Effect of horse sleep behavior on performance in a field-side spatial reversal learning test

Reduced rapid-eye-movement (REM) sleep is associated with impaired learning in many species. We developed a reversal learning test (RLT) suitable for field conditions to explore this association in 16 healthy horses. Nocturnal REM-like sleep behavior was recorded five times for 48 h over a six-week period. The horses performed a target training task followed by an RLT using two objects. When the horses reached a predefined frequency of touching the object, the spatial location was reversed. Mean test parameters were statistically analysed using GENLIN models, longitudinal continuous variables were analysed using linear repeated measures models, and dichotomous repeated measures were analysed using GEE models and Kaplan-Meier method. Altogether 15/16 horses completed RLT by reversing three or more times. Most errors occurred before the second and third reversals. Overall, REM sleep duration (mean ± SE) was 46.1 ± 8.7 min. However, there were ten horses that exhibited REM-like sleep for less than 30 min (10.6 ± 2.2 min, range 0.0–28.0 min), while six horses exhibited REM-like sleep for at least 30 min (42.3 ± 4.8 min, range 36.6–65.8 min). Longer REM sleep was associated with a greater number of reversals (p = 0.04), while no relationship was found with error rate. Survival analysis further indicated a significant difference in progression probability between groups, with horses with shorter REM sleep having a 50% chance of progressing after five reversals, compared to six reversals for longer REM sleepers (p < 0.05). We present here a method to test horses’ reversal learning ability on site over a single day. We found that short REM-like sleep duration without clinical signs of sleep disturbances in horses was associated with poorer performance and lower perseverance during the RLT.