UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The fountainhead of the river of time Van Wijk, Brendan 1996

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

Item Metadata


831-ubc_1996-0288.pdf [ 4.24MB ]
JSON: 831-1.0087143.json
JSON-LD: 831-1.0087143-ld.json
RDF/XML (Pretty): 831-1.0087143-rdf.xml
RDF/JSON: 831-1.0087143-rdf.json
Turtle: 831-1.0087143-turtle.txt
N-Triples: 831-1.0087143-rdf-ntriples.txt
Original Record: 831-1.0087143-source.json
Full Text

Full Text

THE FOUNTAINHEAD OF THE RIVER OF TIME by B R E N D A N V A N WUK B.Sc , Simon Fraser University, 1992 PBD, Simon Fraser University, 1994 A THESIS SUBMITTED IN F U L F I L L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R OF ARTS in THE F A C U L T Y OF G R A D U A T E 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 April 1996 © Brendan Van Wijk, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of F WAosft^Wj The University of British Columbia Vancouver, Canada Date Ayr Z3>/SC? DE-6 (2/88) ABSTRACT The brute sensation that time passes, or flows, is a compelling feature of our daily experience. We have a sense that there is a "moving now" that glides inexorably into the future, a sense so strong and basic that only a philosopher is likely to question whether what is given in our experience of time is best understood as anything akin to motion. However, questions of this kind are not without motivation. The prevailing philosophical and scientific conceptions of time are of structures in which the notion of objective "passage", or "becoming", has no obvious place. The phenomenology of time, and the metaphysics of time, are in apparent conflict. Why should time seem to flow, i f indeed static models of time are the closest approximations we have of the truth? Attempts to resolve this conflict in the literature to date have been few and preliminary. It is possible, however, to reconcile the apparent flow of time with an objective model in which the notion of passage is entirely neglected. Firstly, an examination of an idealized structure, the experience-at-a-time, leads to its characterization as a single, enduring set of temporally-related phenomena, whose varying locations are instants of time. The change of this set's position within our scheme of temporal representation as a function of time is perceived as motion, of sorts. Secondly, an examination of certain very basic factors that contribute to the long-term success of a population of biological organisms, as dictated by selection pressures within an environment in which certain key processes are de facto irreversible, provides an explanation of why it is that the "now" should be perceived to move into the future, and not into the past. Combining the results of these examinations ii yields an account of time's peculiar phenomenological character, in a world in which time is most conveniently treated as a static structure. iii TABLE OF CONTENTS Abstract i i Table of Contents iv List of Figures v 1. A First Pass 1 2. Person Stages in the World Block 8 3. The Structure of an Experience-at-a-Time 12 4. The Fork Principle 24 4.1 Recording Devices 35 4.2 Informational Content 42 4.3 Pre-Recording Devices 47 4.4 Knowledge Asymmetry 49 5. Evolutionary Biology from A to Z, and Z to A 53 5.1 Biological Imperatives 56 5.2 Nomic Imperatives 58 6. Ashley and Bradley 62 6.1 Sensation, Information and Action 68 7. Bringing It A l l Together 79 8. The Trouble with Causation 83 Bibliography 91 iv LIST OF FIGURES 2.1 A Schematic Rendering of the World Block 10 3.1 Three Successive Experiences-at-a-Time 15 4.1 Two Correlated Event Types 31 4.2 Bottom-Up Recording Devices 37 4.3 Top-Down Recording Devices 49 v • A First Pass Each of us is aware of a practically irresistible sensation that time passes, or flows. But while this sensation rules our daily experience, it's not so easy to express in words without resorting to metaphor. It may seem right, to some, to affirm that we and the world around us are forever in motion into the future. Others might prefer the claim that events parade by us into ever more distant reaches of the past. Still others would regard these two assertions as equivalent, and strive for a more discerning analysis of just what it is that we mean when we speak of "coming into being" and "passing away". Intimately connected with our perception of flowing time is the common belief that the past, present, and future are realms that are fundamentally unalike. The future feels to us to be a wide and mysterious realm pregnant with possibility, its final form waiting upon the decisions we make in the present, while the past is commonly regarded as over and done with, a closed case beyond our ability to influence. Appropriate to such a belief is the fact that our most prominent hopes and fears are oriented into the future, and seldom in the reverse direction. Wil l I have food and shelter in years to come? Wil l my dreams come true? Wil l my prospects improve by moving to a new city? These are questions that most any of us would be well motivated to answer. A bleak forecast of what the future holds can and does provoke considerable anxiety, regardless of whether one's life to date has been an uninterrupted parade of success and delight. Similarly, a cheerful forecast can be quite titillating, despite a singularly disagreeable past. However, analogous inquiries directed into the past are usually not nearly so pressing, except insofar as their answers inform the decisions we must make in the future. Did I have food 1 and shelter in bygone years? Did my dreams come true? Were my prospects improved by moving to a new city? Even i f the answers to these time-reversed questions should all turn out to be resoundingly negative, one would not expect to be especially anxious about this fact, so long as one's present circumstances are conspicuously rosy, and one's prospects for the future are expected to be too. Why should the mere fact of time reversal result in such drastic fluctuations in our value judgments? One may suspect that a fully satisfactory treatment of this and affiliated questions would consist in numerous interrelated elements. These elements would likely include (1) a clear statement about the nature of time itself, plus an account of (2) the dependence of our epistemological powers and psychological biases upon the de facto law-like conditions prevailing in our part of the universe, and (3) the dependence of our concepts of causation and action upon our lopsided temporal attitudes. After all, as will become clear before all is done, the future-bias of our psychology has a fundamental impact upon many of our deepest beliefs and most familiar concepts. But be that as it may, all such esoteric considerations aside, it seems plain on an artless intuitive level that the future is so much more important than the past simply because the past isn't the direction in which time goes. In this paper I ' l l be concerned with resolving the putative conflict between the phenomenology of time on the one hand, and the metaphysics of time on the other. On the one hand, the brute sensation that time passes, or flows, is an irresistible and omnipresent feature of our experience. But on the other, the prevailing philosophical and scientific conceptions of time are of structures that not only perform their theoretical roles without requiring some notion of objectively lapsing time to be built into them, but also resist the straightforward inclusion of such 2 features. These are the so-called static models of time, in which all events at all places and times are equally real. Space and time are united into space-time, and the universe is modeled as a block of sorts (with suitably numerous dimensions), in which the histories of all of the universe's denizens from beginning to end are strung out along the temporal axis like the strands of a petrified web.1 Attempts to resolve the conflict between static models and the phenomenology of the "moving now" have been, to date, few and preliminary in the literature. I therefore propose in this paper to reconcile the apparent flow of time with an objective model in which the notion of flux is entirely neglected. I avail myself of the following principal resources: (1) the structure of experiences-at-a-time, as elucidated by W. J. Friedman (1990), Izchak Miller (1984), and Paul Horwich (1987), (2) the de facto irreversibility of numerous processes in our part of the universe, explicable as a prevalence of so-called fork asymmetries (Horwich, 1987), and (3) an examination of the manner in which the imperatives of evolutionary biology are most likely to be enforced among the denizens of a fork-asymmetric world. A study of certain very basic factors that contribute to the long-term success of a population of biological entities, as dictated by selection pressures unique to an environment in which certain key processes are de facto irreversible, will bear directly on the issue at hand. The result will be an account of time's peculiar phenomenological character, in a world in which time is most conveniently treated as a static structure. 1 Space-time models differ from one another with respect to the amount of structure they support. The model I will be using to illustrate my points will lack such relativistic refinements as curvature and semi-Euclidean geometry, but will come furnished with a full slate of metrical structure, such as simultaneity classes and a privileged class of parallel lines. 3 In the course of my investigation, I will simply take as given much that is controversial about the nature of time (and related concepts). Some of my more important assumptions are these: (1) I assume that the dimension of time, like the dimensions of space, is isotropic. In space, the result of an isolated experiment is independent of the direction in which the apparatus happens to be facing at the time. No one direction is intrinsically different from any other. Similarly, with respect to time, i f there is any difference between the past and future directions, it's not in virtue of any deep metaphysical distinction between them. Models in which future time is represented as branched2 in order to avoid the odious (to some) result that all future events are "settled" are omitted from consideration. The past differs from the future only in virtue of the fact that many patterns of events in our world happen to run from past to future far more often than they do from future to past, and vice versa. One may attempt to explain such circumstantial asymmetry without appeal to temporal anisotropy. (2) I assume that time is a static continuum, in which an objective notion of "lapsing" time need never make an appearance. Successive instants, and the various events located at them, all share precisely the same ontological status. There is no restless metaphysical property of "presentness" that ascends through the strata of simultaneity classes, lighting up the contents of successive times. The so-called A -theorists of time often maintain that presentness is a monadic, metaphysical property that picks out successive classes of 2 Such as in A. Prior's (1967) formalism, the idea of which is that the truth of a proposition is relative to a moment and a history. To the extent that indeterminism is true, Prior proposes, the sum of all past and present events does not "settle" (deterministically) all future events. Therefore the truth of all statements about the future is not settled at the present moment. However, in the static block model I will be 4 simultaneous events3, and that pastness and futurity are properties that admit of degrees, distinguishing metaphysically classes of events that are earlier or later, respectively, than the class that is present. Against the cogency of this view McTaggart (1908) has offered an influential argument (despite his ultimate failure to secure the notorious conclusion he sought). Throughout this paper, I ' l l confirm my allegiance to the 5-theorists' view, and maintain that pastness, presentness and futurity are not monadic properties but relations, holding eternally between events spread out in space and time. The tenseless relation "simultaneous with", and the set of relations of the form "earlier than" and "later than", suffice to fix the temporal properties of any concrete object. (3) I take it that the relation between causes and effects may, with equal legitimacy (if not facility), be viewed as running either from earlier to later times, or from later to earlier times. I maintain that the direction in which causation runs is, at root, a matter of convention. Accordingly, I try to minimize the use of those sorts of terms that emphasize the "power" of causes to "give rise to" their effects, or those that suggest that a cause is somehow "more ontologically basic" than its effect. Terms such as "produce", "compel", "make", or even "cause" itself are lingua non grata, except when I feel that the cost of their avoidance to clarity and brevity would be too steep. Instead, I speak as much as possible in terms of readily time-reversible statistical correlations and deterministic relations between event-types. The upshot of this assumption that is most important to considering, the settledness of future events is merely a statement of logical fact independent of the truth of determinism. 3 As Howard Stein argues in his (1968), it is a mistake to assign temporal relations (such as simultaneity) to events that are space-like separated in a special relativistic spacetime, and then imbue the relata with a metaphysical property (such as presentness). On Stein's account, only a single point event located at the conjunction of the vertices of past and future light cones may possess the property of presentness, not 5 our purposes here is that we can't distinguish between past and future directions, and our attitudes regarding them, simply by invoking the concept of causation. (4) Finally, I take a person or self to be a temporally extended object that consists of an indefinite number of person stages (q.v. Lewis, 1986). My account will include no appeal to an enduring "sel f that is wholly present at successive times, directly apprehending the lapsing of time. If it is to be found anywhere, the illusion of lapsing time must be found in the structure of experience itself. While each of my assumptions is controversial to the last detail, they are such that I have left myself with the most frugal stock of conceptual resources with which to craft an account of time's illusory passage. The picture I will seek to draw is simple, yet (I trust) not misleadingly so. As my theory will be rendered with a view toward maximum compatibility with an austere scientific conception of the world, the reader will note that I treat many commonsense notions with a certain dismissive brevity. My illiberality in this respect is methodological. My object is to construct a theory of illusory time flow without helping myself to concepts that consist, in many cases, in ill-defined clusters of beliefs whose presence in a theory befogs more than it clarifies. I therefore help myself to as few conceptual resources as is feasible. I begin with the world block, in which time is represented as the vertical dimension, and space as the two horizontal dimensions. Physical phenomena are represented as static, temporally extended structures imbedded in the block - space-time "worms", i f you will . Biological entities, both individuals and whole populations, number among the simultaneity classes. Similarly, the legitimate domain of pastness and futurity is confined to the past and future light cones, respectively, of a "present" event. 6 objects more or less aligned along the temporal axis. For ease of reference, I ' l l refer to the future direction as we know it as upward, and to the past direction as downward. From now on, I ' l l be consistent in my use of the terms "up", "down", "top" and "bottom" to distinguish between the two temporal directions and to specify the relative positions of events located along time's axis. Looking at the block, then, there are two ways that a chronologically ordered story may be told about series of successive events therein: from the bottom up, or from the top down. We, as entities embedded in the block ourselves, come hard-wired (as I shall argue) with a predisposition to tell the sorts of stories that run from the bottom up. But as imaginary observers external to the block, scrutinizing the world sub specie aeternitatis (q.v. Huw Price's (1996) "view from nowhen", ch. 1), we must abandon all such biases. It's all too easy to allow our lopsided temporal attitudes to colour our attempts to account for the illusion of lapsing time, where they do explanatory work that they cannot legitimately be made to do, on pain of circular reasoning or begging the question. A fully satisfactory account of the illusion must have equal explanatory power whether rendered from a bottom-up, top-down, or orientation-independent perspective. 7 • Person Stages in the World Block A few more words about the model I ' l l be using during the course of coming discussions would be well worth saying. Purely for the sake of easy visualization, I ' l l pretend that the world consists in only two spatial dimensions and one temporal dimension, as mentioned in chapter 1. Thus, successive moments of the world's history are like flat sheets or disjoined frames of a motion picture stacked on top of each other, face-to-face, and fused into a continuous solid. Together, space and time make a static, three-dimensional block, the world block as I shall henceforth refer to it (see Figure 2.1 below). A n object extended in time describes a continuous path of stages that runs more or less parallel to the time axis. This path is commonly referred to as the object's world-line. Motion, in this model, is represented by the slope of a world-line. Motionless objects are characterized by vertical world-lines, while objects whose positions change as a function of time will have world-lines that slope away from the time axis at a proportionate angle. Change, in this model, consists in nothing more than this: i f the facts that describe one stage of an object located at one point in time are not the same as the facts that describe another stage of that same object located at another point in time, then the object may be said to have changed, from one time to the next.4 The world block itself, however, is not subject to change in any intelligible sense. It is important to emphasize that the asymmetry of the past and future directions is a function not of any intrinsic anisotropy in the temporal dimension itself, but rather only 4 Compare B. Russell's (1915) At-At theory, in which it is proposed that motion consists in nothing more than an object's being at certain positions at certain times such that the trajectory is continuous. 8 of the asymmetrical patterns of events within the block. As will be discussed in chapter 4, some investigators have proposed that our cosmic neighborhood is bounded, at bottom, by a condition of micro-chaos, the attendant condition of which is a prevalence of so-called fork asymmetries in the region above the boundary. Despite the temporal symmetry of physical laws, there is a prevalence, in our neck of the universe, of processes the time-reverses of which we never see. We never see glasses spontaneously assembling from shards on the floor and leaping up onto tables, and we never see swimming pools spewing people onto the tips of diving boards, settling into a state of tranquillity immediately thereafter. Vastly many processes both familiar and unfamiliar are de facto irreversible. Exploring the consequences of this fact about our neighborhood will provide crucial insight into the puzzle of illusory time flow. The objects of particular interest to me here are persons. A garden-variety person may be regarded as a succession of temporally contiguous person stages, a la Lewisian Stage Theory (Lewis, 1986). So, a whole person is a structure that stretches not merely from head to toe and fingertip to fingertip and front to back, but also from birth to death, and is present only in part at any one instant of time. Of course, just how one goes about carving up a person's world-line into discrete stages is a pretty arbitrary business, inasmuch as one stage passes smoothly into the next, leaving few if any conspicuous milestones to suggest a fitting location for a line. However, viewing a person as segmented in time is a useful approximation of continuity. One may make the approximation approach actuality about as closely as one may wish, by decreasing the temporal "thickness" of stages and thereby increasing their number. 9 Figure 2.1 A schematic rendering of the world block, in which a series of person stages stands roughly parallel, with the time axis. I take it for granted that all mental phenomena are supervenient on the physical operations of the brain. That is, there can be no variations in a person's mental content without there also being a variation in his physical (specifically, neurophysiological) make-up. Determining just which mental object maps onto which neurophysiological process, is a knotty point that is fortunately nonessential to the aim of this paper. However, just as I resolve persons into stages along the time axis, so too do I resolve their streams of experience. This simplification is at least superficially akin to dividing a continuously changing image into a series of still frames. In segmenting the stream of experience so, in treating it as i f it were a series of discrete, instantaneous frames of a motion picture, there is surely a great deal of nuance that is lost.5 Indeed, even the simple act of assigning simultaneity to the components of an experience-at-a-time is problematic. Dennett (1991) has argued that to exactly fix the time of occurrence of a conscious event is not possible even in principle, just as it is not possible to exactly fix its spatial position. 5 In fact, Dennett (1991, chapter 11) has argued that the stream of consciousness is actually riddled with gaps, but that we are oblivious to this fact for the reason that "an absence of information is not the same as 10 Nevertheless, the static frame model enables us to decant some cloudy bath water while holding on to the baby that is of immediate interest to us here. This move is an important preliminary to the rendering of an idealized structure, the experience-at-a-time, presented in the following chapter. As a final note, I will say that it's not uncommon among those who have not delved deeply into the philosophy of mind implicitly to treat the "se l f as i f it were a unitary spectator resident in a body like a coachman in a carriage, privy to certain events transpiring therein. The self seemingly monitors the influx of sensations that the body provides, as i f said sensations travel in from the body's sundry precincts and unite, in consciousness, into a single coherent picture.6 The illusion of the central spectator is deeply rooted, and its implicit presence can work to the detriment even of theories devised by those with considerable expertise in the philosophy of mind. However, within the purview of this investigation, it will be assumed that there is no unitary self that is wholly present at successive instants of time, and which may be thought to directly apprehend time's passage. Rather, I maintain simply that there are successive experiences-at-a-time that supervene upon successive person stages, and that there is no further spectator that observes the evolution of one "frame" into the next. The illusion of lapsing time must be found in the static structure of the experiential frames themselves, i f anywhere - not in the brute "passage" of one into the next. information about an absence". That is, we are aware of no gaps because there is nothing in them of which to be aware. 6 cf. Dennett's (1991) Cartesian Theater. 11 • The Structure of an Experience-at-a-Time A prominent and fundamental feature of one's stream of experience is its temporal structure. Some phenomena are represented as occurring presently, others as having occurred earlier, and still others as yet to occur. As discussed in chapter 2 of W. J. Friedman's (1990), the temporal structure of experience is maintained by means of several coactive mechanisms. I turn first to the faculty of memory. It is generally agreed that there is no biological date-stamping machine at work in the brain, and that memory is not simply sequential, the order of its contents being fixed by the order in which they first arrive. The three principal mechanisms that cooperate to establish time of occurrence are as follows. Firstly, the strength of a memory trace typically diminishes with time, such that trace strength serves as a crude guide to the order of occurrence. Secondly, much of the temporal structure of our memories is reconstructed by means of inference - our knowledge of natural and social time patterns provides frames on which to arrange an otherwise largely haphazard assemblage of information. And thirdly, a reminding mechanism is at play. New experiences often call to mind previous, related events, and the relative order of the two is stored as an additional piece of information. Anticipation is the future-oriented counterpart of memory. The fidelity and effective range of anticipation is, as a rule, far inferior to that of its past-oriented cousin. This disparity in fidelity contributes to the not uncommon attitude that anticipation is, in large part, a mere exercise in imagination informed by knowledge of past trends, and that it does not have bona fide future events as its content - at least not in the same way that memory has past events as content - since, as is widely held by non-philosophers (and 12 many philosophers), the future isn't settled. However, in the world block, all events are determinate7, and their pastness, presentness or futurity consists only in their being located below, level with, or above the time at which the assignment of one of the three temporal properties is made, respectively. Anticipation, then, will be treated here as a very limited faculty analogous to memory that has bona fide future events as content. This is not to deny that many events projected into the future are the products of the imagination, and not anticipation per se. One may vividly imagine some event as occurring in the future, and one may just as readily imagine some event as having occurred in the past. In a non-ideal system, it's not exactly clear where to draw the line between genuine recollection and anticipation on the one hand and creative embellishment on the other. Nevertheless, we need only concern ourselves with paradigm cases here. The mechanisms whereby the futurity of an event is represented are similar to those at work in the faculty of memory. In general, the most vividly anticipated events are those that are just around the comer, so to speak, and accuracy drops off very rapidly as a function of time. We anticipate the position of the bird wheeling in the sky a second from now far more vividly than we do its position several minutes hence. This time dependency may be the basis of a sorting function analogous to the strength mechanism that helps to order memory. As regards the remaining two mechanisms, only inference may be expected to play a significant role. The chronological order of anticipated events 71 should clarify that by "determinate", I mean only that a proposition that refers to an event located at any point in space-time has a determinate truth-value. I do not mean to suggest that the event in question is determined - i.e., required to occur by the laws of nature, given the state of the world at lower or higher times. An event may or may not be determined in this sense, but I need make no commitment to that effect yet. 13 is as likely to be shaped by inference as those that we remember, since we seldom have reason to expect that the natural and social patterns that have held in the past won't continue to hold in the future. Cross-referencing, however, may be expected to play a very minor role. The field of anticipated events is greatly impoverished compared to the cornucopia of memory, such that we are that much less likely to note salient similarities between anticipated events and log this as a separate piece of information. With the foregoing sketch of time-ordering strategies in mind, consider the following rendering of a highly idealized structure, the experience-at-a-time. Experiences-at-a-time are not unlike the concrete basic units chosen for Rudolph Carnap's particularistic system, as presented in his Der logische Aufbau der Welt (2nd ed., 1961, sections 67 and 109). These units, called elementarerlebnisse (or erlebs, for short), are momentary cross sections of the total stream of experience, together exhausting its full temporal length. Carnap's erlebs (and my experiences-at-a-time) are limited to a least perceivable segment of time, but are otherwise unlimited except by the bounds of immediate experience itself. Their selection as basic units or "ground elements" does not imply that they are actually separate units marked off in experience, but merely that assertions can be made about, and relating, such places in the stream of experience. Carnap preferred erlebs as the ground elements of his system because they seemed to be the closest practicable approximation to what is given - namely, a single (ostensibly) unbroken stream of experience. Yet while erlebs are minimal basic units, not further divisible into particles of which erlebs are themselves constructs, my experiences-at-a-time are divisible into sets of phenomena that are represented as occurring at various times. Thus, experiences-at-a-14 time have a temporal structure, and their basic constituent particles are sets of phenomena that are represented as standing in various temporal relations to one another. On Horwich's account (derived from Izchak Miller's (1984) elucidation of Husserl's Phenomenology of Internal Time Consciousness (1928)), the experience-at-a-time is a complex array of representations. The elements of the array, ... o, p, q..., are each sets of phenomena that are represented as occurring at a particular time: some now, some in the future, and some in the past. The diagram below schematizes a series of three experiences-at-a-time stacked one atop the other, like consecutive person stages embedded in the world block. q r s t u p q r s t - o — p — q — r — s - } experience at time t+l n o p q r m n o p q p q r s t o p q r s - n — o — p — q — r - } experience at time tO m n o p q 1 m n o p o p q r s n o p q r - m — n — o — p — q - } experience at time t-\ 1 m n o p k 1 m n o Fig. 3.1. Three successive experiences-at-a-time.8 The vertical dimension corresponds to actual time. The x axis corresponds to first order temporal representation, while the y axis corresponds to second order temporal representation. For example, at <l,0,rO>, I anticipate q will be present one unit from now, while at <-l,2,tf)>, I anticipate that two units from now, I'll remember q as having been present one unit ago. Let's turn our attention to the array located at time tO, and consider only its middle row, described by <x,0,tO>. The element <0,OyO> of the row, the set of phenomena p, is represented as occurring now. In a static space-time model, the term "now" is understood 15 to have indexical meaning. That is, its meaning changes depending on the circumstances of its utterance. In the world block, which comes equipped with simultaneity classes, the term "now" picks out all events that occur simultaneously with the utterance of the term. In general, as I gaze out at the world, one set of phenomena will always be represented as occurring "now" within any one experience-at-a-time, and the set that is so represented will usually vary from one time to the next. But my total experience at any given time t consists in more than just a blink of sensation isolated from the stream of all events past and future. A n experience-at-a-time is a complex of temporal representations, for it consists also in anticipations of future phenomena, and recollections of past phenomena. Consider the set q located at <l,0,rf)>, for example; q is represented as occurring one unit later than p. That is, at time tO, I anticipate the presentness of q one unit from now. Similarly, the set o at <-l,0,rf)> is represented as occurring one unit earlier than p. At time tO, then, I remember o's having been present one unit ago. And so on for the other phenomena, each of which is represented as occurring at times that are increasingly remote from p. This much, at least, is straightforward enough. But in general, the row <x,0,tO> isn't enough to tell the whole phenomenological story at AO. So far, we have that for the particular set of phenomena p, there may be times at which I anticipate its being present (such as at <1,0,M>), times at which I represent it as being present (such as at <0,0,/0>), and times at which I remember its having been present (such as at <-l,0,M-l>). But it may also be true of p that there are times at which I anticipate rememberingp (i.e. I anticipate 8 The notation I use bears coincidental resemblance to a notation devised by John Bigelow in his (1991) to capture the idea of transience. 16 that in the future, I will remember p), and remember anticipating p (i.e. I remember that in the past, I anticipated p). Or, more generally, "I anticipate that things I now remember will, at some later time, be remembered as having occurred even further in the past, that the things I now sense will be remembered, and that some things I now anticipate will then be expected to happen not so far in the future. And I remember that many of the things I now remember were once remembered as not having happened so far in the past, that the things I now sense were once anticipated, and that the things I now anticipate were then expected to happen even further in the future than they now are"9 So, the temporally-ordered string of phenomena <x,0,f> is itself projected both into the future and into the past, thus generating the remaining rows of the array at time t. Each row of the array for which y = +/- 1, 2, 3, ... is a representation of <x,0,f> projected to a different temporal perspective. Thus, in general, the anticipation, recollection, and present experience of each of ... o, p, q... are themselves the objects of anticipation, recollection, and present experience at the time /. Of course, the three-dimensional array illustrated above is an idealization. Very often, the phenomena that I anticipate will be present at some point t + n in the future will turn out to be quite unlike the phenomena that, at t + n, I actually represent as being present. We are, after all, routinely surprised during the course of daily life, and a susceptibility to surprise is incompatible with an infallible faculty of anticipation. Similar concerns pertain to the faculty of memory. Memory, while greatly superior to anticipation with respect to scope, fidelity and quantity of content, also suffers from distortions in numerous instances, and as a record it is incomplete. In the ideal case described above, sets of phenomena o, p, q, ... have been identified at a particular point in time, and then re-identified at other points in time. The 9 Paul Horwich, Asymmetries in Time (Cambridge, Mass.: The MIT Press, 1987), p. 35. (Hereafter cited as AT.) 17 only difference from one frame to the next is the fact that o, p, q, ... have been collectively shifted to new positions in the scheme of temporal representation. If this picture were a faithful rendering of real experiences-at-a-time, then we should all expect to have perfect recall, and perfect prescience. But we have no such powers. Clearly, this model breaks down if we insist on using only the same set of phenomenological letters (namely, o, p, q, ...) across all values of t. It also breaks down if we extend the array too far along the x and y axes. And yet, despite the obvious departures from ideality of actual experiences-at-a-time, Horwich's account distills an important structural feature. This feature may be revealed most clearly if we reformulate Horwich's account without the use of tensed language, and without the use of phenomenologically loaded terms such as "memory" and "anticipation". For our present purposes, it would be well to ignore the phenomenological asymmetry of memory and anticipation in order to highlight the functional symmetry that underlies them. The feature common to both memory and anticipation is the fact that they are representations of states of affairs that are projected to occur at times removed from "now". Consider, then, the following account derived from Horwich's, in which the terms "up" and "down" pick out the two possible directions in which a representation may be temporally projected. It is unnecessary, in this particular case, to specify whether "up" or "down" corresponds to what we know as "past" or "future". "I upwardly project that at some higher time, things I now downwardly project are then projected even farther down, things I now represent are downwardly projected, and some things I now upwardly project are then projected not so far up. And I downwardly project that at some lower time, things I now upwardly project are then projected even farther up, things I now represent are upwardly projected, and some things I now downwardly project are projected not so far down." 18 As shown, the past / future asymmetry that so dominates our everyday experience does not emerge from the underlying complex structure of experiences-at-a-time alone. The structure we have devised here is plainly time-symmetric. Neither, for that matter, does asymmetry emerge from a simple comparison of the content of upward projection versus downward projection (namely, the elements o, p, q, ...). As we move up or down the array from one experience-at-a-time to the next, the set p (say) simply shifts along the x axis in the appropriate direction. In the case of an ideal array, i f there is a time at which p is projected to occur one unit earlier, there is also a time at which p is projected to occur one unit later. The reasons behind our overpowering bottom-up bias will begin to emerge only after we've indulged in a little physics in chapter 4, and we've considered the nomological and de facto law-like factors responsible for the deep differences between memory and anticipation. Nevertheless, to the extent that our actual phenomenologies are faithful to the ideal structure proposed by Horwich, a crucial element of a fully satisfactory account has been brought to light. Adolph Griinbaum (2nd ed., 1973, chapter 10) foreshadowed this development when he proposed that because the word "now" functions somewhat like a noun, we are naturally disposed to regard it as standing for a single entity whose varying locations are instants of time. Indeed, it is a very large part of our everyday cognitive practice to identify objects at a given time, and then re-identify them at other (later) times, and keep track of the discrepancies, i f any, between their initial and subsequent positions. In any given experience at a time, an object O may be represented as being at position x presently, being at {x-2 one unit ago, x-4 two units ago, ...}, and being at {x+2 one unit from now, x+4 two units from now, ...}. Thus, at any one time, we remember where an 19 object has been, we know where it is, and we anticipate where it will be. Hence, we are capable of perceiving O's motion, a feat that would not be possible i f our experiences-at-a-time included only information about O's present whereabouts (where the "present" is just the time at which the experience in question occurs). But the direction of motion of any object is, by definition, the set of its locations ordered according to the times at which it possesses them. So, on Griinbaum's account, we come to think of now as an entity that glides tirelessly into the future. While Griinbaum's solution may not prick right to the core of the illusion of flowing time (insofar as it seems to neglect the raw feel of flowing time), we may nevertheless wring considerable mileage from his central idea. To see how this is so, consider once more the two-dimensional array of phenomenological elements, o, p, q, .... From one experience-at-a-time to the next, the positions of the elements relative to one another remain constant. However, the position of the array relative to the origin of "here and now" systematically shifts from one person stage to the next. That is, the same contents are successively observed from different temporal vantages. Much as in the case of the object O above, within any given experience-at-a-time there is a string of anticipations, present experiences and memories all represented as occurring presently. That is, I am presently experiencing, remembering and anticipating various phenomena. However, at the same time, I also remember that the same (or at least a similar) string has occurred at different temporal positions in the past, when some of what now seems past to me then seemed to be present or future. Similarly, I anticipate that the same or a similar string will occur at different temporal positions in the future, when some of what now seems future to me will seem present or past. Thus, I am aware of the string of phenomena as a 20 singular, enduring entity, whose properties are at least similar from one moment to the next, except only for the fact that it is offset relatively to my phenomenological origin. I perceive as motion the change of temporal perspective as a function of time. I am at liberty to view my sense of now as a point that is in upward motion relative to a fixed phenomenological array, or as a point that remains fixed relative to an array that is in downward motion. It remains, however, to explain why it is that "now" should seem to glide into the future, rather than into the past. As time "goes by", we are aware of an accumulation of knowledge as our fund of memories fattens. But why not decreasing knowledge? A satisfactory resolution of this problem is not trivial, and will be the focus of discussion for the remainder of this paper. Before moving on to the next chapter, I will sketch one possible solution to this problem that has been suggested by Horwich. He has proposed that the perception of a future-bound "moving now" may be explained as a combination of brute phenomenological asymmetry and linguistic convention. Firstly, relative to the direction in which the present is perceived to move, memory is intuitively regarded as "backward looking", and anticipation is "forward looking". Here, time is resolved into two apparently anisotropic poles in virtue of the phenomenological difference between memory and anticipation. The past is what we remember, and the future is what we anticipate. And secondly, a phenomenon p is perceived to be "earlier than" q i f p is always represented as having either less futurity or more pastness (i.e. i f p is always represented as occurring in the nearer future or farther in the past, respectively) than q. But, by convention, change is defined as occurring from an earlier stage of a process to a 21 later stage of that process. Horwich proposes to account for the direction in which the present seems to move simply by drawing our attention to this deeply rooted convention. It would seem, however, that Horwich's appeal to linguistic convention is unsatisfactory on two counts. Firstly, Horwich goes on to show in later chapters how the terms earlier and later may be reduced to "earlier relative to XY" and "earlier relative to YX\ such that it becomes possible to express the laws of nature without using a fundamental, asymmetric temporal relation {AT, 42-43). In a parallel fashion, the phenomenological sense of earlier and later invoked in his account of the moving now may be reduced to "earlier relative to anticipation-memory" (i.e., a scheme in which p is anticipated before it is remembered) and "earlier relative to memory-anticipation" (i.e., a scheme in which p is remembered before it is anticipated), respectively. It is unclear, on Horwich's account, what non-arbitrary reason we have to define change as occurring from a stage that is "earlier relative to anticipation-memory" to a stage that is "earlier relative to memory-anticipation", and not the other way around. Secondly, Horwich's hypothesis runs counter to our intuition that our perception of a moving now is, in an important sense, prior to linguistic convention. We imagine that the perception of a moving now would be similar for any organism equipped with a cognitive apparatus similar to ours, regardless of the linguistic structure imposed upon the raw phenomenology. Explaining the direction of time's apparent flow by appeal to linguistic convention seems to have the cart before the horse. We want to say that we adopt the convention that we do because of the direction in which time seems to flow, not the reverse. Placing convention at the base of our explanatory tower seems to suggest an objectionable aspect of arbitrariness about a basic and overpowering feature of our 22 phenomenology. If we decided to set up our conventions otherwise, would time then be perceived to flow in the opposite direction? One suspects that we would simply come to use different language to describe the same underlying phenomenological state of affairs. A fully satisfactory account of the direction of time's apparent flow must explain why the sense of "earlier" captured by "earlier relative to anticipation-memory" is to be non-arbitrarily preferred over the sense captured by "earlier relative to memory-anticipation". Furthermore, the account must do justice to our intuition that our perception of a moving now is prior to linguistic convention. Both of these requirements may be fulfilled, I believe, by providing an account drawn from evolutionary considerations of why it is that we care so much more about the future than the past.10 Thus, the arbitrariness that undermines an account that comes to rest on linguistic convention is averted. 1 0 In fact, Horwich himself anticipates the necessity of the strategy I have chosen, observing that "any. deeper explanation of the value asymmetry calls for rigorous work in evolutionary biology" (AT, 198). 23 4 . The Fork Principle As mentioned earlier, our cosmic neighborhood is characterized by a prevalence of de facto irreversible processes. Reading from the bottom up, we never see spherical light waves converging on a point source, nor hoary oaks dwindling to acorns, nor barium and krypton colliding amid an influx of energy to yield uranium 235. These processes and vastly many others are merely de facto irreversible (contrasted with nomologically irreversible), for there is nothing in the canons of natural law that prohibit such processes from running with equal frequency from the top down, instead of exclusively from the bottom up. Another way to capture this feature of our neck of the cosmos is by appeal to fork asymmetries. Whenever two event-types are correlated with one another, they are very likely to be imbedded in a V-shaped chain of nomological determination, but need not be imbedded in a A-shaped pattern. In other words, for any two correlated event types, there is very probably a common determining antecedent, but not necessarily any common determining, consequent." Consider thunder and lightning, two correlated events each of which is improbable. The probability of both events occurring together, P(T&L), is much higher than we would expect were we simply to calculate the product of their respective probabilities: P(T)-P(L). We tend to attribute the strong correlation between thunder and 1 1 For ease of reference, consistently with the convention that "down" corresponds to what we familiarly know as the past and "up" corresponds to the future, I will here establish the convention that an "antecedent" is an event that occurs at a position on the time axis that is' below the point of reference, while a "consequent" is an event that is located above. 24 lightning to the commonness of their determining antecedent (the "cause", in the usual parlance), but not to the commonness of their determining consequent (the "effect(s)"). On Horwich's account, the prevalence of bottom-up fork asymmetries is an attendant condition of the asymmetric boundary conditions of our cosmic neighborhood. At bottom, our neighborhood is bounded by a condition of micro-chaos, or so the story goes; yet at top, this condition does not obtain. In the case of an eternally isolated system, statistical mechanics assumes that every possible microstate will appear with the same frequency, such that the entropy of the system is just as likely to decrease as it is to increase. But let us suppose, as have Boltzman (1898) and Reichenbach (1956) and others, that our entire neighborhood has fluctuated into a state of low entropy. Let us suppose further that none of the branch systems in our neighborhood is perfectly shielded from random external perturbations ("cosmic input noise"). This random external input emerges, bottom-up, from the state of micro-chaos that bounds our neighborhood (and presumably the universe at large) at bottom. It so happens that an attendant condition of the bottom-up convergence of this "noise" upon any given branch system is an extremely low probability that the initial micro-state of the system will be the one required to lead to a decrease in entropy at a higher position in time. Therefore, the overwhelming majority of the processes that we actually see are those that lead to bottom-up entropy increases. The foregoing is a speculative model, far from entrenched in contemporary cosmological dogma. Horwich himself seems interested in carrying forward only the more general hypothesis that it is some feature of the initial conditions of the universe (plus the weak coupling of all systems to the rest of the universe) that determines the basic time-asymmetric phenomenon. Let us return to the fork principle, that is explained 25 in terms of this general hypothesis, and that will prove to be a central feature of my account of illusory time flow. Horwich's presentation of the principle is not without softnesses and obscurities. Given that I intend to ground my account of the psychological arrow of time in the fork principle (in some form or another), I will be at pains to fortify Horwich's presentation against criticism, or subject it to revision when necessary, in order to ensure that the fork principle will stand under the weight that I would place on it. Given correlated event types A and B, Horwich's fork principle maintains that there must be an event token of a third event type C below a given occurrence of A and B such that the following probabilistic relations are satisfied: Assumption 1: P(A) and P(B) are small. This is just to say that neither A nor B is an event type that often occurs. Assumption 2: P(A&B) » P(A)P(B). That is, there is a correlation between the occurrences of event types A and B. The occurrence of one is very often attended by the occurrence of the other. Assumption 3: P(A/Q » P(A), P(BIC) » P(B). It so happens that A occurs much more often when a C-event also occurs lower down, and so does B. Assumption 4: P(A&B/-Q = P(A/-QP(B/-Q. In the absence of a C-event lower down, occurrences of A and B are statistically independent of one another. That is, in the absence of C, we ought not to expect a correlation between occurrences of A and B. It would be well to look a little more closely at these crucial points. We have that there is a range of alternative event types C l , C2, ... , Cn, whose occurrence is often accompanied by the joint occurrence of A and B higher up. In contrast, there need be no 26 range of event types, C l * , C2* , . . . , Cn*, whose instances are almost always accompanied by the joint occurrence of A and B lower down. Rather, it may well be that in a substantial proportion of cases in which A and B co-occur there is no unified determining consequent, C*. Note that A and B may co-occur in the absence of its characteristic unified antecedent - but this is rare. Note also that each C-event is a unified determining antecedent (or, more simply, a 'unified cause') of A and B, in the sense that it cannot be split into two parts such that A is the determining consequent of one part and B is the determining consequent of the other. The unity of a C-event is a crucial feature, being what distinguishes it substantially from a C*-event. The potency of an objection to the fork principle that I will consider later in this section turns on whether a convincing account of what "unity" amounts to in this context can be provided. First, however, I will attempt to contend with a more immediate objection. Steven Savitt, in his (1991) critical notice of Horwich's (1987), argues that the fork principle is in at least one sense too strong, in that it seems to preclude the possibility of correlations lacking unified antecedents - that is, it doesn't make proper allowance for the occurrence of genuine coincidences. Horwich claims that "an uncaused correlation of A and B could occur only if their causal antecedents were correlated; and this would eventually entail a correlation among initial conditions, which is inconsistent with the hypothesis of initial microscopic chaos" (AT, 74, italics mine). But i f one allows that some correlations are genuinely accidental, lacking a unified causal ancestry, then one cannot also maintain that a necessary condition of correlation between event types is a correlation between causal antecedents. Consider any two event types, D and E, that are statistically independent of one another, such as the number of pips on the top faces of a pair of six-sided dice. We 27 should expect that i f each die is properly balanced, any one face should come out on top just as often as any other - at least in the long run. Furthermore, we expect that whether the first die comes up a snake eye (say) should have no bearing on whether the second die will too, except in very contrived circumstances (such as both being rolled with exacting precision by the same mechanical rolling device). Suppose, however, that it should turn out that for each of ten flips, both dice come up with snake eyes. By hypothesis, we suppose further that there is no unified antecedent to which we might point in order to explain this strong correlation. The synchronicity of the two dice is a genuine coincidence. An improbable turn of events, yes - about 1 in 1015, in fact - but not impossible. If the two dice are in fact effectively isolated from one another, statistical mechanics does indeed require that the synchronicity of their states should wash out - but only over the long run. There is no prohibition against "local synchronicity" between events lacking unified antecedents. As we continue rolling indefinitely many times, we expect that the states of the dice will eventually stand revealed as completely uncoupled. But some comparatively short strings of trials within the series at large may well exemplify a strong positive correlation, and others a strong negative correlation. Are any of Horwich's key assumptions incompatible with "locally clumpy" randomness (in other words, genuine coincidences)? The probability of a die coming up a snake eye on any trial is 1 in 6, satisfying (1). It turns out that each of the ten times we rolled the dice, they both came up snake eyes - a strong correlation that satisfies (2). We agreed that the existence of a properly functioning mechanical rolling device or the like would be attended by a higher frequency of snake-eyes higher up. This device is one of a range of hypothetical C-events 28 that may satisfy (3). Note that the absence of such a device in the case presently under consideration does not bear upon (3)'s general truth, for (3) states only that a C-event is a sufficient condition for a higher frequency of A and B, not a necessary one. Finally, assumption (4) states only that in the absence of a C-event lower down, event types A and B are statistically independent of one another. But as just earlier mentioned, statistical independence does not in itself preclude local correlation. It would seem that Horwich's unfortunate use of the phrase "only i f ought to be understood to express not a strictly necessary condition, but rather only a highly probable condition. Indeed, Horwich claims that genuine coincidences are only rare, not impossible. Yet as Savitt points out, it is legitimate to ask how rare, and why. On a guileless level, it seems evident that ten consecutive pairs of snake eyes is rare simply because the likelihood of its occurrence is about 1 in 1015, unless there's a unified determining antecedent involved (such as a mechanical rolling device). On a deeper level, however, it would be unsatisfactory simply to point to the fact that there are strong probabilistic connections between correlated events and unified antecedents. Fortunately, Horwich does more than simply point to the existence of such connections; he appeals to the hypothesis of initial micro-chaos and random cosmic input noise in order to account for their prevalence. The question now becomes, does Horwich succeed in establishing the connection between the fork principle and the postulated initial conditions of the universe? Some commentators have expressed doubts. As Arntzenius (1993) argues, i f the universe is deterministic, then if there is a characteristic event or state below the occurrence of A and B, there will also be a characteristic state above. This state, call it C*, bears the same 29 probabilistic relations to A and B as C does. Therefore, since each common cause is matched by a common effect (no matter which way the relation of causation is taken to run), there can be no fork asymmetry at all in a deterministic world. Horwich's response (in an unpublished A P A symposium) is that the "common effect", the state C*, would not be unified in the way that he intended. What, then, is the sense of "unified" that is needed to systematically distinguish C from C*? Consider once again the case of thunder and lightning, two highly correlated event-types having a characteristic event, C, below them. Lightning is not an altogether simple phenomenon; reading from the bottom up, an electron avalanche initiates a faintly luminous, comparatively slow "step leader" that staggers earthward, creating a channel of ionized air (plasma) that paves the way for a swift, powerful return stroke. A single flash of lightning commonly consists in several rapid strokes. Each subsequent return stroke is preceded by its own "dart leader" that moves down through the plasma more rapidly than the inaugural avalanche. The energy dissipated in lightning goes mostly into heat, and secondarily into visible light and radio waves. Once the current has passed, the air in the column is at an extremely high temperature and a correspondingly high pressure, and expands explosively, thus generating the shock wave that gradually changes into the sound waves we hear as thunder. Consistently with our earlier example, we may isolate for consideration the correlated occurrence of the two kinds of wave fronts - namely, thunder and lightning - and designate them as A and B, respectively. It now falls to the proponent of the fork principle to show how the characteristic determining antecedent C of A and B is unified in a way that distinguishes it substantially from the characteristic determining consequent, C*, in a deterministic world. 30 Various atmospheric and terrestrial states containing: the dissipated energy of thunder and lightning (looking bottom-up) the potential energy of thunder and lightning (looking top-down) jl Thunder and Lightning Current in a plasma column Fig. 4.1. Two correlated event types, A and B, together with their characteristic antecedent C and their characteristic consequent C*. Despite the overall complexity of the lightning stroke, it is possible to identify a comparatively simple core phenomenon that determines both A and B from the bottom up. At root, it is the passage of a powerful electrical current through a narrowly circumscribed column of plasma that determines the occurrence of both thunder and lightning higher up the time axis. So, let C be the current of electrons in a plasma column. A deterministic consequence of C is the bottom-up dissipation of a large amount of energy, some in the form of light (the lightning flash), and some in the form of heat, resulting in high pressure and an attendant shock wave (the thunderclap). Determinism entails that immediately after the occurrence of A and B, there wil l be a state every bit as characteristic as the state immediately preceding. But conceding this point need not be devastating to the proponent of the fork principle. Peculiar to C is the fact that it is not further divisible into two logically, nomologically, and causally independent parts, one of which is the characteristic determinant of the lightning flash and the other of which is the characteristic determinant of the thunderclap. A prodigious flood of electrons is the determining antecedent of both. That is, C is the determining antecedent of A, B, and A&B. But let us turn now to the case of C*. We suppose that 31 determinism entails that each of A, B, and A&B will have characteristic consequents. Let us suppose that (1) a determining consequent of a thunderclap is a more or less radially symmetric shock wave that has dissipated in the atmosphere and in whatever objects it happens to intercept, and that (2) a determining consequent of a lightning flash is a more or less radially symmetric distribution of light waves that has been scattered and absorbed by various obstructions, and that (3) a joint occurrence of A and B is determined by a combination of the two aforementioned conditions. Note, however, that i f we change temporal gears and run the usual causal story in reverse, we have that the event A is caused by a collapsing shock wave, and the event B is caused by a collapsing distribution of light waves, and that A&B go on to jointly cause C. The top-down chains of determination culminating in A and B are distinct and independent of one another in a way that their bottom-up chains are not. Consider the analogous counterfactuals: " i f C, then A&B", and " i f C*, then A&B". If C is a current of electrons in a plasma column, and C* is a state including both a dissipated shock wave and a certain scattered and absorbed distribution of light waves, then so far as determinism is concerned, both statements are generally true. But unlike C, C* is divisible into two logically, nomologically, and causally independent conditions - a dissipated shock wave on the one hand, and a distribution of scattered and absorbed light waves on the other. One condition determines A, and the other independently determines B. The chains of determination of which these conditions are temporal parts become unified only at the common determinant, C. The case of thunder and lightning is, I believe, is a clear cut example of the sort of "unity" that Horwich has in mind, and which, i f an example of a sufficiently general 32 phenomenon, establishes the fork asymmetry in a deterministic world despite Arntzenius' objection. It may be granted that i f determinism is true, characteristic antecedents will be matched by characteristic consequents. The characteristic consequent, i f sufficiently large in scope, will include all those states of affairs necessary to determine A&B from the top down. But unlike the characteristic antecedent, the consequent is typically divisible into logically, nomologically, and causally independent parts, each of which suffices to determine one of A or B, but not both. The antecedent, by contrast, is not divisible into parts that are independent in this way, each of which suffices to determine one of A or B. The normal fork is therefore a time-asymmetric phenomenon, and one may hope to explain other time-asymmetric phenomena in terms of it. I have not dealt with all objections to the fork principle here. I will note in passing that Healey (1991) has pointed out that i f the world is indeterministic, there seems to be no way to rule out the possibility that at some stage a correlated A and B may simply lack causal antecedents. In this event, the connection between the postulated initial conditions of the universe and the fork principle breaks down. But deeper still runs the objection that the initial micro-chaos and random cosmic input noise themselves stand in need of further explanation. Why indeed should such conditions obtain, in an isotropic spacetime? Lawrence Sklar (1974, section 5, and unpublished) argues that so long as the postulate of initial micro-chaos is left unexplained, the weak coupling approach does not supply a satisfactory solution to the puzzle of asymmetry. Sklar's challenge is an important one, and an attempt to meet it would require a whole-hearted plunge into cosmological issues 33 beyond the scope of my competency.12 Yet, with Horwich, I ' l l adopt the stance that despite the present unavailability of a deeper foundation, the importance of the explanatory work done by the postulate of initial micro-chaos should not be underestimated. I shall hereafter take as an assumption the general hypothesis that it is some feature of the initial conditions of the universe (plus the weak coupling of all systems to the rest of the universe) that determines the basic time-asymmetric phenomenon. The most important item of information to be borne in mind throughout subsequent discussion is the fact that in our region of the universe, bottom-up fork-asymmetric processes are highly probable, and their top-down counterparts are highly improbable. And, since biological activities number among the processes that are subject to this de facto law, we may expect the prevalence of fork-asymmetries to be readily legible in the make-up of biological organisms, from the physiological level right up to the phenomenological level. Most particularly, we may expect bottom-up fork-asymmetry to figure prominently in the differentiation of memory from anticipation. 1 2 1 , for one, favour a model like that which Thomas Gold proposed in the early 1960s. In a Gold universe, the Big Bang is paired with a Big Crunch, and a condition of low entropy ("smoothness") obtains at both ends. In the spirit of the world block, one might try to picture such a state of affairs as a solid sphere between whose two poles extends a continuum disc-like stages (of various diameters, corresponding to the size of the universe at various times). In the lower hemisphere, the sundry arrows of time (psychological, epistemological, radiative and thermodynamic) are all oriented more or less upward, while the arrows in the upper hemisphere are oriented more or less downward. At the (perhaps fuzzily-defined) equator of the "world ball", macroscopic entropy tends to be at a maximum, and the cosmic input noise arriving from the direction of one pole resembles that arriving from the other with respect to implicit order. As one approaches the equator, I would argue, all the arrows of time fade and ultimately disappear. Thus, the equator of the Gold universe is no place for the ordinary operation of familiar systems, nor any time-reversed analogues. Far from the equator, however, systems in either hemisphere experience a de facto asymmetric cosmic environment and temporal arrows are robust. Gold's proposal has the merit that it is innocent of the temporal double-standard inherent in supposing that a certain boundary condition obtains at one end of the universe, but not at the other. 34 4.1 Recording Devices The fork principle seems to hold out the promise that many, i f not all, time-asymmetric phenomena may be explained in terms of it. Of greatest interest to us here is the dramatic time-asymmetry of human psychology. This lopsidedness reveals itself in the asymmetry of our knowledge (we seem to know a great deal less about the future than we do about the past), the asymmetry of our attitudes (we prefer the good times to be in the future, and the bad times - i f any - to be in the past), and the asymmetry of our deliberation and action (we deliberate and act for the sake of the future, not the past). One may expect that a treatment of each of these phenomena, and the relations between them, is central to a satisfactory account of illusory time flow. Firstly, then, I ' l l look at the asymmetry of our knowledge, and see how successfully it may be grounded in the fork principle. Horwich's discussion of recording and pre-recording devices in a fork-asymmetric world is a helpful beginning (though not without its obscurities, as will be dealt with in due course). Horwich presents an idealized picture of the recording system S, which is capable of entering into any of a range of mutually exclusive states: SO, SI, SI, and so on. S will pass from its neutral state 5*0 into the state Sk i f and only i f the external condition Ck should occur in its vicinity, and will remain in Sk at all later times (i.e., the informative state is stable). Ck may be said to be the canonical antecedent of Sk, for i f S is ever observed to be in Sk, it is certain (at least in the ideal case) that at some earlier time, the conditions surrounding S were Ck. 35 In real systems, of course, the state Sk isn't always perfectly stable. Recordings may sometimes be lost. Also, Sk may sometimes be associated with conditions other than its canonical antecedent. S may occasionally ww-record. And finally, sometimes a canonical antecedent Ck will occur, but Sk won't (perhaps because S isn't sensitive enough), such that a recording is missed. Even so, the occurrence of Sk at least makes the occurrence of Ck lower down very probable. Alternatively, we may say that Sk is a reliable symptom of Ck. Instances of bottom-up recording devices are so abundant as to be practically ubiquitous. Examples include footprints, fossils, tape recordings, and of course memory. Horwich seeks to account for the prevalence of such devices by appealing to the fork principle, underlaid as it is by the postulate of initial micro-chaos. At bottom, our neighborhood is bounded by a condition of micro-chaos, with the result that apart from the conditions C l , C2, . . . , all other external forces (cosmic input noise) impinging on the recording system from the bottom up are random. In the case of the bottom-up system S, whose informative state Sk occurs higher up the time axis than the point at which the condition Ck occurs, the input of random cosmic noise has no tendency to lead to any particular states. In the unlikely event that subsequent to being exposed to Ck, S should happen to enter some informative state other than Sk (a mis-recording), or fail to enter any informative state (a missed recording), no one state is more likely to turn up than any other. Horwich takes as an assumption that the informative states 51, 52,... of S occur disproportionately often (relative to S"s various non-informative states). On the basis of this assumption, he goes on to maintain that the only likely explanation of a recording 36 system's proclivity for a certain informative state is the earlier occurrence of the corresponding canonical antecedent. Thus, the fork principle is brought to bear to explain the clustering of informative states. In his words: ... it is important to recognize that the phenomenon of recording is an instance of the pattern of events that is known... as a "normal fork".... A recording system, S, gets into each of its informative states, SI, 52,..., much more often than it gets into its noninformative states - those that are not associated with any particular environmental circumstances. And this heavy clustering constitutes a correlation that is explained by the frequent presence of prior circumstances C l , C2, .... Thus, the association of S being in informative state Sk and prior condition Ck, which is essential to the performance of recording systems, is an instance of the general fact that correlations are causally explicable. (AT, 85) Recording Now Now Figure 4.2 Bottom-up recording devices. However, Horwich's assumption of a heavy clustering of informative states appears to be less than plausible. A typical recording system may well spend the majority of its existence in non-informative states, such as the beach that never suffers a footprint or the book in which nothing is ever written, or it may get into a non-informative state more often than it gets into an informative state, such as in Savitt's (1990) example, a video cassette that is erased after each distinct program is recorded on it, such that it is blank more times than it contains any one particular record. It seems not to be a general truth that any given recording system gets into its informative states more often than it gets into its non-informative states. It is even less plausible to suppose that S gets into each of 37 its informative states with disproportionate frequency, for presumably there are vastly many informative states into which a typical recording machine could enter, but never does. Of all the programs that could be recorded on a video cassette, for example, only a tiny fraction of them are likely to actually make it on. Horwich's derivation of the conclusion that P(CklSk) = 1 (i.e. that Sk is a reliable symptom of Ck) rests on the assumption that P(Sk)»P(Sz), which is just to say that there is a relatively high frequency of informative states oVer noninformative states. This preponderance, i f real, would be something that would need to be explained. In light of the preceding objections, however, I think a more fruitful strategy would be to proceed not from the dubious assumption of a preponderance of informative states, nor even from the correlations holding between sets of devices that have recorded the same event (as Savitt proposes), but from the fact of Sk's order. The most salient feature of the informative state Sk is its improbable order relative to its non-informative states. The state Sk is a branch system that typically constitutes a marked departure from entropic equilibrium with the main system. The footprint in the sand, for example, is a rare, improbable, ordered formation on a beach otherwise smoothed by the surf, and so too is the orientation of magnetic particles on a video tape that had been hitherto blank. And most relevantly to the subject central to this paper, a memory trace in a human brain is, I submit, likewise an improbable ordered state that stands in need of explanation. So far, the fork principle has been used as an explanatory tool to account for the correlations between event types. Thunder and lightning, for example, are two types of event that frequently occur together. This correlation is explained by appeal to a shared, 38 characteristic antecedent event type (a strong current in a plasma column). A single coincidence generally doesn't suffice to motivate us to look for a characteristic determining antecedent, since a single coincidence doesn't suffice to establish an improbable correlation. Nevertheless, the fork principle may be brought to bear to illuminate singular events (event tokens) as well, provided that the event is in itself an instance of a sufficiently improbable (set of) correlation(s). The state of being highly ordered (having low entropy) is often just such an instance. In chapter IV of his (1971), Hans Reichenbach extends the thermodynamic concepts applicable to systems of molecules to systems of macro-objects (such as grains of sand and playing cards). Under a certain rule, different arrangements of the macro-objects may be indistinguishable (relative to some group of persons, perhaps). The class of such rule-indistinguishable arrangements constitutes a state of the system of objects, and its members are said to have the same order. The number of same-order arrangements in a state is a measure of the macroprobability or macroentropy of that state. Yet there is an important difference between microprocesses and macroprocesses, for while the constant collision of molecules serves to keep objects naturally shuffled at the micro-level, often no such mechanism is at play at the macro-level. Or, when there is a mechanism, it operates only very slowly (compared to the rate of micro-shuffling). So, i f a set of macro-objects enter into a state of improbable order, it is likely to persist longer (i.e. be more stable) than a similarly improbable state holding among a set of molecules. Records, then, are "ordered macro-arrangements the order of which is preserved; they are frozen order, so to speak" (1971, p. 151). If we regard records as branch systems - nearly isolated subsystems of the universe at large - then we note that each is characterized by the same 39 sequence of events: interaction with an ordered system, a series of (slowly) less ordered states, equilibrium and/or reabsorption into the main system. As it turns out, the interaction is always lower on the time axis than the equilibrium state, but the fact that it is "lower" is only a function of the time ordering convention we have adopted (i.e. the world block could just as easily be turned upside-down, such that what we familiarly know as the past would become "up" and our psychological arrow of time would point from the top down). If we choose to have the direction of increasing macroentropy coincide with the direction of psychological time, then it is a matter of fact that the stable ordered states of more or less isolated macrosystems record past rather than future interactions. Horwich accounts for thermodynamic irreversibility as a special case of the fork asymmetry. As he explains it, the nonoccurrence of entropy-decreasing branch systems is due to the absence of any correlation between the internal states of the system and the forces impinging on it from outside. In his words: The random nature of continual disturbing forces will overwhelm any systematic bias in the distribution of microstates produced by the "creating" initial reaction. And as long as this distribution is not correlated with the outside noise - as long as the internal state of the system and the external interference are not miraculously coordinated with one another - then the probability is very small that on any given occasion the initial state will be the one required to generate an entropy drop. (AT, 71) The appearance of an ordered macro-state in the midst of a system that has otherwise achieved equilibrium (such as a footprint in the otherwise smooth sand) usually signals a local correlation between the external forces impinging on the system and certain of the system's internal microstates. Unlike a beach smoothed by the surf - a macrostate compatible with a large number of arrangements of the component grains of sand - a footprint is improbable and ordered. As Reichenbach puts it, interaction explains order. 40 If we find a set of objects in a low macroentropic state, we may infer with a high degree of reliability that those objects interacted lower down with another system of comparable or higher order (such as a foot). The correlation between the strong external forces and the internal microstates of the soon-to-be-ordered grains of sand leads to a state of lower macroentropy. If one grants the feasibility of accounting for local drops in macroentropy in terms of a local correlation between external forces and internal microstates (a correlation that, in the case of nearly isolated systems only weakly coupled to random cosmic input noise, does not obtain), a new way to entrench the operation of recording devices in the fork asymmetry becomes available. In light of Savitt's criticisms, Horwich's derivation of the result P(Ck/Sk) = 1 from a relatively high frequency of informative states over noninformative states (plus a randomness assumption) appears to be unsound. However, the analogous conclusion that P(C/A&B) = 1 (which says that nearly all co-occurrences of A and B are explained by the presence of a common antecedent determining condition Q , is a felicitous substitute. Cosmic input noise, as discussed earlier, tends not to favor the occurrence of any one state in particular. Therefore, much as in the case of thunder and lightning, an attendant condition of the order-producing correlation between external forces and the internal microstates of certain macro-objects is very likely to be a highly ordered determining antecedent, but not necessarily an ordered determining consequent, due to the asymmetry in the boundary conditions of the universe. Taking the general result, P(CIA8cB) = 1, and substituting (1) Ck for C (given the assumption that only the canonical condition Ck is likely to knock S into Sk), and (2) the ordered informative state 41 Sk for the set of correlated events A&B, we return to the desired result, P(CklSk) = 1. iSA: is a reliable symptom of Ck. 4.2 Informational Content A deficiency in Reichenbach's account that Horwich wishes to remedy is the fact that from the occurrence of Sk, we usually learn more than merely that S has interacted with some more ordered state in the past. We often learn very detailed information. In his attempt to accommodate this fact Horwich provides a way to connect epistemological time asymmetry to other epistemological problems and theories. Savitt (1990), for example, notes that the informational content of Sk (in the ideal case) exactly parallels the sense of "informational content" defined in chapter 3 of Dretske (1981), for the conditional probability of Ck, given Sk, is 1 (that is, P(CklSk) =1). From Sk we can learn that Ck, if our belief that Ck is caused or causally sustained (typically by observing Sk) by the information persisting with indefinite stability in Sk. However, if it is allowed that S can sometimes get into Sk without Ck having occurred lower down (a mis-recording), then the conditional probability of P(CklSk) drops below 1 (however slightly). For Savitt, this result destroys the informational content of Sk, for as he says, "I think it is true, as Dretske insists, that one cannot then leam that Ck from Sk no matter how close to 1 P(CklSk) is, just as I cannot know that my lottery ticket will lose in a fair lottery no matter how close to 1 the probability that it will lose" (1990, p. 322). Savitt's objection turns on a sense of "knowledge" that seems stronger than the sense in which the term is ordinarily used. My response to this objection follows the lines of J. 42 L. Austin's (1961) contention that typical philosophical assessments of our knowledge of the world distort or abandon our normal, everyday practices. Suppose, to use Austin's example, that I claim to know that there is a goldfinch in my garden. Yet as soon as I do so, an acquaintance voices the challenge, "How do you know?". I might defend my knowledge claim in a number of ways. I might mention how I have learned about goldfinches, or note what it is about this particular bird that makes me think it is a goldfinch. If, for example, I cite as my grounds for claiming to know the bird is a goldfinch the fact that it has a red head, my acquaintance might well object that my grounds aren't good enough, since lots of different birds have red heads. My acquaintance has raised a possibility which, i f actual, would imply that I don't in fact know that the bird is a goldfinch. Fair enough. But suppose that even i f I were to counter this objection with further evidence, my acquaintance were to keep on objecting "That's not enough", ad infinitum. As Austin points out, we all ordinarily accept that: (a) If you say "That's not enough", then you must have in mind some more or less definite lack.... If there is no definite lack, which you are at least prepared to specify on being pressed, then it's silly (outrageous) just to go on saying "That's not enough". (b) Enough is enough: it doesn't mean everything. Enough means enough to show that (within reason, and for present intents and purposes) it "can't" be anything else, there is no room for an alternative, competing description of it. It does not mean, for example, enough to show it isn't a stuffed goldfinch. (OM, 52.) When philosophers go on to raise questions about "reality" ("But how do you know it's a real goldfinch?") they intend to question the reliability of the "facts" put forward in support of the original claim to know. That, of course, is something we do in everyday life too. Yet Austin thinks that philosophers sometimes question our grounds for making a knowledge claim without specifying or limiting the ways in which the goldfinch might 1 3 J. L. Austin, 'Other Minds', in his Philosophical Papers (Oxford, 1961). (Hereafter cited as OM.) 43 not be real, and without having some special reason for suggesting that the specified possibility might obtain. It is only in special circumstances that certain kinds of possibilities are relevant to claims to know something (this, despite the omnipresence of the bromide, "Hey, you never know!").14 The distinction that I would like to make explicit here is nothing new in the field of epistemology, but is relevant to my objectives: the distinction between a justified belief on the one hand, and an instance of strict knowledge on the other. In everyday practice, these two concepts are seldom distinguished. Normally we take it for granted that what we are justified in believing is in fact true, and what we are justified in disbelieving is in fact false. The adequacy of one's grounds for making a knowledge claim, the point at which "enough is enough", usually varies according to the nature of one's present intents and purposes. Some circumstances call for very careful scrutiny of one's grounds (such as when the cost of making an error is high), while others motivate comparatively little self-justification. If I claim to know that Ck in light of the observed fact that Sk, citing as my grounds that Sk is a reliable (albeit imperfect) symptom of Ck, then ordinarily, an objection of the form "That's not enough" is not a successful challenge unless the critic has in mind (1) some specific condition or limited range of conditions that would imply the falsehood of my claim to know, and (2) some grounds for supposing that (at least one of) those conditions actually obtains. The question "How do you know" is not ordinarily a successful challenge i f it is based only on general human fallibility, or, for that matter, on small departures from the ideal, P(CklSk) = 1. 1 4 Dretske does not deny this claim; see chapter 5 of his (1981) for his "relevant alternatives" defense of knowledge against skepticism. 44 Savitt claims that " i f mis-recording is a serious, rather than a skeptical, possibility, then the states of S are powerless to convey knowledge of the past" (1990, p. 322). Granting that the determining antecedent of Sk may, on rare occasions, be other than Ck, this fact does not ordinarily threaten the justifiability of my belief that Ck on the basis of Sk, unless I have reason to suppose that an alternative condition has in fact obtained. The probability of winning BC's Lotto 6/49 with a single ticket, for example, is about 7x10"8. While of course it's not likely that anyone would buy a lottery ticket i f knowing the improbability of winning were tantamount to knowing that one wouldn't win, this example is nevertheless somewhat deceptive, for I submit that a good number of our everyday knowledge claims (those for which there is quite a bit less at stake than a 6/49 jackpot) have a better than one in ten million chance of being wrong. In the particular case of the lottery, it is (1) our precise awareness of a condition which, i f it obtained, would imply that I didn't really know that I wouldn't win the lottery (namely, an agreement between the numbers that are drawn and those that are on my ticket), together with (2) our awareness of numerous actual instances of people winning lotteries, that introduces a sliver of daylight between the usually conflated concepts of "justified belief and "knowledge", and makes us hesitant to embrace the claim, "I know that I won't win." In many other cases with similar probabilities of error, provided that the falsifying alternatives aren't so obvious and that actual occurrences of those alternatives aren't so familiar, one might expect knowledge claims to be a good deal less tentative. The justifiability of having a given belief doesn't normally hinge on a capacity to establish the unimpeachability of one's evidence. It hinges merely on the reliability of 45 one's evidence, together with the absence of evidence suggesting the actual occurrence of an improbable, but nevertheless possible, alternative. Indeed, as Horwich points out, although we have tended to describe the epistemological asymmetry as a striking difference between our knowledge of the past and of the future, it seems plausible that the underlying asymmetry here has primarily to do with justification and with the reliability of mechanisms that link our beliefs to the world. (AT, 81) If P(Ck/Sk) = .99, say, then there is a .01 probability that some determining antecedent other than Ck has occurred. That is, i f P(CklSk) is less than 1, then Sk is indeed a symptom of not only Ck, but also all alternative determining antecedents. However, i f P(CklSk) is nearly 1, then Sk provides excellently reliable evidence of C&'s occurrence, and quite wnreliable evidence of the occurrence of any possible alternative. To the extent that a challenge to our ordinary knowledge claims requires, in order to be successful, grounds for supposing that a possible alternative has in fact obtained, then the mere fact that P(CklSk) falls a little short of 1 fails to satisfy this condition. What is relevant to the evolutionary theme I will be developing hereafter is the justification that Sk provides to believe in Ck, not some capacity to establish that the conditional probability of Ck, given Sk, is 1. Put another way, whether an article of information is strictly speaking known or not is a side issue in the context of evolutionary imperatives. The term "knowledge" need not make another appearance in subsequent discussion; instead, talk about "degrees of belief will serve my ends just as well. A successful organism generally can't afford to eliminate all improbable, albeit still possible, alternative conditions from which Sk may have arisen. The fact that PiCklSk) = 1 is generally an adequate and reasonable foundation on which to base behaviour, in the context of one's pragmatic intents and purposes. 46 Even if there are legitimate epistemological issues remaining that I have failed to come to terms with here,15 and that Horwich's account of recording devices is inadequate to handle, I may leave them aside without weakening the case I wish to build. What is relevant to my present purposes is the conditions under which a belief in Ck is justified, regardless of whether one really "knows" that Ck. And in general, the fact that P(Ck/Sk) s 1, in the absence of a (serious, rather than a skeptical) reason to suppose that ~Ck, satisfies those conditions. 4.3 Pre-Recording Devices As earlier noted, instances of bottom-up recording devices are so abundant as to be practically ubiquitous. Yet by contrast, examples of top-down recording devices (what we would call />re-recording devices, from our usual bottom-up perspective) are conspicuously lacking. In a world full of barometers and newspapers and footprints, prognosticating crystal balls are the stuff of fantasy. Why such an imbalance? Horwich grounds his answer in the boundary conditions of the universe. At top, he proposes, the boundary condition of micro-chaos does not obtain. While the initial conditions of the universe may be random, the final conditions are not. Let's consider another ideal recording device much like the sort earlier discussed, except that it records from the top down. Thus, this system makes records of events that occur higher in the world block than the corresponding informative state, not lower. We agree to call this time-reversed system S*. It is important to bear in mind that what we refer to as input from the bottom-up perspective becomes output from the top down perspective, 1 5 See chapter 3 of Dretske's (1981), for a definition of "informational content" that requires that the conditional probability of Ck, given Sk, is 1. 47 and vice versa. However, unlike the cosmic input noise that impinges upon the system S from below (which is random), the cosmic input that impinges upon the system S* from above may have implicit order. Horwich presents a picture of a universe that initially contains large-scale inhomogeneities of energy (macroscopic order) but that, subject to that constraint, is as chaotic as possible on the small scale (microscopic disorder). Yet, as time goes on and the universe approaches a state of equilibrium, the macroscopic order dissipates, but implicit microscopic order accumulates. The cosmic noise that travels down from the universe's upper boundary (or its equator, i f it is a Gold universe - see footnote 12, p. 35) is likewise implicitly ordered. As a consequence, it may well be that the internal microstates of S and the external forces arriving from the top down will be coordinated with each other in such a way as to determine a macroentropic drop farther down. Thus, a condition that was necessary for the success of S as a recording device does not obtain for S*. It may no longer be assumed that the cosmic input is just as likely to lead to one informative state as any other. Before, we could conclude that i f 5 were to enter one of its informative states SI, 5 2 , i t would be unlikely to have been preceded by anything other than its canonical determinant. The reason for this was the fact that random input noise was very unlikely to be coordinated with the internal microstates of S in such as way as to lead to a macroentropic drop. But looking from the top down, i f it should happen that S* enters some ordered state Sk*, then this occurrence may simply be the result of the coordination between implicitly ordered top-down noise and the internal microstates of S*. There is no regularity to the effect that correlated events are always associated with a characteristic antecedent event (where "antecedent", in this case, now refers to positions higher up the 48 time axis). We cannot similarly conclude that from the top down, the ordered states of S* are the attendant conditions only of particular types of "prior" circumstances. Prerecording _ |_ \ System Now Now Not especially improbable, since / can have implicit order Figure 4.3 Top-down recording devices 4.4 Knowledge Asymmetry It would seem that there are at least two distinct but closely related ways to approach the issue of knowledge asymmetry. One approach, by far the most common in the literature, involves an examination of what kinds of inferences about the past and future we are entitled to make on the basis of evidence available to us at any one time. For example, one may attempt to build a theory on the premise that the world abounds with traces of the past but no (or few) traces of the future, such that we are entitled to make inferences about the condition of the past on the basis of available traces, but not the reverse. However, one of several criticisms that may be brought to bear against this ingenuous view is this: i f we know that Ck and Sk are attendant conditions of one another, then we are entitled to infer not only that Ck has occurred on the basis of Sk, but also that Sk will occur on the basis of Ck. The behaviour of properly operating recording devices is predictable, such that it is unclear how the assumption that a recording system is likely to 49 get knocked into Sk i f and only i f Ck occurs lower down leads to our possession of more knowledge of the past than of the future. The second approach, the approach that will provide deeper insight into the psychological arrow of time, involves regarding the brain itself as a recording device in which traces accumulate from the bottom up and linger with varying degrees of stability. With this strategy, the focus is on the informative states Sk that occur within the brain itself, not on all those informative states occurring in the world without, on the basis of which we may (or may not) make inferences. It is easy not to pay sufficient attention to the fact that the very acts of receiving sensations, manipulating them in the brain, and making inferences on basis of the results are themselves processes constrained not only by nomic conditions, but also by the de facto irreversibility implied by the fork principle. Epistemologists who seek to account for our knowledge asymmetry in terms of what sorts of inferences are licensed and which are not are often at risk of allowing our time bias to slip into their accounts through the back door. I want to claim that the epistemological arrow of time (and, at root, the psychological arrow) is a function not of the fact that the various informative states present in the external world license more inferences about the past than about the future, but rather of the very spatiotemporal structure of the processes that take place in the brain (and in the world at large). An informative state Sk of the brain, and of recording devices in general, is a persisting structure. Moreover, the brain is capable of accommodating numerous informative states at once. These states accumulate from the bottom up. Along its temporal length, the brain undergoes various interactions with its environment, an attendant condition of which is a persisting informative state above the point of 50 interaction, but not below. To the extent that the stream of experience is divisible into stages, so too is Sk. Any given experience at-a-time, then, will harbor stages of any of a number of informative states. These stages will be traces, or reliable symptoms, of events occurring lower down the time axis, but not necessarily of events occurring higher up. The relative time order of these traces is established with variable degrees of accuracy by means of the ordering mechanisms sketched in chapter 3. As earlier discussed, the makeup of our cosmic neighborhood is such that any brains resident therein are vastly much more likely to succeed at recording events from the bottom up rather than from the top down. Speaking in evolutionary terms, we expect that organisms will exemplify a capacity for bottom-up recording that is greatly more efficient than any analogous downward-oriented faculty. Again, while the environment does not spontaneously generate misleading records from the bottom up, it can easily generate misleading records from the top down, due to an asymmetry in the boundary conditions of the universe. These results lead us to expect (1) a prevalence of memories and a comparative paucity of anticipations within any given experience-at-a-time, and (2) that the process of anticipation is likely best understood as a form of bottom-up inference based on familiarity with natural and social patterns, rather than a genuine instance of top-down recording. Here, then, we have the rudiments of an account of the brute phenomenological dissimilarity between memory and anticipation, providing at least a partial answer to the question of why the future should seem to be so much more open and mysterious than the past. We are that much closer to providing an account of the value asymmetry that characterizes our view of the future and the past. In the following chapter, I will show 51 that the evolutionary pressures peculiar to a de facto asymmetric world naturally predispose us to exhibit a concern about the future that is deeply different from our attitude about the past. In short, I will attempt to provide an account of the psychological arrow of time, grounded in both the fork principle and in evolutionary biology. Having accomplished that much, I will go on to show why it is naturally to be preferred that the meanings of "earlier" and "later" should be determined relative to the direction given by anticipation-memory, rather than the reverse.. 52 • Evolutionary Biology from A to Z, and Z to A A n ample assortment of biological organisms presently inhabits our world. And while each species may be well adapted to exploit the particular ecological niche in which it has entrenched itself, all are alike in that they occupy an environment in which bottom-up fork asymmetries prevail. What kinds of fundamental features, then, would we expect to be common to all (or virtually all) organisms residing in such an environment? Organisms are highly ordered collections of matter. Such collections don't just spontaneously assemble out of scattered bits from the bottom up (with the probable exception of the primordial precursors of life). Generally, an organism's appearance on the scene requires some highly ordered mechanism - such as the reproductive faculties of parent organisms - to be in place first, the natural operation of which leads to genesis. From the top down, however, we note that organisms and other highly ordered structures routinely assemble out of the ambient flotsam. Hereafter, I will be making an especially scrupulous habit of flipping the stories I tell on their temporal heads - i.e., considering series of events from both a bottom-up and a top-down perspective - so that ultimately, I may develop a picture of evolutionary imperatives that is neutral with respect to orientation in time. With that goal in mind, let's walk through some straightforward evolutionary truisms. If an organism has a finite life-span, then its world-line will have two termini in the world block: an upper and a lower one. As mentioned just earlier, the lower terminus will be associated almost invariably with a highly ordered mechanism the operation of which, when read from the bottom up, leads to the transmission of genetic material from a parent organism to its offspring, and 53 the initiation of the growth process. Bottom-up reproduction may take place via any of a number of patterns: mitotic, parthenogenetic, hermaphroditic, oviparous, placental, and so on. Diverse as these strategies are, they are a scanty selection compared to the variety of demises available to the typical organism - variety of such breadth as seems to obscure the inevitability of death, at least in many a human eye. Reading from the top-down, we see a vastly diverse array of origins. An organism's upper terminus need not be, and indeed frequently is not, associated with any highly ordered mechanism. A given organism may, for example, be secreted by the intestinal tract of a predecessor, or it may assemble spontaneously out of scattered organic bits, or so on. Always, however, the termination of the organism's span at bottom is associated with a highly ordered mechanism the operation of which, when read from the top down, results in the culmination of the dematuration process, and the absorption of genetic material from the organism by a successor. A typical biological organism shares some deep but perhaps unobvious similarities with the recording devices discussed in chapter 4. In the case of the bottom-up recording system S that enters the ordered informative state Sk, we concluded that it is very unlikely that anything but the canonical antecedent, Ck, could have led to 5"s being in that state. However, in the case of the top-down recording system S* that also enters the informative state Sk, we concluded that we have no particular reason to suppose that Sk is likely to have been "preceded" by any characteristic event, since the cosmic input noise that impinges upon S* from the top down may have implicit order. Therefore, while the state Sk is extremely unlikely to come into being from the bottom up except in virtue of the 54 prior occurrence of Ck, it may come into being from the top down in virtue of any of a number of conditions. The informative state Sk is usually a stable, improbably ordered structure, a departure from macroentropic equilibrium with the main system. In this basic respect, it is not unlike a biological organism O, which is itself a stable, especially highly ordered structure. Also, we find that occurrences of O are extremely likely to be associated with a characteristic event at bottom - namely, a reproductive (/ absorptive) mechanism. However, occurrences of O are not especially likely to be associated with a characteristic event at top. In other words, while O is extremely unlikely to come into being from the bottom up except in virtue of the prior occurrence of its canonical antecedent, it may come into being from the top down in virtue of any of a number of "prior" conditions. One may see that biological reproduction is fork asymmetric in the same basic way that recording devices are. O's similarity to Sk in this respect is not merely coincidental; it is a consequence of the same fundamental cosmic conditions that are responsible for a prevalence of fork asymmetries in general. A viable, indefinitely-persisting population of typical organisms describes a network of sorts in space-time, frequently forking from the bottom up. Each of the network's branches is the world-line of an organism, virtually always conjoined at bottom to the world-line of a genetically similar organism, and terminating at top amid various circumstances. Beyond these basic features, the bio-network will occur in different overall patterns, depending on the range of bottom-up reproductive strategies employed by the population in question. The bio-network on Earth is at least three and a half billion years tall from bottom to the present time. Without knowing how close to the top we ourselves are located, it may well be 55 substantially taller all told, perhaps spreading across the long voids of space in higher reaches of time. Though the bio-network is a sprawling crazy-quilt, characterized by multifarious genetic strata and branch systems - some widespread, and some confined, some diffuse, and some densely knit - we find that in general, the number of termini in any suitably large region will be twice the number of conjunctions (i.e. fissions / fusions of successors and predecessors). And regardless of the diversity of the reproductive strategies incorporated into its weave, the bio-network must everywhere jointly satisfy two sets of imperatives in order to fortify its probability as a persisting fixture of the world block. 5.1 Biological Imperatives Firstly, the bio-network must satisfy biological imperatives. Barring those sorts of (mythical) organisms neither of whose termini conjoin with that of a genetically similar predecessor (/ successor) (such as organisms that emerge from a process of spontaneous generation at both termini), organisms with finite longevity must reproduce themselves either from the bottom up or the top down in order for the species to persist indefinitely. The kinds of organisms with which we are acquainted in this fork-asymmetric world propagate their own species from the bottom up.1 6 For the purposes of this paper, we may restrict our attention to the sorts of reproductive patterns common to the class of mammals, since it is, after all, a human phenomenological perplexity that is the focus of this paper. 1 6 An interesting case might be made for cannibalism as an isolated example of top-down same-species propagation. 56 A t least in the case of placental animals, the reproductive imperative tends to be better satisfied by maximizing the temporal extent of an organism's reproductively active phase. The longer the period during which an organism is capable of reproducing, the more opportunities it w i l l tend to have to reproduce during its life-span. In the absence of any significant over-population pressures, and all other things being equal (such as menstrual cycles, gestation periods, rates of maturation, and average life-span), a population whose members are characterized by longer fertile periods is l ikely also to be characterized by more reproductive episodes than a population whose members have comparatively narrow reproductive windows. O f course, different organisms pursue different reproductive strategies, even among the class of mammals. Certain mammals mature rapidly and produce large numbers of under-developed offspring after brief gestation periods, while certain others mature more slowly and produce a smaller number of better-developed offspring after longer gestation periods. Each strategy comes with its own set of advantages and disadvantages, and there are niches enough in nature to accommodate both. Nevertheless, in the absence of significant environmental constraints, the truism that having more opportunities to reproduce tends to promote reproduction holds. Through what means, then, might the maximization of an organism's reproductive phase be accomplished? Part of the answer, we expect, w i l l be found in the genetic make-up of the organism in question: in the absence of artificial intervention, the physiology of an organism imposes approximate limits upon the maximum potential extent of its reproductive viability. Another part of the answer, however, seems so obvious that only in a discussion such as this need it be explicitly mentioned: the organism must survive, at 57 least long enough to fulfill some or all of its reproductive potential. If an organism expires prior to or during its reproductively viable phase (speaking from the bottom up), or i f it assembles after or during its absorptively viable phase (speaking from the top down), that organism's opportunities to reproduce (/ absorb) will be either nullified or reduced. Maximizing longevity tends to maximize reproductive potential no matter in which temporal direction the process runs (again, all other features of the reproductive process being roughly equal), and thus tends to bolster the probability of the species' persistence in space-time.17 The question now becomes this: how can a disposition to survive be built into an organism occupying a fork-asymmetric world? A n answer to this question will lay the foundation of the value asymmetry that characterizes our view of time, and thus supply the last element required for an explanation of why the "moving now" is perceived to I glide into the future, and not into the past. 5.2 Nomic Imperatives The time-reversible laws of the world block are such that the interactions of microscopic entities are limited to only a certain number or range of nomically possible (or, i f preferred, probable) kinds. The bio-network must respect the kinds of patterns that are available to its constituents at the micro-level. It is a deceptively simple matter to merely stipulate that in order for a population of mammals, say, to be likely to persist indefinitely, its members must tend to avoid obstacles from the bottom up that would otherwise interfere with the continuation of their overall biological organization. We 1 7 The effects of a high death rate can be partially offset by a high birth rate, but as a consequence, a higher 58 might simply imagine a web of organisms imbedded inside the world block, and arrange it so that each member swerves deftly around precipitous drops, hungry predators and whatever else from the bottom up. It's easy to spot many of the behavioural patterns an organism should and shouldn't explore in order to fulfill its biological imperative comfortably. We indulge in offhand counterfactual commentaries upon such patterns as a matter of course: i f Rover hadn't eaten those poison berries, he would still be alive today. However, it's at least as easy to forget that Rover is essentially an elaborate mechanism, whose constituents are but spinning physical bits whose motions are limited to the patterns licensed by Newtonian, quantum mechanical, and / or statistical law (or, more broadly, natural law). Finding a precise arrangement of constituents that is consistent with both (1) the demands of natural law at the micro-level, and (2) the macroscopic structure that we need to build (namely, an organism that tends to persist long enough to fulfill its biological imperatives) is a task that is truly astronomical in its complexity. We, as engineers outside of the world block, cannot arbitrarily position an object inside it without taking the overall world pattern that pervades the block into account. Even a comparatively slight adjustment to one constituent or another will likely necessitate a cascade of adjustments in order to preserve the global nomic consistency of the world pattern in the block, adjustments whose scope will tend to be exceedingly far-reaching, especially when medium-scale adjustments are involved. Counterfactual speculation of the form, " i f this event had been so and thus, then that event would have proportion of the population will be sexually immature, and in circumstances where over-population pressures are significant, the higher proportion of sexually inactive members will burden the population. 59 turned out thus and so", almost always proceeds in disregard of this fact. Indeed, i f we were to suppose that at the bottom of the block there is a comparatively simple original event such as a Big Bang, it may be that there are only a small number of distinct, nomically possible world patterns that may fit in the block and be causally consistent with the fixed original event (here, I neglect the margin of "play" introduced into a world pattern by quantum indeterminacy). ) A viable organism must be so constituted that when it is opportunely situated within a certain range of environments, the maintenance of its organization is the most probable outcome of the nomically licensed motion of its constituents. Put another way, the organism must be engineered to sustain itself, such that when it causally "plugged into" a particular locality of the world block, and "unfolded" consistently with (1) its inbuilt blueprint of ultimately mechanical dispositions, and (2) the patterns of both natural and de facto law, one finds that its organization does indeed tend to persist long enough to fulfill its biological imperative. Crucial to most organisms' success is an innate disposition to manifest some sort of avoidance behaviour, either from the bottom up or the top down (or perhaps both). Across the length of an organism's life-span, its stages are characterized by a more or less stable biological organization. In the case of humans, our organization includes such features as the normal placement and operation of our organs, maintenance of a certain homeostatic balance, and so on. However, nomic conditions being what they are, certain interactions between us and our environment are such that the successful maintenance of our bio-organization in such circumstances would be very improbable. Conversely, certain other interactions are such that i f we didn 7 have them, maintenance would once 6 0 again be improbable. We must be engineered in such a way that our world-lines tend not to enter into those sorts of interactions which, were they entered into, would result in disruption, and intercept those interactions which, when they are intercepted, promote maintenance. Thus, the longest space-time worm possible may be woven into the underlying, nomically rigid matrix, fulfilling the biological imperative. As it turns out, virtually all of our avoidance behaviour is upwardly (i.e. future) directed. That is, we anticipate trouble in the future both near and far, and we act to avoid it. If we see a bus hurtling toward us, we leap. If we anticipate that the value of our stocks will only diminish hereafter, we sell. Seldom if ever do we apply ourselves to bodily avoiding unpleasant circumstances remembered to have occurred in past. If the dog has already bitten us, that's that - there's no effacing that fact from the tablet of history. Better to invest one's energy in locating a Band-Aid. Similarly, i f we took a beating on our stocks in the past, no amount of buying or selling is going to alter that fact, though future successes may help us to assuage our chagrin. Or so we are naturally disposed to believe. The most constructive thing we can do about past disasters, we feel, is to avoid dwelling on the unsavory memory of them in the future. Only the future, hypnotically indistinct, filled with menace and promise at once, is a fitting theatre of action. The past is gone; after all, there's no use crying over spilled milk. How puzzling that attitude is, when viewed as taking place in a world in which all events in space-time are equally determinate, and time is isotropic. 61 • Ashley and Bradley As earlier discussed, there are many common processes the time-reverses of which we simply never see. This, despite the time symmetry of natural laws. The expansion of a gas into an available volume, the diffusion of one liquid into another, and the expansion of spherical wave fronts from a point source, are examples of processes that are de facto irreversible. As a further example that is more relevant to the evolutionary theme that I have been developing here, a dog's bite is never observed to instantly heal the tooth-shaped sore that has developed over a period of time upon a person's arm. As hypothetical observers outside the world block, we are at liberty to read a sequence of events from the bottom up or from the top down. But in general, i f "down" corresponds to what we familiarly know as the past and "up" is the direction in which the future extends, reading a sequence from future-to-past will yield a story that will seem rather bizarre when compared to a reading that runs from past-to-future. The vast majority of the stories that we actually tell "from the midst" run from the bottom up. However, given my presupposition that time is isotropic, it ought to be possible to discuss the impact of evolutionary imperatives in terms that are neutral with respect to time's apparent arrow. As with most stories about temporally-ordered sequences of macroscopic events, the story of life as told from the top down is radically unlike its bottom-up version. Even i f we were simply to reverse the behaviours of a population of organisms as i f running a motion picture backwards, the resultant activity would be completely alien to our experience. We would see organisms emerging spontaneously into being from the most unlikely places. A l l but the simplest organisms 62 would meet the same fate, dwindling and discorporating at last into a pair of genomes absorbed into the tissues of their successors. Ingestion and excretion would proceed by methods outrageous to the bottom-up eye. But in order to get the top-down version right, it is not enough merely to stand a sequence of behaviours on its temporal head. The reversed motion picture metaphor is a deceptively simple method by which to wrap the imagination around a time-reversed story. Most of us have seen a motion picture so reversed, and we have each been bemused by the strangeness of the story thus presented. However, as we sit and watch, our sense of the moving now is still oriented from the bottom up. Light streams bottom-up from the screen and bombards our retinas, and signals are piped through our optic nerves into our brains. Sound too is transmitted and received normally, though we may have difficulty recognizing words uttered backwards. A l l information-bearing signals converge upon our world-lines from below as they have always done. The only difference between a reversed film and real life, although eye-catching, is nevertheless merely cosmetic: the incoming signals inform us of a pattern of events unlike those to which we have become accustomed. To get the top-down version right, it is necessary also to reverse the functions of all biological mechanisms. Some such reversals are obvious. Ingestion becomes elimination. The expenditure of energy becomes the stockpiling of energy. Reproduction becomes absorption, and so forth. Consider, for example, the world-line of organism A, that spans times t\ to tl. At any time t between times /I and t2, certain other world-lines will be converging upon, intersecting, diverging from, or laying parallel to A. But i f "convergence" is to be understood as "moving toward" and "divergence" is to be 63 understood as "moving away from", we are required to impart one of two possible directions to the time axis in order to perform a systematic discrimination between "convergence" and "divergence". From a bottom up perspective, a V-pattern is read as divergence, and a A pattern is read as convergence. From a top-down perspective, the reverse is true. Suppose, for example, that we wished to describe the observable history of one particular resident of the world block named Ashley, whose world-line lays approximately parallel to the temporal axis. Suppose further that we wished to tell a story not only from the bottom-up and the top-down perspective in turn, but also from a perspective that is neutral with respect to the two. How would the corresponding stories run? The Bottom-Up Perspective: The initial constituents of Ashley's world-line emerge from a predecessor's uterus at the bottom, and his remaining constituents disperse in the earth at the top. During one period of time along the Ashley's temporal length, his person-stages are seen to be in a state of comparative well-being. But shortly thereafter, Ashley's world-line and that of a certain dog may be seen to converge, and Ashley is observed to exhibit avoidance behaviour. Ultimately, despite Ashley's efforts, Ashley's world-line and the dog's intersect. Immediately subsequent to the intersection, Ashley's arm-stages are characterized by a trauma, and Ashley continues to exhibit avoidance behaviour until finally he and the dog diverge. Subsequent to the event, his supervening mental stages are characterized by a vivid recollection of having been bitten by a dog. The severity of the injury to Ashley's arm-stages gradually diminishes the farther up from the intersection of Ashley and the dog we look, until he returns to a state of comparative well-being. 64 The Top-Down Perspective: The initial constituents of Ashley's world-line are secreted by the earth at the top, and its remaining constituents are absorbed by a successor's uterus at the bottom. During one period of time along Ashley's temporal length, his arm-stages are seen to be characterized by a degradation whose severity gradually increases the farther down Ashley's world-line we look. Ashley's mental stages are characterized by a vivid "anticipation" (i.e., a downward projection) of being bitten by a dog. In time, Ashley's world-line and that of a certain dog may be seen to converge. Ashley is observed to exhibit something not unlike bottom-up pursuit behaviour, and the dog exhibits avoidance behaviour. Ultimately, despite the dog's efforts, Ashley's world-line and the dog's intersect. Immediately subsequent to the intersection, Ashley's arm-stages are characterized by a state of comparative well-being, and Ashley continues to exhibit pursuit behaviour until finally he and the dog diverge. Subsequently, he retains no "mnemonic record" (upward projection) of having been bitten. The Direction-Neutral Perspective: Ashley's world-line is characterized by two termini, the bottom-most demarked by a connection with uterus stages and the top-most demarked by a connection with earth-stages. At one point, Ashley's world-line intersects with that of a dog. Also, a certain series of Ashley's arm-stages are characterized by an injury. The bottom-most terminus of that series, where the injury is most severe, is well-defined and coincident with the intersection of Ashley and the dog. The top-most terminus, where the injury is imperceptible, is not well-defined. Connecting the termini of the injury is a continuum of intermediately-injured arm-stages. Above the bottom-most terminus of the injured series, 65 Ashley's supervenient mental stages are characterized by a downwardly-projected record of his intersection with the dog. These three stories tell a limited tale. They focus only on a series of observable episodes in Ashley's life viewed from one perspective or another, and neglect the consequences of flipping the temporal orientation of Ashley's stream of experience. A close look at the reversed operation of sensory organs, and the reversed motion of those signals that inform us about the state of the world, will prove to be crucial to understanding the value asymmetry that characterizes our view of the past and the future. For, as will be later discussed, the efficacy of avoidance behaviour hinges on its being informed by the condition of the world without, and this is possible only (or at least principally) in virtue of the bottom-up operation of sensory organs. Nevertheless, an important feature does stand out at this preliminary stage, one that is relevant to predicting the impact of evolutionary imperatives upon the structure of our attitudes. In a fork-asymmetric environment, the pathology of trauma is most commonly as follows. The bottom-most terminus of a span of damaged stages tends to be well-defined and coincident with the intersection of the organism with another macroscopic body, and tends to be the point of maximum trauma. The top-most terminus, on the other hand, tends to be poorly defined, and is usually the point of minimum trauma. Except in the case of injuries that only worsen from the bottom up and are ultimately fatal, traumas generally diminish in severity from the bottom-up, proceeding from a point of maximum damage through a more or less continuous series of intermediate stages until the injury becomes imperceptible (or until the curve more or less flattens before rehabilitation is 66 complete, as with traumas that result in permanent disability). The interesting fact to note is that looking from the bottom up, a dog-bite is the vividly remembered but typically unforeseen cause of a trauma, while looking from the top down, a dog-bite is a vividly foreseen but typically soon forgotten relief from a trauma. From the top-down perspective, the truism "time heals all wounds" becomes law-like in its obviousness. As with so many other states of affairs in the world, a particular type of trauma is usually associated with a characteristic event at bottom, but no particular characteristic event at top. Note also that the aforementioned exception, namely the class of injuries that do not mend appreciably and are often ultimately fatal, are relevantly similar to less threatening mishaps in that from the top-down perspective, relief from them may likewise be vividly "anticipated" (barring another class of exceptions, consisting of defects present from birth). Let's introduce the phenomenological asymmetry of memory and anticipation. We now have that avoidance behaviour is directed toward that temporal pole for which we possess a very poor record ("up"). An upward-reading person seldom foresees either the dangerous situations he will have to avoid or how ably he will manage to avoid them (if at all). The time-reversed process, some strange sort of pursuit behaviour, is directed toward that pole for which we have a very good record ("down"). A downward-reading person will typically foresee when and where he will find relief from his traumas, such as they are. Such a person will also know that no matter how severe the trauma is from which he presently suffers, be it a nick or a burn or what have you, he will always find relief from it eventually (again, barring defects present until uterine-absorption), since he will know that all members of his kind invariably meet their end in the uterus of a 67 successor. Thus, even i f its record of the "future" were not so richly detailed, a downward-reading organism that is confident in its anticipation that there is only one kind of characteristic fate that awaits all of its kind can afford to be apathetic about the prospect of future encounters. This attitude maps neatly onto the indifference we upward-reading organisms feel toward injuries that have occurred in the past. 6.1 Sensation, Information and Action Speculating about the psychology and phenomenology of persons with a reversed perception of the moving now is a dicey business at best. For one thing, given our initial assumption of the truth of physicalism and supervenience, there can be no variations in a person's mental content or operation without there also being a variation in his physical (specifically, neurophysiological) make-up. Actually reversing a person's perception of time's passage, i f possible at all, would require some adjustment to be made to his neurophysiological make-up. But let us suppose, for the purposes of illustration, that it is possible to make just such an adjustment, at least in principle. Let's take a duplicate set of Ashley's brain-stages, and modify them such that only the perception of the moving now is reversed, while all behaviours and other mental operations and attitudes remain unperturbed. We will call this backward person Bradley. Let's now consider Bradley's experiences, who is Ashley's perfect duplicate except only for the detail that he perceives the present to move from the top down instead of from the bottom up. Given the unlikelihood of my initial assumption (namely, the possibility of manipulating brain stages such that only the perception of the moving now is affected), I ' l l not pretend to have devised a rigorous retro-phenomenology for Bradley. Nevertheless, my tale will 68 suffice to press home an essential point about the role of sensation and information-bearing signals in the expression of rational behaviour. As far as his physiology is concerned, Bradley's backward story is starting to become familiar. His body absorbs heat and moisture day and night. His muscles are factories in which energy-rich organic molecules are manufactured. He visits the men's room periodically in order to assimilate a depleted matrix on which to accrete biosynthetic products, which is subsequently disgorged, assembled into some low macroentropic form or another in his mouth, and deposited on his plate. A time-reversed tale of Bradley's corresponding mental activity, however, is less obvious. Bradley habitually holds a newspaper in front of him while he disgorges. At such times, most of his foresight of world events for the next day (the next one lower down, that is) drains out through his eyes. While Bradley's foresight of "future" events (for him) becomes increasingly vivid as time goes by, knowledge of "past" events nevertheless rapidly and almost totally evaporates from his brain. As his brain systematically jettisons its fund of information, it sends signals that excite his eyes and ears. Bradley's retinas emit streams of light, and the bones in his ears emit little vibrations, that each issue forth into the world in time to participate in larger convergent wave fronts. His eyes and ears are organs whose function is to eject his knowledge into the world, until at last he has no more. But after the newspaper has been set down, the world events that will take place the day below tomorrow are still foreseen in detail and accepted, as are the events of days even further removed in the future, albeit with diminishing clarity. In due course, each such article of foreknowledge will be effaced at the disgorging table. 69 Bradley's downward-oriented future holds no surprises for him. Even i f he can't foresee exactly when and into which pen the ink stain will retreat from his shirt, he can foresee clearly enough that it won't be there a day from now. His prescience informs him that while he will have to put up with a few more traumas before he is ultimately absorbed by his mother, he will be certain to find the stick that will instantly mend the bruise that gradually worsens on his backside, and the dog that will heal the tooth-shaped sore that will eventually develop on his arm. What threat to his survival could the future possibly hold, since he knows that each and every living thing that will ever assemble on the planet (except possibly for the very last) will live long enough to be absorbed by its successors? Ah, but the past... a realm whose very amorphousness is a source of both wonder and fear. Bradley is always busy jettisoning some backward-looking question or another from his mind. Had he ever been married? Had his wife ever managed to absorb children? Probably in days gone by he had once known the answers to these questions, but that knowledge is forever denied him now. Had he arisen from the earth decades ago, or had he come alive only yesterday, emerging from an uncrumpling car? Bradley wil l never find answers to his backward-looking questions. Just as well that his inquiries systematically evaporate from his stream of consciousness. It never occurs to Bradley to put an end to his habitual knowledge-destroying behaviour. The future always fulfills the vivid, richly detailed prophecies horded in his head with such fidelity that it would be absurd to rail against it. Indeed, he carries a certificate in his wallet which bears the date of his own absorption. And yet the past is an unrelenting source of anxiety. 70 In a world in which houses and cars and all manner of wares are observed to rally from their dilapidated states with the passage of time until the day they attain faultless condition and are promptly dismantled, Bradley's mental life is a story of relentless unraveling. Though he has not the faintest recollection of his birth, in truth he assembled under the earth several decades ago. He was exhumed as a wrinkled gray man with a long life before him. He and his wife, who was due to be exhumed three years later, would live to absorb genomes from five children. His biological imperative was fulfilled admirably. But now, there is no knowing that, though Bradley foresees that his future is filled with anxious past-directed questions. The whole of his life is spent waiting complacently for his prophecies to be fulfilled. His life is not about making plans and fulfilling them. His life is about undoing everything that Ashley would recognize as a worthwhile accomplishment, and bearing passive witness as his plans unravel in his head, until even the germs of his ideas are gone. If Bradley were indeed Ashley's perfect duplicate, except only for the fact that Bradley perceives the present to move from the top down instead of from the bottom up, then it would look to us as i f Bradley's life were bereft of action, or reaction, in any recognizable form. He simply drifts quietly into the arms of his vividly anticipated fate. For when does a coherent signal ever emerge from his past, on the basis of which he might plan and execute a rational action? He only participates, in a small way, in transmitting coherent signals into his future by agency of his eyes and ears and certain other organs, squandering his spontaneously amassed wealth of knowledge in the process. 71 As earlier discussed, the key to sustaining the biological organization of a living being across the longest possible temporal extent is a built-in disposition to react to its environment. However, a disposition to simply flail about randomly in the event of a stimulus is unlikely to promote long-term success. A n organism's repertoire of reactions, in order to have a high probability of securing salutary results, must be appropriate to the condition of the world around it. The condition of the world must therefore have some causal impact upon the behavioural strategies of the organism in question. And for this to be possible, the organism must come equipped with some means by which to gather needed intelligence from the external world. The world communicates its condition from one region to another by means of coherent signals, such as the bottom-up expanding wave fronts of infra-red and visible radiation that are a reliable symptom of the presence of a fire, and the bottom-up expanding compression wave that is a reliable symptom of the snapping of a twig. The organism must come equipped to intercept certain of these signals, and to respond to them appropriately. Our cosmic neighborhood, as discussed in chapter 4, is replete with de facto irreversible processes. Highly correlated events are virtually always associated with a characteristic event at bottom, but no particular characteristic event at top. Such coherent signals as are available to be acted upon occur in a V-shaped pattern in the world block, which means that they are divergent from the bottom up, and convergent from the top down. But top-down convergent signals are useless for the purposes of informing the actions of reactive organisms in our part of the world. Eyes and ears and the like do not receive the convergent signal in the top-down scenario; they merely participate in its transmission from the top down. The organism does not acquire information by means of 72 this process, but loses it. That Bradley possesses no sensory organs with which to receive any sort of top-down signals is unsurprising from an evolutionary standpoint, for since the coherent signals traveling from the top down almost always converge on their destinations (and don't diverge from their sources), Bradley is fated to be a transmitter instead of a receiver with respect to them. His life is spent ejecting his spontaneously amassed knowledge out into the world. The notion of "avoidance behaviour" makes no sense when we search for it from the top down. Rather, it is the class of bottom-up divergent signals that provides reliable information in our fork-asymmetric neighborhood. We know that the informative state Sk of any recording system only occurs above the characteristic event of which Sk is a reliable symptom. In like fashion, the informative mental state Sk is itself the characteristic vertex of some upwardly oriented forks. Some behaviours situated higher up the organism's world-line are reliable symptoms of Sk, in that they are very unlikely to occur without Sk also occurring below them. However, behaviours lower down tend not to have a strong correlation with Sk. Therefore, only behaviours occurring above Sk have any strong likelihood of being informed by that mental state. Given also that (1) as a simple point of logic, a stimulus cannot be reacted to "prior to" (in one sense or the other) its being received by the agent, and that (2) a behaviour, in order to be properly deemed a "reaction", must be informed by the stimulus to which it is (allegedly) a reaction, it follows that our reaction to a given stimulus is as a rule located above the point at which the stimulus is received. That is, the reaction occurs in Ashley's future, and in Bradley's past. It may seem at first glance that avoidance behaviour makes as much sense from the top down as it does from the bottom up, except for the fact that from the top down, it 73 looks like pursuit behaviour. This resemblance, however, is merely specious, reinforced by the natural temptation to think of top-down behaviour as it would be revealed to us by backwards-running movie projector. Actions located below the informative state Sk are not informed by Sk, not in the crucial way that certain actions above are, and so cannot properly be considered as reactions to any "earlier" (speaking from the top down) informative state. Reading from the bottom up, our lives are spent reacting to coherent signals emerging from our near and not-so-near past. The past provides excellent evidence of its contents. Our ongoing reactions to stimuli are informed by subsequent stimuli, and may be adjusted accordingly in mid-flight, as it were. To expend energy in defiance of the superior evidence the past provides is manifestly counter-evolutionary. If an organism were not innately disposed to accept any one source of information as reliable, i f it did not base its behaviour upon the perception of informative states in the world that are reliable symptoms of characteristic events located farther down time's axis, then its behaviours would cease to be responsive to the condition of the world without, and its repertoire of strategies would perforce be limited to random flailing. But random flailing contributes nothing to the maximization of the organism's longevity. If the coherent divergent signals emerging from the past will not serve as a reliable source of information, what could? Moreover, and perhaps more importantly, the past is virtually empty of threats to our bottom-up fecundity. As mentioned earlier, there is only one de facto inevitable demise that awaits us in that direction: absorption by a successor. Speaking from an evolutionary standpoint, we can afford to be complacent about the inevitable and universal terminus 74 below. However, the informative states resident in a typical brain provide poor evidence of the future's contents. For all the organism can know at any point in time (given the severe curbs placed upon its ability to "pre-record" events higher up), its bio-organization will discontinue the next day, or the day after, or years from now. On my view, whether or not an organism will terminate on a certain day is a determinate tenseless fact that holds true eternally. However, at most earlier times, the same fact is epistemologically /^determinate. That is, there is usually no way of knowing in advance whether this contingent fact is true or false. We cannot afford to be complacent about the future as we can about the past, for our brains do not reliably accumulate traces of its contents. In order for an organism to maximize its longevity with reliability, it must be hard-wired in such a way that the reception of coherent signals at one person stage will result, quite mechanically at root, in certain appropriate modifications to later person stages piled thereon. Subsequent signals will inform further modifications higher up, such that the organism's space-time worm, informed of the near past at most every stage of its life by divergent bottom-up signals, may be observed to bend around circumstances which, had they not been avoided, would not likely have been salutary to the continuation of its bio-organization. A l l other conditions being the same, the same organism minus its inbuilt avoidance mechanisms will tend to have a far shorter life-span than its fully-equipped counterpart. This more or less constant readiness to react to unforeseen signals with avoidance behaviour i f the signals express the probability of danger, or seeking behaviour i f the signals express the probability of gain, lies at the root of our future-oriented anxiety and aspiration. At most any stage of our life, we are disposed to react in 75 any of a vast number of ways to unexpected information in order to secure future ends, but never to secure past ends. I say again that on my view, future events are just as determinate as past or present events. There are those who would argue that the determinacy of future events is incompatible with freedom of action. But avoidance and/or seeking behaviour, and freedom of action in general, isn't about changing the future. What will happen at midnight tomorrow simply is whatever will happen at midnight tomorrow. There can be no question of the class of "midnight-tomorrow events" being first one thing and then another, as would have to be the case i f we want to claim that the class of midnight-tomorrow events is something that can be intelligibly said to change. For example, it makes no sense to suppose that numbered among the midnight-tomorrow events is a nuclear apocalypse, right up until I shoot the madman who would otherwise have pushed the button, at which point midnight-tomorrow becomes characterized by peace on earth. Suppose that I shoot the madman at noon tomorrow. Do I mean to say that the class pf midnight-tomorrow events changes at noon-tomorrow? If so, we arrive at the absurd claim that the events located at one moment of time "change" at some other period of time. This claim is rather different than the account of change offered earlier, namely that it consists in nothing more than this: i f the facts that describe one stage of an object located at one point in time are not the same as the facts that describe another stage of that same object located at another point in time, then the object may be said to have changed, from one time to the next. It is things that change over time, not events that 76 change at a time . There is only ever one (set of) thing(s) that occurs at midnight tomorrow.19 Rather than having an ability to change the future, I submit, we desire that the form of future events be at least to some degree causally responsible to the form of our present actions. In other words, we want to be able to say truly that the future will turn out the way it does at least partly because of the way we act in the here and now. The determinacy of future events in no way interferes with or cheapens the fulfillment of this desire. To the extent that the universe is constrained to hang together in accordance with the laws of nature, statistical mechanics or what have you, then it is legitimate to assert that later events happen as they do because of earlier events, regardless of the temporal direction in which we want to run our story. But since in our cosmic neighborhood many ordered states and correlated events are associated with characteristic events at bottom but no particular characteristic events at top, the two sorts of causal stories that would seem to be the most natural to tell would be one of efficient causation from the bottom up, or one of final or teleological causation from the top down. And given the deeply, biologically embedded disposition to value the events surrounding the upper reaches of our worldlines over those located farther down, the bottom-up variant is naturally to be preferred. 1 8 McTaggart's (1908) failure to recognize this distinction undermines his attempt to secure the conclusion that time is unreal. See Horwich's treatment of this fallacy in AT, pp. 20-25 1 9 So-called tree models of reality - in which only past and present events exist, and future ones do not - is alternative against the cogency of which I will not have an opportunity to argue here. I will point out simply that the key assumption of tree-model advocates is that there is no totality of facts that describes the world from a tenseless viewpoint, and express my opinion that such a metaphysical stance seems bizarre and ad hoc, unless motivated by considerations other than a desire to make room for a "moving now". 77 7. Bringing it All Together In chapter 3, I claimed that a fully satisfactory account of the direction of time's apparent flow must explain why the sense of "earlier" captured by "earlier relative to anticipation-memory" is to be non-arbitrarily preferred over the sense captured by "earlier relative to memory-anticipation". I believe that I have accomplished that end. We have that: ( 1 ) A typical person's mental stages are characterized by various ordered informative states. These states are associated with a characteristic interaction at bottom, but not necessarily at top. (2) From a bottom-up perspective, sense organs receive information in the form of coherent divergent signals traveling up from characteristic antecedents below. The informative mental state Sk is a reliable symptom of the event, Ck. (3) From a top-down perspective, however, sense organs expel information by participating in the transmission of coherent signals that subsequently converge on characteristic events below. The informative mental state Sk emerges not from the reception of a coherent signal, but from the bombardment of implicitly ordered top-down cosmic noise. (4) Avoidance behaviour, in order to substantially enhance the probability of longevity, cannot be random; it must be informed by the condition of the world without. Or, more generally, a behaviour cannot properly be considered reactive without there being a "prior" (from one perspective or the other) condition that informs the behaviour. 78 ( 5 ) The informative mental state Sk is itself the characteristic vertex of some upwardly oriented forks. Some behaviours higher up are reliable symptoms of Sk, in that they are very unlikely to occur without Sk also occurring below them. However, behaviours lower down tend not to have a strong correlation with Sk. One condition cannot be said to be "informed" by another i f there is not some probabilistic correlation between the two. Therefore, only behaviours occurring above Sk have any strong likelihood of being informed by that mental state. (6) In general, a behavioural sequence may be read from either the bottom up or the top down. However, the correlation that holds between informative mental states and behaviours higher up, together with the absence of a correlation between informative mental states and behaviours lower down, requires that in order for a behavioural sequence to be interpreted as reactive, it must be read only from the bottom up. Hence the upward orientation of our actions and reactions. (7) To the extent that emotional attitudes may be reduced to sets of behavioural dispositions, then our future-directed anxiety may be understood in terms of our around-the-clock readiness to engage in future-directed reactive behaviour, based on information received from characteristic antecedent conditions. It seems apparent that upward-directed anxiety goes a long way toward fulfilling the biological imperative. One might imagine, not unreasonably I think, that much of our psychological structure is but an adornment upon the contours of brute biological imperatives, and shapes itself around their fulfillment. The anxiety that may be expected to afflict the upward-reading organism is whether it will live long enough to fulfill the 7 9 imperative, and the corresponding anxiety that may be expected to afflict the downward-reading organism is whether it has lived long enough. But such speculation is not essential to the point I wish to make. Psychology aside, a population's prospects for long-term success tend to be fortified when its members' average extension in time is maximized (or, more precisely, when the average temporal extent of its members' reproductive phase is maximized). The asymmetric arrangement of material events in an otherwise isotropic spacetime is such that threats to an organism's longevity converge upon its person from the direction of the past, not the reverse. I have attempted to show why, i f the stages of human mental machinery are hard-wired with upward-directed anxiety and downward-directed complacency, the population's longevity amid a fork-asymmetric arrangement of events will tend to be maximized. Built-in bottom-up anxiety and avoidance behaviour makes it possible for a longer spacetime worm to "fit" within a matrix of overall natural law, unlike the organism that is built in such a way that it does not tend to veer around potentially life-terminating circumstances. Given that (1) a string of temporally-projected phenomena may be perceived as an enduring entity viewed from different temporal perspectives, (2) the change of temporal perspective as a function of time may be perceived as 'motion', of sorts, (3) reactive behaviour may only be properly understood to operate from the bottom-up, and (4) the upward orientation of reactive behaviour suffices to establish our bottom-up bias, it is to be expected that "the present" should be perceived to move from the bottom up - that is, in the direction that coincides with the sense of "earlier" captured by "earlier relative to anticipation-memory" (i.e. a scheme in which p is anticipated before it is remembered), rather than the reverse. The 'moving now', then, is an artifact of the sort of hard-wiring 80 that equips an organism for longevity in our sort of asymmetric environment. Time is perceived to flow toward that temporal pole about which we have been designed to be anxious. 81 8. The Trouble With Causation I have striven to plumb the puzzle of our lopsided temporal attitudes - and their contribution to our perception of a "moving now" - right down to the underlying physics and evolutionary biology. Many other investigators, however, have opted to approach the issue of value-asymmetry and related subjects at a more analytic level. It is not uncommon among those interested in prying at the tangle of our time-related attitudes to approach the problem from the perspective of a "rational agent". Questions are often posed in the form: what is the rational basis for such-and-such an action? Or, what is the rational basis for holding such-and-such an attitude? For example, what is the rational basis for acting for the sake of the future, as opposed to the past? Similarly, is it rational to fear (one's own) death? If it is, then why not also fear the expanse of time that yawns at the opposite terminus of one's existence: namely, birth? One's life, after all, is an uncomfortably succinct string of events, pressed between two vast tracts of time in which we do not exist. Is there a relevant difference between the past and future directions, such that there are rational grounds for valuing the contents of one realm over those of the other? Writers who have approached value-asymmetry at the rational level include Parfit (1984), Nagel (1979), Brueckner and Fischer (1986), McMahan (1988), Glannon (1994) and, millennia hence, Epicurus and Lucretius. I wish to turn my attention for a moment upon Walter Glannon's (1994) article, "Temporal Asymmetry, Life, and Death", for in it I find an example of an argument built upon an insufficiently considered foundation of physics and evolutionary biology. In particular, the argument relies on the concept of 82 causation in a way that suggests it is a well understood and straightforward notion. On the contrary: I will argue that causation is a concept that is either vitiated by a stipulative analytic element, or rooted in an inconveniently nebulous cluster of beliefs. Too often, causation is viewed as a basic and obvious feature of our world, and it is invoked to explain phenomena which, more appropriately, ought to be invoked to explain causation. Where Horwich has been admirably assiduous, other writers have been, I think, remiss. The focus of Glannon's paper is a comparison of the rational grounds for concern about post-mortem (or future) non-existence versus pre-natal (or past) non-existence. He argues for a surprising conclusion: that past non-existence, and not future non-existence, is a fitting object of rational concern. His argument proceeds as follows: (1) It is rational for one to be concerned now about something relevantly like what one is not concerned about only if one justifiably believes that the two are relevantly different. (2) Past nonexistence is relevantly different from future non-existence insofar as states of affairs that obtain during the former period can affect persons adversely in respects in which states of affairs obtaining during the latter period cannot, (asymmetry thesis) (3) One is justified in believing that past non-existence is relevantly different from future non-existence. (4) It is rational to be concerned now about what can actually affect a person adversely. (5) Therefore, although it is not rational for one to be concerned now about one's future non-existence, it may be rational for one to be concerned about one's past non-existence. (1994, p. 242) Glannon's argument hinges on the unidirectionality of causation. Expanding on premise (2), we have that on the one hand, persons may experience the temporally remote effects caused by states of affairs that obtain prior to their conception and birth, while on the other hand, states of affairs that might obtain in the remote future after our death cannot affect us in the course of our lives. This, in virtue of the fact that causes precede (or are simultaneous with) their effects; but effects cannot precede their causes. Since effects can occur both proximally to and remotely from their causes, and since causes precede their effects, the past can produce effects in the present and the future. By 83 contrast, the future cannot produce effects on our lives in the present, because this would involve backward causation with effects at the macrophysical level, a circumstance that Glannon dismisses as a dubious hypothesis at best. The unidirectionality of causation, as it is invoked here, is an a priori, analytic feature of the concept, a feature that receives no elaboration in an empirical light. In fact, Glannon denies himself access to empirical resources when, earlier in his discussion, he embraces the view that the proverbial "arrow of time" is at bottom epiphenomenal to, or a by-product of, our psychology. Once we have ceased to exist as persons, he maintains, "any putative temporal asymmetry between past and future dissolves, since the asymmetry itself is a function of a person's psychology" (1994, p. 238). The problem with this view, however, is that i f there is nothing more to temporal asymmetry than a psychological illusion, then what remains to account for the asymmetry of causation? Not only does Glannon's stance rule out straightforward temporal anisotropy (that is, the metaphysical distinctness of past and future) as a basis for causal asymmetry, but it also appears to attach no importance to any other condition extrinsic to a person's psychology, including the de facto irreversibility that occurs in a region of space-time bounded, at bottom, by a condition of micro-chaos. Once both metaphysical and circumstantial asymmetry are rejected (or ignored), no substantial empirical basis for causal unidirectionality remains, and we are left with a "truth by definition" or "truth by convention", against the possibility of which Quine (1951) has persuasively argued. For, insofar as causation is a kind of determination, it is time-symmetric, because all the candidate relations of determination - laws of nature, necessity, contiguity, probabilistic 8 4 connection, and so on - are themselves time-symmetric. Therefore, without appeal to empirical resources, the time order of causation must be added as an a priori accessory. Glannon may well amend his argument to the effect that psychological asymmetry should itself be properly understood as a function of de facto irreversibility (or some such), and thus supply himself with a substantial empirical foundation on which to ground an a posteriori account of causal asymmetry. Yet Glannon seems disinclined to do so. Observing that most of us care more about experiencing future pleasure and avoiding future pain than about past pleasure and past pain that we have already experienced or avoided, he seeks to explain this fact with the further observation that "we are experiencing subjects who generally anticipate what will occur in our lived future with greater concern than the concern we have when we reflect on what has occurred in our lived past" (1994, p. 238). This observation, however, succeeds only in restating the fact that at least psychologically, the past and future are distinct, and fails to explain the phenomenon in terms of whatever underlying conditions may be responsible for it. Be that as it may, even i f a satisfactory empirical account of psychological asymmetry were supplied, there would still remain the problem that in an isotropic space-time, the meanings of the terms "earlier" and "later" are established relative to ordered sequences of events in space-time. Horwich has shown how the terms earlier and later may be reduced to "earlier relative to X T ' and "earlier relative to YX\ such that it becomes possible to express the laws of nature without using a fundamental, asymmetric temporal relation. Since the past and future directions are metaphysically indistinct, we have no metaphysical grounds to select one or the other as the earlier direction. And, without a whole-hearted plunge into the nebulae of beliefs constitutive of our notions 85 about causation (beliefs fortified by both the observation of probabilistic connections, and our subjective experience of deliberation, knowledge and control), we have no non-stipulative grounds for preferring what we familiarly know as the future as the direction into which causal relations run. Still, Glannon might simply adopt the stance that it so happens that in our part of the universe, the great majority of causal relations run from the past to the future, rather than the reverse. In this case, there is still a stipulative ingredient present in the concept of causation, but the door is open to provide an a posteriori basis for it. Glannon himself need not take long strides through that door, citing instead the almost universal agreement that causes always or almost always precede their effects from the bottom-up. It would seem that the burden of proof falls to whoever wishes to suggest otherwise. Yet it is here that I would voice a methodological objection, to the effect that when asymmetric temporal attitudes are at issue, it is a mistake to place causal notions at (or even near) the base of our explanatory towers. Like Horwich, I maintain that our belief in the unidirectionality of causation is derivative from various sources both objective and psychological. On the objective front, our concept of causation is derivative from our observation of (time-reversible) probabilistic connections. On the psychological front, the concept is derivative from our subjective experience of (de facto irreversible) deliberation, knowledge, and control. But probabilistic connections do not suffice in themselves to fix the direction of causation, for (as mentioned at the end of Chapter 6), it would still remain open to us to tell a story of efficient causation from the bottom up, or one of final or teleological causation from the top down. And insofar as our belief in the future-orientation of causation is a product of the brute asymmetry of our experience 86 (specifically, our time-asymmetric experience of deliberation, knowledge, and our sense of "control"), then we gain nothing by turning around and invoking the concept of unidirectional causality in order to illuminate the facets of our time-asymmetric psychology. Causal decision theorists, such as L . J. Savage (1972), Robert Stalnaker (1980), Allan Gibbard and Bi l l Harper (1978), Nancy Cartwright (1979), Brian Skyrms (1980), and David Lewis (1981), seek to build on our intuition that we act for the sake of the future - and that we care more about the future - simply because our actions can effect only those events that they precede (or are simultaneous with). In other words, we act for the sake of the future because we act for the sake of events that we can cause, and those can only be in the future, in virtue of the very nature of causation. This way of putting the matter tends to receive a warm endorsement from our intuitions, and as such it seems to be an eminently plausible place to begin. However, the plausibility is merely specious. Again, the unidirectionality of causation does not emerge from time-symmetric deterministic relations alone, be they laws of nature, necessity, contiguity, or probabilistic connection. We may supply the concept with its direction in one of two ways. Firstly, we may simply stipulate that causation runs from the bottom-up. But this tactic is unsatisfactory, for we end up with the unilluminating fiat: "we act for the sake of the future because we act for the sake of events that we can cause, and those can only be in the future, by definition.'''' The explanatory potency of truths by definition is wanting, as Quine has shown. Or secondly, we may root the direction of causation in our psychological biases. A more fruitful move as far as advancing our understanding of causation goes, but one that disqualifies the concept from playing a fundamental 87 explanatory role in decision theory. For, we now have that: "we act for the sake of the future because we act for the sake of events that we can cause, and those can only be in the future, because it seems to us that our experience of deliberation and control is directed into the future.''' In other words, we act for the sake of the future because it seems to us that the future is the only direction for the sake of which we can act. This is a singularly unhelpful observation. In addition to the questions I have already raised, one may well also ask: (1) Why must it be that the desirable states for the sake of which we are acting are the effects of our actions and not the causes of our actions? So long as the desirable states for the sake of which we are acting actually and reliably occur, what does it matter that the arrow of causation points from the act to the desirable state, or from the desirable state to the act? ( 2 ) Similarly, i f it should be that two events are probabilistically connected, what makes one of them the "cause" and the other the "effect"? No mere probabilistic connection nor time-symmetric body of natural law will suffice to settle these questions. What remains is appeal to psychological factors. But, i f the direction of causation is cashed out as a function of psychological biases, then psychological biases cannot then be cashed out as a function of unidirectional causation. What remains for the illumination of our time-asymmetric attitudes are bare probabilistic connections, stripped of causal clutter. It is for this reason that when discussing relations between events, I have striven to speak in terms of probabilistic connections alone, and I present no one class of events as "ontologically more basic" than any other. Only in this 88 way, I believe, may an account of time's compelling and enigmatic "feel" be truly explanatory. 89 Bibliography Arntzenius, F.; 1990, "Physics and Common Causes". Synthese, 82, pp. 77-96 Austin, J. L.; 1961, Philosophical Papers. Oxford: Oxford University Press. Bigelow, J.; 1991, "Worlds Enough for Time", Nous, 25(1), pp. 1-19, Mr 91. Boltzmann, L.; 1898, Lectures on Gas Theory. Translated by G. S. Brush. Berkeley: University of California Press, 1964. Carnap, R.; 1928, Der logische Aujbau der Welt, 2nd edition, Hamburg: Felix Meiner Verlag, 1961 Carrwright, N. ; 1979, "Causal Laws and Effective Strategies." Nous 419-438. Reprinted in How the Laws of Physics Lie. Oxford: Clarendon Press, 1983. Dennett, D.; 1991, Consciousness Explained. Boston, Mass.: Little, Brown and Company. Dretske, F.; 1981, Knowledge and the Flow of Information. Cambridge, Mass.: The MIT Press. Friedman, W. J.; 1990, About Time: Inventing the Fourth Dimension. Cambridge, Mass.: The MIT Press. A Bradford Book. Gibbard, A., and Harper, W.; 1978. "Counterfactuals and Two Kinds of Expected Unity." In Foundations and Applications of Decision Theory. Edited by C. Hooker et al. Western Ontario Series in the Philosophy of Science. Vol. 13. Dordecht: Reidel. Reprinted in Ifs. Edited by W. L. Harper et al. Western Ontario Series in the Philosophy of Science.Vol. 15. Dordecht: Reidel, 1980. Glannon, W., 1994, "Temporal Asymmetry, Life, and Death". American Philosophical Quarterly, Jul 01 1994 v 31 n 3 Gold, T.; 1962, "The Arrow of Time," American Journal of Physics, 30, 403-410. Goodman, N.; 1951, The Structure of Appearance. 2nd edition, New York: The Bobbs-Merril Company, Inc., 1966 Grunbaum, A.; 1962, Philosophical Problems of Space and Time. New York: Knopf: 2nd edition, Dordrecht: Reidel, 1973 Horwich, P.; 1987, Asymmetries in Time. Cambridge, Mass.: The MIT Press. A Bradford Book. Husserl, E.; 1928, The Phenomenology of Internal Time-Consciousness. Translated by J. S. Churchill. Bloomington: University of Indiana Press. Lewis, D.; 1981, "Causal Decision Theory." Australasian Journal of Philosophy 59: 5-30 ; 1986, On the Plurality of Worlds. New York, NY: Basil Blackwell Inc. McTaggart, J. E.; 1908, "The Unreality of Time." Mind, New Series, No. 68, Oct 1908. Miller, I.; 1984, Husserl, Perception, and the Awareness of Time. Cambridge, Mass.: The MIT Press. A Bradford Book. 90 Parfit, D.; 1985, Reasons and Persons. Oxford: Oxford University Press. Price, H.; 1996, Time's Arrow & Archimedes' Point. Oxford: Oxford University Press. Prior, A.; 1967, Past, Present and Future. Oxford: Oxford University Press. Quine, W.V.O.; 1951, "Two Dogmas of Empiricism." Reprinted in From a Logical Point of View. New York: Harper and Row, 1963. Reichenbach, H.; 1956, The Direction of Time. Berkeley: University of California Press. Russell, B.; 1915, "On the Experience of Time." The Monist, 24, pp. 212-33 Savage, L. J.; 1972, The Foundations of Statistics. 2nd rev. Ed. New York: Dover; 1st, New York, Wiley, 1954. Savitt, S; 1990, "Epistemological Time Asymmetry". PSA, 1990 v 1 pp. 317-324 ; 1991, "Critical Notice: Paul Horwich, Asymmetries in Time". Canadian Journal of Philosophy, Sept 01 1991 v21 pp. 399-417 ; (Forthcoming) "Survey Article: The Direction of Time" Skyrms, B.; 1980. Causal Necessity. New Haven: Yale University Press. Stalnaker, R.; 1980, Letter to David Lewis. In Ifs. Edited by W. L. Harper et al. Dordecht: Reidel. Letter written in 1972. Stein, H; 1968, "On Einstein-Minkowski Space-time". The Journal of Philosophy, 1968 v 65 pp. 5-23 Sklar, L.; 1974, Space, Time and Spacetime. Berkeley: University of California Press. ; "Horwich on the Origin of Entropic Asymmetry" (unpublished). 91 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items