Murray Singer

Inferring consequences in story comprehension

The study asked whether readers infer the consequences of events described in stories. Forward and backward inferences were distinguished: only backward inferences contribute to the coherence of a message. The subjects read stories of 9 to 11 sentences and then answered eight questions. The time needed to answer forward inference questions was about the same as for questions paraphrasing the story, but over 0.2 seconds longer than for questions repeating part of the story and for backward inference questions. It was concluded that backward consequence inferences are more reliably drawn during the course of reading than are forward consequence inferences.

Constructing and Validating Motive Bridging Inferences

Understanding Jane left early for the birthday party, She spent an hour shopping at the mall requires detecting that the first statement motivates the second. The validation model states that before accepting this bridging inference, the reader validates it with reference to relevant knowledge. In particular, a mediating idea is first derived from the text outcome and its candidate motive. If the mediating idea is supported by general knowledge, then the inference has been validated. In tests of this anaylsis, experimental subjects read motive or control sequences and then answered questions probing the knowledge hypothesized to validate the motive inferences, such as Do birthday parties involve presents? Five experiments confirmed that understanding motive sequences facilitates validating knowledge. A control procedure also refuted a priming counterexplanation of these effects (Experiment 1). Validation processing obtained for motive-outcome statements separated by two to four sentences in coherent sequences (Experiments 2 to 4). Inferred and explicit validating knowledge had a similar representational status (Experiment 3). Whereas proofreading abolished the validation effect, a reading strategy promoting causal processing did not enhance it (Experiment 4). A delayed priming procedure indicated that validating knowledge is integrated with the text representation (Experiment 5). The implications of these findings for the constructionist and minimal inference analyses were explored. The validation effects were simulated using construction-integration model.

Individual differences in bridging inference processes

The role of individual differences in bridging-inference processing was studied. Students (n=135) read passages of short to moderate length. After each one, they answered corresponding questions about inferences that bridged causally related ideas that were either near or far apart in the text. The main hypothesis was that local bridging-inference processing is facilitated by the reader’s predisposition to access pertinent knowledge during comprehension Regression analyses provided support for this proposal and indicated that greater working-memory capacity and vocabulary knowledge promote inference processing. The following relationships between the predictors and inference processing were proposed: Knowledge access promotes the co-occurrence in working memory of the text ideas and knowledge needed to compute the bridge. Working-memory capacity enhances the likelihood that needed antecedent ideas will be available to the bridging processes. Vocabulary knowledge may promote inference processing because-unfamiliar word meanings place more demands on working-memory resources than do familiar meanings.

The role of working memory capacity and knowledge access in text inference processing

Individual differences in drawing bridging inferences during text comprehension were examined. We measured reader differences in working memory capacity, using the reading span task, and in access to relevant knowledge, using Potts and Peterson’s (1985) integration task. The dependent measure of greatest concern was answer time about facts posited to validate the bridging inference. Reading span and access were negligibly correlated, an outcome that supports their independence. Answer times were lower both for high reading span and high-access readers. In addition, readers who were either high on both reader traits or low on both traits exhibited qualitatively different inference effects from the typical pattern. It is proposed that knowledge access during comprehension is facilitated by the extraction of integrated situation models from text and that it is individuals with efficient reading processes who can construct these models.

Constructing inferences in expository text comprehension

Prior studies have suggested that causal bridging inferences are constructed during the comprehension of narrative text (Singer, Halldorson, Lear, & Andrusiak, 1992) but not expository text (Noordman, Vonk, & Kempff, 1992). Experiment 1 revealed that results consistent with both of these conclusions can be measured simultaneously. In Experiments 2 to 4, subjects read expository texts only, and, after each one, answered a question that probed the inference of interest. For Noordman et al.’s original materials, the answer time patterns replicated a result pattern diagnostic of the failure to, compute bridging inferences. However, a pattern consistent with on‐line inference computation was detected with a new set of materials, as long as the readers were not rushed in their examination of the text (Experiments 2 and 4). It was concluded that, whereas, inference processing may be impeded by the relative lack of familiarity of the content of expository text, it is not strictly precluded. Rather, the results in...

The role of case‐filling inferences in the coherence of brief passages∗

Three experiments were conducted to determine whether inferences which function so as to preserve the coherence of a passage are drawn in the course of comprehension. The study focused upon the implications of verbs and verb phrases concerning the concepts that fill the related agent, patient, and instrument cases. In all three experiments, judgments about the “necessary” inferences were compared with those about directly expressed ideas and about inferences that were possible but not crucial for coherence. In experiment 1, the rated recognition of necessary inference tests was intermediate to the scores for direct ideas and unnecessary inferences. Experiment 2 introduced an important control for physical similarity. The rated recognition of necessary inferences was comparable to that for “direct” tests, and higher than the unnecessary inference scores. In Experiment 3, subjects needed the same amount of time to verify necessary inference and direct tests, and 245 msec more for unnecessary inferences. It ...

Psychology of language : an introduction to sentence and discourse processes

This comprehensive volume addresses the central issues of sentence and discourse processes, with particular emphasis placed on reading and listening comprehension. The text material is accessible to both upper-level undergraduate and graduate students and informative for professionals and educators. In this regard, this uncommon volume identifies the logic of both the specific experimental manipulations that are described, and the more general on-line and memory measures frequently invoked. The principles presented in the text are supported by hundreds of numbered and unnumbered examples, and by precise tables and figures.

Processes of inference during sentence encoding

Four experiments were designed to determine whether implicit instruments are “computed” and stored during sentence encoding. Subjects read pairs of related sentences and indicated when they understood the second member of each pair. In Experiment 1, response times were longer for test sentences that, relative to the antecedent, mentioned implicit rather than explicit instruments. Experiment 2 revealed this difference to be stable over a range of reading times and two phases of practice. The results were interpreted as suggesting that the inferences in question are not computed, or at most, are partially drawn during encoding In Experiment 3, the procedure was employed to distinghish four degrees of relation between propositions and the inferences they permit. Experiment 4 examined the agent, patient, and instrument cases. It was shown that subjects need more time to verify both true and false inference tests than to verify their direct counterparts.

Effect of delay on recognition decisions: Evidence for a criterion shift

Recent evidence indicates that in intermixed recognition testing of different stimulus classes, people can apply different decision criteria (acriterion shift) to stimulus classes distinguished by the study-test delay (Singer, Gagnon, & Richards, 2002), but not by a conspicuous strength manipulation (Stretch & Wixted, 1998b). In an attempt to reconcile these differences, we applied Singer et al.’s text retrieval method to word recognition. People first studied blocked items from each of five categories. After a delay, five new category lists were presented. After each one, the participants recognized intermixed targets and distractors from the current category and one of the earlier ones. At delays of up to 40 min, the answering criteria for immediate and delayed categories were indistinguishable. At delays of 2 days, in contrast, however, bothyes—no and confidence-rating data indicated that more lenient criteria were applied to delayed than to immediate test items. This suggests that people can use the delay between study and test to flexibly adjust the decision criteria of word recognition.

Validation in Reading Comprehension

Language comprehension involves analysis at the level of the word, sentence, and message and the integration of message meaning with the prior discourse and world knowledge. Contemporary research converges on another facet of comprehension: the validation of message consistency. Existing evidence already favors several principles in validation of reading and listening. Validation is initiated immediately and is routine rather than requiring intentional strategies. Successful validation is a precondition to updating the situational representation of the message. Validation applies to discourse inferences as well as explicit assertions. Finally, the memory-retrieval processes that enable validation closely resemble those of intentional discourse memory. Competing observations of people’s validation failures are proposed to systematically stem from features of the message, understander, and comprehension task. Therefore, theoretical analysis that accommodates both successful and deficient language validation ought to be attainable.