Joint stimulus control in a temporal discrimination task

Exploring boundary conditions of the single-code/default strategy in pigeons

To investigate the extent of adoption of more efficient coding strategies, pigeons learned, in three experiments, a symbolic matching-to-sample task that featured an asymmetric sample-comparison mapping. In all experiments, one comparison was correct following one of the samples (one-to-one mapping), and another comparison was correct following the remaining samples (many-to-one mapping). The experiments differed in sample number; Experiment 1 featured three samples, Experiment 2 five samples, and Experiment 3 seven samples. Our goal was to assess the adoption of a single-code/default coding strategy, which establishes two response rules: one rule specific to the sample mapped one-to-one (the single code), and another rule to be applied following any other sample (the default rule). Alternatively, the animals could establish more response rules, one per sample. Thus, the single-code/default strategy allows learning a task via a reduced number of response rules, and the more samples are mapped many-to-one, the greater the savings it allows. As such, the three experiments should progressively be more amenable to the adoption of this strategy. Overall, the adoption of a single-code/default strategy was not widespread. When taken together with previous results, the present study suggests that the amount of training may affect the coding strategy pigeons adopt. Additionally, our results underscore that individual differences are a fundamental aspect to consider when studying learning flexibility.

Sample-comparison mapping and joint stimulus control

A matching-to-sample task with a 3-sample, 2-comparison mapping has been found to engender joint control by the stimuli that signaled the samples (keylight) and the inter-trial interval (houselight), with a trade-off in the degree of control exerted by each stimulus. To learn about the boundary conditions for the establishment of that joint stimulus control, we trained pigeons in a similar task, but featuring a one-to-one sample-comparison mapping, with two samples and two comparisons. To assess their relative influences, we ran two tests where each of the stimuli was removed: in one test, no sample keylight was presented, and in the other, the ITI was spent in darkness. Results were consistent with both stimuli influencing choice but there was no clear evidence of a trade-off between them. These results suggest that sample-comparison mapping and/or total number of samples may influence how an animal directs its attention to the stimuli available.

Searching for nothing: placing zero on the temporal continuum

Generalization allows responses acquired in one situation to be transferred to similar situations. For temporal stimuli, a discontinuity has been found between zero and non-zero durations: responses in trials with no (or 0-s) stimuli and in trials with very short stimuli differ more than what would be expected by generalization. This discontinuity may happen because 0-s durations do not belong to the same continuum as non-zero durations. Alternatively, the discontinuity may be due to generalization decrement effects: a 0-s stimulus differs from a short stimulus not only in duration, but also in its presence, thus leading to greater differences in performance. Aiming to reduce differences between trials with and without a stimulus, we used two procedures to test whether a potential reduction in generalization decrement would bring performance following zero and non-zero durations closer. In both procedures, there was a reduction in the discontinuity between 0-s and short durations, supporting the hypothesis that 0-s durations are integrated in the temporal subjective continuum.

Choice biases in no-sample and delay testing in pigeons (Columba livia)

In experimental tasks that involve stimuli that vary along a quantitative continuum, some choice biases are commonly found. Take, for instance, a matching-to-sample task where animals must, following the presentation of sample stimuli (that differ in duration), choose between two or more comparison stimuli. In tests where no sample is presented there is usually a bias towards the comparison that is correct following the shortest sample. To examine some aspects of these choice biases, pigeons were trained in a symbolic matching-to-sample task with two durations of keylight as samples, where key pecking had to be maintained during sample presentation. Firstly, even though animals were required to attend to the sample, a preference for the "short" comparison in no-sample testing was found. This result disproves an account where this effect was hypothesized to happen due to non-programmed learning resulting from the animals failing to attend to some trials. Secondly, even though a bias for "short" was found in both no-sample and delay testing, the extent of the biases differed between tasks, thus suggesting that forgetting the sample presented during a delay does not necessarily land the animal in a state similar to presenting no sample at all to begin with.