Single-Stage Prediction Models Do Not Explain the Magnitude of Syntactic Disambiguation Difficulty

The structure and statistics of language jointly shape cross-frequency neural dynamics during spoken language comprehension

Humans excel at extracting structurally-determined meaning from speech despite inherent physical variability. This study explores the brain's ability to predict and understand spoken language robustly. It investigates the relationship between structural and statistical language knowledge in brain dynamics, focusing on phase and amplitude modulation. Using syntactic features from constituent hierarchies and surface statistics from a transformer model as predictors of forward encoding models, we reconstructed cross-frequency neural dynamics from MEG data during audiobook listening. Our findings challenge a strict separation of linguistic structure and statistics in the brain, with both aiding neural signal reconstruction. Syntactic features have a more temporally spread impact, and both word entropy and the number of closing syntactic constituents are linked to the phase-amplitude coupling of neural dynamics, implying a role in temporal prediction and cortical oscillation alignment during speech processing. Our results indicate that structured and statistical information jointly shape neural dynamics during spoken language comprehension and suggest an integration process via a cross-frequency coupling mechanism.

Grammar and Expectation in Active Dependency Resolution: Experimental and Modeling Evidence From Norwegian

Filler-gap dependency resolution is often characterized as an active process. We probed the mechanisms that determine where and why comprehenders posit gaps during incremental processing using Norwegian as our test language. First, we investigated why active filler-gap dependency resolution is suspended inside island domains like embedded questions in some languages. Processing-based accounts hold that resource limitations prevent gap-filling in embedded questions across languages, while grammar-based accounts predict that active gap-filling is only blocked in languages where embedded questions are grammatical islands. In a self-paced reading study, we find that Norwegian participants exhibit filled-gap effects inside embedded questions, which are not islands in the language. The findings are consistent with grammar-based, but not processing, accounts. Second, we asked if active filler-gap processing can be understood as a special case of probabilistic ambiguity resolution within an expectation-based framework. To do so, we tested whether word-by-word surprisal values from a neural language model could predict the location and magnitude of filled-gap effects in our behavioral data. We find that surprisal accurately tracks the location of filled-gap effects but severely underestimates their magnitude. This suggests either that mechanisms above and beyond probabilistic ambiguity resolution are required to fully explain active gap-filling behavior or that surprisal values derived from long-short term memory are not good proxies for humans' incremental expectations during filler-gap resolution.

Language in Brains, Minds, and Machines

It has long been argued that only humans could produce and understand language. But now, for the first time, artificial language models (LMs) achieve this feat. Here we survey the new purchase LMs are providing on the question of how language is implemented in the brain. We discuss why, a priori, LMs might be expected to share similarities with the human language system. We then summarize evidence that LMs represent linguistic information similarly enough to humans to enable relatively accurate brain encoding and decoding during language processing. Finally, we examine which LM properties-their architecture, task performance, or training-are critical for capturing human neural responses to language and review studies using LMs as in silico model organisms for testing hypotheses about language. These ongoing investigations bring us closer to understanding the representations and processes that underlie our ability to comprehend sentences and express thoughts in language.

Cloze probability, predictability ratings, and computational estimates for 205 English sentences, aligned with existing EEG and reading time data

We release a database of cloze probability values, predictability ratings, and computational estimates for a sample of 205 English sentences (1726 words), aligned with previously released word-by-word reading time data (both self-paced reading and eye-movement records; Frank et al., Behavior Research Methods, 45(4), 1182-1190. 2013) and EEG responses (Frank et al., Brain and Language, 140, 1-11. 2015). Our analyses show that predictability ratings are the best predictors of the EEG signal (N400, P600, LAN) self-paced reading times, and eye movement patterns, when spillover effects are taken into account. The computational estimates are particularly effective at explaining variance in the eye-tracking data without spillover. Cloze probability estimates have decent overall psychometric accuracy and are the best predictors of early fixation patterns (first fixation duration). Our results indicate that the choice of the best measurement of word predictability in context critically depends on the processing index being considered.

Clinical efficacy of pre-trained large language models through the lens of aphasia

The rapid development of large language models (LLMs) motivates us to explore how such state-of-the-art natural language processing systems can inform aphasia research. What kind of language indices can we derive from a pre-trained LLM? How do they differ from or relate to the existing language features in aphasia? To what extent can LLMs serve as an interpretable and effective diagnostic and measurement tool in a clinical context? To investigate these questions, we constructed predictive and correlational models, which utilize mean surprisals from LLMs as predictor variables. Using AphasiaBank archived data, we validated our models' efficacy in aphasia diagnosis, measurement, and prediction. Our finding is that LLMs-surprisals can effectively detect the presence of aphasia and different natures of the disorder, LLMs in conjunction with the existing language indices improve models' efficacy in subtyping aphasia, and LLMs-surprisals can capture common agrammatic deficits at both word and sentence level. Overall, LLMs have potential to advance automatic and precise aphasia prediction. A natural language processing pipeline can be greatly benefitted from integrating LLMs, enabling us to refine models of existing language disorders, such as aphasia.

MCP: Multi-Chicken Pose Estimation Based on Transfer Learning

Poultry managers can better understand the state of poultry through poultry behavior analysis. As one of the key steps in behavior analysis, the accurate estimation of poultry posture is the focus of this research. This study mainly analyzes a top-down pose estimation method of multiple chickens. Therefore, we propose the "multi-chicken pose" (MCP), a pose estimation system for multiple chickens through deep learning. Firstly, we find the position of each chicken from the image via the chicken detector; then, an estimate of the pose of each chicken is made using a pose estimation network, which is based on transfer learning. On this basis, the pixel error (PE), root mean square error (RMSE), and image quantity distribution of key points are analyzed according to the improved chicken keypoint similarity (CKS). The experimental results show that the algorithm scores in different evaluation metrics are a mean average precision (mAP) of 0.652, a mean average recall (mAR) of 0.742, a percentage of correct keypoints (PCKs) of 0.789, and an RMSE of 17.30 pixels. To the best of our knowledge, this is the first time that transfer learning has been used for the pose estimation of multiple chickens as objects. The method can provide a new path for future poultry behavior analysis.

Dissociating language and thought in large language models

Large language models (LLMs) have come closest among all models to date to mastering human language, yet opinions about their linguistic and cognitive capabilities remain split. Here, we evaluate LLMs using a distinction between formal linguistic competence (knowledge of linguistic rules and patterns) and functional linguistic competence (understanding and using language in the world). We ground this distinction in human neuroscience, which has shown that formal and functional competence rely on different neural mechanisms. Although LLMs are surprisingly good at formal competence, their performance on functional competence tasks remains spotty and often requires specialized fine-tuning and/or coupling with external modules. We posit that models that use language in human-like ways would need to master both of these competence types, which, in turn, could require the emergence of separate mechanisms specialized for formal versus functional linguistic competence.

Category Locality Theory: A unified account of locality effects in sentence comprehension

In real-time sentence comprehension, the comprehender is often required to establish syntactic dependencies between words that are linearly distant. Major models of sentence comprehension assume that longer dependencies are more difficult to process because of working memory limitations. While the expected effect of distance on reading times (locality effect) has been robustly observed in certain constructions, such as relative clauses in English, its generalizability to a wider range of constructions has been empirically questioned. The current study proposes a new metric of syntactic distance that capitalizes on the flexible constituency of Combinatory Categorial Grammar (CCG), and argues that it offers a unified account of the locality effects. It is shown that this metric correctly predicts both the presence of the locality effect in English relative clauses and its absence in verb-final languages, without assuming language- or dependency-specific differences in the sensitivity to the locality effect. It is further shown that the CCG-based distance is a significant predictor of the self-paced reading times from an English corpus, even when other known predictors such as dependency-based locality and surprisal are taken into account. These results suggest that human sentence comprehension involves rapid integration of input words into efficiently compressed syntactic representations, and CCG is a plausible theory of the grammar that subserves this process.

Surprisal From Language Models Can Predict ERPs in Processing Predicate-Argument Structures Only if Enriched by an Agent Preference Principle

Language models based on artificial neural networks increasingly capture key aspects of how humans process sentences. Most notably, model-based surprisals predict event-related potentials such as N400 amplitudes during parsing. Assuming that these models represent realistic estimates of human linguistic experience, their success in modeling language processing raises the possibility that the human processing system relies on no other principles than the general architecture of language models and on sufficient linguistic input. Here, we test this hypothesis on N400 effects observed during the processing of verb-final sentences in German, Basque, and Hindi. By stacking Bayesian generalised additive models, we show that, in each language, N400 amplitudes and topographies in the region of the verb are best predicted when model-based surprisals are complemented by an Agent Preference principle that transiently interprets initial role-ambiguous noun phrases as agents, leading to reanalysis when this interpretation fails. Our findings demonstrate the need for this principle independently of usage frequencies and structural differences between languages. The principle has an unequal force, however. Compared to surprisal, its effect is weakest in German, stronger in Hindi, and still stronger in Basque. This gradient is correlated with the extent to which grammars allow unmarked NPs to be patients, a structural feature that boosts reanalysis effects. We conclude that language models gain more neurobiological plausibility by incorporating an Agent Preference. Conversely, theories of human processing profit from incorporating surprisal estimates in addition to principles like the Agent Preference, which arguably have distinct evolutionary roots.

Language prediction in monolingual and bilingual speakers: an EEG study

Prediction of upcoming words is thought to be crucial for language comprehension. Here, we are asking whether bilingualism entails changes to the electrophysiological substrates of prediction. Prior findings leave it open whether monolingual and bilingual speakers predict upcoming words to the same extent and in the same manner. We address this issue with a naturalistic approach, employing an information-theoretic metric, surprisal, to predict and contrast the N400 brain potential in monolingual and bilingual speakers. We recruited 18 Iranian Azeri-Persian bilingual speakers and 22 Persian monolingual speakers. Subjects listened to a story in Persian while their electroencephalogram (EEG) was recorded. Bayesian item-level analysis was used. While in monolingual speakers N400 was sensitive to information-theoretic properties of both the current and previous words, in bilingual speakers N400 reflected the properties of the previous word only. Our findings show evidence for a processing delay in bilingual speakers which is consistent with prior research.

Causality and signalling of garden-path sentences

Sheaves are mathematical objects that describe the globally compatible data associated with open sets of a topological space. Original examples of sheaves were continuous functions; later they also became powerful tools in algebraic geometry, as well as logic and set theory. More recently, sheaves have been applied to the theory of contextuality in quantum mechanics. Whenever the local data are not necessarily compatible, sheaves are replaced by the simpler setting of presheaves. In previous work, we used presheaves to model lexically ambiguous phrases in natural language and identified the order of their disambiguation. In the work presented here, we model syntactic ambiguities and study a phenomenon in human parsing called garden-pathing. It has been shown that the information-theoretic quantity known as 'surprisal' correlates with human reading times in natural language but fails to do so in garden-path sentences. We compute the degree of signalling in our presheaves using probabilities from the large language model BERT and evaluate predictions on two psycholinguistic datasets. Our degree of signalling outperforms surprisal in two ways: (i) it distinguishes between hard and easy garden-path sentences (with a [Formula: see text]-value [Formula: see text]), whereas existing work could not, (ii) its garden-path effect is larger in one of the datasets (32 ms versus 8.75 ms per word), leading to better prediction accuracies. This article is part of the theme issue 'Quantum contextuality, causality and freedom of choice'.

Word Frequency and Predictability Dissociate in Naturalistic Reading

Many studies of human language processing have shown that readers slow down at less frequent or less predictable words, but there is debate about whether frequency and predictability effects reflect separable cognitive phenomena: are cognitive operations that retrieve words from the mental lexicon based on sensory cues distinct from those that predict upcoming words based on context? Previous evidence for a frequency-predictability dissociation is mostly based on small samples (both for estimating predictability and frequency and for testing their effects on human behavior), artificial materials (e.g., isolated constructed sentences), and implausible modeling assumptions (discrete-time dynamics, linearity, additivity, constant variance, and invariance over time), which raises the question: do frequency and predictability dissociate in ordinary language comprehension, such as story reading? This study leverages recent progress in open data and computational modeling to address this question at scale. A large collection of naturalistic reading data (six datasets, >2.2 M datapoints) is analyzed using nonlinear continuous-time regression, and frequency and predictability are estimated using statistical language models trained on more data than is currently typical in psycholinguistics. Despite the use of naturalistic data, strong predictability estimates, and flexible regression models, results converge with earlier experimental studies in supporting dissociable and additive frequency and predictability effects.

Large-scale evidence for logarithmic effects of word predictability on reading time

During real-time language comprehension, our minds rapidly decode complex meanings from sequences of words. The difficulty of doing so is known to be related to words' contextual predictability, but what cognitive processes do these predictability effects reflect? In one view, predictability effects reflect facilitation due to anticipatory processing of words that are predictable from context. This view predicts a linear effect of predictability on processing demand. In another view, predictability effects reflect the costs of probabilistic inference over sentence interpretations. This view predicts either a logarithmic or a superlogarithmic effect of predictability on processing demand, depending on whether it assumes pressures toward a uniform distribution of information over time. The empirical record is currently mixed. Here, we revisit this question at scale: We analyze six reading datasets, estimate next-word probabilities with diverse statistical language models, and model reading times using recent advances in nonlinear regression. Results support a logarithmic effect of word predictability on processing difficulty, which favors probabilistic inference as a key component of human language processing.

Does entropy modulate the prediction of German long-distance verb particles?

In this paper we examine the effect of uncertainty on readers' predictions about meaning. In particular, we were interested in how uncertainty might influence the likelihood of committing to a specific sentence meaning. We conducted two event-related potential (ERP) experiments using particle verbs such as turn down and manipulated uncertainty by constraining the context such that readers could be either highly certain about the identity of a distant verb particle, such as turn the bed […] down, or less certain due to competing particles, such as turn the music […] up/down. The study was conducted in German, where verb particles appear clause-finally and may be separated from the verb by a large amount of material. We hypothesised that this separation would encourage readers to predict the particle, and that high certainty would make prediction of a specific particle more likely than lower certainty. If a specific particle was predicted, this would reflect a strong commitment to sentence meaning that should incur a higher processing cost if the prediction is wrong. If a specific particle was less likely to be predicted, commitment should be weaker and the processing cost of a wrong prediction lower. If true, this could suggest that uncertainty discourages predictions via an unacceptable cost-benefit ratio. However, given the clear predictions made by the literature, it was surprisingly unclear whether the uncertainty manipulation affected the two ERP components studied, the N400 and the PNP. Bayes factor analyses showed that evidence for our a priori hypothesised effect sizes was inconclusive, although there was decisive evidence against a priori hypothesised effect sizes larger than 1μV for the N400 and larger than 3μV for the PNP. We attribute the inconclusive finding to the properties of verb-particle dependencies that differ from the verb-noun dependencies in which the N400 and PNP are often studied.

Estimating the True Cost of Garden Pathing: A Computational Model of Latent Cognitive Processes

What is the processing cost of being garden-pathed by a temporary syntactic ambiguity? We argue that comparing average reading times in garden-path versus non-garden-path sentences is not enough to answer this question. Trial-level contaminants such as inattention, the fact that garden pathing may occur non-deterministically in the ambiguous condition, and "triage" (rejecting the sentence without reanalysis; Fodor & Inoue, 2000) lead to systematic underestimates of the true cost of garden pathing. Furthermore, the "pure" garden-path effect due to encountering an unexpected word needs to be separated from the additional cost of syntactic reanalysis. To get more realistic estimates for the individual processing costs of garden pathing and syntactic reanalysis, we implement a novel computational model that includes trial-level contaminants as probabilistically occurring latent cognitive processes. The model shows a good predictive fit to existing reading time and judgment data. Furthermore, the latent-process approach captures differences between noun phrase/zero complement (NP/Z) garden-path sentences and semantically biased reduced relative clause (RRC) garden-path sentences: The NP/Z garden path occurs nearly deterministically but can be mostly eliminated by adding a comma. By contrast, the RRC garden path occurs with a lower probability, but disambiguation via semantic plausibility is not always effective.