Syntactic Input Research Articles

Effect of Healthy Aging and Gender on Syntactic Input Processing: A P600 Event-Related Potential Study.

This study aimed to investigate the effect of healthy aging and gender, as well as the interaction, thereof, on syntactic input processing during sentence comprehension. This was achieved through the recording of the P600 event-related potential. Sixty Flemish (native speakers of Dutch) participants (30 men and 30 women), equally distributed into three age groups (young, middle-aged, and older adults), were subjected to a visually presented word order violation task under simultaneous electro-encephalography recording. The task contained 60 sentences, of which half were grammatical and half contained a word order violation. P600 responses were analyzed for amplitude, latency, topographical distribution, and source localization. Regarding the effect of healthy aging, no age-related differences were found for the amplitude, onset latency, and topographical distribution of the P600 effect (difference wave). Although aging effects on the P600 effect amplitude were absent, a reduced P600 amplitude in response to both the grammatical and ungrammatical sentences was found, next to a reduced overall degree of source activation in linguistic regions of interest. Also, a reduced behavioral accuracy in response to the word order violation was observed in the older adults group. Regarding the effect of gender, females exhibited a larger P600 effect amplitude and a reduced behavioral accuracy compared to males. No gender-related differences were found for P600 effect onset latency, topographical distribution, and source activation. While this study demonstrates no effect of aging on the P600 effect, the lower behavioral response and absence of any activation shift argues against functional compensation. Moreover, although increased neural activation in women combined with their reduced behavioral accuracy may indicate the use of different cognitive strategies in men and women, source localization analysis could not objectify this hypothesis.

A prosodic theory of possible ellipsis remnants

We formulate a new generalization of the distribution of ellipsis remnants. Ellipsis cannot strand functional material to the exclusion of a potential prosodic host (the Stranding Generalization). Explaining the Stranding Generalization requires a theory of ellipsis in which the prosodic needs of ellipsis remnants can be taken into account. Drawing on Match Theory (Selkirk 2009, 2011), we develop an account which locates the computation of ellipsis in the syntax-prosody mapping. Specifically, ellipsis results from an optional reranking of a constraint (DestressGiven), which forces reduction of semantically recoverable material, over Match constraints governing the realization of syntactic elements in prosodic structure. The Stranding Generalization is shown to follow from independently motivated prosodic well-formedness constraints, which in the relevant cases cannot be reconciled with the ranking responsible for ellipsis. The broader implications of our analysis, if successful, is that it motivates a view of ellipsis whereby any constraints on ellipsis beyond semantic recoverability are the result of competition between candidates for the possible phonological output of the syntactic input.

A combined syntactic-semantic embedding model based on lexicalized tree-adjoining grammar

This paper presents a joint syntactic-semantic embedding model which not only uses syntactic information to enrich the word embeddings but also generates distributed representations for the syntactic structures themselves. The syntactic input to our model comes from a Lexicalized Tree-Adjoining Grammar parser. The word embeddings from our model outperform the Skip-gram embeddings in several word similarity and sentiment classification experiments. The syntactic structure embeddings help improve a transition-based dependency parser by a clear margin.

Syntactic processing in L2 depends on perceived reliability of the input: Evidence from P600 responses to correct input.

In 3 ERP experiments, we investigated how experienced L2 speakers process natural and correct syntactic input that deviates from their own, sometimes incorrect, syntactic representations. Our previous study (Lemhöfer, Schriefers, & Indefrey, 2014) had shown that L2 speakers do engage in native-like syntactic processing of gender agreement but base this processing on their own idiosyncratic (and sometimes incorrect) grammars. However, as in other standard ERP studies, but different from realistic L2 input, the materials in that study contained a large proportion of incorrect sentences. In the present study, German speakers of Dutch read exclusively objectively correct Dutch sentences that did or did not contain subjective determiner "errors" (e.g., de boot "the boat," which conflicts with the intuition of many German speakers that the correct phrase should be het boot). During reading for comprehension (Experiment 1), no syntax-related ERP responses for subjectively incorrect compared to correct phrases were observed. The same was true even when participants explicitly attended to and learned from the determiners in the sentences (Experiment 2). Only when participants judged the correctness of determiners in each sentence (Experiment 3) did a clear P600 appear. These results suggest that the full and native-like use of subjective grammars, as reflected in the P600 to subjective violations, occurs only when speakers have reason to mistrust the grammaticality of the input, either because of the nature of the task (grammaticality judgments) or because of the salient presence of incorrect sentences. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Culturally embedded schemata for false belief reasoning

I argue that both language acquisition and cultural and social factors contribute to the formation of schemata that facilitate false belief reasoning. While the proposal for an active role of language acquisition in this sense has been partially advanced by several voices in the mentalizing debate, I argue that other accounts addressing this issue present some shortcomings. Specifically, I analyze the existing proposals distinguishing between “structure-oriented” views (with the most relevant representative in de Villiers and de Villiers in Complements enable representation of the contents of false beliefs: the evolution of a theory of theory of mind. Language Acquisition, Foster-Cohen, 2009; de Villiers in Int Rev Gen Linguistics 117(11):1858–1878, 2007; de Villiers in Why language matters for theory of mind. Oxford University Press, New York, pp 186–219, 2005), that stress the role of language as a set of rules providing syntactic input and providing a representational format, and “cultural/social-oriented views”, that stress the role of social interaction (among others Nelson in Why language matters for theory of mind. Oxford University Press, Oxford, 2005; Hutto in Folk psychological narratives: the sociocultural basis of understanding reasons. Cambridge University Press, Cambridge, 2008b). Starting from the analysis of these views, I defend my own account of the role of language acquisition in aiding false belief reasoning. I argue that language acquisition plays a pivotal role in the formation of schemata used by pre-schoolers to pass the false belief task. I propose a specific learning mechanism for exploiting linguistic information that taps into specific cognitive abilities, thus making a very concrete suggestion about the role of linguistic input in specific cultural contexts for the development of false belief skills.

How readers process syntactic input depends on their goals

During reading, the recognition of words is influenced by the syntactic compatibility of surrounding words: a sentence-superiority effect. However, when the goal is to make syntactic categorization decisions about single target words, these decisions are influenced by the syntactic congruency rather than compatibility of surrounding words. Although both these premises imply that readers can extract syntactic information from multiple words in parallel, they also suggest that how the brain organizes syntactic input—and consequently how surrounding stimuli affect word recognition—depends on the reader's top-down goals. The present study provides a direct test of this conception. Participants were offered nouns and verbs amidst a grammatical context (‘this horse fell’) and ungrammatical context (‘fell horse this’). Using a conditional task setup, we manipulated the amount of emphasis put on respectively sentences and single words. In two blocks readers were instructed to make sentence grammaticality judgments only if the middle word was respectively noun or verb; in two other blocks readers were instructed to syntactically categorize the middle word only if the sentence was respectively correct or incorrect. We established an interaction effect whereby the impact of grammatical correctness on syntactic categorization decisions was greater than the effect of grammatical correctness per se. This first sentence-superiority effect in the categorization of single words, combined with the absence of this effect in prior flanker studies, leads us to surmise that word-to-word syntactic constraints only operate if the reader is engaged in sentence processing.

The influence of syntactic expectations on reading comprehension is malleable and strategic: An eye-tracking study of English dative alternation.

Language processing is incremental and inherently predictive. Against this theoretic backdrop, we investigated the role of upcoming structural information in the comprehension of the English dative alternation. The use of eye-tracking enabled us to examine both the time course and locus of the effect associated with (a) structural expectations based on a lifetime of experience with language, and (b) rapid adaptation of the reader to the local statistics of the experiment. We quantified (a) as a verb subcategorization bias toward dative alternatives, and (b) as distributional biases in the syntactic input during the experiment. A reliable facilitatory effect of the verb bias was only observed in the double-object datives and only in the disambiguation region of the second object. Furthermore, structural priming led to an earlier locus of the verb bias effect, suggesting an interaction between (a) and (b). Our results offer a new outlook on the utilization of syntactic expectations during reading, in conjunction with rapid adaptation to the immediate linguistic environment. We demonstrate that this utilization is both malleable and strategic. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Dependency or Span, End-to-End Uniform Semantic Role Labeling

Semantic role labeling (SRL) aims to discover the predicateargument structure of a sentence. End-to-end SRL without syntactic input has received great attention. However, most of them focus on either span-based or dependency-based semantic representation form and only show specific model optimization respectively. Meanwhile, handling these two SRL tasks uniformly was less successful. This paper presents an end-to-end model for both dependency and span SRL with a unified argument representation to deal with two different types of argument annotations in a uniform fashion. Furthermore, we jointly predict all predicates and arguments, especially including long-term ignored predicate identification subtask. Our single model achieves new state-of-the-art results on both span (CoNLL 2005, 2012) and dependency (CoNLL 2008, 2009) SRL benchmarks.

Prosodic Noun Incorporation: The Relationship between Prosody and Argument Structure in Niuean

AbstractThis paper argues for a uniform X0‐raising analysis of VSO and VOS word order in Niuean. Accounting for VSO via X0raising has a strong foundation in the V1 literature; the claim that X0raising can also underlie VOS word order—as in Niuean's pseudo‐noun‐incorporation construction—is more controversial. This paper explains the VOS order of Niuean pseudo noun incorporation by appealing to a condition on prosodic well‐formedness,Argument‐φ, that requires a head and its internal argument(s) to form a unique phonological phrase. In order to satisfy this requirement, the incorporated argument undergoes prosodic restructuring into a position adjacent to the verb at PF. Since the verb arrives at its clause‐initial position via X0raising, the syntactic input to the prosodic grammar is VSO, while the prosodic output is VOS.

Three-Verb Clusters in Interference Frisian: A Stochastic Model over Sequential Syntactic Input.

Abstract Interference Frisian (IF) is a variety of Frisian, spoken by mostly younger speakers, which is heavily influenced by Dutch. IF exhibits all six logically possible word orders in a cluster of three verbs. This phenomenon has been researched by Koeneman and Postma (2006), who argue for a parameter theory, which leaves frequency differences between various orders unexplained. Rejecting Koeneman and Postma's parameter theory, but accepting their conclusion that Dutch (and Frisian) data are input for the grammar of IF, we will argue that the word order preferences of speakers of IF are determined by frequency and similarity. More specifically, three-verb clusters in IF are sensitive to: their linear left-to-right similarity to two-verb clusters and three-verb clusters in Frisian and in Dutch; the (estimated) frequency of two- and three-verb clusters in Frisian and Dutch. The model will be shown to work best if Dutch and Frisian, and two- and three-verb clusters, have equal impact factors. If different impact factors are taken, the model's predictions do not change substantially, testifying to its robustness. This analysis is in line with recent ideas that the sequential nature of human speech is more important to syntactic processes than commonly assumed, and that less burden need be put on the hierarchical dimension of syntactic structure.

Dependency-based n-gram models for general purpose sentence realisation

AbstractThis paper presents a general-purpose, wide-coverage, probabilistic sentence generator based on dependency n-gram models. This is particularly interesting as many semantic or abstract syntactic input specifications for sentence realisation can be represented as labelled bi-lexical dependencies or typed predicate-argument structures. Our generation method captures the mapping between semantic representations and surface forms by linearising a set of dependencies directly, rather than via the application of grammar rules as in more traditional chart-style or unification-based generators. In contrast to conventional n-gram language models over surface word forms, we exploit structural information and various linguistic features inherent in the dependency representations to constrain the generation space and improve the generation quality. A series of experiments shows that dependency-based n-gram models generalise well to different languages (English and Chinese) and representations (LFG and CoNLL). Compared with state-of-the-art generation systems, our general-purpose sentence realiser is highly competitive with the added advantages of being simple, fast, robust and accurate.

A contextual approach to facilitating word retrieval in agrammatic aphasia

Background: Although virtually all individuals with aphasia demonstrate problems with word retrieval, this symptom might arise for different reasons in individuals with different types of aphasia. A trade‐off of dependence on semantic and syntactic information is hypothesised to underlie dissociations in word retrieval shown by fluent anomic and non‐fluent agrammatic speakers. This division of labour predicts that strengthening semantic input will improve word retrieval for those with anomic aphasia, whereas strengthening syntactic input through contextual cues will improve word retrieval for those with agrammatic aphasia. I would like to thank the two participants in this study for their patience and perseverance. Several students helped with the collection and analysis of data, particularly (in chronological order) Rebekah Abel, Michelle Bullock, Ellen Marschner, Becky Bartlett, and Anna Happ. Finally, I would like to express my gratitude to ASHA's special interest Division 2 for the financial and moral support provided to carry out this study. Aims: To explore the outcome of a new treatment approach that proposes to improve word retrieval in individuals with agrammatic aphasia by implicitly strengthening the links between target words and associated words which co‐occur in a connected speech context. Methods & Procedures: The outcomes of two therapy approaches were examined in two participants with agrammatic aphasia. One therapy approach focused on explicitly training semantic features of target words; the other, more novel, approach focused on implicitly strengthening contextual associations through story telling and retelling. It was predicted that the latter approach would result in greater benefits for the two agrammatic participants. Outcomes & Results: Although both therapy approaches appeared to be effective, the predicted advantage of the contextually based approach was not found. The evolution of the error patterns throughout the treatment was examined to help understand the mechanisms underlying the improvements shown for each participant. Conclusions: A novel treatment involving training words in a story context was shown to result in improved word retrieval for two participants with agrammatic aphasia. The merits of the approach are discussed, relative to more traditional explicit word retrieval therapy approaches.

Semantic weight and verb retrieval in aphasia

Individuals with agrammatic aphasia may have difficulty with verb production in comparison to nouns. Additionally, they may have greater difficulty producing verbs that have fewer semantic components (i.e., are semantically “light”) compared to verbs that have greater semantic weight. A connectionist verb-production model proposed by Gordon and Dell (2003) learns through error correction to “divide the labor” between syntax and semantics. Verbs that are semantically heavier come to depend less on syntax and more on semantics. For lighter verbs, the reverse is true. We performed this study to clarify the role of semantic weight in aphasic verb production and to test the prediction from Gordon and Dell that a brain lesion that impairs the syntactic input to verb retrieval will impair lighter verbs more than heavier ones. Consistent with this prediction, we found that the decrement for lighter verbs was present in a group with agrammatic aphasia but not in a matched group without agrammatism.

Dividing things up: The semantics of or and the modal/or interaction

In this paper, the meanings of sentences containing the word or and a modal verb are used to arrive at a novel account of the meaning of or coordinations. It is proposed that or coordinations denote sets whose members are the denotations of the disjuncts; and that the truth conditions of sentences containing or coordinations require the existence of some set made available by the semantic environment which can be ‘divided up’ in accordance with the disjuncts. The relevant notion of ‘dividing things up’ is made explicit in the paper. Detailed attention is given to the question of how the proposed truth conditions are derived from the syntactic input. The account offered allows for the derivation of both the disjunctive and the non-disjunctive readings of modal/or sentences, including the much-discussed ‘free choice’ readings of may/or sentences.

Activations of “motor” and other non-language structures during sentence comprehension

In this paper we report the results of an experiment in which subjects read syntactically unambiguous and ambiguous sentences which were disambiguated after several words to the less likely possibility. Understanding such sentences involves building an initial structure, inhibiting the non-preferred structure, detecting that later input is incompatible with the initial structure, and reactivating the alternative structure. The ambiguous sentences activated four areas more than the unambiguous sentences. These areas are the left inferior frontal gyrus (IFG), the right basal ganglia (BG), the right posterior dorsal cerebellum (CB) and the left median superior frontal gyrus (SFG). The left IFG is normally activated when syntactic processing complexity is increased and probably supports that function in the current study as well. We discuss four hypotheses concerning how these areas may support comprehension of syntactically ambiguous sentences. (1) The left IFG, right CB and BG could support articulatory rehearsal used to support the processing of ambiguous sentences. This seems unlikely since the activation pattern associated with articulatory rehearsal in other studies is not similar to that seen here. (2) The CB acts as an error detector in motor processing. Error detection is important for recognizing that the wrong sentence structure has been chosen initially. (3) The BG acts to select and sequence movements in the motor domain and in cognitive domains may serve to inhibit competing and completed plans which is not unlike inhibiting the initially non-preferred structure or “unchoosing” the initial choice when incompatible syntactic input is received. (4) The left median SFG is relevant for the evaluation of plausibility. Evaluating the plausibility of the two possibilities provides an important basis for choosing between them. The notion of the use of domain general cognitive processes to support a linguistic process is in line with recent suggestions that the a given area may subserve a specific cognitive task because it carries out an appropriate sort of computation rather than because it supports a specific cognitive domain.

Multiple lexical selection and parallelism in Hungarian VP ellipsis

We propose that forward and backward ellipsis should both be regardedas the non-insertion of the phonological form under the terminal nodes offully specied lexical-figrammatical feature matrices. We conceive of the creationof structures with VP ellipsis as the multiple selection from the lexiconof the lexical items relevant for the licensing and execution of ellipsisand enumerating them at the syntactic input in such away that they can satisfythe constraints of structural and referential parallelism. It is in this casethat the option of not inserting the phonological form may be chosen.

The representation and processing of coreference in discourse

A model is presented that addresses both the distribution and comprehension of different forms of referring expressions in language. This model is expressed in a formalism (Kamp & Reyle, 1993) that uses interpretive rules to map syntactic representations onto representations of discourse. Basic interpretive rules are developed for names, pronouns, definite descriptions, and quantified descriptions. These rules are triggered by syntactic input and interact dynamically with representations of discourse to establish reference and coreference. This interaction determines the ease with which coreference can be established for different linguistic forms given the existing discourse context. The performance of the model approximates that observed in studies of intuitive judgments of grammaticality and studies using online measures of language comprehension. The model uses the same basic interpretive mechanisms for coreference within and between sentences, thereby linking the domain traditionally studied by generative linguists to domains that have been of concern primarily to psychologists and computational linguists.

The Representation and Processing of Coreference in Discourse

A model is presented that addresses both the distribution and comprehension of different forms of referring expressions in language. This model is expressed in a formalism (Kamp & Reyle, 1993) that uses interpretive rules to map syntactic representations onto representations of discourse. Basic interpretive rules are developed for names, pronouns, definite descriptions, and quantified descriptions. These rules are triggered by syntactic input and interact dynamically with representations of discourse to establish reference and coreference. This interaction determines the ease with which coreference can be established for different linguistic forms given the existing discourse context. The performance of the model approximates that observed in studies of intuitive judgments of grammaticality and studies using online measures of language comprehension. The model uses the same basic interpretive mechanisms for coreference within and between sentences, thereby linking the domain traditionally studied by generative linguists to domains that have been of concern primarily to psychologists and computational linguists.

Semantic criteria and syntactic input in the acquisition of verb argument structure

Semantic criteria and syntactic input in the acquisition of verb argument structure