A simple modelling approach was used to investigate the influence of habitat fragmentation on the distributions and regional population sizes of passerines breeding in 145 small woods in an arable landscape. Observed distributions in three consecutive years were compared with those predicted by random selection of woods by pairs (based on a Poisson distribution), and by a simulation model based on the area of habitat available and estimates of individual species minimum area requirements and territory sizes. None of the species studied were distributed randomly; some were distributed according to the area of woodland available, whereas others preferred either small or large woods. However, species distributions were influenced by their regional population sizes. When species population sizes were small, observed distributions could be predicted using algorithms based on area alone, but as numbers increased, algorithms using territory parameters were required. An absence of detectable territorial interaction suggested that space for more pairs of the species concerned had been available in our woods. Of the 14 species investigated, the models for only two consistently required the use of territory parameters, indicating that the available habitat could probably have accommodated more pairs of most of the species. Thus habitat fragmentation may reduce regional population sizes more than expected from the loss of habitat alone, to the extent that patches of suitable habitat may be unused, or occupied at low density.