Acquisition, generalization, and contextual control of taxonomic and thematic relational responding

In three experiments, college students performed either 2- or 3-comparison conditional discriminations (arbitrary matching to sample) that utilized 64 different configurations composed of drawings. Within each configuration, one comparison related to the sample taxonomically, one related thematically, and, where there was a third comparison, it did not relate to the sample. In training phases, subjects received positive verbal feedback for selections of either the taxonomic or thematic comparisons. Between training phases, subjects responded to novel configurations similar to those of training phases on which they received no feedback for their selections. For some subjects, one cycle through all the phases ended the experiment (Experiment 1); for others, in a second cycle, verbal feedback was reversed to follow selections based on the other relation and all phases were repeated (Experiment 2); and for 1 subject, contextual stimuli indicated which relation would lead to positive verbal feedback for each selection (Experiment 3). On test phases, the selections of all subjects became increasingly consistent with verbal feedback during training while the contextual stimuli reliably enabled the appropriate relations. These results suggest that human subjects respond relationally on this task, and that such relational responding can be contextually controlled.