Search

While AI models of sentence comprehension reach high, human-like performance levels, the architecture and neural functioning of these network models are not biologically plausible. We propose a neurobiologically inspired neural network model for sentence comprehension that includes modules representing a mental lexicon, syntactic processing, and semantic processing. Alongside the hierarchical module structure, the developmental trial-and-error (or iterative engineering) process for building the model resulted in the need for two parallel processing pathways, a content-related path that directly forwards lexical information to semantic sentence processing, and a sequence-related path, unfolding the syntactic structure of the sentence. A semantic processing module integrates the information from both pathways. At the model’s highest processing level, the information processed in both pathways allows for thematic role assignment, or semantic event specification, for the sentence. [ABSTRACT FROM AUTHOR]

Recent artificial neural networks that process natural language achieve unprecedented performance in tasks requiring sentence-level understanding. As such, they could be interesting models of the integration of linguistic information in the human brain. We review works that compare these artificial language models with human brain activity and we assess the extent to which this approach has improved our understanding of the neural processes involved in natural language comprehension. Two main results emerge. First, the neural representation of word meaning aligns with the context-dependent, dense word vectors used by the artificial neural networks. Second, the processing hierarchy that emerges within artificial neural networks broadly matches the brain, but is surprisingly inconsistent across studies. We discuss current challenges in establishing artificial neural networks as process models of natural language comprehension. We suggest exploiting the highly structured representational geometry of artificial neural networks when mapping representations to brain data. [ABSTRACT FROM AUTHOR]

Empirical studies have found a processing asymmetry between Chinese subject-extracted relative clauses (SRCs) and object-extracted relative clauses (ORCs). Still, there is no consensus on how this SRC-ORC asymmetry occurs. Thus, aiming to elucidate how the neural activity, in the forms of both event-related potentials (ERPs) and brain oscillations (i.e. event-related synchronisation/desynchronisation, ERS/ERD), attuned to sentences with different levels of processing difficulty, we conducted an electroencephalography (EEG) study to examine the comprehension of Chinese SRCs and ORCs. The results showed an N400 and a P600 effect when comparing SRCs and ORCs. Simultaneously, delta ERS was associated with N400 during the processing of both types of relative clauses and theta ERS with P600 during the processing of SRCs. By incorporating the ERP and ERS indexes, we propose that the dissociation between the integration and retrieval effort involved in sentence comprehension may account for the processing asymmetry between sentences. [ABSTRACT FROM AUTHOR]

Comprehenders across languages tend to interpret role-ambiguous arguments as the subject or the agent of a sentence during parsing. However, the evidence for such a subject/agent preference rests on the comprehension of transitive, active-voice sentences where agents/subjects canonically precede patients/objects. The evidence is thus potentially confounded by the canonical order of arguments. Transitive sentence stimuli additionally conflate the semantic agent role and the syntactic subject function. We resolve these two confounds in an experiment on the comprehension of intransitive sentences in Basque. When exposed to sentence-initial role-ambiguous arguments, comprehenders preferentially interpreted these as agents and had to revise their interpretation when the verb disambiguated to patient-initial readings. The revision was reflected in an N400 component in ERPs and a decrease in power in the alpha and lower beta bands. This finding suggests that sentence processing is guided by a top-down heuristic to interpret ambiguous arguments as agents, independently of word order and independently of transitivity. [ABSTRACT FROM AUTHOR]

In this study, we investigated whether different morphosyntactic features, i.e. number and gender, play a role during the adult online comprehension of subject relative clauses (SRC) and object relative clauses (ORC), in Italian. This study was inspired by developmental studies showing that children struggle with ORC compared to SRC; yet, ORC comprehension improves if the head and the subject of the RC mismatch in relevant morphosyntactic features (e.g. number but not gender in Italian, based on the featural Relativized Minimality principle, fRM). We found that Italian adults read ORC more slowly than SRC verbs; moreover, ORC verbs were read faster in the head-subject number mismatch condition, while there was no facilitation in the head-subject gender mismatch condition, in line with developmental studies and fRM. We conclude that online parsing is feature-sensitive, that features are not all equally "relevant", and that current models should be refined to account for these differences. [ABSTRACT FROM AUTHOR]

Prediction facilitates word processing in the moment, but the longer-term consequences of prediction remain unclear. We investigated whether prediction error during language encoding enhances memory for words later on. German-speaking participants read sentences in which the gender marking of the pre-nominal article was consistent or inconsistent with the predictable noun. During subsequent word recognition, we probed participants' recognition memory for predictable and unpredictable nouns. Our results indicate that individuals who demonstrated early prediction error during sentence reading, showed enhanced recognition memory for nouns overall. Results from an exploratory step-wise regression showed that prenominal prediction error and general reading speed were the best proxies for recognition memory. Hence, prediction error may facilitate recognition by furnishing memory traces built during initial reading of the sentences. Results are discussed in the light of hypotheses positing that predictable words show a memory disadvantage because they are processed less thoroughly. [ABSTRACT FROM AUTHOR]

The effective use of preverbal linguistic cues to make successful clause-final verbal prediction as well as robust maintenance of such predictions has been argued to be a cross-linguistic generalisation for SOV languages such as German and Japanese. In this paper, we show that native speakers of Hindi (an SOV language) falter in forming a clause-final structure in the presence of a centre-embedded relative clause with a non-canonical word order. In particular, the fallibility of the parser is illustrated by the formation of a grammatically illicit locally coherent parse during online processing. Such a parse should not be formed if the grammatically licit matrix clause final structure was being successfully formed. The formation of a locally coherent parse is further illustrated by probing various syntactic dependencies via targeted questions. We show that the parser's susceptibility to form such structures is not driven by top-down processing, rather the effect can only be explained through a bottom-up parsing approach. Further, our investigation suggests that while plausibility is essential, presence of overt agreement features might not be necessary for forming a locally coherent parse in Hindi. These results go against top-down proposals to local coherence such as lossy surprisal and are consistent with the good-enough processing model to comprehension while only partially supporting the SOPARSE account. The work highlights how top-down processing and bottom-up information interact during sentence comprehension in SOV languages – prediction suffers with increased complexity of the preverbal linguistic environment. [ABSTRACT FROM AUTHOR]

Sentence comprehension can be facilitated when readers anticipate the upcoming word. Notwithstanding, it remains uncertain if only the most expected word is anticipated, as postulated by the serial graded hypothesis, or if all probable words are pre-activated, as proposed by the parallel probabilistic hypothesis. To test these contrasting accounts, we compared the processing of expected and unexpected words with second-best words, i.e. the second most expected word in a sentence. The results, from 30 participants, revealed a graded facilitation effect for the expected words, indexed by the N400 mean amplitude, which was the least negative for the most expected words, intermediate for second-best words, and most negative for unexpected words. The Post-N400 Positivity analysis did not reveal any significant effects. The facilitation effect found for the most expected and second-best words suggests that readers can pre-activate multiple candidates during sentence comprehension. [ABSTRACT FROM AUTHOR]

Reprocessing has been speculated to be a late higher-level cognitive process in the P600 time window in event-related potential (ERP) studies. However, uncertainties regarding the P600 as an index of reprocessing have obscured the role of reprocessing in sentence comprehension. In this study, we compared meaningless pseudowords with or without a homophonic repairing clue (homophonic pseudoword vs. ordinary pseudoword) and semantically-plus-syntactically violated real words (doubly violated) to investigate the role of reprocessing in language comprehension. All anomalous conditions elicited significant typical central-parietal P600 effects. The homophonic pseudoword condition elicited a significantly larger P600 effect than the other two anomalous conditions. No significant difference was found between the P600 effects in the ordinary pseudoword and the doubly violated real-word conditions. These results demonstrate an association between reprocessing and the observed P600 effects, supporting an active role of reprocessing in sentence comprehension. Possible implications for language processing models are also discussed. [ABSTRACT FROM AUTHOR]

Given the association between multisensory information and inner attention, the quality of multisensory information is hypothesised to modulate the neuronal activity associated with language comprehension. To verify, we investigated the interaction effect between multisensory quality (sensory enrichment) and semantic (in)congruency during L2 sentence comprehension. English words were selected for the subjects to learn according to the sensory-based emotioncy model. The words were embedded in an acceptability judgment task with 216 sentences under congruent and semantically incongruent conditions. Twenty-two subjects performed the task while their EEG was being recorded. Based on the time–frequency analysis results, in the 300–850 ms window, only lower delta power (1.7–2.2 Hz) showed an enhanced increase after sentences with semantic incongruency. This increase was more pronounced in the sentences with limited multisensory quality. To make up for the insufficient multisensory knowledge and maintain concentration, irrelevant cortical areas were deactivated as a result of lower delta wave synchronisation. [ABSTRACT FROM AUTHOR]

Real-time speech comprehension is challenging because communicatively relevant information is distributed throughout the entire utterance. In five mouse tracking experiments on German and American English, we probe if listeners, in principle, use non-local, early intonational information to anticipate upcoming referents. Listeners had to select a speaker-intended referent with their mouse guided by intonational cues, allowing them to anticipate their decision by moving their hand toward the referent prior to lexical disambiguation. While German listeners (Exps. 1–3) seemed to ignore early pitch cues, American English listeners (Exps. 4–5) were in principle able to use these early pitch cues to anticipate upcoming referents. However, many listeners showed no indication of doing so. These results suggest that there are important positional asymmetries in the way intonational information is integrated, with early information being paid less attention to than later cues in the utterance. Open data, scripts, and materials can be retrieved here: . [ABSTRACT FROM AUTHOR]

Older adults often experience difficulties comprehending speech in noisy backgrounds, which hearing loss does not fully explain. It remains unknown how cognitive abilities, brain networks, and age-related hearing loss may uniquely contribute to speech in noise comprehension at the sentence level. In 31 older adults, using cognitive measures and resting-state fMRI, we investigated the cognitive and neural predictors of speech comprehension with energetic (broadband noise) and informational masking (multi-speakers) effects. Better hearing thresholds and greater working memory abilities were associated with better speech comprehension with energetic masking. Conversely, faster processing speed and stronger functional connectivity between frontoparietal and language networks were associated with better speech comprehension with informational masking. Our findings highlight the importance of the frontoparietal network in older adults' ability to comprehend speech in multi-speaker backgrounds, and that hearing loss and working memory in older adults contributes to speech comprehension abilities related to energetic, but not informational masking. [ABSTRACT FROM AUTHOR]

Any proposal for predictive language comprehension must address receipt of less expected information. While a relationship between the N400 and sentence predictability is well established, a clear picture is still emerging of the link between post-N400 positivities (PNPs) and processing of semantically unexpected words, as well as any relation to other not-specifically-linguistic and/or syntactic late positivities. The current study employs event-related brain potentials (ERPs) to congruent and anomalous words to assess the impacts of semantic relatedness and contextual plausibility on processing unpredictable sentences. We observe PNPs with different scalp topographies to plausible unexpected words unrelated to predictable continuations (anterior PNP) and to anomalous words, regardless of, but delayed by, relatedness (posterior PNP). We offer functional explanations that reconcile inconsistencies with reported PNP findings and place added constraints on the anterior PNP's proposed link to inhibitory processing. We also suggest a testable general cognitive account for the posterior PNP. [ABSTRACT FROM AUTHOR]

The current study examined the influence of contexts' emotional arousal on prediction and integration of words during sentence comprehension. In sentence context, we manipulated the emotional arousal of verbs and the predictability of their following neutral nouns. At the anticipatory stage of the upcoming nouns, a larger sustained negativity was elicited in the highly constraining than the weakly constraining sentences for the neutral verbs but not for the emotional verbs. The results reflect reduced processing loads in response to the pre-activation of the highly predictable nouns when the emotional verbs automatically captured sufficient attentional resources. At the integration stage of the nouns, we found reduced N400 and LPC amplitudes for the highly versus weakly predictable nouns in both the emotionally arousing and non-arousing conditions. The findings suggest that the emotional arousal of context interacts with the prediction of upcoming words, but not the integration of the bottom-up inputs. [ABSTRACT FROM AUTHOR]

Previous cross-modal priming studies showed that lexical decisions to words after a pronoun were facilitated when these words were semantically related to the pronoun's antecedent. These studies suggested that semantic priming effectively measured antecedent retrieval during coreference. We examined whether these effects extended to implicit reading comprehension using the N400 response. The results of three experiments did not yield strong evidence of semantic facilitation due to coreference. Further, the comparison with two additional experiments showed that N400 facilitation effects were reduced in sentences (vs. word pair paradigms) and were modulated by the case morphology of the prime word. We propose that priming effects in cross-modal experiments may have resulted from task-related strategies. More generally, the impact of sentence context and morphological information on priming effects suggests that they may depend on the extent to which the upcoming input is predicted, rather than automatic spreading activation between semantically related words. [ABSTRACT FROM AUTHOR]

German particle verbs consist of a base and a particle, two constituents which occupy separate positions in main clauses, but share one lexical entry. It is still unclear if the combination of particles and bases during sentence comprehension is lexical, syntactic or dual in nature. Using behavioural and ERP measurements, we investigated lexical access and sentence integration of split particle verbs in German two-argument sentences. Our results show that the integration of split particle verbs violating sentence structure or lexical constraints leads to both lexical and syntactic processing difficulty. This extends earlier comparable findings reporting only lexical access difficulties, and suggests that the parse is not immediately abandoned upon encountering a nonexistent particle verb. The integration of grammatical particle verbs assigning lexical case did not lead to measurable processing difficulties. We discuss the impact of this finding for current accounts of the role of lexical case marking in sentence comprehension. [ABSTRACT FROM AUTHOR]

Several recent event-related potential (ERP) studies have observed a left (anterior) negativity (L(A)N) in (morpho)syntactically well-formed but semantically anomalous sentences, which are often referred to as semantic reversal anomalies (e.g. The window closes someone). Such a L(A)N elicitation for semantic reversal anomalies is not expected under the widely held assumption that LANs are associated with morphosyntactic processing loads and are insensitive to semantic anomalies. This raises the empirical question of why semantic reversal anomalies elicit a L(A)N effect. One possible explanation for this observation can be presented in relation the multi-stream processing model, according to which the independent semantic processing stream challenges an analysis outputted by the morphosyntactic processing stream, leading to a misperception of morphosyntactic ill-formedness. Alternatively, the L(A)N effect may reflect a mismatch between expected and actual syntactic structures. As an inanimate subject triggers the expectation of an intransitive structure rather than an active transitive structure, the input of a transitive verb should violate such an expectation. The present study tests these two hypotheses by manipulating the temporal predictabilities of verbs to examine the underlying cognitive processes of L(A)N effects in semantic reversal anomalies. The results reveal a L(A)N effect for semantic reversal anomalies only when prediction of a verb type was possible, in favour of the latter prediction-based view. The implications of this finding are discussed with respect to predictive processing mechanisms and the architecture of processing systems. [ABSTRACT FROM AUTHOR]

Despite strong evidence for prediction during language comprehension, the underlying mechanisms, and the extent to which they are specific to language, remain unclear. Re-analysing an event-related potentials study, we examined responses in the time-frequency domain to expected and unexpected (but plausible) words in strongly and weakly constraining sentences, and found results similar to those reported in nonverbal domains. Relative to expected words, unexpected words elicited an increase in the theta band (4–7 Hz) in strongly constraining contexts, suggesting the involvement of control processes to deal with the consequences of having a prediction disconfirmed. Prior to critical word onset, strongly constraining sentences exhibited a decrease in the alpha band (8–12 Hz) relative to weakly constraining sentences, suggesting that comprehenders can take advantage of predictive sentence contexts to prepare for the input. The results suggest that the brain recruits domain-general preparation and control mechanisms when making and assessing predictions during sentence comprehension. [ABSTRACT FROM PUBLISHER]

Various findings suggest that once a verb is accessed, all of its complementation options are activated. This fMRI study examined whether all the complementation options are activated even in contexts where this seems unnecessary. We examined whether introducing the selected complement prior to the verb (in topicalised sentences) still involves the activation of all complementation options. We performed ROI analyses in the left STG, a brain region that has been linked to the processing of argument structure and the number of complementation options. In this region, multiple-option verbs elicited greater activations compared with one-option verbs, both when the complement appeared after the verb and when it appeared pre-verbally. This suggests encapsulated lexical retrieval of the verb, which involves exhaustive activation of all its complementation options when the verb is accessed. [ABSTRACT FROM PUBLISHER]

Comprehension of classifier-noun sequences was examined in separate studies in English and Mandarin by comparing event-related brain potential (ERP) responses to classifier-noun matches (a sheet of paper) and mismatches (a sheet of coffee) embedded in sentences. One goal was to determine which ERP components are sensitive to such mismatches, as a clue about the nature of the underlying combinatorial processes. Another goal was to examine effects of classifier constraint strength (a piece of …  vs.a sheet of … ) on anticipation of a subsequent noun. Results were similar in the two languages, which is remarkable given substantial differences between them in classifier usage. In both languages, nouns evoked larger N400s in mismatching classifier-noun sequences, suggesting that combinatorial processing was primarily semantic, and general classifiers evoked a larger sustained frontal negativity than specific classifiers starting 200 milliseconds after classifier onset, reflecting effects of constraint strength on anticipation of the upcoming noun. [ABSTRACT FROM AUTHOR]

In order to study the difficulties experienced during sentence comprehension in a foreign language (L2), we investigated semantic and world knowledge information retrieval in L2 comprehenders. Event-related potentials (ERP) were collected in late learners of English whose native language is Spanish, performing a sentence reading task in English. We investigated the mean amplitude of the P2 and N400 ERP components elicited by the critical word of sentences in three conditions: (1) correct,, (2) semantic violation, and (3) world knowledge violation (semantically acceptable but factually untrue). In the N400 window, ERP modulations elicited by semantic and world knowledge violations had similar amplitudes, as previously observed in LI comprehenders. However, semantic violations failed to modulate P2 mean amplitude as it did in native speakers. These results suggest that, whilst L2 and LI readers similarly integrate world knowledge and semantic information, L2 readers lack fast semantic access at around 200 ms after the onset of a critical word. [ABSTRACT FROM AUTHOR]

Relative clauses (RCs) in Chinese are prenominal. In object-modifying, object-extracted RCs (e.g. Click on [RC the ball broke] window, meaning 'Click on the window [RC that the ball broke]), the ambiguous status of the local noun ball and the long-distance attachment of the head noun window into the main verb appear to make online parsing of Chinese RCs particularly difficult. By interposing mismatching classifiers and the passive marker BEI into the RC sentences, we investigated whether the presence of incomplete heads would add storage costs, as predicted by the Dependency Locality Theory (DLT), or would serve as retrieval cues to help pre-build the RC structure, as predicted by the cue-based retrieval theory. Results from a visual world eye-tracking experiment and a self-paced reading showed that Chinese comprehenders are able to use BEI cues and the mismatching classifier (albeit to a less extent) to pre-build RC structure, providing support for the cue-based retrieval theory. [ABSTRACT FROM AUTHOR]

Two individual differences experiments examined the relationship between executive control and the revision of misinterpretations in sentence comprehension. Garden-path sentences were used as they often lead to initial misinterpretations, necessitating revision during comprehension. In addition to garden-path revision, verbal and non-verbal executive controls were assessed by using the verbal and non-verbal version of the Stroop task. Experiment 1 showed that garden-path revision errors in a grammaticality judgement task correlated with verbal Stroop interference errors. Experiment 2 further showed that the time taken to revise the garden-path interpretation correlated with the time taken to resolve verbal Stroop interference, but not with the time taken to resolve non-verbal Stroop interference. Together, the results argue for a role of executive control, which is possibly domain-specific, in the revision of misinterpretations during sentence comprehension. [ABSTRACT FROM AUTHOR]