Abstract: This study investigates the ability of horses to recall a feeding event in a two-point choice apparatus. Twelve horses were individually tested whereby they were maintained immobile in a test arena and visually and aurally experienced the delivery of food into one of two feed goals. The horses were then released to make their choice in two experimental contexts: immediate release after experiencing the delivery of food, and release 10 s after food delivery. Each horse performed 40 immediate-release (IR) trials, followed by forty 10-s release trials over a 3-day period. In addition, the same horses were tested 3 months later in the spring with the same number and sequence of trials. Results were analysed by log-linear analysis of frequencies. Results showed that while horses were able to achieve the correct feed goal choice in the immediate-release trials, they were unsuccessful with the 10-s release trials. This suggests that there are limitations in recall abilities in horses, in that they may not possess a prospective type of memory. There are welfare and training implications in these findings concerning the effects of overestimating the mental abilities of horses during training and the effects of delays in reinforcements.