The traditional view of language comprehension is that the meaning of a sentence is composed of the meaning of each word combined into a fully specified syntactic structure.These processes are assumed to be generally completed fully and automatically.However, there is increasing evidence that these processes may, in some circumstances, not be completed fully, and the resultant representation, underspecified.This is taken as evidence for shallow processing and is best typified, we argue, when readers fail to detect semantically anomalous words in a sentence.For example, when asked, “how many animals did Moses take on the Ark?” readers often incorrectly answer “two” failing to notice that it was Noah and not Moses who built the Ark.There has been surprisingly little work carried out on the on-line processing of these types of anomalies, and the differences in processing when anomalies are detected or missed.This thesis presents a series of studies, including four eye-tracking and one ERP study that investigates the nature of shallow processing as evidenced when participants report, or fail to report, hard-to-detect semantic anomalies.The main findings are that semantic anomaly detection is not immediate, but slightly delayed.Anomaly detection results in severe disruption in the eye movement data, and a late positivity in ERPs.There was some evidence that non-detected anomalies were processed unconsciously in both the eye movement record or in ERPs, however effects were weak and require replication.The rate of anomaly detection is also shown to be modulated by processing load and experimental task instructions.The discussion considers what these results reveal about the nature of shallow processing.
PDF
download
878987797
028-85220240
