Investigating the Extent to which Distributional Semantic Models Capture a Broad Range of Semantic Relations.

Distributional semantic models (DSMs) are a primary method for distilling semantic information from corpora. However, a key question remains: What types of semantic relations among words do DSMs detect? Prior work typically has addressed this question using limited human data that are restricted to semantic similarity and/or general semantic relatedness. We tested eight DSMs that are popular in current cognitive and psycholinguistic research (positive pointwise mutual information; global vectors; and three variations each of Skip-gram and continuous bag of words (CBOW) using word, context, and mean embeddings) on a theoretically motivated, rich set of semantic relations involving words from multiple syntactic classes and spanning the abstract-concrete continuum (19 sets of ratings). We found that, overall, the DSMs are best at capturing overall semantic similarity and also can capture verb-noun thematic role relations and noun-noun event-based relations that play important roles in sentence comprehension. Interestingly, Skip-gram and CBOW performed the best in terms of capturing similarity, whereas GloVe dominated the thematic role and event-based relations. We discuss the theoretical and practical implications of our results, make recommendations for users of these models, and demonstrate significant differences in model performance on event-based relations.

Patterns of hemispheric asymmetry provide evidence dissociating the semantic and syntactic P600.

To understand how neural networks in the left (LH) and right (RH) cerebral hemispheres contribute to different aspects of language comprehension, in two experiments we measured event-related potentials (ERPs) as right-handed participants read sentences, some of which contained morphosyntactic and thematic role violations. Replicating prior work (Kuperberg et al., 2006), in Experiment 1 thematic role violations elicited both an N400 and a (semantic) P600 effect. Morphosyntactic violations elicited effects that differed as a function of participants' familial sinistrality (the presence [FS+] or absence [FS-] of a left-handed biological relative): FS+ participants showed a (syntactic) P600 effect whereas FS- participants showed a biphasic N400 and P600 response. To assess whether this difference reflects different underlying patterns of lateralization, in Experiment 2 target words were presented using visual half-field (VF) presentation. Indeed, for morphosyntactic violations, the FS- group elicited an asymmetric pattern, showing a P600 effect only with LH-biased presentation and an N400 effect in both VFs (cf. Lee and Federmeier, 2015). In contrast, FS+ participants showed a bilateral (N400-only) response pattern. This provides further evidence of FS-based differences in hemispheric contributions to syntactic processing. Strikingly, we found that, when lateralized, thematic role violations did not elicit a P600 effect, suggesting that this effect requires contributions from both hemispheres. The different response patterns for morphosyntactic and thematic role animacy violations across FS and VF also point to a processing difference in the comprehension mechanisms underlying the semantic and syntactic P600, which had heretofore been assumed to be variants of the same component.

Wrong or right? Brain potentials reveal hemispheric asymmetries to semantic relations during word-by-word sentence reading as a function of (fictional) knowledge.

How does individual-level variation in experience and knowledge influence neural mechanisms important for real-time language comprehension? We used event-related brain potentials (ERPs) combined with lateralized visual field (VF) presentation of sentence-final words to examine asymmetries in hemispheric processing as individuals who varied in their knowledge of the fictional world of Harry Potter (HP) read sentences about general topics and HP. HP sentence endings were either contextually supported, unrelated anomalies, or semantically related anomalies. Amongst HP experts, we observed contextual support effects with presentation to both left and right VFs, but related anomaly effects only with presentation to the left VF (right hemisphere). Our findings are in line with accounts in which the left hemisphere (LH) activates relatively narrow/specific semantic information and the RH activates a broader range of semantic relations, including those relating to the construction of mental/situation models, as individuals attempt to comprehend sentences, one word at a time. We suggest that RH-biased hemispheric asymmetries in processing related (but linguistically inappropriate) words in written sentences may emerge as a function of each individual's degree of relevant knowledge. We tentatively hypothesize that content experts may optimize hemispheric differences in scope of semantic activation to maximize both precision (in the LH) and flexibility (in the RH) during language comprehension.

Thematic and other semantic relations central to abstract (and concrete) concepts.

In this article, we discuss multiple types of meaningful (semantic) relations underlying abstract (as compared to concrete) concepts. We adopt the viewpoint that words act as cues to meaning (Elman in Ment Lexicon 6(1):1-34, 2011; Lupyan and Lewis in Lang Cogn Neurosci 34(10):1319-1337, 2019), which is dependent on the dynamic contents of a comprehender's mental model of the situation. This view foregrounds the importance of both linguistic and real-world context as individuals make sense of words, flexibly access relevant knowledge, and understand described events and situations. We discuss theories of, and experimental work on, abstract concepts through the lens of the importance of thematic and other semantic relations. We then tie these findings to the sentence processing literature in which such meaningful relations within sentential contexts are often experimentally manipulated. In this literature, some specific classes/types of abstract words have been studied, although not comprehensively, and with limited connection to the literature on knowledge underlying abstract concepts reviewed herein. We conclude by arguing that the ways in which humans understand relatively more abstract concepts, in particular, can be informed by the careful study of words presented not in isolation, but rather in situational and linguistic contexts, and as a function of individual differences in knowledge, goals, and beliefs.

Harry Potter and the Chamber of What?: The impact of what individuals know on word processing during reading.

During reading, effects of contextual support indexed by N400-a brain potential sensitive to semantic activation/retrieval-amplitude are presumably mediated by comprehenders' world knowledge. Moreover, variability in knowledge may influence the contents, timing, and mechanisms of what is brought to mind during real-time sentence processing. Since it is infeasible to assess the entirety of each individual's knowledge, we investigated a limited domain-the narrative world of Harry Potter (HP). We recorded event-related brain potentials while participants read sentences ending in words more/less contextually supported. For sentences about HP, but not about general topics, contextual N400 effects were graded according to individual participants' HP knowledge. Our results not only confirm that context affects semantic processing by ~250 ms or earlier, on average, but empirically demonstrate what has until now been assumed-that N400 context effects are a function of each individual's knowledge, which here is highly correlated with their reading experience.

To catch a Snitch: Brain potentials reveal variability in the functional organization of (fictional) world knowledge during reading.

We harnessed the temporal sensitivity of event-related brain potentials (ERPs) alongside individual differences in Harry Potter (HP) knowledge to investigate the extent to which the availability and timing of information relevant for real-time written word processing are influenced by variation in domain knowledge. We manipulated meaningful (category, event) relationships between sentence fragments about HP stories and their sentence final words. During word-by-word reading, N400 amplitudes to (a) linguistically supported and (b) unsupported but meaningfully related, but not to (c) unsupported, unrelated sentence endings varied with HP domain knowledge. Single-trial analyses revealed that only the N400s to linguistically supported (but not to either type of unsupported) sentence-final words varied as a function of whether individuals knew (or could remember) the correct (supported) ending for each HP "fact." We conclude that the quick availability of information relevant for word understanding in sentences is a function of individuals' knowledge of both specific facts and the domain to which the facts belong. During written sentence processing, as domain knowledge increases, it is clearly evident that individuals can make use of the relevant knowledge systematically organized around themes, events, and categories in that domain, to the extent they have it.

Lumos!: Electrophysiological tracking of (wizarding) world knowledge use during reading.

In Troyer and Kutas (2018), individual differences in knowledge of the world of Harry Potter (HP) rapidly modulated individuals' average electrical brain potentials to contextually supported words in sentence endings. Using advances in single-trial electroencephalogram analysis, we examined whether this relationship is strictly a result of domain knowledge mediating the proportion of facts each participant knew; we find it is not. Participants read sentences ending in a contextually supported word, reporting online whether they had known each fact. Participants' reports correlated with HP domain knowledge and reliably modulated event-related brain potentials to sentence-final words within 250 ms. Critically, domain knowledge had a dissociable influence in the same time window for endings that participants reported not having known and/or were less likely to be known/remembered across participants. We hypothesize that knowledge impacts written word processing primarily by affecting the neural processes of (implicit) retrieval from long-term memory (LTM): Greater knowledge eases otherwise difficult retrieval processes. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Maternal Socioeconomic Status Influences the Range of Expectations During Language Comprehension in Adulthood.

In infancy, maternal socioeconomic status (SES) is associated with real-time language processing skills, but whether or not (and if so, how) this relationship carries into adulthood is unknown. We explored the effects of maternal SES in college-aged adults on eye-tracked, spoken sentence comprehension tasks using the visual world paradigm. When sentences ended in highly plausible, expected target nouns (Exp. 1), higher SES was associated with a greater likelihood of considering alternative endings related to the action of the sentence. Moreover, for unexpected sentence endings (Exp. 2), individuals from higher SES backgrounds were sensitive to whether the ending was action-related (plausible) or unrelated (implausible), showing a benefit for plausible endings. Individuals from lower SES backgrounds did not show this advantage. This suggests maternal SES can influence the dynamics of sentence processing even in adulthood, with consequences for processing unexpected content. These findings highlight the importance of early lexical experience for adult language skills.

Elaboration over a Discourse Facilitates Retrieval in Sentence Processing.

Language comprehension requires access to stored knowledge and the ability to combine knowledge in new, meaningful ways. Previous work has shown that processing linguistically more complex expressions ('Texas cattle rancher' vs. 'rancher') leads to slow-downs in reading during initial processing, possibly reflecting effort in combining information. Conversely, when this information must subsequently be retrieved (as in filler-gap constructions), processing is facilitated for more complex expressions, possibly because more semantic cues are available during retrieval. To follow up on this hypothesis, we tested whether information distributed across a short discourse can similarly provide effective cues for retrieval. Participants read texts introducing two referents (e.g., two senators), one of whom was described in greater detail than the other (e.g., 'The Democrat had voted for one of the senators, and the Republican had voted for the other, a man from Ohio who was running for president'). The final sentence (e.g., 'The senator who the {Republican/Democrat}had voted for…') contained a relative clause picking out either the Many-Cue referent (with 'Republican') or the One-Cue referent (with 'Democrat'). We predicted facilitated retrieval (faster reading times) for the Many-Cue condition at the verb region ('had voted for'), where readers could understand that 'The senator' is the object of the verb. As predicted, this pattern was observed at the retrieval region and continued throughout the rest of the sentence. Participants also completed the Author/Magazine Recognition Tests (ART/MRT; Stanovich and West, 1989), providing a proxy for world knowledge. Since higher ART/MRT scores may index (a) greater experience accessing relevant knowledge and/or (b) richer/more highly structured representations in semantic memory, we predicted it would be positively associated with effects of elaboration on retrieval. We did not observe the predicted interaction between ART/MRT scores and Cue condition at the retrieval region, though ART/MRT interacted with Cue condition in other locations in the sentence. In sum, we found that providing more elaborative information over the course of a text can facilitate retrieval for referents, consistent with a framework in which referential elaboration over a discourse and not just local linguistic information directly impacts information retrieval during sentence processing.

Action verbs are processed differently in metaphorical and literal sentences depending on the semantic match of visual primes.

Language comprehension requires rapid and flexible access to information stored in long-term memory, likely influenced by activation of rich world knowledge and by brain systems that support the processing of sensorimotor content. We hypothesized that while literal language about biological motion might rely on neurocognitive representations of biological motion specific to the details of the actions described, metaphors rely on more generic representations of motion. In a priming and self-paced reading paradigm, participants saw video clips or images of (a) an intact point-light walker or (b) a scrambled control and read sentences containing literal or metaphoric uses of biological motion verbs either closely or distantly related to the depicted action (walking). We predicted that reading times for literal and metaphorical sentences would show differential sensitivity to the match between the verb and the visual prime. In Experiment 1, we observed interactions between the prime type (walker or scrambled video) and the verb type (close or distant match) for both literal and metaphorical sentences, but with strikingly different patterns. We found no difference in the verb region of literal sentences for Close-Match verbs after walker or scrambled motion primes, but Distant-Match verbs were read more quickly following walker primes. For metaphorical sentences, the results were roughly reversed, with Distant-Match verbs being read more slowly following a walker compared to scrambled motion. In Experiment 2, we observed a similar pattern following still image primes, though critical interactions emerged later in the sentence. We interpret these findings as evidence for shared recruitment of cognitive and neural mechanisms for processing visual and verbal biological motion information. Metaphoric language using biological motion verbs may recruit neurocognitive mechanisms similar to those used in processing literal language but be represented in a less-specific way.

Pre-processing in sentence comprehension: Sensitivity to likely upcoming meaning and structure.

For more than a decade, views of sentence comprehension have been shifting toward wider acceptance of a role for linguistic pre-processing-that is, anticipation, expectancy, (neural) pre-activation, or prediction-of upcoming semantic content and syntactic structure. In this survey, we begin by examining the implications of each of these "brands" of predictive comprehension, including the issue of potential costs and consequences to not encountering highly constrained sentence input. We then describe a number of studies (many using online methodologies) that provide results consistent with prospective sensitivity to various grains and levels of semantic and syntactic information, acknowledging that such pre-processing is likely to occur in other linguistic and extralinguistic domains, as well. This review of anticipatory findings also includes some discussion on the relationship of priming to prediction. We conclude with a brief examination of some possible limits to prediction, and with a suggestion for future work to probe whether and how various strands of prediction may integrate during real-time comprehension.

Virtual EEG: A Software-Based Electroencephalogram Designed for Undergraduate Neuroscience-Related Courses.

A current topic in neuroscience addresses the link between brain activity and visual awareness. The electroencephalogram (EEG), which uses non-invasive high temporal resolution scalp recordings to measure brain activity, is a common tool used to probe this question. EEG recordings, however, are difficult to implement in the curriculum of laboratory-based courses. Thus, undergraduate students often lack experience with EEG experiments. We report here an EEG program (Virtual EEG) that can be used in undergraduate courses to analyze averaged EEG data, termed Event Related Potentials (ERPs). The program was designed so that students can generate hypothesis-driven studies that address how the brain encodes categories of visual stimuli. The Virtual EEG is a large database of EEG recordings consisting of 32 channels taken from real human subjects who were shown 256 pictures of visual stimuli. The program provides a number of possible ways to group the stimuli. After selecting the appropriate stimuli, the program constructs graphs of the ERPs. The channels can be selected for statistical analysis. Because the program uses real data, students are encouraged to interpret their results in light of previously published work. Thus, students have the opportunity to discover something new about how the brain processes visual information. This article also includes a tutorial and summarizes the results of an assessment survey. Finally, we include information regarding the companion Virtual EEG website. The Virtual EEG has been used successfully for the past six years at Indiana University with over a thousand undergraduate students in a research methods course, and the assessment results illustrate its strengths and limitations.