1001Ecology,84(4), 2003, pp. 1001–1011q2003 by the Ecological Society of AmericaMAPPING THE ASSEMBLY OF PROTIST COMMUNITIES IN MICROCOSMSPHILIPH. WARREN,1,3RICHARDLAW,2ANDANITAJ. WEATHERBY1,41Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN UK2Biology Department, York University, P.O. Box 373, York, YO10 5YW UKAbstract.Theoretical models of community assembly predict a variety of possibleassembly phenomena, but evidence for them in natural systems is hard to obtain becauseof problems of temporal scale and replication. We used laboratory experiments toinvestigatethe potential assembly pathways in a small pool of protist species. Mapping the paths ofassembly involved two steps: testing the ability of every possible species combination topersist, and testing each persistent community for its capacity to be invaded and changedby every other species from the pool. We then compared the properties of assembly in theexperimental system with those predicted by theoretical models.We found that the number of community states and assembly paths was much smallerthan theoretically possible, but that the system nonetheless had quite a complex range ofassembly behaviors. There were many alternative paths to most communities, but overall,the species pool had just one state to which the paths eventually led. This absorbing stateinvolved an oscillation between two communities, driven in part by a catalytic species thatcould persist in neither of them. Catalytic species have the property of invasion, changingthe resident community, and then going extinct. In the species pool used here, only thepredatory species acted as catalysts. Several communities capable of long-term persistencewere unreachable by sequential assembly. One community had the property that it couldnot be reassembled from the species it contained (a ‘‘Humpty-Dumpty’’ community).These results are for the most part in keeping with those from numerical simulationsof community assembly, although there are certain discrepancies and new phenomena. Theresults also highlight the potential importance of long-term transient dynamics on assemblyin natural systems.Key words: community assembly; invasion; microcosm; protist.INTRODUCTIONThe arrival of a new species in a community maycause changes in species composition both through theaddition of the species itself to the community and alsothrough extinction of species already present. A seriesof such changes, brought about by sequential arrivalof species from a regional pool, can be thought of asa process of community assembly (Luh and Pimm1993, Law 1999). Viewed in this way, community as-sembly emphasizes changes in the community state,rather than embracing all evidence for pattern in com-munity structure, the broader context in which the termassembly is sometimes used (e.g., Diamond 1975,Drake 1990a,Belyea and Lancaster 1999, Weiher andKeddy 1999). Knowledge of the paths by which com-munities are assembled helps ecologists to understandthe role of history in shaping current communities, andmay be important for community restoration (Drake1990a,Pimm 1991, Lockwood and Pimm 1999, Weiherand Keddy 1999).Manuscript received 27 September 2001; revised 6 August2002; accepted 17 August 2002. Corresponding Editor (ad hoc):M. Holyoak.3E-mail: [email protected] address: The Ponds Conservation Trust: Policyand Research, c/o Oxford Brookes University, Gipsy Lane,Headington, Oxford OX3 0BP UK.Our aim here is to describe the network of assemblypaths (which we term the assembly graph) for a poolof six protist species in a laboratory system, using datafrom two experiments that tested the persistence andinvasibility properties of species combinations fromthepool. The experiments assume a separation in timescales, invasion by new species being rare events onthe time scale of population dynamics. This makes ex-periments feasible because resident population densi-ties should be relatively close to an attractor when newspecies arrive, rather than at arbitrary densities thatoccur during the transient dynamics. The first experi-ment (Weatherby et al. 1998) examined all possiblecombinations of species from the pool and identifiedthe sets of species able to persist in the long term. Thesecond experiment challenged each of these persistentcommunities with separate introductions of each of theother species, to see which species could invade (Lawet al. 2000) and to determine the species compositionto which the new community settled. By putting to-gether the results from these two experiments, it ispossible to map all the potential paths for communityassembly as a set of persistent communities with in-vasions driving the transitions between states. We be-lieve this is the first time a complete assembly graphhas been constructed in this way, although see Zim-mermann et al. (2003) for a slightly different approach1002PHILIP H. WARREN ET AL.Ecology, Vol. 84, No. 4to the same problem, using data from several differentexperiments.We use the assembly graph to address a number ofquestions about possible assembly behaviors. Thequestions stem largely from numerical analysesof com-munity assembly, which make use of several methodsto find the new community following an invasion: sta-tistical likelihood of transitions between states (Luhand Pimm 1993), numerical integration (Case 1990),existence of a locally stable equilibrium point (e.g.,Post and Pimm 1983, Drake 1990b), and permanence(Law and Morton 1993, 1996). Permanence is a globalcriterion for persistence of a set of species, roughlythat the density of each tends to increase when thespecies is sufficiently rare (for a formal definition, seeHofbauer and Sigmund [1988:97, 160]). The questionswe address are as follows:To what extent are species that can initially invadea system present in the resulting persistent communi-ty?—An invader here is a speciesthat initiallyincreaseswhen introduced in small numbers; it may or may notpersist in the longer term. The resulting persistent com-munity is the set of species to which the system settlesin the long term. Numerical studies of Lotka-Volterrasystems using permanence found that invaders are pre-sent in the resulting persistent community (Law andMorton 1996); in a broader context this property mayor may not hold.Are there catalytic species?—These are species thatinvade, change the community, and then go extinct,never forming part of a persistent community arisingfrom their