The cost of surface change in recognition memory for pictures and words
In Part I, I present a series of five experiments which explored the effects of changing the surface form of study stimuli on recognition memory. Study items were presented in either picture or word form. Test items were presented in either the same or different form. Surface change cost was defined as the difference between recognition performance for same and different form items.
Strong surface change cost effects were observed in both accuracy and RT (Experiment 1). The cost tended to be somewhat larger when attention was focused on the sensory features of stimuli at study either by repeated exposure (Experiment 2) or by a surface encoding task (Experiment 3). Cost was eliminated by preexposing subjects to picture-word pairs (Experiment 4) and by inducing subjects to generate the alternate form of the study stimulus (Experiment 5). Cost was more reliably observed for items studied as pictures than for items studied as words. Results are discussed in terms of the cognitive mechanisms underlying recognition memory and the distinction between explicit and implicit memory.
A problem arose in interpreting the results of the experiments in Part I using empirical based analyses because the typical mirror pattern of hit and false alarm rates was violated. Using a source monitoring paradigm in Experiment 5, I found evidence suggesting that the violation of the mirror effect could be attributed to a particular response bias which operates in surface change experiments.
In Part II, I present a model-based analysis of Experiment 5 based on Batchelder and Riefer's (1990) multinomial processing model for source monitoring. The model based analysis is a critical supplement to the empirical based analysis because it provides independent measures of old/new recognition, source discrimination, and response biases. The model produced an excellent fit to the data and successfully predicted the theoretically expected effects of the independent variables on old/new recognition and source discrimination. The model based analysis also provided evidence confirming the prominent role of response bias in surface change experiments.