The capacity to exhibit generalized sameness-difference judgments is a hallmark of cognition that is regularly exhibited by humans. As yet, that capacity has not been well documented in New World monkeys such as the capuchin (Cebus apel/a). This article presents data obtained with 6 capuchi n monkeys with a variety of procedures that might lead to gEmeralized identity matching-to-sample (MTS) in this species, reporting part of a research program conducted to evaluate methods for assessing the species' relational learning capacity. Our working hypothesis is that past failures to demonstrate relational learning have been caused by procedural insufficiency rather than a lack of capacity. Thus far, 6 capuchin monkeys have been test, ed for generalized identity MTS. The apparatus was a touchscreen-equipped microcomputer-controlled experimental chamber. Eleven sets of 3 visual stimuli (black shapes on gray backgrounds) were used. The general procedure was comprised of 4 phases: (a) simple discriminations, (b) repeated shifts of simple discriminations, (c) identity MTS training, and (d) generalized identity MTS tests . Every subject was exposed to each of the phasE3s. Positive results

This paper reports use of sample stimulus control shaping procedures to teach arbitrary matching-to-sample to 2 capuchin monkeys (Cebus apella). The procedures started with identity matching-to-sample. During shaping, stimulus features of the sample were altered gradually, rendering samples and comparisons increasingly physically dissimilar. The objective was to transform identity matching into arbitrary matching (i.e., matching not based on common physical features of the sample and comparison stimuli). Experiment 1 used a two-comparison procedure. The shaping procedure was ultimately effective, but occasional high error rates at certain program steps inspired a follow-up study. Experiment 2 used the same basic approach, but with a three-comparison matching task. During shaping, the monkey performed accurately until the final steps of the program. Subsequent experimentation tested the hypothesis that the decrease in accuracy was due to restricted stimulus control by sample stimulus features that had not yet been changed in the shaping program. Results were consistent with this hypothesis, thus suggesting a new approach that may transform the sample stimulus control shaping procedure from a sometimes useful laboratory tool to a more general approach to teaching the first instance of arbitrary matching performances to participants who show protracted difficulties in learning such performances.