From the lexicon to expectations about kinds: a role for associative learning

In the novel noun generalization task, 2 1/2-year-old children display generalized expectations about how solid and nonsolid things are named, extending names for never-before-encountered solids by shape and for never-before-encountered nonsolids by material. This distinction between solids and nonsolids has been interpreted in terms of an ontological distinction between objects and substances. Nine simulations and behavioral experiments tested the hypothesis that these expectations arise from the correlations characterizing early learned noun categories. In the simulation studies, connectionist networks were trained on noun vocabularies modeled after those of children. These networks formed generalized expectations about solids and nonsolids that match children's performances in the novel noun generalization task in the very different languages of English and Japanese. The simulations also generate new predictions supported by new experiments with children. Implications are discussed in terms of children's development of distinctions between kinds of categories and in terms of the nature of this knowledge.     

Statistical regularities in vocabulary guide language acquisition in connectionist models and 15-20-month-olds

This research tested the hypothesis that young children's bias to generalize names for solid objects by shape is the product of statistical regularities among nouns in the early productive vocabulary. Data from a 4-layer Hopfield network suggested that the statistical regularities in the early noun vocabulary are strong enough to create a shape bias, and that the shape bias is overgeneralized to nonsolid stimuli. A 2nd simulation suggested that this overgeneralization is due to the dominance of names for shape-based categories in the early noun vocabulary. Two subsequent longitudinal experiments tested whether it is possible to create word learning biases in children. Children 15-20 months old were given intensive naming experiences with 12 noun categories typical of the types of categories children learn to name early. The children developed a precocious shape bias that was overgeneralized to naming nonsolid substances; they also showed accelerated vocabulary development. Children taught an atypical set of nouns or no new nouns did not develop a shape bias and did not show accelerated vocabulary development.     

Naming solids and nonsolids: Children's default construals

Four experiments used a free-naming task to examine children's and adults' default construals of solids and nonsolids. In Experiment 1,4-year-old children viewed entities presented in familiar geometric shapes (e.g., square, triangle), without touching them. One half saw solids (e.g., a square made of wood); the other half saw nonsol ids matched carefully in shape (e.g., a square with the same dimensions but made of peanut butter). To tap their default construals, children were simply asked, “What is that?” Answers varied sharply with the type of stimulus. If the entity was solid, children tended to provide an individual-related word (e.g., “a square”), even if they also knew a substance-related word (e.g., “wood”). But if the stimulus was nonsolid, children tended to give a substance-related word (e.g., “peanut butter”), even if they also knew an individual-related word (e.g., “a square”). These words were usually common nouns produced in appropriate sentential contexts, suggesting that 4-year-olds represented the words as naming kinds. The same pattern of results obtained in Experiments 2 and 3, which were modified replications of Experiment 1. The results of Experiment 4 replicated the main findings of Experiment 1 using adults as participants. The studies suggest that, as a default, 4-year-olds conceptualize solids and nonsolids in (a) fundamentally distinct, (b) kind-based (rather than perceptual property-based), and (c) adult-like ways.     

Attentive tracking of objects versus substances

Recent research in vision science, infant cognition, and word learning suggests a special role for the processing of discrete objects. But what counts as an object? Answers to this question often depend on contrasting object-based processing with the processing of spatial areas or unbound visual features. In infant cognition and word learning, though, another salient contrast has been between rigid cohesive objects and nonsolid substances. Whereas objects may move from one location to another, a nonsolid substance must pour from one location to another. In the study reported here, we explored whether attentive tracking processes are sensitive to dynamic information of this type. Using a multiple-object tracking task, we found that subjects could easily track four items in a display of eight identical unpredictably moving entities that moved as discrete objects from one location to another, but could not track similar entities that noncohesively "poured" from one location to another-even when the items in both conditions followed the same trajectories at the same speeds. Other conditions revealed that this inability to track multiple "substances" stemmed not from violations of rigidity or cohesiveness per se, because subjects were able to track multiple noncohesive collections and multiple nonrigid deforming objects. Rather, the impairment was due to the dynamic extension and contraction during the substancelike motion, which rendered the location of the entity ambiguous. These results demonstrate a convergence between processes of midlevel adult vision and infant cognition, and in general help to clarify what can count as a persisting dynamic object of attention.     

Individual differences in the shape bias in preschool children with specific language impairment and typical language development: theoretical and clinical implications

We investigated whether preschool children with specific language impairment (SLI) exhibit the shape bias in word learning: the bias to generalize based on shape rather than size, color, or texture in an object naming context (‘This is a wek; find another wek’) but not in a non‐naming similarity classification context (‘See this? Which one goes with this one?’). Fifty‐four preschool children (16 with SLI, 16 children with typical language [TL] in an equated control group, and 22 additional children with TL included in individual differences analyses but not group comparisons) completed a battery of linguistic and cognitive assessments and two experiments. In Experiment 1, children made generalization choices in object naming and similarity classification contexts on separate days, from options similar to a target object in shape, color, or texture. On average, TL children exhibited the shape bias in an object naming context, but children with SLI did not. In Experiment 2, we tested whether the failure to exhibit the shape bias might be linked to ability to detect systematicities in the visual domain. Experiment 2 supported this hypothesis, in that children with SLI failed to learn simple paired visual associations that were readily learned by children with TL. Analyses of individual differences in the two studies revealed that visual paired‐associate learning predicted degree of shape bias in children with SLI and TL better than any other measure of nonverbal intelligence or standard assessments of language ability. We discuss theoretical and clinical implications.     

Infant sustained attention but not joint attention to objects at 9 months predicts vocabulary at 12 and 15 months

Vocabulary differences early in development are highly predictive of later language learning as well as achievement in school. Early word learning emerges in the context of tightly coupled social interactions between the early learner and a mature partner. In the present study, we develop and apply a novel paradigm—dual head‐mounted eye tracking—to record momentary gaze data from both parents and infants during free‐flowing toy‐play contexts. With fine‐grained sequential patterns extracted from continuous gaze streams, we objectively measure both joint attention and sustained attention as parents and 9‐month‐old infants played with objects and as parents named objects during play. We show that both joint attention and infant sustained attention predicted vocabulary sizes at 12 and 15 months, but infant sustained attention in the context of joint attention, not joint attention itself, is the stronger unique predictor of later vocabulary size. Joint attention may predict word learning because joint attention supports infant attention to the named object.     

Language-relative construal of individuation constrained by universal ontology: revisiting language universals and linguistic relativity

Objects and substances bear fundamentally different ontologies. In this article, we examine the relations between language, the ontological distinction with respect to individuation, and the world. Specifically, in cross‐linguistic developmental studies that follow Imai and Gentner (1997) , we examine the question of whether language influences our thought in different forms, like (1) whether the language‐specific construal of entities found in a word extension context ( Imai & Gentner, 1997 ) is also found in a nonlinguistic classification context; (2) whether the presence of labels per se, independent of the count‐mass syntax, fosters ontology‐based classification; (3) in what way, if at all, the count‐mass syntax that accompanies a label changes English speakers' default construal of a given entity? On the basis of the results, we argue that the ontological distinction concerning individuation is universally shared and functions as a constraint on early learning of words. At the same time, language influences one's construal of entities cross‐lingistically and developmentally, and causes a temporary change of construal within a single language. We provide a detailed discussion of how each of these three ways language may affect the construal of entities, and discuss how our universally possessed knowledge interacts with language both within a single language and in cross‐linguistic context.     

Statistical word learning at scale: the baby's view is better

A key question in early word learning is how children cope with the uncertainty in natural naming events. One potential mechanism for uncertainty reduction is cross‐situational word learning – tracking word/object co‐occurrence statistics across naming events. But empirical and computational analyses of cross‐situational learning have made strong assumptions about the nature of naming event ambiguity, assumptions that have been challenged by recent analyses of natural naming events. This paper shows that learning from ambiguous natural naming events depends on perspective. Natural naming events from parent–child interactions were recorded from both a third‐person tripod‐mounted camera and from a head‐mounted camera that produced a ‘child's‐eye’ view. Following the human simulation paradigm, adults were asked to learn artificial language labels by integrating across the most ambiguous of these naming events. Significant learning was found only from the child's perspective, pointing to the importance of considering statistical learning from an embodied perspective.     

Ontological categories guide young children's inductions of word meaning: object terms and substance terms

Three experiments assessed the possibility, suggested by Quine (1960, 1969) among others, that the ontology underlying natural language is induced in the course of language learning, rather than constraining learning from the beginning. Specifically, we assessed whether the ontological distinction between objects and non-solid substances conditions projection of word meanings prior to the child's mastery of count/mass syntax. Experiments 1 and 2 contrasted unfamiliar objects with unfamiliar substances in a word-learning task. Two-year-old subjects' projection of the novel word to new objects respected the shape and number of the original referent. In contrast, their projection of new words for non-solid substances ignored shape and number. There were no effects of the child's knowledge of count/mass syntax, nor of the syntactic context in which the new word was presented. Experiment 3 revealed that children's natural biases in the absence of naming do not lead to the same pattern of results. We argue that these data militate against Quine's conjecture.     

Embodied attention and word learning by toddlers

Many theories of early word learning begin with the uncertainty inherent to learning a word from its co-occurrence with a visual scene. However, the relevant visual scene for infant word learning is neither from the adult theorist's view nor the mature partner's view, but is rather from the learner's personal view. Here we show that when 18-month old infants interacted with objects in play with their parents, they created moments in which a single object was visually dominant. If parents named the object during these moments of bottom-up selectivity, later forced-choice tests showed that infants learned the name, but did not when naming occurred during a less visually selective moment. The momentary visual input for parents and toddlers was captured via head cameras placed low on each participant's forehead as parents played with and named objects for their infant. Frame-by-frame analyses of the head camera images at and around naming moments were conducted to determine the visual properties at input that were associated with learning. The analyses indicated that learning occurred when bottom-up visual information was clean and uncluttered. The sensory-motor behaviors of infants and parents were also analyzed to determine how their actions on the objects may have created these optimal visual moments for learning. The results are discussed with respect to early word learning, embodied attention, and the social role of parents in early word learning.     

Age and age of acquisition: An evaluation of the cumulative frequency hypothesis

An important determinant of picture and word naming speed is the age at which the names were learned (age of acquisition). Two related interpretations of these effects are that they reflect differences between words in their cumulative frequency of use, or that they reflect differences in the amount of time early and lateacquired words have spent in lexical memory. Both theories predict that differences between early and late-acquired words will be less apparent in older than younger adults. Two experiments are reported in which younger and older adults read words varying in age of acquisition or frequency, or named objects varying in age of acquisition. There was an observed effect of word frequency only for young adults' word naming. In contrast, strong age of acquisition effects were found for both the young and the old participants. The implications of these results for theories of how age of acquisition might affect lexical processing are discussed.     

Domain effects in lexical development: learning words for foods and toys

We investigated 3-year-olds’ and adults’ use of domain cues in learning words for solid and nonsolid material entities. In Experiments 1 and 2, participants heard a novel neutral noun (e.g., “my X”) for a standard solid or nonsolid entity described as either a toy or a food. They then were asked to extend the word to one of two other entities. Both options matched the standard in solidity; but one differed from it in an object-relevant property (shape) and the other in a substance-relevant property (color, texture, or smell). Both children and adults were more likely to select the same-shaped entity if the standard was (1) solid than if it was nonsolid, and (2) described as a toy than if it was described as a food. Their interpretations of novel words for material entities were thus affected not only by perceptual information (about solidity) but also by conceptual information (about domain). In Experiment 3, the novel noun was presented in a syntactic context that suggested the solid entity should be interpreted as an object (e.g., “an X”) and that the nonsolid entity should be interpreted as a substance (e.g., “some X”). For adults, these changes largely eliminated the effect of the entity’s domain (toy, food) on interpretation. We interpret these findings in terms of the proposal that domain cues, like solidity cues, furnish information about whether an entity’s structure should be thought of as arbitrary or nonarbitrary and, hence, about whether a word should be interpreted as naming an object or a substance construal.     

Working Memory Capacity and Time Course of Predictive Inferences

A naming task assessed activation of inference concepts during reading. A predicting, or a control, context sentence was followed by a target word to be named, which represented the predicted event or an inconsistent event. The interval between the end of the context and the onset of the target word varied between 50 and 1050-msec. Individual differences in working memory capacity were assessed by the reading span task. As reflected by facilitation in naming latencies in the predicting condition, relative to the control condition, (a) inferences were not made within the first 50-msec after the context, regardless of reading span; (b) only the high-span participants drew inferences within a 550-msec interval; and, (c) both the high-and the low-span participants generated them within a 1050-msec interval. These results indicate that high working memory capacity accelerates the time course of predictive inferences, although they do not become automatic. We propose that this effect occurs because these inferences involve time-consuming elaborations that place demands on the effortful and limited resources of working memory. Deficiencies in word knowledge, speed of lexical access, or comprehension of explicit information do not account for low-span readers' difficulties in generating predictive inferences.     

Working memory capacity and time course of predictive inferences

A naming task assessed activation of inference concepts during reading. A predicting, or a control, context sentence was followed by a target word to be named, which represented the predicted event or an inconsistent event. The interval between the end of the context and the onset of the target word varied between 50 and 1050 msec. Individual differences in working memory capacity were assessed by the reading span task. As reflected by facilitation in naming latencies in the predicting condition, relative to the control condition, (a) inferences were not made within the first 50 msec after the context, regardless of reading span; (b) only the high-span participants drew inferences within a 550-msec interval; and, (c) both the high- and the low-span participants generated them within a 1050-msec interval. These results indicate that high working memory capacity accelerates the time course of predictive inferences, although they do not become automatic. We propose that this effect occurs because these inferences involve time-consuming elaborations that place demands on the effortful and limited resources of working memory. Deficiencies in word knowledge, speed of lexical access, or comprehension of explicit information do not account for low-span readers' difficulties in generating predictive inferences.     

Children show limited movement repertoire when learning a novel motor skill

Examining age differences in motor learning using real‐world tasks is often problematic due to task novelty and biomechanical confounds. Here, we investigated how children and adults acquire a novel motor skill in a virtual environment. Participants of three different age groups (9‐year‐olds, 12‐year‐olds, and adults) learned to use their upper body movements to control a cursor on a computer screen. Results showed that 9‐year‐old and 12‐year‐old children showed poorer ability to control the cursor at the end of practice. Critically, when we investigated the movement coordination, we found that the lower task performance of children was associated with limited exploration of their movement repertoire. These results reveal the critical role of motor exploration in understanding developmental differences in motor learning.     

Processes and pathways in development via digital media: Examples from word learning

Word learning unfolds over multiple, cascading pathways which support in-the-moment processing and learning. The process is refined with each exposure to a word, and exposures to new words occur across a variety of forms and contexts. However, as children are exposed to more and more digital media, the ways in which children encounter, learn, and build on their vocabulary is also shifting. These shifts represent changes in context, content, and at the level of the child that can lead to negative outcomes. Less work, however, has discussed what these differences mean for how things change in the underlying developmental cascade and learning processes. Here, we suggest that the increasing presence of digital media may shift the developmental pathways for learning (the chain of events that support future learning) but not necessarily the developmental processes (the mechanisms underlying learning). Moreover, the interaction of the two may lead to different behavior and outcomes for learning in a digital era. We argue it is imperative for researchers to not only study how digital media differs from everyday learning, but directly measure if the well-worn pathways, processes, and variables found with decades of research with real items translate to a digital media era.     

Detailed behavioral analysis as a window into cross-situational word learning

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.     

Co‐occurrence statistics as a language‐dependent cue for speech segmentation

To what extent can language acquisition be explained in terms of different associative learning mechanisms? It has been hypothesized that distributional regularities in spoken languages are strong enough to elicit statistical learning about dependencies among speech units. Distributional regularities could be a useful cue for word learning even without rich language‐specific knowledge. However, it is not clear how strong and reliable the distributional cues are that humans might use to segment speech. We investigate cross‐linguistic viability of different statistical learning strategies by analyzing child‐directed speech corpora from nine languages and by modeling possible statistics‐based speech segmentations. We show that languages vary as to which statistical segmentation strategies are most successful. The variability of the results can be partially explained by systematic differences between languages, such as rhythmical differences. The results confirm previous findings that different statistical learning strategies are successful in different languages and suggest that infants may have to primarily rely on non‐statistical cues when they begin their process of speech segmentation.     

The evocative power of words: activation of concepts by verbal and nonverbal means

A major part of learning a language is learning to map spoken words onto objects in the environment. An open question is what are the consequences of this learning for cognition and perception? Here, we present a series of experiments that examine effects of verbal labels on the activation of conceptual information as measured through picture verification tasks. We find that verbal cues, such as the word "cat," lead to faster and more accurate verification of congruent objects and rejection of incongruent objects than do either nonverbal cues, such as the sound of a cat meowing, or words that do not directly refer to the object, such as the word "meowing." This label advantage does not arise from verbal labels being more familiar or easier to process than other cues, and it does extends to newly learned labels and sounds. Despite having equivalent facility in learning associations between novel objects and labels or sounds, conceptual information is activated more effectively through verbal means than through nonverbal means. Thus, rather than simply accessing nonverbal concepts, language activates aspects of a conceptual representation in a particularly effective way. We offer preliminary support that representations activated via verbal means are more categorical and show greater consistency between subjects. These results inform the understanding of how human cognition is shaped by language and hint at effects that different patterns of naming can have on conceptual structure.     

A Computational Model of the Self‐Teaching Hypothesis Based on the Dual‐Route Cascaded Model of Reading

The self‐teaching hypothesis describes how children progress toward skilled sight‐word reading. It proposes that children do this via phonological recoding with assistance from contextual cues, to identify the target pronunciation for a novel letter string, and in so doing create an opportunity to self‐teach new orthographic knowledge. We present a new computational implementation of self‐teaching within the dual‐route cascaded (DRC) model of reading aloud, and we explore how decoding and contextual cues can work together to enable accurate self‐teaching under a variety of circumstances. The new model (ST‐DRC) uses DRC’s sublexical route and the interactivity between the lexical and sublexical routes to simulate phonological recoding. Known spoken words are activated in response to novel printed words, triggering an opportunity for orthographic learning, which is the basis for skilled sight‐word reading. ST‐DRC also includes new computational mechanisms for simulating how contextual information aids word identification, and it demonstrates how partial decoding and ambiguous context interact to achieve irregular‐word learning. Beyond modeling orthographic learning and self‐teaching, ST‐DRC’s performance suggests new avenues for empirical research on how difficult word classes such as homographs and potentiophones are learned.