Garden-path Sentences Research Articles

On-Line Reflexive Resolution in Japanese Causative Garden Path Sentences

This study explored the time course in which lexical information in causative verbs influences the identification of the antecedent of a Japanese reflexive pronoun, jibun. Since verb information specifies that the reflexive is bound to the indirect object despite its being ordinarily associated with the subject in Japanese sentences, this requires parsers to revise the syntactic representation they had previously formed. 42 students were required to identify quickly the antecedent when given a marker which was placed immediately after the reflexive, immediately at the end of the sentence following a verb, or 4 sec. after the end of the sentence. Correct identification was almost nil immediately after the reflexive, increasing to 20.3% correct immediately after the end of the sentence and 52.4% correct 4 sec. after the end of the sentence. This finding indicates that the verb-control information is not used immediately when parsing sentences in which the verb occurs at the end of the sentence. Thus, the products of moment-to-moment initial syntactic computation are not immediately mapped onto the final semantic interpretations.

Processing empty categories: A parallel approach

In many current theories of human sentence processing, the mechanisms and principles proposed to account for data related to processing (1) long-distance dependencies and (2) structural ambiguities, such as garden-path sentences, are independent, despite the fact that deciding whether or not to posit a gap is just a special case of ambiguity resolution. In this paper we demonstrate how the parallel parsing theory proposed by Gibson (1991, in press)—which was developed to account for nongap ambiguity resolution data—also explains a number of gap-positing facts, without additional strategies. In particular, we show how this theory correctly explains filled-object-gap effects and the lack of filled-subject-gap effects in English, as well as certain gap-processing effects in Dutch.

Regressive eye movements and sentence parsing: on the use of regression-contingent analyses.

Altmann, Garnham, and Dennis (1992) recently advocated the use of a regression-contingent analysis in eye movement studies dealing with parsing of sentences containing temporary structural ambiguities. Using experimental sentences preceded by referentially supportive contexts, they demonstrated a different pattern of results when regressive eye movements were eliminated from the data set than when regressions were included. The regression-contingent analysis that they used involves an implicit assumption that regressive eye movements are a necessary consequence of subjects' being garden pathed. We report some data which demonstrate that garden path effects are sometimes even stronger in the absence of regressions and, thus, argue that readers can be garden pathed without making regressions.

Brain potentials elicited by garden-path sentences: Evidence of the application of verb information during parsing.

Event-related potentials were recorded from 13 scalp locations while participants read sentences containing a syntactic ambiguity. In Experiment 1, syntactically disambiguating words that were inconsistent with the favored syntactic analysis elicited a positive-going brain potential (P600). Experiment 2 examined whether syntactic ambiguities are resolved by application of a phrase-structure-based minimal attachment principle or by word-specific subcategorization information. P600 amplitude was a function of subcategorization biases rather than syntactic complexity. These findings indicate that such biases exist and can influence the parser under certain conditions and that P600 amplitude is a function of the perceived syntactic well-formedness of the sentence.

Unimmediate construction of syntactic structure for garden path sentences in Japanese

While self-paced reading and eye-fixation have been used as on-line measures of sentence processing, they provide only indirect evidence as to the exact syntactic state during sentence processing. Two experiments were designed to fill this gap, where garden path sentences in Japanese were used and the marker demanding reanalysis of a previously established syntactic structure appeared at the end of these sentences. Forty-eight students were given a probe verb from an embedded clause and required to quickly identify the subject whose syntactic role could be confirmed only when they later encountered the marker. Findings showed that correct identification was below 50% immediately after the marker (0-sec delay) and increased gradually, reaching 82% at the delay of 4 sec. However, perfect identification was not obtained even for the latter delay condition. The findings indicate that parsers cannot immediately reconstruct the previous syntactic structure but need a certain period of time to reach a final decision. They also suggest that a distinction should be clarified between the moment-to-moment syntactic computation being made and the syntatic structure being constructed during sentence processing.

Parameters in the theory of sentence processing: Minimal Commitment theory goes east

This paper introduces the Minimal Commitment theory. This theory is a subspecies of deterministic parsers. The theory builds representations where immediate dominance and precedence relations are unspecified. Justification that this approach is psycholinguistically justified comes from its ability to provide a cross-linguistically valid theory of Garden path sentences in English and Japanese.

Parallel parsing: Evidence from reactivation in garden-path sentences

An all-visual, on-line, lexical priming technique was used to investigate whether the human sentence processor computes syntactic representations serially or in parallel. Structurally ambiguous garden-path sentences such as the following were studied: “The psychologist told the wife that the man bumpedthat her car was stolen.” Despite the strong preference for the sentential-complement reading of the ambiguous region (italicized), a reactivation effect for the head of the relative clause (wife) was observed immediately following the presentation of the embedded verb (bumped), suggesting that the relative clause analysis is also computed. This finding is taken as evidence for parallel parsing since both possible analyses were shown to have a processing reflection simultaneously: Computation of the sentential-complement analysis was demonstrated by the fact that readers garden-path when that analysis turns out to be incorrect, and computation of the relative clause analysis was demonstrated by the reactivation effect.

Reading processes during syntactic analysis and reanalysis.

Abstract We conducted two experiments to examine the processing of garden path sentences such as While the boy scratched the dog yawned loudly. We varied the ambiguous phrase of such sentences so that it was either short (the dog), long due to the inclusion of a relative clause (the dog that was hairy), or long due to the inclusion of prenominal adjectives (the big and hairy dog). The subjects' task was to read each sentence while their eye movements were monitored and then judge whether the sentence was grammatical. We found that early closure sentences were judged grammatical less often than were late closure sentences, and sentences with a long ambiguous phrase, particularly one made long through the insertion of a relative clause, were judged grammatical less often than were those with a short ambiguous phrase. The reading time data, however, showed a curious pattern. Reading times on the disambiguating word of the sentences (yawned in the above example) were longer for early closure than for late closure sentences, as expected, but reading times were shorter for sentences with a long ambiguous phrase. We argue that the latter finding reflects subjects' tendency to read more quickly as they proceed through a sentence. We suggest that researchers interpret any studies showing differences in the processing of syntactically easy and difficult constructions with extreme caution, unless the words being compared in these studies are preceded by the same number of words in the different experimental conditions.Resume Nous avons effectue deux experiences sur le traitement des phrases trompe - l'oeil telles While the boy scratched the dog yawned loudly. Nous avons varie la formulation ambigue de ces phrases pour qu'elle soit breve (the dog), longue en raison de l'insertion d'une proposition relative (the dog that was hairy) ou longue par l'insertion d'adjectifs anteposes (the big and hairy dog). Les sujets devaient lire chacune des phrases pendant que leurs mouvements oculaires etaient enregistres, puis ils devaient juger de leur grammaticalite. Nous avons constate que les phrases a clo@ture ha@tive etaient jugees moins souvent grammaticales que les phrases reposant sur le principe de clo@ture tardive. Nous avons egalement constate que les phrases ayant une formulation longue et ambigue, en particulier lorsqu'une proposition relative y etait inseree, etaient jugees grammaticales moins souvent que celles dont la formulation etait courte et ambigue. Un patron etrange ressort des temps de lecture. Comme prevu, les temps de lecture enregistres pour le terme desambiguisant des phrases (yawned dans l'exemple cidessus) etaient plus longs dans le cas des phrases a clo@ture ha@tive; en revanche, ils etaient plus courts pour les phrases longues et ambigues. Ce resultat montre, a notre avis, que la vitesse de lecture des sujets tend a s'accroi@tre d'un bout a l'autre de la phrase. Nous invitons les chercheurs a interpreter avec force circonspection toutes les etudes revelant des differences dans le traitement des constructions syntaxiquement complexes et simples, a moins que les mots qui y sont compares soient precedes du me@me nombre de mots dans les differentes conditions experimentales.Reading involves a number of inter - related processes. Readers must access the lexical reprentations of words and combine them so as to reflect the overall meaning of the sentence. For example, consider the ambiguous sentence Mary saw the boy with binoculars. The phrase with binoculars could function either as the modifier of the verb phrase that includes saw, indicating that Mary was the one in possession of the binoculars, or as the modifier of boy, in which case the boy is the one with the binoculars. These different syntactic attachments of the phrase with binoculars have different semantic consequences, resulting in different truth conditions for the two possible semantic interpretations of the ambiguous sentence. An important process during reading, then, is to create phrases and assign them to their appropriate grammatical roles, a process we will refer to as parsing. …

Avoiding the garden path: Eye movements in context

Pragmatic factors, such as referential context, influence the decisions of the syntactic processor. At issue, however, has been whether such effects take place in the first or second pass analysis of the sentence. It has been suggested that eye movement studies are the only appropriate means for deciding between first and second pass effects. In this paper, we report two experiments using ambiguous relative/complement sentences and unambiguous controls. In Experiment 1 we show that referential context eliminates all the first pass reading time differences that are indicative of a garden path to the relative continuation in the null context. We observe, however, that the context does not eliminate the increased proportion of regressions from that disambiguating continuation. We therefore introduce a regression-contingent analysis of the first pass reading times and show that this new measure provides an important tool for aiding in the interpretation of the apparently conflicting data. Experiment 2 investigated whether the results of Experiment 1 were an artifact of the kinds of questions about the contexts that were asked in order to encourage subjects to attend to the contexts. The results demonstrated that the use of explicity referential questions had little effect. There was some small evidence for a garden path effect in this second experiment, but the regression-contingent measure enabled us to locate all garden path effects in only a small proportion of trials and to conclude that context does influence the initial decisions of the syntactic processor.

Parsing in discourse: Context effects and their limits

Can discourse context override syntactic parsing preferences? This question is important for models that assume an autonomous syntax. Three experiments addressed this question by embedding two different types of garden path sentences in biasing contexts. For sentences with postnominal prepositional phrases (PP) such as The boy hit the girl with the boomerang/smile we found context override of the preference to attach the PP high to the verb. This result was obtained both in self-paced reading (Experiments 1 & 3) and in an eye-movement study (Experiment 2). By contrast, sentences with reduced relatives such as The coffee spilled on the rug was difficult for her to conceal produced longer reading times than control sentences even in context (Experiment 3). The selectivity of context override requires a model of parsing sensitive to differences among syntactic structures. One such model (The Cooperative Language Processor) is discussed in light of these results.

Finger movements in braille reading: The effect of local ambiguity

The hand movements of blind readers were recorded while they read sentences containing temporary local ambiguities, designed to induce garden path effects upon disambiguation. A syntactically ambiguous beginning (“The spy saw the cop with…”) was compatible with attachment of the following prepositional phrase (PP) either to the verb (“saw”) or the the noun (“cop”), but the content of the PP was inconsistent with verb phrase attachment in NPA sentences (“with a revolver”), not in VPA sentences (“with binoculars”). Mean total scanning time per character increased in the PP region of NPA sentences, not of VPA sentences. The effect, however, depended on both the content of the PP and its predictability, and was significant by items only when both factors favoured VPA. The slowing down over the PP was entirely due to increased incidence of regressions to that region, and first-pass scanning speed was unaffected by sentence type. In studies with visual reading, disambiguation has been found to have a large effect on first-pass scanning. The results confirm previous suggestions that regressions are the main repair device available to the braille reader to cope with local increases in processing load.

Recovery from misanalyses of garden-path sentences

In five experiments we examined the way in which readers reanalyze garden-path sentences, using grammaticality judgments as the dependent measure. The stimuli were twoclause sentences containing an ambiguous noun phrase which could function as either the object of the first clause or the subject of the second. Prior research has shown that the former analysis is generally preferred. In the first two experiments, we varied the number of words in the ambiguous phrase and found that reanalysis of garden-path sentences was more difficult with a longer ambiguous phrase. The third experiment established that this effect of phrase length is not attributable to the greater syntactic complexity of longer phrases. The fourth and fifth experiments demonstrated that the effect of phrase length is attributable to increasing the distance from the head of the ambiguous phrase to the disambiguating word of the graden-path sentence: Ambiguous phrases made long by the addition of prenominal adjectives were easy for the parser to reanalyze, but phrases made long by the addition of postnominal modifying prepositional phrases (Experiment 3) or relative clauses (Experiments 4 and 5) were hard for the parser to reanalyze. From these results, we argue that sentence comprehension requires the creation of phrase structure and the assignment of thematic roles to phrases, with the assignment taking place at the phrasal head. Reanalysis is affected by the ease with which thematic roles can be reassigned to misanalyzed phrases.

Context and distance-to-disambiguation effects in ambiguity resolution: Evidence from grammaticality judgments of garden path sentences

This study investigated whether single or multiple structures are constructed at the level of syntactic analysis for ambiguous strings. To this end, three experiments were conducted to determine the effects of various types of snytactic and nonsyntactic information on grammaticality judgments for garden path sentences. The results of Experiment 1, using a speeded grammaticality judgment task, were that context and length of the ambiguous region influenced the probability of calling a garden path sentence grammatical. In Experiment 2, even when given unlimited time to make grammaticality judgments, 40% of all long garden path sentences that required a less preferred analysis were called ungrammatical. Experiment 3 extended the length effects found in Experiments 1 and 2 to some additional types of garden path sentences. The results support a parsing model that is influenced by context and that constructs only a single interpretation of an ambiguous string at a time.

The role of specific information about the verb in parsing sentences with local structural ambiguity

Three experiments were conducted to determine whether individual words can influence the course of parsing in an on-line subject-paced reading task. The materials were sentences with local structural ambiguities. The results of Experiment I showed that when one verb in a sentence was replaced by another, this produced a garden-path effect. This suggests that the identity of the verb may have influenced the choice of structure for the ambiguous phrase. Experiments II and III were conducted to rule out alternative explanations based on possible segmentation artifacts and on changes in the overall plausibility of the sentences. The evidence supported the initial interpretation in terms of a general and robust parsing bias. The findings throw light on the kinds of information that the parser would have to act on when processing the verb.

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 processor's 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.

Predicting garden path sentences

This work is an investigation into part of the human sentence parsing mechanism (HSPM). The major test of the psychological validity of any model of the HSPM is that it fail on precisely those sentences that humans find to be garden paths. It is hypothesized that the HSPM consists of at least two processes. We call the first process the syntactic processor, and the second will be known as the semantic processor. It is hypothesized that the syntactic processor is unconscious, deterministic and fast, but limited. While most ambiguities are resolved on the basis of syntactic information, when the syntactic processor can no longer guarantee a correct analysis, semantic information is used to help resolve the ambiguity. This model leads to a better prediction and explanation of which sentences will cause people to garden path.

A Theory of Syntactic Recognition for Natural Language

A Theory of Syntactic Recognition for Natural Language makes use of the hypothesis that the syntax of any natural language can be parsed by a mechanism which operates deterministically in that it does not simulate a nondeterministic machine. Basing his research strictly on English, Marcus sets forth some principles of processing that interact to offer explanations for some fundamental properties of language. He shows that language should have a certain design in order to be efficiently processed by the system he has constructed; specifically, rules should have certain locality properties, left-right asymmetries, and hierarchic structures that enter into rule types in special ways.Included in this volume are sections on the Determinism Hypothesis, Historical Perspective, the Grammar Interpreter, Structure of Grammar, Capturing Linguistic Generalizations, The Grammar Interpreter and Chomsky's Constraints, Parsing Relative Clauses, Parsing Noun Phrases, Differential Diagnosis and Garden Path Sentences, and the Necessity of Some Semantic/Syntactic Interactions.