The effects of associative congruence and dimensional similarity of samples in delayed matching by pigeons.

Pigeons were trained with a many-to-one mapping of sample stimuli onto comparison stimuli. The red comparison stimulus was correct after red, vertical line or circle sample stimuli, and the green comparison was correct after green, horizontal line or triangle sample stimuli. On control trials, only a single sample stimulus was presented. On congruent trials, two sample stimuli were presented and both were associated with the same comparison response. On incongruent trials, two sample stimuli were presented and they were associated with different comparison responses. The two sample stimuli on congruent and incongruent trials were either from the same stimulus dimension or from different stimulus dimensions. Matching accuracy on congruent trials was equivalent to that on control trials regardless of whether the two samples were from the same or from different stimulus dimensions. Matching accuracy on incongruent trials was significantly lower than on control trials. In addition, accuracy was equivalent when the incongruent samples were from the same dimension and when they were different dimensions. These findings were robust across various interstimulus intervals and retention intervals in three experiments. The results indicated that pigeons, unlike monkeys, are insensitive to the dimensional similarity of sample stimuli. The data suggest that pigeons do not retrospectively code samples in a many-to-one procedure, rather they are consistent with the notion that associating more than one sample with the same comparison produces a common, shared memory code for the various samples. This common code could be an intermediate sample code ("Sample A", "Sample B"), or a prospective response code ("Peck Red", "Peck Green").