(1) The data of the budburst of lateral shoots on young Picea sitchensis (Bong.) Carr. trees can be estimated from mean daily temperatures using the model: T = a + b exp (rC), where T is day degrees > 5.degree. C accumulated from 1 February and C is number of chill days .ltoreq. 5.degree. C counted from 1 November (Cannell and Smith 1983). (2) Clonal differences in the relative values of a, b and r were determined by (i) bringing ramets of twelve clones into a warm glasshouse after they had experienced different numbers of chill days, and measuring clonal differences in the form of the relationship between T and C, and (ii) calculating the actual T and C values corresponding to the dates of budburst of different clones growing in the field. It was found that r values were very similar between clones, so clonal differences could be simulated by changing a and b. (3) Using different values of a and b in the model, dates of budburst, and hence return times of potentially damaging frosts at budburst, were estimated at Eskdalemuir and Braemar (Scottish upland meteorological sites with long records). Budburst could readily be delayed by 7 days by selecting late-flushing clones, but this would only increase the return time of frost damage from once every 3 to once every 4 years, owing to the high incidence of late frosts. (4) Early-flushing clones were no more likely to be frost damaged than clones with average dates of flushing. The T and C requirements of early-flushing clones tended to adjust their dates of budburst to avoid frosts that would be encountered if budburst occurred on the same day each year.