Measurements of the thermal phyllochron in wheat using air temperature show a systematic variation with sowing date, with long phyllochrons for crops sown in autumn and winter, but shorter phyllochrons for sowings outside that range. Based on the hypothesis that the phyllochron is constant, two sources of error for the apparent variation in phyllochron are investigated, namely the site of temperature measurement (air or soil) and the implicit assumption that leaf appearance rate responds linearly to temperature. Measurements of the phyllochron from three sowings per annum for 6 years, from controlled environments, and from a sequence of sowings through an annual cycle are presented. The field results show similar systematic variation to that found elsewhere. However, for the annual cycle, calculations of the phyllochron based on soil temperature at 2 cm depth were much more stable than those based on air temperature. In addition, leaf appearance rate was linearly related to 2 cm soil temperatures across all environments, with a 0 °C intercept, and the optimum temperature was still not reached in the highest temperature treatment (30.9 °C soil temperature). Following earlier published work, we then show that the assumption that the phyllochron is constant, but that the controlling temperature at the apex is different from air temperature, requires a correction of less than 1 °C when sowing dates do not stray too much outside the range of normal practice. We show that the correction to air temperature is systematically related to sowing date, and a suitable correction can be made by using daylength some time after emergence as a predictive surrogate for seasonal variation in the air–apex temperature difference. We suggest that a model using such a predictor will improve predictions of leaf appearance rate over earlier functional models. However, we suggest that the observed response is caused by the physics of the system rather than the biology.