highly significantP < 0.00001
Results remained highly significant when excluding the cross-fostered bats ( n = 11), which were only tested in year 3 ( P < 0.00001; learning : 0.19±17 vs. 0.45 ±0.19 error rates, respectively, R 2 =0.72, p < 0.00001, mixed GLM with learning set as the response variable, the origin as a fixed factor and the bat’s ID as a random effect, n = 65).