The role of verb argument structure in sentence processing: Distinguishing between syntactic and semantic effects

This paper describes an ongoing research program designed to investigate how syntactic and semantic aspects of lexical information become available to the sentence processing system. The two experiments described here distinguished between syntactic and semantic representations by using cross-modal naming and lexical decision in a new way. The relationship between the main verb and the probe word was varied such that the probe word met either the syntactic criteria to be an argument, the semantic criteria, neither, or both the syntactic and semantic criteria. Lexical decision times were sensitive to both syntactic and semantic congruity, while naming times were sensitive only to syntactic congruity. The two tasks were then used to investigate syntactic and semantic representations when verb argument structure was ambiguous. Subcategorized structures were constructed without regard for biasing context, but the contextually inappropriate thematic frame was ruled out while the inappropriate syntactic frame was still available.     

Analysis of the conceptual structure of natural and artifactual categories in Alzheimer's disease patients

This work has two goals, the first one is to study in detail the conceptual structure of some natural and artifactual categories, by using multidimensional scaling (MDS). According to our theoretical approximation, conceptual structure is composed of semantic features, and these features have different entities and nature. As an index of relevance in mental representation of semantic categories, we took the frequency of production of semantic features on a free verbal production task. Our second goal was to compare the structure of conceptual representations in two populations, healthy elderly and Alzheimer's disease (AD) patients, assuming that conceptual representation is impaired in this last population. We assume that the impairment in the conceptual representations of AD patients will show, when the disease reaches a determined level, loosing specific features. This impairment will have serious consequences in the whole conceptual structure.     

Learning words and rules: abstract knowledge of word order in early sentence comprehension

Children quickly acquire basic grammatical facts about their native language. Does this early syntactic knowledge involve knowledge of words or rules? According to lexical accounts of acquisition, abstract syntactic and semantic categories are not primitive to the language-acquisition system; thus, early language comprehension and production are based on verb-specific knowledge. The present experiments challenge this account: We probed the abstractness of young children's knowledge of syntax by testing whether 25- and 21-month-olds extend their knowledge of English word order to new verbs. In four experiments, children used word order appropriately to interpret transitive sentences containing novel verbs. These findings demonstrate that although toddlers have much to learn about their native languages, they represent language experience in terms of an abstract mental vocabulary. These abstract representations allow children to rapidly detect general patterns in their native language, and thus to learn rules as well as words from the start.     

The comprehension of anomalous sentences: evidence from structural priming

We report three experiments investigating how people process anomalous sentences, in particular those in which the anomaly is associated with the verb. We contrast two accounts for the processing of such anomalous sentences: a syntactic account, in which the representations constructed for anomalous sentences are similar in nature to the ones constructed for well-formed sentences; and a semantic account, in which the representations constructed for anomalous sentences are erroneous, or altogether missing, and interpretation is achieved on the basis of semantic representations instead. To distinguish between these accounts, we used structural priming. First, we ruled out the possibility that anomaly per se influences the magnitude of the priming effect: Prime sentences with morphologically incorrect verbs produced similarly enhanced priming (lexical boost) to sentences with the same correct verbs (Exp. 1). Second, we found that prime sentences with a novel verb (Exp. 2) or a semantically and syntactically incongruent verb (Exp. 3) produced a priming effect, which was the same as that produced by well-formed sentences. In accord with the syntactic account, we conclude that the syntactic representations of anomalous sentences are similar to those constructed for well-formed sentences. Our results furthermore suggest that lexically-independent syntactic information is robust enough to produce well-formed syntactic representations during processing without requiring aid from lexically-based syntactic information.     

Mapping sensorimotor sequences to word sequences: A connectionist model of language acquisition and sentence generation

In this article we present a neural network model of sentence generation. The network has both technical and conceptual innovations. Its main technical novelty is in its semantic representations: the messages which form the input to the network are structured as sequences, so that message elements are delivered to the network one at a time. Rather than learning to linearise a static semantic representation as a sequence of words, our network rehearses a sequence of semantic signals, and learns to generate words from selected signals. Conceptually, the network's use of rehearsed sequences of semantic signals is motivated by work in embodied cognition, which posits that the structure of semantic representations has its origin in the serial structure of sensorimotor processing. The rich sequential structure of the network's semantic inputs also allows it to incorporate certain Chomskyan ideas about innate syntactic knowledge and parameter-setting, as well as a more empiricist account of the acquisition of idiomatic syntactic constructions.     

Lexical encoding of event participant information.

The typical understanding of a short passive sentence like The ship was sunk is that it was sunk by someone. We investigated whether unexpressed agent information is introduced via semantic argument information associated with the lexical representations of verbs, or instead via conceptually derived inferences. We demonstrate, in self-paced reading and eye-monitoring studies, that implicit agents are derived from lexical rather than conceptual sources and that verb argument structure information is accessed as soon as a verb is recognized.     

The organization and dissolution of semantic-conceptual knowledge: Is the ‘amodal hub’ the only plausible model?

In recent years, the anatomical and functional bases of conceptual activity have attracted a growing interest. In particular, Patterson and Lambon-Ralph have proposed the existence, in the anterior parts of the temporal lobes, of a mechanism (the 'amodal semantic hub') supporting the interactive activation of semantic representations in all modalities and for all semantic categories. The aim of then present paper is to discuss this model, arguing against the notion of an 'amodal' semantic hub, because we maintain, in agreement with the Damasio's construct of 'higher-order convergence zone', that a continuum exists between perceptual information and conceptual representations, whereas the 'amodal' account views perceptual informations only as a channel through which abstract semantic knowledge can be activated. According to our model, semantic organization can be better explained by two orthogonal higher-order convergence systems, concerning, on one hand, the right vs. left hemisphere and, on the other hand, the ventral vs. dorsal processing pathways. This model posits that conceptual representations may be mainly based upon perceptual activities in the right hemisphere and upon verbal mediation in the left side of the brain. It also assumes that conceptual knowledge based on the convergence of highly processed visual information with other perceptual data (and mainly concerning living categories) may be bilaterally represented in the anterior parts of the temporal lobes, whereas knowledge based on the integration of visual data with action schemata (namely knowledge of actions, body parts and artefacts) may be more represented in the left fronto-temporo-parietal areas.     

Using Language to Navigate the Infant Mind

ABSTRACT— How do infants represent objects, actions, and relations in events? In this review, we discuss an approach to studying this question that begins with linguistic theory—specifically, semantic structures in language. On the basis of recent research exploring infant cognition and prominent linguistic analyses, we examine whether infants' representations of motion events are articulated in terms of the components proposed by Talmy (1985 ; e.g., path, manner) and whether infants' event representations are defined in terms of broad semantic roles (agent, patient, source, goal) as proposed by Jackendoff (1990) and Dowty (1991) . We show how recent findings in infant cognition are consistent with the idea that the infant's representation of events is a close reflection of the linguistic categories. We especially highlight research that is explicitly guided by linguistic categories likely to have correlates in nonlinguistic cognition to illustrate the usefulness of using language to pose questions about early conceptual representations.     

Predicted errors in children's early sentence comprehension

Children use syntax to interpret sentences and learn verbs; this is syntactic bootstrapping. The structure-mapping account of early syntactic bootstrapping proposes that a partial representation of sentence structure, the set of nouns occurring with the verb, guides initial interpretation and provides an abstract format for new learning. This account predicts early successes, but also telltale errors: Toddlers should be unable to tell transitive sentences from other sentences containing two nouns. In testing this prediction, we capitalized on evidence that 21-month-olds use what they have learned about noun order in English sentences to understand new transitive verbs. In two experiments, 21-month-olds applied this noun-order knowledge to two-noun intransitive sentences, mistakenly assigning different interpretations to "The boy and the girl are gorping!" and "The girl and the boy are gorping!". This suggests that toddlers exploit partial representations of sentence structure to guide sentence interpretation; these sparse representations are useful, but error-prone.     

Finding Structure in Time

Time underlies many interesting human behaviors. Thus, the question of how to represent time in connectionist models is very important. One approach is to represent time implicitly by its effects on processing rather than explicitly (as in a spatial representation). The current report develops a proposal along these lines first described by Jordan (1986) which involves the use of recurrent links in order to provide networks with a dynamic memory. In this approach, hidden unit patterns are fed back to themselves: the internal representations which develop thus reflect task demands in the context of prior internal states. A set of simulations is reported which range from relatively simple problems (temporal version of XOR) to discovering syntactic/semantic features for words. The networks are able to learn interesting internal representations which incorporate task demands with memory demands: indeed, in this approach the notion of memory is inextricably bound up with task processing. These representations reveal a rich structure, which allows them to be highly context-dependent, while also expressing generalizations across classes of items. These representations suggest a method for representing lexical categories and the type/token distinction.     

From concepts to lexical items.

In this paper we address the question how in language production conceptual structures are mapped onto lexical items. First we describe the lexical system in a fairly abstract way. Such a system consists of, among other things, a fixed set of basic lexical entries characterized by four groups of information: phonetic form, grammatical features, argument structure, and semantic form. A crucial assumption of the paper is that the meaning in a lexical entry has a complex internal structure composed of more primitive elements (decomposition). Some aspects of argument structure and semantic form and their interaction are discussed with respect to the issue of synonymy. We propose two different mappings involved in lexical access. One maps conceptual structures to semantic forms, and the other maps semantic forms to conceptual structures. Both mappings are context dependent and are many-to-many mappings. We present an elaboration of Levelt's (1989) model in which these processes interact with the grammatical encoder and the mental lexicon. Then we address the consequences of decomposition for processing models, especially the nature of the input of lexical access and the time course. Processing models that use the activation metaphor may have difficulties accounting for certain phenomena where a certain lemma triggers not one, but two or more word forms that have to be produced with other word forms in between.     

熟练潮-普双言者的语义通达机制:来自通道比较的证据

采用跨语言启动的范畴判断范式,通过2个实验,考察了熟练潮-普双言者在听觉通道内和听-视跨通道条件下的语义通达机制。结果发现,无论是以潮州话词语为启动词,还是以普通话词语为启动词,当词对具有高语义相关时,启动词均对目标词产生了显著的跨语言启动效应;在听觉通道内,两种语言之间的启动量无差异,而在听-视跨通道条件下,L2对L1的启动量显著大于L1对L2的启动量。这些结果表明,熟练潮-普双言者共享语义(概念)表征,词汇表征既相互独立又互相联系,L1与L2的词汇均能直接通达语义表征,但其联系强度及其相互之间的词汇表征联系强度均存在通道差异性。整个研究结果支持Kroll等人的修正分级模型。     

Brain responses to filled gaps

An unresolved issue in the study of sentence comprehension is whether the process of gap-filling is mediated by the construction of empty categories (traces), or whether the parser relates fillers directly to the associated verb's argument structure. We conducted an event-related potentials (ERP) study that used the violation paradigm to examine the time course and spatial distribution of brain responses to ungrammatically filled gaps. The results indicate that the earliest brain response to the violation is an early left anterior negativity (eLAN). This ERP indexes an early phase of pure syntactic structure building, temporally preceding ERPs that reflect semantic integration and argument structure satisfaction. The finding is interpreted as evidence that gap-filling is mediated by structurally predicted empty categories, rather than directly by argument structure operations.     

Meaning from syntax: Evidence from 2-year-olds

When toddlers view an event while hearing a novel verb, the verb’s syntactic context has been shown to help them identify its meaning. The current work takes this finding one step further to reveal that even in the absence of an accompanying event, syntactic information supports toddlers’ identification of verb meaning. Two-year-olds were first introduced to dialogues incorporating novel verbs either in transitive or intransitive sentences, but in the absence of any relevant referent scenes (see Yuan & Fisher, 2009). Next, toddlers viewed two candidate scenes: (a) two participants performing synchronous actions, (b) two participants performing a causative action. When asked to “find mooping”, toddlers who had heard transitive sentences chose the causative scene; those who had heard intransitive sentences did not. These results demonstrate that 2-year-olds infer important components of meaning from syntactic structure alone, using it to direct their subsequent search for a referent in a visual scene.     

The vocabulary spurt predicts the emergence of backward semantic inhibition in 18‐month‐old toddlers

The current study examines the relationship between 18‐month‐old toddlers’ vocabulary size and their ability to inhibit attention to no‐longer relevant information using the backward semantic inhibition paradigm. When adults switch attention from one semantic category to another, the former and no‐longer‐relevant semantic category becomes inhibited, and subsequent attention to an item that belongs to the inhibited semantic category is impaired. Here we demonstrate that 18‐month‐olds can inhibit attention to no‐longer relevant semantic categories, but only if they have a relatively large vocabulary. These findings suggest that an increased number of items (word knowledge) in the toddler lexical‐semantic system during the “vocabulary spurt” at 18‐months may be an important driving force behind the emergence of a semantic inhibitory mechanism. Possessing more words in the mental lexicon likely results in the formation of inhibitory links between words, which allow toddlers to select and deselect words and concepts more efficiently. Our findings highlight the role of vocabulary growth in the development of inhibitory processes in the emerging lexical‐semantic system.     

Similarity Judgment Within and Across Categories: A Comprehensive Model Comparison

Similarity is one of the most important relations humans perceive, arguably subserving category learning and categorization, generalization and discrimination, judgment and decision making, and other cognitive functions. Researchers have proposed a wide range of representations and metrics that could be at play in similarity judgment, yet have not comprehensively compared the power of these representations and metrics for predicting similarity within and across different semantic categories. We performed such a comparison by pairing nine prominent vector semantic representations with seven established similarity metrics that could operate on these representations, as well as supervised methods for dimensional weighting in the similarity function. This approach yields a factorial model structure with 126 distinct representation-metric pairs, which we tested on a novel dataset of similarity judgments between pairs of cohyponymic words in eight categories. We found that cosine similarity and Pearson correlation were the overall best performing unweighted similarity functions, and that word vectors derived from free association norms often outperformed word vectors derived from text (including those specialized for similarity). Importantly, models that used human similarity judgments to learn category-specific weights on dimensions yielded substantially better predictions than all unweighted approaches across all types of similarity functions and representations, although dimension weights did not generalize well across semantic categories, suggesting strong category context effects in similarity judgment. We discuss implications of these results for cognitive modeling and natural language processing, as well as for theories of the representations and metrics involved in similarity.     

Reply to David Kemmerer's "a critique of Mark D. Allen's 'the preservation of verb subcategory knowledge in a spoken language comprehension deficit'"

Allen [Allen, M. D. (2005). The preservation of verb subcategory knowledge in a spoken language comprehension deficit. Brain and Language, 95, 255-264] presents evidence from a single patient, WBN, to motivate a theory of lexical processing and representation in which syntactic information may be encoded and retrieved independently of semantic information. In his critique, Kemmerer argues that because Allen depended entirely on preposition-based verb subcategory violations to test WBN's knowledge of correct argument structure, his results, at best, address a "strawman" theory. This argument rests on the assumption that preposition subcategory options are superficial syntactic phenomena which are not represented by argument structure proper. We demonstrate that preposition subcategory is in fact treated as semantically determined argument structure in the theories that Allen evaluated, and thus far from irrelevant. In further discussion of grammatically relevant versus irrelevant semantic features, Kemmerer offers a review of his own studies. However, due to an important design shortcoming in these experiments, we remain unconvinced. Reemphasizing the fact the Allen (2005) never claimed to rule out all semantic contributions to syntax, we propose an improvement in Kemmerer's approach that might provide more satisfactory evidence on the distinction between the kinds of relevant versus irrelevant features his studies have addressed.     

Linking somatic and symbolic representation in semantic memory: the dynamic multilevel reactivation framework

Biological plausibility is an essential constraint for any viable model of semantic memory. Yet, we have only the most rudimentary understanding of how the human brain conducts abstract symbolic transformations that underlie word and object meaning. Neuroscience has evolved a sophisticated arsenal of techniques for elucidating the architecture of conceptual representation. Nevertheless, theoretical convergence remains elusive. Here we describe several contrastive approaches to the organization of semantic knowledge, and in turn we offer our own perspective on two recurring questions in semantic memory research: (1) to what extent are conceptual representations mediated by sensorimotor knowledge (i.e., to what degree is semantic memory embodied)? (2) How might an embodied semantic system represent abstract concepts such as modularity, symbol, or proposition? To address these questions, we review the merits of sensorimotor (i.e., embodied) and amodal (i.e., disembodied) semantic theories and address the neurobiological constraints underlying each. We conclude that the shortcomings of both perspectives in their extreme forms necessitate a hybrid middle ground. We accordingly propose the Dynamic Multilevel Reactivation Framework—an integrative model predicated upon flexible interplay between sensorimotor and amodal symbolic representations mediated by multiple cortical hubs. We discuss applications of the dynamic multilevel reactivation framework to abstract and concrete concept representation and describe how a multidimensional conceptual topography based on emotion, sensation, and magnitude can successfully frame a semantic space containing meanings for both abstract and concrete words. The consideration of ‘abstract conceptual features’ does not diminish the role of logical and/or executive processing in activating, manipulating and using information stored in conceptual representations. Rather, it proposes that the materials upon which these processes operate necessarily combine pure sensorimotor information and higher-order cognitive dimensions involved in symbolic representation.     

Distinguishing Language from Thought: Experimental Evidence That Syntax Is Lexically Rather Than Conceptually Represented

It is generally assumed that syntax is represented linguistically rather than conceptually, consistent with the more general view that language and thought are coded separately. This claim is widely defended on logical grounds, but it has received little experimental support. In the present study, we asked Spanish and English speakers to make semantic and syntactic categorizations for pictures and their corresponding names. Consistent with past results, latencies to semantically categorize pictures and words were similar. The new finding is that participants were faster to make syntactic decisions for words compared with pictures, suggesting that syntactic features such as grammatical gender and the count-mass distinction are more closely linked to lexical than conceptual representations.