Search

Large Language Models (LLMs) have been criticized for failing to connect linguistic meaning to the world—for failing to solve the "symbol grounding problem." Multimodal Large Language Models (MLLMs) offer a potential solution to this challenge by combining linguistic representations and processing with other modalities. However, much is still unknown about exactly how and to what degree MLLMs integrate their distinct modalities—and whether the way they do so mirrors the mechanisms believed to underpin grounding in humans. In humans, it has been hypothesized that linguistic meaning is grounded through "embodied simulation," the activation of sensorimotor and affective representations reflecting described experiences. Across four pre-registered studies, we adapt experimental techniques originally developed to investigate embodied simulation in human comprehenders to ask whether MLLMs are sensitive to sensorimotor features that are implied but not explicit in descriptions of an event. In Experiment 1, we find sensitivity to some features (color and shape) but not others (size, orientation, and volume). In Experiment 2, we identify likely bottlenecks to explain an MLLM's lack of sensitivity. In Experiment 3, we find that despite sensitivity to implicit sensorimotor features, MLLMs cannot fully account for human behavior on the same task. Finally, in Experiment 4, we compare the psychometric predictive power of different MLLM architectures and find that ViLT, a single-stream architecture, is more predictive of human responses to one sensorimotor feature (shape) than CLIP, a dual-encoder architecture—despite being trained on orders of magnitude less data. These results reveal strengths and limitations in the ability of current MLLMs to integrate language with other modalities, and also shed light on the likely mechanisms underlying human language comprehension. [ABSTRACT FROM AUTHOR]

Transformer language models have received widespread public attention, yet their generated text is often surprising even to NLP researchers. In this survey, we discuss over 250 recent studies of English language model behavior before task-specific fine-tuning. Language models possess basic capabilities in syntax, semantics, pragmatics, world knowledge, and reasoning, but these capabilities are sensitive to specific inputs and surface features. Despite dramatic increases in generated text quality as models scale to hundreds of billions of parameters, the models are still prone to unfactual responses, commonsense errors, memorized text, and social biases. Many of these weaknesses can be framed as over-generalizations or under-generalizations of learned patterns in text. We synthesize recent results to highlight what is currently known about large language model capabilities, thus providing a resource for applied work and for research in adjacent fields that use language models. [ABSTRACT FROM AUTHOR]

Most words have multiple meanings, but there are foundationally distinct accounts for this. Categorical theories posit that humans maintain discrete entries for distinct word meanings, as in a dictionary. Continuous ones eschew discrete sense representations, arguing that word meanings are best characterized as trajectories through a continuous state space. Both kinds of approach face empirical challenges. In response, we introduce two novel "hybrid" theories, which reconcile discrete sense representations with a continuous view of word meaning. We then report on two behavioral experiments, pairing them with an analytical approach relying on neural language models to test these competing accounts. The experimental results are best explained by one of the novel hybrid accounts, which posits both distinct sense representations and a continuous meaning space. This hybrid account accommodates both the dynamic, context-dependent nature of word meaning, as well as the behavioral evidence for category-like structure in human lexical knowledge. We further develop and quantify the predictive power of several computational implementations of this hybrid account. These results raise questions for future research on lexical ambiguity, such as why and when discrete sense representations might emerge in the first place. They also connect to more general questions about the role of discrete versus gradient representations in cognitive processes and suggest that at least in this case, the best explanation is one that integrates both factors: Word meaning is both categorical and continuous. [ABSTRACT FROM AUTHOR]

Speakers enjoy considerable flexibility in how they refer to a given referent––referring expressions can vary in their form (e.g., "she" vs. "the cat"), their length (e.g., "the (big) (orange) cat"), and more. What factors drive a speaker's decisions about how to refer, and how do these decisions shape a comprehender's ability to resolve the intended referent? Answering either question presents a methodological challenge; researchers must strike a balance between experimental control and ecological validity. In this paper, we introduce the SCARFS (Spontaneous, Controlled Acts of Reference between Friends and Strangers) Database: a corpus of approximately 20,000 English nominal referring expressions (NREs), produced in the context of a communication game. For each NRE, the corpus lists the concept the speaker was trying to convey (from a set of 471 possible target concepts), formal properties of the NRE (e.g., its length), the relationship between the interlocutors (i.e., friend vs. stranger), and the communicative outcome (i.e., whether the expression was successfully resolved). Researchers from diverse disciplines may use this resource to answer questions about how speakers refer and how comprehenders resolve their intended referent––as well as other fundamental questions about dialogic speech, such as whether and how speakers tailor their utterances to the identity of their interlocutor, how second-degree associations are generated, and the predictors of communicative success. [ABSTRACT FROM AUTHOR]

People often speak indirectly. For example, "It's cold in here" might be intended not only as a comment on the temperature but also as a request to turn on the heater. How are comprehenders' inferences about a speaker's intentions informed by their ability to reason about the speaker's mental states, that is, mentalizing? We introduce a mechanistic framework by which mentalizing might be recruited for pragmatic inference and then ask the following: Is mentalizing recruited primarily for sampling mental state information or also for the deployment of that information for pragmatic inference? We find that the role of mentalizing is modulated by how explicitly a task involves knowledge. Mentalizing correlates with task performance when comprehenders are asked to sample and report mental state information (Experiment 1b) or when given mental state information explicitly and asked to make an inference (Experiment 2, Explicit); in contrast, mentalizing does not correlate with task performance, or correlates only weakly, when participants are given mental state information implicitly and asked to make a pragmatic inference (Experiment 1a and Experiment 2, Implicit). These results suggest that mentalizing is recruited flexibly, allowing comprehenders to construct meaning from underspecified input. [ABSTRACT FROM AUTHOR]

People often speak ambiguously, as in the case of indirect requests. Certain indirect requests are conventional and thus straightforward to interpret, such as "Can you turn on the heater?", but others require substantial additional inference, such as "It's cold in here." How do comprehenders make inferences about a speaker's intentions? And what makes a comprehender more or less successful? Here, we explore the hypothesis that comprehenders do so in part by mentalizing—encoding what the speaker knows or believes about the world—and that differences in mentalizing ability predict how well comprehenders draw these inferences. In Experiment 1, we find that comprehenders' pragmatic interpretations are significantly influenced by a speaker's inferable knowledge states. In Experiment 2, we find that variability in this effect is explained by individual differences in comprehenders' mentalizing ability. Finally, in Experiment 3, we find that both effects are robust across different dependent measures of inference. [ABSTRACT FROM AUTHOR]

What concepts and words do communicative gestures activate in the minds of people who view them? It’s widely believed that many gestures grow from iconic origins—they look like what they mean—but also that at some point they may become emblematic—conventionalized as culturally agreed-upon symbols. How long do links between physical movements of the body and the things in the world they denote persist in the minds of gesture-users? A pair of experiments asks this question for the Middle-Finger, a cross-culturally recognized obscene gesture. The prevailing view is that the gesture originates in a phallic symbol. Yet it is now predominantly used as an emblematic gesture displaying contempt (among other things). It is currently unknown whether the iconic origins of gestures persist through the emblematic stage in the minds of gesture users. Two experiments tested the hypothesis that viewing the Middle-Finger primes thoughts about penises or the word penis. The results showed that the Middle-Finger induced no priming of penis compared with control, unlike another obscene penis-representing gesture (Finger-Bang), which did. This suggests that the Middle-Finger no longer activates thoughts of penises in the minds of contemporary American English speakers. Emblematic gestures with iconic origins may undergo historical change not just in the functions they serve but also in the effects they have on the minds of people who use them. [ABSTRACT FROM AUTHOR]

Languages around the world use a recurring strategy to discuss abstract concepts: describe them metaphorically, borrowing language from more concrete domains. We "plan ahead" to the future, "count up" to higher numbers, and "warm" to new friends. Past work has found that these ways of talking have implications for how we think, so that shared systems of linguistic metaphors can produce shared conceptualizations. On the other hand, these systematic linguistic metaphors might not just be the cause but also the effect of shared, non‐linguistic ways of thinking. Here, we present a case study of a variety of American English in which a shared, non‐linguistic conceptualization of time has become crystallized as a new system of linguistic metaphors. Speakers of various languages, including English, conceptualize time as a lateral timeline, with the past leftward and the future rightward. Until now, this conceptualization has not been documented in the speech of any language. In two studies, we document how members of the U.S. military, but not U.S. civilians, talk about time using conventionalized lateral metaphors (e.g., "move the meeting right" to mean "move the meeting later"). We argue that, under the right cultural circumstances, implicit mental representations become conventionalized metaphors in language. [ABSTRACT FROM AUTHOR]

People use space in a variety of ways to structure their thoughts about time. The present report focuses on the different ways that space is employed when reasoning about deictic (past/future relationships) and sequence (earlier/later relationships) time. In the first study, we show that deictic and sequence time are aligned along the lateral axis in a manner consistent with previous work, with past and earlier events associated with left space and future and later events associated with right space. However, the alignment of time with space is different along the sagittal axis. Participants associated future events and earlier events—not later events—with the space in front of their body and past and later events with the space behind, consistent with the sagittal spatial terms (e.g., ahead, in front of) that we use to talk about deictic and sequence time. In the second study, we show that these associations between sequence time and sagittal space are sensitive to person-perspective. This suggests that the particular space-time associations observed in English speakers are influenced by a variety of different spatial properties, including spatial location and perspective. [ABSTRACT FROM AUTHOR]

In contrast to clear stimulatory effects of rising temperature, recent studies of the effects of CO2 on planktonic bacteria have reported conflicting results. To better understand the potential impact of predicted climate scenarios on the development and performance of bacterial communities, we performed bifactorial mesocosm experiments (pCO2 and temperature) with Baltic Sea water, during a diatom dominated bloom in autumn and a mixed phytoplankton bloom in summer. The development of bacterial community composition (BCC) followed well-known algal bloom dynamics. A principal coordinate analysis (PCoA) of bacterial OTUs (operational taxonomic units) revealed that phytoplankton succession and temperature were the major variables structuring the bacterial community whereas the impact of pCO2 was weak. Prokaryotic abundance and carbon production, and organic matter concentration and composition were partly affected by temperature but not by increased pCO2. However, pCO2 did have significant and potentially direct effects on the relative abundance of several dominant OTUs; in some cases, these effects were accompanied by an antagonistic impact of temperature. Our results suggest the necessity of high-resolution BCC analyses and statistical analyses at the OTU level to detect the strong impact of CO2 on specific bacterial groups, which in turn might also influence specific organic matter degradation processes. [ABSTRACT FROM AUTHOR]

Embodied approaches to comprehension propose that understanding language entails performing mental simulations of its content. The evidence, however, is mixed. Action-sentence Compatibility Effect studies (Glenberg and Kaschak 2002) report mental simulation of motor actions during processing of motion language. But the same studies find no evidence that language comprehenders perform spatial simulations of the corresponding locations. This challenges simulation- based approaches. If locations are not represented in simulation, but are still understood, then simulation may be unnecessary for understanding. We conducted a Location-sentence Compatibility experiment, to determine whether understanders mentally simulate locations. People did indeed simulate locations, but only when sentences used progressive (and not perfect) grammatical aspect. Moreover, mental simulations of locations differed for language about concrete versus abstract events. These findings substantiate the role of mental simulation in language understanding, while highlighting the importance of the grammatical form of utterances as well as their content. [ABSTRACT FROM AUTHOR]

Many animals can be trained to perform novel tasks. People, too, can be trained, but sometime in early childhood people transition from being trainable to something qualitatively more powerful-being programmable. We argue that such programmability constitutes a leap in the way that organisms learn, interact, and transmit knowledge, and that what facilitates or enables this programmability is the learning and use of language. We then examine how language programs the mind and argue that it does so through the manipulation of embodied, sensorimotor representations. The role language plays in controlling mental representations offers important insights for understanding its origin and evolution. [ABSTRACT FROM AUTHOR]

Human languages evolve to make communication more efficient. But efficiency creates trade-offs: what is efficient for speakers is not always efficient for comprehenders. How do languages balance these competing pressures? We focus on Zipf's meaning-frequency law, the observation that frequent wordforms have more meanings. On the one hand, this law could reflect a speaker-oriented pressure to reuse frequent wordforms. Yet human languages still maintain thousands of distinct wordforms, suggesting a countervailing, comprehender-oriented pressure. What balance of these pressures produces Zipf's meaning-frequency law? Using a neutral baseline, we find that frequent wordforms in real lexica have fewer homophones than predicted by their phonotactic structure: real lexica favor a comprehender-oriented pressure to reduce the cost of frequent disambiguation. These results help clarify the evolutionary drive for efficiency: human languages are subject to competing pressures for efficient communication, the relative magnitudes of which reveal how individual-level cognitive constraints shape languages over time. [ABSTRACT FROM AUTHOR]

A recent pyrosequencing study along the whole Baltic Sea salinity transect identified members of the V errucomicrobia class S partobacteria as an important component of Baltic Sea bacterioplankton. In this study, catalysed reporter deposition-fluorescence in situ hybridization was used for cellular quantification. The published probes VER47 and SPA714 were optimized for samples from the Baltic Sea and a new, specific probe ( SPA476) was used to quantify the dominant spartobacterial lineage ' LD29'. The results confirmed that in the brackish surface waters of the Baltic Sea S partobacteria comprise an important component, constituting up to 12% of all bacteria. The positive correlation and physical association of S partobacteria with phytoplankton suggest their involvement in the utilization of phytoplankton-derived organic matter in the Baltic Sea. [ABSTRACT FROM AUTHOR]

While we often talk about time using spatial terms, experimental investigation of space-time associations has focused primarily on the space in front of the participant. This has had two consequences: the disregard of the space behind the participant (exploited in language and gesture) and the creation of potential task demands produced by spatialized manual button-presses. We introduce and test a new paradigm that uses auditory stimuli and vocal responses to address these issues. Participants made temporal judgments about deictic or sequential relationships presented auditorily along a body-centered sagittal or transversal axis. Results involving the transversal axis replicated previous work while sagittal axis results were surprising. Deictic judgments did not use the sagittal axis but sequential judgments did, in a previously undocumented way. Participants associated earlier judgments with the space in front of them and later judgments with the space behind them. These findings, using a new approach, provide evidence that different time concepts recruit space differently, mediated by meaning, stimulus modality and response mode. [ABSTRACT FROM AUTHOR]

Abstract: Language comprehenders can mentally simulate perceptual and motor features of scenes they hear or read about (Barsalou, 1999; Glenberg & Kaschak, 2002; Zwaan, Stanfield, & Yaxley, 2002). Recent research shows that these simulations adopt a particular perspective (Borghi, Glenberg, & Kaschak, 2004; Brunyé, Ditman, Mahoney, Augustyn, & Taylor, 2009). Moreover, features of utterances influence the perspective that comprehenders are led to adopt. For instance, language about you primes a participant visual perspective, while third person he and she prime an observer perspective. But what role does perspectival mental simulation play in the comprehension of person? On the one hand, the different perspectives adopted during language understanding could be necessary for successfully determining the meaning of an utterance. However, current empirical evidence is also compatible with the possibility that adopting a perspective in mental simulation is not essential to comprehending who did what to whom. If the latter is the case, then we should be able to find cases where language comprehenders understand who did what to whom without measurably performing mental simulation from a particular perspective. A candidate language that might display such a case is Japanese, where grammatical subject pronouns can be omitted when the subject is inferable from context. We replicated a previously used method for assessing perspectival mental simulation during language comprehension, but tailored it to Japanese. The results showed that when pronouns were present, like in English, sentences facilitated identification of an image matching the proposed perspective associated with the mentioned pronoun. This replicated the previous finding for English. But when the subject pronoun was omitted, so that the sentence did not explicitly mention the subject, there was no such effect. Nonetheless, native comprehenders of Japanese automatically and easily tracked who the subjects of the sentences with omitted subjects were. Together, these findings suggest that while grammatical person modulates visual perspective in mental simulation, visual perspective is not necessary for successful identification and representation of event participants. [Copyright &y& Elsevier]

Performing two cognitive tasks at the same time can degrade performance for either domain-general reasons (e.g., both tasks require attention) or domain-specific reasons (e.g., both tasks require visual working memory). We tested predictions of these two accounts of interference on the task of driving while using language, a naturally occurring dual task. Using language and driving a vehicle use different perceptual and motor skills. As a consequence, a domain-general explanation for interference in this dual task appears most plausible. However, recent evidence from the language processing literature suggests that when people use language with motor content (e.g., language about actions) or visual content (e.g., language about visible objects and events), they engage their motor and perceptual systems in ways specifically reflecting the actions and percepts that the language is about. This raises the possibility that language might interfere with driving for domain-specific reasons when the language has visual or motor content. To test this, we had participants drive a simulated vehicle while simultaneously answering true-false statements that had motor, visual, or abstract content. A domain-general explanation for interference would predict greater distraction in each of these three conditions compared with control, while a domain-specific explanation would predict greater interference in the motor and visual conditions. Both of these predictions were borne out but on different measures of distraction, suggesting that language-driven distraction during driving and dual tasks involving language in general may be the result not only of domain-general causes but also specific interference caused by linguistic content. [ABSTRACT FROM AUTHOR]

Language comprehenders activate mental representations of sensorimotor experiences related to the content of utterances they process. However, it is still unclear whether these sensorimotor simulations are driven by associations with words or by a more complex process of meaning composition into larger linguistic expressions, such as sentences. In two experiments, we investigated whether comprehenders indeed create sentence-based simulations. Materials were constructed such that simulation effects could only emerge from sentence meaning and not from word-based associations alone. We additionally asked when during sentence processing these simulations are constructed, using a garden-path paradigm. Participants read either a garden-path sentence (e.g., "As Mary ate the egg was in the fridge") or a corresponding unambiguous control with the same meaning and words (e.g., "The egg was in the fridge as Mary ate"). Participants then judged whether a depicted entity was mentioned in the sentence or not. In both experiments, picture response times were faster when the picture was compatible (vs. incompatible) with the sentence-based interpretation of the target entity (e.g., both for garden-path and control sentence: an unpeeled egg), suggesting that participants created simulations based on the sentence content and only operating over the sentence as a whole. [ABSTRACT FROM AUTHOR]

Grammatical aspect is a pervasive linguistic device that, according to linguistic analyses, allows speakers to encode different ways of construing events. For instance, the progressive ( I was writing a book) is thought to reflect increased focus on the internal details of an event, as contrasted with the perfect ( I had written a book). However, experimental evidence that speakers describing events using progressive versus non-progressive aspect are in fact thinking about the same events differently is lacking. We used co-speech gesture as a means to investigate what speakers' event representations are like when they produce progressive versus non-progressive utterances. We found that progressive event descriptions were accompanied by longer-lasting and more complex gestures, but only when participants described events originally presented to them in the progressive. This evidence suggests that people are actually construing events differently when they use different grammatical aspects, but that the aspect originally used to encode the events plays a role as well. [ABSTRACT FROM AUTHOR]

Abstract: When processing sentences about perceptible scenes and performable actions, language understanders activate perceptual and motor systems to perform mental simulations of those events. But little is known about exactly what linguistic elements activate modality-specific systems during language processing. While it is known that content words, like nouns and verbs, influence the content of a mental simulation, the role of grammar is less well understood. We investigate the role of grammatical markers in mental simulation through two experiments in which we manipulate the meanings of sentences by modifying the grammatical aspect they use. Using the Action-sentence Compatibility Effect (ACE) methodology [Glenberg, A., Kaschak, M. (2002). Grounding language in action. Psychonomic Bulletin and Review, 9, 558–565], we show that progressive sentences about hand motion facilitate manual action in the same direction, while perfect sentences that are identical in every way except their aspect do not. The broader implication of this finding for language processing is that while content words tell understanders what to mentally simulate and what brain regions to use in performing these simulations, grammatical constructions such as aspect modulate how those simulations are performed. [Copyright &y& Elsevier]

There is mounting evidence that language comprehension involves the activation of mental imagery of the content of utterances (Barsalou, 1999; Bergen, Chang, & Narayan, 2004; Bergen, Narayan, & Feldman, 2003; Narayan, Bergen, & Weinberg, 2004; Richardson, Spivey, McRae, & Barsalou, 2003; Stanfield & Zwaan, 2001; Zwaan, Stanfield, & Yaxley, 2002). This imagery can have motor or perceptual content. Three main questions about the process remain under-explored, however. First, are lexical associations with perception or motion sufficient to yield mental simulation, or is the integration of lexical semantics into larger structures, like sentences, necessary? Second, what linguistic elements (e.g., verbs, nouns, etc.) trigger mental simulations? Third, how detailed are the visual simulations that are performed? A series of behavioral experiments address these questions, using a visual object categorization task to investigate whether up- or down-related language selectively interferes with visual processing in the same part of the visual field (following Richardson et al., 2003). The results demonstrate that either subject nouns or main verbs can trigger visual imagery, but only when used in literal sentences about real space—metaphorical language does not yield significant effects—which implies that it is the comprehension of the sentence as a whole and not simply lexical associations that yields imagery effects. These studies also show that the evoked imagery contains detail as to the part of the visual field where the described scene would take place. [ABSTRACT FROM AUTHOR]

English deictic and existential there -constructions have been analyzed as constituting a single radial category of form-meaning pairings, related through motivated links, such as metaphor (Lakoff 1987). By comparison, existentials and deictic demonstratives in French make use of two distinct radial categories. The current study analyzes the varied senses of French deictic demonstratives ( voilà ‘there is’ and voici ‘here is’) and the existential ( il y a ‘there is’). We argue that the syntactic behavior of each of their senses is best explained by the semantic and pragmatic function of that sense, in combination with constraints imposed by its relation to other senses. A cross-linguistic comparison of the deictic demonstrative and existential constructions in French and English supports this claim: despite the different historical origins of these forms in the two languages, they display a strikingly similar array of uses and formal constraints. The parallel evolution of deictics and existentials in these two languages is interpreted as a case of convergent evolution of linguistic items, much like convergent evolution in biological species. [ABSTRACT FROM AUTHOR]

The psychological reality of English phonaesthemes is demonstrated through a priming experiment with native speakers of American English. Phonaesthemes are well-represented sound-meaning pairings, such as English gl-, which occurs in numerous words with meanings relating to light and vision. In the experiment, phonaesthemes, despite being noncompositional in nature, displayed priming effects much like those that have been reported for compositional morphemes. These effects could not be explained as the result of semantic or phonological priming, either alone or in combination. The results support a view of the lexicon in which shared form and meaning across words is a key factor in their relatedness, and in which morphological composition is not required for internal word structure to play a role in language processing. [ABSTRACT FROM AUTHOR]

For many scientific and engineering applications, it is often desirable to use unstructured grids to represent complex geometries. Unfortunately, the data structures required to represent discretizations on such grids typically result in extremely inefficient performance on current high-performance architectures. Here, we introduce a grid framework using patch-wise, regular refinement that retains the flexibility of unstructured grids, while achieving performance comparable to that seen with purely structured grids. This approach leads to a grid hierarchy suitable for use with geometric multigrid methods, thus combining asymptotically optimal algorithms with extremely efficient data structures to obtain a powerful technique for large scale simulations. Copyright © 2004 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Human languages are replete with ambiguity. This is most evident in homophony--where two or more words sound the same, but carry distinct meanings. For example, the wordform "bark" can denote either the sound produced by a dog or the protective outer sheath of a tree trunk. Why would a system evolved for efficient, effective communication display rampant ambiguity? Some accounts argue that ambiguity is actually a design feature of human communication systems, allowing languages to recycle their most optimal wordforms (those which are short, frequent, and phonotactically well-formed) for multiple meanings. We test this claim by constructing five series of artificial lexica matched for the phonotactics and distribution of word lengths found in five real languages (English, German, Dutch, French, and Japanese), and comparing both the quantity and concentration of homophony across the real and artificial lexica. Surprisingly, we find that the artificial lexica exhibit higher upper-bounds on homophony than their real counterparts, and that homophony is even more likely to be found among short, phonotactically plausible wordforms in the artificial than in the real lexica. These results suggest that homophony in real languages is not directly selected for, but rather, that it emerges as a natural consequence of other features of a language. In fact, homophony may even be selected against in real languages, producing lexica that better conform to other requirements of humans who need to use them. Finally, we explore the hypothesis that this is achieved by "smoothing" out dense concentrations of homophones across lexical neighborhoods, resulting in comparatively more minimal pairs in real lexica. [ABSTRACT FROM AUTHOR]

Although identifying the referents of single words is often cited as a key challenge for getting word learning off the ground, it overlooks the fact that young learners consistently encounter words in the context of other words. How does this company help or hinder word learning? Prior investigations into early word learning from children's real‐world language input have yielded conflicting results, with some influential findings suggesting an advantage for words that keep a diverse company of other words, and others suggesting the opposite. Here, we sought to triangulate the source of this conflict, comparing different measures of diversity and approaches to controlling for correlated effects of word frequency across multiple languages. The results were striking: while different diversity measures on their own yielded conflicting results, once nonlinear relationships with word frequency were controlled, we found convergent evidence that contextual consistency supports early word learning. Research Highlights: The words children learn occur in a sea of other words. The company words keep ranges from highly variable to highly consistent and circumscribed.Prior findings conflict over whether variability versus consistency helps early word learning.Accounting for correlated effects of word frequency resolved the conflict across multiple languages.Results reveal convergent evidence that consistency helps early word learning. [ABSTRACT FROM AUTHOR]

Humans can attribute beliefs to others. However, it is unknown to what extent this ability results from an innate biological endowment or from experience accrued through child development, particularly exposure to language describing others' mental states. We test the viability of the language exposure hypothesis by assessing whether models exposed to large quantities of human language display sensitivity to the implied knowledge states of characters in written passages. In pre‐registered analyses, we present a linguistic version of the False Belief Task to both human participants and a large language model, GPT‐3. Both are sensitive to others' beliefs, but while the language model significantly exceeds chance behavior, it does not perform as well as the humans nor does it explain the full extent of their behavior—despite being exposed to more language than a human would in a lifetime. This suggests that while statistical learning from language exposure may in part explain how humans develop the ability to reason about the mental states of others, other mechanisms are also responsible. [ABSTRACT FROM AUTHOR]

Ambiguity pervades language. The sentence "My office is really hot" could be interpreted as a complaint about the temperature or as an indirect request to turn on the air conditioning. How do comprehenders determine a speaker's intended interpretation? One possibility is that speakers and comprehenders exploit prosody to overcome the pragmatic ambiguity inherent in indirect requests. In a pre-registered behavioral experiment, we find that human listeners can successfully determine whether a given utterance was intended as a request at a rate above chance (55%), above and beyond the prior probability of a given sentence being interpreted as a request. Moreover, we find that a classifier equipped with seven acoustic features can detect the original intent of an utterance with 65% accuracy. Finally, consistent with past work, the duration, pitch, and pitch slope of an utterance emerge both as significant correlates of a speaker's original intent and as predictors of comprehenders' pragmatic interpretation. These results suggest that human and machine comprehenders alike can use prosody to enrich the meaning of ambiguous utterances, such as indirect requests. [ABSTRACT FROM AUTHOR]

People speak metaphorically about abstract concepts—for instance, a person can be “full of love” or “have a lot of love to give.” Over the past decade, research has begun to focus on how metaphors are processed during language comprehension. Much of this work suggests that understanding a metaphorical expression involves activating brain and body systems involved in perception and motor control. However, no research to date has asked whether the same is true while speakers produce language. We address this gap using a sentence production task. Its results demonstrate that visually activating a concrete source domain can trigger the use of metaphorical language drawn from that same concrete domain, even in sentences that are thematically unrelated to the primes, a metaphorical priming effect . This effect suggests that conceptual metaphors play a part in language production. It also shows that activation in the perceptual system that is not part of an intended message can nevertheless influence sentence formulation. [ABSTRACT FROM AUTHOR]

Mathematics requires precise inferences about abstract objects inaccessible to perception. How is this possible? One proposal is that mathematical reasoning, while concerned with entirely abstract objects, nevertheless relies on neural resources specialized for interacting with the world—in other words, mathematics may be grounded in spatial or sensorimotor systems. Mental arithmetic, for instance, could involve shifts in spatial attention along a mental “number-line”, the product of cultural artefacts and practices that systematically spatialize number and arithmetic. Here, we investigate this hypothesized spatial processing during exact, symbolic arithmetic (e.g., 4 + 3 = 7). Participants added and subtracted single-digit numbers and selected the exact solution from responses in the top corners of a computer monitor. While they made their selections using a computer mouse, we recorded the movement of their hand as indexed by the streamingx,ycoordinates of the computer mouse cursor. As predicted, hand movements during addition and subtraction were systematically deflected toward the right and the left, respectively, as if calculation involved simultaneously simulating motion along a left-to-right mental number-line. This spatial–arithmetical bias, moreover, was distinct from—but correlated with—individuals' spatial–numerical biases (i.e., spatial–numerical association of response codes, SNARC, effect). These results are the first evidence that exact, symbolic arithmetic prompts systematic spatial processing associated with mental calculation. We discuss the possibility that mathematical calculation relies, in part, on an integrated system of spatial processes. [ABSTRACT FROM AUTHOR]

The Action-sentence Compatibility Effect (ACE) is a well-known demonstration of the role of motor activity in the comprehension of language. Participants are asked to make sensibility judgments on sentences by producing movements toward the body or away from the body. The ACE is the finding that movements are faster when the direction of the movement (e.g., toward) matches the direction of the action in the to-be-judged sentence (e.g., Art gave you the pen describes action toward you). We report on a pre-registered, multi-lab replication of one version of the ACE. The results show that none of the 18 labs involved in the study observed a reliable ACE, and that the meta-analytic estimate of the size of the ACE was essentially zero. [ABSTRACT FROM AUTHOR]