Discriminability measures such as d′ and log d become infinite when performance is extremely accurate and no errors are recorded. Different arbitrary corrections can be applied to produce finite values, but how well do these values estimate true performance? To answer this question, we directly calculated the effects of a range of different corrections on the sampling distributions of d ′ and log d . Many arbitrary corrections produced better estimates of discriminability than did the intuitively plausible technique of rerunning problem conditions. We concluded that when it is not possible to run more trials and when other techniques are not appropriate, the best correction overall is to add a correction constant between 0.25 and 0.5 to all response counts, regardless of their value.

Traditional theories of delayed matching-to-sample performance do not predict that accuracy will improve when absolute levels of reinforcement are increased. This prediction emerges only when reinforcement context is considered (J. A. Nevin, M. Davison, A. L. Odum, & T. A. Shahan, 2007). To provide quantitative data, the authors factorially manipulated between conditions the probability and duration of reinforcement for correct choices by pigeons. In Experiment 1, increasing the value of either variable improved initial discriminability of the forgetting functions, but did not affect the rate of forgetting. In Experiment 2, initial discriminability covaried with changes in choice immediacy and trial completion rate, suggesting a relationship with response strength consistent with Nevin et al.'s behavioral momentum model. Adding reinforcement context to K. G. White and J. T. Wixted's (1999) model also generates predictions consistent with the present experiments and with the effects of manipulating extraneous reinforcement. The inclusion of reinforcement context thus improves predictions of delayed matching-to-sample performance.

Two experiments examined whether postsample signals of reinforcer probability or magnitude affected the accuracy of delayed matching to sample in pigeons. On each trial, red or green choice responses that matched red or green stimuli seen shortly before a variable retention interval were reinforced with wheat access. In Experiment 1, the reinforcer probability was either 0.2 or 1.0 for both red and green responses. Reinforcer probability was signaled by line or cross symbols that appeared after the sample had been presented. In Experiment 2, all correct responses were reinforced, and the signaled reinforcer durations were 1.0 s and 4.5 s. Matching was more accurate when larger or more probable reinforcers were signaled, independently of retention interval duration. Because signals were presented postsample, the effects were not the result of differential attention to the sample.

Performance measures such as log d and d' aim to measure stimulus discriminability independently of response bias in conditional discrimination tasks, including the yes/no signal-detection procedure. However, they assume only one dimension of bias (e.g., response color) and do not account for bias on additional dimensions (e.g., response side). Such bias reduces log d, thus violating the statistical independence of discriminability and bias measurements. We modified log d to account for side bias and reanalyzed previous side-biased data. With strong side bias, the modified log d differed enough from the standard log d to potentially alter the conclusions of an experiment. Simulations showed that the modified log d produces discriminability estimates that are more accurate and bias-independent than the standard log d calculation.

In two experiments, pigeons' responding on an extraneous task was explicitly reinforced during delayed matching-to-sample trials. In Experiment 1, red or green sample stimuli were followed by retention intervals of 0.2, 1, 4, or 12 sec, during which pecks to a white center key were reinforced with 2.5-sec access to wheat according to extinction, variable-interval 30-sec, and variable-interval 15-sec schedules in different conditions. A proportion of .2, .5, .7, or .9 of subsequent red or green choice responses that matched the sample were reinforced with 3-sec access to wheat. The result was that increasing center key reinforcement, or reducing reinforcer probability, lowered overall accuracy. Initial discriminability fell, but with no change in the rate of forgetting. In Experiment 2, initial discriminability was affected by extraneous reinforcers that were contingent on center key pecking, but not by noncontingent reinforcers. A plausible conclusion is that initial discriminability decreases when reinforcers strengthen competing behaviors.