjunctive steps in his schemata, as evidenced by, for example, N E O - D A R W I N I A N S E L E C T I O N itself (47). As a result, these two main processes (competition and adaptation to food availability) will best be understood as part of a disjunctive account of that part of the schema that deals with ecological conditions. Furthermore, Kitcher considers schemata that exhibit explanatory extensions to demonstrate 239 progress over the schemata that they extend (Kitcher 1993, 110). He suggests, for instance, that neo-Darwinian selection has been refined and deepened by a better understanding of ecology (51), and this sort of deepening and refinement counts for him as a kind of "explanatory progress [which] involves the improvement of the schema itself (109). So we can understand this as a subtype of his explanatory progress. Thus, putting this sort of refinement into the terms of his explanatory schemata, Kitcher says that this is to be "understood as embedding NEO-DARWINIAN SELECTION within a yet more encompassing schema, one that uses insights of contemporary ecology to replace step (3) (the analysis of the ecological conditions) with general patterns of explanation in terms of costs and benefits from which differential fitnesses of different variants may be derived" (Kitcher 1993, 52). In (3') and (3") I am taking into account ecological discoveries that demonstrably bear on the evolution of a group of organisms, and expanding his premise (3) on the basis of them, so it would count for Kitcher as explanatory progress as much as the two examples he gives. I want briefly to mention here some of the other players in the Galapagos finch competition controversy, although I will return to them later in the context of other kinds of progress. The arguments of Bowman would act as defense of (3')(a) and (3")(a). Recall that not all of the stochastic arguments were arguments only against competition, but those are the ones we will be addressing here. Similar results apply here as applied in the case of Lack. That is, in terms of the schemata accepted by consensus practice, it would remain unchanged by their work, even though they made what would seem to be progressive and innovative arguments for the importance of adaptation to floristic resources and random colonization of the islands. Further, Strong et al. were arguing against (b)'s applying in a 240 particular instantiation, whereas this was not to deny that it would be operative in other instantiations of neo-Darwinian selection: they admit that their results do not rule out competition, but only show that it is less common in nature than usually assumed (Strong et al. 1979, 910). Again, all of these considerations are aspects of scientific progress that changes to the schema are blind to. To incorporate their position of the denial of (b), one would have to make a hybrid of the NEO-DARWINIAN SELECTION schema and an instantiation, something that Kitcher's machinery does not allow. I return below to the question of whether the insights of the stochastic modelers are to be captured by other aspects of Kitcher's progress. 8.2.5 Schematizing Optimality and Adaptive Radiation It is likely that NEO-DARWINIAN SELECTION with the addition of (3') or something like (3") would be accepted by all evolutionary biologists working today, but perhaps not at the time of Lack (1947); as suggested earlier, competition probably did not enter consensus practice until after Lack's work, and Lack's work was instrumental in bringing (3")(b) into consensus practice. Lack himself would likely have accepted (3") as well. So far, then, there has been little in the way of explanatory progress in the present case study, except just out of view of the case study I detailed here, i.e., in the discipline of evolutionary biology more generally. That is, the improvements in the schemata of evolutionary biology are not to be found yet in any of our scientists' work, although their acceptance into consensus practice had much to do, I claim, with the evidence and arguments adduced by them. Nevertheless, the episodes 241 from Lack through Schluter and Grant (1984) do admit of Kitcherian progress of other kinds, as detailed below. Here is another candidate extension of Kitcher's NEO-DARWINIAN SELECTION schema, this one taking optimality and adaptive landscapes into account: (3"') An analysis of the ecological conditions and the effects on their bearers of Pi, ... , Pkis to include: 80 (a) that there is an optimal phenotypic adaptation 0(P\>0> Pk) to a resource, such that P\,... ,Pk which are closer to the optimum will give 81 their bearers correspondingly higher associated w's relative to O (b) in addition to (a), there is an optimal adaptation O' (Pi > O' > Pk) to a different resource such that bearers of Pi, ... , Pk, other than those in (a), which are closer to the optimum O' will give their bearers correspondingly higher associated w's relative to 0'\ furthermore, O lies some minimum distance m,(m>0) away from O', or (c) other ecological factors, and the effects of these ecological conditions on their bearers of Pi, ... , Pk. Here the minimum distance m is intended to capture the notion of the gap caused by the adaptive valley between peaks. In Schluter and Grant (1984), the expected population densities result would be germane to (3"')(a), and the five-way test would be more relevant to (b). Once again, as we saw with Lack, the arguments given in that paper are not best 242 captured by the machinery of schemata, and the explanatory progress that is parasitic upon it. Schluter would accept something like this extension. However, just as we saw with the extensions that Lack would have accepted, Schluter is not the first to accept this schema. It is difficult to pinpoint exactly when something like (3"') would have entered consensus practice, but it would have occurred sometime after Simpson's (1944) discussion of metaphorical adaptive landscapes. Given that the metaphor of adaptive landscapes is so prevalent in evolutionary heuristic explanation, it would appear that something like this schema would have entered consensus practice certainly before Schluter and Grant (1984). Without making an in-depth historical survey of its use as a heuristic, it would be impossible to narrow the time frame very far. This might be considered a criticism one could make of Kitcher, but putting an exact date on discoveries and when they become commonly accepted by the scientific community is a ubiquitous problem in the history and philosophy of science. What should be clear about (3"') however, is that it is an improvement over (3"), even though this is not capturable by Kitcher's explanatory progress. To review, Kitcher has four kinds of explanatory progress, and the one that has been at issue in the foregoing discussion has been explanatory extension. It is the fourth kind that Kitcher mentions, "when the picture of dependencies is embedded within some larger scheme" (Kitcher 1993, 110). In principle, some of the other kinds of explanatory progress could apply simultaneously. However, in this case none of them apply either. They were the introduction of correct schemata, the elimination of incorrect schemata, and their generalization such that 243 they could "deal... with a broader class of instances" (109-110). Both (3") and (3"') are presumably correct, and each would deal with the same class of instances. It appears that Kitcher's kinds of improvements in schemata will not be fine-grained enough to accommodate progress in reconceptualizations of existing schemata. Perhaps this is not surprising given that explanatory progress for Kitcher is intended to capture the increase in scientists' understanding of the dependencies of phenomena. Whether one describes adaptation in terms of individual traits or in terms of those traits approaching optima, this increase in theoretical elegance has no bearing on dependencies in nature. 8.3 Lack's Contribution to Progress in Accepted Statements As we saw in the previous sections, there were some candidate kinds of progress that did not seem capturable by the sorts of progress Kitcher most emphasizes. I will address these in this section, and introduce other progressive aspects of the Galapagos finch competition controversy that are amenable to similar treatment. Lack's major contribution is best understood as helping to make competition part of the consensus practice of evolutionary biologists. In other words, one of his contributions was in giving evidence and argument in defense of a schema that might already have been understood as part of consensus practice in the time of Darwin. Sulloway notes that the so-called Darwin finches were not even mentioned in The Origin of Species (1982, 41). So the application of the competition hypothesis to the Galapagos finches did not occur until the work of Lack, although Stresemann (1936) mentions allopatric speciation in the finches 244 followed by a maintenance of species separation at secondary contact (Lack 1947, 132). So Lack did not only, as we have seen earlier, contribute to getting the Galapagos finches accepted as one of the paradigmatic instantiations of adaptive radiation and SIMPLE INDIVIDUAL SELECTION. He also suggested that (in effect) the finches constituted an instantiation of the competition extension of the two evolutionary schemata Kitcher details, gave convincing evidence for this and, moreover, was the first to suggest it. Although we have seen that in the case of the granivorous ground finches, he did not suggest adaptation to food resources, he gave much evidence and argument that adaptation to food resources had been operative in the evolution of some of the other finch species. He was, further, the first to suggest that genetic drift applied to some of the finches, although he only gestured toward this in 1947, rather than arguing for it. Let us review from Chapter VII the relationship between schemata, instantiations, significant questions and progress in accepted statements. Instantiations are phenomena that make or demonstrate schemata to be true or approximately true. Instantiations can be represented by conjunctions of statements. Where Kitcher would consider the statements that make up those instantiations to be significant, I use the term "significant instantiations." "[QJuestions are intrinsically significant when (a) answers to them would exhibit the possibility of instantiating an accepted schema" (114), with the caveat that where there are already paradigmatic instantiations of existing schemata in consensus practice, "[m]any questions to which an available schema could be directed are not regarded as significant" (113). Correspondingly, answers to these latter questions would not count as significant statements, since a significant statement is an answer to a significant question (118). What 245 all this means is that instantiations of schemata are to count as significant instantiations for Kitcher at least where there have not been too many paradigm instantiations of that same schema. Furthermore, one of the ways that progress is made in statements accepted by consensus practice is in the addition of significant true statements. Accordingly, it is to be counted as progress when we have significant instantiations, instantiations of existing schemata before there is a glut of such paradigmatic instantiations. Although we saw at the end of the last chapter that Kitcher's commitments on whether the accumulation of significant instantiations counts as progress seemed to conflict, I will here proceed for the time being on the assumption that an accumulation of significant statements and instantiations will count as progress for Kitcher. We can count Lack's argument that the finch fauna has been shaped by competitive forces as leading to progress in accepted statements, specifically conjunctions of statements representing instantiations. As we have seen in the last chapter, paradigmatic instantiations are significant, and progress in accepted statements is made by means of their acceptance into consensus practice. The significance of the "leading to" clause above is that even though we can assume that the Galapagos finches entered the consensus practice of evolutionary biologists, this is something that the field as a whole did, not something that Lack did. Lack strengthened both the schema and instantiation by means of his evidence and arguments that this particular instantiation did fit the competition case, but this strengthening is a different thing than its being accepted into consensus practice, even though the two are related. That is, progress was constituted by the acceptance of the finch 246 instantiation as a paradigmatic one, and evolutionary biologists did so largely, Sulloway argues, on the basis of Lack's work. One presumes that it is because Lack's evidence and arguments were persuasive that the field of evolutionary biology accepted this paradigmatic instantiation into consensus practice. The deficit in the account is that Lack's work itself does not get to count as progress in the acceptance of statements, at least. One might want to say that Lack himself made progress in obtaining the evidence he did and, in particular, in presenting that evidence as part of his argument that competition had shaped the finch fauna (instantiation)-but Kitcher's account does not give us this. In summary, we can say that Lack's contribution in terms of Kitcher's progress was in advancing true significant statements in themselves counting as localized progress which led to the acceptance of a significant instantiation into consensus practice, which counts as progress in accepted statements. Furthermore, he was the first to suggest that particular instantiation, a precursor to the progress that would be made when it was accepted into the consensus practice of the field. 8.4 Bowman's Contributions to Progress Bowman's work contributed to the acceptance of true statements about the finches, at least within a subdiscipline of those who were interested in that particular instantiation. Although his main argument was that competition was not needed to explain the radiation of the Galapagos finches, he brought to light new evidence, new true statements about the finches, in making his case. We can next ask whether these statements were also significant in 2 4 7 Kitcher's sense. Significant statements answer significant questions, and intrinsically significant questions all relate either to new instantiations of accepted schemata or to defending already-accepted instantiations. The new evidence that Bowman supplies does not, then, count among significant statements accepted even by a subdiscipline of evolutionary biologists because Bowman's evidence fulfils neither of these requirements. Presumably Bowman's evidential statements will not count as instrumentally significant either, since then they would have to be intrinsically significant to some other field, and it is difficult to imagine what field they would be significant to other than evolutionary biology and ecology. The final way that significance accrues to questions is derivatively, by standing in a sequence that ultimately answers an intrinsically significant primary question (Kitcher 1993, 119). Again, since the finches are already an accepted instantiation, questions that indirectly bear on directly significant questions surrounding that instantiation have already been answered. There is one small exception to the non-significance of the evidential statements and argumentative statements that Bowman gave regarding the finches. We have seen that Lack was responsible for the acceptance of the finch instantiation as paradigmatic, and this applies to both the adaptation and the competition aspects of the ecological premise of the revised evolutionary schemata. If we are to split hairs, however, we could separate the instantiation that Lack endorsed into two parts (although it is doubtful that consensus practice would have been this precise). Lack suggested specifically that adaptation to food resources did not apply to three sympatric Geospizae. Bowman's position was that it applied to all of them. So strictly, three finch species were not part of the instantiation with respect 248 to the adaptation disjunct of the ecological premise, that is, the instantiation of the finches which was already made by Lack. So Bowman's suggestion that the proper instantiation was to include these three finch species under the adaptation disjunct and not the competition disjunct would have counted as significant because it gave a new (partial) instantiation. However, if we take the conclusions of Schluter and Grant (1984) to be correct, Bowman's partial instantiation does not count as true. Therefore, once again, it does not count as progress for Kitcher. Of course, even if it is permissible to look at overlapping instantiations in this way, none of this is reflected in any changes to the ecological premise, which is an inclusive disjunction. What Bowman did was to make a lot more known about the finches, although none of his evidential statements will count on Kitcher's account as progress. He accumulated a lot of data, but Kitcher does not count an accumulation of non-significant statements to be progressive. I will have more to say below about this. Bowman may also have had a positive influence in the argumentative process surrounding the competition question in the finches. That is, his strong negative views on the matter may have had an important effect in generating scientific response. This would not count as a merely external influence, as it is internal to science itself, and part of the argumentative structure of science. 8.5 The Contributions of Abbott, Abbott and Grant Abbott et al. can be seen further to extend the collection of true statements about the finches. They present evidence and argument for competition, but even where those are true 249 statements, they do not count as significant in the way that Lack's contributions did because they shore up an instantiation that Lack had already brought into consensus practice. Their evidence and arguments demonstrating the importance of adaptation to food resources strengthened both the schemata by giving more arguments for the correctness of the finch instantiation, and the instantiation itself. Neither of these is captured by Kitcher's machinery. 8.6 Progress applied to the Stochastic Theorists Next I turn to those who argued against competition by suggesting that the distribution of the finches could be accounted for by stochastic effects. They did not gather evidence from the field as Bowman did, except to compare a few data points to their computer-generated results. Computer modeling might be understood either as a sort of experimental technique, or as a kind of argumentation. If it is thought of as argument, then we can analyze it in terms of whether those arguments constitute true significant statements. Presumably the statements themselves about the results of the modeling count as true. For the same reasons as we saw in the case of Bowman, however, they will not be significant. Let us look at the modeling as a kind of experiment or technique, then, and assess it in Kitcher's terms. Experimental progress is made when "the increased power of instruments and techniques ... deliver improved statements" (Kitcher 1993, 117). Furthermore, a mixture of techniques is admissible, and it is "possible that an instrument or technique could yield better results in some instances but worse in others" (123) and still 250 count as an improved technique or instrument. It seems from all this that the way to assess whether a technique is an improvement is to see whether it leads to significant true statements more often than not, and more often than predecessor techniques. It is too early to be able to assess this in the case of the stochastic models. It appears that, to the extent that their results show that competition need not be invoked in the Galapagos finch radiation (rather than just that it does not happen all that frequently), they are wrong. Because these specific techniques were designed with a very specific case in mind, it is likely that they will not be used enough times to make the relevant assessment. Of course, the broader technique of comparing to a null hypothesis, and even finding ways to model the null hypothesis and the target hypothesis by means of computer programs is not new to this case study. 8.7 Sch lu te r a n d G r a n t It is in relation to Schluter's work that most of the criticisms of Kitcher can be advanced. We saw earlier that the shift from (3") to (3"'), while it appeared to be a heuristically valuable, and hence progressive, reconceptualization, did not constitute explanatory progress for Kitcher. It would not count as conceptual progress either-even though initially it might seem to be an expansion of concepts-because it does not involve a change in reference potential of a kind term. It is a sort of change in the way scientists would describe the same extension, although not by means of something so simple as changing a single term. Kitcher begins his section on conceptual progress with a suggestion that progress is made when scientists "are able to provide more adequate specifications of [their] referents" 251 (95-6). One referent associated with (3"') that might initially seem to be a candidate for an increase in intension is "phenotypic trait," which is represented by all of Pi, ... , Pk-However, it is not phenotypic traits whose intension is being changed by the reconceptualization in terms of optima. There is another kind of progress Kitcher allows that would capture the reconceptualization demonstrated in moving from (3") to (3"'). It is one of the subtypes of progress in the statements scientists accept: namely, "reconceptualiz[ing] already accepted truths" (117). Furthermore, once this reconceptualization in terms of optimal values for adaptive phenotypic traits has taken place, it changes the kind of question that scientists can ask. They can then refer explicitly to optimals. This shift then makes Kitcher's erotetic progress possible in the current case, although it does not constitute erotetic progress in itself. There are, however, other aspects of the Schluter and Grant paper which are not so easily accommodated by Kitcher's categories of progress. Three main results of Schluter and Grant (1984) were the expected population densities graphs, the five-way test of ecological factors affecting the finches, and the additional ecological data gathered regarding the finches. For the same reason that the extensive data collected by Bowman and Abbott et al. did not count toward progress in the statements accepted by scientists, neither do the evidential statements of Schluter and Grant (1984). Furthermore, even if we were to take the evidence of Schluter and Grant as contributing to the reconceptualization of competition and adaptation, we have already seen that this reconceptualization did not count as explanatory progress, or progress with respect 252 to the relevant schema. As such, statements associated with it would not be counted as significant, since any questions to which they were answers would not count as significant. Therefore, this would not involve progress in terms of accepted statements. "Estimating adaptive landscapes from environments" (Schluter 2000, 111) was Schluter's most well-known result (Schluter 1999). Within Kitcher's purview it counts as evidence toward the Galapagos finch instantiation. It is relevant to the adaptation disjunct of our extended schematization (3"'), although as we have seen, the acceptance of even that reformulation as part of evolutionary consensus practice occurred prior to the Schluter and Grant (1984) paper. Futuyma suggests that "The thrust of this example is that in at least some instances, (1) the distributions of species are affected by those of competing species, and (2) the numbers and distributions of species in a locality are predictable from ecological information" (Futuyma 1998, 219). These, then, should be counted among the important elements of this result and, I would suggest, progressive aspects of the Galapagos finch competition controversy. One can emphasize here that these aspects of the importance of the expected densities result hinge upon their demonstration in Schluter and Grant (1984). However, it is difficult to see how Kitcher would accommodate these elements within his taxonomy of progress. First, Schluter and Grant's (1984) paper demonstrated that competition had shaped the evolution of the finches, i.e., (1). Thus their work could be seen as contributing to the acceptance by evolutionary biologists of some new statements regarding the finch instantiation, but they would not be significant because they did not introduce a new instantiation. Therefore, progress would not accrue to this aspect of the study via Kitcher's 25.3 progress in accepted statements, as with other components of the case study that we have so far examined. (2) bears on one of the most significant aspects of this case: the demonstration of the ability of evolutionary ecologists to predict the phenotypes of species given sufficient information. Schluter turned the metaphorical, heuristic tool of adaptive landscapes into a predictive tool. Kitcher emphasizes explanation, but not prediction, at least explicitly. Consensus practice improves and progresses by gathering more significant statements, and one tool to this end, although Kitcher does not seem to say so, may be reliable predictive models. The closest element in Kitcher's account to predictive models would seem to be what he refers to as techniques. As we have recently seen, they constitute experimental progress when they "deliver improved statements" (117). The graphic output of the expected population densities formula could, by a stretch of the imagination, be thought of as "statements" or at least as representable by statements. This underscores another deficiency in Kitcher, his statement-based analysis, where statements may not always be the best way to understand progressive and other scientific results. Nonetheless, if we are to translate the graphs into statements about expected population density of the ground finches, we see that some of predicted "statements" are "true." That is, the peaks in expected population density occur very close to the mean beak depths of the species actually present on those islands. Here it is not appropriate to make the criticism I adduced in the case of the stochastic theorists, viz., that we do not have enough cases to be able to decide whether it leads more frequently to significant true statements than predecessor techniques. First, there 254 are no predecessor techniques relevantly similar (which will be another aspect of the case that I will deal with at greater length below). Second, the comparison of these graphs to the actual finch species present on the islands yielded true "statements" if it is permissible to translate the graphs into verbal descriptions. However, the question, as ever, is whether they are significant as well. I would suggest not, for similar reasons as mentioned earlier. Putting data into a formula that takes all the main relevant ecological quantities and interactions, and that then mathematically gives a predictive output does not seem appropriately called experimental progress. It appears therefore that Kitcher had something very different in mind when he designated the category of experimental progress. I conclude then that although it is possible to stretch Kitcher's notion of experimental progress so that it encompasses this example, that it is not really appropriately applied to it. At any rate, I would argue that the importance of this predictive formula transcends just the production of new, true statements (whether significant or not). It is nothing short of a demonstration that evolutionary ecology obeys codifiable laws and is sufficiently precise so as to be able to generate accurate predictions when sufficient data is taken into account. Another way of stretching Kitcher's account may possibly address these values. The predictive formula might count for Kitcher as an instantiation of a broader scientific schema that somehow schematizes the notion that in order to count as scientific, a science ought to be able to predict phenomena of certain kinds in the domain with which it is concerned. Whether this sort of commitment would be part of the consensus practice of all scientists, and whether Kitcher would be amenable to such broad science-wide schemata are open 255 questions. Certainly evolutionary biologists have often argued against the relevance of this sort of commitment to what they study. Wright's original metaphorical adaptive landscapes have undergone a couple of conceptual restructurings over the years, as detailed in Chapter II. I am not here concerned with the population biological interpretations, except as part of what the notion of adaptive landscapes has led to or refers to in the theoretical landscape. However, Schluter has taken this heuristic tool for describing adaptive radiations and given it teeth. He has done this in two ways: first, by a sort of mathematical reconceptualizing of what an adaptive landscape can be and, second, by using it to predict the depths of beaks that should be found on specific islands and the combinations of beak depths of the corresponding species there. We have already examined the second of these. Let us turn the first. What I am calling a reconceptualization of adaptive landscapes is not to be understood as the reconceptualization that Kitcher mentions as a subtype of progress in accepted statements. That was to "reconceptualize already accepted truths" (117). Here there is no accepted truth that is being reconceptualized by the reordering of mathematical relations between the variables of the formula that gives the expected population densities result. The resulting graphs track some truths that scientists already knew: the species of Geospiza present and their mean beak depths on fifteen islands. However, the formula itself is brand new and, furthermore, is based on Wright's metaphorical, pre-mathematicized notion of adaptive landscapes. I suggested in Section 8.1 that the relevant terms in the Galapagos finch controversy would not count as standard natural kind terms.Adaptive landscapes are not a standard sort of kind, and so not amenable to Kitcher's conceptual 256 progress. Also, the mathematization relates to schemata only in that it provides additional evidence and argument for the Tightness of adaptation's application to the finch instantiation. It may be a sort of instrumental technique as we have seen, but this way of looking at it does not apply to it qua comparisons to earlier notions of adaptive landscape. The most perspicuous thing to say about it would be that it is a conceptual advance or extension of older notions of adaptive landscapes by means of mathematical rigorization. Kitcher does not have the machinery to regard it this way, however. It would seem that this reconceptualization of adaptive landscapes, which I claim as progressive in the resource competition case, is not to be understood as progressive in Kitcher's terms. It remains to examine whether the computer modeling of expected densities might somehow be understood as erotetic progress. First, to count toward erotetic progress, questions have to be genuinely significant (Kitcher 1993, 114). I will assess this momentarily. To expand the notion of question, one might think of a computer model as a way of asking a question or a meta-question. In the first model (of the five-way test), for instance, they designed the model to output what the result would be if they assumed "all phenotypes are equally likely successfully to colonize or evolve where expected population density is positive; competition and variation in food supply play no role" (Schluter and Grant 1984, 185). The answer to this "question" was then compared to the actual species distribution present and found to be wanting. All of the models taken together could be thought of as a sort of meta-question asking which scenario best predicted the actual finch species distribution. Schluter and Grant suggest that evidence is often equivocal between the stochastic, floristic and competition hypotheses, and add that their five-way test is an 257 improvement over earlier forms of adjudicating between these (176). So they might be amenable to thinking of the five-way modeling project as a sort of meta-question which acts as a better way of asking the question of which forces were at work in the ground finch radiation. One kind of erotetic progress that might apply here then is "progress by posing more tractable questions" (Kitcher 1993,114). The other candidate for erotetic progress in the present case is the breaking down of existing questions into subquestions (115). If we take the over-arching question to be something like, "Which processes of the following have been most relevant in the finch radiation?" we can assess this for significance. The best candidate for the kind of significant question it could be is the second kind of intrinsic significance: when the answers to it "would exhibit the possibility of instantiating an accepted schema in apparently problematic instances" (Kitcher 1993, 114). I will take it as strictly false that an answer to our meta-question would count as instantiating an accepted schema in a problematic instance. It is true that the finch instantiation has been problematic in that the relative importance of two different ecological factors-adaptation to food resources and competition between congeners-has been debated throughout the Galapagos finch competition controversy. However, the instantiation itself is accepted as a paradigmatic one of selection, adaptive radiation and (by many) of competition's influence. That the instantiation applies to the NEO-DARWINIAN SELECTION schema is not in question. Furthermore, because of the disjunctive nature of the ecological premise, whether adaptation alone or adaptation with competition are proved to be operative in the present instantiation does not matter for the purposes of whether the finches instantiate the schema: there is sufficient leeway provided by the disjunction in either of (3") or (3"'). This may 258 constitute an argument against the wisdom of having disjunctive premises allowed in schemata, but Kitcher demonstrates that he accepts these (eg. Kitcher 1993,47). A further aspect of both the expected population densities mathematization and the five-way test is their innovativeness and ingenuity. Schluter came up with both while in the Galapagos, although he subsequently found that there had been a couple of papers that could be thought of as precursors (Schluter 2001). One of these, a mathematical coevolutionary model to be found in Roughgarden (1976) was, in any case, too abstract to apply to the ground finches, so in response Schluter "invent[ed] some simple rules for competition and implement[ed] them in a computer model" (ibid.). As we have seen, Schluter was not the first to use computer modeling in order to generate an ecological distribution, so in that sense, his contribution was not innovative. However, he took the model beyond what Roughgarden had done, and applied it to the finch instantiation by making the processes involved less abstract. There was another precursor as well (Case 1979), but that researcher did not deal as thoroughly with the data as Schluter and Grant did in their (1984) paper (Schluter 2001). At any rate, Schluter came up with the idea of using the available data to predict expected population densities in isolation from these others who were historically prior, but whom Schluter was not aware of until after his own insight. The closest Kitcher comes to capturing the progressive innovation within his framework is in intrinsic significance. He says "Intrinsic significance ... often accrues to those questions that seem hard to answer-questions that challenge the ingenuity of a scientist" (113). Ingenuity or innovativeness looks to be, for Kitcher, a marker of intrinsic significance, but it does not count, in itself, as progress. 259 The five-way test is, fiortherrnore, an advance in experimental technique or method. Let us see whether it counts as such according to Kitcher's definitions. Unlike the graphical results of expected population densities, here we have explicit statements as results of the modeling. To count as experimental-instrumental progress, a technique, I argued earlier, had to lead to more true significant statements than its predecessors. While these models' positive result supporting competition seems a particularly strong one here, and while this result is presumably true, it is not the first such support. Other techniques such as simple observation of a gap between beak depths in sympatric species have also suggested competition. The superiority of the current computer modeling technique does not seem to be a qualitative one in terms of more true statements, but perhaps a quantitative one which is to be assessed in some other way. Again, significance would be lacking in the resulting statements because they only demonstrate an already-accepted instantiation for an already-accepted schema. 8.8 Significance and Accumulativity As we have seen in Chapter VII, although Kitcher does make room for the accumulation of true significant statements as counting as progress, he does not stress it as one of the most significant kinds of progress. This may be because he finds accumulation of answers to significant questions such an obvious form of, and prerequisite to, progress, that he thinks his commitment to it should be obvious. By way of reminder, he does consider the accumulation of significant questions to be a form of erotetic progress (Kitcher 1993, 114). 260 The other possibility and a more likely one, as I have demonstrated in Chapter VII, is that he does not think of accumulativity as a very important kind of progress at all. We have seen, in this chapter, multiple cases in which the collection of additional important information about the nature of evolutionary ecology and the Galapagos finches did not count for Kitcher as progress because they were not significant according to his definition. What I would argue against Kitcher is that his view of significance leaves out the kind of progress that is an increase in information. What I want to suggest is that more knowledge and understanding about finches is valuable in its own right. Kitcher makes a good first stab at significance by offering the analysis he does-as I agree that there are truths that are not significant in the sense of contributions to science insofar as progress is concerned-but he has not gone all the way yet. There needs to be some sort of distinction between important and trivial truths, with the latter not counting toward progress, but I submit that Kitcher has left out too much that is progressive by having the requirements on significance that he does. I am tempted to say that statements concerning the 412th decimal place of pi do not constitute important truths unless, of course, for some reason it derives significance through a chain leading back to some explanatory schema. So what one can say is that being the answer to a significant question in Kitcher's sense is sufficient (but not necessary) for the importance of a truth, and hence the progressiveness that accrues to an accumulation of more of these truths about the world. More truths of a kind such as statements concerning the 412th decimal place of pi do not count as progress. More important truths do, so there is a very important problem here of how to delineate the boundary between the two. One can 261 say that boundaries are allowed to be fuzzy so long as there is some clarity when it comes to non-borderline cases. I want to have more finch data counting as progress, whereas Kitcher's account does not allow this. Critics might suggest here that I count too many statements as important and thus as progressive. For example, they might suggest that more finch data are just as uninteresting as the 412th decimal place of pi, and add that I am biased because I think animals and what they do are more intrinsically interesting than numbers. A principled answer to that would be valuable, but in the meantime, I would rather err on the side of inclusiveness of new knowledge statements. 262 Chapter IX Conclusions In this concluding chapter I compare Lakatos, Laudan and Kitcher in terms of the major abilities and inabilities of their accounts to accommodate progress in the Chapter II case study. I conclude that the adequacies and deficiencies of their accounts arise from their detailed, individual commitments. In the following section, I detail three main ways in which none of the three adequately accounts for the progress associated with the fine-grained structure of the Galapagos finch controversy. Thereafter I detail a criticism that might be adduced on the basis of the case's being consistent with Kuhnian normal science. Finally I summarize my conclusions and present the beginnings of a positive account of scientific progress. 9.1 Comparison of the Three Accounts In this section I detail something of a hodge podge of virtues and vices of the accounts under consideration: some of the other main criticisms that their application to the case has highlighted, as well as the virtues of these accounts in terms of the progress that they are able to accommodate. In the next section I will detail the kinds of progress that the three philosophers are unanimously unable to accommodate 263 One way in which there could be a mismatch between these accounts of progress and the progress evident in a given case study is that these accounts might isolate progress where other principles suggest to us that there is none. Even so, all of the epistemic progress suggested by any of the philosophers assayed here counts as real progress in my view. As a result, I now turn to the kinds of progress these accounts do and do not illuminate in our case study. 9.1.1 Laudan The application of Laudan's problem-solving methodology to our case study yielded the surprising result of suggesting that the elements of the case made little or no progress. Laudan accommodates something like progress in the solving of individual problems; he says that when a theory solves an "initial empirical problem, ... to that extent, we can say that 'progress' has been made" (Laudan 1977, 67). Certainly in this sense there was progress made in the scientific episodes studied, as every researcher or group isolated problems and gave solutions to them. However, because true progress for Laudan is an irreducibly comparative matter (Laudan 1977, 120), this preliminary gesture at a definition of progress does not count by itself as major, or significant, instance of epistemic progress. Laudan even allows that it is in theory possible to "construct something like a progressive ranking of all research traditions at a given time" but, even so, this constitutes only "an approximate determination of the effectiveness of a research tradition" (145). This approximation would be possible even in the case of assessing a single research tradition in isolation such as neo-264 Darwinianism, so it is possible in principle to make this approximation if one were to conduct the onerous procedure of adding up and assessing the weights of all of the micro-problems solved by the tradition. However, we see once again that in spite of his allowance for an approximate ranking of the progress internal to a research tradition, it is not possible for Laudan to assess true progress except comparatively. The one possible exception to the non-progressiveness of the Galapagos finch competition controversy in Laudan's terms is found in the work of Abbott, Abbott and Grant. As we saw in Chapter VI, before Abbott et al. it was not possible to demonstrate anomalies for either Bowman or Lack. A demonstration of progress in Laudan's account requires demonstrating that one theory or research tradition has greater problem-solving effectiveness than another (Laudan 1977, 68; 120). In turn, to determine problem-solving effectiveness requires "assessing the number and importance of the empirical problems which the theory solves and deducting therefrom the number and importance of the anomalies and conceptual problems which the theory generates" (68). Because Bowman's and Lack's work could be understood as assessing all of the same problems, and as solving them, on Laudan's view those solutions would not count as more numerous or more strongly weighted for either Lack's or Bowman's theory. As for conceptual problems presented by the two theorists' solutions, internal conceptual problems were seen not properly to apply to Lack's and Bowman's solutions because, although they could be understood to conflict, this is unavoidable given that they are part of a disjunctive account of the kinds of process that apply in these kinds of cases. As for external conceptual problems, it would be difficult to imagine theories from other domains that they conflict 265 with, other than ones that all of the neo-Darwinian tradition would conflict with (i.e., such as creationism). Conflicts arising from such large-scale inconsistencies between neo-Darwinism and other traditions would presumably count equally for Bowman and Lack. For comparative purposes, then, this state of affairs leaves us with only anomalies to assess; the theory with the smaller number and weight of anomalies would be the superior one and thus progressive. Although I did not assess the relative number and weights of anomalies presented by Abbott et al. for each of Bowman's and Lack's theories, this could, in principle, be done. However, I questioned the relevance of doing so in any case because Abbott et al. and those following them accepted that both hypotheses apply to the finch radiation in the Galapagos, so the two hypotheses are not rightly understood as competing. So we see that even in the one case in which Laudan's machinery allows us to view the extended scientific episode as in any way progressive, there are reasons to question the relevance of such progress determination. One of the recurring criticisms that arose in applying Laudan to these accounts was that he provides insufficient guidelines as to what will count as an adequate solution. There are examples in Chapter VI in which it might have been possible to assess the relative merits (and hence progressiveness) of differing solutions to problems if Laudan had taken a 82 more realist approach. He stresses that assessments of progress must be made comparatively between theories and research traditions, but never between theory and extra-theoretic reality. While it is true that comparisons with reality present some significant problems, this potentially leaves Laudan in the unenviable position of sometimes