400
w.w.L. Au
targets. However, the results also show the animal did not discriminate the
targets when the probes were in the "flat face forward" orientation.
Backscatter measurements indicated high variability in the echo amplitudes for the cubes and cylinders in the flat face forward orientation
and low variability for the cylinder in the upright position. The standard
deviations of 15 target strength measurements on the three cubes with
the flat faces forward were 2.7, 5.9, and 6.2dB compared with 1.2, 1.0,
and 1.0dB for the cubes standing upright. The targets were removed and
placed back on the holder after each measurement to simulate the experimental situation. Measurements from cylinders with a flat face forward
yielded similar results as the cubes in a flat face forward orientation.
Standard deviations of 2.6, 4.9, and 5.9 dB were obtained. These differences
in variability paralleled the performance of the dolphin. The animal most
likely received echoes varying in amplitude when scanning across the flat
surfaces of the cubes or the tops of the cylinders and received relatively
uniform amplitude echoes when scanning across the curved portion of the
cylinders.
5. Multidimensional Discrimination
Roitblat et al. (1990) performed a matching-to-sample experiment with an
echolocating T. truncatus, in which the dolphin was required to match
targets from a sample array with targets from three comparison arrays, each
array having similar targets. A target set consisted of a large hollow PVC
tube (lO-cm outer diameter, 25-cm length, .3-cm wall), a small hollow PVC
tube (7.5-cm outer diameter, 15-cm length, .3-cm wall), a solid aluminum
cone (10-cm base, lO-cm length), and a 5-cm waterfilled sphere. The sample
array was located 4.8m directly ahead of the animal's station and the comparison arrays were suspended from a bar located 3.9 m from the observing aperture and spaced apart from each other. The blindfolded dolphin
first echolocated a sample target while in a hoop station. Then after backing
out of the station, one target from each comparison array was then lowered
into the water; one of the targets matched the sample and two were different. The dolphin reentered the station and echolocated the targets of the
comparison arrays and responded by touching one of three response
paddles to indicate the location of the matching target.
The dolphin's choice accuracy during the final 48 sessions of the experiment averaged 94.5% correct. Roitblat, Penner, and Nachtigall (1990) were
able to apply a sequential sample model using Bayesian decision rules and
obtained results that corresponded well with the dolphin's performance in
the matching-to-sample task. Differences in the target echoes using simulated dolphin signals were obvious. Easily discriminable targets were
deliberately used to demonstrate that a dolphin could grasp the concepts
of same-difference and matching-to-sample in a sonar modality (Nachtigall
w.w.L. Au
targets. However, the results also show the animal did not discriminate the
targets when the probes were in the "flat face forward" orientation.
Backscatter measurements indicated high variability in the echo amplitudes for the cubes and cylinders in the flat face forward orientation
and low variability for the cylinder in the upright position. The standard
deviations of 15 target strength measurements on the three cubes with
the flat faces forward were 2.7, 5.9, and 6.2dB compared with 1.2, 1.0,
and 1.0dB for the cubes standing upright. The targets were removed and
placed back on the holder after each measurement to simulate the experimental situation. Measurements from cylinders with a flat face forward
yielded similar results as the cubes in a flat face forward orientation.
Standard deviations of 2.6, 4.9, and 5.9 dB were obtained. These differences
in variability paralleled the performance of the dolphin. The animal most
likely received echoes varying in amplitude when scanning across the flat
surfaces of the cubes or the tops of the cylinders and received relatively
uniform amplitude echoes when scanning across the curved portion of the
cylinders.
5. Multidimensional Discrimination
Roitblat et al. (1990) performed a matching-to-sample experiment with an
echolocating T. truncatus, in which the dolphin was required to match
targets from a sample array with targets from three comparison arrays, each
array having similar targets. A target set consisted of a large hollow PVC
tube (lO-cm outer diameter, 25-cm length, .3-cm wall), a small hollow PVC
tube (7.5-cm outer diameter, 15-cm length, .3-cm wall), a solid aluminum
cone (10-cm base, lO-cm length), and a 5-cm waterfilled sphere. The sample
array was located 4.8m directly ahead of the animal's station and the comparison arrays were suspended from a bar located 3.9 m from the observing aperture and spaced apart from each other. The blindfolded dolphin
first echolocated a sample target while in a hoop station. Then after backing
out of the station, one target from each comparison array was then lowered
into the water; one of the targets matched the sample and two were different. The dolphin reentered the station and echolocated the targets of the
comparison arrays and responded by touching one of three response
paddles to indicate the location of the matching target.
The dolphin's choice accuracy during the final 48 sessions of the experiment averaged 94.5% correct. Roitblat, Penner, and Nachtigall (1990) were
able to apply a sequential sample model using Bayesian decision rules and
obtained results that corresponded well with the dolphin's performance in
the matching-to-sample task. Differences in the target echoes using simulated dolphin signals were obvious. Easily discriminable targets were
deliberately used to demonstrate that a dolphin could grasp the concepts
of same-difference and matching-to-sample in a sonar modality (Nachtigall