Distribution of trials and intertrial retention in delayed matching to sample with pigeons.

On the basis of the notion that the events of a delayed matching to sample (DMTS) trial would be forgotten over a long intertrial interval (ITI), it was predicted that a 20-sec ITI would lead to better performance than a 1-sec ITI with a random order of DMTS trials but that a 1-sec ITI would be superior to a 20-sec ITI with a homogeneous order of DMTS trials. Contrary to this prediction, it was found in Experiments 1 and 2 that a 20-sec ITI led to a higher accuracy than a 1-sec ITI with both random and homogeneous orders. In Experiment 3, daily reversal learning was used to show that varying the ITI was directly opposite effects on DMTS with homogeneous orders of trials and on simultaneous discrimination learning; a 1-sec ITI facilitated discrimination learning and hindered DMTS, relative to a 20-sec ITI. Analyses of two-trial sequences within the first three experiments indicated that memories of Trial n - 1 influenced performance on Trial n; pigeons' choice on Trial n reflected a preference for the pattern presented on or for the position of the key most recently pecked on Trial n - 1 regardless of the outcome (reinforced or nonreinforced) of Trial n - 1. In Experiment 4, performance on homogeneous DMTS trials embedded within random DMTS trials, using stimuli from another dimension, was studied. Performance on embedded trials was equally good with massed and spaced distribution of trials and was superior to performance on control trials. It is suggested that the carry-over of memory of preceding trials provides information necessary for successful performance in discrimination learning but acts to interfere with processing of the sample stimulus in DMTS.