Recruitment algorithms in forest gap models are examined with particular regard to their suitability for simulating forest ecosystem responses to a changing climate. The traditional formulation of recruitment is found limiting in three areas. First, the aggregation of different regeneration stages (seed production, dispersal, storage, germination and seedling establishment) is likely to result in less accurate predictions of responses as compared to treating each stage separately. Second, the relatedassumptions that seeds of all species are uniformly available and that environmental conditions are homogeneous, are likely to cause overestimates of future species diversity and forest migration rates. Third, interactions between herbivores (ungulates and insect pests) and forest vegetation are a big unknown with potentially serious impacts in many regions. Possible strategies for developing better gap model representations for the climate-sensitive aspects of each of these key areas are discussed. A working example of a relatively new model that addresses some of these limitations is also presented for each case. We conclude that better models of regeneration processes are desirable for predicting effects of climate change, but that it is presently impossible to determine what improvements can be expected without carrying out rigorous tests for each new formulation.