1. According to the ideal-free distribution (IFD), individuals within a population are free to select habitats that maximize their chances of success. Assuming knowledge of habitat quality, the IFD predicts that average fitness will be approximately equal among individuals and between habitats, while density varies (density-dependent habitat selection). Populations are often assumed to follow an IFD, although this assumption is rarely tested with empirical data, and may be incorrect when territoriality indicates habitat selection tactics that deviate from the IFD (e.g. ideal despotic distribution or ideal preemptive distribution). 2. When territoriality influences habitat selection, species’ density will not directly reflect components of fitness such as reproductive success or survival. In such cases, assuming an IFD can lead to false conclusions about habitat quality. We tested theoretical models of density-dependent habitat selection on a species known to exhibit territorial behavior in order to determine whether commonly applied habitat models are appropriate under these circumstances. 3. We combined long-term radio telemetry and census data from gray wolves (Canis lupus) in the Upper Peninsula of Michigan, USA to relate spatiotemporal variability in wolf density to underlying classifications of habitat within a hierarchical state-space modeling framework. We then iteratively applied isodar analysis to evaluate which distribution of habitat selection best described this recolonizing wolf population. 4. The wolf population in our study expanded by >1000% during our study (~ 50 to > 600 individuals), and density-dependent habitat selection was most consistent with the ideal preemptive distribution, as opposed to the ideal-free or ideal-despotic alternatives. 5. Population density of terrestrial carnivores may not be positively correlated with the fitness value of their habitats, and density-dependent habitat selection patterns may help to explain complex predator-prey dynamics and cascading indirect effects. Source-sink population dynamics appear likely when species exhibit rapid growth and occupy interspersed habitats of contrasting quality. These conditions are likely and have implications for large carnivores in many systems, such as areas in North America and Europe where large predator species are currently recolonizing their former ranges.