HOLISM REVISITED - EVIDENCE FOR PARALLEL INDEPENDENT WORD-LEVEL AND LETTER-LEVEL PROCESSORS DURING WORD RECOGNITION

This study tested the predictions of the parallel input serial analysis (PISA) model (Allen & Madden, 1990), the interactive activation model (McClelland & Rumelhart, 1981), and the activation-verification model (Paap, Newsome, McDonald, & Schvaneveldt, 1982) by using lexical-decision (Experiment 1) and letter-identification (Experiments 1-4) tasks. For the lexical-decision task, reaction time decreased monotonically as word frequency increased. However, letter identification took longer for medium-high-frequency words than for very-high-frequency, low-frequency, and very-low-frequency words (i.e., a partial word frequency disadvantage). For letter identification, there was a word inferiority effect for medium-high-frequency words (Experiments 2-4). These word frequency disadvantages and word inferiority effects support the PISA model but not the interactive activation model or the activation-verification model.