Processing local and unbounded dependencies: A unified account

This paper proposes an account of incremental sentence processing and the initial stage of syntactic ambiguity resolution, based on the claim that the processor seeks to provide semantic interpretations for sentence fragments as soon as it possibly can. In this model, there is no fundamental distinction between local and unbounded dependencies. The processor employs a version of categorial grammar based on dependency grammar, in which dependency constituents are derived from dependencies between words and are permitted to overlap. The processor seeks to form dependency constituents as soon as it can, and to give interpretations to these fragments immediately. The initial stage of ambiguity resolution is determined by theprinciple of dependency formation, under which the processor automatically chooses an analysis that allows a single dependency constituent to be formed in preference to one that does not. The motivation is semantic: Such an analysis maximizes the amount of incremental interpretation that is possible. But if more than one analysis is compatible with the formation of a single constituent, the processor can appeal to a range of sources of nonsyntactic information in making its choice. I show how this account can capture a range of psycholinguistic evidence without positing any fundamental distinction between local and unbounded dependencies.     

Early, Mandatory, Pragmatic Processing

The experiment reported in this paper used a delayed same/different sentence matching task with concurrent measurement of eye movements to investigate three questions: whether pragmatic plausibility effects are restricted to certain phrasal environments; how rapidly such effects are shown in on-line sentence processing; and whether they are a product of optional, high-level, inferential processes. The results clearly show that plausibility effects are not restricted to low–level phrasal units and that they appear to arise as a necessary consequence of the process responsible for deriving basic sentence meaning. The rapid and highly localized nature of the effects supports a view of sentence processing involving incremental interpretation of the earliest available syntactic representations. We argue that the apparently mandatory nature of plausibility effects, coupled with their insensitivity to repetition context, presents difficulties for both modular and interactive views of sentence processing.     

A probabilistic model of semantic plausibility in sentence processing

Experimental research shows that human sentence processing uses information from different levels of linguistic analysis, for example, lexical and syntactic preferences as well as semantic plausibility. Existing computational models of human sentence processing, however, have focused primarily on lexico-syntactic factors. Those models that do account for semantic plausibility effects lack a general model of human plausibility intuitions at the sentence level. Within a probabilistic framework, we propose a wide-coverage model that both assigns thematic roles to verb–argument pairs and determines a preferred interpretation by evaluating the plausibility of the resulting ( verb , role , argument ) triples. The model is trained on a corpus of role-annotated language data. We also present a transparent integration of the semantic model with an incremental probabilistic parser. We demonstrate that both the semantic plausibility model and the combined syntax/semantics model predict judgment and reading time data from the experimental literature.     

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.     

A dual system for producing self-repairs in spontaneous speech: Evidence from experimentally elicited corrections

This paper presents a cognitive theory on the production and shaping of self-repairs during speaking. In an extensive experimental study, a new technique is tried out: artificial elicitation of self-repairs. The data clearly indicate that two mechanisms for computing the shape of self-repairs should be distinguished. One is based on the repair strategy called reformulation, the second one on lemma substitution. W. Levelt's (1983, Cognition, 14, 41–104) well-formedness rule, which connects self-repairs to coordinate structures, is shown to apply only to reformulations. In case of lemma substitution, a totally different set of rules is at work. The linguistic unit of central importance in reformulations is the major syntactic constituent; in lemma substitutions it is a prosodie unit, the phonological phrase. A parametrization of the model yielded a very satisfactory fit between observed and reconstructed scores. © 1987 Academic Press, Inc.     

Question production in agrammatism: the tree pruning hypothesis

This study investigated question production in agrammatic aphasia, focusing on the comparison between Wh questions and yes/no questions and on the interaction between the question production deficit and language-specific properties. A total of 16 agrammatic aphasics (13 Hebrew speakers, 2 Palestinian Arabic speakers, and 1 English speaker) participated in the study, which included sentence elicitation and repetition tasks. In addition, the patients' spontaneous speech, containing 2272 utterances, was analyzed. The main findings were that Hebrew- and Arabic-speaking agrammatics encounter severe difficulties in Wh question production but retain the ability to produce yes/no questions. English-speaking agrammatics do not show this dissociation and can form neither Wh nor yes/no questions. These dissociations as well as the error pattern, are explained by reference to the Tree Pruning Hypothesis, according to which the highest nodes of the syntactic tree, which are required for Wh questions in Hebrew, Arabic, and English and for yes/no questions in English, are impaired or inaccessible in agrammatism.     

On-line syntactic processing in aphasia: studies with auditory moving window presentation

Twenty-eight aphasic patients with left hemisphere strokes and matched control subjects were tested on an auditory moving windows task in which successive phrases of a sentence were presented in response to subjects' self-paced button presses and subjects made timed judgments regarding the plausibility of each sentence. Pairs of sentences were presented that differed in syntactic complexity. Patients made more errors and/or took longer in making the plausibility judgments than controls, and were more affected than controls by the syntactic complexity of a sentence in these judgments. Normal subjects showed effects of syntactic structure in self-paced listening. On-line syntactic effects differed in patients as a function of their comprehension level. High-performing patients showed the same effects as normal control subjects; low performing patients did not show the same effects of syntactic structure. On-line syntactic effects also differed in patients as a function of their clinical diagnosis. Broca's aphasic patients' on-line performances suggested that they were not processing complex syntactic structures on-line, while fluent aphasics' performances suggested that their comprehension impairment occurred after on-line processing was accomplished. The results indicate that many aphasic patients retain their ability to process syntactic structure on-line, and that different groups of patients with syntactic comprehension disorders show different patterns of on-line syntactic processing.     

Age-related influences on lexical selection and orthographic encoding during homophone spelling

Two experiments investigated age differences in how semantic, syntactic, and orthographic factors influence the production of homophone spelling errors in sentence contexts. Younger and older adults typed auditorily presented sentences containing homophone targets (e.g., blew) that were categorized as having a regular spelling (EW) or an irregular spelling (UE). In Experiment 1, homophones were preceded by an unrelated word, a semantic prime that was congruent with the target's meaning in the sentence (e.g., wind), or a semantic prime incongruent with the target's meaning (e.g., sky) and instead related to the competitor homophone. Experiment 2 manipulated the target's part of speech, where target and competitor homophones shared or differed in part of speech. For both age groups, significant semantic priming occurred, where homophone errors decreased following congruent semantic primes and increased following incongruent primes. However, priming only occurred when homophones shared part of speech. Further, both age groups made more errors on homophones with an irregular than a regular spelling, and this regularity effect was smaller for older adults when homophones shared part of speech. Contrary to many spoken production tasks, older adults made fewer errors overall than younger adults. These findings demonstrate age preservation in lexical selection but age differences in orthographic encoding, resulting in older adults producing fewer errors because of reduced activation to competitor homophones. These findings also illustrate that syntactic factors, such as part of speech, can influence the spellings of individual words.     

Interaction with context during human sentence processing

Psychological theories of natural language processing have usually assumed that the sentence processor resolves local syntactic ambiguities by selecting a single analysis on the basis of structural criteria such as Frazier's (1978) “minimal attachment.” According to such theories, alternative analyses will only be attempted if the initial analysis subsequently proves inconsistent with the context. (See also Ferreira & Clifton, 1986; Ford, Bresnan, & Kaplan, 1982; Rayner, Carlson, & Frazier, 1983). An alternative hypothesis exists, however: If sentences are understood incrementally, more or less word-by-word (Marlsen-Wilson, 1973, 1975), then syntactic processing can in principle exploit the fact that interpretations are available, using them “interactively” to select among alternative syntactic analyses on the basis of their plausibility with respect to the context. The present paper considers possible architectures for such incremental and interactive sentence processors, and argues for an architecture     

Building Syntactic Structure in Speaking

We investigated how people produce simple and complex phrases in speaking using a newly developed immediate recall task. People read and tried to memorize a target sentence, then read a prime sentence, then did a distractor task involving the prime sentence. Despite the delay and activity between memory and recall, people could still recall the target sentence although the syntactic form of the recalled sentence was influenced by the syntactic form of the prime sentence. This result replicates the syntactic priming effect found with other experimental paradigms. Using this task, we tested how people used abstract syntactic plans to produce simple and complex noun phrases. We found syntactic priming both when targets and prime sentences matched in complexity and when they did not match, suggesting that simple and complex noun phrases are built by the same syntactic routines during speech production.     

Building syntactic structures in speaking: a bilingual exploration

In a series of three experiments we investigated syntactic priming using a sentence recall task. Participants read and memorized a target sentence for later recall. After reading a prime sentence and engaging in a distraction task, they were asked to produce the target sentence aloud. Earlier investigations have shown that this task is sensitive to a syntactic priming effect. That is, the syntactic form of the prime sentence sometimes influences the syntactic form of the recalled target. In this paper we report on a variation on this task, using Spanish-English bilingual participants. In the first two experiments we replicated the prepositional phrase priming effect using English target sentences and Spanish prime sentences. In the final experiment we investigated two additional syntactic forms, using Spanish target sentences and English prime sentences. Implications for models of syntax generation and bilingual speech production are discussed.     

Markers of Topical Discourse in Child‐Directed Speech

Although the language we encounter is typically embedded in rich discourse contexts, many existing models of processing focus largely on phenomena that occur sentence‐internally. Similarly, most work on children's language learning does not consider how information can accumulate as a discourse progresses. Research in pragmatics, however, points to ways in which each subsequent utterance provides new opportunities for listeners to infer speaker meaning. Such inferences allow the listener to build up a representation of the speakers' intended topic and more generally to identify relationships, structures, and messages that extend across multiple utterances. We address this issue by analyzing a video corpus of child–caregiver interactions. We use topic continuity as an index of discourse structure, examining how caregivers introduce and discuss objects across utterances. For the analysis, utterances are grouped into topical discourse sequences using three annotation strategies: raw annotations of speakers' referents, the output of a model that groups utterances based on those annotations, and the judgments of human coders. We analyze how the lexical, syntactic, and social properties of caregiver–child interaction change over the course of a sequence of topically related utterances. Our findings suggest that many cues used to signal topicality in adult discourse are also available in child‐directed speech.     

Effects of dentalization and devoicing of the /z/ phoneme on listeners' judgment of a speaker's sexual orientation

25 undergraduate listeners judged that a male speaker with normal speech, who dentalized (lisped) or devoiced the /z/ phoneme in the context of a sentence, sounded significantly more "gay" in terms of judged sexual orientation than did the same speaker producing /z/ without phonological processes. Speech-language pathologists should be aware of possible social consequences of speech production disorders.     

The processing of extraposed structures in English

In most languages, most of the syntactic dependency relations found in any given sentence are projective: the word-word dependencies in the sentence do not cross each other. Some syntactic dependency relations, however, are non-projective: some of their word-word dependencies cross each other. Non-projective dependencies are both rarer and more computationally complex than projective dependencies; hence, it is of natural interest to investigate whether there are any processing costs specific to non-projective dependencies, and whether factors known to influence processing of projective dependencies also affect non-projective dependency processing. We report three self-paced reading studies, together with corpus and sentence completion studies, investigating the comprehension difficulty associated with the non-projective dependencies created by the extraposition of relative clauses in English. We find that extraposition over either verbs or prepositional phrases creates comprehension difficulty, and that this difficulty is consistent with probabilistic syntactic expectations estimated from corpora. Furthermore, we find that manipulating the expectation that a given noun will have a postmodifying relative clause can modulate and even neutralize the difficulty associated with extraposition. Our experiments rule out accounts based purely on derivational complexity and/or dependency locality in terms of linear positioning. Our results demonstrate that comprehenders maintain probabilistic syntactic expectations that persist beyond projective-dependency structures, and suggest that it may be possible to explain observed patterns of comprehension difficulty associated with extraposition entirely through probabilistic expectations.     

句法结构和动词重复对汉语句子口语产生中句法启动效应的影响

研究中采用启动范式和图片描述任务, 利用句法选择比率和句子产生潜伏期的两个指标, 考察了启动句句法结构、动词是否相同和延时对汉语口语句子产生中句法启动效应的影响。结果发现在选择比率上, 句法结构产生的启动效应随延时变化保持稳定, 由启动句和目标句动词相同导致的启动效应的提高量(词汇增强效应)快速衰退, 首次为句子产生过程的句法选择阶段和计划阶段提供了证据。在句子产生潜伏期上, 仅发现启动句和目标句句法结构相同时缩短了句子产生潜伏期, 这可能是由于启动句句法结构激活水平的提高, 使得说话者在目标句产生的句法计划和选择阶段都更快引起的。选择阶段的加工影响句法选择比率, 而选择阶段和计划阶段共同影响了句子产生的潜伏期, 实验结果支持了两阶段竞争理论的观点。     

Semantic Factors in the Production of Number Agreement

This paper examines the role of semantic factors in the production of subject–verb number agreement. As an ostensibly grammatical process, number agreement provides an interesting case for examining the flow and interaction of semantic and syntactic information through the language-production system. Using a sentence-completion task, agreement errors can be elicited from subjects by presenting them with sentence fragments containing a complex noun-phrase, in which the nonhead noun is plural (e.g., The key to the cabinets...WERE missing.). Previous research has demonstrated that the probability of making an error can be affected by varying the properties of the nouns in the complex noun phrase. By investigating which variables do and do not affect error rates, constraints on the flow of information through the production system can be inferred. In three experiments, we investigated the possible effects of three different semantic manipulations of the nouns in the complex NP: animacy, semantic overlap, and plausibility of modification by the sentence predicate. We found that both animacy and semantic relatedness had reliable effects on error rates, indicating that the mechanism involved in implementing agreement cannot be blind to semantic information. However, the plausibility with which each noun could serve as the subject of the sentence predicate had no effect on error rates. Taken together, these results suggest that while semantic information is visible to the agreement mechanism, there are still constraints on when this information can affect the process. Specifically, it may be the case that only information contained within the complex NP is considered for the purposes of implementing agreement.     

Language acquisition following hemidecortication: Linguistic superiority of the left over the right hemisphere

The language development of three 9- and 10-year-old children possessing only a right or a left hemisphere was studied. Surgical removal of one brain half antedated the beginning of speech, so each child has acquired speech and language with only one hemisphere. Different configurations of language skill have developed in the two isolated hemispheres: phonemic and semantic abilities are similarly developed but syntactic competence has been asymmetrically acquired. In relation to the left, the right brain half is deficient in understanding auditory language, especially when meaning is conveyed by syntactic diversity; detecting and correcting errors of surface syntactic structure; repeating stylistically permuted sentences; producing tag questions which match the grammatical features of a heard statement; determining sentence implication; integrating semantic and syntactic information to replace missing pronouns; and performing judgments of word interrelationships in sentences. Language development in an isolated right hemisphere, even under seizure-free conditions, results in incomplete language acquisition.     

Language integration in bilingual sentence production

To what extent are processes used in sentence production integrated between the different languages of a bilingual and to what extent are they kept separate? We consider three models that differ in their assumptions about the degree of integration: De Bot's [De Bot, K. (1992). A bilingual production model: Levelt's Speaking model adapted. Applied Linguistics, 13, 1-24] bilingual blueprint of the speaker, Ullman's [Ullman, M. T. (2001). The neural basis of lexicon and grammar in first and second language: The declarative/procedural model. Bilingualism: Language and Cognition, 4, 105-122] declarative/procedural model of bilingualism, and Hartsuiker et al.'s [Hartsuiker, R. J., Pickering, M. J., & Veltkamp, E. (2004). Is syntax separate or shared between languages? Cross-linguistic syntactic priming in Spanish/English bilinguals. Psychological Science, 15, 409-414] integrated model. A review of the evidence from bilingual sentence production studies shows that Hartsuiker et al.'s predictions are supported, but argues against the other two models. We discuss some repercussions for bilingual language use.