Paired Comparisons Effectively Drive the Learning of Multi-Category Perceptual Learning

Considerable research shows the value of comparisons in perceptual and category learning. In earlier work, it was shown that mixing simultaneous comparison trials with single-item classification trials in an adaptive learning framework enhanced multi-category perceptual learning (Jacoby, Massey, & Kellman, 2021). Here we tested the effectiveness of perceptual learning based entirely on simultaneous paired comparison trials. In a face identification paradigm with 22 categories, we compared an All Comparisons condition with Single Item and Dual Item Classification conditions, where the latter had trials resembling comparison trials but required two identification responses rather than selecting between items. Conditions were equated on time invested in learning (40 min). Several different face images of the same individual appeared across trials for each of the 22 categories. Learning was assessed in two posttests, one immediately after training and one after a one-week delay. All conditions used a common classification-based assessment containing some images previously seen in learning and some novel images for each person category. Since paired comparison trials had a chance accuracy of .5 and seemed less demanding than identification of a person from among 22 possibilities in the other conditions, we wondered if the All Comparisons condition would produce robust learning on a par with the other conditions. Results: All conditions resulted in successful learning and transfer of person recognition to novel exemplars. Comparison trials were completed more quickly than either classification format, and participants in the All Comparisons condition completed significantly more learning trials as a result. Average accuracy on both the immediate and delayed assessments did not differ reliably between conditions. These results suggest that with training time equated, a learning intervention comprised completely of relatively undemanding paired comparison trials are as effective as the more demanding classification-based trials in producing the learning and subsequent transfer of multiple categories.