Does incidental training increase the prevalence of overall similarity classification? A re-examination of Kemler Nelson (1984)

Kemler Nelson (1984) reported that incidental training, relative to intentional training, increased the prevalence of overall similarity classification, supporting a non-deliberative account of overall similarity sorting. However, the analysis conducted by Kemler Nelson (1984) does not adequately distinguish between usage of an overall similarity classification strategy and single-attribute strategies. The current study replicates Kemler Nelson’s (1984) experiment, seeking to test the original conclusions using a more rigorous analysis. The current study approximates the original experimental procedure, using almost identical stimuli and a longer, modified test phase. Results replicate those found by Kemler Nelson (1984) when the original analysis is applied; however the model-based analysis suggest an overall similarity classification strategy is used rarely and that incidental training increases the prevalence of suboptimal single-attribute strategies. These results imply that overall similarity classification may be more deliberative than previously thought.