Evaluation of computational models of infant language development against robust empirical data from meta-analyses: what, why, and how?

Computational models of child language development can help us understand the cognitive underpinnings of the language learning process. One advantage of computational modeling is that is has the potential to address multiple aspects of language learning within a single learning architecture. If successful, such integrated models would help to pave the way for a more comprehensive and mechanistic understanding of language development. However, in order to develop more accurate, holistic, and hence impactful models of infant language learning, the research on models also requires model evaluation practices that allow comparison of model behavior to empirical data from infants across a range of language capabilities. Moreover, there is a need for practices that can compare developmental trajectories of infants to those of models as a function of language experience. The present study aims to take the first steps to address these needs. More specifically, we will introduce the concept of comparing models with large-scale cumulative empirical data from infants, as quantified by meta-analyses conducted across a large number of individual behavioral studies. We start by formalizing the connection between measurable model and human behavior, and then present a basic conceptual framework for meta-analytic evaluation of computational models together with basic guidelines intended as a starting point for later work in this direction. We exemplify the meta-analytic model evaluation approach with two modeling experiments on infant-directed speech preference and native/non-native vowel discrimination. We also discuss the advantages, challenges, and potential future directions of meta-analytic evaluation practices.

The Effects of Age, Dosage, and Poverty on Second Language Learning through SparkLingTM in Infant Education Centers in Madrid, Spain

The first 1000 days represent a unique window of opportunity for second language learning. In two recent studies we demonstrated that Spanish infants’ use of second-language (L2) English productive vocabulary and early utterances rapidly increased through the play-based, interactive and highly social SparkLingTM Intervention, which consists of an evidence-based method and curriculum stemming from decades of research on infant language development. Analyzing an expanded and more diverse sample of Spanish infants (n = 414; age 9–33 months) who received the SparkLingTM Intervention, this study examines the variability in L2 production, which was assessed via first-person LENA audio recordings. Infants’ age significantly and positively correlated with L2 production, demonstrating an advantage for older infants in the sample. While overall socioeconomic status (SES) was not related to L2 production, very young infants (under two years) who lived in high poverty homes showed faster increases in English production compared to peers who lived in moderate poverty homes. Infants’ attendance in the program (“dosage”) was also predictive of their L2 production outcomes. Infants across SES have the capacity to begin acquiring two languages in early education classrooms with SparkLingTM through one-hour/day sessions in social environments that engages them through frequent high-quality language input.

Relationship Between Parent Vowel Hyperarticulation in Infant-Directed Speech and Infant Phonetic Complexity on the Level of Conversational Turns

When speaking to infants, parents typically use infant-directed speech, a speech register that in several aspects differs from that directed to adults. Vowel hyperarticulation, that is, extreme articulation of vowels, is one characteristic sometimes found in infant-directed speech, and it has been suggested that there exists a relationship between how much vowel hyperarticulation parents use when speaking to their infant and infant language development. In this study, the relationship between parent vowel hyperarticulation and phonetic complexity of infant vocalizations is investigated. Previous research has shown that on the level of subject means, a positive correlational relationship exists. However, the previous findings do not provide information about the directionality of that relationship. In this study the relationship is investigated on a conversational turn level, which makes it possible to draw conclusions on whether the behavior of the infant is impacting the parent, the behavior of the parent is impacting the infant, or both. Parent vowel hyperarticulation was quantified using the vhh-index, a measure that allows vowel hyperarticulation to be estimated for individual vowel tokens. Phonetic complexity of infant vocalizations was calculated using the Word Complexity Measure for Swedish. Findings were unexpected in that a negative relationship was found between parent vowel hyperarticulation and phonetic complexity of the immediately following infant vocalization. Directionality was suggested by the fact that no such relationship was found between infant phonetic complexity and vowel hyperarticulation of the immediately following parent utterance. A potential explanation for these results is that high degrees of vowel hyperarticulation either provide, or co-occur with, large amounts of phonetic and/or linguistic information, which may occupy processing resources to an extent that affects production of the next vocalization.

Multimodal attention creates the visual input for infant word learning

Infant language acquisition is fundamentally an embodied process, relying on the body to select information from the learning environment. Infants show their attention to an object not merely by gazing at the object, but also through orienting their body towards the object and generating various types of manual actions on the object, such as holding, touching, and shaking. The goal of the present study was to examine how multimodal attention shapes infant word learning in real-time. Infants and their parents played in a home-like lab with unfamiliar objects with assigned labels. While playing, participants wore wireless head-mounted eye trackers to capture visual attention. Infants were then tested on their knowledge of the new words. We identified all the utterances in which parents labeled the learned or not learned objects and analyzed infant multimodal attention during and around labeling. We found that proportion of time spent in hand-eye coordination predicted learning outcomes. To understand the learning advantage hand-eye coordination creates, we compared the size of objects in the infant’s field of view. Although there were no differences in object size between learned and not learned labeling utterances, hand-eye coordination created the most informative views. Together, these results suggest that in-the-moment word learning may be driven by the greater access to informative object views that hand-eye coordination affords.

An Integrated Perspective on Spatio-Temporal Attention and Infant Language Acquisition

Efficiency in the early ability to switch attention toward competing visual stimuli (spatial attention) may be linked to future ability to detect rapid acoustic changes in linguistic stimuli (temporal attention). To test this hypothesis, we compared individual performances in the same cohort of Italian-learning infants in two separate tasks: (i) an overlap task, measuring disengagement efficiency for visual stimuli at 4 months (Experiment 1), and (ii) an auditory discrimination task for trochaic syllabic sequences at 7 months (Experiment 2). Our results indicate that an infant’s efficiency in processing competing information in the visual field (i.e., visuospatial attention; Exp. 1) correlates with the subsequent ability to orient temporal attention toward relevant acoustic changes in the speech signal (i.e., temporal attention; Exp. 2). These results point out the involvement of domain-general attentional processes (not specific to language or the sensorial domain) playing a pivotal role in the development of early language skills in infancy.

Higher Utilization of Social Services Is Associated with Higher Language Scores in Children from Deeply Impoverished Urban Families

Language development has been consistently linked with socioeconomic status (SES), with children from lower SES backgrounds at higher risk for language delays. The objective of this study is to investigate the relationship between familial social service use and language development during the first year of life. Thirty-one low-income mothers and their infants were recruited from the New York metropolitan area. Mothers provided information about demographics and utilization of social services (Women, Infants, and Children (WIC), food stamps, Medicaid, and public housing). Infant language skills were assessed using the Preschool Language Scale. Multiple linear regressions were used to investigate the relationship between social service use and language skills. We found that the number of social services utilized was not an overall significant linear predictor of language skills. However, social service use interacted with poverty level to predict language skills. Specifically, for families living in deep poverty, higher service use significantly predicted higher infant language scores (β = 3.4, p = 0.005). These results suggest that social services may be an appropriate target to help narrow socioeconomic disparities in language development.