Participants
Four adult male bottlenose dolphins (Tursiops truncatus), Eagle, Quick, Tino, and Peace, participated in the present experiments. They lived as a group in the Port of Nagoya Public Aquarium in Nagoya City, Aichi, Japan. They usually received four 15-min sessions of husbandry, performance, and cognitive training per day by the human trainers (Tomonaga et al. 2010, 2014a, b).
Experimental setting
Experiments were conducted in a pool (elliptical shape, 34 × 18 m and 9 m in depth) adjacent to their home pool (Figure 5). At the one side of the pool, a blue opaque panel (PVC, 90 cm wide, 180 cm high) was set (Figure 2). Identical buckets with rewards (pieces of fish) were placed on the floor 1 m to the left- and right-side edges of the panel. Another bucket was placed at the opposite side of the pool.
Procedure
Preliminary training
Each dolphin was initially given preliminary training sessions, which consisted of five trials. The first four were baseline trials, and the last one was the test trial. Before the onset of the trial, the main trainer stood in front of the panel. Once the dolphin stayed in front of the panel, the trainer moved to left or right and showed a gestural sign for action (e.g., jump, flipper-shaking) by the dolphin. If the dolphin performed the corresponding action correctly, a whistle was sounded, and a food reward was given. After the reward, the trainer made the dolphin move to the front of the panel and stay there, and the trainer also stayed in front of the panel. In the baseline trials, only the main trainer appeared. In the test trial, two persons (of the same sex) stood in front of the panel simultaneously. One of them was the main trainer, and the other was a dummy person (but familiar to the dolphin). Each person then walked to the left or right along the pool side, and they met at the opposite side of the pool. In daily routine training, when trainers walked along the pool side, the dolphins always followed them. Thus, in this test setting, the dolphin also followed one of the persons. When the dolphin came to the opposite side where the third bucket was located, the main trainer performed a gestural command to the dolphin. The dolphin was rewarded only when it successfully performed the action corresponding to the gestural sign, irrespective of the following response. Thus, following responses were not explicitly and differentially reinforced. Each dolphin was given eight preliminary training sessions.
During preliminary training sessions completed before each test session, all dolphins followed the main trainers on every trial.
Different-clothing test
After completing the preliminary training, each dolphin was given 24 test sessions (i.e., 24 test trials). Figure 5 shows the flow of the baseline and test trials. The difference between the preliminary training and the subsequent test trials was that in the latter, the trainer always hid behind the panel. In the different-clothing test trials, the two people wore different clothes to each other, and the main trainer also changed clothes randomly from session to session (Figure 2). The side from which the main trainer appeared (which also set the direction in which the trainer then walked) was counterbalanced and randomized. During the preliminary training and testing sessions, each dolphin–trainer pairing was fixed. Dummy persons were changed from session to session.
Same-clothing test
One year after the different-clothing test, we conducted an additional test in which the two persons were of the same sex and of similar height, and they wore identical clothing (Figure 4). A different trainer from the first test was assigned to each dolphin. Before the test sessions, the dolphins were given 12 preliminary training sessions in which the two persons stood in front of the panel; however, due to the limitations imposed by exhibition schedules, only three dolphins (Eagle, Quick, and Tino) participated in this test; they were given 13, 15, and 16 sessions, respectively.
Throughout the experiments, dolphins responded accurately to the gestural commands given by the trainer (97.9% accuracy, averaged across dolphins).