The effect of response requirement and target probability on the performance of dogs during a go/no-go scent detection task

Abstract

As the response threshold required to obtain a reinforcer is increased, the likelihood that the reinforcer will be obtained typically decreases. When a response is reinforced in the presence of a stimulus and not in its absence (i.e., a discriminated operant), the response threshold appears to influence the probability that the response will occur in the presence of the discriminative stimulus. In addition, when the target probability during a task is low, the number of responses performed typically decreases. We examined these two factors, response requirement and target prevalence, in a scent-detection task with domestic dogs where amyl acetate solutions were presented to each dog in an automated apparatus. The dogs were trained to place their nose in a sample port where they had access to samples, and to hold their nose in the port to indicate the sample was positive or to activate a limit switch when the sample was negative (a go/no-go scent detection task). During Experiment One, dogs were presented with various indication thresholds (seconds) which were manipulated systematically to evaluate the influence of response requirement on hit rate and correct rejection rate. During Experiment Two, target prevalence was manipulated systematically. Increasing indication threshold had a significant effect on hit rate when large increase in correct rejection rate as the response requirement was increased in the lower range and high correct rejection rate across most other values tested with a slight drop in hit rate as the response requirement was increased. Experiment Two revealed a small partial eta squared effect on accuracy when target prevalence was increased. Limitations and suggestions for future research are discussed.