• Campbell, Steven P.
  Department of Wildlife Ecology, University of Maine, Orono, Maine
  
• Witham, Jack W.
  Holt Research Forest, University of Maine, Arrowsic, Maine
  
• Hunter Jr., Malcolm L.
  Department of Wildlife Ecology, University of Maine, Orono, Maine
  

Stochasticity is rarely explicitly investigated as a determinant of patterns of habitat use, even though evidence of its influence would undermine support for deterministic models of habitat selection. To assess the role of stochasticity in generating patterns in habitat use, we compared observed patterns of year-to-year variation in habitat use of 20 bird species over five years with patterns randomly generated from null models with different biological constraints (i.e., abundance, territoriality, exclusion of unused habitat, site fidelity, and habitat preference). We were able to recreate patterns of variability in spatial distributions with one or more null models for 18 species, suggesting an important role of stochasticity. Although year- to-year variability could be modeled as a purely stochastic process for only four species, null models in which random variability was constrained by the influences of vegetation or site fidelity collectively matched the observed patterns of variability of all 18 species. Support for these two models suggested that factors that increase the likelihood of returning to a previously used area and thus limit the spatial extent of random territory placement can generate patterns similar to those we observed. Greater support for models with preference constraints (i.e., models that included bird–vegetation associations) over models with avoidance constraints (i.e., models that excluded areas that were never occupied) indicated that preference was the primary mode of habitat selection. Stochastic patterns in habitat use may have resulted from a weak relationship between habitat selection and habitat quality, influence of non-vegetation habitat features (e.g., climatic fluctuations), nonequilibrium conditions, and a greater influence of stochastic variation at small spatial and temporal scales. Alternatively, responses to environmental cues may have been deterministic, but in summation, these responses may have created variability in spatial distribution patterns that was indistinguishable from random variation. In such cases, we suggest that it may be easier to model variation in habitat use as a random process rather than using deterministic models that account for a multitude of environmental factors influencing habitat selection of birds. Regardless, stochasticity may play a larger role in generating patterns of habitat use than is often considered, and it deserves wider attention.