Semantic constraints and judged preference for interpretations of ambiguous sentences

It is proposed that the degree of sensibleness of sentences is determined by semantic constraints which may be more or less satisfied. Such continuous semantic constraints were examined in two experiments in which subjects judged the likelihood of obtaining each of the interpretations of ambiguous sentences. The sentences were factorially generated by independently varying the degree to which semantic constraints for each interpretation were satisfied. In one experiment, the semantic constraints were manipulated by varying critical words within the ambiguous sentence; in the other experiment, a preceding context sentence was used. The results of both experiments supported the hypotheses that the judged likelihood was a direct function of the relative sensibleness of the interpretations, that semantic constraints determined the degree of sensibleness of each interpretation, and that these semantic constraints are continuous restrictions which are independent of each other and stable from sentence to sentence in which they occur.     

Sentence comprehension in Parkinson's disease: the role of attention and memory.

Sentence comprehension is a complex process involving at least attentional, memory, grammatical, and semantic components. We report three experiments designed to evaluate the impairments underlying sentence comprehension difficulties in nondemented patients with Parkinson's disease (PD). In the first experiment, we asked patients to answer simple questions about sentences which varied in terms of grammatical complexity and semantic constraint. We found that PD patients are significantly compromised in their ability to perform this task. Their difficulties became more prominent as grammatical complexity increased, but they were significantly assisted by semantic constraints that limited possible interpretations of a sentence. Analyses of individual patient profiles revealed heterogeneous performance across the group of PD patients and somewhat inconsistent performance for patients across testing sessions. In the second experiment, we tested the possibility that patients' heterogeneous performance on the sentence comprehension task is due to an impairment in memory or attention, cognitive domains known to be compromised in some PD patients. Although PD patients and control subjects differed on one memory measure, there were no significant correlations between attention and memory performance and the results of the sentence comprehension task. In the final experiment, we manipulated the sentences used in the first experiment in a fashion that stressed the need for memory and attention in a sentence. The results indicated that PD patients are significantly compromised in their ability to attend to certain critical grammatical features of a sentence. A regression analysis identified specific grammatical, semantic, and attentional mechanisms as significant contributors to PD patients' overall sentence comprehension, accounting for over 97% of the variance in their performance. We conclude that there are multiple sources of cognitive difficulty underlying PD patients' sentence comprehension impairment.     

The conceptual base view of categorization

An experiment was designed to show that some categories, called Type C categories, are mediated by an abstract, interpretively derived conceptual base. To this end, each of four groups of subjects ranked 10 sentences (instances) in terms of how well they illustrated the figurative meaning of a proverb (Proverb group), or how well they illustrated the meaning of an excellent interpretation of the proverb (Excellent Interpretation group), or a poor interpretation of the proverb (Poor Interpretation group), or an unspecified, unstated underlying meaning (Control group). The Excellent Interpretation groups' rankings correlated highly with standard ranks established by the Proverb group, but the Poor Interpretation group's and the Control group's ranking were uncorrelated with these two group's rankings. Apparently, the subjects in the Proverb group accomplished their rankings by using a conceptual base or microtheory similar in meaning to the interpretation used by the Excellent Interpretation group. Discussion centered on the question of whether the Classical, Probabilistic, or Exemplar Views of categorization (Smith & Medin, 1981) could account for the results. It was argued that they could not, basically because Type C categories are more dependent upon interpretive processes than the more perceptually based Type P categories to which these views have traditionally been applied.IBM.     

Levels of processing of normal and ambiguous sentences in different contexts

Summary An experiment was conducted to investigate the effect of depth of semantic analysis on the recall of sentences presented for comprehension. The depth of semantic analysis was varied by presenting 48 subjects with 24 unambiguous or lexically ambiguous sentences that were either preceded by a picture or not. Each picture showed either one or both interpretations of the respective ambiguous sentence. The sentence remained on display until the subject had pressed a key to indicate that he had understood its meaning. After the presentation of all the sentences the subjects were tested for recall. Ambiguous sentences were equally well understood as unambiguous sentences, but were better recalled when their ambiguity had been noticed. The subject's awareness of sentence ambiguity, and hence the depth of semantic analysis, was found to depend on the pictorial context in which the sentences were presentend. The pictorial context was also found to affect the depth of processing of unambiguous sentences, which, when presented without a picture, were more time-consuming in comprehension and less well recalled than when preceded by a picture. These findings provide the background for a discussion of the interrelations between the comprehension of sentences, the depth of their semantic processing, and the recall of these sentences.The author gratefully acknowledges the assistance of Lorenz Sichelschmidt in collecting and analyzing data and B. Jankowski in the translation of this paper from an original German version     

The anomaly of anomaly

In order to determine whether mechanically generated sentences, often referred to in the literature as anomalous, are in fact anomalous, subjects in the present experiment were asked to do two different tasks: a paired-associates learning task involving anomalous sentences as responses, and a second task in which they were asked to interpret such sentences. These tasks were counterbalanced across different groups of subjects. Results of the interpretation task showed that a large proportion of subjects were able to give interpretations for anomalous sentences, while learning results indicated that anomalous sentences were more difficult to anticipate than natural sentencesonly when task order was Learn first, Interpret second. In the order Interpret-Learn, differences in ease of learning between anomalous and naturally occurring sentences did not appear. The results of this study demonstrate that anomalous sentences are interpretable, that a range of difficulty for such sentences can be established, that anomalous sentences are learned as easily as naturally occurring ones after interpretation, and that many of the interpretations given to such sentences are metaphoric in character. These findings are discussed in terms of their implications for models of lexical organization.     

Imagery effects and semantic similarity in sentence recognition memory

The idea that subjects often use imagery to discriminate semantically similar sentences was tested in three experiments. In the first experiment, subjects heard subject-verb-object sentences in the context of either a comprehension task or an image-generation task. Their memory for the sentences was tested using a two-alternative forced-choice recognition test in which different types of distractor sentence were used. A sentence semantically similar to the target sentence was one type; a sentence with the same subject and object nouns as the target sentence, but dissimilar in meaning, was another type; and a sentence similar in meaning to one of the stimulus sentences, but not to the target sentence, was a third type. The results showed that the image-generation instructions enhanced later recognition performance, but only for semantically similar test items. A second experiment showed that this finding only holds for high-imagery sentences containing concrete noun concepts. A third experiment demonstrated that the enhanced recognition performance could not be accounted for in terms of a semantic model of test-item discrimination. Collectively, the results were interpreted as providing evidence for the notion that subjects discriminate the semantically similar test items by elaborating the sentence encoding through image processing.     

Agrammatism in a case-inflected language: Comprehension of agent-object relations

Yugoslav agrammatic aphasics and normal control subjects were tested for comprehension of agent-object relations in a series of simple Serbo-Croatian sentences consisting of two nouns (N) and a transitive action verb (V). The availability of nominative and accusative case inflections and a semantic contrast was systematically varied across sentences. Sentences were also varied with respect to the two sequences NVN and VNN. An analysis of subjects' agent-object assignments yielded the following results: While Broca's aphasics did show some sensitivity to case inflections, their ability to process such cues was greatly impaired relative to normal subjects, for whom morphological cues were almost completely deterministic. To a lesser degree, Broca's aphasics were impaired in their ability to employ a strategy of “choose the first noun as agent” when case inflections and semantic contrasts were not available. While Broca's aphasics showed no impairment in their ability to exploit semantic contrasts for agent-object assignment, there was no absolute compensatory increase in the degree to which they relied on semantic cues. Differences in word sequence had no effect on agent-object assignment in Broca's aphasics. Finally, Broca's aphasics frequently responded unsystematically when cues to agent-object relations occurred in isolation or in competition with one another, but when there was a convergence of cues their performance approached that of normal subjects. This result was interpreted in terms of an accessing hypothesis.     

Syntactic form and semantic function

The experiments reported here were designed to investigate the effects on linguistic performance of varying the interaction between the syntactic form of sentences and their semantic function. The experimental task required subjects to decide whether pairs of sentences had the same or a different meaning. The results of Experiment I confirmed the prediction that the times taken to decide about pairs of affirmative and negative sentences would be shorter when the negative was performing its natural function of signalling a change of meaning. To a lesser extent, performance on pairs of active and passive sentences was facilitated when the two sentences meant the same thing. These results were found both with “meaningful” sentence material and with abstract x-y sentences. A second experiment provided a control for the possibility that the results were due to syntactic derivational factors rather than to the semantic function interaction.     

The detection of semantic illusions: Task-specific effects for similarity and position of distorted terms

The effect of task demands on the detection of semantic illusions was investigated. In Exp. 1, subjects were given a detection task with different instructions for accuracy. Less illusions occurred under instructions that stressed accuracy, indicating strategic control of detection rates. In Exp. 2, sentences with dissimilar distorted terms resulted in shorter latencies than sentences with similar distorted terms in a detection task, but in longer response times in a question-answering task. In Exp. 3, the similarity effect was found to vary with the position of the distorted term in combination with task demands. In a verification task, the similarity effect did not differ for the beginning or the end of sentences. In a question-answering task, a significant similarity effect was observed only for distorted terms at the beginning of sentences. We argue that the results indicate minimal depth of semantic processing with respect to different task requirements. Implications for different theoretical accounts of semantic illusions are discussed.     

Semantic memory for metaphor: The conceptual base hypothesis

The idea that people can encode and use an extremely abstract and general form of a complex linguistic (proverb) input-a conceptual base-was examined in two experiments. In Experiment I, each proverb was accompanied by either a conceptually related (good, mediocre, or poor) or an unrelated interpretation. The related interpretations were more effective recall prompts than were the unrelated interpretations, but only for high-imagery proverbs. In Experiment II, subjects wrote interpretations of the proverbs and then received either the proverb subject-noun or a brief story as a prompt. As was the case for the interpretations in Experiment I, the stories did not share any major vocabulary or propositional structure with their proverb source. Nonetheless, the stories were as effective as the nouns. Also, quality of proverb interpretation and of recall performance were positively related, with the correlations involving low-imagery proverbs, and stories, tending to be higher. Both experiments provided support for the conceptual-base notion, and underlined the importance of interpretive context, but more definitive evidence is needed.     

Aptness in metaphor

We propose that in understanding a metaphor, an individual sees a concept from one class or domain in terms of its similarity in two different respects to a concept from another class or domain. The two kinds of similarity are within-domain similarity, or the degree to which two concepts occupy similar positions with respect to their own class or domain; and between-domain similarity, or the degree to which the classes or domains occupied by the concepts are themselves similar. To test this dual notion of similarity, we obtained ratings of the aptness of 64 metaphors from one group of subjects and ratings of their comprehensibility from another group of subjects. The terms of the metaphors had been scaled (based on the ratings of pretest subjects) to give measures of distance both within and between domains. Aptness of metaphors related positively to betweendomain distance, negatively to within-domain distance, and not at all to overall distance. Metaphors are thus perceived as more apt to the extent that their terms occupy similar positions within domains that are not very similar to each other. Comprehensibility also related to aptness. In a second experiment, subjects ranked a set of terms as possible completions for metaphors. For both groups of subjects in this experiment, the rank of an alternative's within-domain distance correlated with its relative popularity. Quantitative models, patterned after a model proposed by D. Rumelhart and A. A. Abrahamson (Cognitive Psychology, 1973, 5, 1–28) for analogical reasoning, afforded significant prediction of the choices of the group of subjects in which all the possible completions of a metaphor were from a single domain. The same models did not predict the choices of a group of subjects in which the possible completions of the metaphors were from different domains.     

Sentence production and working memory

Previous research has shown that subjects can make up sentences faster from related noun pairs than from unrelated pairs. Two experiments tested the proposal that for the related pairs a subject simply has to access stored information but must make up an appropriate relation for unrelated pairs. In both studies the subjects were asked on some trials to remember a digit preload while they made up their sentences. Contrary to an initial prediction, the latencies were faster rather than slower with a preload. Furthermore, a surprise recall test at the end of Experiment 2 showed that there was no difference in the recall of related and unrelated pairs. The results were interpreted as evidence that semantic work was necessary for both types of noun pairs, and that the subjects did less semantic work with a digit preload. This interpretation was further supported by the finding that with the preload the subjects produced a narrower range of semantic relations between the noun pairs.     

Syntactic and semantic contributions to sentence comprehension in agrammatism

Five aphasic subjects, who demonstrated agrammatic speech, and eight control subjects were presented with a sentence-picture matching task in which the factors of syntactic complexity, semantic reversibility, and sentence plausibility were independently varied. A subset of the sentences was patterned after that presented by A. Caramazza and E. Zurif (1976, Brain and Language, 5, 572-583) who concluded that Broca's aphasics rely on semantic constraints instead of syntactic information for sentence comprehension. Our aphasic subjects showed the same pattern of performance on this subset of sentences. However, the results from the full set of sentence materials we tested show that the aphasic subjects could perform some sentence level syntactic analyses, even when syntactic information conflicted with semantic constraints. The aphasic subjects correctly interpreted most active and passive sentences. They failed, however, to assign thematic roles and adjectives in center-embedded relative sentences, and instead relied on nonsyntactic information. These results show that both semantic and syntactic information contributed to sentence comprehension in the aphasic subjects we tested, in contrast to previous claims that syntactic and semantic processes are completely dissociated in this population.     

Ontological realism and sentential form

The standard argument for the existence of distinctively mathematical objects like numbers has two main premises: (i) some mathematical claims are true, and (ii) the truth of those claims requires the existence of distinctively mathematical objects. Most nominalists deny (i). Those who deny (ii) typically reject Quine’s criterion of ontological commitment. I target a different assumption in a standard type of semantic argument for (ii). Benacerraf’s semantic argument, for example, relies on the claim that two sentences, one about numbers and the other about cities, have the same grammatical form. He makes this claim on the grounds that the two sentences are superficially similar. I argue that these grounds are not sufficient. Other sentences with the same superficial form appear to have different grammatical forms. I offer two plausible interpretations of Benacerraf’s number sentence that make use of plural quantification. These interpretations appear not to incur ontological commitments to distinctively mathematical objects, even assuming Quine’s criterion. Such interpretations open a new, plural strategy for the mathematical nominalist.     

Social inference and individual differences in aggression: Evidence for spontaneous judgments of hostility

This research examined the hypothesis that aggressive vs. nonaggressive individuals differ in their spontaneous trait inferences, i. e., inferences made without any conscious intention of inferring characteristics of an actor. We anticipated that spontaneous processing conditions would be more revealing of aggressive/nonaggressive differences than would conditions that prompt deliberate inference processes. We used a cued-recall paradigm. Aggressive and nonaggressive subjects were instructed to memorize sentences that were open to either hostile or nonhostile interpretations. Sentence recall was then cued by either hostile dispositional terms or by words that were linked semantically to the element of the sentences. Within the spontaneous inference condition, semantic cues prompted twice as much recall as hostile cues among nonaggressive subjects, whereas dispositional cues aided recall more than semantic cues among aggressive subjects. As predicted, within the delinerate inference conditions there were no aggressive/nonaggressive differences. The nature of spontaneous vs. deliberate inferential processes and the advantages of spontaneous inference paradigms for testing predictions about schema-based processing in aggression are discussed. © 1995 Wiley-Liss, Inc.     

Normalization in paraphrase and recall: Effects of processing,grammatical article type,and retention interval

Investigations of semantic normalization of discourse have not generally controlled for perceived connectiveness of materials, depth of processing, and retention interval. The current experiment investigated semantic normalization in paraphrase and recall as a function of whether sentences were paraphrased or merely copied, of whether sentences were perceived by subjects to be connected or unrelated, and of retention interval. Results indicated that significantly more normalization occurred in paraphrase than in recall, that more normalization and better recall occurred when sentences were perceived as related, and that there was more normalization and better recall when sentences were paraphrased than when they were copied. No significant differences in normalization were found as a function of retention interval. Results are discussed in terms of depth of processing and a need for learners to integrate congruent and discrepant material into cohesive entities by use of grammatical articles, article changes, and intrusions that psychologically connect unrelated sentences.     

Does the Quine/duhem Thesis Prevent Us from Defining Analyticity?

Quine claims that holism (i.e., the Quine-Duhem thesis) prevents us from defining synonymy and analyticity (section 2). In Word and Object, he dismisses a notion of synonymy which works well even if holism is true. The notion goes back to a proposal from Grice and Strawson and runs thus: R and S are synonymous iff for all sentences T we have that the logical conjunction of R and T is stimulus-synonymous to that of S and T. Whereas Grice and Strawson did not attempt to defend this definition, I try to show that it indeed gives us a satisfactory account of synonymy. Contrary to Quine, the notion is tighter than stimulus-synonymy – particularly when applied to sentences with less than critical semantic mass (section 3). Now according to Quine, analyticity could be defined in terms of synonymy, if synonymy were to make sense: A sentence is analytic iff synonymous to self-conditionals. This leads us to the following notion of analyticity: S is analytic iff, for all sentences T, the logical conjunction of S and T is stimulus-synonymous to T; an analytic sentence does not change the semantic mass of any theory to which it may be conjoined (section 4). This notion is tighter than Quine's stimulus-analyticity; unlike stimulus-analyticity, it does not apply to those sentences from the very center of our theories which can be assented to come what may, even though they are not synthetic in the intuitive sense (section 5).     

Semantic distance and single cue use in category search

Perceptual classification may be based either on the physical features of target and background items or on the semantic attributes of the presented items. In this paper we used enumeration tasks to study the role of semantic features in a categorial classification task. This means that subjects were asked to count the number of target words in a display belonging to one semantic category among a number of background items of other categories. Our goal was to study the decision logic in category search by manipulating target background conditions and the semantic distance between target and background classes. In the first experiment we found that the larger the semantic distance between targets and background words, the easier it was to find the targets. In the second experiment we found a "pop-out" effect, in which subjects could use and benefit from a single distinctive semantic feature, "part-likeness", in categorial classification. The results of the two experiments imply that the categorization decision logic is basically the same in physical and semantic perceptual classification.     

Bias in interpretation of ambiguous sentences related to threat in anxiety

In the 1st of 2 experiments, currently clinically anxious, recovered clinically anxious, and normal control subjects were presented with a mixture of unambiguous and ambiguous sentences; both threatening and nonthreatening interpretations were possible for the latter. A subsequent recognition-memory test indicated that the currently anxious subjects were more likely than normal control and recovered anxious subjects to interpret the ambiguous sentences in a threatening fashion rather than in a nonthreatening fashion. This suggests that the biased interpretation of ambiguity found in currently anxious subjects reflected their anxious mood state. A 2nd experiment established that the difference in interpretative processes between currently anxious and control subjects was not due to response bias and that the interpretative bias was a reasonably general one.     

The role of imagery,analogy, and instantiation in proverb comprehension

The role of imagery, analogy, and instantiation in understanding proverbs was examined in the light of the Conceptual Base Theory (CBT). The CBT proposes that proverb comprehension involves four phases and that (a) after Problem Recognition, imagery is often used during the Literal Transformation Phase with the intention of constructing a figurative meaning but that imagery does not enter into this meaning; (b) figurative meaning is the solution to a four-part, a:b::c:d, analogy, which occurs during the Figurative Phase; (c) the Instantiation Phase can be entered only if a figurative meaning is constructed, but once instantiation occurs it serves to refine this meaning. A transfer paradigm was used to test these hypotheses. During acquisition, the subjects were presented proverbs and given rating tasks in connection with one of the following: pictures that illustrated the literal proverb information (picture groups), instructions to image the literal proverb information (imagery groups), or analogies involving literala andb terms drawn from the proverb andc andd terms that constituted an interpretation of the proverb (analogy group), concrete verbal instances that illustrated the proverbs' figurative meanings (instance group), or an analogy and an instance (analogy-instance group). During transfer, all subjects attempted to distinguish between novel sentences that were either positive or negative instances of the figurative meaning of the acquisition proverbs. The picture and imagery groups, which were assumed to have used imagery without the intention of constructing a figurative meaning, performed at chance. The analogy group performed above chance and on a par with the instance group subjects, who theoretically had constructed and solved an implicit analogy. The analogy-instance group showed superior performance. Discussion centered on the pattern of the results in relation to the CBT and on the role of imagery and analogy in understanding proverbs as opposed to metaphors.The authors wish to thank Suzanne Leake for help in designing the experiment, preparing materials, and running subjects.