We present evidence that the comprehension of illustrated text makes use of the visuospatial sketchpad component of working memory. The results from three experiments demonstrate that the comprehending of texts that are accompanied by pictures interferes with the performance of a spatial short-term memory task more than does the comprehending of texts that have no pictures. A fourth experiment demonstrates that the interference is found only when there is a requirement to comprehend the text; that is, the simple presentation of the texts and their pictures did not degrade performance on the spatial task. Finally, we show that the comprehension of illustrated texts does not differentially interfere with a verbal short-term memory task. These results are consistent with the claim that illustrations accompanying a text encourage the formation of a spatial mental model with the use of the working memory's sketchpad.Pictures increase comprehension of texts in virtually all situations (Levie & Lentz, 1982). Our goal was to investigate one component of the process by which this might occur, the use of the visuospatial component of working memory (Baddeley, 1992) to construct spatial mental models. We tested this idea in a series of experiments in which we used concurrent task methodologies. Our basic finding was that the comprehending of texts with pictures disrupted the performance of a spatial concurrent task, but had no effect on the performance of a verbal concurrent task. Not only was this disruption selective in regard to the concurrent task that was disrupted, it also occurred only within the context of a comprehension task.In most theories of text comprehension, it is agreed that comprehension requires the construction of a referential representation of the meaning of the text: a representation that includes the objects and events being described. Such a representation adds to the literal meaning of the text by incorporating relevant world knowledge. In different theories, referential representations have been given various names, such as situational representations (Perrig & Kintsch, 1985) and mental models (Johnson-Laird, 1983 is about rather than the text itself (Glenberg, Meyer, & Lindem, 1987), so that the structure of the objects being described takes precedence over the structure of the text. Our conception of mental models (Glenberg, Kruley, & Langston, in press) differs from other referential representations in that we take seriously the claim that mental models are analogical. Thus, we are particularly interested in texts that describe spatial layouts or that describe dimensions that are isomorphic to spatial layouts. In these domains, a mental model is hypothesized to consist of representational elements arrayed in the spatial medium of the visuospatial component of working memory.Support for this formulation is provided by some recent research. Glenberg and Langston (1992) considered several ways in which pictures could aid comprehension, including that pictures simply repeat information, that pictu...

Pictures enhance our comprehension of written texts, but the perceptual and cognitive processes that underlie this effect have not been identified. Because integrating the information contained in a text places demands on working memory, the effect of a picture may be to expand the functional capacity of working memory and thereby to facilitate comprehension. Reasoning thus, we predicted that the availability of a diagram would interact with the difficulty of resolving anaphoric references in texts. The resolution of an anaphor distant from its antecedent (which should stress working memory) should benefit greatly from the presentation of a picture, whereas the resolution of an anaphor near to its antecedent should benefit less from the presentation of a picture. Picture availability and distance separating the anaphor from its antecedent were manipulated in experiments involving both cumulative and moving window presentations of texts. Although picture presence and ease of anaphor resolution significantly improved comprehension of the material, no evidence was found for an interaction of these factors. The results are interpreted as consistent both with dual code theory and with aspects of working memory management that do not involve anaphor resolution.

Two experiments were conducted to probe for performance differences between early-semester subjects and late-semester subjects in an introductory psychology subject pool. In Experiment 1A, hypotheses regarding changes in performance included differential effort, changes in subjects' attention to the task, and changes in subjects' sensitivity. A signal detection paradigm was used, and 278 subjects were sampled over the course of the semester. No evidence of changes in subject performance was found, in spite of sufficient power to detect relevant patterns of change. The results of a replication (N = 135) were consistent with those ofExperiment 1A. Experiment 2 (N = 118) employed a text comprehension task to examine possible differences in subject performance with a more cognitive task. Again, there was no evidence to support the idea that latesemester subjects were performing differently from early-semester subjects.In the field of psychology, details of experimental design are thoroughly scrutinized. Particularly when the experimental task is complex, great care must be taken to ensure that the subject is attempting to follow the directions ofthe experimenter. A foundational assumption for any experiment not explicitly investigating subject differences is that the population of subjects sampled is homogeneous. Systematic changes in subject behavior across the semester could influence the pattern of results or reduce the power of an experimental manipulation. Additionally, if systematically different groups are sampled, the generalizability of the conclusions may be seriously reduced.Interestingly, the belief is common among researchers utilizing an undergraduate subject pool that students do vary in important respects across the semester. Subjects participating in the early weeks of the semester are usually believed to follow directions more closely or attend more fully during the experiment.Several studies have been conducted yielding significant differences between early-semester subjects and latesemester subjects on such demographic attributes as academic orientation and locus of control (Evans & Donnerstein, 1974), preoccupation with death (Blatt & Quinlan, 1967), and respect for the scientific enterprise (Holmes, 1967). However, absent theoretical connections between such indices and performance for a given task, these reThe authors would like to thank Arthur Glenberg, Yei-Yu Yeh, Mike Hogan, Steve Morgan, and Bob Kachelski for their helpful comments on an earlier version ofthe manuscript. We thank John Theios and Arthur Glenberg for the use of their lab space and equiprnent in conducting the experiment. The order of the first and second and the third and fourth authors was determined by a coin f1ip. Peter Kruley was partialiy supported by AASERT Grant F4962D-92-J-Q31O from the Air Force Office of Sponsored Research during preparation of this report. Requests for reprints may be sent to William Langston, Department of Psychology, 1202 W. Johnson, Madison, WI 53706. sults provide little support for the inf...