Previous studies have detected an effect of earlier temperatures on the incidence of campylobacteriosis in humans, but without adjustment for earlier numbers of cases of the disease. We estimated the effect of temperature on the number of cases notified by week in Montreal, Canada, from 1 January 1990 to 26 March 2006, simultaneously with the effect of the numbers of cases notified in the preceding weeks. The current campylobacteriosis count (week 0) was modelled by negative binomial regression, with earlier weekly average temperatures and earlier counts as predictors. Secular trends were accounted for by cubic spline functions and seasonal variations by sine-cosine functions. Indicator variables identified weeks with fewer than 5 working days. In the final statistical model, a 1°C increase in temperature above 10°C during any of weeks -1 to -6 was associated with a 0.8% (95% CI: 0.3% to 1.3%) increase in the current count. For each additional notified case during any of weeks -1 to -5 or -9 to -12, the increase in the current count was approximately 0.5% (95% CI: 0.2% to 1.0%). Thus, earlier temperatures and earlier counts have independent effects, that of temperatures being the larger one. The temperature effect is too small to require short term public health planning. However, in Montreal, an increase in average temperature of the order of 4.5°C, forecast by some for 2055, could produce a 23% increase in incidence, resulting in about 4,000 excess cases per year.