Children have difficulty comprehending novel verbs in the double object dative as compared to the prepositional dative. We explored this pattern with 3 and 4 year olds. In Experiment 1, we replicated the documented difficulty with the double object frame, even though we provided more contextual support. In Experiment 2, we tested a novel hypothesis that children would comprehend novel verbs in, and generalize them to, the double object frame if they were first familiarized to the verbs in the prepositional (...) frame. They did, suggesting that part of their difficulty with the double object frame is due to uncertainty about a new verb's semantic/syntactic properties, information that the easy-to-comprehend prepositional frame provides. The benefits of training were short-lived, however; children again struggled after a 2-h delay. The results are discussed in the context of mechanisms underlying verb acquisition. (shrink)

In this paper, we describe a model that learns semantic representations from the distributional statistics of language. This model, however, goes beyond the common bag‐of‐words paradigm, and infers semantic representations by taking into account the inherent sequential nature of linguistic data. The model we describe, which we refer to as a Hidden Markov Topics model, is a natural extension of the current state of the art in Bayesian bag‐of‐words models, that is, the Topics model of Griffiths, Steyvers, and Tenenbaum (2007), (...) preserving its strengths while extending its scope to incorporate more fine‐grained linguistic information. (shrink)

Previous research on lexical development has aimed to identify the factors that enable accurate initial word-referent mappings based on the assumption that the accuracy of initial word-referent associations is critical for word learning. The present study challenges this assumption. Adult English speakers learned an artificial language within a cross-situational learning paradigm. Visual fixation data were used to assess the direction of visual attention. Participants whose longest fixations in the initial trials fell more often on distracter images performed significantly better at (...) test than participants whose longest fixations fell more often on referent images. Thus, inaccurate initial word-referent mappings may actually benefit learning. (shrink)

Recent laboratory experiments have shown that both infant and adult learners can acquire word‐referent mappings using cross‐situational statistics. The vast majority of the work on this topic has used unfamiliar objects presented on neutral backgrounds as the visual contexts for word learning. However, these laboratory contexts are much different than the real‐world contexts in which learning occurs. Thus, the feasibility of generalizing cross‐situational learning beyond the laboratory is in question. Adapting the Human Simulation Paradigm, we conducted a series of experiments (...) examining cross‐situational learning from children's egocentric videos captured during naturalistic play. Focusing on individually ambiguous naming moments that naturally occur during toy play, we asked how statistical learning unfolds in real time through accumulating cross‐situational statistics in naturalistic contexts. We found that even when learning situations were individually ambiguous, learners’ performance gradually improved over time. This improvement was driven in part by learners’ use of partial knowledge acquired from previous learning situations, even when they had not yet discovered correct word‐object mappings. These results suggest that word learning is a continuous process by means of real‐time information integration. (shrink)

Yang returns to the problem of referential ambiguity, addressed in the opening paper by Gleitman and Trueswell. Using a computational approach, he argues that “big data” approaches to resolving referential ambiguity are destined to fail, because of the inevitable computational explosion needed to keep track of contextual associations present when a word is uttered. Yang tests several computational models, two of which depend on one‐trial learning, as described in Gleitman and Trueswell’s paper. He concludes that such models outperform cross‐situational learning (...) models, thereby formalizing, reinforcing, and advancing the conclusions offered by Gleitman and Trueswell. (shrink)

Propositional attitude verbs, such as think and want, have long held interest for both theoretical linguists and language acquisitionists because their syntactic, semantic, and pragmatic properties display complex interactions that have proven difficult to fully capture from either perspective. This paper explores the granularity with which these verbs’ semantic and pragmatic properties are recoverable from their syntactic distributions, using three behavioral experiments aimed at explicitly quantifying the relationship between these two sets of properties. Experiment 1 gathers a measure of 30 (...) propositional attitude verbs’ syntactic distributions using an acceptability judgment task. Experiments 2a and 2b gather measures of semantic similarity between those same verbs using a generalized semantic discrimination task and an ordinal scale task, respectively. Two kinds of analyses are conducted on the data from these experiments. The first compares both the acceptability judgments and the semantic similarity judgments to previous classifications derived from the syntax and semantics literature. The second kind compares the acceptability judgments to the semantic similarity judgments directly. Through these comparisons, we show that there is quite fine-grained information about propositional attitude verbs’ semantics carried in their syntactic distributions—whether one considers the sorts of discrete qualitative classifications that linguists traditionally work with or the sorts of continuous quantitative classifications that can be derived experimentally. (shrink)

Cross-situational learning has recently gained attention as a plausible candidate for the mechanism that underlies the learning of word-meaning mappings. In a recent study, Blythe and colleagues have studied how many trials are theoretically required to learn a human-sized lexicon using cross-situational learning. They show that the level of referential uncertainty exposed to learners could be relatively large. However, one of the assumptions they made in designing their mathematical model is questionable. Although they rightfully assumed that words are distributed according (...) to Zipf's law, they applied a uniform distribution of meanings. In this article, Zipf's law is also applied to the distribution of meanings, and it is shown that under this condition, cross-situational learning can only be plausible when referential uncertainty is sufficiently small. It is concluded that cross-situational learning is a plausible learning mechanism but needs to be guided by heuristics that aid word learners with reducing referential uncertainty. (shrink)

Word learning depends critically on the use of linguistic context to constrain the likely meanings of words. However, the mechanisms by which children infer word meaning from linguistic context are still poorly understood. In this study, we asked whether adults (n = 58) and 2‐ to 6‐year‐old children (n = 180) use discourse coherence relations (i.e., the meaningful relationships between elements within a discourse) to constrain their interpretation of novel words. Specifically, we showed participants videos of novel animals exchanging objects. (...) These videos were accompanied by a linguistic description of the events in which we manipulated a single word within a sentence (and vs. because) in order to alter the causal and temporal relations between the events in the discourse (e.g., “One animal handed the baby to the other animal [and/because] the baby started crying in the talfa's arms”). We then asked participants which animal (the giver or the receiver) was the referent of the novel word. Across two experiments, we found evidence that young children used the causal and temporal relations in each discourse to constrain their interpretations of novel words. (shrink)

Recent research has demonstrated that word learners can determine word-referent mappings by tracking co-occurrences across multiple ambiguous naming events. The current study addresses the mechanisms underlying this capacity to learn words cross-situationally. This replication and extension of Yu and Smith (2007) investigates the factors influencing both successful cross-situational word learning and mis-mappings. Item analysis and error patterns revealed that the co-occurrence structure of the learning environment as well as the context of the testing environment jointly affected learning across observations. Learners (...) also adopted an exclusion strategy, which contributed conjointly with statistical tracking to performance. Implications for our understanding of the processes underlying cross-situational word learning are discussed. (shrink)

We evaluate here the performance of four models of cross-situational word learning: two global models, which extract and retain multiple referential alternatives from each word occurrence; and two local models, which extract just a single referent from each occurrence. One of these local models, dubbed Pursuit, uses an associative learning mechanism to estimate word-referent probability but pursues and tests the best referent-meaning at any given time. Pursuit is found to perform as well as global models under many conditions extracted from (...) naturalistic corpora of parent-child interactions, even though the model maintains far less information than global models. Moreover, Pursuit is found to best capture human experimental findings from several relevant cross-situational word-learning experiments, including those of Yu and Smith (), the paradigm example of a finding believed to support fully global cross-situational models. Implications and limitations of these results are discussed, most notably that the model characterizes only the earliest stages of word learning, when reliance on the co-occurring referent world is at its greatest. (shrink)

Cross-situational learning is a mechanism for learning the meaning of words across multiple exposures, despite exposure-by-exposure uncertainty as to the word's true meaning. We present experimental evidence showing that humans learn words effectively using cross-situational learning, even at high levels of referential uncertainty. Both overall success rates and the time taken to learn words are affected by the degree of referential uncertainty, with greater referential uncertainty leading to less reliable, slower learning. Words are also learned less successfully and more slowly (...) if they are presented interleaved with occurrences of other words, although this effect is relatively weak. We present additional analyses of participants’ trial-by-trial behavior showing that participants make use of various cross-situational learning strategies, depending on the difficulty of the word-learning task. When referential uncertainty is low, participants generally apply a rigorous eliminative approach to cross-situational learning. When referential uncertainty is high, or exposures to different words are interleaved, participants apply a frequentist approximation to this eliminative approach. We further suggest that these two ways of exploiting cross-situational information reside on a continuum of learning strategies, underpinned by a single simple associative learning mechanism. (shrink)

Abstract Since their inception, distributional models of semantics have been criticized as inadequate cognitive theories of human semantic learning and representation. A principal challenge is that the representations derived by distributional models are purely symbolic and are not grounded in perception and action; this challenge has led many to favor feature-based models of semantic representation. We argue that the amount of perceptual and other semantic information that can be learned from purely distributional statistics has been underappreciated. We compare the representations (...) of three feature-based and nine distributional models using a semantic clustering task. Several distributional models demonstrated semantic clustering comparable with clustering-based on feature-based representations. Furthermore, when trained on child-directed speech, the same distributional models perform as well as sensorimotor-based feature representations of children’s lexical semantic knowledge. These results suggest that, to a large extent, information relevant for extracting semantic categories is redundantly coded in perceptual and linguistic experience. Detailed analyses of the semantic clusters of the feature-based and distributional models also reveal that the models make use of complementary cues to semantic organization from the two data streams. Rather than conceptualizing feature-based and distributional models as competing theories, we argue that future focus should be on understanding the cognitive mechanisms humans use to integrate the two sources. (shrink)

What makes a word easy to learn? Early‐learned words are frequent and tend to name concrete referents. But words typically do not occur in isolation. Some words are predictable from their contexts; others are less so. Here, we investigate whether predictability relates to when children start producing different words (age of acquisition; AoA). We operationalized predictability in terms of a word's surprisal in child‐directed speech, computed using n‐gram and long‐short‐term‐memory (LSTM) language models. Predictability derived from LSTMs was generally a better (...) predictor than predictability derived from n‐gram models. Across five languages, average surprisal was positively correlated with the AoA of predicates and function words but not nouns. Controlling for concreteness and word frequency, more predictable predicates and function words were learned earlier. Differences in predictability between languages were associated with cross‐linguistic differences in AoA: the same word (when it was a predicate) was produced earlier in languages where the word was more predictable. (shrink)

Cognitive Science, Volume 46, Issue 1, January 2022.

Mental-content verbs such as think, believe, imagine and hope seem to pose special problems for the young language learner. One possible explanation for these diYculties is that the concepts that these verbs express are hard to grasp and therefore their acquisition must await relevant conceptual development. According to a diVerent, perhaps complementary, proposal, a major contributor to the diYculty of these items lies with the informational requirements for identifying them from the contexts in which they appear. The experiments reported here (...) explore the implications of these proposals by investigating the contribution of observational and linguistic cues to the acquisition of mental predicate vocabulary. We Wrst demonstrate that particular observed situations can be helpful in prompting reference to mental contents, speciWcally, contexts that include a salient and/or unusual mental state such as a false belief. We then compare the potency of such observational support to the reliability of alternate or concomitant syntactic information (e.g., sentential complementation) in tasks where both children and adults are asked to hypothesize the meaning of novel verbs. The Wndings support the eYcacy of false belief situations for increasing the saliency of mental state descriptions, but also show that.. (shrink)

Recently, philosophers have offered compelling reasons to think that demonstratives are best represented as variables, sensitive not to the context of utterance, but to a variable assignment. Variablists typically explain familiar intuitions about demonstratives—intuitions that suggest that what is said by way of a demonstrative sentence varies systematically over contexts—by claiming that contexts initialize a particular assignment of values to variables. I argue that we do not need to link context and the assignment parameter in this way, and that we (...) would do better not to. (shrink)

Learning to map words onto their referents is difficult, because there are multiple possibilities for forming these mappings. Cross-situational learning studies have shown that word-object mappings can be learned across multiple situations, as can verbs when presented in a syntactic context. However, these previous studies have presented either nouns or verbs in ambiguous contexts and thus bypass much of the complexity of multiple grammatical categories in speech. We show that noun word learning in adults is robust when objects are moving, (...) and that verbs can also be learned from similar scenes without additional syntactic information. Furthermore, we show that both nouns and verbs can be acquired simultaneously, thus resolving category-level as well as individual word-level ambiguity. However, nouns were learned more quickly than verbs, and we discuss this in light of previous studies investigating the noun advantage in word learning. (shrink)

We present a series of three analyses of young children's linguistic input to determine the distributional information it could plausibly offer to the process of grammatical category learning. Each analysis was conducted on four separate corpora from the CHILDES database (MacWhinney, 2000) of speech directed to children under 2;5. We showthat, in accord with other findings, a distributional analysis which categorizeswords based on their co‐occurrence patterns with surroundingwords successfully categorizes the majority of nouns and verbs. In Analyses 2 and 3, (...) we attempt to make our analyses more closely relevant to natural language acquisition by adopting more realistic assumptions about howyoung children represent their input. In Analysis 2, we limit the distributional context by imposing phrase structure boundaries, and find that categorization improves even beyond that obtained from less limited contexts. In Analysis 3, we reduce the representation of input elements which young children might not fully process and we find that categorization is not adversely affected: Although noun categorization is worse than in Analyses 1 and 2, it is still good; and verb categorization actually improves. Overall, successful categorization of nouns and verbs is maintained across all analyses. These results provide promising support for theories of grammatical category formation involving distributional analysis, as long as these analyses are combined with appropriate assumptions about the child learner's computational biases and capabilities. (shrink)

Evidence from numerous studies using the visual world paradigm has revealed both that spoken language can rapidly guide attention in a related visual scene and that scene information can immediately influence comprehension processes. These findings motivated the coordinated interplay account (Knoeferle & Crocker, 2006) of situated comprehension, which claims that utterance‐mediated attention crucially underlies this closely coordinated interaction of language and scene processing. We present a recurrent sigma‐pi neural network that models the rapid use of scene information, exploiting an utterance‐mediated (...) attentional mechanism that directly instantiates the CIA. The model is shown to achieve high levels of performance (both with and without scene contexts), while also exhibiting hallmark behaviors of situated comprehension, such as incremental processing, anticipation of appropriate role fillers, as well as the immediate use, and priority, of depicted event information through the coordinated use of utterance‐mediated attention to the scene. (shrink)

Lidz also ponders the theory of syntactic bootstrapping by asking why it is that learners are considered to retain little of extralinguistic environments (i.e., their observations) during word learning, while being able to retain detailed representations of linguistic context, for example, the multiple syntactic frames in which a verb appears. Lidz argues that learners value syntactic information over extralinguistic context from the beginning of learning, consistent with syntactic bootstrapping as a key device for verb learning.

The paper investigates the interaction of lexical and constructional meaning in valency coercion processing, and the effect of (in)compatibility between verb and construction for its successful resolution (Perek, Florent & Martin Hilpert. 2014. Constructional tolerance: Cross-linguistic differences in the acceptability of non-conventional uses of constructions. Constructions and Frames 6(2). 266–304; Yoon, Soyeon. 2019. Coercion and language change: A usage-based approach. Linguistic Research 36(1). 111–139). We present an online experiment on valency coercion (the first one on Italian), by means of a (...) semantic priming protocol inspired by Johnson, Matt A. & Adele E. Goldberg. 2013. Evidence for automatic accessing of constructional meaning: Jabberwocky sentences prime associated verbs. Language & Cognitive Processes 28(10). 1439–1452. We test priming effects with a lexical decision task which presents different target verbs preceded by coercion instances of four Italian argument structure constructions, which serve as primes. Three types of verbs serve as target: lexical associate (LA), construction associate (CA), and unrelated (U) verbs. LAs are semantically similar to the main verb of the prime sentence, whereas CAs are prototypical verbs associated to the prime construction. U verbs serve as a mean of comparison for the two categories of interest. Results confirm that processing of valency coercion requires an integration of both lexical and constructional semantics. Moreover, compatibility is also found to influence coercion resolution. Specifically, constructional priming is primary and independent from compatibility. A secondary priming effect for LA verbs is also found, which suggests a contribution of lexical semantics in coercion resolution – especially for low-compatibility coercion coinages. (shrink)

By the time they reach early adulthood, English speakers are familiar with the meaning of thousands of words. In the last decades, computational simulations known as distributional semantic models have demonstrated that it is possible to induce word meaning representations solely from word co-occurrence statistics extracted from a large amount of text. However, while these models learn in batch mode from large corpora, human word learning proceeds incrementally after minimal exposure to new words. In this study, we run a set (...) of experiments investigating whether minimal distributional evidence from very short passages suffices to trigger successful word learning in subjects, testing their linguistic and visual intuitions about the concepts associated with new words. After confirming that subjects are indeed very efficient distributional learners even from small amounts of evidence, we test a DSM on the same multimodal task, finding that it behaves in a remarkable human-like way. We conclude that DSMs provide a convincing computational account of word learning even at the early stages in which a word is first encountered, and the way they build meaning representations can offer new insights into human language acquisition. (shrink)

A variety of mechanisms contribute to word learning. Learners can track co-occurring words and referents across situations in a bottom-up manner. Equally, they can exploit sentential contexts, relying on top–down information such as verb–argument relations and world knowledge, offering immediate constraints on meaning. When combined, CSWL and SLCL potentially modulate each other's influence, revealing how word learners deal with multiple mechanisms simultaneously: Do they use all mechanisms? Prefer one? Is their strategy context dependent? Three experiments conducted with adult learners reveal (...) that learners prioritize SLCL over CSWL. CSWL is applied in addition to SLCL only if SLCL is not perfectly disambiguating, thereby complementing or competing with it. These studies demonstrate the importance of investigating word-learning mechanisms simultaneously, revealing important characteristics of their interaction in more naturalistic learning environments. (shrink)

Why are some words easier to learn than others? And what enables the eventual learning of the more difficult words? These questions were addressed for nouns using a paradigm in which adults were exposed to naturalistic maternal input that was manipulated to simulate access to several different information sources, both alone and in combination: observation of the extralinguistic contexts in which the target word was used, the words that co‐occurred with the target word, and the target word's syntactic context. Words (...) that were not accurately identified from observation alone were both abstract (e.g., music) and concrete (e.g., tail). Whether a noun could be learned from observation depended on whether it labeled a basic‐level object category (BLOC). However, the difference between BLOC labels and non‐BLOC labels was eliminated when observation was supplemented with linguistic context. Thus, although BLOC labels can be learned from observation alone, non‐BLOC labels require richer linguistic context. These findings support a model of vocabulary growth in which an important role is played by changes in the information to which learners have access. (shrink)

Prior research has shown that people can learn many nouns from a short series of ambiguous situations containing multiple words and objects. For successful cross-situational learning, people must approximately track which words and referents co-occur most frequently. This study investigates the effects of allowing some word-referent pairs to appear more frequently than others, as is true in real-world learning environments. Surprisingly, high-frequency pairs are not always learned better, but can also boost learning of other pairs. Using a recent associative model, (...) we explain how mixing pairs of different frequencies can bootstrap late learning of the low-frequency pairs based on early learning of higher frequency pairs. We also manipulate contextual diversity, the number of pairs a given pair appears with across training, since it is naturalistically confounded with frequency. The associative model has competing familiarity and uncertainty biases, and their interaction is able to capture the individual and combined effects of frequency and contextual diversity on human learning. Two other recent word-learning models do not account for the behavioral findings. (shrink)

How do children learn the meanings of propositional attitude verbs? We argue that children use information contained in both syntactic distribution and pragmatic function to zero in on the appropriate meanings. Specifically, we identify a potentially universal link between semantic subclasses of attitude verbs, their syntactic distribution and the kinds of indirect speech acts they can be used to perform. As a result, children can use the syntax as evidence about the meaning, which in turn constrains the kinds of pragmatic (...) enrichments they do and do not make in understanding these verbs in conversation. (shrink)

Gleitman and Trueswell’s “The easy words” forms a pair with their earlier paper, “Hard words,” completing a circle in which the authors ask how “easy” words (e.g., concrete nouns) are learned. They take up the hypothesis of “cross‐situational learning,” and argue that accumulating observations actually hinders learning if the mechanism requires holding all exemplars in memory over time. They present an alternative hypothesis, “Propose but Verify,” wherein people use one‐trial learning to confirm or disconfirm their current hypothesis—a mechanism distinctly different (...) from cross‐situational learning. (shrink)

Across languages, children map words to meaning with great efficiency, despite a seemingly unconstrained space of potential mappings. The literature on how children do this is primarily limited to spoken language. This leaves a gap in our understanding of sign language acquisition, because several of the hypothesized mechanisms that children use are visual (e.g., visual attention to the referent), and sign languages are perceived in the visual modality. Here, we used the Human Simulation Paradigm in American Sign Language (ASL) to (...) determine potential cues to word learning. Sign‐naïve adult participants viewed video clips of parent–child interactions in ASL, and at a designated point, had to guess what ASL sign the parent produced. Across two studies, we demonstrate that referential clarity in ASL interactions is characterized by access to information about word class and referent presence (for verbs), similarly to spoken language. Unlike spoken language, iconicity is a cue to word meaning in ASL, although this is not always a fruitful cue. We also present evidence that verbs are highlighted well in the input, relative to spoken English. The results shed light on both similarities and differences in the information that learners may have access to in acquiring signed versus spoken languages. (shrink)