1. Four models for predicting budburst in northern hardwood trees, based on response to spring warming alone, or with the response to spring warming modified by winter chilling and photoperiod, were compared. An historical, 18-year budburst record, and artificial datasets with budburst dates generated according to each of four conceptual models, were used to analyse the abilities of the models to predict budburst dates. 2. The four models all gave better predictions than could be obtained by taking the average date of budburst of a species. The historical budburst dates were most accurately predicted by models based only on spring warming from a fixed start date, or from a start date determined by the satisfaction of a chilling requirement