Automatic syntactic computation and use of semantic information during sentence comprehension

Summary The paper represents a contribution to the issue of the differential effects of syntactic, semantic and pragmatic information in the process of sentence comprehension. The results of four experiments are reported, in which sentences containing syntactic, semantic and pragmatic violations were presented with different tasks to see how these violations affected language comprehension, how easily they were detected and whether they affected sentence comprehension even when they were not explicitly recognized. The main results can be summarized as follows. First, the most relevant cues used to comprehend a sentence are not so much syntactic, but semantic or pragmatic. Second, with difficult or anomalous linguistic input, syntactic cues become more important. Third, even when the reader is not aware of syntactic violations, these seem to affect his processing of the linguistic input. This suggests the hypothesis of an automatic computation of syntactic information, the output of which is, however, used only when evidence coming from other sources, semantic or pragmatic, is insufficient or not clear enough.     

A model of the time course and content of reading

This paper describes a computer simulation of reading that is strongly driven by eye fixation data from human readers. The simulation, READER, is a natural language understanding system that reads a text word by word and whose processing cycles on each word have some correspondence with the human gaze duration on that word. READER operates within a newly developed information processing architecture, a Collaborative, Activation-based, Production System (CAPS) that permits the modeling of the temporal properties of human comprehension. CAPS allows for concurrent, collaborative execution of processes operating at different levels of analysis. As READER encounters each successive word, the word is operated on by processes at the levels of word encoding, lexical access, syntactic and semantic analysis, and referential and schema-level processes. Like human readers, READER uses a strategy of immediacy of comprehension, attempting to interpret each word as soon as it is encountered, rather than unnecessarily buffering information. A major contribution of this simulation is its use of human performance characteristics in constraining and determining the model's mechanisms.     

Repetition in the transcortical aphasias

Transcortical aphasic patients were assessed on a repetition task comprised of both well-formed and deviant sentences. The patients faithfully repeated those sentences that were factually incorrect but grammatically well formed, and those sentences that were ungrammatical because of selection restriction violations. In contrast, presented with sentences featuring only minor syntactic violations (e.g., lack of number agreement), the patients spontaneously, and without awareness, resisted exact repetition; moreover, the resultant changes most often served to correct the syntactic deviations. Repetitions of sentences of this latter type also revealed a difference between the patients as a function of level of comprehension. The transcortical motor aphasic patient was influenced by semantic factors in his alterations of the minor syntactic violations, while the transcortical sensory aphasic was not. These findings are discussed in relation to the notion of autonomous syntactic processing. They are taken to suggest that syntactic facts such as number agreement are represented at a level in the processing chain that is distinct from that of semantic representation.     

中文阅读之句法分析历程初探

该文报告的是香港中文大学心理学系的认知心理实验室中最近完成的两个与句法分析有关的实验。第一个实验探讨阅读中文时词汇、句法以及语意资料之触发和运用的时序问题。第二个实验研究的是中文读者所使用的句法分析策略。在两个实验中,基本的作业是让被试阅读在电脑屏幕上呈现的短文。文中每个字的阅读时间则由电脑记录作为反映阅读理解即时处理历程的指标。文中除了报告实验的结果外,还依据这些结果讨论了读者在阅读中文时所使用的句法分析策略以及句法资料的触发和使用历程。     

Aphasics' selective deficits in appreciating grammatical agreements

This study attempted to dissociate in aphasic patients different aspects of grammatical capacity. Subjects were asked to evaluate sentences containing violations of agreements between pronouns and verbs. Agreements which are considered primarily structural in nature were violated (e.g., a surface object case like “them” placed in a surface subject sentential slot) while other violated agreements seemed to involve both structural and semantic information (e.g., lack of agreement between a pronoun and a verb in terms of the number of people performing the act). These violations were couched in one of three types of sentential frames which varied in terms of syntactic complexity (e.g., active declarative vs. passive syntactic voices). The results revealed that Broca's aphasics found agreement violations difficult to detect in complex syntactic frames. They were quite successful at detecting even violations of the largely structural “case” type of agreement, however, when couched in simpler syntactic frames. Fluent aphasics encountered more difficulty detecting violations of agreements involving both semantic and structural information than agreements which were primarily structural in nature, that is, regardless of the syntactic frame. These unique performance profiles suggest that Broca's aphasics may be agrammatic only with respect to certain aspects of a sentence's structure, and that fluent aphasics may also experience some selective—but different—grammatical deficits.     

句子理解的关键——对句法和语义关系的再探讨

对于句子理解中句法和语义的关系,传统的模块化观点认为,句法信息优先于语义信息,句法加工支配着句子的理解过程。近期,这种强调句法优先的理论受到了一系列ERP（Event-related potentials）研究结果的挑战,这些研究从不同角度论证了语义信息对句法加工及句子整合加工的重要作用。基于这些研究发现,研究者提出句子理解过程中句法和语义的关系并非一成不变：一些情况下可能是句法支配语义加工,而另一些情况下则可能是语义引导句法加工     

Canceling updating in the comprehension of counterfactuals embedded in narratives

Participants were given counterfactual sentences--for example, "If Mary had won the lottery she would have bought a Mercedes car"--or factual sentences--for example, "Because Mary won the lottery, she bought a Mercedes car"--embedded in short narratives. Reading times showed that readers were immediately sensitive to the special status of counterfactual information (Experiment 1). In addition, probe-recognition latencies demonstrated that old information was more accessible in counterfactual than in factual stories, and new information was equally accessible in both kinds of stories (Experiment 2). However, after reading additional clauses, new information became less accessible in counterfactual than in factual stories (Experiment 3). These results suggest that counterfactual events are momentarily represented but are later suppressed and the readers' attention goes back to previous events in the story.     

An event-related fMRI study of syntactic and semantic violations

We used event-related functional magnetic resonance imaging to identify brain regions involved in syntactic and semantic processing. Healthy adult males read well-formed sentences randomly intermixed with sentences which either contained violations of syntactic structure or were semantically implausible. Reading anomalous sentences, as compared to well-formed sentences, yielded distinct patterns of activation for the two violation types. Syntactic violations elicited significantly greater activation than semantic violations primarily in superior frontal cortex. Semantically incongruent sentences elicited greater activation than syntactic violations in the left hippocampal and parahippocampal gyri, the angular gyri bilaterally, the right middle temporal gyrus, and the left inferior frontal sulcus. These results demonstrate that syntactic and semantic processing result in nonidentical patterns of activation, including greater frontal engagement during syntactic processing and larger increases in temporal and temporo-parietal regions during semantic analyses.     

Dissociation of algorithmic and heuristic processes in language comprehension: Evidence from aphasia

Three groups of aphasic patients, Broca's, Conduction, and Wernicke's, and a nonaphasic patients control group were tested for comprehension of object-relative center-embedded sentences. The sentences were of three types: sentences in which semantic constraints between words allowed the subjects to assign a correct semantic reading of the sentence without decoding the syntax, sentences in which semantic constraints were relaxed and for which a correct reading was only possible with knowledge of syntactic relationships among words, and sentences which described highly improbable events. The subjects' task was to choose which of two pictures captured the meaning expressed in the sentence. Broca's and Conduction aphasics performed near perfectly on sentences where they could use semantic information. Their performance dropped to chance when they had to use syntactic information. These results support a neuropsychological dissociation of heuristic and algorithmic processes based primarily, though not exclusively, on semantic and syntactic information, respectively.     

Comprehending Semantic and Grammatical Violations in Italian. N400 and P600 Comparison with Visual and Auditory Stimuli

Arguments about the existence of language-specific neural systems and specifically about the independence of syntactic and semantic processing have focused on the event-related brain measures (ERPs) as tool to monitoring moment-by-moment the cognitive processes underlaid. In the present experiments, the available evidence indicates that the ERP response to semantic anomalies is at least partially distinct from the ERP response to syntactic anomalies and that two distinct processes are activated in sentences comprehension. ERPs were recorded from 10 electrodes while subjects read (Experiment 1) or listened (Experiment 2) to sentences containing semantic or syntactic violations. Final-words that were inconsistent with the sentence context elicited a negative-going wave at about 400 ms poststimulus, whereas penultimate-word incongruous with the grammatical structure (subject–verb non-agreement) elicited a positive-going wave about 600 ms poststimulus. No differences based on the perceptual modality of the stimulus (visual or auditory) nor different ERP correlates as a function of task-relevance (explicit/implicit task induction) were found. The implications of our results for Italian language are explained.     

汉、英句法结构相似性与第二语言的熟练度对第二语言句法加工的影响

采用事件相关电位（ERP）技术,考察了汉、英句法结构相似性与第二语言（英语）熟练度对第二语言句法加工的影响。实验材料包含与汉语句法结构相似和不同的两种英语句法结构违反的句子。结果发现：与正确句子相比,相似句法违反句在高熟练组被试中,引发正向的ERP波形偏转（150-800ms）,而在低熟练组被试中,只引发了降低的N400（300-550ms）效应;对于不同句法违反句,无论高低熟悉程度的被试,均未出现显著的ERP效应。研究结果表明,第二语言句法与母语的相似性越高,加工起来更容易;第二语言熟练度水平对第二语言句法加工具有一定的促进作用,但受到两种语言句法结构相似性的影响,相似性越高,熟练度的促进作用越大。     

Reply to David Kemmerer's "a critique of Mark D. Allen's 'the preservation of verb subcategory knowledge in a spoken language comprehension deficit'"

Allen [Allen, M. D. (2005). The preservation of verb subcategory knowledge in a spoken language comprehension deficit. Brain and Language, 95, 255-264] presents evidence from a single patient, WBN, to motivate a theory of lexical processing and representation in which syntactic information may be encoded and retrieved independently of semantic information. In his critique, Kemmerer argues that because Allen depended entirely on preposition-based verb subcategory violations to test WBN's knowledge of correct argument structure, his results, at best, address a "strawman" theory. This argument rests on the assumption that preposition subcategory options are superficial syntactic phenomena which are not represented by argument structure proper. We demonstrate that preposition subcategory is in fact treated as semantically determined argument structure in the theories that Allen evaluated, and thus far from irrelevant. In further discussion of grammatically relevant versus irrelevant semantic features, Kemmerer offers a review of his own studies. However, due to an important design shortcoming in these experiments, we remain unconvinced. Reemphasizing the fact the Allen (2005) never claimed to rule out all semantic contributions to syntax, we propose an improvement in Kemmerer's approach that might provide more satisfactory evidence on the distinction between the kinds of relevant versus irrelevant features his studies have addressed.     

Lateralization of auditory language functions: a dynamic dual pathway model

Spoken language comprehension requires the coordination of different subprocesses in time. After the initial acoustic analysis the system has to extract segmental information such as phonemes, syntactic elements and lexical-semantic elements as well as suprasegmental information such as accentuation and intonational phrases, i.e., prosody. According to the dynamic dual pathway model of auditory language comprehension syntactic and semantic information are primarily processed in a left hemispheric temporo-frontal pathway including separate circuits for syntactic and semantic information whereas sentence level prosody is processed in a right hemispheric temporo-frontal pathway. The relative lateralization of these functions occurs as a result of stimulus properties and processing demands. The observed interaction between syntactic and prosodic information during auditory sentence comprehension is attributed to dynamic interactions between the two hemispheres.     

句法和语义在汉语简单句理解中的作用

两个实验探讨句法和语义在汉语简单句加工中的作用及其相互关系。实验一采用施事判断任务,发现句法作用与语义可逆性的作用都很显著;实验二采用快速句子判断任务,发现句法作用是主要的,语义可逆性无显著作用;两个实验都发现句法与语义可逆性之间的交互作用不显著。这些结果表明：（１）在汉语简单句理解的过程中句法因素的作用是主要的,语义因素的作用是次要的,只要当实验任务更多地强调语义分析时,语义因素才有显著的作用。     

The effect of lexical, pragmatic, and morphological violations on reading time and deviance ratings of English and German sentences

Effects on reading time and deviance ratings of word-choice violations were studied at the surface code, propositional textbase, and situation model levels of representation in German and English sentences. Lexical violations (propositional textbase level) such as The housewife massaged the bread dough were rated as more deviant from normal language than were pragmatic violations (situation model level) such as The police officer shot the parking violator, although the pragmatic violations took longer to comprehend. The addition of a morphological (surface code) violation through the wrong form of an article (e.g., a ugly car) decreased reading time but only in cases where that material was deeply processed and where the morphology carried substantial syntactic information (e.g., in German, the definite article carries case, gender, and number information). The results thus confirmed the operation of comprehension strategies at multiple levels of representation and identified some cross-linguistic generality in comprehension and some language differences.     

Making sense of sentences in schizophrenia: electrophysiological evidence for abnormal interactions between semantic and syntactic processing

Event-related potentials to critical verbs were measured as patients with schizophrenia and healthy controls read sentences word by word. Relative to their preceding context, critical verbs were (a) congruous, (b) incongruous and semantically unrelated to individual preceding words (pragmatic-semantic violations), (c) incongruous but semantically related to individual preceding words (animacy-semantic violations), or (d) syntactically anomalous. The N400 was modulated normally in patients, suggesting that semantic integration between individual words within sentences was normal in schizophrenia. The amplitude of the P600 to both syntactic and animacy-semantic violations was reduced in patients relative to controls. The authors suggest that, in schizophrenia, an abnormality in combining semantic and syntactic information online to build up propositional meaning leaves sentence processing to be primarily driven by semantic relationships between individual words.     

The Time Course of Syntactic Activation During Language Processing: A Model Based on Neuropsychological and Neurophysiological Data

This paper presents a model describing the temporal and neurotopological structure of syntactic processes during comprehension. It postulates three distinct phases of language comprehension, two of which are primarily syntactic in nature. During the first phase the parser assigns the initial syntactic structure on the basis of word category information. These early structural processes are assumed to be subserved by the anterior parts of the left hemisphere, as event-related brain potentials show this area to be maximally activated when phrase structure violations are processed and as circumscribed lesions in this area lead to an impairment of the on-line structural assignment. During the second phase lexical-semantic and verb-argument structure information is processed. This phase is neurophysiologically manifest in a negative component in the event-related brain potential around 400 ms after stimulus onset which is distributed over the left and right temporo-parietal areas when lexical-semantic information is processed and over left anterior areas when verb-argument structure information is processed. During the third phase the parser tries to map the initial syntactic structure onto the available lexical-semantic and verb-argument structure information. In case of an unsuccessful match between the two types of information reanalyses may become necessary. These processes of structural reanalysis are correlated with a centroparietally distributed late positive component in the event-related brain potential. The different temporal and topographical patterns of the event-related brain potential. as well as some aspects of aphasics′ comprehension behavior are taken to support the view that these different processing phases are distinct and that the left anterior cortex, in particular, is responsible for the on-line assignment of syntactic structure.     

语义和句法信息在副中央凹加工中的作用

读者能够从副中央凹中提取到什么类型的信息, 是当前阅读眼动研究领域关注的焦点问题.当前普遍认为低水平信息(如正字法信息)可以在副中央凹中得以加工, 但是高水平的信息(如语义和句法信息)能否从副中央凹中进行提取存在争议.本文总结了近年来高水平信息在副中央凹加工的研究进展, 包括拼音文字和非拼音文字(如中文)阅读过程中语义和句法预视效益的研究现状及影响因素, 当前眼动控制模型(如E-Z读者和SWIFT)对预视效益的解释和不足之处, 最后提出未来关于语义和句法预视加工的研究方向.