Ecology, 89(8), 2008, pp. 2172–2180 Ó 2008 by the Ecological Society of America PERTURBATIONS ALTER COMMUNITY CONVERGENCE, DIVERGENCE, AND FORMATION OF MULTIPLE COMMUNITY STATES GREGORY R. HOUSEMAN, 1,5 GARY G. MITTELBACH, 1,2 HEATHER L. REYNOLDS, 1,3 AND KATHERINE L. GROSS 1,4 1 W. K. Kellogg Biological Station, Michigan State University, Hickory Corners, Michigan 49060-9516 USA 2 Department of Zoology, Michigan State University, East Lansing, Michigan 48824 USA 3 Department of Biology, Jordan Hall 142, Indiana University, Bloomington, Indiana 47405-7005 USA 4 Department of Plant Biology, Michigan State University, East Lansing, Michigan 48824 USA Abstract. Environmental perturbations (e.g., disturbance, fertilization) commonly shift communities to a new mean state, but much less is known about their effects on the variability (dispersion) of communities around the mean, particularly when perturbations are combined. Community dispersion may increase or decrease (representing a divergence or convergence among communities) if changing environmental conditions alter species interactions or magnify small initial differences that develop during community assembly. We used data from an experimental study of disturbance and fertilization in a low-productivity grassland to test how these two perturbations affect patterns of species composition and abundance. We found that a one-time biomass reduction decreased community dispersion, which persisted over four growing seasons. Conversely, continuous fertilization increased community dispersion and, when combined with disturbance, led to the formation of three distinct community states. These results illustrate that perturbations can have differing effects on community dispersion. Attention to the variance in community responses to perturbations lends insight into how ecological interactions determine community structure, which may be missed when focusing only on mean responses. Furthermore, multiple perturbations may have complex effects on community dispersion, yielding convergence or divergence patterns that are difficult to predict based on analysis of single factors. Key words: alternate states; community analysis; community assembly; dispersion; disturbance; fertilization; grasslands, southwest Michigan, USA; perturbation; priority effects. INTRODUCTION Predicting how communities respond to changing environmental conditions is an important goal of ecology that has become even more pressing in the face of increasing environmental stressors associated with global change (Balmford and Bond 2005). While most theoretical and empirical studies of communities focus on the mean response of communities to changing environmental conditions, the variability in community responses may also depend on environmental context (Warwick and Clarke 1993, Suding et al. 2004). We know relatively little, however, about the nature of this variation in community response. Variance may occur in both independent and depen- dent variables, with potentially important consequences for ecological communities (Palmer and Dixon 1990, Adler et al. 2001, Benedetti-Cecchi 2003). Spatial or temporal variability in an independent variable can alter population dynamics (Nisbet and Gurney 1982), affect species interactions (Chesson 2000, Brassil 2006), and lead to changes in community composition and stability (Collins 2000). For example, temporal variation in rainfall patterns may alter species diversity in grasslands (Knapp et al. 2002). Variance in a dependent variable can provide insight into how and why a system responds to ecological processes (Benedetti-Cecchi 2003). For exam- ple, patterns of spatial heterogeneity may reveal scale- dependent processes (Palmer et al. 1997, Adler et al. 2001, Collins and Smith 2006), and the degree of temporal and spatial variation observed among exper- imental replicates measures the predictability of com- munity and ecosystem responses to environmental perturbations (Carpenter and Brock 2006). Community responses to environmental perturbations are often expressed as changes in species composition and abundance, and these multi-species responses can be quantified using a variety of multivariate techniques (McCune and Grace 2002). For example, community similarity based on species composition and abundance is commonly used to examine the mean response (group centroid in multivariate space), however, it also can be used to quantify the variation (dispersion of replicates around the group centroid). Perturbations may cause a change in the mean community state with no effect on dispersion (i.e., all replicate communities respond similarly; Fig. 1A). Alternatively, an environmental perturbation may decrease or increase community Manuscript received 26 July 2007; revised 29 November 2007; accepted 2 January 2008. Corresponding Editor: N. J. Gotelli. 5 Present address: Department of Ecology and Evolution- ary Biology, University of Kansas, Lawrence, Kansas 66045 USA. E-mail: houseman@ku.edu 2172