In a symbolic matching-to-sample task, pigeons learned to discriminate between 5 and 15 key pecks (samples): different choices were correct following the smaller and the larger response requirements. Subsequently, accuracy was tested in delayed matching, with the delay spent in darkness, contrarily to previous studies, that used illuminated delays. On average, delayed choices reflected indifference between the choices, but individual analyses showed different biases, replicating previous findings. It has been suggested that the end result of a delay may be similar to presenting no sample to begin with, so we compared preferences following a delay and following trials where no pecks were required. Performance in the two situations differed and, on zero-peck trials, a bias towards the “small” choice was found. Finally, to assess if the “small” bias was due to stimulus generalization, we compared zero-peck trials and trials with small response requirements (ranging from one to four) and found a discontinuity between zero and non-zero samples that may seem to be at odds with a generalization account.

在一个符号匹配到样本的任务中，鸽子学会了区分 5 和 15 个关键啄（样本）：随着较小和较大的响应要求，不同的选择是正确的。随后，在延迟匹配中测试了准确性，延迟在黑暗中花费，这与之前使用照明延迟的研究相反。平均而言，延迟选择反映了选择之间的差异，但个别分析显示出不同的偏见，复制了之前的发现。有人提出延迟的最终结果可能类似于一开始就没有提供样本，因此我们比较了延迟后的偏好和不需要啄食的试验后的偏好。这两种情况下的表现不同，在零啄试验中，发现了对“小”选择的偏见。最后，