Processing syntactically ambiguous sentences: Evidence from semantic priming

In this paper, we report the results of a study which investigates the processing of syntactically ambiguous sentences. We examined the processing of sentences in which an embedded clause is interpretable as either a complement clause or as a relative clause, as in, for example,The receptionist informed the doctor that the journalist had phoned about the events. The embedded clause in such sentences is typically analyzed as a complement to the verbinformed, rather than as a relative clause modifyingthe doctor. A number of models parsing predict this is the only analysis ever considered, while others predict that both interpretations are computed in parallel. Using a cross-model semantic priming technique, we probed for activation ofdoctor just after the embedded verb. Since only the relative clause analysis contains a connection betweenthe doctor and the embedded verb, we expected reactivation ofdoctor at that point only if the relative clause analysis were a viable option. Our results suggest that this is the case: Compared to priming in an ambiguous control sentence, a significant reactivation effect was obtained. These results are argued to support a model of parsing in which attachment of a clause may be delayed.The research reported here was supported in part by a grant from the McDonnell-Pew Cognitive Neurosciences Program, in part by grant DC-01409 (a research and training grant funded by the National Institute on Deafness and Other Communicative Disorders), and in part by a SERC postdoctoral fellowship B/90/ITF/293. We are grateful to Andrew Barss, Paul Bloom, Merrill Garrett, David Swinney, and an anonymous reviewer for comments on an earlier draft of this paper, and to Paul Gorrell for extremely useful discussion of parsing issues. We thank Suzanne Delaney for giving voice to our stimulus sentences. The results of this study were presented at the Fifth Annual CUNY Sentence Processing Conference, New York, March 1992.Order of authors is alphabetical.     

Syntactic parsing as revealed by brain responses: First-pass and second-pass parsing processes

This paper reviews a series of electrophysiological experiments on syntactic processing against the background of a psycholinguistic two-stage model of parsing. The data reveal two event-related brain potential components in correlation with syntactic processes: an early left anterior negativity and a late centro-parietal positivity. It is argued that these two components can be correlated with two separate stages of syntactic processing: the early left anterior negativity reflecting first-pass parsing processes and the late positivity reflecting second-pass parsing processes possibly including processes of reanalyses.     

Linguistic rhythm guides parsing decisions in written sentence comprehension

Various recent studies attest that reading involves creating an implicit prosodic representation of the written text which may systematically affect the resolution of syntactic ambiguities in sentence comprehension. Research up to now suggests that implicit prosody itself depends on a partial syntactic analysis of the text, raising the question of whether implicit prosody contributes to the parsing process, or whether it merely interprets the syntactic analysis. The present reading experiments examine the influence of stress-based linguistic rhythm on the resolution of local lexical-syntactic ambiguities in German. Both speech production data from unprepared oral reading and eye-tracking results from silent reading demonstrate that readers favor syntactic analyses that allow for a prosodic representation in which stressed and unstressed syllables alternate rhythmically. The findings contribute evidence confirming immediate and guiding effects of linguistic rhythm on the earliest stages of syntactic parsing in reading.     

Learning To Parse?

A strong claim about human sentence comprehension is that the processing mechanism is fully innate and applies differently to different languages only to the extent that their grammars differ. If so, there is hope for an explanatory project which attributes all parsing strategies to fundamental design characteristics of the parsing device. However, the whole explanatory program is in peril because of the discovery (Cuetos & Mitchell, 1988) that Late Closure is not universal: Spanish, and also Dutch and other languages, favor Early Closure (high attachment) where English favors Late Closure flow attachment). I argue that the universal parser can weather this storm. Exceptions to Late Closure in Spanish and other languages are observed only in one construction (a relative clause attaching into a complex noun phrase [NP]), which is borderline in English too. For other constructions, low attachment is preferred in all languages tested. I propose that what differentiates the complex NP construction is the heaviness of the attachee compared to that of the host configuration. A relative clause is a heavy attachee, and the lower NP alone is small as a host; the relative is therefore better balanced if the whole complex NP is its host. A wide range of facts is accounted for by the principle that a constituent likes to have a sister of its own size. Light constituents will tend to attach low, and heavy ones to attach high, since larger constituents are dominated by higher nodes. A preference for balanced weight is familiar from work on prosodic phrasing. I suggest, therefore, that prosodic processing occurs in parallel with syntactic processing (even in reading) and influences structural ambiguity resolution. Height of attachment ambiguities are resolved by the prosodically motivated same-size-sister constraint. The exceptional behavior of English may be due to its prosodic packaging of a relative pronoun with the adjacent noun, overriding the balance tendency. If this explanation is correct, it is possible that all cross-language variations in parsing preferences are due to cross-language variations in the prosodic component of the competence grammar.     

The influence of prosodic structure on the resolution of temporary syntactic closure ambiguities

This paper investigates the influence of prosodic structure on the process of sentence comprehension, with a specific focus on the relative contributions of syntactic and prosodic information to the resolution of temporary syntactic closure ambiguities. We argue that prosodic structure provides an initial memory representation for spoken sentences, and that information from this prosodic representation is available to inform syntactic parsing decisions. This view makes three predictions for the processing of temporary syntactic ambiguity: 1. When prosodic and syntactic boundaries coincide, syntactic processing should be facilitated. 2. When prosodic boundaries are placed at misleading points in syntactic structure, syntactic processing should show interference effects. 3. The processing difficulties that have been reliably demonstrated in reading experiments for syntactically complex sentences should disappear when those sentences are presented with a felicitous prosodic structure in listening experiments. These predictions were confirmed by series of experiments measuring end-of-sentence comprehension time and cross-modal naming time for sentences with temporary syntactic closure ambiguities. Sentences with coinciding or conflicting prosodic and syntactic boundaries were compared to a prosodic baseline condition.This research was supported in part by NIMH grant R29 MH51768 to the first author and NIMH grant T32 MH19729 to the Northeastern University Psychology Department.     

Prosody and parsing in coordination structures

The effect of prosodic boundary cues on the off-line disambiguation and on-line parsing of coordination structures was examined. It was found that relative clauses were attached to coordinated object noun phrases in preference to second conjuncts in sentences like: The lawyer greeted the powerful barrister and the wise judge who was/were walking to the courtroom. Naive speakers signalled the syntactic contrast between the two structures by a prosodic break between the conjuncts when the relative clause was attached to the second conjunct. Listeners were able to use this prosodic information in both off-line syntactic disambiguation and on-line syntactic parsing. The findings are compatible with a model in which prosody has a strong immediate effect on parsing. It is argued that the current experimental design has avoided confounds present in earlier studies on the on-line integration of prosodic and syntactic information.     

Syntactic structure assembly in human parsing: a computational model based on competitive inhibition and a lexicalist grammar

We present the design, implementation and simulation results of a psycholinguistic model of human syntactic processing that meets major empirical criteria. The parser operates in conjunction with a lexicalist grammar and is driven by syntactic information associated with heads of phrases. The dynamics of the model are based on competition by lateral inhibition ('competitive inhibition'). Input words activate lexical frames (i.e. elementary trees anchored to input words) in the mental lexicon, and a network of candidate 'unification links' is set up between frame nodes. These links represent tentative attachments that are graded rather than all-or-none. Candidate links that, due to grammatical or 'treehood' constraints, are incompatible, compete for inclusion in the final syntactic tree by sending each other inhibitory signals that reduce the competitor's attachment strength. The outcome of these local and simultaneous competitions is controlled by dynamic parameters, in particular by the Entry Activation and the Activation Decay rate of syntactic nodes, and by the Strength and Strength Build-up rate of Unification links. In case of a successful parse, a single syntactic tree is returned that covers the whole input string and consists of lexical frames connected by winning Unification links. Simulations are reported of a significant range of psycholinguistic parsing phenomena in both normal and aphasic speakers of English: (i) various effects of linguistic complexity (single versus double, center versus right-hand self-embeddings of relative clauses; the difference between relative clauses with subject and object extraction; the contrast between a complement clause embedded within a relative clause versus a relative clause embedded within a complement clause); (ii) effects of local and global ambiguity, and of word-class and syntactic ambiguity (including recency and length effects); (iii) certain difficulty-of-reanalysis effects (contrasts between local ambiguities that are easy to resolve versus ones that lead to serious garden-path effects); (iv) effects of agrammatism on parsing performance, in particular the performance of various groups of aphasic patients on several sentence types.     

The relationship between the frequency and the processing complexity of linguistic structure

In this paper the relative frequencies of the possible resolutions of ambiguities involving noun phrase attachment sites are compared to the results of off-line psycholinguistic measurements of syntactic complexity. A lack of correlation between the two is observed. It is therefore argued that the comprehension system is distinct from what is driving the frequencies in the corpora. A production heuristic separate from the comprehension system is proposed to account for the observed frequencies.We would like to thank Susanne Tunstall for providing linguistic judgments in the corpus analyses reported in Sec. 3.3. We also wish to thank the following people for helpful, comments and discussions on earlier presentations of this paper: the audience at the Eighth Human Sentence Processing Conference, in Tucson Arizona, Maria Babyonyshev, Tom Bever, Martin Corley, Danny Fox, Don Mitchell, Edson Miyamoto, Janet Nicol, Colin Phillips, Gregg Solomon, Michael Spivey-Knowlton, James Thomas, and Tom Wasow. All remaining errors are our own. The second author was supported by the Research Training Grant Language: Acquisition and Computation awarded by the National Science Foundation (U.S.) to the Massachusetts Institute of Technology (DIR 9113607), and by a fellowship from the Social Sciences and Humanities Research Council of Canada.     

Verb argument information used in a prodrop language: An experimental study in Japanese

Japanese is an SOV, head-final language with phonologically null arguments. Due to these characteristics, any (or all) of the arguments classified by a verb may be null. Consequently, verb argument information which Japanese readers/listeners receive overtly is sometimes incomplete. Furthermore, the arguments in a string may or may not belong to the same clause, and the disambiguating verb argument information does not become available until the end of a clause. These characteristics pose a question of whether or not verb argument information is a useful source of information for on-line syntactic processing in Japanese. An experiment using three types of complex NPs was designed specifically to examine whether Japanese readers use the argument structure information, or alternatively, they simply use minimal attachment in on-line processing of syntactically ambiguous structures. The reading times of three types of complex NPs suggest that native speakers of Japanese indeed utilize verb argument information, and they do so immediately. This finding supports the hypothesis that the Japanese language is processed in the similar way to English in that verb argument information functions as an important source of information in on-line processing. The difference between processing English and processing Japanese is that the former has a mechanism which keeps track of null arguments as well as overt ones.This research was supported in part by a grant to Mineharu Nakayama by the College of Humanitiesm The Ohio State University. A portion of this paper was presented at the CUNY Sentence Processing Conference (1992). I would like to thank Paul G. Gorrell, Julie E. Boland, Laurie A. Stowe, Mineharu Nakayama, and Jerry Packard for their useful comments and discussions.     

Grammatical determinants of ambiguous pronoun resolution

Crawley et al. (1990) argue for the primacy of a subject assignment strategy for pronoun assignment during reading, and against the notion of parallel function (Sheldon, 1974). However, most of their items deviated from parallel structure, and none included subject pronouns. In four experiments with subject and nonsubject pronouns, strong parallel function effects emerge when a potential antecedent has the same syntactic role as the pronoun and when the two clauses have the same attachment site and constituent structure. Attachment nonparallelism causes the greatest ambiguity, while the other types lead to more subject assignment overall, although there is always an overlaid parallel function effect. These observations support a model of pronoun assignment according to which potential antecedents are checked for morphological, syntactic and semantic feature matches with the pronoun, and priming/reactivation of syntactic structure across clauses facilitates parallel assignment.This research was supported in part by research grant 410-89-0395 from the Social Sciences and Humanities Research Council of Canada. A preliminary version of the paper was presented at the 1992 meetings of the Canadian Linguistic Association in Charlottetown, Prince Edward Island. I would like to thank Elizabeth Cowper, Keren Rice, Rosemary Stevenson, and an anonymous reviewer for their comments.     

Rare Constructions Are More Often Sentence‐Initial

Main clause phenomena (MCPs) are syntactic constructions that occur predominantly or exclusively in main clauses. I propose a processing explanation for MCPs. Sentence processing is easiest at the beginning of the sentence (requiring less search); this follows naturally from widely held assumptions about sentence processing. Because of this, a wider variety of constructions can be allowed at the beginning of the sentence without overwhelming the sentence‐processing mechanism. Unlike pragmatic and grammatical accounts of MCPs, the processing account predicts avoidance of MCPs in non‐initial main clauses (non‐initial coordinate clauses and premodified clauses). A corpus study supports these predictions, but it is somewhat inconclusive. A further corpus study examines another type of syntactic construction, premodifying adjunct phrases (“openers”); the prediction here is that less common types of opener will be especially avoided in non‐initial contexts. The prediction is confirmed, supporting the processing view of rare constructions.     

Competition and recency in a hybrid network model of syntactic disambiguation

The competitive attachment model of human parsing is a hybrid connectionist architecture consisting of a distributed feature passing method for establishing syntactic relations within the network, and a numeric competition mechanism for resolving ambiguities, which applies to all syntactic relations. Because the approach employs a uniform mechanism for establishing syntactic relations, and a single competition mechanism for disambiguation, the model can capture general behaviors of the human parser that hold across a range of syntactic constructions. In particular, attachment and binding relations are similarly processed and are therefore subject to the very same influences of disambuguation and processing over time. An important influence on the competitive disambiguation process is distance within the network. Decay of numeric activation, along with distributed feature passing through the network structure, has an unavoidable effect on the outcome of attachment and binding competitions. Inherent properties of the model thus lead to a principled explanation of recency effects in the human parsing of both attachment and filler/gap ambiguities.     

Subject-Object Ambiguities in German Embedded Clauses: An Across-the-Board Comparison

This paper examines the processing of embedded clauses in German which are ambiguous between a subject-before-object and an object-before-subject order. In an experiment using a speeded grammaticality judgment task, four types of locally ambiguous clauses were compared: (i) sentences involving movement of a definite noun phrase (NP), (ii) sentences involving pronoun movement, (iii) relative clauses, and (iv) embedded questions. We found that readers were consistently garden-pathed in the object-before-subject condition, regardless of sentence type. Furthermore, there were considerable differences with respect to garden-path strength. The garden-path effect was strongest for sentences involving scrambling. In addition, sentences involving pronoun movement induced more processing difficulty than embedded questions and relative clauses. We argue that our findings can be best explained within a serial processing model that acknowledges both syntactic and nonsyntactic influences on reanalysis and that can account for graded effects of garden-path strength.     

Gender Processing in Speech Production: Evidence from German Speech Errors

The present study investigates the relationship between gender processing and the properties of speech errors. Studying German noun substitution errors it was found that intended and intruded nouns were more often of the same grammatical gender than one would expect by chance (identical gender effect). In the present study, German slips of the tongue were investigated on the assumption that the occurrence of the identical gender effect depends on the processing level, where the error arises. The syntactic context preceding errors of nonidentical gender was also explored. In the German language, interactions of nonidentical gender nouns often result in agreement violations. It can be shown that gender congruency between nouns and preceding articles also depends on the processing level at which the noun error occurs. The results are consistent with two-stage models of lexical retrieval and speech production, according to which the syntactic information of a noun is only represented during the first stage of lexical access.     

Studying the grammatical aspects of word recognition: lexical priming,parsing, and syntactic ambiguity resolution

Two experiments are reported examining the relationship between lexical and syntactic processing during language comprehension, combining techniques common to the on-line study of syntactic ambiguity resolution with priming techniques common to the study of lexical processing. By manipulating grammatical properties of lexical primes, we explore how lexically based knowledge is activated and guides combinatory sentence processing. Particularly, we find that nouns (like verbs, see Trueswell & Kim, 1998) can activate detailed lexically specific syntactic information and that these representations guide the resolution of relevant syntactic ambiguities pertaining to verb argument structure. These findings suggest that certain principles of knowledge representation common to theories of lexical knowledge—such as overlapping and distributed representations—also characterize grammatical knowledge. Additionally, observations from an auditory comprehension study suggest similar conclusions about the lexical nature of parsing in spoken language comprehension. They also suggest that thematic role and syntactic preferences are activated during word recognition and that both influence combinatory processing.     

Ambiguity in the brain: what brain imaging reveals about the processing of syntactically ambiguous sentences

Two fMRI studies investigated the time course and amplitude of brain activity in language-related areas during the processing of syntactically ambiguous sentences. In Experiment 1, higher levels of activation were found during the reading of unpreferred syntactic structures as well as more complex structures. In Experiments 2A and 2B higher levels of brain activation were found for ambiguous sentences compared with unambiguous sentences matched for syntactic complexity, even when the ambiguities were resolved in favor of the preferred syntactic construction (despite the absence of this difference in previous reading time results). Although results can be reconciled with either serial or parallel models of sentence parsing, they arguably fit better into the parallel framework. Serial models can admittedly be made consistent but only by including a parallel component. The fMRI data indicate the involvement of a parallel component in syntactic parsing that might be either a selection mechanism or a construction of multiple parses.     

Constraint satisfaction as a theory of sentence processing

Various problems with the constraint satisfaction model are discussed. It is argued that the empirical evidence presented in support of the model does not concern predictions of the model that diverge from those of depth-first (one analysis at a time) models. Several methodological problems are also noted. As a theory of sentence processing, the model is inadequate. It fails to account for the assignment of local structure, global structure, structure involving discontinuous dependencies, long-distance dependencies, and adjunct phrases. It makes incorrect predictions about the timing of syntactic analysis. Further, because syntactic structure is available only through activation of syntactic projections stored in the lexical entry of words, the model leaves entirely unexplained the myriad psycholinguistic findings demonstrating independence of lexical and syntactic structure (in Event Related Potential studies, code-switching, pure syntactic priming, etc). Finally, the model is not restrictive or explanatory, providing an account that largely consists ofpost hoc correlations between frequency counts or subjects ratings of sentences and processing time data for the same sentences.The present paper is an excerpt from Sentence (Re-)Analysis presented at the 1994 CUNY Conference. It was supported by NIH Grant HD18708 to Clifton and Frazier and DBS9121375 to Rayner and Frazier. I am very grateful to Chuck Clifton and to an anonymous reviewer for comments on an earlier draft of this paper.     

Prosodic facilitation in the resolution of syntactic ambiguities in subjects with left and right hemisphere damage

The purpose of this study was to determine if prosody facilitates the comprehension of sentences containing temporary syntactic ambiguities in control, and left (LHD) and right hemisphere damaged (RHD) subjects. To test for effects of prosodic facilitation, sentences were created where prosodic boundaries coincided with (cooperating), were absent (baseline), or conflicted (conflicting) with syntactic boundaries in three response times (RTs) experiments. Despite differences in overall RTs and response accuracy for each group, all three groups responded faster and more accurately to sentences in the cooperating than in the baseline and conflicting conditions across experiments, indicating that prosody facilitates syntactic parsing in brain-damaged subjects just as it does with normal control subjects. Results are discussed in relation to psycholinguistic theories of syntactic parsing and neurolinguistic theories of hemispheric specialization in processing the acoustic properties of prosodic structures.     

Recycling prosodic boundaries

The present study investigates the role of prosodic structure in selecting a syntactic analysis at different stages of parsing in silent reading of Japanese relative clauses. Experiments 1 and 2 (sentence-completion questionnaires) revealed an effect of the length of the sentence-initial constituent on the resolution of a clause boundary ambiguity in Japanese. Experiment 3 (fragment-reading) showed that this length manipulation is also reflected in prosodic phrasing in speech. Its influence on ambiguity resolution is attributed to recycling of prosodic boundaries established during the first-pass parse. This explanation is based on the implicit prosody proposals of Bader (1998) and Fodor (1998). Experiment 4 (self-paced reading) demonstrated the immediacy of the influence on ambiguity resolution on-line. Experiment 5 (self-paced reading) found support for the additional prediction that when no boundary is available to be recycled, processing the relative clause construction is more difficult.