Top-Down Effects on Multiple Meaning Access within and between Languages.

Abstract

This research investigates context effects on multiple meaning access during word recognition. Previous monolingual word recognition research suggests that multiple meanings of homographs are temporarily activated. In disambiguating context, Reordered Access predicts multiple meaning activation, while Selective Access predicts single meaning activation. The difference arises from differences in their predictions for contextuallyinappropriate meanings: Reordered Access predicts no context effects, and Selective Access predicts suppression due to context. Two eyetracking during listening experiments showed that top-down context both increased activation of the appropriate meaning of a homophone and decreased activation of the inappropriate meaning, however, multiple meanings were still activated. Thus, a strict form of neither Reordered Access or Selective Access can account for the present results.

Most previous research on context effects on homophone resolution assumed that participants fully engaged in the sentence processing tasks and fully understood the sentence contexts. However, if this assumption is invalid, the conclusions of previous studies may also be invalid. Two naming experiments investigated motivational effects (monetary compensation, supervision, feedback) on homograph meaning resolution. The results indicated that participant motivation increased overall task performance, but did not reliably affect homograph meaning activation. Previous bilingual research has found that word-initial cohort competitors from multiple languages are activated, even in monolingual contexts. BIA+ and BIMOLA both account for multiple language activation, but differ in how context affects the nontarget language. BIA+ assumes that lexicons of multiple languages are integrated; context affects words in both languages simultaneously. In contrast, BIMOLA assumes that lexicons of multiple languages are stored in different language networks; context effects can be selective to one language. Three eyetracking-during-listening experiments showed that biasing context increased activation of the target language meaning, but did not affect the nontarget language activation. Thus, context effects on multiple language activation are language-selective, although multiple languages are activated, supporting the BIMOLA. The present set of experiments demonstrated that regardless of surrounding context, multiple meanings and multiple languages are activated. Biasing context plays a role in modulating lexical activation, both facilitating appropriate meanings and inhibiting inappropriate meanings. However, context effects modulate meaning activation only in the target language.