Low temperatures during the flowering period of cereals can lead to floret sterility and yield reduction, resulting in economic losses in Australian crops. In this study we investigated the physiological bases of yield determination in a historic set of wheat cultivars grown under frost-prone field conditions in Southern Australia. We tested the hypothesis that selection for yield inadvertently improved frost tolerance. We measured yield and yield components, including the distribution of grains within the spike, in a factorial experiment combining twelve wheat cultivars (released from 1973 to 2015), two sowing dates (19th April and 22nd May 2017), and two thermal regimes, ambient control and frost-protected. To protect crops from frost we used moveable, lightweight passive heating systems before each frost event ( 0°C). Phenotypic plasticity of yield, grain number and grain weight were analysed. We found a positive relationship between phenotypic plasticity of grain yield and phenotypic plasticity of grain number, but no correlation between yield and year of release. Across cultivars, the average numbers of grains per spike was 35.1 ± 0.3 in frost-protected crops compared to 16.1±0.1 in frosted controls in the first sowing, and 29.7 ± 0.3 and 22.9 ± 0.2 respectively in the second sowing. Preventing frost improved spike fertility by increasing the proportion of grains in distal positions within spikelets in relation to controls.