Sentence memory: A theoretical analysis

Abstract

How sentences from a discourse are recognized or verified can be explained by combining theories of item recognition derived from list-learning experiments with hypotheses about the representation of text in memory within the framework of the construction-integration model of discourse comprehension. The implications of such a theory of sentence recognition are worked out for two experimental situations. In the first experiment, subjects read brief texts and were then tested for recognition with verbatim old sentences, paraphrases, inferences, and contextually related and unrelated new distractor sentences after delays from 0 to 4 days. Differential decay rates for the wording and meaning of the text and for situational information were observed. The theory provides a good quantitative account of the data. In the second experiment, the speed-accuracy trade-off in sentence verification for two subject groups with different prior knowledge was studied for old verbatim sentences and inferences. Qualitative predictions derived from the theory with the parameter estimates from the first study were in agreement with the data. Readers without an adequate situational understanding (novices) were found to make quick judgments based on surface and textbase characteristics of the test sentences, while experts in addition utilized their situation model successfully, which required more processing time. © 1990 Academic Press, inc.A large number of experiments on recognition memory exist for lists of words or pictures. Several models of recognition memory are available today which account very well for most of the phenomena observed in these experiments. Can these theories also account for experimental data when the materials used are not lists of items, but coherent discourse? By combining the essential features of current models of recognition memory developed for list- learning studies with a model of discourse comprehension and assumptions about the representation of discourse in memory, a model of sentence recognition and sentence verification can be obtained that accounts for major features of sentence-recognition data. Thus, we do not propose developing a new model for sentence memory. Instead, we shall combine existing models of listlearning and text-comprehension processes to derive a theoretical analysis of sentence recognition and verification.We begin by comparing three current models of item recognition (Gillund & Shiffrin, 1984;Hintzman, 1988;Murdock, 1982) and determine their common essential features, to be used for modelling sentence memory. We will then review some notions

Eyewitness memory: Balancing the accuracy, precision and quantity of information through metacognitive monitoring and control

“…Also, studies examining memory over time have shown that memory for fine grained information deteriorates relatively rapidly, while memory for more coarse grained information endures (e.g., Christiaansen, 1980;Conway, Cohen, & Stanhope, 1991;Koriat, Levy-Sadot, Edry, & de Marcas, 2003;Stanhope, Cohen, & Conway, 1993). For example, memory for gist, or general meaning, endures longer than memory for surface form and verbatim information (e.g., Kintsch, Welsch, Schmalhofer, & Zimmy, 1990). There is also evidence that memory for categories is retained longer than memory for items (Dorfman & Mandler, 1994), and that there is a retention advantage of basic level information (Pansky & Koriat, 2004).…”

Constructing arguments from multiple sources: Tasks that promote understanding and not just memory for text.

“…Because these tasks require students to go beyond the information they were given, inference and analogy tasks rely on the development of a situation model (Kintsch, Welsch, Schmalhofer, & Zimny, 1990). Thus, if the more transformed, integrated, and causal essays are related to the development of a model of the text, then students who write such essays should perform better on inference and analogy tasks, suggesting they have indeed gained a conceptual understanding of the subject matter.…”

Recognition of script-typical versus script-atypical information: Effects of cognitive elaboration

“…Therefore, source memory should be relatively constant across item types and presentation forms. More specifically, source memory is likely to be moderate or poor for all items because memory for surface information vanishes rather quickly when attention is directed toward semantic aspects of texts (see, e.g., Alba & Hasher, 1983;Kintsch & Ericsson, 1996;Kintsch, Welsch, Schmalhofer, & Zimny, 1990;Schmalhofer & Glavanov, 1986).…”