We determined the effect of an H 1 receptor antagonist on the functional recovery of Carassius auratus submitted to telencephalic ablation. Five days after surgery the fish underwent a spatial-choice learning paradigm test. The fish, weighing 6-12 g, were divided into four groups: telencephalic ablation (A) or sham lesion (S) and saline (SAL) or chlorpheniramine (CPA, ip, 16 mg/kg). For eight consecutive days each animal was trained individually in sessions separated by 24 h (alternate days). Training trials (T1-T8) consisted of finding the food in one of the feeders, which were randomly blocked for each subject. Animals received an intraperitoneal injection of SAL or CPA 10 min after the training trials. The time spent by the animals in each group to find the food (latency) was analyzed separately at T1 and T8 by the Kruskal-Wallis test, followed by the Student Newman-Keuls test. At T1 the latencies (mean ± SEM) of the A-SAL (586.3 ± 13.6) and A-CPA (600 ± 0) groups were significantly longer than those of the S-SAL (226.14 ± 61.15) and S-CPA (356.33 ± 68.8) groups. At T8, the latencies of the A-CPA group (510.11 ± 62.2) remained higher than those of the other groups, all of which showed significantly shorter latencies (A-SAL = 301.91 ± 78.32; S-CPA = 191.58 ± 73.03; S-SAL = 90.28 ± 41) compared with T1. These results support evidence that training can lead to functional recovery of spatial-choice learning in telencephalonless fish and also that the antagonist of the H 1 receptor impairs it.