Quantitative Linguistic Predictors of Infants' Learning of Specific English Words

When is a word in good company for learning?

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.

Wordform variability in infants' language environment and its effects on early word learning

Most research regarding early word learning in English tends to make the simplifying assumption that there exists a one-to-one mapping between concrete objects and their labels. In the current work, we provide evidence that runs counter to this assumption, aligning English with more morphologically-rich languages. We suggest that even in a morphologically-poor language like English, real world language input to infants does not provide tidy 1-to-1 mappings. Instead, infants encounter many variant wordforms for familiar nouns (e.g. dog∼doggy∼dogs). We explore this wordform variability in 44 English-learning infants' naturalistic environments using a longitudinal corpus of infant-available speech. We look at both the frequency and composition of wordform variability. We find two broad categories of variability: referent-changing alterations, where words were pluralized or compounded (e.g. coat∼raincoats); and wordplay, where words changed form without a notable change in referent (e.g. bird∼birdie). We further find that wordplay occurs with a limited number of lemmas that are usually early-learned, high-frequency, and shorter. When looking at all wordform variability, we find that individual words with higher levels of wordform variability are learned earlier than words with fewer wordforms, over and above the effect of frequency.

It's All in the Interaction: Early Acquired Words Are Both Frequent and Highly Imageable

Prior studies have found that children are more likely to learn words that are frequent in the input and highly imageable. Many theories of word learning, however, predict that these variables should interact, particularly early in development: frequency of a form is of little use if you cannot infer its meaning, and a concrete word cannot be acquired if you never hear it. The present study explores this interaction, how it changes over time and its relationship to syntactic category effects in children acquiring American English. We analyzed 1461 monolingual English-speaking children aged 1;4-2;6 from the MB-CDI norming study (Fenson et al., 1994). Word frequency was estimated from the CHILDES database, and imageability was measured using adult ratings. There was a strong over-additive interaction between frequency and imageability, such that children were more likely to learn a word if it was both highly imageable and very frequent. This interaction was larger in younger children than in older children. There were reliable differences between syntactic categories independent of frequency and imageability, which did not interact with age. These findings are consistent with theories in which children's early words are acquired by mapping frequent word forms onto concrete, perceptually available referents, such that highly frequent items are only acquired if they are also imageable, and vice versa.

Touch to learn: Multisensory input supports word learning and processing

Infants experience language in rich multisensory environments. For example, they may first be exposed to the word applesauce while touching, tasting, smelling, and seeing applesauce. In three experiments using different methods we asked whether the number of distinct senses linked with the semantic features of objects would impact word recognition and learning. Specifically, in Experiment 1 we asked whether words linked with more multisensory experiences were learned earlier than words linked fewer multisensory experiences. In Experiment 2, we asked whether 2-year-olds' known words linked with more multisensory experiences were better recognized than those linked with fewer. Finally, in Experiment 3, we taught 2-year-olds labels for novel objects that were linked with either just visual or visual and tactile experiences and asked whether this impacted their ability to learn the new label-to-object mappings. Results converge to support an account in which richer multisensory experiences better support word learning. We discuss two pathways through which rich multisensory experiences might support word learning.

Realistic and broad-scope learning simulations: first results and challenges

There is a current 'theory crisis' in language acquisition research, resulting from fragmentation both at the level of the approaches and the linguistic level studied. We identify a need for integrative approaches that go beyond these limitations, and propose to analyse the strengths and weaknesses of current theoretical approaches of language acquisition. In particular, we advocate that language learning simulations, if they integrate realistic input and multiple levels of language, have the potential to contribute significantly to our understanding of language acquisition. We then review recent results obtained through such language learning simulations. Finally, we propose some guidelines for the community to build better simulations.

Consequences of phonological variation for algorithmic word segmentation

Over the first year, infants begin to learn the words of their language. Previous work suggests that certain statistical regularities in speech could help infants segment the speech stream into words, thereby forming a proto-lexicon that could support learning of the eventual vocabulary. However, computational models of word segmentation have typically been tested using language input that is much less variable than actual speech is. We show that using actual, transcribed pronunciations rather than dictionary pronunciations of the same speech leads to worse segmentation performance across models. We also find that phonologically variable input poses serious problems for lexicon building, because even correctly segmented word forms exhibit a complex, many-to-many relationship with speakers' intended words. Many phonologically distinct word forms were actually the same intended word, and many identical transcriptions came from different intended words. The fact that previous models appear to have substantially overestimated the utility of simple statistical heuristics suggests a need to consider the formation of the lexicon in infancy differently.

The infant's view redefines the problem of referential uncertainty in early word learning

The learning of first object names is deemed a hard problem due to the uncertainty inherent in mapping a heard name to the intended referent in a cluttered and variable world. However, human infants readily solve this problem. Despite considerable theoretical discussion, relatively little is known about the uncertainty infants face in the real world. We used head-mounted eye tracking during parent-infant toy play and quantified the uncertainty by measuring the distribution of infant attention to the potential referents when a parent named both familiar and unfamiliar toy objects. The results show that infant gaze upon hearing an object name is often directed to a single referent which is equally likely to be a wrong competitor or the intended target. This bimodal gaze distribution clarifies and redefines the uncertainty problem and constrains possible solutions.

Toward a “Standard Model” of Early Language Learning

A standard model is a theoretical framework that synthesizes observables into a quantitative consensus. Have researchers made progress toward this kind of synthesis for children’s early language learning? Many computational models of early vocabulary learning assume that individual words are learned through an accumulation of environmental input. This assumption is also implicit in empirical work that emphasizes links between language input and learning outcomes. However, models have typically focused on average performance, whereas empirical work has focused on variability. To model individual variability, we relate the tradition of research on accumulator models to item response theory models from psychometrics. This formal connection reveals that currently available data sets do not allow researchers to test the resulting models fully, illustrating a critical need for theory to contribute to shaping new data collection and creating and testing an eventual standard model.

The Growth of Children's Semantic and Phonological Networks: Insight From 10 Languages

Children tend to produce words earlier when they are connected to a variety of other words along the phonological and semantic dimensions. Though these semantic and phonological connectivity effects have been extensively documented, little is known about their underlying developmental mechanism. One possibility is that learning is driven by lexical network growth where highly connected words in the child's early lexicon enable learning of similar words. Another possibility is that learning is driven by highly connected words in the external learning environment, instead of highly connected words in the early internal lexicon. The present study tests both scenarios systematically in both the phonological and semantic domains across 10 languages. We show that phonological and semantic connectivity in the learning environment drives growth in both production- and comprehension-based vocabularies, even controlling for word frequency and length. This pattern of findings suggests a word learning process where children harness their statistical learning abilities to detect and learn highly connected words in the learning environment.

Tailoring the Input to Children's Needs: The Use of Fine Lexical Tuning in Speech Directed to Normally Hearing Children and Children With Cochlear Implants

Purpose: The aim of the present study was to explore fine lexical tuning in Dutch infant-directed speech (IDS) addressed to congenitally deaf infants who received a cochlear implant (CI) early in life (<2 years of age) in comparison with children with normal hearing (NH). The longitudinal pattern of parents' utterance length in the initial stages of the child's lexical development was examined. Parents' utterances containing the words the children eventually acquired in the earliest developmental stages were selected and their MLU (Mean Length of Utterance) was measured.

Method: Transcriptions of monthly recordings of spontaneous interactions of 10 CI children and 30 NH children with their parents were analyzed. The children with CI were followed from the moment their device was switched on, and the NH children from the age of 6 months onwards. A total of 57,846 utterances of parents of CI children and 149,468 utterances of parents of NH children were analyzed.

Results: IDS addressed to children with NH and children with CI exhibits fine lexical tuning: parents adjust the MLU of the utterances that contain the words that children are on the verge of producing themselves. More specifically, the parents' mean length of those utterances decreased in relation to the point when the children began using the item. Consequently, the number of occurrences in isolation of the lexical item increased. The speech addressed to all the children exhibited this phenomenon, but it was significantly more strongly present in speech addressed to the children with CI.

Conclusions: The speech addressed to children with NH and CI is characterized by fine lexical tuning and a high incidence of single-word utterances in the period leading up to the children's first use of words in speech production. Notwithstanding striking commonalities, IDS addressed to children with a hearing impairment is markedly different, which suggests that parents take this specific character of the children into account.

Fine lexical tuning in infant directed speech to typically developing children

Do parents fine-tune the MLU of utterances with a particular word as the word is on the verge of appearing in the child's production? We analyzed a corpus of spontaneous interactions of 30 dyads. The children were in the initial stages of their lexical development, and the parents' utterances containing the words the children eventually acquired were selected. The main finding is that the MLU of the parental utterances containing the target words gradually decreased up to the point of the children's first production of those words. This suggests that parents fine-tune their utterances to support the children's linguistic development.

Adjacent and Non-Adjacent Word Contexts Both Predict Age of Acquisition of English Words: A Distributional Corpus Analysis of Child-Directed Speech

Children show a remarkable degree of consistency in learning some words earlier than others. What patterns of word usage predict variations among words in age of acquisition? We use distributional analysis of a naturalistic corpus of child-directed speech to create quantitative features representing natural variability in word contexts. We evaluate two sets of features: One set is generated from the distribution of words into frames defined by the two adjacent words. These features primarily encode syntactic aspects of word usage. The other set is generated from non-adjacent co-occurrences between words. These features encode complementary thematic aspects of word usage. Regression models using these distributional features to predict age of acquisition of 656 early-acquired English words indicate that both types of features improve predictions over simpler models based on frequency and appearance in salient or simple utterance contexts. Syntactic features were stronger predictors of children's production than comprehension, whereas thematic features were stronger predictors of comprehension. Overall, earlier acquisition was predicted by features representing frames that select for nouns and verbs, and by thematic content related to food and face-to-face play topics; later acquisition was predicted by features representing frames that select for pronouns and question words, and by content related to narratives and object play.

Two-year-old children's processing of two-word sequences occurring 19 or more times per million and their influence on subsequent word learning

We know that 8-month-old infants track the statistical properties of a series of syllables and that 2- and 3-year-old children process familiar phrases more efficiently than unfamiliar phrases, but less is known about the intermediary level of two-word sequences. In Study 1, 2-year-olds (N = 45, mean age = 651 days) heard two-word sequences consisting of a prime word followed by a noun, with two pictures appearing on the screen (depicting the noun and a distractor). Eye tracking showed that children looked more quickly at the noun picture for two-word sequences occurring an average of 19 times per million and 206 times per million in child-directed speech than for novel sequences. In Study 2, corpus analyses showed that 2-year-olds' noun learning increased in line with the frequency of the two-word sequence that preceded it in caregiver speech utterances. This effect holds even after controlling nouns for frequency in caregiver speech, phonemic length, neighborhood density, phonotactic probability, and concreteness and after removing nouns produced in isolation by caregivers and nouns produced by children before being produced by caregivers. These studies show that young children's language processing is facilitated by known two-word sequences, allowing children to focus on more novel aspects of the utterance. Such efficiencies are far-reaching because nearly two thirds of child-directed utterances contain two-word sequences with frequencies of 19 or more per million.

Degree and not type of iconicity affects sign language vocabulary acquisition

Lexical iconicity-signs or words that resemble their meaning-is overrepresented in children's early vocabularies. Embodied theories of language acquisition predict that symbols are more learnable when they are grounded in a child's firsthand experiences. As such, pantomimic iconic signs, which use the signer's body to represent a body, might be more readily learned than other types of iconic signs. Alternatively, the structure mapping theory of iconicity predicts that learners are sensitive to the amount of overlap between form and meaning. In this exploratory study of early vocabulary development in American Sign Language (ASL), we asked whether type of iconicity predicts sign acquisition above and beyond degree of iconicity. We also controlled for concreteness and relevance to babies, two possible confounding factors. Highly concrete referents and concepts that are germane to babies may be amenable to iconic mappings. We reanalyzed a previously published set of ASL Communicative Development Inventory (CDI) reports from 58 deaf children learning ASL from their deaf parents (Anderson & Reilly, 2002). Pantomimic signs were more iconic than other types of iconic signs (perceptual, both pantomimic and perceptual, or arbitrary), but type of iconicity had no effect on acquisition. Children may not make use of the special status of pantomimic elements of signs. Their vocabularies are, however, shaped by degree of iconicity, which aligns with a structure mapping theory of iconicity, though other explanations are also compatible (e.g., iconicity in child-directed signing). Previously demonstrated effects of type of iconicity may be an artifact of the increased degree of iconicity among pantomimic signs. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Consistency and Variability in Children's Word Learning Across Languages

Why do children learn some words earlier than others? The order in which words are acquired can provide clues about the mechanisms of word learning. In a large-scale corpus analysis, we use parent-report data from over 32,000 children to estimate the acquisition trajectories of around 400 words in each of 10 languages, predicting them on the basis of independently derived properties of the words’ linguistic environment (from corpora) and meaning (from adult judgments). We examine the consistency and variability of these predictors across languages, by lexical category, and over development. The patterning of predictors across languages is quite similar, suggesting similar processes in operation. In contrast, the patterning of predictors across different lexical categories is distinct, in line with theories that posit different factors at play in the acquisition of content words and function words. By leveraging data at a significantly larger scale than previous work, our analyses identify candidate generalizations about the processes underlying word learning across languages.

Segmentability Differences Between Child-Directed and Adult-Directed Speech: A Systematic Test With an Ecologically Valid Corpus

Previous computational modeling suggests it is much easier to segment words from child-directed speech (CDS) than adult-directed speech (ADS). However, this conclusion is based on data collected in the laboratory, with CDS from play sessions and ADS between a parent and an experimenter, which may not be representative of ecologically collected CDS and ADS. Fully naturalistic ADS and CDS collected with a nonintrusive recording device as the child went about her day were analyzed with a diverse set of algorithms. The difference between registers was small compared to differences between algorithms; it reduced when corpora were matched, and it even reversed under some conditions. These results highlight the interest of studying learnability using naturalistic corpora and diverse algorithmic definitions.

Lexical Learning May Contribute to Phonetic Learning in Infants: A Corpus Analysis of Maternal Spanish

In their first year, infants begin to learn the speech sounds of their language. This process is typically modeled as an unsupervised clustering problem in which phonetically similar speech-sound tokens are grouped into phonetic categories by infants using their domain-general inference abilities. We argue here that maternal speech is too phonetically variable for this account to be plausible, and we provide phonetic evidence from Spanish showing that infant-directed Spanish vowels are more readily clustered over word types than over vowel tokens. The results suggest that infants' early adaptation to native-language phonetics depends on their word-form lexicon, implicating a much wider range of potential sources of influence on infants' developmental trajectories in language learning.