To assess the contention of many people with diabetes that they gain weight in winter and lose weight in summer we have investigated 10836 weights recorded during all the routine clinic visits of 3119 patients (age 18-93 years, 51.7% male, 31.2% diet controlled, 32.8% on oral hypoglycaemic agents, 36.0% on insulin)between February 1992 and July 1995. Efforts to fit a sinewaveform to the weight data (raw and log transformed) plotted against time of the year were unsuccessful and a smoothed mean weight plotted over time closely matched the overall mean. Linear modelling was applied to data from each clinic visit relating log transformed weight to age, gender, diabetes therapy, random plasma glucose and the interactions of these terms. The best fitting model accounted for 23.3% of the observed variability with gender, diabetes therapy, age, age squared and interactions between these as the key factors; interpatient variation accounted for over 95% of the residual variance from the model. Addition of time of year as a continuous variable or as 12 discrete factors (months) did not improve the fit of the model. We conclude that there is no general tendency to seasonal weight variation in people with diabetes.