Abstract
In 1975, Norman and Bobrow proposed a “data-limited” and “resource-limited” model for speech processing. Processing is considered “resource-limited” when performance in a task can be improved by exerting more effort, and “data-limited” when it cannot. Exceedingly difficult conditions may produce a “tipping point” in effort after which processing shifts from being resource-limited to data-limited. As a result, cognitive effort declines while performance becomes stagnant at floor levels. Likewise, clear conditions may eventually produce a tipping point in effort because individuals can maintain high levels of performance regardless of the effort they apply to the task. This forms an inverted U-shaped function of effort across the continuum of task difficulty. The recently introduced DRL model, an expansion of Norman and Bobrow’s original framework, predicts that processing becomes “language-limited” at the upper asymptote of task performance. Therefore, performance and effort depend on the linguistic complexity of the stimulus. The two experiments in this study tested the predictions of the DRL model. In Experiment 1, young adults were asked to listen to and recall syntactically complex object-relative (OR) sentences and simpler subject-relative (SR) sentences time-compressed to 10 different speech rates. Results showed the expected psychophysical function, where performance was lowest at the fastest speech rate, increased linearly as speech became less time-compressed, and plateaued at high levels when speech was uncompressed. Differences in SR and OR performance were greatest at this point, consistent with the predictions of the DRL model. Experiment 2 measured changes in participants’ pupil size as an index of their cognitive effort across speech rates and found the expected inverted U-shaped function for SR sentences. However, the more syntactically complex OR sentences did not follow the expected inverted U, as effort increased for these sentences even in the language-limited region. Results are discussed in terms of the strengths and weaknesses of the DRL model.