Retrieval cues and syntactic ambiguity resolution: Speed-accuracy tradeoff evidence

Successful language comprehension often involves coping with lexical and syntactic ambiguity, and sometimes, recovering from misanalysis of the input. Syntactic ambiguity resolution has been shown throughout the literature to result in increases reaction time compared to unambiguous sentences, a fact which has shaped debates about architectures and mechanisms in sentence processing. However, increased reaction time can be caused either by a decrease in true processing speed, or by a decrease in the quality or quantity of information needed to reach criterion when making a response. Thus, increased reaction time to syntactic ambiguity could reflect differences in representational quality or multiple reanalysis attempts, or both. Current cue-based accounts of sentence processing predict that cues at the point where misanalysis becomes apparent (e.g., onset of second verb) may aid in reanalysis (e.g., retrieval of the correct subject). We used the speed-accuracy tradeoff procedure (SAT) to orthogonally derive estimates of processing speed and accuracy. We manipulated ambiguity and the semantic similarity between a disambiguating verb and the nouns already present in the sentence. Estimates of processing speed (SAT rate) indicated that, on average, ambiguous conditions took 250ms longer to interpret than unambiguous controls, demonstrating that reanalysis does increase veridical processing time. No interaction between cue diagnosticity and ambiguity was observed on speed or accuracy, but verbs more strongly related to the correct subject increased accuracy, regardless of ambiguity. These findings are consistent with a language processing architecture where cue-driven retrieval operations give rise to interpretation, and wherein diagnostic cues aid retrieval, regardless of parsing difficulty or structural uncertainty.