Selection-based versus topography-based responding: An important distinction for stimulus equivalence?

English-speaking subjects were taught 16 English-French word pairs. Within any given trial, one word from each pair was presented; for eight items, subjects were to select its counterpart from an array of words in the other language (selection-based training), and for the other eight items there was no array and subjects were to type its counterpart (topography-based training). In Experiment 1, all items were trained from French to English, and later, interspersed no-feedback probe trials tested for the emergence of the reversed relations. Half of the eight selection-trained items were tested in the selection-based mode and half were tested in the topography-based mode; similarly, half of the eight topography-trained items were tested in the selection-based mode and half were tested in the topography-based mode. On the first reversal test trial, all 7 subjects scored 100% correct for the selection-tested items; in contrast, 5 of the 7 subjects scored 0% or near 0% correct for the topography-tested items, which improved to varying degrees with repeated testing. The training response mode affected neither acquisition rate nor reversal test trial performance. In Experiment 2, all items were tested in the topography-based mode only, and subjects were exposed to nine consecutive reversal test trials prior to interspersed probe testing. Improved accuracy across reversal test trials was not observed until the conditions of probe testing were instated, an indication that reexposure to the trained relations was a crucial component of delayed emergence. In Experiment 3, all items were trained from English to French, resulting in subjects typing a familiar rather than an unfamiliar word on reversal test trials. Accuracy on reversal test trials was considerably better than in the previous two experiments. We discuss the implications of our findings for stimulus equivalence research.