This data set contains raw data and best linear unbiased estimates (BLUEs) for ~300 wheat varieties grown under contrasting day lengths. Two hundred and eighty-six varieties from the OzWheat diversity panel were germinated under ‘long’ (12hr light) and ’short’ (8hr light) photoperiod regimes. Both the long and short-day treatments were carried out in a double coated plastic growth house with temperature control based at the CSIRO Black Mountain Innovation Precinct in Canberra, Australia. For the duration of each day length treatment temperature was maintained at 20oC and plants experienced normal sun light, which was supplemented with artificial lighting under fluorescent lamps delivering 300 µmol m-2 s-1 outside daylight hours. The long and short-day treatments were sown on the 30th November 2015 and 6th of June 2016 and harvested by the 6th of May and 9th of November 2016 respectively. In each treatment 1716 pots were grown across 9 benches within a 12 row and 16 column grid. Plants were watered via an automated irrigation system every 2 days. Seven traits were recorded by manual assessment twice weekly, from the 5th week post sowing which captured variation in phenology staging and spike development. These were number of days to first node detectable (Z31), number of days to anthesis (Z61), height (cm), spike length (cm), spikelets per spike and number of empty spikelets at maturity. Variation in phenology and spike traits was first assessed for incorrect entries and outlying values. Extreme outliers (i.e. less than 1\% of data) were excluded from the analyses. All traits adhered to expectations of normality. Trait data were analysed using a linear mixed model implemented in ASReml-R Release 3 (Butler et al. 2009, R Development Core Team, 2015). As in a typical repeated measures analysis we modelled 'variety' as a random treatment effect. Within each day length experiment trait data were analysed using the general mixed model framework. Broad sense heritability, or repeatability of variety means (Pmr), was estimated as the proportion of the total phenotypic variance explained by the random variety (genetic) effect variance (sigma^2), following Falconer and Mackay (1996). Further, variety level best linear unbiased estimates (BLUEs) were obtained for each trait by re-running the above model while fitting variety as a fixed effect. Adjusted variety means were extracted from this model using the predict() function classifying on 'variety' in ASReml-R. This approach was taken so as to avoid shrinkage of random effect estimates that would subsequently be applied in association analysis.