Guillette Lauren M, Farrell Tara M, Hoeschele Marisa, Nickerson Carly M, Dawson Michael R W, Sturdy Christopher B
Department of Psychology, University of Alberta, Edmonton, Alberta, Canada T6G 2E9.
J Comp Psychol. 2010 Feb;124(1):109-15. doi: 10.1037/a0017741.
We report on operant conditioning and artificial neural network (ANN) simulations aimed at further elucidating mechanisms of black-capped chickadee chick-a-dee call note category perception. Specifically, we tested for differences in the speed of acquisition among different discrimination tasks and, in two selected discrimination groups, searched for evidence of peak shift. Earlier, unreported ANN data were instrumental in providing the motivation for the current set of studies with chickadees and are provided here. The ANNs revealed differences in the speed of learning among note-type discrimination groups that is related to the degree of perceptual similarity among the three note types tested (i.e., A, B, and C notes). In many respects, bird and network results were in agreement (i.e., in the observation of peak shift in the same group), but they also differed in important ways (i.e., all discrimination groups showed differences in speed of learning in simulations but not in chickadees). We suggest that the start, peak and end frequency of the chick-a portion of chick-a-dee call notes, which form a graded but overlapping continuum, may drive the peak shift observed.
我们报告了旨在进一步阐明黑顶山雀 chick-a-dee 叫声类别感知机制的操作性条件反射和人工神经网络(ANN)模拟。具体而言,我们测试了不同辨别任务之间习得速度的差异,并在两个选定的辨别组中寻找峰值转移的证据。早期未报告的人工神经网络数据为当前这组关于山雀的研究提供了动力,现于此呈现。人工神经网络揭示了音符类型辨别组之间学习速度的差异,这与所测试的三种音符类型(即 A、B 和 C 音符)之间的感知相似程度有关。在许多方面,鸟类和网络的结果是一致的(即在同一组中观察到峰值转移),但它们也在重要方面存在差异(即所有辨别组在模拟中显示出学习速度的差异,但在山雀中没有)。我们认为,构成渐变但重叠连续体的 chick-a-dee 叫声中 chick-a 部分的起始、峰值和结束频率,可能驱动了所观察到的峰值转移。