Cornish, Ian Martin (1984) Memory for story-like material. Doctoral thesis, Durham University.
A series of experiments was performed using the free recall of short story-like passages and a variety of analytic techniques to investigate two aspects of the mental representation of text: the nature of the encoding, and the structural relations among information in memory. The verbatim component of recall was the most variable across several factors, declined fastest after a moderate interval but was unresponsive to the recall accuracy demanded by instructions. It seemed to represent the most accurate information rather than any specifically verbatim encoding, Qualitative analysis demonstrated that verbatim recall was strongly determined by lexical and contextual constraints. Substitutions tended to be higher frequency words, supporting semantic decomposition and the loss of finer components of meaning. The mental representation appeared to consist of information on a number of levels of detail and accuracy, with no evidence for discrete propositions. The literature indicates two alternative approaches to structural factors in discourse comprehension. 'Text-led' theories employ structures peculiar to text and stress causal relations as organising factors. 'Knowledge-led' theories base organisation on the structure of corresponding information in semantic memory and predict that thematic relations, whether causal or not, will dominate memory. Evidence from clause recall contingencies and simple cluster analyses supported the 'knowledge-led' position. The differential behaviour of 'narrative' and 'nodal', apparently organised around verbs (activities) and nouns (actors or objects) respectively, was also consistent with knowledge-led processing. The implications of the results for selective processing during comprehension, and for the role of working memory were discussed. An attempt was also made to identify possible semantic memory structures which might be responsible for organising the episodic representation of information derived from text. Though still sketchy, the present framework is consistent with several recent lines of research and provides direction for future investigation.
|Item Type:||Thesis (Doctoral)|
|Award:||Doctor of Philosophy|
|Copyright:||Copyright of this thesis is held by the author|
|Deposited On:||15 May 2013 15:46|