Theories Of Sentence Processing Research Articles

Philippine Psycholinguistics

Over the last decade, there has been a slow but steady accumulation of psycholinguistic research focusing on typologically diverse languages. In this review, we provide an overview of the psycholinguistic research on Philippine languages at the sentence level. We first discuss the grammatical features of these languages that figure prominently in existing research. We identify four linguistic domains that have received attention from language researchers and summarize the empirical terrain. We advance two claims that emerge across these different domains: ( a) The agent-first pressure plays a central role in many of the findings, and ( b) the generalization that the patient argument is the syntactically privileged argument cannot be reduced to frequency, but instead is an emergent phenomenon caused by the alignment of competing pressures toward an optimal candidate. We connect these language-specific claims to language-general theories of sentence processing.

The Plausibility of Sampling as an Algorithmic Theory of Sentence Processing.

Words that are more surprising given context take longer to process. However, no incremental parsing algorithm has been shown to directly predict this phenomenon. In this work, we focus on a class of algorithms whose runtime does naturally scale in surprisal-those that involve repeatedly sampling from the prior. Our first contribution is to show that simple examples of such algorithms predict runtime to increase superlinearly with surprisal, and also predict variance in runtime to increase. These two predictions stand in contrast with literature on surprisal theory (Hale, 2001; Levy, 2008a) which assumes that the expected processing cost increases linearly with surprisal, and makes no prediction about variance. In the second part of this paper, we conduct an empirical study of the relationship between surprisal and reading time, using a collection of modern language models to estimate surprisal. We find that with better language models, reading time increases superlinearly in surprisal, and also that variance increases. These results are consistent with the predictions of sampling-based algorithms.

Similarity-Based Interference in Sentence Comprehension in Aphasia: a Computational Evaluation of Two Models of Cue-Based Retrieval

Sentence comprehension requires the listener to link incoming words with short-term memory representations in order to build linguistic dependencies. The cue-based retrieval theory of sentence processing predicts that the retrieval of these memory representations is affected by similarity-based interference. We present the first large-scale computational evaluation of interference effects in two models of sentence processing — the activation-based model and a modification of the direct-access model — in individuals with aphasia (IWA) and control participants in German. The parameters of the models are linked to prominent theories of processing deficits in aphasia, and the models are tested against two linguistic constructions in German: pronoun resolution and relative clauses. The data come from a visual-world eye-tracking experiment combined with a sentence-picture matching task. The results show that both control participants and IWA are susceptible to retrieval interference, and that a combination of theoretical explanations (intermittent deficiencies, slow syntax, and resource reduction) can explain IWA’s deficits in sentence processing. Model comparisons reveal that both models have a similar predictive performance in pronoun resolution, but the activation-based model outperforms the direct-access model in relative clauses.

Predictive Sentence Processing at Speed: Evidence from Online Mouse Cursor Tracking.

Three online mouse cursor-tracking experiments investigated predictive sentence processing at speed. Participants viewed visual arrays with objects like a bike and kite while hearing predictive sentences like, "What the man will ride, which is shown on this page, is the bike," or non-predictive sentences like, "What the man will spot, which is shown on this page, is the bike." Based on the selectional restrictions of "ride" (i.e., vs. "spot"), participants made mouse cursor movements to the bike before hearing the noun "bike." Compellingly, this effect was observed at speech rates of ∼3 (Experiment 1), ∼6 (Experiment 2), and ∼9 (Experiment 3) syllables/s. While prior research suggests striking limits on prediction, these results highlight temporal dynamics that may impact comprehenders' ability to preactivate information when hearing impressively rapid speech. Implications for theories of sentence processing are discussed.

Does a walk-through video help the parser down the garden-path? A visually enhanced self-paced reading study in Dutch.

The human language processing mechanism assigns a structure to the incoming materials as they unfold. There is evidence that the parser prefers some attachment types over others; however, theories of sentence processing are still in dispute over the stage at which each source of information contributes to the parsing system. The present study aims to identify the nature of initial parsing decisions during sentence processing through manipulating attachment type and verbs' argument structure. To this end, we designed a self-paced reading task using globally ambiguous constructions in Dutch. The structures included double locative prepositional phrases (PPs) where the first PP could attach both to the verb (high attachment) and the noun preceding it (low attachment). To disambiguate the structures, we presented a visual context in the form of short animation clips prior to each reading task. Furthermore, we manipulated the argument structure of the sentences using 2- and 3-argument verbs. The results showed that parsing decisions were influenced by contextual cues depending on the argument structure of the verb. That is, the visual context overcame the preference for high attachment only in the case of 2-argument verbs, while this preference persisted in structures including 3-argument verbs as represented by longer reading times for the low attachment interpretations. These findings can be taken as evidence that our language processing system actively integrates information from linguistic and non-linguistic sources from the initial stages of analysis to build up meaning. We discuss our findings in light of serial and parallel models of sentence processing.

Effects of world knowledge on the prediction of upcoming verbs: an eye-tracking study

Some theories of sentence processing make a distinction between two kinds of meaning: a linguistic meaning encoded at the lexicon (i.e., selectional restrictions), and an extralinguistic knowledge derived from our everyday experiences (i.e., world knowledge). According to such theories, the former meaning is privileged over the latter in terms of the time-course of its access and influence during on-line language comprehension. The present study aims to examine whether world knowledge anomalies (that do not violate selectional restrictions) are rapidly detected during online sentence processing. In an eye-tracking experiment, we used materials in which the likelihood of a specific verb (entrevistar or secuestrar, the Spanish translations for to interview and to kidnap) depended on the agent of the event (periodista or terrorista, the Spanish translations for journalist and terrorist). The results showed an effect of typicality in regression path duration and total reading times at both the verb region and the spillover region, thus providing evidence that world knowledge is rapidly accessed and used during on-line sentence comprehension.

Individual Differences in Cue Weighting in Sentence Comprehension: An Evaluation Using Approximate Bayesian Computation.

Cue-based retrieval theories of sentence processing assume that syntactic dependencies are resolved through a content-addressable search process. An important recent claim is that in certain dependency types, the retrieval cues are weighted such that one cue dominates. This cue-weighting proposal aims to explain the observed average behavior, but here we show that there is systematic individual-level variation in cue weighting. Using the Lewis and Vasishth cue-based retrieval model, we estimated individual-level parameters for reading speed and cue weighting using 13 published datasets; hierarchical approximate Bayesian computation (ABC) was used to estimate the parameters. The modeling reveals a nuanced picture of cue weighting: we find support for the idea that some participants weight cues differentially, but not all participants do. Only fast readers tend to have the predicted higher weighting for structural cues, suggesting that reading proficiency (approximated here by reading speed) might be associated with cue weighting. A broader achievement of the work is to demonstrate how individual differences can be investigated in computational models of sentence processing without compromising the complexity of the model.

Case-matching effects under clausal ellipsis and the cue-based theory of sentence processing

This paper is concerned with case-matching effects under clausal ellipsis. We begin by considering available crosslinguistic data that indicate that variation in case marking on a fragment is delimited by the argument structure of the lexical head that assigns case to the fragment’s correlate in the antecedent clause. We then offer experimental evidence for a case-matching preference in Korean when a fragment and its correlate may differ in case marking. This case-matching preference corresponds to a known case of mandatory case-matching in Hungarian, but their relationship is not predicted by any of the existing syntactic accounts of case-matching effects under clausal ellipsis. We propose a novel perspective on fragments that derives case-matching effects, including optional and mandatory case matching, from the predictions of cue-based retrieval. Two further acceptability judgment studies are offered in support of our proposal.

Comparison of Structural Parsers and Neural Language Models as Surprisal Estimators.

Expectation-based theories of sentence processing posit that processing difficulty is determined by predictability in context. While predictability quantified via surprisal has gained empirical support, this representation-agnostic measure leaves open the question of how to best approximate the human comprehender's latent probability model. This article first describes an incremental left-corner parser that incorporates information about common linguistic abstractions such as syntactic categories, predicate-argument structure, and morphological rules as a computational-level model of sentence processing. The article then evaluates a variety of structural parsers and deep neural language models as cognitive models of sentence processing by comparing the predictive power of their surprisal estimates on self-paced reading, eye-tracking, and fMRI data collected during real-time language processing. The results show that surprisal estimates from the proposed left-corner processing model deliver comparable and often superior fits to self-paced reading and eye-tracking data when compared to those from neural language models trained on much more data. This may suggest that the strong linguistic generalizations made by the proposed processing model may help predict humanlike processing costs that manifest in latency-based measures, even when the amount of training data is limited. Additionally, experiments using Transformer-based language models sharing the same primary architecture and training data show a surprising negative correlation between parameter count and fit to self-paced reading and eye-tracking data. These findings suggest that large-scale neural language models are making weaker generalizations based on patterns of lexical items rather than stronger, more humanlike generalizations based on linguistic structure.

Word Order Variation is Partially Constrained by Syntactic Complexity.

Previous work suggests that when speakers linearize syntactic structures, they place longer and more complex dependents further away from the head word to which they belong than shorter and simpler dependents, and that they do so with increasing rigidity the longer expressions get, for example, longer objects tend to be placed further away from their verb, and with less variation. Current theories of sentence processing furthermore make competing predictions on whether longer expressions are preferentially placed as early or as late as possible. Here we test these predictions using hierarchical distributional regression models that allow estimates of word order and word order variation at the level of individual dependencies in corpora from 71 languages, while controlling for confounding effects from the type of dependency (e.g., subject vs. object), and the type of clause (main vs. subordinate) involved as well as from trends that are characteristic of individual languages, language families, and language contact areas. Our results show the expected correlations of length with position and variation only for two out of six dependency types (obliques and nominal modifiers) and no difference between clause types. These findings challenge received theories of across‐the‐board effects of complexity on word order and word order variation and call for theoretical models that relativize effects to specific kinds of syntactic structures and dependencies.

On the grammaticality of morphosyntactically reduced remnants in Polish sluicing

Abstract This paper explores the grammaticality status of reduced sluicing remnants (i.e., remnants realized as NPs due to preposition drop) in Polish. We provide experimental evidence that reduced remnants are variously acceptable in a specific environment (where there is a prior explicit correspondent in the antecedent clause) and are as unacceptable as ungrammatical structures elsewhere. We interpret this pattern as reflecting elaboration effects (i.e., effects that the degree of elaboration of explicit correspondents has on the acceptability of reduced remnants) that follow from the cue-based retrieval theory of sentence processing. Our data support the option of treating reduced remnants as ungrammatical but sometimes acceptable and the option of treating them as grammatical but sometimes degraded, and we discuss how they fit into the current theories of clausal ellipsis.

Capitalization interacts with syntactic complexity.

We investigated whether readers use the low-level cue of proper noun capitalization in the parafovea to infer syntactic category, and whether this results in an early update of the representation of a sentence's syntactic structure. Participants read sentences containing either a subject relative or object relative clause, in which the relative clause's overt argument was a proper noun (e.g., The tall lanky guard who alerted Charlie/Charlie alerted to the danger was young) across three experiments. In Experiment 1 these sentences were presented in normal sentence casing or entirely in upper case. In Experiment 2 participants received either valid or invalid parafoveal previews of the relative clause. In Experiment 3 participants viewed relative clauses in only normal conditions. We hypothesized that we would observe relative clause effects (i.e., inflated fixation times for object relative clauses) while readers were still fixated on the word who, if readers use capitalization to infer a parafoveal word's syntactic class. This would constitute a syntactic parafoveal-on-foveal effect. Furthermore, we hypothesized that this effect should be influenced by sentence casing in Experiment 1 (with no cue for syntactic category being available in upper case sentences) but not by parafoveal preview validity of the target words. We observed syntactic parafoveal-on-foveal effects in Experiment 1 and 3, and a Bayesian analysis of the combined data from all three experiments. These effects seemed to be influenced more by noun capitalization than lexical processing. We discuss our findings in relation to models of eye movement control and sentence processing theories. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

The Influence of Globally Ungrammatical Local Syntactic Constraints on Real-Time Sentence Comprehension: Evidence From the Visual World Paradigm and Reading.

We investigated the influence of globally ungrammatical local syntactic constraints on sentence comprehension, as well as the corresponding activation of global and local representations. In Experiment 1, participants viewed visual scenes with objects like a carousel and motorbike while hearing sentences with noun phrase (NP) or verb phrase (VP) modifiers like "The girl who likes the man (from London/very much) will ride the carousel." In both cases, "girl" and "ride" predicted carousel as the direct object; however, the locally coherent combination "the man from London will ride…" in NP cases alternatively predicted motorbike. During "ride," local constraints, although ruled out by the global constraints, influenced prediction as strongly as global constraints: While motorbike was fixated less than carousel in VP cases, it was fixated as much as carousel in NP cases. In Experiment 2, these local constraints likewise slowed reading times. We discuss implications for theories of sentence processing.

Locus of difficulty in processing L2 English object relative clauses: A study with Korean university students

Previous work on L2 sentence processing shows that L2 learners are sensitive to the greater processing complexity of object relative clauses as compared with subject relative clauses in English as are native speakers of the language. The exact locus of this effect within the relative clause, however, is yet to be empirically determined. In a self-paced reading paradigm with 27 Korean university students, English object relative clauses with an animate subject were contrasted with those that contained an inanimate subject (e.g. The woman that the teenager on the car injured ... vs. The woman that the accident on the road inured ...). Comparable simple sentences embedded as an extraposed that-clause were also tested as a control (e.g., It was verified that the teenager/accident on the road injured the woman ...). The results suggest that the processing difficulty of object relative clauses materializes before the relative clause verb, and that animacy effects appear at both the subject and verb of the relative clause but in different patterns. The findings are discussed in relation to sentence processing theories.

Retrieval cues and syntactic ambiguity resolution: speed-accuracy tradeoff evidence

ABSTRACTLanguage comprehension involves coping with ambiguity and recovering from misanalysis. Syntactic ambiguity resolution is associated with increased reading times, a classic finding that has shaped theories of sentence processing. However, reaction times conflate the time it takes a process to complete with the quality of the behavior-related information available to the system. We therefore used the speed-accuracy tradeoff procedure (SAT) to derive orthogonal estimates of processing time and interpretation accuracy, and tested whether stronger retrieval cues (via semantic relatedness: neighed->horse vs. fell->horse) aid interpretation during recovery. On average, ambiguous sentences took 250ms longer (SAT rate) to interpret than unambiguous controls, demonstrating veridical differences in processing time. Retrieval cues more strongly related to the true subject always increased accuracy, regardless of ambiguity. These findings are consistent with a language processing architecture where cue-driven operations give rise to interpretation, and wherein diagnostic cues aid retrieval, regardless of parsing difficulty or structural uncertainty.

Exhaustive semantic activation for reading ambiguous verbs in Chinese sentences

Studies of lexical ambiguity resolution in sentential contexts have not sufficiently considered the relatedness among an ambiguous word’s meanings as a predicting factor for semantic activation. To better understand the relation between lexical access and discourse processing and the effect of semantic relatedness on lexical ambiguity resolution, a cross-modal lexical priming experiment focusing on Mandarin ambiguous verbs of varying degrees of semantic relatedness was conducted. The results indicated that both meanings of an ambiguous verb were activated regardless of contextual biases and the degrees of semantic relatedness between the associated meanings. Taken together with previous research, the present study suggests that the meanings of an ambiguous word (i.e., homophonic homographs, which share both phonological and orthographic representations) are co-activated exhaustively if they are syntactically licensed by the context. These results thus support the exhaustive semantic activation model of lexical ambiguity resolution and the syntax-first theory of sentence processing.

The Processing of Plural Expressions

AbstractThis paper reviews the psycholinguistic work on the processing of plurals and offers a novel account of how plural expressions are represented during comprehension. The evidence suggests that there are three distinct types of plural representation. The first type of representation is completely underspecified such that the plural expression may denote a singleton. The second type of representation specifies that the plural expression refers to multiple entities without individuating any of the entities. Finally, the last type of plural expression is fully specified in that the number of entities is known and the entities are differentiated. This account of plural representation is consistent with the semantic proposal that plurals are left unmarked for number compared to singulars. This proposal is also consistent with theories of sentence processing that allow some semantic information to be left unresolved during comprehension.

Rational integration of noisy evidence and prior semantic expectations in sentence interpretation

Sentence processing theories typically assume that the input to our language processing mechanisms is an error-free sequence of words. However, this assumption is an oversimplification because noise is present in typical language use (for instance, due to a noisy environment, producer errors, or perceiver errors). A complete theory of human sentence comprehension therefore needs to explain how humans understand language given imperfect input. Indeed, like many cognitive systems, language processing mechanisms may even be "well designed"--in this case for the task of recovering intended meaning from noisy utterances. In particular, comprehension mechanisms may be sensitive to the types of information that an idealized statistical comprehender would be sensitive to. Here, we evaluate four predictions about such a rational (Bayesian) noisy-channel language comprehender in a sentence comprehension task: (i) semantic cues should pull sentence interpretation towards plausible meanings, especially if the wording of the more plausible meaning is close to the observed utterance in terms of the number of edits; (ii) this process should asymmetrically treat insertions and deletions due to the Bayesian "size principle"; such nonliteral interpretation of sentences should (iii) increase with the perceived noise rate of the communicative situation and (iv) decrease if semantically anomalous meanings are more likely to be communicated. These predictions are borne out, strongly suggesting that human language relies on rational statistical inference over a noisy channel.

Processing Chinese relative clauses in context

This paper presents a self-paced reading experiment comparing the processing of subject-extracted relative clauses (SRCs) and object-extracted relative clauses (ORCs) in supportive contexts in Chinese. It is argued that lack of a consistent pattern in the literature for the comparison between Chinese SRCs and ORCs is due to potential temporary ambiguity in these constructions in null contexts. By placing the materials in contexts biased towards a relative clause (RC) interpretation, we limit the effects of temporary ambiguity. The results of the experiment demonstrate that SRCs are read more slowly than ORCs in supportive contexts. These results provide evidence for working memory-based sentence processing theories whereby processing difficulty increases for connecting sentence elements that are further apart. Some convergent evidence that strengthens these conclusions comes from recent research on aphasic populations where a dissociation between English and Chinese RC processing has been revealed: whereas English aphasic patients have more difficulty with ORCs and Chinese aphasic patients have more difficulty with SRCs (Su, Lee, & Chung, 2007). Taken together, these results support the idea that sentence processing is constrained by working memory limitations.

The rhymes that the reader perused confused the meaning: Phonological effects during on-line sentence comprehension

Research on written language comprehension has generally assumed that the phonological properties of a word have little effect on sentence comprehension beyond the processes of word recognition. Two experiments investigated this assumption. Participants silently read relative clauses in which two pairs of words either did or did not have a high degree of phonological overlap. Participants were slower reading and less accurate comprehending the overlap sentences compared to the non-overlapping controls, even though sentences were matched for plausibility and differed by only two words across overlap conditions. A comparison across experiments showed that the overlap effects were larger in the more difficult object relative than in subject relative sentences. The reading patterns showed that phonological representations affect not only memory for recently encountered sentences but also the developing sentence interpretation during on-line processing. Implications for theories of sentence processing and memory are discussed.