What is learned in concept identification: Evidence for Two-Stage theories from an analysis of response subsequences

Direct tests of One-Stage and Two-Stage theories of discrimination learning and concept identification by means of shift experiments yield ambiguous results due to interpretational problems. These difficulties can be avoided by analysing the original acquisition data. A test of the theories is suggested for concept identification experiments which is based on an analysis of the dependence between appropriately constructed response subsequences. Predictions for the theoretical alternatives are derived from a generalized all-or-none model. Results from a concept identification study with adult subjects reject One-Stage theories and support Two-Stage theories.