A field-based experimental study was conducted in 50 × 50 m square plots to investigate the behaviour of free-spreading fires in wheat to quantify the effect of crop condition (i.e. harvested, unharvested and harvested and baled) on the propagation rate of fires and their associated flame characteristics, and to evaluate the adequacy of existing operational prediction models used in these fuel types. The dataset of 45 fires ranged from 2.4 to 10.2 km h−1 in their forward rate of fire spread and 3860 and 28 000 kW m−1 in fireline intensity. Rate of fire spread and flame heights differed significantly between crop conditions, with the unharvested condition yielding the fastest spreading fires and tallest flames and the baled condition having the slowest moving fires and lowest flames. Rate of fire spread in the three crop conditions corresponded directly with the outputs from the models of Cheney et al. (1998) for grass fires: unharvested wheat → natural grass; harvested wheat (~0.3 m tall stubble) → grazed or cut grass; and baled wheat (<0.1 m tall stubble) → eaten-out grass. These models produced mean absolute percent errors between 21% and 25% with reduced bias, a result on par with the most accurate published fire spread model evaluations.