Contribution of working memory processes to relational matching-to-sample performance in baboons (Papio papio).
Prior research suggests that variability discrimination is basic to same-different conceptualization (Young and Wasserman, 2001). In that research, people were trained with 16-item arrays; this training might have encouraged people to use perceptual variability to solve the task. Here, two groups of participants were trained with either 2- or 16-item Same and Different arrays (Groups 2 and 16, respectively). Participants had to learn which of two arbitrary responses was correct for the arrays without being told about the "sameness" or "differentness" of the stimuli. Surprisingly, 52% of participants in Group 2 did not learn the discrimination compared to only 21% of participants in Group 16; also, learners in Group 16 reached higher accuracy levels sooner and their choice responding was faster than learners in Group 2. A large disparity in the variability (measured by entropy) between the Same and Different arrays evidently helped participants to learn the same-different task. As well, in Group 16, we found the same two patterns of performance-Categorical and Continuous-as in prior research (Castro et al., 2006; Young and Wasserman, 2001). In Group 2, we again found the Categorical cluster, but we lost the genuine Continuous cluster and we observed a novel strategy: some participants developed a highly inclusive notion of "sameness" that applied to any array containing at least two identical icons. These findings indicate that individuals may deploy a multiplicity of possible strategies when learning a seemingly simple same-different discrimination.