Lyn Frazier

Making and Correcting Errors during Sentence Comprehension: Eye Movements in the Analysis of Structurally Ambiguous Sentences

Eye movements were recorded as subjects read sentences containing temporary structural ambiguities. In accord with the garden-path theory of sentence comprehension, shorter reading times were found for sentences conforming to certain independently motivated parsing strategies (late closure and minimal attachment) than for comparable sentences which violate these strategies. Further, longer fixation durations were associated with the very first fixation in the region of the sentence which disambiguated the sentence, suggesting that the human sentence-parsing mechanism operates in a rather systematic fashion, immediately computing the structural consequences of fixated material for the analysis of preceding material. The pattern of regressive eye movements did not conform to the view that the parsing mechanism automatically returns to the beginning of the sentence to revise an incorrect analysis of linguistic material nor did it support the view that the parsing mechanism systematically backtracks through the sentence until the source of the erroneous analysis is located. Rather, the pattern of regressions indicated that the parsing mechanism typically engages in selective reanalysis, exploiting whatever information it has available about the type of error it has committed to guide its reanalysis attempts. Finally, it is emphasized that an understanding of the parsers revision procedures is essential to an explanation of why certain linguistic structures cannot be successfully parsed by humans.

The sausage machine: A new two-stage parsing model

Abstract It is proposed that the human sentence parsing device assigns phrase structure to word strings in two steps. The first stage parser assigns lexical and phrasal nodes to substrings of roughly six words. The second stage parser then adds higher nodes to link these phrasal packages together into a complete phrase marker. This model of the parser is compared with ATN models, and with the two-stage models of Kimball (1973) and Fodor, Bever and Garrett (1974). Our assumption that the units which are shunted from the first stage to the second stage are defined by their length, rather than by their syntactic type, explains the effects of constituent length on perceptual complexity in center embedded sentences and in sentences of the kind that fall under Kimballs principle of Right Association. The particular division of labor between the two parsing units allows us to explain, without appeal to any ad hoc parsing strategies, why the parser makes certain ‘shortsighted’ errors even though, in general, it is able to make intelligent use of all the information that is available to it.

The interaction of syntax and semantics during sentence processing: eye movements in the analysis of semantically biased sentences

Two experiments explored the effects of semantic and pragmatic information on the syntactic analysis of ambiguous sentences. In both experiments, eye movements were recorded as subjects read structurally ambiguous sentences. The first experiment showed that the relative plausibility of two possible real world events does not influence the language processors choice of an initial syntactic analysis of an ambiguous string: clear garden-path effects were observed in both relatively plausible and relatively implausible reduced relative clauses, while no garden-path effect was observed for simple active clauses. The second experiment showed that semantic and pragmatic considerations do govern the ultimate analysis of an ambiguous sentence but that structural preferences, and their agreement with the final analysis, governed reading times. Reading times were significantly shorter for sentences where the pragmatically more plausible analysis of the sentence coincided with the analysis preferred on purely structural grounds than for sentences where the pragmatically preferred analysis conflicted with the structurally preferred analysis. Taken together, the results of the two experiments argue for the existence of distinct processors in the human sentence comprehension mechanism.

Filler driven parsing: A study of gap filling in dutch

Simple Dutch declaratives and questions were presented in a grammaticality judgment task to assess the validity of the “active filler strategy” (AFS). The AFS predicts that moved constituents, such as the initial constituent of a Dutch sentence, will be assigned to the leftmost possible gap. This results in better performance on subject-initial than on object-initial sentences. These predictions were confirmed, supporting a “filler” driven account of gap filling where gaps may be postulated before or without identifying a missing constituent in the input string. The results argue against the existence of a bottom-up parser in which the presence of a gap is detected only when lexically present local phrases have been parsed.

Comprehending Sentences with Long-Distance Dependencies

In the process of comprehending a sentence, a reader or listener identifies its grammatical constituents and their relationships. The resulting grammatical analysis is eventually combined with lexical information and world knowledge to arrive at the message conveyed by the sentence. We propose that one or more distinct components of the human cognitive system are responsible for identifying the grammatical characteristics of a sentence (see Berwick and Weinberg, 1983, 1984; J. A. Fodor, 1983; J. D. Fodor, 1979; Forster, 1979, for similar views).

Filling gaps: Decision principles and structure in sentence comprehension

Abstract The correct grammatical characterization of sentences containing filler-gap dependencies is a topic of considerable theoretical interest in linguistics. In some grammatical frameworks, these dependencies are representef in terms of conditions on the permissible indexing of structures (or alternative structure evaluation conditions) which a representation must adhere to in order to be well-formed. In other frameworks, constraints on permissible filler-gap dependencies are simply inherent in the set of phrase structure rules contained in the grammar of a language. The processing of sentences with multiple (potential) filler-gap dependencies was investigated in two experiments. The first experiment provided evidence for three claims. First, the human sentence processor abides by a strategy of assigning the most recent potential filler to a gap. Hence, ‘recent filler’ sentences where this assignment proves to be correct takes less time to comprehend than ‘distant filler’ sentences where this decision turns out to be incorrect. Second, the recent filler strategy is itself just a special case of a more general strategy of assigning the most salient potential filler to a gap. Third, unambiguous sentences in which a filler-gap assignment is disambiguated by ‘control’ information specified by individual verbs gives rise to the same recent filler errors as ambiguous sentences. This suggests that tentative filler-gap assignments are made by the processor before all of the relevant constraints on permissible filler-gap dependencies are consulted by the processor. The second experiment tested an alternative hypothesis that the more complex ‘distant filler’ sentences took longer to comprehend in the first experiment only because these sentences often contained verbs which license two adjacent gaps. The experiment showed that there was a significant recent filler effect in sentences that did not contain adjacent gaps and that this effect did not interact with verb class. The finding that the processor delays use of verb-control information is extremely surprising. It may be explained by the fact that this information is only relevant to one type of gap (‘equi-gaps’) and what type of gap the processor is dealing with often can not be determined unambiguously at the time when it initially encounters a gap in its left-to-right processing of a sentence. If our interpretation of these findings is correct, they argue for a considerable amount of structure in the sentence comprehension system. Further, they favor a view of sentence processing in which processing operations involving constraints on the permissible indexing (or evaluation) of structures lag behind the processors structure building operations. Hence, the results favor those grammatical theories which preserve this distinction over grammatical theories which provide a uniform characterization of all syntactic well-formedness conditions.

Language processing and language acquisition

The Grammatical Nature of the Acquisition Sequence: Adjoin-a and the Formation of Relative Clauses.- The Status of Grammatical Default Systems: Comments on Lebeaux.- On Unparsable Input in Language Acquisition.- Logical and Psychological Constraints on the Acquisition of Syntax.- How to Make Parameters Work: Comments on Valian.- On Parameter Setting and Parsing: Predictions for Cross-Linguistic Differences in Adult and Child Processing.- Comments on Mazuka and Lusts paper.- Parameters and Parameter-Setting in a Phrase Structure Grammar.- The Acquisition of Long-distance Rules.- Child Grammars - Radically Different, or More of the Same?: Comments on de Villiers, Roeper and Vainikka.- The Processing and Acquisition of Control Structures by Young Children.- Intuitions, Category and Structure: Comments on McDaniel and Cairns.- Visiting Relatives in Italy.- Obeying the Binding Theory.- Knowledge Integration in Processing and Acquisition: Comments on Grimshaw and Rosen.- List of First Authors.

Resolution of syntactic category ambiguities: Eye movements in parsing lexically ambiguous sentences☆

Three experiments explored the effects of the interaction of lexical and syntactic processes during language comprehension. Subjects read sentences containing lexical items that resulted in syntactic category ambiguities (e.g., desert trains, where desert can be a noun or an adjective and trains can be a verb or a noun). During reading, eye movements were monitored as a reflection of on-line parsing activities. The experiments tested alternative hypotheses about how the processor resolves syntactic category ambiguities. All experiments supported a delay strategy in which the processor delays assigning an analysis to a categorially ambiguous string until it receives disambiguating information dictating the correct analysis of the string. The implications of the results for a general theory of sentence comprehension are discussed.

Lexical Expectations in Sentence Comprehension.

Verbs, as a syntactic category, accept noun phrase complements (direct objects). Some specific verbs require such complements, others prohibit them, while still others differ in the likelihood that they will be used with direct objects. Two experiments demonstrated that readers use specific lexical information in comprehending sentences to anticipate and prepare for the appearance of lexical noun phrases and to postulate “gaps” that are associated with “fillers.” One experiment used an on-line secondary task to demonstrate the effects of violating a lexically based syntactic expectation, while a second experiment used an end-of-the-sentence grammaticality judgment task to demonstrate a similar effect on gap filling. The experiments demonstrated that lexically based expectations involve the use of information about appropriate syntactic or thematic categories, and provided evidence for the use of pragmatic information as well. The question of how lexical information interacts with other types of information in sentence processing was raised.

Parsing temporarily ambiguous complements

Holmes, Kennedy and Murray (1987) recently claimed that the empirical support for the Minimal Attachment Strategy of sentence parsing had been weakened by results they reported. They found that reading time for an ambiguous string of words did not decrease when it was preceded by an overt complementizer, which should have disambiguated it. Thus, they suggested that results that we (Frazier and Rayner, 1982) earlier attributed to Minimal Attachment were not due to “garden-path” effects, but rather reflected the extra complexity caused by having to process two sets of clausal relations instead of just one. In the present experiment, we replicated their experiment using eye movement data rather than the subject-paced reading task they used. We found that readers processed Nonminimal Attachment sentences with overt complementizers considerably faster than those without a complementizer. Our results showed that the complexity of Nonminimal Attachment sentences cannot be attributed to their clausal status per se. Differences between the tasks that might contribute to the different pattern of results across the experiments are discussed.