JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Submitted April 9, 2007; Accepted September 13, 2007; Electronically published December 14, 2007 Online enhancement: appendix.abstract: Food web assembly algorithms show great promise for investigating issues involving the dynamics of whole webs, such as succession, rehabilitation, and invasibility. Permanence, which requires that all species densities remain positive and finite, has been suggested as a good stability constraint. This study tests the validity of the permanence constraint by comparing real webs and model webs from the literature to the predictions of three assembly algorithms: one constrained by permanence and feasibility, one constrained by feasibility alone, and one with no dynamical constraint. It is found that the addition of the permanence constraint does not improve the predictive ability of the algorithm. Its main effect is to increase the efficiency of species selected for the web. Dynamically constrained webs have lower connectance and indistinct trophic levels compared to real webs and webs from other models, which is a consequence of omitting species' physiology. Although webs are less likely to be permanent if they have high omnivory and cycling, the web-building process circumvents this constraint. The challenges of testing and justifying system-level hypotheses, including isolating and detecting their effects, are discussed.Keywords: food webs, stability, permanence, feasibility, niche model, nested-hierarchy model.A food web-building algorithm is a series of repeating steps by which model food webs can be created. Starting with one or a few species, the algorithm permits new species to invade and some species to be lost at each time Weatherby et al. 1998;Chase 2003;Fukami 2004). Because assembly algorithms simulate invasion and extinction in a whole food web, they hold great promise for answering questions about succession, rehabilitation, and invasibility. The rules governing which food webs survive intact to the next time step are often based on dynamical constraints (though not always; e.g., Luh and Pimm 1993). For example, food webs may be constrained to be feasible (i.e., all species have a positive steady state biomass [e.g., Tregonning and Roberts 1979]) or locally stable (e.g., Post and Pimm 1983). This article focuses on a dynamical constraint called permanence. A food web is permanent if the densities of all species remain positive and finite and it can recover after those densities are brought arbitrarily close to 0 (Hofbauer and Sigmund 1988). Permanence has the advantage over local stability in that it does not presuppose the existence of a single steady state but does allow for complex trajectories within the phase space, such a...