An Overview: Real and Apparent Patterns in Community Structure

ROBERT M. MAY

Biology Department, Princeton University, Princeton, New Jersey 08544

"Two noticeable characteristics of papers recently published are the widespread interest in field quantitative methods in the study of population density, rates of spread, fluctuation, reproduction, feeding, or mortality; and an increasing awareness of evolutionary dynamic problems such as selection and competition At the same time there is a general adoption and a tightening up of the statistical treatment of ecological data, which, though entirely sound and necessary, would become a partly bad feature if it tended to exclude the equally valuable type of observations on the pattern of nature, and the habitats and distribution of animals, that ecologists and naturalists can contribute." (Elton, 1947)

INTRODUCTION

In 1944, the British Ecological Society held a symposium on "The Ecology of Closely Allied Species," at which Lack, Elton, Varley, and others used various lines of evidence to argue that competition is a major factor in structuring plant and animal communities. Others argued to the contrary, with Diver contending that "the mathematical and experimental approaches had been dangerously oversimplified and omitted consideration of many factors [including] sources of energy and their relative availability, predator attack, mobility, population structure and growth, individual growth rate and bulk, relation of life cycle to annual cycle, range of tolerance, means of dispersal, and the like" (Anon., 1944). He concluded "there was little direct evidence that cohabitation or separation of related species was determined by space and food, since other factors usually kept populations below the point at which serious pressure was developed." Broadly similar themes dominated the celebrated Cold Spring Harbor Symposium in 1957, with some arguing that density-dependent effects arising from biological interactions are of predominant importance in setting population levels, while others argued the importance of the density-independent regulatory effects caused by the weather and other environmental factors. The Brookhaven Symposium of 1969 on "Diversity and Stability in Ecological Systems" again drew together many of the contemporary leaders of the subject; I think it gave a less polarized and more synthetic account of the issues, although (as the title itself suggests) there may have been too much of a tendency to view communities as orderly, patterned "systems." The present volume stems from a conference held at Wakulla Springs in 1981, and the same themes still interweave, albeit now greatly enriched by a rapidly expanding body of field observations, carefully planned experimental manipulations in the field and laboratory, and more rigorous techniques of statistical evaluation of the data. Whether these themes are drawing toward their resolution, or whether we are still in the opening passages of the work, is for the reader of this volume to decide.

An eager and naive pattern-seeker might note that these landmark meetings are regularly spaced, with a 12-year period, and might even go on to speculate on the underlying cause of this cycle (12 years is roughly the time from entering graduate school to the tenure decision?). This is silly. The "cycle" does, however, serve to illustrate one central concern of the Wakulla Springs Conference: given some apparent pattern in the organization of an ecological community, does it really derive from biological interactions among and within species? Or is it the sort of coincidence one often finds when the data are few? Or may the investigator have produced it, unconsciously, by making observations or designing experiments to conform to a preconceived notion? Or may the pattern simply be a statistical property of the system — a true pattern, but having no biological significance? Such questions, involving the disentangling of real from apparent patterns, occur in many other areas of science (see, for example, the debate between Arp and Bahcall (1973) on whether there are, or are not, significant spatial associations between astronomical objects with large but different red shifts); the questions are rarely easy to answer.

The apparent 12-year cycle is spoiled by, inter alia, the meeting on the "Ecology and Evolution of Communities" held in 1973 as a memorial to Robert MacArthur (Cody and Diamond, 1975a). Unlike the more wide-ranging meetings mentioned above, this one was mainly concerned with those areas, and that style, where MacArthur's own contributions had been so stimulating: field observations and theoretical models aimed at understanding how communities are structured by biological interactions, particularly competition. In some ways, the 1981 Wakulla Springs meeting represents a healthy reaction against too enthusiastic and uncritical an acceptance of some of the "pattern-seeking" field and theoretical studies of the two past decades. One technique, set out in some recent papers in this volume and elsewhere, is to construct "neutral models" or "null hypotheses," which aim to elucidate those apparent patterns that might be exhibited when comparisons are made among a given set of communities, assuming that some, or all, classes of biological interactions are absent. Insofar as comparative studies of real communities yield patterns beyond those found in such neutral models, one can be more confident that community structure is indeed being forged by particular kinds of interactions among species. More specifically, neutral models are being employed in an effort to see whether biologically based patterns must necessarily be attributed to competition, or whether predation or other effects (including, I would emphasize, pathogens or parasites) could equally well be the cause.

Although the phrases null/neutral hypothesis/model are characteristic of the relatively recent literature, the essential ideas have been applied in various contexts in community ecology over the years. (Indeed, any assignment of a significance level to a regression line is explicitly a statement about the rejection of a null hypothesis.) Unfortunately, an appropriate null hypothesis is not always easy to construct; in particular, a proper neutral model may require data that simply are not available. It is just as easy — and just as foolish — to construct an inappropriate or misleading neutral model as it is glibly to deduce evidence for competition from data that are susceptible to other interpretations.

In what follows, I briefly outline a miscellany of examples in which (whether or not the contemporary phraseology is employed) neutral models have been used in the hope of elucidating theoretical or empirical aspects of community structure; some of these examples are covered in more detail elsewhere in this volume, although most are not. Most of these vignettes are complex, and do not admit of Manichean division into white-hatted and black-hatted people. To my mind, no simple moral emerges from these tales, other than the broad injunction that alternative explanations in general, and appropriate neutral models in particular, should always be kept in view when experiments are designed or data analyzed.

I must emphasize, most strongly, that the contents of this introductory chapter do not accurately reflect the amounts of time spent on various topics at the conference itself. One of the main aims of the conference was to focus on analytically designed field studies that test theoretical ideas. Many of the papers did just this, presenting interesting and previously unpublished data (as, for example, in the chapters by Lawton, by Strong, by Rey, and by Grant and Schluter). Although such case studies predominated at the conference, and form the bulk of this consequent book, many participants' clearest memories will be of the disagreements — good-humored but nonetheless sharp — over theoretical and methodological issues; my chapter dwells exclusively on these issues.

NUMBER OF SPECIES PER GENUS

At the British Ecological Society meeting in 1944, Elton presented an analysis of 55 animal communities (including some parasite ones) and 27 plant communities, each from a relatively small geographical area. He showed that, in these communities, the average number of species per genus (the "S/G ratio") was markedly smaller than that for faunal lists from any large region, and attributed this difference to "existing or historical effects of competition between species of the same genus, resulting in a strong tendency for the species of any genus to be distributed as ecotypes in different habitats, or if not, to be unable to coexist permanently on the same area of the same habitat" (Elton, 1946).

C. B. Williams (1964), however, subsequently pointed out that it is a property of the statistical distribution of species among genera that, as the number of species and genera in a sample decrease (as they will when one goes from a larger region to a smaller), the ratio S/G will decrease. More recently, Grant (1966b) and Moreau (1966) have sought to find evidence for competition in the smaller S/G ratios observed for birds on islands or in restricted habitats, while Simberloff (1970) has given an incisive analysis (including extensive numerical simulations) to show that just such decreases in S/G with decreasing S are mathematical properties of the S-G distribution.

This cautionary tale is fairly straightforward. Although the observed S/G pattern appears to be just what one would expect if communities are structured by competition, closer examination shows the pattern to be primarily a statistical artifact, a mathematical property of the way the average S/G ratio varies with S. Insofar as the observed S/G patterns do differ slightly from mathematical expectation, the S/G ratios in restricted habitats appear to be relatively high rather than relatively low (the data points tend to lie slightly above the line derived from the null hypothesis); a more full discussion is given by Strong (1980) and by Simberloff in this volume. Notice that, as Simberloff has repeatedly emphasized, the explanation of the S/G pattern by a null hypothesis does not mean that competition is necessarily unimportant in determining which species co-occur in the communities studied by Elton, Grant, Moreau, and others; rather, it means this particular line of inquiry simply sheds little light.

STABILITY AND COMPLEXITY

The idea that complex ecosystems, with many species and a rich web of interactions, should be more stable than simple ones is an intuitively appealing one; it may seem that a community is better able to cope with disturbance if there are many alternative pathways along which energy and nutrients may flow. Elton (1958) advanced a set of six arguments in support of this notion that complexity begets stability. One of the six was a theoretical argument and consisted of the observation that mathematical models of simple prey-predator associations exhibit instability (Elton had in mind the neutrally stable Lotka-Volterra model for one prey and one predator, and the unstable Nicholson-Bailey model for host-parasitoid interactions). Whatever the status of the other five arguments (May, 1973, pp. 37–40, 173), this theoretical observation is meaningless until one has determined the stability properties of the analogous models with many predators and many prey. Such multispecies models turn out, in general, to be less stable the more species are present. That is, increasing dynamical stability is not a general mathematical consequence of increasing complexity; rather, the contrary is true.

I think this example belongs in a broad discussion of the uses of null hypotheses, because it provides an illuminating instance where an attractive idea was long accepted (and still is in many Introductory Biology texts) on the basis of logically incomplete arguments. Real communities, of course, are not random selections from the universe of general mathematical models, and the current task is to try to understand the special structural features that complex ecosystems may possess to help them reconcile stability with complexity. Are apparently complex tropical ecosystems actually constituted of many loosely coupled subsystems (Gilbert, 1977; Root, 1973)? Do dynamical considerations constrain the length of trophic chains (Pimm and Lawton, 1977; DeAngelis et al., 1978; Lawton and Pimm, 1978)? Is "donor control" (DeAngelis, 1975), or the character of predators' functional responses (Nunney, 1980), or some other feature, crucial in distinguishing real food webs from those that may seem possible in general? Or may it be that complex ecosystems really are typically more fragile, being found only in environments where disturbances are typically less severe or more localized than is the case for simpler ecosystems (Wolda, 1978; May, 1979)?

The example is also interesting for the light it sheds on the generation of hypotheses, null and otherwise. It is lunacy to imagine that the dynamical behavior of real communities bears anything but the vaguest metaphorical relation to the linearized stability properties of the conventional "community matrix" (Levins, 1975; May, 1973). But analyses of abstract community matrices have led to the generation of new ideas and the framing of testable hypotheses, such as those about the patterns of connectance in real food webs (Yodzis, 1980; Rejmánek and Stary, 1979), about the lengths of trophic chains, about the structuring of communities in terms of subunits or guilds, and so on. Some of this work is developed more fully in the chapters by Pimm, Auerbach, and Lawton.