We present statistical methods appropriate for the analysis of seedling recruitment data. The techniques are applied to the data collected by H. A. Roberts and his colleagues. The main result of this analysis is that seed banks do not in general decay according to a negative exponential pattern and that a wide range of recruitment patterns occur. We explore the effects of unobserved mortality on age-specific recruitment. We also demonstrate the difficulty of interpreting age-specific patterns of recruitment in terms of the behavior of individual seeds. This difficulty is a result of unobserved seed mortality and between-seed variability, either of which can qualitatively change the shape of the recruitment curve. A simple model of the recruitment process is presented, and it is demonstrated that density dependence can generate spurious patterns of age-specific recruitment.