Publications

Abstract

Human language stands out in the natural world as a biological signal that uses a structured system to combine the meanings of small linguistic units (e.g., words) into larger constituents (e.g., phrases and sentences). However, the physical dynamics of speech (or sign) do not stand in a one-to-one relationship with the meanings listeners perceive. Instead, listeners infer meaning based on their knowledge of the language. The neural readouts of the perceptual and cognitive processes underlying these inferences are still poorly understood. In the present study, we used scalp electroencephalography (EEG) to compare the neural response to phrases (e.g., the red vase) and sentences (e.g., the vase is red), which were close in semantic meaning and had been synthesized to be physically indistinguishable. Differences in structure were well captured in the reorganization of neural phase responses in delta (approximately <2 Hz) and theta bands (approximately 2 to 7 Hz),and in power and power connectivity changes in the alpha band (approximately 7.5 to 13.5 Hz). Consistent with predictions from a computational model, sentences showed more power, more power connectivity, and more phase synchronization than phrases did. Theta–gamma phase–amplitude coupling occurred, but did not differ between the syntactic structures. Spectral–temporal response function (STRF) modeling revealed different encoding states for phrases and sentences, over and above the acoustically driven neural response. Our findings provide a comprehensive description of how the brain encodes and separates linguistic structures in the dynamics of neural responses. They imply that phase synchronization and strength of connectivity are readouts for the constituent structure of language. The results provide a novel basis for future neurophysiological research on linguistic structure representation in the brain, and, together with our simulations, support time-based binding as a mechanism of structure encoding in neural dynamics.

Abstract

It has long been recognised that phrases and sentences are organised hierarchically, but many computational models of language treat them as sequences of words without computing constituent structure. Against this background, we conducted two experiments which showed that participants interpret ambiguous noun phrases, such as second blue ball, in terms of their abstract hierarchical structure rather than their linear surface order. When a neural network model was tested on this task, it could simulate such “hierarchical” behaviour. However, when we changed the training data such that they were not entirely unambiguous anymore, the model stopped generalising in a human-like way. It did not systematically generalise to novel items, and when it was trained on ambiguous trials, it strongly favoured the linear interpretation. We argue that these models should be endowed with a bias to make generalisations over hierarchical structure in order to be cognitively adequate models of human language.

Abstract

Recent research has established that cortical activity “tracks” the presentation rate of syntactic phrases in continuous speech, even though phrases are abstract units that do not have direct correlates in the acoustic signal. We investigated whether cortical tracking of phrase structures is modulated by the extent to which these structures compositionally determine meaning. To this end, we recorded electroencephalography (EEG) of 38 native speakers who listened to naturally spoken Dutch stimuli in different conditions, which parametrically modulated the degree to which syntactic structure and lexical semantics determine sentence meaning. Tracking was quantified through mutual information between the EEG data and either the speech envelopes or abstract annotations of syntax, all of which were filtered in the frequency band corresponding to the presentation rate of phrases (1.1–2.1 Hz). Overall, these mutual information analyses showed stronger tracking of phrases in regular sentences than in stimuli whose lexical-syntactic content is reduced, but no consistent differences in tracking between sentences and stimuli that contain a combination of syntactic structure and lexical content. While there were no effects of compositional meaning on the degree of phrase-structure tracking, analyses of event-related potentials elicited by sentence-final words did reveal meaning-induced differences between conditions. Our findings suggest that cortical tracking of structure in sentences indexes the internal generation of this structure, a process that is modulated by the properties of its input, but not by the compositional interpretation of its output.

Abstract

People readily generalize knowledge to novel domains and stimuli. We present a theory, instantiated in a computational model, based on the idea that cross-domain generalization in humans is a case of analogical inference over structured (i.e., symbolic) relational representations. The model is an extension of the Learning and Inference with Schemas and Analogy (LISA; Hummel & Holyoak, 1997, 2003) and Discovery of Relations by Analogy (DORA; Doumas et al., 2008) models of relational inference and learning. The resulting model learns both the content and format (i.e., structure) of relational representations from nonrelational inputs without supervision, when augmented with the capacity for reinforcement learning it leverages these representations to learn about individual domains, and then generalizes to new domains on the first exposure (i.e., zero-shot learning) via analogical inference. We demonstrate the capacity of the model to learn structured relational representations from a variety of simple visual stimuli, and to perform cross-domain generalization between video games (Breakout and Pong) and between several psychological tasks. We demonstrate that the model’s trajectory closely mirrors the trajectory of children as they learn about relations, accounting for phenomena from the literature on the development of children’s reasoning and analogy making. The model’s ability to generalize between domains demonstrates the flexibility afforded by representing domains in terms of their underlying relational structure, rather than simply in terms of the statistical relations between their inputs and outputs.

Abstract

Linguistic phrases are tracked in sentences even though there is no one-to-one acoustic phrase marker in the physical signal. This phenomenon suggests an automatic tracking of abstract linguistic structure that is endogenously generated by the brain. However, all studies investigating linguistic tracking compare conditions where either relevant information at linguistic timescales is available, or where this information is absent altogether (e.g., sentences versus word lists during passive listening). It is therefore unclear whether tracking at phrasal timescales is related to the content of language, or rather, results as a consequence of attending to the timescales that happen to match behaviourally relevant information. To investigate this question, we presented participants with sentences and word lists while recording their brain activity with magnetoencephalography (MEG). Participants performed passive, syllable, word, and word-combination tasks corresponding to attending to four different rates: one they would naturally attend to, syllable-rates, word-rates, and phrasal-rates, respectively. We replicated overall findings of stronger phrasal-rate tracking measured with mutual information for sentences compared to word lists across the classical language network. However, in the inferior frontal gyrus (IFG) we found a task effect suggesting stronger phrasal-rate tracking during the word-combination task independent of the presence of linguistic structure, as well as stronger delta-band connectivity during this task. These results suggest that extracting linguistic information at phrasal rates occurs automatically with or without the presence of an additional task, but also that IFG might be important for temporal integration across various perceptual domains.

Abstract

Sentences contain structure that determines their meaning beyond that of individual words. An influential study by Ding and colleagues (2016) used frequency tagging of phrases and sentences to show that the human brain is sensitive to structure by finding peaks of neural power at the rate at which structures were presented. Since then, there has been a rich debate on how to best explain this pattern of results with profound impact on the language sciences. Models that use hierarchical structure building, as well as models based on associative sequence processing, can predict the neural response, creating an inferential impasse as to which class of models explains the nature of the linguistic computations reflected in the neural readout. In the current manuscript, we discuss pitfalls and common fallacies seen in the conclusions drawn in the literature illustrated by various simulations. We conclude that inferring the neural operations of sentence processing based on these neural data, and any like it, alone, is insufficient. We discuss how to best evaluate models and how to approach the modeling of neural readouts to sentence processing in a manner that remains faithful to cognitive, neural, and linguistic principles.

Abstract

In this study, we compared the performance of a long short-term memory (LSTM) neural network to the behavior of human participants on a language task that requires hierarchically structured knowledge. We show that humans interpret ambiguous noun phrases, such as second blue ball, in line with their hierarchical constituent structure. LSTMs, instead, only do
so after unambiguous training, and they do not systematically generalize to novel items. Overall, the results of our simulations indicate that a model can behave hierarchically without relying on hierarchical constituent structure.

Abstract

How a system represents information tightly constrains the kinds of problems it can solve. Humans routinely solve problems that appear to require abstract representations of stimulus properties and relations. How we acquire such representations has central importance in an account of human cognition. We briefly describe a theory of how a system can learn invariant responses to instances of similarity and relative magnitude, and how structured, relational representations can be learned from initially unstructured inputs. Two operations, comparing distributed representations and learning from the concomitant network dynamics in time, underpin the ability to learn these representations and to respond to invariance in the environment. Comparing analog representations of absolute magnitude produces invariant signals that carry information about similarity and relative magnitude. We describe how a system can then use this information to bootstrap learning structured (i.e., symbolic) concepts of relative magnitude from experience without assuming such representations a priori.

Abstract

Psychology endeavors to develop theories of human capacities and behaviors on the basis of a variety of methodologies and dependent measures. We argue that one of the most divisive factors in psychological science is whether researchers choose to use computational modeling of theories (over and above data) during the scientific-inference process. Modeling is undervalued yet holds promise for advancing psychological science. The inherent demands of computational modeling guide us toward better science by forcing us to conceptually analyze, specify, and formalize intuitions that otherwise remain unexamined—what we dub open theory. Constraining our inference process through modeling enables us to build explanatory and predictive theories. Here, we present scientific inference in psychology as a path function in which each step shapes the next. Computational modeling can constrain these steps, thus advancing scientific inference over and above the stewardship of experimental practice (e.g., preregistration). If psychology continues to eschew computational modeling, we predict more replicability crises and persistent failure at coherent theory building. This is because without formal modeling we lack open and transparent theorizing. We also explain how to formalize, specify, and implement a computational model, emphasizing that the advantages of modeling can be achieved by anyone with benefit to all.

Abstract

Whether neural networks can capture relational knowledge is a matter of long-standing controversy. Recently, some researchers have argued that (1) classic connectionist models can handle relational structure and (2) the success of deep learning approaches to natural language processing suggests that structured representations are unnecessary to model human language. We tested the Story Gestalt model, a classic connectionist model of text comprehension, and a Sequence-to-Sequence with Attention model, a modern deep learning architecture for natural language processing. Both models were trained to answer questions about stories based on abstract thematic roles. Two simulations varied the statistical structure of new stories while keeping their relational structure intact. The performance of each model fell below chance at least under one manipulation. We argue that both models fail our tests because they can't perform dynamic binding. These results cast doubts on the suitability of traditional neural networks for explaining relational reasoning and language processing phenomena.

Abstract

Neuronal oscillations putatively track speech in order to optimize sensory processing. However, it is unclear how isochronous brain oscillations can track pseudo-rhythmic speech input. Here we propose that oscillations can track pseudo-rhythmic speech when considering that speech time is dependent on content-based predictions flowing from internal language models. We show that temporal dynamics of speech are dependent on the predictability of words in a sentence. A computational model including oscillations, feedback, and inhibition is able to track pseudo-rhythmic speech input. As the model processes, it generates temporal phase codes, which are a candidate mechanism for carrying information forward in time. The model is optimally sensitive to the natural temporal speech dynamics and can explain empirical data on temporal speech illusions. Our results suggest that speech tracking does not have to rely only on the acoustics but could also exploit ongoing interactions between oscillations and constraints flowing from internal language models.

Abstract

Successful dependency resolution during language comprehension relies on accessing certain representations in memory, and not others. We recently reported event-related potential (ERP) evidence that syntactically unavailable, intervening attractor-nouns interfered during comprehension of Spanish noun-phrase ellipsis (the determiner otra/otro): grammatically correct determiners that mismatched the gender of attractor-nouns elicited a sustained negativity as also observed for incorrect determiners (Martin, Nieuwland, & Carreiras, 2012). The current study sought to extend this novel finding in sentences containing object-extracted relative clauses, where the antecedent may be less prominent. Whereas correct determiners that matched the gender of attractor-nouns now elicited an early anterior negativity as also observed for mismatching determiners, the previously reported interaction pattern was replicated in P600 responses to subsequent words. Our results suggest that structural and gender information is simultaneously taken into account, providing further evidence for retrieval interference during comprehension of grammatical sentences.