Njegovan M, Ito S, Mewhort D, Weisman R
Department of Psychology, Queen's University at Kingston, Ontario, Canada.
J Exp Psychol Anim Behav Process. 1995 Jan;21(1):33-42.
We trained songbirds and humans in go/no-go discriminations among 27 tones. In the compact discrimination, S + s formed a contiguous middle range (3080-4040 Hz), and S-s formed contiguous lower (2000-2960 Hz) and upper (4160-5120 Hz) ranges. In the distributed discrimination, S + s were spread across all 3 ranges. Songbirds acquired the compact discrimination more quickly and with higher accuracy than humans. Songbirds acquired the distributed discrimination only after much extended training; humans did not acquire the distributed discrimination. Compact groups (birds and humans) accurately classified test tones spaced 60 Hz from the training tones, but the distributed groups did not. A single reversal in discrimination between tones on the boundary between the lower S- and middle S + ranges did not propagate to all the tones in either range. A neural network model provided an account of the classification of tones in songbirds and humans.
我们对鸣禽和人类进行了27种音调的“是/否”辨别训练。在紧凑辨别任务中，S +（刺激信号）形成了一个连续的中间范围（3080 - 4040赫兹），而S -（非刺激信号）形成了连续的较低范围（2000 - 2960赫兹）和较高范围（4160 - 5120赫兹）。在分布式辨别任务中，S +分布在所有三个范围内。鸣禽比人类更快且更准确地学会了紧凑辨别任务。鸣禽只有在经过大量延长训练后才学会了分布式辨别任务；人类则没有学会分布式辨别任务。紧凑组（鸟类和人类）能够准确地对与训练音调相差60赫兹的测试音调进行分类，但分布式组则不能。在较低S -范围和中间S +范围之间的边界上，音调辨别中的单次反转并没有传播到任何一个范围内的所有音调。一个神经网络模型解释了鸣禽和人类对音调的分类情况。
