Mechanisms of adaptation in (healthy and aphasic) noisy-channel comprehension

点击翻译按钮获取中文摘要

The broad objective of the proposed research and training is to understand the mechanisms that underlie efficient language comprehension and how these are altered in persons with language impairment, specifically aphasia. Recent theories of language processing have proposed that communication is typically subject to noise (e.g., Gibson, Bergen, & Piantadosi, 2013)—in the form of speaker error, environmental noise, or listener misperception—and that comprehenders employ rational integration of prior knowledge and noisy evidence to infer the intended meaning of the corrupted message. Further, Gibson, Sandberg, Fedorenko, Bergen, & Kiran (in press) propose that failures to use syntactic cues and an over-reliance on plausibility information by persons with agrammatic aphasia can be explained by the same noisy-channel framework, rather than a syntactic deficit. On this account, persons with aphasia simply have a higher base level of noise in their representation of how a message is likely to be corrupted. These novel proposals raise important questions about language processing in healthy individuals, as well as persons with aphasia. The proposed work will examine the learning and memory mechanisms that support the representation of the noise model. Aim 1. To determine whether listeners adapt to specific properties of the noise and to what extent the representations are tied to the context in which they are experienced. Aim 2. To investigate whether the representation of noise and adaptation mechanisms differ between healthy individuals and persons with agrammatic aphasia. In the proposed experiments, we will address these aims by comparing persons with agrammatic aphasia to neurologically healthy controls and brain damaged controls. Participants will be exposed to sentences, some of which contain errors (i.e., noise), and asked to interpret their meaning. The critical sentences will be grammatically correct but implausible based on world knowledge (e.g. “The mother gave the candle the daughter.”). If participants assume the intended sentence was actually, “The mother gave the candle to the daughter,” that indicates that, in their model of the noise, the deletion of one word is very likely. The noise will be manipulated in various ways to test whether participants can learn specific features of the noise (e.g., that some errors are more likely than others), bind features of the noise to the context (e.g., that some errors are more likely but only given a particular speaker), or flexibly exert control over their own adaptation (e.g., adjust their expectations if they are told the speaker's errors were due to an external disturbance that has been removed). Findings from this research will bring us one step closer to understanding how the language processing system, in healthy individuals, accomplishes such efficient communication in the face of variable noise. Furthermore, understanding how these noise-adaptation mechanisms operate in aphasia will contribute to the development of treatments for persons with aphasia.