The effect of semantic relatedness of ambiguous words in the free recall task

Abstract

Years of empirical work have shown that words with multiple related senses (polysemy) are processed faster than unambiguous words, whereas the ones with multiple unrelated meanings (homonymy) are processed slower (Rodd, Gaskell, & Marslen-Wilson, 2002). This was true for the lexical decision task (LDT), however, if different experimental task was employed, presence of those effects would vary. To account for this, Semantic Settling Dynamics (SSD) model was developed (Armstrong & Plaut, 2016) which relies on the hypothesis that different amounts of semantic demands between experimental tasks are the cause of varying effects. Model predicts that polysemes produce a large effect in early phases that decreases with further processing. In order to test this prediction further, we described polysemy by entropy – measure that combines number of senses with the balance of their probabilities. In two experiments, we used the same strategy of prolonging of the processing: we compared a visual LDT with its slowed down version. In our first experiment, slowing down was attempted by reducing stimulus-background contrast and in the second one, we compared visual and auditory LDT (Armstrong & Plaut, 2016; Medeiros & Armstrong, 2017). First manipulation failed to prolong processing, although expected facilitatory entropy effect was present in both experimental conditions (high contrast: b = -.011, S.E. = .004, df = 220.4, t = -2.41, p = .017; reduced contrast: b = -.012, S.E. = .003, df = 232.4, t = -2.748, p = .006). Second manipulation substantially prolonged the processing and the interaction between entropy and the experimental manipulation was significant (b = .022, S.E. = .003, df = 139.485, t = 8.039, p = .000). In visual LDT, expected entropy effect was found (b = -024, S.E. = .020, df = 142.885, t = 30.425, p = .000), while in the auditory LDT (prolonged), there was no such effect (b = -.003, S.E. = .005, df = 166.99, t = -.589, p = .557). Both results are in the accordance with model predictions – when there was no prolonging of the processing, there was no change in entropy effect, while in the prolonged condition the effect disappeared. According to model background, auditory lexical decision should still produce polysemy effect, while in our experiment no effect was noted in that condition. These results implicate that SSD model may predict modulation of the entropy effect, however, it is unable to predict the exact effect change which is a great limitation of its predictive power.