Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Morphological variation in a biotically patchy environment : evidence from a pasture population of Trifolium… Evans, Richard C. 1986

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

Item Metadata


831-UBC_1986_A6_7 E92.pdf [ 4.04MB ]
JSON: 831-1.0096690.json
JSON-LD: 831-1.0096690-ld.json
RDF/XML (Pretty): 831-1.0096690-rdf.xml
RDF/JSON: 831-1.0096690-rdf.json
Turtle: 831-1.0096690-turtle.txt
N-Triples: 831-1.0096690-rdf-ntriples.txt
Original Record: 831-1.0096690-source.json
Full Text

Full Text

M O R P H O L O G I C A L V A R I A T I O N IN A B I O T I C A L L Y P A T C H Y E N V I R O N M E N T : E V I D E N C E F R O M A P A S T U R E P O P U L A T I O N OF TRIFOLIUM REPENS L. by RICHARD C. EVANS A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE F A C U L T Y OF GRADUATE STUDEES Botany Department We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA October, 1986 © Richard C. Evans, 1986 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t of the requirements f o r an advanced degree a t the U n i v e r s i t y o f B r i t i s h Columbia, I agree t h a t the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and study. I f u r t h e r agree t h a t p e r m i s s i o n f o r e x t e n s i v e copying o f t h i s t h e s i s f o r s c h o l a r l y purposes may be granted by the head o f my department o r by h i s or her r e p r e s e n t a t i v e s . I t i s understood t h a t copying or p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be allowed without my w r i t t e n p e r m i s s i o n . _ Botany Department of The U n i v e r s i t y of B r i t i s h Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 15 n r . t n h P r , 1 Qfifi ABSTRACT The relationship between morphological variability and biotic environmental heterogeneity was studied in a pasture population of Trifolium repens. It had been argued that the unexpectedly high levels of variation in T. repens could be maintained by diversifying selection. The mosaic of neighbours (perennial grasses) with which T. repens coexists constitutes a prominent element of biotic patchiness that may lead to sorting among T. repens genotypes on the basis of neighbour-specific compatibilities. A variation study was conducted on a set of 400 individuals of T. repens collected on a neighbour-specific basis from a 43 year old pasture and grown for one season under common garden conditions. A significant proportion of the variation in a set of twelve morphological characters was accounted for by the neighbour with which the individual of T. repens had been growing in the field. The actual amount of variation accounted for, however, was low (6-19%). It was concluded that although diversifying selection could be operating in the pasture, it is not of primary importance in the maintenance of variation in this population. A repeat study was carried out after the plants had spent two years in the common garden. None of the earlier among-neighbour differences in morphology were retained. I concluded that the original results reflected developmental differences carried over from the pasture. ii TABLE OF CONTENTS Abstract ii List of Tables iv Acknowledgements v Preface vi I. Overview 1 A. Variation in Trifolium repens 1 B. Genotype dynamics and variation in clonal populations 2 C. Environmental heterogeneity and diversifying selection 6 D. Environmental heterogeneity in pastures 8 E. The clonal growth habit and environmental heterogeneity 12 F. Diversifying selection in the pasture 15 II. Common garden study, 1982 17 A. Materials and Methods 17 1. The Pasture 17 2. Collection and Propagation of Material 20 3. Assessment of variation 22 B. Analyses and results 25 C. Discussion 37 III. Common garden study, 1984 47 A. Carry-over effects and the common garden method 47 B. Materials and methods 51 C. Analyses and results 53 D. Discussion 60 IV. Diversifying selection and Trifolium repens: a reconsideration 66 References 72 iii List of Tables Table I. Composition of the "Highland Forage Mix" used in seeding a pasture in 1977 18 Table II. List of species present in the study site, a 42 year old pasture in the lower Fraser Valley in British Columbia 19 Table HI. List of characters measured in the 1982 common garden study of variation in a pasture population of Trifolium repens 23 Table IV. Stratification of the sample population of Trifolium repens by original neighbours 25 Table V. Tests of homogeneity of variances for twelve morphological characters of Trifolium repens. Raw and log transformed data from the 1982 common garden study. ... 27 Table VI. Summary of analyses of variance for 12 morphological characters from four neighbour-specific groups of Trifolium repens. Data from the 1982 common garden study 29 Table VII. Variance components for four morphological characters from a pasture population of Trifolium repens. Data from the 1982 common garden study 31 Table VIII. Correlations among 12 morphological characters of Trifolium repens. Data from the 1982 common garden study 32 Table IX. Principal components analysis of 12 morphological characters of Trifolium repens. Data from the 1982 common garden study 35 Table X . Summary of analysis of variance of principal component scores for four neighbour-specific groups of Trifolium repens. Data from 1982 common garden study 35 Table XI. List of characters measured in the 1984 common garden study of variation in a pasture population of Trifolium repens 52 Table XII. Summary of analyses of variance for 12 morphological characters from four neighbour-specific groups of Trifolium repens. Data from the 1984 common garden study 54 Table XIII. Variance components for four morphological characters from a pasture population of Trifolium repens. Data from the 1984 common garden study 55 Table XV . Principal components analysis of 12 morphological characters of Trifolium repens. Data from the 1984 common garden study 58 Table XVI . Summary of analysis of variance of principal component scores for four neighbour-specific groups of Trifolium repens. Data from 1984 common garden study 59 iv A C K N O W L E D G E M E N T S I would like to extend my sincerest thanks to my supervisor, Dr. Roy Turkington for his guidance and enthusiasm. I am also grateful to my commitee members; Dr. Jack Maze (scepticism) and Dr. Judy Meyers (optimism) for playing the antipodes of ecology, and Dr. Fred Ganders for his part in the final acts and for his tolerance of last minute impositions. I wish to acknowledge the counsel and encouragement of my fellow students, notably Lonnie Aarssen, Roberta Parish, Loyal Mehrhoff, Rob Scagel, John Spence, and Gary Armstrong. The financial support provided by Dr. Turkington and by the Botany Department was greatly appreciated. I also wish to thank Mr. Bi l l Chard and Ms. Mary Chard for the access to their pastures. Finally, I am especially grateful to my wife, Janet, for her continued support and perseverance. v PREFACE An important area of emphasis in plant ecology over the last 60 years (since Turesson 1922) has been the study of genetic variation in species and populations (see reviews by Heslop-Harrison 1964; Langlet 1971; Briggs & Walters 1984). Recently, attention has been drawn to the relationships between events at the level of individuals and patterns of variation at the level of populations (Harper 1967, 1982; Dirzo & Sarukhan 1984). Much of this interest has stemmed from the desire to understand evolutionary patterns (especially speciation). The synthetic theory of evolution holds that large scale evolutionary phenomena may be explained by reference to processes affecting individuals in small populations. This postulated connection between population-level events and species-level patterns has never been clearly established, and is currently under debate. There are, however, clear connections between individual-level events and patterns of local (micro-)evolution (Jain & Bradshaw 1966; Bradshaw 1972; Antonovics, 1976, 1984; Hamrick 1982; Turkington & Aarssen 1984). Processes originally proposed to account for among-species patterns (natural selection, genetic drift, migration) can also be applied to the organization and dynamics of genetic variation within populations. Features of populations and communities can often be more convincingly tied to individual-level processes than can features of species. Population-level demography and genetics thus retain their interest for ecologists despite the lack of a definite linkage with species-level evolution. Much of the interest in genetic variation in populations has centered on the vi amount of variation and its maintenance (Lewontin 1974a; Brown 1979; Hamrick, Linhart & Mitton 1979; Ennos 1983). Spieth (1979), for example has referred to the explanation of the high levels of variation within populations as "the central problem in population genetics". A potential contributing factor in the maintenance of variation involves environmental heterogeneity and diversifying selection (Hedrick, Ginevan & Ewing 1976; Ennos 1983). Diversifying selection has drawn attention because it is clearly involved in the fine-scale differentiation of plant populations distributed across sharp environmental transitions such as from heavy metal polluted soil near mines to non-polluted pastures (Jain & Bradshaw 1966; Antonovics, Bradshaw & Turner 1971). Even finer scale differentiation associated with mosaic environments has also been demonstrated (Snaydon & Davies 1972, 1976, 1982; Turkington & Harper 1979c). These results support the suggestion that environmental heterogeneity and diversifying selection might be commonly involved with the maintenance of variation in plant populations. Diversifying selection has been specifically invoked as a possible general explanation for the maintenance of variation in populations of Trifolium repens in permanent pastures (Burdon 1980a). These populations have been of interest because the high levels of variation recorded in them appear to contradict expectations for variation in primarily clonal populations. The work to be reported here investigates the possibility that variation for a set of morphological characters in a pasture population of T. repens is maintained by diversifying selection associated with a mosaic of patches of grass. In the first part of the thesis, the arguments leading to the prediction of maintenance of vii variation by diversifying selection in T. repens are reviewed. The results of a common garden study of the relationship between morphological variation and the distribution of the grasses is then reported. The second part of the thesis investigates the possibility that the patterns of variation detected in the common garden represented developmental effects carried over from the field rather than genetically based variation. Finally, some of the original arguments on diversifying selection in T. repens populations are reconsidered. viii I. OVERVIEW A. VARIATION IN TRIFOLIUM REPENS. Trifolium repens L., white clover, has been of widespread interest to population biologists (Turkington & Aarssen 1984; Turkington 1985). This common component of pastures and lawns has been extensively studied because of its economic importance as a forage species (Lowe 1970; Wilson 1978; Sackville-Hamilton 1980). Many of these studies have considered within-population variation, and have consistently shown that pasture populations of white clover contain genetic or quantitative variation for a wide array of characters (Burdon 1983; Turkington & Burdon 1983). Characters studied have included several polymorphic traits such as leaf marks (Brewbaker 1955; Carnahan et al. 1955; Corkill 1971; Cahn & Harper 1976), cyanogenic glycosides (Corkill 1952; Jones 1966; Angseesing & Angseesing 1973), incompatibility alleles (Atwood 1942), and isozymes (Gliddon & Trathan 1985). Quantitative variation has beeen shown for physiological characters including disease resistance (Burdon 1980a,b), susceptibility to Rhizobium infection (Mytton 1975), and response to nutrients (Snaydon 1962a; Snaydon & Bradshaw 1962b, 1969). Common garden studies have shown variation in morphology for numerous characters, and in phenology and reproduction (Burdon 1980a; Aarssen & Turkington 1985c). Quantitative variation has also been shown for performance-related characters such as growth rate, rate of stolon extension (Burdon and Harper 1980), flower production (Burdon 1980a; Aarssen & Turkington 1985c) and response to growth with different pasture grasses (Turkington & Harper 1979c; Turkington 1979; Gliddon & Trathan 1985). 1 Overview / 2 The demonstration that a particular plant population is highly variable should not be surprising. Reviews of reported levels of genetic (allozyme) variation have shown that plants as a group contain as much or more variation than other broad groups of organisms such as vertebrates or invertebrates (Brown 1979; Hamrick, Linhart & Mitton 1979; Hamrick 1982). From similar coarse-level comparisons it might be predicted that Trifolium repens, in particular, would have highly variable populations. Trifolium repens is a long-lived perennial with a widespread geographical distribution (Burdon 1983; Turkington & Burdon 1983). Its mode of reproduction is outcrossing enforced by self-incompatibility (Atwood 1942). These characters have been consistently associated with the most variable of plant species (Hamrick, Linhart & Mitton 1979; Loveless & Hamrick 1984). Variation in T. repens has drawn attention because, theoretically, we might predict that a clonal perennial living in pastures would have low genetic variability. Therefore, the observed high levels of variation in these populations require either a special explanation or modifications to the prediction. B. GENOTYPE DYNAMICS AND VARIATION IN CLONAL POPULATIONS. One of the characteristic features of pasture communities is that most of the dominant plants have a well developed capacity for vegetative reproduction or cloning. Vegetative reproduction among pasture plants takes two basic forms, tillering as in many tufted grasses (e.g., Lolium perenne and Dactylis glomerata) and stolon or rhizome formation as in creeping herbs and grasses (e.g., T. repens, Ranunculus repens, and Poa compressa). Studies of the population dynamics of these plants have shown that cloning represents the predominant mode of Overview / 3 reproduction. Seedling establishment is rare in the closed pasture vegetation except on disturbances (Harberd 1961a; Turkington et al. 1979; Parish, in Turkington 1985). Generally, the establishment of clonally produced copies (ramets) of existing genetic individuals (genets) vastly exceeds the establishment of new sexually produced genets. These observations lead to a common assumption that the input of new genets has a minor effect on the population biology of clonal perennials such as T. repens in established pastures. Contingent on the validity of this assumption, some predictions about genotype dynamics can be made. In recently established pastures, clones will be small and numerous and clonal diversity will be high. Maintenance of this diversity, as the pasture ages, will be a function of the persistence of members of the original cohort. Although each genet may, in theory, have an indefinite lifespan, in reality, its persistence will depend on the continual establishment of new ramets as existing ramets senesce. Ramets of pasture perennials typically have short lifespans, < 1 year in T. repens (Chapman 1983; Thorhallsdottir 1983; Turkington 1983a) and 1.2-2.1 years in Ranunculus repens (Sarukhan & Harper 1973). Because of this rapid turnover of ramets, the sizes (numbers of ramets) of clones will be subject to rapid fluctuations. Any condition which depresses ramet production within a clone can cause an immediate decline in the size of a clone. Continued low ramet production will eventually lead to death of the clone. In an older pasture, populations of ramets may be regulated at constant densities (Harper 1977). This has been noted for the clonal grassland perennials Ranunculus repens (Sarukhan & Harper 1973) and Prunella vulgaris (Schmid & Overview / 4 Harper 1985). If the entire ramet population is under density regulation, expansion in size of some clones must then be accompanied by declines in others. Any differences among genets in competitive ability, as demonstrated for T. repens by Aarssen (1983), may find expression as differences in relative sizes of clones. Eventually, a few vigorous clonal families might steadily expand at the expense of less competitive lines and come to dominate the population. As less competitive clones decline in size, some will certainly be lost from the population. Such clonal dominance has been predicted on theoretical grounds (Williams 1975), and might be particularly likely in pasture vegetation which is expected to commonly experience density limitation (Donald 1963; Snaydon 1978). Even without competition a certain base rate of extinction might be expected simply from the random failure of genets to replace themselves. A computer simulation based on demographic data for the clonal Ranunculus repens predicted an exponential decline in the number of original genets in the population in the absence of competition (Soane & Watkinson 1979). Whether genets are lost through stochastic or directional processes, if they are not replaced through seedling recruitment, the genotypic diversity of the population will become progressively depleted, eventually consisting of a few very old, very large clones (Harberd 1961a, 1962, 1963). Numerous studies on population dynamics of pasture perennials have documented a steady, often exponential decline in numbers of genets within a cohort (Langer, Ryle & Jewiss 1964; Antonovics 1972; Kays & Harper 1974). Langer, Ryle & Jewiss (1964) observed ramet and genet dynamics in swards of Phleum pratense Overview / 5 and Festuca pratensis for three years after establishment and noted that "..the swards, which initially contained many plants with few tillers, were changed to swards of few plants with many tillers". In addition, some clonal populations have extremely low genotypic diversity. A series of studies on clonal species, including sand dune populations of T. repens, (Harberd 1961a, 1962, 1963) demonstrated population dominance by a small number of widely distributed clones of great age. Similar low diversity clonal populations have been reported in Pteridium (Oinonen 1967) and Spartina (Silander 1979). Comparisons among populations from pastures of different ages suggest that genotype depletion is a continuing process. McNeilly and Roose (1984) found 36-43 different isozyme genotypes of Lolium perenne per 0.25 m 2 in three ten year old pastures but only 5 genotypes per 0.25 m J in a 40 year old pasture. Clearly, clonal populations in some circumstances do exhibit genotypic depletion. Such depletion should be reflected in overall levels of genetically determined variability, including the types of characters which have been studied for T. repens. The observation of high levels of variation in white clover populations, however, has been interpreted as contradictory to the expectation of genotypic depletion (Harper 1978; Burdon 1980a; Biilow-Olsen, Sackville-Hamilton & Hutchings 1984). Studies on clonal diversity in British pastures have found no evidence that clonal depletion is a dominant trend. Cahn and Harper (1976), using the leaf mark polymorphism, found at least 3-4 morphs per 100 sq cm. The frequency distribution of these morphs showed no evidence of dominance by a few high frequency clones. In the same pasture, Gliddon and Trathan (1985) identified 48-50 isozyme phenotypes per square metre. Burdon (1980a) found that each of Overview / 6 50 ramets collected from the same pasture were phenotypically distinct and concluded that each represented a different genet. Aarssen and Turkington (1985c) studied morphological variation in a series of four pastures in British Columbia (aged 4, 23, 42, and 67 years). Levels of variation were again high, although variances for many of their characters declined across the age series for both T. repens and Lolium perenne. These declines in variances were accompanied, however, by parallel declines in means. Coefficients of variation did not follow the pattern. These results can be interpreted, then, as reflecting shifts in mean values rather than declines in actual levels of genotypic diversity. In some pastures at least, the expectation of low variability has not been met. This has prompted the question; what maintains the high levels of variation observed in these highly clonal populations? One popular suggestion involves environmental heterogeneity and natural selection. It was proposed for T. repens by Turkington and Harper (1979c) and Burdon (1980a), and by Hamrick (1982) to explain the generally high levels of variation in plant populations, and by Snaydon (1985) for variation in pasture species. C. ENVIRONMENTAL HETEROGENEITY AND DIVERSIFYING SELECTION. The term natural selection is derived from the observation that in a population some individuals survive while others do not, or some leave more descendants than others. Differential survivorship and/or reproduction (fitness) may be Overview / 7 randomly distributed among individuals or may be nonrandom with respect to attributes of individuals. Natural selection refers to the events (organism-environment interactions) which cause nonrandom patterns in fitness. As a consequence of natural selection, relative frequencies of genetically based variants or types (nucleotide sequences, proteins, character states, etc.) may change, i.e., evolution by natural selection may occur. One theoretically prominent source of natural selection lies in the structure of environments (Levins 1968; Hedrick, Ginevan & Ewing 1976). Theoretically an environment is termed homogeneous when it presents a single "selective regime" with respect to attributes of the individuals living in it. That is, the most successful individuals in a homogeneous environment will be of a single type. If the attributes involved have a genetic basis, the relative frequency of this type will increase. Population-wide variation for the attributes will decline along with the frequency of alternative types. Alternatively, a heterogeneous environment is composed of a variety of "microhabitats" each presenting a distinct selective regime. If the microhabitats are different enough no single type will have the flexibility to be the most successful across the entire range. In this case, a set of different types may persist, each representing the most successful type in one of the microhabitats. Population-wide variation will be maintained at a level reflecting the number of types and the differences among them. The pattern of natural selection associated with heterogeneous environments is termed diversifying (or disruptive) in contrast with stabilizing or directional selection in homogeneous environments. Overview / 8 The prediction of depletion of variation in clonal populations matches the expectation for variation in general in a homogeneous environment. If the pasture environment is homogeneous (or irrelevant) to clovers, the most successful lines would be of a single type or a random sample. Either of these cases leads to the expected loss of clones. Although pastures may appear superficially homogeneous, elements of fine-scale patchiness are easy to identify. Some of these include soils (Snaydon 1985), topography, and vegetation (Turkington & Harper 1979a). If patchiness constitutes environmental heterogeneity, depletion of variation may be counteracted by diversifying selection. Burdon (1980a) specifically proposed that diversifying selection associated with vegetation patchiness maintains the high levels of variation in T. repens populations. D. ENVIRONMENTAL HETEROGENEITY IN PASTURES. The existence of environmental heterogeneity or differentiated microhabitats is basic to any argument that variation is being maintained by diversifying selection. A study of diversifying selection begins with the identification of a pattern of patchiness in the environment followed by the demonstration that the patchiness represents an actual heterogeneity. This requires that performances of individuals in the field show environment-specific responses. There is no doubt that the pasture environment contains elements of diversity. These may be easily recognizable, such as the patterns of distribution of different species, or variations in topography. Other, more subtle elements may be found in the movements of grazing animals, soil types and availabilities of nutrients, or distributions of microorganisms. Because any of these elements might reasonably Overview / 9 be expected to have an impact on the lives and deaths of plants, each can potentially define a set of microhabitats. Variation in the soil environment, for example, has been shown to represent environmental heterogeneity for T. repens (Snaydon 1962 a,b; Snaydon & Bradshaw 1962a). Although abiotic patchiness exists in pastures (e.g., soils), the diversifying selection arguments, as applied to T. repens have been based primarily on biotic heterogeneity. This has been identified specifically as "the identity, age, and size of neighbours (associated plants) and the degree to which they compete for the same resources" (Burdon 1980a). The most prominent of these neighbours are the perennial grasses which, along with T. repens dominate pasture communities in Great Britain and North America. The local composition of a sward changes from dominance by each grass alone through various mixtures of two or more. This mosaic of variation presents an easily recognisable pattern of habitat variation. The dense and continuous nature of pasture vegetation ensures that T. repens plants must exist in close association with the grasses. We might then predict strong selection on T. repens individuals for characters influencing compatibilities (persistence, growth, and reproduction in association with neighbours) with the grasses. In addition, the lives and deaths of plants are influenced differentially by different neighbours, (Donald 1963; Harper 1977). The neighbour mosaic therefore represents a potential source for diversifying selection on T. repens for characters influencing compatibilities with specific species of grasses. Its use to explain the maintenance of variation in Trifolium repens, however, requires that the anthropocentric "easily recognised pattern" translates to an actual heterogeneity in the environments of clovers. Overview / 10 The pasture grasses differ in a number of characters which could influence the compatibilities of T. repens with them. These include characters such as morphology, resource use, and seasonality in growth patterns. In one pasture in British Columbia, for example, the grasses include three tufted, patch-forming species {Lolium perenne L., Holcus lanatus L., and Dactylis glomerata L.) and one stoloniferous, spreading species (Poa compressa L.). Among the tufted species D. glomerata tends to form small dense clumps whereas L. perenne and H. lanatus patches tend to be larger but less exclusive. In addition, the grass patches may differ in a variety of ways which might have indirect effects on clovers. These include attractiveness and palatability to grazers, and differences in habitat for small grazing mammals and invertebrates. Numerous studies have considered aspects of the responses of white clover to pasture grasses (review in Chestnutt & Lowe 1970). Clatworthy (1960, cited in Harper 1977) grew ramets of Trifolium repens in combination with varying densities of L. perenne and Agrostis capillaris. While dry weight, flowering, and stolon production all declined with increasing density of neighbours, there was also a difference in the magnitude of the effect depending on which neighbour was used. Solangaarachchi (1985) showed that branching in T. repens was strongly influenced by the physical structure of grass patches. Stolon branching and adventitious rooting occurred at a lower rate in densely structured patches (H. lanatus and A. capillaris) relative to more loosely structured patches (L. perenne and Cynosurus cristatus). This was attributed both to physical impedence and to changes in light quantity and quality. Overview / 11 Turkington (1983 a,b) studied the demography of leaves and flowers of T. repens grown in plots of several different grass species. In these experiments the age structures of the leaf and flower populations, stolon extension rates, and total dry weight production all responded strongly to the identity of the neighbours; leaf and flower flux rates were less influenced. In a related experiment (Turkington 1983c), dry weight allocations to various plant parts were measured. One of the test individuals showed little response to changes in its neighbour environment. The other showed a differential response in its allocation to inflorescences, stolons, and in its total dry weight production. Differential responses of clover to grass neighbours have also been demonstrated in the field. Turkington et al. (1979) and Turkington and Harper (1979c) transplanted ramets of T. repens into patches of pasture dominated by four different grasses. They found significant differences in dry weight production of ramets among the four neighbour environments. Comparable results from a concurrent greenhouse experiment (Turkington & Harper 1979c) strengthened the conclusion that the response was primarily a neighbour effect rather than a soil effect. In many cases, therefore, and for some response measures, T. repens is responsive to changes in its neighbours. The in situ transplants of Turkington and Harper (1979c) in particular support the interpretation that the mosaic of grass neighbours constitutes a heterogeneous environment for clovers as well as Overview / 12 for ecologists. Several of these studies have also shown that T. repens populations are variable for these "neighbour relationships". Turkington and Harper (1979c), in addition to demonstrating that pasture grasses differ as habitats for clovers, also found that T. repens individuals differed in their responses to the grasses. The results of Turkington (1983c) also demonstrated differences among individuals in responses to neighbours. The identification of a likely environmental heterogeneity plus related environment-specific responses suggests that the potential for diversifying selection does exist. Therefore, some associated potential for maintenance of variation might also be expected. E. THE CLONAL GROWTH HABIT AND ENVIRONMENTAL HETEROGENEITY Additional support for the diversifying selection argument has been drawn from consideration of the growth habit of T. repens. The modular construction, which is typical of plants, finds its extreme expression in stoloniferous herbs such as T. repens. The clover plant is composed of a branched sequence of functional units (rooted nodes or ramets). Each is potentially capable of independent existence and, in fact, will typically become physically independent through the deaths of adjacent ramets or the breakage of stolon connections. While the entire T. repens clone can persist indefinitely, the ramet is transitory, persisting usually for less than one year (mean lifespan about six months, Thorhallsdottir 1983). The size (number of ramets) of a clone will respond rapidly to any Overview / 13 environmental condition which influences the lives and deaths of ramets. In an environment which is generally unfavourable, or in which new ramets are not consistently produced, a clone may go quickly to extinction. As noted earlier, this contributes to the expectation of genotype depletion within a population of clones in a homogeneous environment. In an environment which is heterogeneous for clovers, however, the clonal habit may enhance the potential for diversifying selection. Through the growth of stolons, T. repens individuals can spread horizontally. Clones can migrate rapidly; a single plant started from a 2 cm stolon tip produced more than 100 new stolon branches for a total of over 20 metres of stolon after five months (Evans unpublished). As a consequence of this capacity for movement, different parts of the same clone will commonly become widely separated. During its lifetime, an individual genet may encounter (and reencounter) the entire range of available habitats in a pasture. As an expanding clone encounters a spatially heterogeneous environment, its ramets may respond differently in different microhabitats. A stolon apex which encounters a favourable microhabitat may respond by producing new ramets or by branching. A second apex which encounters a less favourable microhabitat may respond by dying, or by continuing to grow through the area without branching or rooting (Harvey 1979, cited in Newton 1986; Solangaarachchi 1985; Newton 1986). For example, clover stolons which encounter tightly clumped patches of Dactylis glomerata commonly grow through the patch without rooting. Overview / 14 As a result of differential production of ramets in different microhabitats (patches), the distribution of a clone may come to reflect the distribution of suitable patches. In addition, if the clover population is composed of genotypes which differ in their habitat suitabilities, the distributions of different clones may diverge. In a heterogeneous environment then, clovers may be sorted by habitat suitabilities both within and among clones. This process has been referred to in a anthropocentric sense as "tracking". Harper (1977, p766), for example, has written "the growth form of clover allows individual genets to migrate continuously through the sward, tracking local favourable areas". Salzman (1985) has demonstrated tracking of soil types by clones of Ambrosia psilostachya. Plants grown in a gradient of saline to nonsaline soil showed differential rhizome growth into nonsaline soil. The result of this environmental sorting (or tracking) would amount, of course, to diversifying selection and could clearly act to increase the number of genotypes which are likely to persist in a pasture. As discussed earlier, the neighbour mosaic potentially provides the habitat heterogeneity within which such a pattern of responses might be expected. As noted by Burdon (1980a) "...a genet may spread until it encounters a set of biotic factors to which it is less well adapted ... At this point the original successful genet is supplanted by other genets which are better adapted to the different set of conditions. It is through this interaction of biotic forces at the local environment level that the high diversity of clover genotypes is maintained." Overview / 15 F. DIVERSIFYING SELECTION IN THE PASTURE. If such a process is operating at any level other than ecological speculation there should be evidence of it! Environmental sorting of clones on the basis of neighbour relationships and ramet responses should be reflected in the distribution of clones. That is, there should be a relationship between the neighbour compatibilities of T. repens individuals in the field and their physical neighbours. Evidence for such a relationship was found by Turkington and Harper (1979c) using reciprocal transplants. They collected T. repens growing in close physical association with each of four pasture grasses. Ramets of these individuals were grown in combination with all four of the grasses. The test ramets produced significantly more dry weight when grown in combination with the same grass with which they had originally been associated in the field. These results were obtained both from transplants into the pasture and into greenhouse flats. Further evidence of such fine-scale biotic differentiation was obtained by Aarssen and Turkington (1985b). They repeated the experiment with T. repens collected from patches of a single grass species, Lolium perenne. The results were similar, clovers performed best when grown in combination with the same L.perenne individual with which they had originally been associated in the field. Similar results, also using L.perenne and T. repens were obtained by Gliddon and Trathan (1985) working in a pasture in N. Wales. These results suggest that clover genotypes are sorted in the field on the basis of their compatibilities with their neighbours. That is, neighbour-specific diversifying selection does appear to be occurring in these pastures. The results also support the suggestion that such diversifying selection could be counteracting Overview / 16 the genotypic depletion expected for a clonal population. Turkington and Harper (1979c) concluded "The simplest interpretation of the observations is that the genetic diversity of white clover in this old pasture is maintained in part by diversifying selection from the variety of neighbouring grass species". None of these studies have, however, specifically addressed the maintenance of variation for morphological characters such as were studied by Burdon (1980a) and by Aarssen and Turkington (1985b). Neighbour-specific differentiation was demonstrated for only a single character, dry weight production. The work to be reported here will extend these studies by looking for evidence of similar neighbour-specific patterns in the distribution of variation for a set of morphological characters. II. C O M M O N G A R D E N STUDY, 1982 A. MATERIALS AND METHODS 1. The Pasture The material used in this study was collected from a pasture on the farm of Bil l and Mary Chard, 25704 Fraser Highway, Aldergrove, British Columbia (SW 1/4 Sec. 25, Twp. 10). The farm is located in the Fraser Valley in the Coastal Douglas F ir biogeoclimatic zone (Krajina 1965). This farm has been the site of a number of other studies including Aarssen (1983), Aarssen and Turkington (1985a,b,c), Parish (in Turkington 1985). The farm has been managed for dairy production since the beginning of the century. The pasture is grazed intermittently from spring or midsummer until late fall by a herd of 20 to 30 cows. The cows are rarely present between December and April, because of unfavorable weather conditions. Typically, grazing is also delayed until after a hay crop is harvested in midsummer. The pasture receives no fertilizers other than animal excretions, occasionally supplemented with barnyard manure. A conversion to beef cattle in 1983 was not accompanied by any alteration in management. The pasture was first cleared about 1900 and was last ploughed and seeded in 1939. The seed mixture used was composed of 5-10% Trifolium repens, 15-20% Dactylis glomerata, and 70-80% Buckerfield's "Highland" forage mix. Although the exact composition of the Highland forage mix used in 1939 is unavailable, it has 17 Common garden study, 1982 / 18 not changed much over the years (Richardson Seed Company, personal communication) and that used in a recently established pasture (1977) is given in Table I. Table I. Composition of the "Highland Forage M i x " used in seeding a pasture in 1977. Dactylis glomerata 45% Trifolium 20% Lolium perenne 15% Lolium multiflorum 10% Phleum pratense 5% Trifolium repens 2% Ladino clover (T. repens) 3% The pasture community is presently composed of 15 grasses and 17 other herbaceous species (Table II). The most common of these are Trifolium repens, Lolium perenne, Holcus lanatus, Dactylis glomerata, and Poa compressa. Together, these five species constitute approximately 75% of the total vegetation cover. Percentage cover by species varies both seasonally and annually (Aarssen 1983; Parish, unpublished). Common garden study, 1982 / Table II. List of species present in the study site, a 42 year old pasture in lower Fraser Valley in British Columbia. GRASSES NON-GRASSES Agropyron repens (L.) Beauv. Achillea millefolium L. Agrostis alba L. Carex spp. Alopecuris geniculatus L. Cerastium vulgatum L. Alopecuris pratensis L. Cirsium arvense (L.)Scop. Anthoxanthum odoratum L. Hypochoeris radicata L. Dactylis glomerata L. Juncus spp. Festuca pratensis Huds. Medicago lupulina L. Festuca rubra L. Plantago lanceolata L. Glyceria declinata Breb. Plantago major L. Holcus lanatus L. Ranunculus acris L. Lolium multiflorum Lam. Rumex acetosella L. Lolium perenne L. Rumex crispus L. Phleum pratensc L. Rumex obtusifolius L. Poa compressa L. Stellaria media (L.)Vill. Poa trivialis L. Taraxacum officinale Weber Trifolium pratense L. Trifolium repens L. Data of R. Parish (personal communication) Common garden study, 1982 / 20 Al l four of the common grasses are perennial. Vegetative replication in three (L. perenne, H. lanatus, and D. glomerata) is by tillering, and in the other (P. compressa) by stolons. Al l exhibit varying degrees of patchiness, up to nearly 100% local cover. Patches dominated by L. perenne or H. lanatus may exceed one square metre, while D. glomerata and P. compressa patches are usually smaller. D. glomerata patches are the smallest, usually less than .25 m 2 but are very dense clumps. The sizes and locations of patches do not appear to be stable either annually or seasonally. Percentage cover within patches also appears to vary seasonally. 2. Collection and Propagation of Material Material for the study was collected on May 17 and 18, 1982. One hundred patches of each of the grasses (L. perenne, H. lanatus, D. glomerata, and P. compressa) were identified visually. Ideally, a patch consisted of at least 75% cover of one of the four grasses over an area of .5 m 2 . This goal was consistently met for L. perenne, H. lanatus, and P. compressa patches; D. glomerata patches tended to be smaller in area but had very high percent cover of D. glomerata. One ramet of Trifolium repens was collected from each of the 400 patches. This gave a replication rate of one hundred genets per neighbour. Replication was by patches rather than by repeated collections per patch, in order to minimize the problem of genet duplication (Harberd 1961b). Each ramet consisted of a 4 cm section of stolon apex, including the apical bud and the first node with its leaves. Any roots remaining on the node were removed. Only stolons which were actually rooted in the patch were used, and preference Common garden study, 1982 / 21 was given to stolons with several rooted nodes and/or branches within the patch. However, few rooted clover stolons were found within patches of D. glomerata. Where none were available, a stolon growing through the patch and rooted on both sides was chosen. The collection of 400 ramets was propagated in a common garden at the Plant Sciences Field Station on the University of British Columbia campus. Each ramet was planted separately in a six inch pot with "field station soil". Before planting, ramets were treated with Rootone rooting hormone to stimulate root production. The pots were arranged in groups corresponding to the original neighbour of the clovers. The potted ramets were allowed a six week establishment period. In July, 1982 a fresh cutting was taken from each pot and replanted in a second pot in the same manner as before. The second set of pots was given one fertilization with 1 teaspoon per pot of NPK 20-20-20. During the establishment period, 26 out of the original 400 ramets had died. These were replaced with cuttings from a stock of extra ramets which had been collected at the same time as the originals and maintained in the same manner. After a further ten weeks, the second generation of ramets was harvested (Sept, 1982). The entire plants were removed from the pots and soil washed from the roots. Because of the length of time required for taking measurememts (approx. four weeks), the plants were stored in a cold-room while awaiting processing. Toward the end of the processing period, it was noticed that many of the plants had continued to grow, slowly, while in the cold-room. It was not possible to estimate the amount of growth, but there was no reason to believe Common garden study, 1982 / 22 that it was nonrandomly distributed among the four groups of clovers. Therefore, this growth was considered to be part of experimental error and not a contributor to among-groups variation. Each plant was assessed for a set of characters reflecting mostly sizes and numbers of parts. The character set was based on the characters used in the variation studies of Burdon (1980a), and Aarssen and Turkington (1985c) (Table III). 3. Assessment of variation a. Numbers of Parts Each plant was separated into root and shoot material. Shoot material was further subdivided by its primary stolons. These were identified as branches rising directly from the main taproot. A count was made of the number of primary stolons (PSTOL#) and of the total number of stolons (TSTOL#) including secondary and tertiary branches along with the primaries. The count did not include dormant or unelongated buds (less than 10mm). The number of internodes on the longest primary stolon was also counted (INODE#). b. Sizes of Parts The length of the longest primary stolon (PSTOL.L) and the longest secondary stolon derived from it (SSTOL.L) were measured. Three internodes were measured on the longest stolon (INODE.L). The choice of internodes excluded the actively elongating stolon tip and unelongated internodes at the stolon base. Five leaves were selected from each plant. These were taken from the third to fifth nodes below, the apex and from as many different stolons as possible. The length of Common garden study, 1982 / Table III. List of characters measured in the 1982 common garden study of variation in a pasture population of Trifolium repens. CHARACTER Root dry weight Shoot dry weight Total dry weight Primary stolon number Total stolon number Internode number Length of primary stolon Length of secondary stolon Internode length Petiole length Leaflet width Leaflet length Leaf marks: white chevrons red flecks RTWT SHTWT TWT PSTOL# TSTOL# INODE# PSTOL.L SSTOL.L INODE.L PET.L LF .W LF . L L FMK .W LFMK .R UNITS OF MEASUREMENT 0.01 gm 0.01 gm 0.01 gm 5 mm 5 mm 0.5 mm 1 mm 0.5 mm 0.5 mm a,b a,b a,b a,b a,b a,b a.b a = character used in the variation study of Aarssen and Turkington(1985a). b = character used in the variation study of Burdon (1980a). Common garden study, 1982 / 24 each petiole (PET.L) from the top of the stipule attachment to the base of the leaflets, and the length and width of each terminal leaflet (LF.L, LF.W) were measured. The leaves were collected and pressed just prior to harvesting of the plants. Measurements were made on the pressed leaves. For this reason, the three leaf characters were not affected by the growth that occurred during storage in the cold-room. c. Weights Root and shoot material was dried separately for at least three days at 100 C. Root and shoot dry weights (RTWT, SHTWT) were measured directly and total dry weights (TWT) derived from them. d. Leaf marks Each plant was scored for two types of leaf markings, red flecks (LFMK.R) and white chevrons (LFMK.W). These were treated as separate characters because they are genetically independent (Carnahan et al. 1955; Corkill 1971). The red flecks were recorded as present or absent, and the white chevrons classified into types which were known to be genetically distinct (Brewbaker 1955; Carnahan et al. 1955). Because the expression of leaf mark genotypes is often unclear, classification of phenotypes was done independently by three observers on two occasions. Plants for which a consensus could not be reached were classified as "indistinct". Common garden study, 1982 / 25 B. ANALYSES AND RESULTS Statistical analyses of the data were performed using the Michigan Interactive Data Analysis System (MIDAS) (Fox and Guire 1976) on the Ahmdal 5840 computer at the University of British Columbia Computing Center. The data for each character were stratified according to the original neighbour of each clover. The data set thus had four strata corresponding to the four common grasses; Lolium perenne (TLOL), Holcus lanatus, (THOL), Dactylis glomerata (TDAC), and Poa compressa (TPOA). Each stratum had about 100 T. repens individuals (Table IV). Table IV. Stratification of the sample population of Trifolium repens by original neighbours. ORIGINAL NEIGHBOUR STRATUM NUMBER OF INDIVIDUALS Lolium perenne TLOL 96 Holcus lanatus THOL 100 Dactylis glomerata TDAC 93 Poa compressa TPOA 95 The within group (stratum) variances for all characters were highly heterogeneous (Table V). The data were transformed to base ten logarithms (LOG 10) (Sokol Common garden study, 1982 / 26 and Rohlf 1981). Heterogeneity was reduced in most cases and homoscedasity was obtained for five characters (Table V). Analyses of variance (ANOVA) were carried out on either the raw or transformed data as appropriate (by the homogeneity of variances criterion). The test for the presence of neighbour-specific differentiation was carried out using analyses of variance (ANOVA). The stratification (original neighbour) was treated as a random effect (Model II ANOVA),(Sokal and Rohlf 1981). The random effects model is used for the analysis of variation within and among groups when the causes of differences among group means are unclear. The present situation involved quantitative, highly plastic characters. It was not certain that their expression under common garden conditions was a reliable indicator of their expression in the field (Briggs and Walters 1984; Mitchell-Olds and Rutledge 1986). For this reason, comparisons of means would have been inappropriate and possibly misleading. The use of a fixed effects model (Model I) including comparisons among means would be appropriate for the analysis of data from reciprocal transplants in the actual field environment. Any neighbour-specific differentiation in these characters should be reflected in the data as phenotype-environment correlation. Such a correlation would be indicated by the presence of a significant added variance component in the ANOVA (a significant portion of the total variance lying in the among-groups component). In a Model I ANOVA added variance components are detected by testing the among-group mean squares against the error (within-groups) mean squares. In ten out of the twelve characters a significant added variance component was detected Common garden study, 1982 / 27 Table V. Tests of homogeneity of variances for twelve morphological characters of Trifolium repens. Raw and log transformed data from the 1982 common garden study. CHARACTER RTWT SHTWT TWT PSTOL# TSTOL# INODE# PSTOL.L SSTOL.L INODE.L PET.L LF .W LF . L RAW DATA F 16.36 ** 11.02 ** 10.26 ** 6.95 ** 4.95 ** 3.31 * 4.69 ** 4.02 ** 4.56 ** 7.97 ** 5.84 ** 7.18 ** TRANSFORMED DATA (LOG 10) F 11.14 ** 5.56 ** 5.65 ** 3.87 ** 7.01 ** 5.68 ** 4.00 ** 2.38 NS 2.36 NS 1.64 NS 3.68 * 3.78 * Columns give F-statistic and significances for test of homogeneity of variances of four neighbour-specific groups. ** = p<0.01 * = p<0.05 NS = no significant heterogeneity among variances Common garden study, 1982 / 28 (Table VI). This means that for ten characters the population showed neighbour-specific differentiation. For two characters, INODE# and PSTOL.L, there were no added variance components, hence no evidence of differentiation. The proportion of the total variances represented by the added variance components (percent variation among-groups) were calculated for these characters (Table VI). For the ten characters that showed differentiation, the percent variation among-groups ranged from 2.02% (SSTOL#) to 20.19% (PET.L) with a mean of 7.93%. These amounts reflect the amount of variation in the clover population which is "explained" by the type of grass the clover was growing with and are measures of the strength of the biotic differentiation. The data for four of the characters included several measurements for each individual (PET.L, LF.W, LF .L (five leaves per individual) and INODE.L (three per individual)). In these cases, a within-individual mean square can be calculated and used to estimate a variance component representing among-genets (within-neighbours) variance. Mean squares were obtained from an heirarchical model ANOVA using the ANOVAR program and variance components were estimated from them using the method of Falconer (1981). Significant added variance components (among-genets) were detected for all four characters. The % variance among-genets ranged from 45.1 for L F . L to 73.5 for INODE.L (Table VII). The % variances among-neighbours were also estimated from the ANOVAR Common garden study, 1982 / 29 Table VI. Summary of analyses of variance for 12 morphological characters from four neighbour-specific groups of Trifolium repens. Data from the 1982 common garden study. n MS A MS E S J A SIG % VAR(A) RTWT 96.0 0.559 0.096 0.0048 ** 4.80 SHTWT 96.5 0.686 0.117 0.0059 ** 4.81 TWT 95.8 0.645 0.101 0.0057 * * 5.33 PSTOL# 96.5 0.245 0.036 0.0021 * * 5.73 TSTOL# 96.8 1610 442 12.06 * 2.65 INODE# 96.8 14.8 9.07 0.0596 NS 0.65 PSTOL.L 96.8 0.042 0.026 0.0001 NS 0.66 SSTOL.L 94.3 0.119 0.041 0.0008 * 2.02 INODE.L 96.8 0.078 0.023 0.0006 * 2.42 PET.L 98.0 0.461 0.018 0.0045 * * 20.19 LF .W 89.8 0.062 0.005 0.0006 ** 11.75 LF . L 89.8 0.145 0.006 0.0016 ** 19.61 n = mean # of individuals/group MS = among-groups mean square A MS^ = error mean square S 1 = added variance component = ( MS - MS )/n A A E % VAR(A) = % variance among-groups = S 2 /( S 2 + MS ) A A E Sig = Level of significance from F-test for presence of added variance components ** = p<0.01 * = p<0.05 NS = not significant Common garden study, 1982 / 30 results and were comparable with those that had been obtained from the MIDAS program (Table VI). Differences between the two can be attributed to the subsetting of cases necessary to obtain a balanced data set for ANOVAR and to the reduction of the repeated measures to means for the MIDAS analysis. Some of the characters retained significant amounts of heterogeneity of variances even after transformation (Table V). This situation formally violates the assumptions of analysis of variance, although the effects often do not alter the results obtained (Sokal and Rohlf 1981). To check the ANOVA results for these characters, they were reanalyzed using the Kruskal-Wallis test which is independent of means and variances (Sokal and Rohlf 1981). The results of the Kruskal-Wallis test exactly paralleled the ANOVA results (Table VI). In every case in which the ANOVA detected a significant added variance component, the Kruskal-Wallis test indicated a significant difference among groups. Because the leaf mark characters were categorical rather than continuous, they were analyzed using the nonparametric Kruskal-Wallis test. This method analyzes ranked observations, testing for significant differences among-groups in the rankings. Neither of the characters, L FMK .W nor LFMK.R, showed any evidence of neighbour-specific differentiation. Most of the characters in the study involved sizes and numbers of parts. Because it might be reasonably expected that larger plants would also have more stolons, larger leaves, etc, it is likely that the characters vary in parallel. If this were the case, the analyses of these characters would not be independent and Common garden study, 1982 / 31 Table VII. Variance components for four morphological characters from a pasture population of Trifolium repens. Data from the 1982 common garden study. Neighbour Genet Error PET.L 17.0 (**) 50.9 (**) 32.1 LF.W 8.5 (**) 56.6 (**) 27.5 LF . L 14.1 (**) 45.1 (**) 27.5 INODE.L 2.7 (**) 73.5 (**) 23.8 Entries give % variation accounted for by each source of variation and a test for the presence of an added variance component. ** = p<.01 Sources of variation: Neighbour = among four neighbour-specific groups Genet = among-Trifolium repens individuals (within groups) Error = among-measurements (within-individuals) the use of a series of univariate ANOVAs might not be appropriate. Conclusions based on such an analysis would certainly be weaker than if based on analysis of independent characters. The relationships among the morphological characters were investigated by calculating correlation coefficients (Table VIII). The results of the correlation analysis show clearly that the characters are not independent. A l l values are positive, and all except three are significant (p<0.01). Common garden study, 1982 / 32 Table VIII. Correlations among 12 morphological characters of Trifolium repens. Data from the 1982 common garden study. RTWT 1.000 SHTWT 0.832 1.000 TWT 0.919 0.983 1.000 PSTOL# 0.610 0.560 0.598 1.000 TSTOL# 0.742 0.801 0.813 0.662 1.000 INODE# 0.269 0.306 0.306 0.068 0.282 1.000 PSTOL.L 0.432 0.599 0.567 0.167 0.380 0.515 1.000 SSTOL.L 0.545 0.720 0.690 0.273 0.558 0.384 0.825 1.000 INODE.L 0.390 0.583 0.514 0.227 0.319 0.192 0.814 0.710 PET.L 0.523 0.683 0.625 0.249 0.400 0.109 0.532 0.529 LF .W 0.356 0.422 0.417 0.162 0.273 0.067 0.412 0.395 LF . L 0.367 0.448 0.439 0.200 0.307 0.082 0.461 0.440 RT SHT TWT PSTOL TSTOL INODE PSTOL SSTOL WT WT # # # L L INODE.L 1.000 PET.L 0.583 1.000 LF .W 0.428 0.568 1.000 L F . L 0.481 0.581 0.850 1.000 INODE PET.L LF.W LF . L L Table entries are product moment correlation coefficients. Characters with coefficients greater than 0.139 are significantly correlated (p<.01) Coefficients indicating no significant correlation are printed in dark type, n = 373 Common garden study, 1982 / 33 This high level of intercorrelation suggests that pattern in the data set might be best represented by a single character, probably reflecting plant sizes. Analysis of additional characters correlated with the first would be redundant, adding little interpretable information. For example, PSTOL.L and SSTOL.L are correlated at r=0.825 suggesting that these variables are highly redundant. RTWT, SHTWT, and TWT are also highly intercorrelated (r>0.80). TWT is obviously redundant because it is completely determined by RTWT and SHTWT. Analysis of ten highly correlated characters such as these is only a little more informative than an analysis of the same character ten times. A method of dealing with a set of highly correlated characters is to reduce them to a smaller number of uncorrected variables using a multivariate method, principal components analysis (PCA). The new, composite variables, or principal components, are analogous to coordinate axes in multidimensional space. The principal components are derived from the correlation structure among all of the original characters. Each component describes an independent trend of variation across the entire data set. The location of an individual in multivariate space is defined by its position on the principal component axes. Position on each axis is determined by a combination of its scores for all of the original characters. Each character score is weighted by the amount that the character contributes to the trend of variation being described by that axis. The patterns of variation associated with the various PCA axes can sometimes be interpreted by inspection of the character weightings, the amount of variation explained by the axis, and the Common garden study, 1982 / 34 original correlation matrix. The PCA axes represent a set of independent, composite characters. They can, therefore, be analyzed individually in the same manner as were the original ten characters but without the ambiguities that were introduced by the high levels of intercorrelation. A principal components analysis was performed on the data (Table IX). The first component (PCA1) represents the most prominent trend in the data set (54% of the total variation). This axis was influenced in the same direction by all of the original variables. This is typical for an axis describing variation in size. The strong multivariate association among the variables supports the pattern noted from the correlation analysis; these plants are primarily distinguished by size, all of the measured variables increasing or decreasing together. Plant weights and stolon numbers contribute most strongly to the pattern. An ANOVA was performed on P C A l . As was the case for the majority of the original characters, a significant added variance component was detected (Table X). The interpretation based on the earlier ANOVAs is thus supported. When all twelve characters are reduced to one multivariate character representing the most prominent trend in the data, the neighbour-specific differentiation is retained. The percent variance among-groups for P C A l was 9.73%. This result is also comparable to those obtained from the original characters. The second component (PCA2) represents the next most prominent trend in Common garden study, 1982 / 35 Table IX. Principal components analysis of 12 morphological characters of Trifolium repens. Data from the 1982 common garden study. P C A l PCA2 % VARIATION 54.36 14.32 C U M U L A T I V E 54.36 • 68.68 RTWT 0.322 -0.287 SHTWT 0.362 -0.178 TWT 0.363 -0.220 PSTOL# 0.215 -0.426 TSTOL# 0.300 -0.361 INODE# 0.150 -0.053 PSTOL.L 0.303 0.244 SSTOL.L 0.325 0.095 INODE.L 0.281 0.292 PET.L 0.288 0.225 LF .W 0.234 0.404 LF . L 0.248 0.397 AXIS PCA3 PCA4 PCA5 10.97 6.18 4.32 79.65 85.83 90.15 0.106 -0.068 -0.232 0.022 0.041 -0.224 0.050 0.007 -0.235 0.253 0.079 0.627 0.063 -0.093 0.082 -0.598 -0.668 0.030 -0.403 0.121 0.161 -0.277 0.190 0.064 -0.188 0.479 0.276 0.168 0.156 -0.524 0.373 -0.363 0.123 0.344 -0.310 0.208 Table entries are coefficients for each character for the first five principal components (axes). % VARIATION is the amount of variation in the multivariate data set which is explained by each axis. T A B L E X. Summary of analysis of variance of principal component scores for Common garden study, 1982 / 36 four neighbour-specific groups of Trifolium repens. Data from 1982 common garden study. AXIS MS MS S 2 A SIG %VAR(A) PCA 1 PCA 2 PCA 3 PCA 4 PCA 5 66.16 5.67 13.02 0.04 2.13 6.03 1.69 1.22 0.75 0.50 0.650 0.043 0.128 0.018 NS 9.73 2.49 9.46 0.00 3.36 n = mean # of individuals/group MS = among-groups mean square A M S £ = error mean square S 2 = added variance component = ( MS - MS„ )/n A A E % VAR(A) = % variance among-groups = S 2 /( S 2 + MS ) A A E Sig = Level of significance from F-test for presence of added variance components ** = p<0.01 * = p<0.05 NS = not significant Common garden study, 1982 / 37 variation (14.3% of the total variance, for a total of 68.7%). This axis was associated positively with stolon lengths and leaf sizes and negatively with weights and stolon numbers. This suggests that when the influence of plant size is removed, there is a negative relationship between stolon numbers and stolon lengths. Such a pattern is reminiscent of the guerrilla-phalanx dichotomy in stolon structure described by Lovett-Doust (1981) for Ranunculus repens and by Solangaarachchi (1985) for T. repens. Some clovers (guerrillas) tend to have a few, long, unbranched stolons while others (phalanx) have more numerous, shorter, more highly branched stolons with shorter internodes. ANOVA on PCA2 detected a significant (p<.01) added variance component explaining 2.5% of the variation in axis scores (Table X). The third, fourth, and fifth components accounted for 11.0%, 6.2%, and 4.3% of the variance, respectively, for a five axis total of 90.2% (Table IX). There were no obvious patterns in the relationships among the characters and these three axes. Significant (p<.01) added variance components were present on PCA3 and PCA5 with percent variation-among groups of 9.5% (PCA3) and 3.4% (PCA5) (Table X). ANOVA on PCA4 detected no added variance component. Again, the multivariate analysis generally (for four out of five components) supports the interpretations from the analyses of the original characters, that this population is differentiated morphologically on a neighbour-specific basis. C. DISCUSSION The results of . this study show that some morphological variation in this Trifolium repens population is distributed on a neighbour-specific basis. For ten out of twelve characters, a significant (p<.05) added variance component was Common garden study, 1982 / 38 detected when variation was partitioned according to the species of grass with which the Trifolium repens individuals had been growing in the field (Table VI). When the twelve characters were reduced to principle components significant neighbour-specific variation was also detected on two out of the first three axes (Table IX). The results of Turkington and Harper (1979c) and of Aarssen and Turkington (1985c) relating variation in performance (dry weight production) of clovers to neighbour relationships are paralleled by variation in morphological characters. This pattern is consistent with the expectations of Burdon (1980a) that morphological variation was being maintained by neighbour-specific diversifying selection. These results also confirm the findings of Burdon (1980a) and of Aarssen and Turkington (1985c) that there is a lot of variation in pasture populations of T. repens. For the four characters on which repeated measures were taken, there was a highly significant (p<.01) variance component representing differences among individuals. Although variation in this set of morphological characters appears to be influenced by clover-grass interactions, evidence for any direct role of these characters in determining neighbour compatibilities is circumstantial. It is not possible, using this type of nonmanipulative study, to distinguish direct involvement of a character from passive correlation with the actual determining characters. For example, a highly significant (p<.01) proportion of the variation in petiole length (PET.L) was accounted for by the identity of neighbours. This is not sufficient, however, to establish petiole length as a major determinant of neighbour compatibilities. Although petiole length has been related to differential persistence of clover types in swards of different heights and under different Common garden study, 1982 / 39 grazing intensities (Rhodes & Harris 1979; Davies 1973), such an explanation does not seem likely in this pasture. There are no consistent differences among the grass patches in either height or grazing intensity (Parish, personal communication). Given the data from this study it is as likely that petiole length is merely correlated with neighbour compatibility as that it is a determinant of it. Furthermore, because I have no data on the expressions of these characters in the field, I have no grounds for placing ecological interpretations on the common garden measurements. Evidence addressing both of these criticisms could be gained from reciprocal transplant experiments using clovers with a range of character combinations. Because much of the causal basis of neighbour relationships is likely to be physiological (Turkington 1985), demonstrating a direct role for morphological characters would be difficult. Whether these characters are directly or passively involved with neighbour relationships, there is, nevertheless, a neighbour-specific component to variation. That is, there is more variation in this population, for these characters, than would be expected in a population coexisting with a monoculture of one of the four grasses. Variation is probably maintained at a higher level in this biotically heterogeneous pasture than it would be in a similar but more homogeneous pasture. The demonstration that a neighbour-specific component to variation exists answers only part of the question of maintenance of variation. We also want to know how much of the population-wide variation is acounted for by the neighbour-specific pattern. The relative importance of neighbours as a diversifying Common garden study, 1982 / 40 influence should be reflected in the amount of variation accounted for by the classification into neighbour groups (% variation, Table VI). These amounts indicate that neighbour-specific diversifying selection is not playing a major role in the maintenance of morphological variation in this population. Only for the three leaf characters, petiole length (PET.L), leaflet width (LF.W), and leaflet length (LF.L) was more than 6% of the total variation accounted for. The multivariate analysis reflects the same pattern with less than 10% of the variation accounted for by A N O V A of the first principle component scores (Table X). This means that although neighbour-specific diversifying selection may be having an effect the effect is not strong. Over 90% of the variation in the data set remains unexplained. There is only 6-7% more morphological variation in this population than would be expected for a population in a pasture composed of a single grass species. This can be compared with the 45-70% variance explained by differences among-individuals (within-patches). There is much more variation within the patches than there is among them. Thus, this population would be highly variable even if there were no differences among the patches. This conclusion must be qualified, however, because the estimation of variance components from common garden data may be complicated by genotype-environment interactions (Mitchell-Olds & Rutledge 1986). The variance in field-collected data for these quantitative characters would include an environmental component reflecting the effects of growing in different grass patches. The point of the common garden method is to minimize this environmental component of variation and so isolate the genetic component which Common garden study, 1982 / 41 represents the among-neighbours differentiation. The procedure assumes, however, that different environments will have the same effect on each group of clovers. If the differences among groups are not constant among environments (if there is genotype-environment interaction), differentiation will not be reliably estimated from measurements taken in any common environment (Lewontin 1974b; Mitchell-Olds & Rutledge 1986). Even if the common environment is comparable to one of the field environments, the method is restricted to one of "local" analysis (Lewontin 1974b). In contrast to the common garden design, reciprocal transplant experiments make no assumptions about genotype-environment interactions and also surmount the local analysis problem if transplants are carried out across a representative range of environments. For these reasons, reciprocal transplants are the preferred method for investigating variation when genotype-environment interactions are expected. In the case of T. repens, the neighbour relationships detected by Turkington and Harper (1979c) clearly rule out the assumption of no genotype-environment interactions. Because the present study focussed on the same neighbour relationships, a reciprocal transplant experiment would have allowed for stronger conclusions. This study was originally designed around a reciprocal transplant experiment. Transplants were set out in May, 1983 but initial mortality was so high (>80% after two weeks) that the experiment was abandoned. It is not possible to determine what effect interactions between genotypes and the Common garden study, 1982 / 42 common garden might have had on the estimation of variance components in this data set. It does not seem likely, however, that the highly significant neighbour-specific differentiation was entirely a product of genotype-common environment interaction. On the other hand, any portion of the among-groups variance which does not represent additive genetic variance further reduces the already small amount of variation explained by neighbour-specific diversifying selection. Numerous cases of morphological differences across environmental boundaries have been recorded in herbaceous plant populations (e.g., Watson 1969; Linhart 1974; Silander & Antonovics 1979; Scheiner & Goodnight 1984; Antlfinger 1981; Snaydon & Davies 1976, 1982; Seliskar 1985; Silander 1985). In many of these cases transplants have provided evidence of genetic differentiation among sub-populations. By implication, diversifying selection could often be involved in the maintenance of morphological variation. It is also noteworthy that in several of these studies, no evidence for genetic differentiation was found (Scheiner & Goodnight 1984; Antlfinger 1981; Seliskar 1985). The magnitude of the role of diversifying selection in maintenance of variation has seldom been investigated. Of the studies which showed morphological differentiation, only one included an estimate of the amount of variance associated with environmental heterogeneity (Silander & Antonovics 1979; Silander 1985). Ramets of Spariina patens were reciprocally transplanted into adjacent salt marsh, dune, and swale habitats. For 21 morphological characters, the median % variation accounted for by differentiation among habitats was 4.9% (range from 0-34.1%). This compared with 18.5% among genets, and 59.8% within habitats. These results are quite Common garden study, 1982 / 43 comparable to the present results for Trifolium repens. Again, although diversifying selection appears to be commonly operating in plant populations, the extent of its role in maintenance of variation is not clear, and may not be large. The low proportion of variation associated with the neighbour mosaic does not preclude the possibility that additional variation could be explained by selection associated with other heterogeneities. One other element of patchiness in the pasture environment which could constitute a heterogeneity is soil type. Numerous examples of local scale differentiation of plant populations associated with soil heterogeneity have been recorded (Jain & Bradshaw 1966; Antonovics, Bradshaw & Turner 1971; Snaydon & Davies 1976, 1982). The role of soil patchiness as an agent of diversifying selection on T. repens has been reported by Snaydon (1962, 1971). The ecological role of soil patchiness in pastures was played down by Turkington and Harper (1979a), by Burdon (1980a), and Aarssen (1983) who found low variability for soil factors and few correlations between soil type and species distributions. The reciprocal transplant experiments of Turkington and Harper (1979c), however, do show evidence that both vegetation and soils are associated with differentiation in T. ripens. Trifolium repens individuals transplanted into plots from which the grasses had been removed showed site-specific differentiation paralleling the neighbour-specific differentiation of individuals transplanted into undisturbed vegetation. Because the neighbour-specific differentiation remained when soil heterogeneity was removed in a glasshouse experiment, it is not likely that Common garden study, 1982 / 44 the effect was primarily soil based, as suggested by Snaydon (1985). Additional elements of patchiness could be proposed as environmental heterogeneities (e.g., vegetation density, topography, pathogens, invertebrate grazers, and soil microorganisms). As with soils some of these could potentially be associated with diversifying selection on T. repens populations. Considering the low proportion of variation accounted for by such a prominant element of patchiness as the neighbour mosaic, however, it seems probable that these other factors would play a similarly minor role in maintenance of variation. It appears unlikely that maintenance of variation in T. repens will find a complete explanation based on diversifying selection. Therefore, it seems appropriate to reconsider the potential role played by the introduction of new genetic variation into the population. As has been discussed, the expectation of genotype depletion in clonal populations is tied to the assumption that the rate of replenishment through seedling establishment is negligible. Studies and observations on pasture perennials have consistently supported or confirmed that seedling (genet) establishment is rare relative to ramet establishment. The proposed role for diversifying selection was to retard the rate of genotype depletion to a level that would be offset by the occasional successful seedling. The crucial question, however, of genet input relative to genet extinction has not been commonly examined, presumably due to the difficulty of observing or quantifying genet extinction in a stoloniferous species. The modelling study by Soane and Watkinson (1979) did address this question using demographic data for pasture populations of Ranunculus repens, a clonal perennial with a similar Common garden study, 1982 / 45 growth form to T. repens. Although their computer simulation confirmed the expected decline in numbers of original genets, it also demonstrated that a recruitment rate of only 3% of the existing numbers of genotypes was sufficient to offset the estimated rate of genet loss and so maintain diversity. Their simulation did not provide any evidence for selective maintenance of clonal diversity. They concluded that "even a very low rate of genetic recruitment through seedling establishment is sufficient to maintain a large diversity of clonal families within small areas. Consequently, it may not always be necessary to search for disruptive forces to explain diversity within the field." Soane and Watkinson's (1979) study suggests that the assumption that genet replacement plays a negligible role in the maintenance of diversity in clonal pasture perennials might be unrealistic. A recent study of seedling establishment on disturbances, conducted on the same pasture as the present study, indicated that recruitment of T. repens may also be more frequent than has been generally recognized (Parish, in Turkington 1985). After two years in which no successful T. repens seedlings were detected, in 1985 seedlings appeared at a average of 20 per disturbance. After three months, five T. repens seedlings per m 2 (averaged across the entire pasture) were surviving, representing a potentially massive genotypic input into the population. If this is not a unique event, such episodic recruitment could easily account for the genotypic diversity in T. repens populations, again without resorting to complex, selective explanations. The present results are consistent with this interpretation. Much more of the variation in this population is explained by the establishment of morphologically different individuals than by any subsequent sorting of them through neighbour relationships. Common garden study, 1982 / 46 In addition to seedling recruitment, another potential source of genetic variation in clonal populations is somatic mutation. Although the process is largely unexplored, clonal herbs may be particularly likely to accumulate genetic diversity through somatic mutations. A non-lethal mutation which occurs in any active meristem of a clone may be perpetuated through all of the daughter ramets produced from it. A clonal line may, over a period of years, evolve into a mosaic of genotypes. Such within-individual variation has been documented in Populus (Whitham & Slobodchikoff 1981) and in Hamamelis virginiana (Gill & Halverson 1984) for resistance to insect herbivory. Within clone variation in rate of tiller production by Lolium perenne has been demonstrated using selection experiments (Breese, Hayward & Thomas 1965; Shimamoto & Hayward 1975). Whitham and Slobodchikoff (1981) and Antolin and Strobeck (1985) suggest that the accumulation of somatic mutations may represent an important source of genetic variation within clonal populations. III. COMMON GARDEN STUDY, 1984 A. CARRY-OVER EFFECTS AND THE COMMON GARDEN METHOD The common garden technique has been a fundamental tool of genecology since its inception (Heslop-Harrison 1964; Langlet 1971). Its importance is reflected in a recent quote by Bradshaw (1984) who stated that "The crucial step forward was taken when investigators such as Kerner (1891) and especially Turesson (1922) appreciated the value of the common garden technique to remove direct effects of differing environments and reveal underlying genetically determined variation, an approach so simple that it seems extraordinary it had not been used by Darwin". However, simplicity does not necessarily equate with reliability. Genecological studies which depend, as the present one did, on the common garden method can be criticized on the grounds that the patterns of variation detected do not have a clear genetic basis. At best, variation can be resolved into "genetic" and environmental components but without more formal genetic techniques even this cannot be taken with confidence. Heritability studies using seed progeny or transplants into a variety of environments can often support stronger conclusions. One of the largest difficulties with the common garden method is the uncertainty over the amount of variation that is carried over from the field environment to the garden. The expression of many morphological characters is strongly influenced by the environment in which the plant lives. Genetic interpretation of differences among plants taken directly from their original environments is impossible. The basic assumption of the common garden method is that a common environment will influence all of the plants in the same manner. The 47 Common garden study, 1984 / 48 environmental or developmental component' of variation will then be minimized. Remaining differences among the plants will reflect only genetic differences. The effect of genotype-environment interactions on this assumption of no environmental variation was discussed earlier. The problem of carry-over is also tied to the validity of this assumption. When transplants are collected for a study, the resource levels and physiological state of the material will reflect to some degree the conditions under which the plant was growing. If such a preexisting state influences the plants' performance in the garden, transplants from different environments may express morphological differences which have no genetic basis. Interpretation of such patterns will be confounded by variation "carried over" from the field. Even when seeds are used in place of transplants the problem of carry-over may remain. Because the size and condition of a seed can influence its performance, the environment under which it was produced may precondition it and confound a variation study. Such "maternal effects" have been documented (Schaal 1984). The usual method of addressing carry-over effects in transplanted material is to give the plants an establishment period during which their physiological states ostensibly become adjusted to the garden environment and field-derived resources are used up. This approach is particularly well suited to vegetatively replicating species like Trifolium repens because a transplanted stolon tip can be grown into an entire plant. A subsequent cutting taken from this plant will be formed of tissue grown entirely under the garden conditions. The "second generation" plant grown from this cutting is expected to have lost most of the field effects carried over by its parent. This cloning process can be repeated for as many generations Common garden study, 1984 / 49 as is thought to be necessary. Obviously, the length of the conditioning period or the number of cloning generations required will vary among species. Different investigators have used a variety of conditioning treatments, usually supporting their choice by rationalization rather than with data. In fact, carry-over effects and conditioning treatments have not been widely investigated (but see Warwick & Briggs 1978; Akeroyd & Briggs 1983; Mackenzie 1985) and are often ignored in common garden studies. The patterns of variation described in Part I were those of a set of clones in their second generation in the common garden. The original (first generation) clones had undergone a ten week conditioning period. By the time the measurements were made, the second generation clones had spent an entire season in the common garden. This treatment was comparable to those used on T. repens in the earlier competition studies of Aarssen (1983), the neighbour relationship study of Turkington and Harper (1979c), and the edaphic differentiation studies of Snaydon (1962b) and Snaydon & Bradshaw (1962a,b). Despite the consensus of previous investigators, some doubt remained about the adequacy of the conditioning treatment. The capacity of T. repens to undergo morphological change (plasticity) in response to environmental change is well known. Brougham et al. (1978) noted that such changes could be persistent and suggested caution in the interpretation of experimental results when T. rvpens is grown in different environments. Obviously, if the among-neighbour groups component of variation detected in the 1982 study included persistent environmental variation, the small portion of variation attributed to Common garden study, 1984 / 50 neighbour-specific diversifying selection would have been an overestimate. The second part of this research was undertaken to determine if the original pattern would be retained under a longer conditioning period. Specifically, if the striking neighbour-specific differentiation did, in fact, represent genetic differences it should not 6how much decay over a longer period. Alternatively, if the original conditioning period was inadequate, the differentiation might simply reflect the incomplete loss of developmental responses carried over from the field. In this case, additional time under common garden conditions would allow further expression of a new set of developmental responses, this time reflected as morphological convergence rather than differentiation. Common garden study, 1984 / 51 B. MATERIALS AND METHODS On July 28, 1984 a fresh set of stolon tip cuttings was taken from the 376 surviving clones and replanted in the same manner as described previously. By this time, the clones had been in the common garden for 27 months (from May, 1982) and had undergone two more generations of repropagation for a total of four generations since the original collection. This set of ramets was given the same period of growth (ten weeks) as the 1982 set. The plants were harvested in the fall of 1984 and a set of characters, similar, but not identical to the original set measured (Table XI). Because of the problem of the harvested plants continuing to grow while stored in the cold room, an extended harvest was used. Plants were harvested at a rate that allowed immediate processing in the lab without the need for cold storage. The harvest was begun on Oct 1, 1984 and completed on Nov 10, 1984. Several changes were made in the character set mostly to remove ambiguities in the interpretation of characters. The original character "primary stolon number" (PSTOL#) was eliminated because of difficulty in identification of primary stolons. In many cases, a prominent stolon having secondary and tertiary branches had developed from the base of a primary stolon rather than directly from the main taproot. To define this as a secondary stolon simply because its branch point was above ground rather than below seemed arbitrary. The alternative of including such branches as primaries was not feasible because of the impossibility of defining a consistent criterion for recognizing the point at which a secondary branch becomes a primary. For these reasons, only the total number of stolons (TSTOL#) was recorded. The character "internode number" (INODE#) was also Common garden study, 1984 / 52 Table XI. List of characters measured in the 1984 common garden study of variation in a pasture population of Trifolium repens. CHARACTER UNITS OF M E A S U R E M E N T Root dry weight RTWT 0.01 gm Shoot dry weight SHTWT 0.01 gm Total dry weight TWT 0.01 gm Total stolon number TSTOL# Length of primary stolon PSTOL.L 5 mm Total stolon length TSTOL.L 5 mm Internode length INODE.L 0.5 mm Petiole length PET.L 1 mm Leaflet width LF .W 0.5 mm Leaflet length LF . L 0.5 mm eliminated because of the difficulty of counting the first few internodes which are commonly unelongated. Because of the highly variable number of these unelongated internodes, the character seemed to have little relationship to the structure of the stolon. "Internode length" (INODE.L) was expanded to include three internodes on each of three stolons for a total of nine measurements per plant. One additional measurement, total stolon length (TSTOL.L) was recorded. Common garden study, 1984 / 53 C. ANALYSES AND RESULTS Analyses of variance were performed on the data in the same manner as for the 1982 data (Table XII). Because none of the variables showed significant levels of heterogeneity among variances, no transformations were used. The analyses provided no evidence of differentiation for any of the characters. In no case was a significant added variance component detected. The highest proportion of the variance explained was for PSTOL.L (0.17%). The highly significant differentiation detected in 1982 had completely disappeared. Among-genet (within-neighbour) variance components were calculated for the four characters with repeated measurements (PET.L, LF.W, LF.L, and INODE.L. In contrast to the among-neighbour variance components, the among-genet components remained large (40%-70%) and highly significant (Table XIII). Because the number of measurements of INODE.L had been increased to three on each of three stolons per individual an among-stolon (within-genet) variance component could also be calculated. This component was also highly significant and explained 30% of the variance in INODE.L. Thus, for this character, there was more difference among stolons than there had ever been among neighbours for any character. The correlation structure was similar to that of the 1982 data set (Table XIV). A l l correlations were positive and all except LF .W and LF . L with TSTOL# were significant (p<0.01). Correlations among the seven characters which were measured in both years tended to decline from 1982 to 1984. Correlations among the weights, TWT, RTWT, and SHTWT remained high, however. Common garden study, 1984 / 54 Table XII. Summary of analyses of variance for 12 morphological characters from four neighbour-specific groups of Trifolium repens. Data from the 1984 common garden study. n MS A MS E S 2 A SIG %VARC RTWT 92.3 0.265 0.460 - NS 0.00 SHTWT 92.5 1.11 1.37 - NS 0.00 TWT 92.3 2.13 3.23 - NS 0.00 TSTOL# 92.3 50.8 80.4 - NS 0.00 TSTOL.L 90.5 455000 474000 - NS 0.00 PSTOL.L 92.3 6140 5300 9.09 NS 0.17 INODE.L 89.0 26.2 31.9 NS 0.00 PET.L 91.8 96.4 97.1 - NS 0.00 LF . L 91.3 1.61 2.04 - NS 0.00 LF.W 91.3 1.08 2.89 _ - NS 0.00 n = mean # of individuals/group MS = among-groups mean square A M S £ = error mean square S : = added variance component = ( MS - MS )/n A y A E %VAR(A) = % variance among-groups = S 2 /( S 2 + MS ) A A E Sig = Level of significance from F-test for presence of added variance components. ** = p<0.01 * = p<0.05 NS = not significant Common garden study, 1984 / 55 Table XIII. Variance components for four morphological characters from a pasture population of Trifolium repens. Data from the 1984 common garden study. NEIGHBOUR GENET ERROR STOLON PET.L LF .W LF . L INODE.L 0.0 (NS) 0.0 (NS) 1.3 (NS) 0.0 (NS) 40.3 (**) 70.7 (**) 70.6 (**) 49.7 (**) 59.7 29.3 28.1 20.1 30.2 (**) Entries give % variation accounted for by each source of variation and a test for the presence of an added variance component. ** = p<.01 Sources of variation: Neighbour = among four neighbour-specific groups Genet = among-Trifolium repens individuals (within-groups) Error = among-measurements (within-individuals)(within-stolons for INODE.L) Stolon = among stolons (within-individuals) (for INODE.L only) Common garden study, 1984 / 56 Table XIV. Correlations among 12 morphological characters of Trifolium repens. Data from the 1984 common garden study. RTWT 1.000 SHTWT 0.867 1.000 TWT 0.945 0.982 1.000 TSTOL# 0.723 0.760 0.771 1.000 TSTOL.L 0.733 0.880 0.853 0.836 1.000 PSTOL.L 0.354 0.561 0.501 0.235 0.596 1.000 INODE.L 0.278 0.438 0.392 0.205 0.488 0.811 1.000 PET.L 0.326 0.500 0.451 0.235 0.380 0.477 0.518 1.000 LF . L 0.187 0.219 0.214 0.104 0.142 0.148 0.216 0.316 1.000 LF .W 0.211 0.223 0.226 0.136 0.170 0.184 0.235 0.302 0.837 RTWT SHT TWT TSTOL TSTOL PSTOL INODE PET.L LF . L WT # L L L Table entries are product moment correlation coefficients. Characters with coefficients greater than 0.139 are significantly correlated (p<.01) Coefficients indicating no significant correlation are printed in dark type, n = 342 Common garden study, 1984 / 57 Because the variables remained highly intercorrelated, a principal components analysis was performed (Table XV). As for the 1982 data, the first axis (PCA l , 53.4% of the total variation) was most strongly influenced by plant weights and stolon numbers. ANOVA on P C A l produced the same result as the ANOVAs on the original variables, no added variance components, only 0.69% of the variance accounted for, and no evidence of differentiation (Table XVI). The second and third axes (18.3% and 13.7% of the total variation) were influenced by the relatively low correlations between LF . L and LF .W and the other characters. The stolon structure pattern noted on PCA2 in 1982 appeared again but was less distinct. The fourth and fifth axes accounted for 5.8% and 3.7% of the total variation respectively for a five axis total of 94.9%. As in 1982, there were no obvious patterns on the last two axes. ANOVAs on PCA2-5 again produced no evidence of differentiation (Table XVI). Common garden study, 1984 / 58 Table XV . Principal components analysis of 12 morphological characters of Trifolium repens. Data from the 1984 common garden study. PCA1 PCA2 % VARIATION 53.42 18.30 CUMULAT IVE 53.42 71.72 RTWT 0.366 0.192 SHTWT 0.412 0.135 TWT 0.408 0.161 TSTOL# 0.331 0.273 TSTOL.L 0.392 0.172 PSTOL.L 0.291 -0.144 INODE.L 0.262 -0.238 PET.L 0.255 -0.262 LF . L 0.150 -0.583 LF .W 0.159 -0.574 AXIS PCA3 PCA4 PCA5 13.72 5.84 3.75 85.44 91.28 94.93 0.228 -0.059 -0.514 0.062 -0.084 -0.168 0.126 -0.077 -0.304 0.262 0.097 0.638 -0.009 0.208 0.357 -0.522 0.303 -0.170 -0.534 0.245 0.087 -0.236 -0.861 0.207 0.356 0.090 -0.002 0.347 0.171 0.027 Table entries are coefficients for each character for the first five principal components (axes). % VARIATION is the amount of variation in the multivariate data set which is explained by each axis. Common garden study, 1984 / 59 Table XVI . Summary of analysis of variance of principal component scores for four neighbour-specific groups of Trifolium repens. Data from 1984 common garden study. AXIS MS MS S 2 A SIG % VAR(A) PCA 1 PCA 2 PCA 3 PCA 4 PCA 5 8.487 0.285 0.460 1.480 0.353 5.315 1.844 1.380 0.577 0.365 0.037 0.011 NS NS NS NS NS 0.69 0.00 0.00 1.80 0.00 n = mean # of individuals/group MS = among-groups mean square A MS = error mean square E S 2 = added variance component = ( MS - MS )/n A A E %VAR(A) = % variance among-groups = S 2 /( S 2 + MS^ ) A A E Sig = Level of significance from F-test for presence of added variance components. ** = p<0.01 * = p<0.05 NS = not significant Common garden study, 1984 / 60 D. DISCUSSION The extreme plasticity of Trifolium repens is well known (Broughham et al. 1978; Hi l l 1977) In fact, Bradshaw (1965) referred to the petiole of T. repens as one of the most plastic plant organs known. It is noteworthy that petiole length showed the strongest differentiation of any of the measured characters in the first part of this study. It should not be surprising that T. repens would respond plastically to changes in its neighbour environment. It should also not be surprising that some of the plastic effects produced in different environments were retained for a season in the common garden. What was unexpected, however, was that the carry-over effects would have been so extreme. Where there had been 6-7% differentiation among neighbour-groups after one season, after two years none remained. The clear conclusion is that there is no evidence at all in this data set for morphological differentiation. The apparent differentiation detected in the first part of the study simply reflected the carry-over of environment-specific developmental adjustments (phenotypic plasticity). Over the 30 months of the study, the plants converged morphologically as they readjusted developmentally to the common garden environment. Presumably, if the plants were transplanted back to the field, the neighbour-specific pattern would be regenerated. The morphological convergence among neighbour-groups did not extend to variation among individuals. The among-genets component of variance did not change much between years. Coefficients of variation were also similar in both years. Thus, the convergence represented shifts in mean values for the neighbour-groups rather than an overall decline in variation. Common garden study, 1984 / 61 A number of studies on differentiation have also demonstrated morphological convergence among plants transplanted to a common environment. Watson (1969) collected Potentilla erecta from two adjacent habitats, an Agrostis pasture and a Molinia meadow. Lengths of stems and basal internodes were measured immediately after collection and again after two years in a common garden. Populations from the two habitats differed in the initial measurements by =70% for both characters. Two years later, the differences had declined to =30%. Despite the convergence, the populations remained significantly (p<.01) differentiated, supporting the conclusion that they were genetically different. Warwick and Briggs (1979) collected Achillea millefolium, Bellis perennis, Plantago lanceolata, P. major, and Prunella vulgaris from 40 populations representing four different lawn and grassland habitat types. Leaf sizes and widths, measured immediately after collection showed significant (p<.05) differentiation both among-habitat types and among-populations (within-habitat types) for all five species. After one year in. a common garden, only one of the species, P. major, retained the among-habitat differentiation. A l l retained significant among-individual and among-population variation. After a second year, the among-population differences in P. lanceolata had also disappeared. Akeroyd and Briggs (1983) collected Rumex crispus over two years from 23 sites representing five different habitat types. Measurements of seven morphological characters at the time of collection showed significant differentiation (p<.05) Common garden study, 1984 / 62 among-habitat types for four characters and among-population (within-habitat types) for all seven. Measurements after one year in a common garden on plants collected in the first year showed convergence among-populations but divergence among-habitat types. For the second year's collection, the results were opposite, convergence both among-habitat types and among-populations. The results of Akeroyd and Briggs (1983) show that the pattern of phenotypic response to a common garden may be unpredictable. Both divergence and convergence were noted as well as different responses among years. Seliskar (1985) reciprocally transplanted ramets of five salt marsh species along transects between mudflat and upland habitats. After one year, the ramets transplanted vertically along transects had converged morphologically with ramets transplanted between comparable points on different transects. In these studies, convergence has been noted over periods varying from 6 months to 2 years. In the present study it occurred before 30 months. The conclusion is that if measurements on common garden material are made within the first season after collection, carry-over effects may not have had time to dissipate and differentiation may be overestimated. The time scale for morphological convergence noted in the above studies fits well with the concept of plasticity as reversible, developmental responses to environmental changes occurring within the lifetime of the individual, or in the case of clonal herbs such as T. repens, within the lifetime of the ramet (Bradshaw 1965). This time scale is linked to the largely intuitive treatment of Common garden study, 1984 / 63 carry-over effects in genecological work on perennial herbs. After transplantation to a common environment, plants are generally given "conditioning" periods roughly corresponding to the lifetime of a ramet. The conditioning period is often measured as numbers of "tissue turnovers" or "cloning generations". It is tacitly assumed that any persistent effects of the field environment will be lost with the death of the original material and that development of second generation ramets will be entirely under the influence of the common environment. Phenotypes of second generation ramets thus should not reflect carry-over effects. When in doubt, experimenters often choose to err on the conservative side and add additional time for conditioning (e.g., Snaydon (1971), 2 years for T. repens; Silander (1979), 2 years for Spartina patens; Shaver, Chapin & Billings (1979), 18 months for Carex aquatilis; Turkington (1983a,b), 17 months for T. repens). The studies which have documented convergence suggest that such extended periods are usually sufficient to minimize carry-over. Pilot studies on carry-over are rare, however, leaving most investigators to "assume that this treatment removed any effects of preconditioning by the different field environments" (Shaver, Chapin & Billings 1979). One experimental study of carry-over effects was conducted by Mackenzie (1985). Ramets cloned from Ranunculus repens genets were transplanted into three laboratory environments, grown for three months and then transplanted to a common environment for six months. Highly significant (p<.01) differences were found wtf/ttn-genets for ramets subjected to the different treatments. These results demonstrate that plastic effects that develop quickly (three months) may take longer to disappear (at least six months). Common garden study, 1984 / 64 In the case of T. repens, the short lifespan of the ramet makes it convenient to incorporate several generations of tissue turnover prior to experimental work. Thus, a number of studies of differentiation and variation in Trifolium repens (and other perennial herbs) have been conducted using only brief periods of conditioning but incorporating several tissue turnovers. These include Snaydon (1962b), 3 months, Turkington and Harper (1979c), 3 months, Gliddon and Trathan (1985), 5 months, Aarssen and Turkington (1985c), 4 months, and Aarssen and Turkington (1985b), no conditioning. Burdon (1980a) did not report any conditioning period. Similarly brief periods were used for Bellis perennis and Prunella vulgaris by Schmid (1985), (7 weeks). Lovett-Doust (1981c) found that among-habitat differentiation increased in the common garden for R. repens and thus used no conditioning period. These studies have not generally reported the actual number of tissue turnovers used but generally refer to periods of cloning or of repeated division, implying that several tissue generations had taken place. The present results show that carry-over effects in T. repens can last well beyond two cloning generations. Therefore, an extended period of time should more effectively minimize carry-over than an extended number of cloning generations. Some of the results based on short conditioning periods may have reflected carry-over effects and thus may have overestimated differentiation. It should be noted that while all of the studies cited above used brief pre-experimental conditioning periods, the two which demonstrated neighbour-specific differentiation in Trifolium repens (Turkington and Harper 1979c; Aarssen and Turkington 1985b) both allowed extended periods of growth under experimental conditions (one year) before taking measurements. The results of these two studies, then, may have been less influenced by carry-over than the others Common garden study, 1984 / 65 mentioned. Therefore, their conclusions are not necessarily brought into question by the present results. It should also be noted that these two studies looked at performances of T. repens (dry weight production) in the presence of the actual neighbours (i.e., neighbour-compatibilities were measured directly). Because there is no evidence linking the morphological characters studied here to performances, the lack of neighbour-specific morphological differentiation does not necessarily imply a lack of differentiation for neighbour compatibilities. The implication of my results for genecological work is that short term studies with perennial herbs may not give reliable evidence on differentiation. This applies both to common garden and reciprocal transplant studies. The control of carry-over effects will require more objective treatment than intuitively generated numbers of tissue turnovers. The process of plastic adjustment to experimental environments can be followed by periodic recording of characters beginning at the time of collection. Final measurements should be delayed until after the initial convergence is complete. IV. DIVERSIFYING SELECTION AND TRIFOLIUM REPENS: A RECONSIDERATION These results have confirmed that there is considerable morphological variation in T. repens populations which is stable under extended periods in a common garden, and which may be genetically based. They offer no support however, for the hypothesis that genetically based morphological variation in this population is being maintained by neighbour-specific diversifying selection. There was no evidence for spatial correlation between the patterns of distribution of the grasses and the distribution of persistent variation in the morphology of clovers. Although the data from the first season in the common garden showed a neighbour-specific pattern, little of the population-wide variation was accounted for, and even this later proved to be ephemeral. While T. repens individuals are influenced by changes in their neighbour environment, the effects of neighbours may not extend to ramet dynamics. Responses, at the morphological level, apparently reflect transitory adjustments in physiology and development, not changes in the establishment and survival of ramets. Thus, there does not appear to be a strong linkage beween neighbour relationships and population-wide genetic patterns in morphology. In light of the importance often attributed to environmental heterogeneity as a diversifying influence (e.g., Hamrick 1980; Hedrick, Ginevan & Ewing 1976; Ennos 1983), it seems appropriate to consider why such a prominent element of patchiness as the neighbour mosaic did not turn out to constitute an environmental heterogeneity. Why is morphological variation in this population not strongly shaped by neighbour-specific diversifying selection? The relationship among 66 Diversifying selection and Trifolium. repens: a reconsideration / 67 environmental heterogeneity, diversifying selection, and maintenance of variation has been analysed theoretically (Levins 1969; Hedrick, Ginevan & Ewing 1976; Dickinson & Antonovics 1973). Models have suggested that diversifying selection in a heterogeneous environment will be most effective in maintaining variation when selective potentials among microhabitats (patches) are high and consistent, and individuals spend their entire lives in a single patch (i.e., have coarse grained environments). Bradshaw (1972) has argued that plants are commonly subject to strong selective potentials. This is, in part, a consequence of their immobility and a reflection of the fact that most plants do have coarse-grained environments. Certainly, many plant populations will be subject to diversifying selection. This expectation is borne out by the large number of cases in which plant populations are differentiated in concordance with an environmental heterogeneity (e.g., Jain & Bradshaw 1966; Snaydon & Davies 1976, 1982). For perennials, effects of selection may be cumulative, especially where the conditions of a patch are permanent or repeated annually. The expectation of among-neighbour selection in T. repens was supported by the results of competition experiments. However, the evidence on the effectiveness of such selection is equivocal. Plot and greenhouse studies have commonly, but not invariably shown significant neighbour-specific differences in performance measures such as dry weight or stolon production. Turkington (1983c), for example, found that one of two T. repens individuals responded differently to different neighbours, but the other did not. Solangaarachchi (1985) found that performances of T. Diversifying selection and Trifolium repens: a reconsideration / 68 repens differed among neighbours when stolon lengths were measured but not when measurements were on dry weights of stolons or shoots. In addition, rankings of neighbours on the basis of their effects on the performances of T. repens have sometimes been inconsistent between and even within studies. Solangaarachchi (1985) found significantly greater stolon lengths in T. repens grown in experimental plots of Lolium perenne or Cynosurus cristatus than in plots of Holcus lanatus or Agrvstis capillaris. The transplants of Turkington et al. (1979) into natural plots of the same four grasses produced a similar ranking. The concurrent transplants of Turkington and Harper (1979c) into patches of the same grasses in the same field, however, produced a reversal in the rankings of L. perenne and A. capillaris. Consideration of the patch structure of the pasture generates further doubt as to the likelihood of its being a source of effective diversifying selection. Any single survey of this pasture is likely to reveal distinct patches of some or all of the four grasses. Such was the case in May, 1982 when the material for this study was collected. One hundred patches of each of the four grasses were easily located. Repeated surveys across seasons or years, however, demonstrate that these patches are not stable (Parish, personal communication). The size and composition of patches can vary widely through the year as a result of differences in the active growing times of the grasses. Additionally, the neighbour mosaic varies across years. A survey in May, 1983, (one year after the original collection date), failed to identify a single patch dominated by Poa compressa (personal observation). Similarly, in 1984 Holcus lanatus was nearly absent from the pasture (Parish, personal communication). Such wide variation in species Diversifying selection and Trifolium repens: a reconsideration / 69 composition is not unusual for permanent pastures (Thorhallsdottir 1983; Snaydon 1985). As a consequence, any selective potential associated with neighbour relationships may fluctuate even within the lifetimes of clover ramets. Snaydon (1985) has argued that this temporal variation should enhance the effectiveness of diversifying selection in maintaining variation in pasture populations. Theoretical models, however, suggest the opposite, that the effectiveness of diversifying selection will be diluted by such inconsistencies in strength and direction of selection (Ennos 1983). The second condition for effective maintenance of variation by diversifying selection requires that the environments of individuals be coarse-grained. A consequence of the modular construction of T. repens and its stoloniferous habit is that the environmental grain of a genet can be different than that for a ramet. Because a genet is composed of a number of stolons which may commonly grow across patch boundaries, the environment of a genet will usually be fine grained. In fact, a long lived T. repens genet probably encounters much of the range of microhabitats in a pasture. Thus, the persistence of genets may not depend on conditions within any single patch. In contrast, a ramet does spend its entire life on a single spot. If conditions at that spot do not change during that period of time, the ramet will have a coarse-grained environment. Its persistence and replication will be dependent on conditions at that spot. Thus there is a possibility of strong patch-specific selection among ramets that could lead to sorting of genets among patches and a reduced rate of genet depletion. It is not clear, however, just how coarse-grained the environments of ramets really are. Trifolium repens ramets do not normally become physically independent of the Diversifying selection and Trifolium repens: a reconsideration / 70 parent ramet until after they have become well established (Harvey 1970). It has been demonstrated in various clonal plants that daughter ramets can draw physiological support from their parents (Lovett-Doust 1981a; Hartnett & Bazzaz 1983; Mackenzie 1985). Instead of being dependent on the conditions within a single patch, the performances of T. repens genets may actually reflect the integration of the conditions across several patches. Thus ramets may not have the coarse-grained environment that would be compatible with a diversifying selection scenario. The evidence for clonal integration in T. repens, however, is not conclusive. Harvey (1970) showed that photosynthates moved only in the direction of stolon apices. This suggests that integration would occur only within stolon branches. Ramets on different stolons would not support each other. In contrast, Solangaarachchi (1985) found that manipulations (prevention of root formation) in one part of an interconnected clone infuenced growth and stolon branching in other parts of the clone. Most of the experiments whch have supported the expectation of neighbour-specific selection among ramets have been based on the survival and growth of excised ramets. If, as seems likely, the "regeneration niche" (Grubb 1977) of excised ramets is narrower than that of ramets with intact stolon connections, the performances of excised ramets may not equate with those of intact ramets. These experiments may, then, have overestimated the strength of the neighbour mosaic as an agent of diversifying selection. One relevant experiment was performed by Turkington (1983a) who planted ramets of T. repens at the centers of hexagons of patches of different neighbours including four pasture grasses. Growth of the T. repens ramets into the patches was monitored. Final percent Diversifying selection and Trifolium repens: a reconsideration / 71 cover of T. repens did not vary significantly among the four grass patches. For initial invasion, at least, the differences among grass patches may be irrelevent to established T. repens plants. Because the effectiveness of among-neighbour selection on T. repens ramets may be diluted by instability in patch structure and by among-patch integration, the neighbour mosaic may not really have represented a likely source of diversifying selection for T. repens populations. Bazzaz and Sultan (in press) have also proposed that clonal integration and temporal environmental instability will buffer clonal genets against selection from their biotic environment. They also suggest that it may be unrealistic to expect extensive selective elimination of genets from clonal populations. Thorhallsdottir (1983) noted that T. repens ramets commonly become established on small gaps within patches. She suggested that these gaps were more important than the identity of the patch for the establishment of T. repens ramets (and therefore for the persistence of genets). These considerations reinforce the earlier conclusion that a selective explanation for maintenance of variation in pasture populations of T. repens may be both unlikely and unnecessary. R E F E R E N C E S Aarssen, L. W, (1983). Interactions and coexistance of species in pasture community evolution. Ph.D. Thesis, University of British Columbia. Aarssen, L. W. & Turkington, R. (1985a).Vegetation dynamics and neighbour associations in pasture community evolution. Journal of Ecology, 73,585-603. Aarssen, L.W. & Turkington, R. (1985b). Biotic specialization between neighbouring genotypes in Lolium perenne and Trifolium repens from a permanent pasture. Journal of Ecology, 73,605-614. Aarssen, L.W. & Turkington, R. (1985c). Within species diversity in natural populations of Holcus lanatus, Lolium perenne, and Trifolium repens from four different aged pastures. Journal of Ecology, 73,869-886. Akeroyd, J.R. & Briggs, D. (1983). Genecological studies in Rumex crispus L. I. Garden experiments using transplanted material. New Phytologist, 94,309-323. Angseesing, J . P. A. & Angseesing, W. J . (1973). Field observations on the cyanogenesis polymorphism in Trifolium repens. Heredity, 31,276-282. Antlfinger, A. E. (1981). The genetic basis of microdifferentiation in natural and experimental populations of Borrichia fructescens in relation to salinity. Evolution, 35,1056-1068. Antolin, M. F. & Strobeck, C. (1985). The population genetics of somatic mutations in plants. American Naturalist, 126,52-62. Antonovics, J . (1972). Population dynamics of the grass Anthoxanthum oderatum on a zinc mine. Journal of Ecology, 60,351-366. Antonovics, J . (1976). The input from population genetics: "The new ecological genetics". Systematic Botany, 1,233-245. Antonovics, J . (1984). Genetic variation within populations. Perspectives on Plant Population Ecology (Ed. by R. Dirzo and J . Saruknan), pp.229-241. Sinauer, Sunderland, Mass. Antonovics, J . , Bradshaw, A. D. & Turner, R. G. (1971). Heavy metal tolerance in plants. Advances in Ecological Research, 7,1-85. Atwood, S. S. (1942). Oppositional alleles causing cross incompatibility in Trifolium repens. Genetics, 27,333-338. Bazzaz, F. A. & Sultan, S. E. (in press). Ecological variation and the maintenance of diversity in plant populations. Symposium of the International Society of Biosystematics. Bradshaw, A. D. (1965). Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics, 13,115-155. 72 / 73 Bradshaw, A. D. (1972). Some of the evolutionary consequences of being a plant. Evolutionary Biology, 5,25-47. Bradshaw, A. D. (1984). Ecological significance of genetic variation between populations. Perspectives on Plant Population Ecology (Ed. by R. Dirzo & J . Sarukhan), pp. 213-228. Breese, E. L., Heyward, M. D. & Thomas, A. C. (1965). Somatic selection in perennial ryegrass. Heredity, 20,367-379. Brewbaker, J . L. (1955). Leaf marking in white clover. Journal of Heredity, 46,115-121. Briggs, D. & Walters, S. M. (1984). Plant Variation and Evolution. Cambridge University Press, Cambridge. Broughham, R. W., Ball, P. R. & Williams, W. M. (1978). The ecology and management of white clover based pastures. Plant Relations in Pastures (Ed. by J . R. Williams), pp. 309-324. CSIRO, Melbourne. Brown, A. H. D. (1979). Enzyme polymorphism in plant populations. Theoretical Population Biology, 15,1-42. Bulow-Olsen, A., Sackville-Hamilton, N. R. & Hutchings, M. J . (1984). A study of growth form in genets of Trifolium repens L. as affected by intra- and interplant contacts. Oecologia(Berlin), 61,383-387. Burdon, J . J . (1980a). Intra-specific diversity in a natural population of Trifolium repens. Journal of Ecology, 68,717-735. Burdon, J . J . (1980b). Variations in disease resistance within a population of Trifolium repens. Journal of Ecology, 68,737-744. Burdon, J . J . (1983). Biological flora of the British Isles: Trifolium repens. Journal of Ecology, 71,307-330. Burdon, J . J . & Harper, J . L. (1980). Relative growth rates of individual members of a plant population. Journal of Ecology, 68,953-958. Cahn, M. G. & Harper, J . L. (1976). The biology of the leaf mark polymorphism in Trifolium repens L. I. Distribution of phenotypes at a local scale. Heredity, 37,309-325. Carnahan, H. L., Hil l, H. D., Hanson, A. A. & Brown, K.G. (1955). Inheritance and frequencies of leaf markings in white clover. Journal of Heredity, 46,109-114. Chapman, D. F. (1983). Growth and demography of Trifolium repens stolons in grazed pastures. Journal of Applied Ecology, 20,597-608. Evol. 37:947-956. Charles. A. H. (1964). Differential survival of plant types in swards. Journal of / 74 the British Grassland Society, 16,69-75. Chestnutt, D. M. B. & Lowe, J . (1970). Agronomy of white clover-grass swards. Review. Occasional Symposium of the British Grassland Society, 6,191-213. Clatworthy, J . N. (1960). Studies on the nature of competition between closely related species. Ph.D. Thesis, Univ of Oxford. Corkill, L. (1952). Cyanogenesis in white clover (Trifolium repens L.) VI. Experiments with high-glucoside and glucoside-free strains. New Zealand Journal of Science & Technology Section A, 34,1-16. Corkill, L. (1971). Leaf markings in white clover. Journal of Heredity, 62,307-310. Davies, W. E. (1973). Breeding white clover (Trifolium repens) in relation to companion grass. Report of the Fodder Crops Meeting, Dublin. Dickinson, H. & Antonovics, J . (1973). The effect of environmental heterogeneity on the genetics of finite populations. Genetics, 3,713-735. Dirzo, R. & Saruknan, J . (1984). (Eds.) Perspectives on Plant Population Ecology. Sinauer, Sunderland, Mass. Donald, C. M. (1963). Competition among crop and pasture plants. Advances in Agronomy, 15,1-118. Ennos, R. A. (1983). Maintenance of genetic variation in plant populations. Evolutionary Biology, 16,129-155. Falconer, D.S. (1981). Introduction to Quantitative Genetics, 2nd ed., Longman, New York. Fox, D. J . & Guire, K. E. (1976). Documentation for MIDAS. Statistical Research Laboratory, University of Michigan, Ann Arbor. Gill, D. E. & Halverson, T. G. (1984). Fitness variation among branches within trees. Evolutionary Ecology (Ed. by B. Shorrocks). 23rd Symposium of the British Ecological Society. Blackwell, Oxford. Gliddon, C. & Trathan, P. (1985). Interactions between white clover and perennial ryegrass in an old permanent pasture. Structure and Functioning of Plant Populations. 2. (Ed. by J . Haeck & J . W. Woldendorp), North Holland Publishing Company, Amsterdam. Grubb, P. J . (1977). The maintenance of species richness in plant communities: The importance of the regeneration niche. Biological Review, 52,107-145. Hamrick, J . L. (1982). Plant population genetics and evolution. American Journal of Botany, 69,1685-1693. / 75 Hamrick, J . L., Linhart, Y. B., & Mitton, J . B. (1979). Relationship between life history characteristics and electrophoretically detectable genetic variation in plants. Annual Review of Ecology and Systematica, 10,173-200. Harberd, D. J . (1961a). Observations on population structure and longevity of Festuca rubra L. New Phytologist, 60,184-206. Harberd, D. J . (1961b). The case for extensive rather than intensive sampling in genecology. New Phytologist, 60,325-338. Harberd, D. J . (1962). Some observations on natural clones in Festuca ovina. New Phytologist, 61,85-100. Harberd, D. J . (1963). Observations on natural clones of Trifolium repens. New Phytologist, 62,198-204. Harper, J . L. (1967). A Darwinian approach to plant ecology. Journal of Ecology, 55,247-270. Harper, J . L. (1977). Population Biology of Plants. Academic Press, N.Y. Harper, J . L. (1978). Plant relations in pastures: A keynote address. Plant Relations in Pastures (Ed. by J . R. Wilson), pp. 3-16. CSIRO, Melbourne. Harper, J . L. (1982). After description. The Plant Community as a Working Mechanism. (Ed. by E. I. Newman), Special Publication of the British Ecological Society, 1, pp. 29-52. Blackwell, Oxford. Hartnett, D. C. & Bazzaz, F. A. (1983). Physiological integration among ramets in Solidago canadensis. Ecology, 64,779-788. Harvey, H. J . (1970). Patterns of assimilate translocation in Trifolium repens. Occaisional Symposium of the British Grassland Society, 6,181-186. Harvey, H. J . (1979). The regulation of vegetative reproduction. D. Phil Thesis, University of Wales. Hedrick, P. W., Ginevan, M. E. & Ewing, E. P. (1976). Genetic polymorphism in heterogeneous environments. Annual Review of Ecology and Systematics, 7,1-32. Heslop-Harrison, J . (1964). Forty years of genecology. Advances in Ecological Research, 2,159-247. Hill, J . (1977). Plasticity of white clover grown in competition with ryegrass. Report of the Welsh Plant Breeding Station for 1976, Aberystwyth, pp.24-25. Jain, S. K. & Bradshaw, A. D. (1966). Evolutionary divergence among adjacent plant populations. I. The evidence and its theoretical analysis. Heredity, 21,407-41. / 76 Jones, D. A. (1966). On the polymorphism of cyanogenesis in Trifolium repens L. I. Selection by animals. Canadian Journal of Genetics and Cytology, 8,556-567. Kays, S. & Harper, J . L. (1974). The regulation of plant and tiller density in a grass sward. Journal of Ecology, 62,97-105. Kerner von Marilaun, A. (1891). Pflanzenleben Vol. 2. Leipzig. Krajina, V . J . (1965). Biogeoclimatic zones and classification of British Columbia. Ecology of Western North America, 1,1-17. Langer, R. H. M., Ryle, S. M. & Jewiss, O. R. (1964). The changing plant and tiller populations of timothy and meadow fescue swards. I. Plant survival and the pattern of tillering. Journal of Applied Ecology, 1,197-208. Langlet, O. (1971). Two hundred years of genecology. Taxon, 20,653-722. Levins, R. (1968). Evolution in Changing Environments. Princeton University Press, Princeton. Lewontin, R. C. (1974a). The Genetic Basis of Evolutionary Change. Columbia University Press, New York. Lewontin, R. C. (1974b). The analysis of variance and the analysis of causes. American Journal of Human Genetics, 26,400-411. Linhart, Y. B. (1974). Intrapopulation differentiation in annual plants. I. Veronica peregrina L. raised under noncompetitive conditions. Evolution, 28,232-243. Loveless, M. D. & Hamrick, J . L. (1984). Ecological determinants of genetic structure in plant populations. Annual Review of Ecology and Systematics, 15,65-95. Lovett-Doust, L. (1981a). Intraclonal variation and competition in Ranunculus repens. New Phytologist, 89,495-502. Lovett-Doust, L. (1981b). Population dynamics and clonal specialization in a clonal perennial (Ranunculus repens). I. The dynamics of ramets in contrasting habitats. Journal of Ecology, 69,743-755. Lovett-Doust, L. (1981c). Population dynamics and local specialization in a clonal perennial (Ranunculus repens). H. The dynamics of leaves and a reciprocal transplant-replant experiment. Journal of Ecology, 69,757-768. Lowe, J . (1970). (Ed.) White clover research. Occasional Symposium of the British Grassland Society, 6. Mackenzie, S. (1985). Reciprocal transplantation to study local specialization and the measurement of components of fitness. Ph.D. Thesis, University of Wales. / 77 McNeilly, T. & Roose, M. L. (1984). The distribution of ryegrass genotypes in swards. New Phytologist, 98,503-513. Mitchell-Olds, T. & Rutledge, J . J . (1986). Quantitative genetics in natural plant populations. American Naturalist, 127,379-402. Mytton, L. R. (1975). Plant genotype x Rhizobium strain interactions in white clover. Annals of Applied Biology, 73,329-338. Newton, P. C. D. (1986). The establishment, growth, and fate of white clover plants with special reference to the physiology of stolon growth. Ph.D. Thesis, University of Wales. Oinonen, E. (1967). Sporal regeneration of bracken in Finland in the light of the dimensions and age of its clones. Acta Forestalia Fennica, 83,3-96. Rhodes, I. & Harris, W. (1979). The nature and basis of differences in sward composition and yield in ryegrass-white clover mixtures. Changes in Sward Composition and Productivity (Ed. by A. H. Charles & R. J . Haggar), British Grassland Society, Hurley, pp 58-60. Sackville-Hamilton, N. R. (1980). Variation and adaptation in wild populations of white clover (Trifolium repens) in E. Anglia. Ph.D. Thesis, Univ of Cambridge. Salzman, A. (1985). Habitat selection in a clonal plant. Science, 228,603-604. Saruknan, J . & Harper, J . L. (1973). Studies on plant demography: Ranunculus repens L., R. bulbosus L., and R. acris L. I. Population flux and survivorship. Journal of Ecology, 61,675-716. Schaal, B. A. (1984). Life history variation, natural selection, and maternal effects in plant populations. Perspectives on Plant Population Ecology (Ed. by R. Dirzo & J . Sarukhan), pp. 188-206, Sinauer, Sunderland, Mass. Scheiner, S. M. & Goodnight, C. J . (1984). The comparison of phenotypic plasticity and genetic variation in populations of the grass Danthonia spicata. Evolution, 38,845-855. Schmid, B. (1985). Clonal growth in grassland perennials. III. Genetic variation within and between populations of Bellis perennis and Prunella vulgaris. Journal of Ecology, 73,819-830. Schmid, B. & Harper, J . L. (1985). Clonal growth in grassland perennials. I. Density and pattern-dependent competition between plants with different growth forms. Journal of Ecology, 73, 793-808. Seliskar, D. M. (1985). Effect of reciprocal transplanting between extremes of plant zones on morphometric plasticity of five plant species in an Oregon salt marsh. Canadian Journal of Botany, 63,2254-2262. / 78 Shaver, G. R., Chapin, F. S. & Billings, W. D. (1979). Ecotypic differentiation in Carex aquatilis on ice wedge polygons in the Alaska coastal tundra. Journal of Ecology, 67,1025-1046. Shimamoto, Y. & Hayward, M. D. (1975). Somatic variation in Lolium perenne. Heredity, 34,225-230. Silander, J . A. (1979). Microevolution and clone structure in Spartina patens. Science, 203,658-660. Silander, J . A. (1985). The genetic basis of the ecological amplitude of Spartina patens. II. Variance and covariance analysis. Evolution, 39,1034-1052. Silander, J . A. & Antonovics, J . (1979). The genetic basis of the ecological amplitude of Spartina patens. I. Morphometric and physiological traits. Evolution, 33,1114-1127. Snaydon, R. W. (1962a). Microdistribution of Trifolium repens L. and its relation to soil factors. Journal of Ecology, 50,133-143. Snaydon, R. W. (1962b). The growth and competitive ability of contrasting populations of Trifolium repens L. on calcareous and acid soils. Journal of Ecology, 50,439-447. Snaydon, R. W. (1971). An analysis of competition between plants of Trifolium repens L. collected from contrasting soils. Journal of Applied Ecology, 8,687-697. Snaydon, R. W. (1978). Genetic change in pasture populations. Plant Relations in Pastures (Ed. by J . R. Wilson), CSIRO. East Melbourne. Snaydon, R. W. (1985). Aspects of the ecological genetics of pasture species. Structure and Functioning of Plant Populations. 2. (Ed. by J . Haeck & J . W. Woldendorp), North Holland Publishing Company, Amsterdam. Snaydon, R. W. & Bradshaw, A. D. (1962a). The performance and survival of contrasting natural populations of white clover when planted into an upland Festuca I Agrostis sward. Journal of the British Grassland Society, 17,113-118. Snaydon, R. A. & Bradshaw, A. D. (1962b). Differences between natural populations of Trifolium repens L. in response to mineral nutrients. I. Phosphate. Journal of Experimental Botany, 13,422-434. Snaydon, R. W. & Bradshaw, A. D. (1969). Differences between natural populations of Trifolium repens L. in response to mineral nutrients. II. Calcium, magnesium, and phosphorus. Journal of Applied Ecology, 6,185-202. Snaydon, R. W. & Davies, M. S. (1972). Rapid population differentiation in a mosaic environment. II. Morphological differentiation in Anthoxanthum / 79 odoratum. Evolution, 26,390-405. Snaydon, R. W. & Davies, M. S. (1976). Rapid population differentiation in a mosaic environment. IV. Populations of Anthoxanthum odoratum at sharp boundaries. Heredity, 37,9-25. Snaydon, R. W. & Davies, T. M. (1982). Rapid divergence of plant populations in response to recent changes in soil conditions. Evolution, 36,289-297. Soane, I. D. & Watkinson, A .R. (1979). Clonal variation in populations of Ranunculus repens. New Phytologist, 82,557-573. Sokol, R. R. & Rohlf, F. J . (1981). Biometry, (2nd ed.). Freeman, San Fransisco. Solangaarachchi, S. M. (1985). The nature and control of branching pattern in white clover (Trifolium repens L.). Ph.D. Thesis, University of Wales. Spieth, P. T. (1979). Environmental heterogeneity: A problem of contradictory selective pressures, gene flow, and local polymorphism. American Naturalist, 113,247-260. Thorhallsdottir, T. E. (1983). The dynamics of a grassland community with special reference to five grasses and white clover. Ph.D. Thesis, University of Wales. Turesson, G. (1922). The genecological response of the plant species to the habitat. Hereditas, 3,211-350. Turkington, R. (1979). Neighbour relationships in grass-legume communities. IV. Fine scale biotic differentiation. Canadian Journal of Botany, 57,2711-2716. Turkington, R. (1983a). Leaf and flower demography of Trifolium repens L. I. Growth in mixture with grasses. New Phytologist, 93,599-616. Turkington, R. (1983b). Leaf and flower demography of Trifolium repens L. II. Locally differentiated populations. New Phytologist, 93,617-631. Turkington, R. (1983c). Plasticity in growth and patterns of dry matter distribution of two genotypes of T rifolium repens grown in different environments of neighbours. Canadian Journal of Botany , 61,2186-2194. Turkington, R. (1985). Variation and differentiation in populations of Trifolium repens in permanent pastures. Studies on Plant Demography: A Festschrift for John L. Harper (Ed. by J . White), pp. 69-82. Academic Press, London. Turkington, R. & Aarssen, L. W. (1984). Local scale differentiation as a result of competitive interactions. Perspectives on Plant Population Ecology (Ed. by R. Dirso & J . Sarukhan), Sinauer, Sunderland, Mass. Turkington R. & Burdon, J . J . (1983). The Biology of Canadian Weeds. 57. Trifolium repens L. Canadian Journal of Plant Science, 63,243-266. / 80 Turkington, R., Cahn, M. A. Vardy, A. & Harper, J . L. (1979). The growth, distribution, and neighbour relationships of Trifolium repens in permanent pastures. HI. The establishment and growth of Trifolium repens. in natural and disturbed sites. Journal of Ecology, 67,231-243. Turkington, R. & Harper, J . L. (1979a). The growth, distribution, and neighbour relationships of Trifolium repens in permanent pastures. I. ordination, pattern, and contact. Journal of Ecology, 67,201-218. Turkington, R. & Harper, J . L. (1979b). The growth, distribution, and neighbour relationships of Trifolium repens in permanent pastures. II. Inter- and intra-specific contact. Journal of Ecology, 67,219-230. Turkington, R. & Harper, J . L. (1979c). The growth, distribution, and neighbour relationships of Trifolium repens in permanent pastures. IV. Fine scale biotic differentiation. Journal of Ecology, 67,244-254. Warwick, S. I. & Briggs, D. (1979). The genecology of lawn weeds. III. Cultivation experiments with Achillea millifolium L., Bellis perennis L., Plantago lanceolata L., Plantago major L., and Prunella vulgaris L. collected from lawns and contrasting grassland habitats. New Phytologist , 83,509-536. Watson, P. (1969). Evolution in closely adjacent plant populations. VI. An entomophilous species, Potentilla erecta, in two contrasting habitats. Heredity, 24,407-422. Whitham, T. C. & Slobodchikoff, C. N. (1981). Evolution by individuals, plant herbivore interactions, and mosaics of genetic variability: The adaptive significance of somatic mutations in plants. Oecologia(Berl), 49,287-292. Williams, G. C. (1975). Sex and Evolution. Princeton University Press, Princeton. Wilson, J . R. (1978). (Ed.) Plant Relations in Pastures. CSIRO, Melbourne. 


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