I used Box–Jenkins transfer function models to analyze the relationship between water temperature and Maine lobster catch and catch-per-unit-effort (CPUE). I first modelled catch and CPUE with univariate autoregressive – integrated moving average (ARIMA) models to provide a basis for comparison with transfer function models. Time series models were constructed for annual Maine lobster landings during two periods: 1928–85 and 1945–85; catches during the latter period were assumed to be less dependent on changes in fishing effort. I also modelled annual CPUE for the period 1930–85 and monthly landings for 1968–85. Landings and CPUE for 1986 were held in reserve to check forecast estimates. An immediate temperature effect (lag 0–1 yr) was demonstrated for each annual series. This result is consistent with known aspects of lobster biology; activity levels and hence vulnerability to capture increase with water temperature. In addition, the probability of molting increases with increasing water temperature, affecting the short-term supply of legal-sized lobsters. A significant effect of temperature at a 6-yr lag was also indicated, but only for the 1945–85 catch series. Delayed effects of this type may indicate environmental influences on natality or survival during early life history stages. Time series models for the Maine lobster fishery provided forecasts for 1986 catch and CPUE which differed by no more than 4% of the actual 1986 levels.