Complement Coercion: Distinguishing Between Type-Shifting and Pragmatic Inferencing

Modeling Brain Representations of Words' Concreteness in Context Using GPT-2 and Human Ratings

The meaning of most words in language depends on their context. Understanding how the human brain extracts contextualized meaning, and identifying where in the brain this takes place, remain important scientific challenges. But technological and computational advances in neuroscience and artificial intelligence now provide unprecedented opportunities to study the human brain in action as language is read and understood. Recent contextualized language models seem to be able to capture homonymic meaning variation ("bat", in a baseball vs. a vampire context), as well as more nuanced differences of meaning-for example, polysemous words such as "book", which can be interpreted in distinct but related senses ("explain a book", information, vs. "open a book", object) whose differences are fine-grained. We study these subtle differences in lexical meaning along the concrete/abstract dimension, as they are triggered by verb-noun semantic composition. We analyze functional magnetic resonance imaging (fMRI) activations elicited by Italian verb phrases containing nouns whose interpretation is affected by the verb to different degrees. By using a contextualized language model and human concreteness ratings, we shed light on where in the brain such fine-grained meaning variation takes place and how it is coded. Our results show that phrase concreteness judgments and the contextualized model can predict BOLD activation associated with semantic composition within the language network. Importantly, representations derived from a complex, nonlinear composition process consistently outperform simpler composition approaches. This is compatible with a holistic view of semantic composition in the brain, where semantic representations are modified by the process of composition itself. When looking at individual brain areas, we find that encoding performance is statistically significant, although with differing patterns of results, suggesting differential involvement, in the posterior superior temporal sulcus, inferior frontal gyrus and anterior temporal lobe, and in motor areas previously associated with processing of concreteness/abstractness.

Solving the elusiveness of word meanings: two arguments for a continuous meaning space for language

I explore the hypothesis that the experience of meaning discreteness when we think about the "meaning" of a word is a "communicative" illusion. The illusion is created by processing-contextual constraints that impose disambiguation on the semantic input making salient a specific interpretation within a conceptual space that is otherwise continuous. It is this salience that we experience as discreteness. The understanding of word meaning as non-discrete raises the question of what is context; what are the mechanisms of constraint that it imposes and what is the nature of the conceptual space with which pronunciations (i.e., visual/oral signs) associate themselves. I address these questions by leveraging an algebraic continuous system for word meaning that is itself constrained by two fundamental parameters: control-asymmetry and connectedness. I evaluate this model by meeting two challenges to word meaning discreteness (1) cases where the same pronunciation is associated with multiple senses that are nonetheless interdependent, e.g., English "smoke," and (2) cases where the same pronunciation is associated with a family of meanings, minimally distinct from each other organized as a "cline," e.g., English "have." These cases are not marginal-they are ubiquitous in languages across the world. Any model that captures them is accounting for the meaning system for language. At the heart of the argumentation is the demonstration of how the parameterized space naturally organizes these kinds of cases without appeal for further categorization or segmentation of any kind. From this, I conclude that discreteness in word meaning is epiphenomenal: it is the experience of salience produced by contextual constraints. And that this is possible because, by and large, every time that we become consciously aware of the conceptual structure associated with a pronunciation, i.e., its meaning, we do so under real-time processing conditions which are biased toward producing a specific interpretation in reference to a specific situation in the world. Supporting it is a parameterized space that gives rise to lexico-conceptual representations: generalized algebraic structures necessary for the identification, processing, and encoding of an individual's understanding of the world.

Expressions with Aspectual Verbs Elicit Slower Reading Times than Those with Psychological Verbs: An Eye-Tracking Study in Mandarin Chinese

Research over the last 20 years has investigated the processing costs for sentences such as John began the book. Much of this work has conflated sentences with aspectual verbs, like start or finish, with psychological verbs, like enjoy or tolerate. However, recent studies have reported greater costs for aspectual verbs compared to psychological verbs (e.g., Katsika et al. in Ment Lex 7:58-76, 2012; Lai et al. in Compositionality and concepts in linguistics and psychology, 2017). The present paper reports an eye-tracking study that examined the costs of processing both verb types in Mandarin Chinese. The results revealed greater costs both for aspectual verbs compared to controls (John read the book) and for aspectual verbs compared to psychological verbs, reinforcing the claims of the Structured Individual Hypothesis (Piñango and Deo in J Semant 33:359-408, 2016). Strikingly, there was an early effect at the verb for aspectual verbs but not for psychological verbs. We argue that this result, together with previous findings and other conceptual issues, necessitates a conservative modification of the SIH: aspectual verbs are semantically more complex than psychological verbs. This modification retains the core analysis underlying the SIH, but reconciles the SIH with experimental findings by bringing it in line with the view that lexical semantic complexity has immediate consequences in processing (e.g., Brennan and Pylkkänen in Lang Cogn Process 25:777-807, 2010).

Indeterminate and Enriched Propositions in Context Linger: Evidence From an Eye-Tracking False Memory Paradigm

A sentence such as We finished the paper is indeterminate with regards to what we finished doing with the paper. Indeterminate sentences constitute a test case for two major issues regarding language comprehension: (1) how we compose sentence meaning; and (2) what is retained in memory about what we read in context over time. In an eye-tracking experiment, participants read short stories that were unexpectedly followed by one of three recognition probes: (a) an indeterminate sentence (Lisa began the book), that is identical to the one in the story; (b) an enriched but false probe (Lisa began reading the book); and (c) a contextually unrelated probe (Lisa began writing the book). The probes were presented either at the offset of the original indeterminate sentence in context or following additional neutral discourse. We measured accuracy, probe recognition time, and reading times of the probe sentences. Results showed that, at the immediate time point, participants correctly accepted the identical probes with high accuracy and short recognition times, but that this effect reversed to chance-level accuracy and significantly longer recognition times at the delayed time point. We also found that participants falsely accept the enriched probe at both time points 50% of the time. There were no reading-time differences between identical and enriched probes, suggesting that enrichment might not be an early, mandatory process for indeterminate sentences. Overall, results suggest that while context produces an enriched proposition, an unenriched proposition true to the indeterminate sentence also lingers in memory.

Processing of Complement Coercion With Aspectual Verbs in Mandarin Chinese: Evidence From a Self-Paced Reading Study

This study examines whether Chinese complement coercion sentences with aspectual verbs will elicit processing difficulty during real-time comprehension. Complement coercion is a linguistic phenomenon in which certain verbs (e.g., start, enjoy), requiring an event-denoting complement, are combined with an entity-denoting complement (e.g., book), as in The author started a book. Previous studies have reported that the entity-denoting complement elicited processing difficulty following verbs that require event argument compared with verbs that do not (e.g., The author wrote a book). While the processing of complement coercion has been extensively studied in Indo-European languages such as English and German, it is relatively under-researched in Sino-Tibetan languages such as Mandarin Chinese. Given the fact that there are many linguistic elements behaving distinctly in the different language families, for instance, verbs with respect to their semantic properties and syntactic representations of the complement, it is meaningful to investigate whether or not the existing linguistic differences have any effect on the processing of complement coercion in Mandarin. With this research goal, we recorded self-paced reading time of 61 native Mandarin speakers to investigate the processing of the entity-denoting complement in sentences with three different verb types (aspectual verbs which require an event-denoting complement, preferred verbs which denote a preferred interpretation of the aspectual expressions, and non-preferred verbs which denote a non-preferred but plausible interpretation of the aspectual expressions), as exemplified in 顾客开始/填写/查看这份问卷 gù-kè kāi-shǐ/tián-xiě/chá-kàn zhè-fèn wèn-juàn “The customer started/filled in/checked the questionnaire.” It was found that the entity noun complement (e.g., 这份问卷 zhè-fèn wèn-juàn “the questionnaire”) elicited significantly longer reading times in coercion sentences than non-coercion counterparts. The results are compatible with the previous findings in English that complement coercion sentences impose processing cost during real-time comprehension. The study contributes empirical evidence to coercion studies cross-linguistically.

Context Breeds False Memories for Indeterminate Sentences

What are the roles of semantic and pragmatic processes in the interpretation of sentences in context? And how do we attain such interpretations when sentences are deemed indeterminate? Consider a sentence such as "Lisa began the book" which does not overtly express the activity that Lisa began doing with the book. Although it is believed that individuals compute a specified event to enrich the sentential representation - yielding, e.g., "began [reading] the book" - there is no evidence that a default event meaning is attained. Moreover, if indeterminate sentences are enriched, it is not clear where the information required to generate enriched interpretations come from. Experiment 1 showed that, in isolation, there is no default interpretation for indeterminate sentences. The experiment also showed that biasing contexts constrain event interpretations and improve plausibility judgments, suggesting that event representations for indeterminate sentences are generated by context. In Experiment 2, participants heard biasing discourse contexts and later falsely recognized foil sentences containing the biased events ("Lisa began reading the book") at the same proportion and with the same confidence as the original indeterminate sentence ("Lisa began the book"). We suggest that indeterminate sentences trigger event-enriching inferences but only in sufficiently constraining contexts. We also suggest that indeterminate sentences create two memory traces, one for the proposition consistent with the denotational, compositional meaning, and another for the proposition that is enriched pragmatically over time.

Complement Coercion: The Joint Effects of Type and Typicality

Complement coercion (begin a book →reading) involves a type clash between an event-selecting verb and an entity-denoting object, triggering a covert event (reading). Two main factors involved in complement coercion have been investigated: the semantic type of the object (event vs. entity), and the typicality of the covert event (the author began a book →writing). In previous research, reading times have been measured at the object. However, the influence of the typicality of the subject–object combination on processing an aspectual verb such as begin has not been studied. Using a self-paced reading study, we manipulated semantic type and subject–object typicality, exploiting German word order to measure reading times at the aspectual verb. These variables interacted at the target verb. We conclude that both type and typicality probabilistically guide expectations about upcoming input. These results are compatible with an expectation-based view of complement coercion and language comprehension more generally in which there is rapid interaction between what is typically viewed as linguistic knowledge, and what is typically viewed as domain general knowledge about how the world works.

The Neuronal Correlates of Indeterminate Sentence Comprehension: An fMRI Study

Sentences such as The author started the book are indeterminate because they do not make explicit what the subject (the author) started doing with the object (the book). In principle, indeterminate sentences allow for an infinite number of interpretations. One theory, however, assumes that these sentences are resolved by semanticcoercion, a linguistic process that forces the noun book to be interpreted as an activity (e.g., writing the book) or by a process that interpolates this activity information in the resulting enriched semantic composition. An alternative theory, pragmatic, assumes classical semantic composition, whereby meaning arises from the denotation of words and how they are combined syntactically, with enrichment obtained via pragmatic inferences beyond linguistic-semantic processes. Cognitive neuroscience studies investigating the neuroanatomical and functional correlates of indeterminate sentences have shown activations either at the ventromedial pre-frontal cortex (vmPFC) or at the left inferior frontal gyrus (L-IFG). These studies have supported the semantic coercion theory assuming that one of these regions is where enriched semantic composition takes place. Employing functional magnetic resonance imaging (fMRI), we found that indeterminate sentences activate bilaterally the superior temporal gyrus (STG), the right inferior frontal gyrus (R-IFG), and the anterior cingulate cortex (ACC), more so than control sentences (The author wrote the book). Activation of indeterminate sentences exceeded that of anomalous sentences (…drank the book) and engaged more left- and right-hemisphere areas than other sentence types. We suggest that the widespread activations for indeterminate sentences represent the deployment of pragmatic-inferential processes, which seek to enrich sentence content without necessarily resorting to semantic coercion.

Eye-Tracking and Corpus-Based Analyses of Syntax-Semantics Interactions in Complement Coercion

Previous work has shown that the difficulty associated with processing complex semantic expressions is reduced when the critical constituents appear in separate clauses as opposed to when they appear together in the same clause. We investigated this effect further, focusing in particular on complement coercion, in which an event-selecting verb (e.g., began) combines with a complement that represents an entity (e.g., began the memo). Experiment 1 compared reading times for coercion versus control expressions when the critical verb and complement appeared together in a subject-extracted relative clause (SRC) (e.g., The secretary that began/wrote the memo) compared to when they appeared together in a simple sentence. Readers spent more time processing coercion expressions than control expressions, replicating the typical coercion cost. In addition, readers spent less time processing the verb and complement in SRCs than in simple sentences; however, the magnitude of the coercion cost did not depend on sentence structure. In contrast, Experiment 2 showed that the coercion cost was reduced when the complement appeared as the head of an object-extracted relative clause (ORC) (e.g., The memo that the secretary began/wrote) compared to when the constituents appeared together in an SRC. Consistent with the eye-tracking results of Experiment 2, a corpus analysis showed that expressions requiring complement coercion are more frequent when the constituents are separated by the clause boundary of an ORC compared to when they are embedded together within an SRC. The results provide important information about the types of structural configurations that contribute to reduced difficulty with complex semantic expressions, as well as how these processing patterns are reflected in naturally occurring language.

Metonymy as Referential Dependency: Psycholinguistic and Neurolinguistic Arguments for a Unified Linguistic Treatment

We examine metonymy at psycho- and neurolinguistic levels, seeking to adjudicate between two possible processing implementations (one- vs. two-mechanism). We compare highly conventionalized systematic metonymy (producer-for-product: "All freshmen read O'Connell") to lesser-conventionalized circumstantial metonymy ("[a waitress says to another:] 'Table 2 asked for more coffee."'). Whereas these two metonymy types differ in terms of contextual demands, they each reveal a similar dependency between the named and intended conceptual entities (e.g., Jackendoff, 1997; Nunberg, 1979, 1995). We reason that if each metonymy yields a distinct processing time course and substantially non-overlapping preferential localization pattern, it would not only support a two-mechanism view (one lexical, one pragmatic) but would suggest that conventionalization acts as a linguistic categorizer. By contrast, a similar behavior in time course and localization would support a one-mechanism view and the inference that conventionalization acts instead as a modulator of contextual felicitousness, and that differences in interpretation introduced by conventionalization are of degree, not of kind. Results from three paradigms: self-paced reading (SPR), event-related potentials (ERP), and functional magnetic resonance imaging (fMRI), reveal the following: no main effect by condition (metonymy vs. matched literal control) for either metonymy type immediately after the metonymy trigger, and a main effect for only the Circumstantial metonymy one word post-trigger (SPR); a N400 effect across metonymy types and a late positivity for Circumstantial metonymy (ERP); and a highly overlapping activation connecting the left ventrolateral prefrontal cortex and the left dorsolateral prefrontal cortex (fMRI). Altogether, the pattern observed does not reach the threshold required to justify a two-mechanism system. Instead, the pattern is more naturally (and conservatively) understood as resulting from the implementation of a generalized referential dependency mechanism, modulated by degree of context dependence/conventionalization, thus supporting architectures of language whereby "lexical" and "pragmatic" meaning relations are encoded along a cline of contextual underspecification.

Logical metonymy resolution in a words-as-cues framework: evidence from self-paced reading and probe recognition

Logical metonymy resolution (begin a book → begin reading a book or begin writing a book) has traditionally been explained either through complex lexical entries (qualia structures) or through the integration of the implicit event via post-lexical access to world knowledge. We propose that recent work within the words-as-cues paradigm can provide a more dynamic model of logical metonymy, accounting for early and dynamic integration of complex event information depending on previous contextual cues (agent and patient). We first present a self-paced reading experiment on German subordinate sentences, where metonymic sentences and their paraphrased version differ only in the presence or absence of the clause-final target verb (Der Konditor begann die Glasur → Der Konditor begann, die Glasur aufzutragen/The baker began the icing → The baker began spreading the icing). Longer reading times at the target verb position in a high-typicality condition (baker + icing → spread ) compared to a low-typicality (but still plausible) condition (child + icing → spread) suggest that we make use of knowledge activated by lexical cues to build expectations about events. The early and dynamic integration of event knowledge in metonymy interpretation is bolstered by further evidence from a second experiment using the probe recognition paradigm. Presenting covert events as probes following a high-typicality or a low-typicality metonymic sentence (Der Konditor begann die Glasur → AUFTRAGEN/The baker began the icing → SPREAD), we obtain an analogous effect of typicality at 100 ms interstimulus interval.