Null Models in Ecology. – book reviews
Kimberly A. With
Few methodologies in ecology have evoked as much debate, and of such a vitriolic nature, as has the use of null models in the discipline of community ecology. Null models have had such a controversial legacy in ecology primarily because the results of some of the early tests (e.g., communities are merely random assemblages of species drawn from some source pool) contradicted the ruling paradigm of the time in community ecology (i.e., competition theory). Far from fanning the coals of what some have probably hoped was a dead fire, Nicholas Gotelli and Gary Graves present null models as a basic tool for community ecologists in their recently published book, Null models in ecology. Thus, despite the title, the focus of the book is exclusively on community-level questions. The book is a vastly expanded, updated, and more in-depth treatment than the Harvey et al. (1983) review of the same name which was published during the height of the null model controversy (Harvey, P. H., R. K. Colwell, J. W. Silvertown, and R. M. May. 1983. Null models in ecology. Annual Review of Ecology and Systematics 14:189-211). Another text, Neutral models in biology (Nitecki, M. H., and A. Hoffman [eds.]. 1987. Oxford University Press, New York), is broader in scope, encompassing a variety of subjects from neutral models in genetics to paleontological models, but has only a single chapter devoted to community ecology.
Null models in ecology is an ambitious endeavor. Gotelli and Graves have synthesized a voluminous literature base ([greater than] 960 citations) to present a comprehensive analysis of the use and misuse of null models in community ecology. Given the pivotal, if controversial, role that null models have played in the field of community ecology, Gotelli and Graves begin by discussing the philosophical issues surrounding the null model approach in general, in terms of the principles of falsification and parsimony, and review the history and underlying criticisms of null models in community ecology from the earliest models of species/genus ratios in Chapter 1. Here they also attempt to clarify and operationalize the definition of a null model as “a pattern-generating model that is based on randomization of ecological data or random sampling from a known or imagined distribution. . . . The randomization is designed to produce a pattern that would be expected in the absence of a particular ecological mechanism.” Null models are viewed as “thought experiments,” which have a long tradition in the physical sciences, and provide a statistical baseline for comparison in studies where experimental manipulations are impossible or unwieldy and standard statistical procedures are lacking.
The construction of the null model (e.g., appropriateness of randomization algorithms and biological constraints in the model) is at the heart of the debate surrounding the utility of null models in community analysis. How null is null? A null or “neutral” model should not be taken to mean that it is free from bias; any model must make some assumptions and simplifications and these must be considered in terms of how they affect the utility of the model in a given context and thus the interpretation of model results. Wimsatt (1987. False models as means to truer theories. pp. 23-55 in M. H. Nitecki and A. Hoffman leds.]. Neutral Models in Biology. Oxford University Press, New York) commented on the irony that many of the same individuals who advocate the use of appropriate controls and design constraints in field and laboratory experiments condemn models for doing the same thing!
While Gotelli and Graves stop short of claiming that null models have “logical primacy” in the scientific method, they do claim that they should be investigated first “so that stochastic and sampling effects can be distinguished from biologically meaningful patterns” and, more practically, “because they may save a huge amount of time that could be frittered away in search of a nonexistent process or phenomenon.” Although I am an advocate of the null model approach, I fear such a statement is akin to stipulating that “analytical models must always be developed before a computer simulation model is attempted” or that “experimental manipulation is the only way to do robust science.” The appropriateness of the null model approach depends upon the type of question under investigation, and in this at least, Gotelli and Graves present a convincing argument for the potential and diversity of null model applications in community ecology.
The remaining nine chapters each focus on a specific area within community ecology that has employed null model analysis. A considerable portion of each chapter is devoted to a review of the subject, outlining the contributions and shortcomings of previous analyses, as a means of introducing the potential applications of null models to the question. Given that patterns of species diversity are central to community ecology and biogeography, Chapter 2 explores the use of rarefaction as a random sampling procedure for comparing expected species richness among different samples standardized for numbers of individuals or area. Chapter 3 highlights species abundance models (e.g., log normal, geometric series, broken-stick model) as null resource utilization models. Chapters 4 and 5 deal with the plethora of randomization techniques that have been developed to predict the levels of spatial and temporal niche overlap expected in the absence of competition (e.g., expected flowering phenologies of plants in the absence of competition for pollinators). Chapters 6 (size ratios) and 7 (species co-occurrence patterns, such as checkerboard distributions) comprise the bulk of the discussion in this text, because these issues contributed most of the controversy regarding the application of null models in community ecology. The passive sampling hypothesis is presented as a null model of the species-area relationship in Chapter 8 (i.e., the probability that a species occurs on an island is random and proportional to island area). Null models of biogeographic patterns of species occurrence along elevational gradients or among geographic ranges are featured in Chapter 9. Up to this point, almost all of the null models have been designed to test for pattern in the absence of one mechanism, competition. Chapter 10, on food webs, remedies this by focusing on connectance among trophic levels expected in the absence of predation or other biotic interactions, although this chapter also tackles the diversity-stability hypothesis and the null model structure of competition matrices. Null models have led the development of food web theory and subsequently their use in this arena has been less contentious.
Although this is not a “how to null” book, each chapter concludes with a series of recommendations to provide guidelines for those interested in applying null models to a particular question, which is a real boon to the uninitiated attempting to wade through the morass of models that may have been developed in a given area. In the Epilogue, Gotelli and Graves highlight several of the major shortcomings still facing null models in community ecology, such as the proper delineation of species source pools, the need to incorporate sexual size dimorphism and geographic variation in body size explicitly within null model analyses, problems concerning data quality, and recognition that human-caused extinctions on islands may bias the current biogeographical distributional patterns of some species. Finally, the authors cite the lack of commercially available null modeling software as a major deterrent to the widespread application of null models in ecology. This may be true, but widespread availability would increase the risk that null models will be used inappropriately, without any real understanding or appreciation for the underlying assumptions of the model structure which may constrain its application in a given context.
One of the reviewers of the proposal for this book is quoted as admitting that “since I am not a fan of the null model approach, it would not disappoint me if they fail in their enterprise.” I trust this reviewer was indeed disappointed by Gotelli and Grave’s success, who despite their obvious bias in favor of null models, were able to give an even-handed coverage of the major issues that have arisen in community ecology over the past 30 years. Whatever your stand on null models, this book will be of interest and is recommended reading for all community ecologists.
KIMBERLY A. WITH Bowling Green State University Department of Biological Sciences Bowling Green, Ohio 43403
COPYRIGHT 1997 Ecological Society of America
COPYRIGHT 2004 Gale Group
