Larissa K. Samuelson - US grants

DESCRIPTION (Applicant's Abstract): Children are amazing word learners. Even though the number of possible meanings for each novel word is immense, children learn words quickly and with seemingly little effort. Previous research has suggested that children are biased to only consider some of the possible meanings for a new word. However, the origin of word learning biases and the mechanisms by which they operate on a moment-to-moment basis has yet to be determined. The proposed project addresses this gap in our understanding of word learning. In three projects, I will test the hypothesis that word learning biases develop out of statistical regularities in the language and categories children learn. The specific questions addressed are: (1) what are the statistical regularities in the language input of children; (2) can a simple learner of statistical regularities replicate the development of specific word-learning biases; and (3) if the statistical regularities found in the words children know are altered, does the developmental trajectory of specific word learning biases by change? These questions are addressed in neural network simulations of the development of specific word learning biases; experimental studies in which the natural development of a word learning bias is altered by teaching children words; and in modeling of the changes in word learning that arise from the lexical training.

[unreadable] DESCRIPTION (provided by applicant): Learning word meanings in a first language should be a difficult task. The number and range of possible meanings to be learned is immense and, by some arguments, indeterminate from the typical information provided to young children. Despite these challenges, between 18- and 30-months-of-age the typical child's productive vocabulary increases tenfold. Many of the new words children learn at this age are nouns. Previous research suggests that the task of learning new nouns is made easier by biases or constraints that direct children's attention to the correct features of objects for name learning. Despite considerable empirical support for these word learning biases, however, little is known about their origins or the mechanisms by which they are implemented. This is the focus of the present proposal. The present study will test and augment a processed-based developmental account of how perceptual word learning biases develop. Specifically, the empirical studies proposed here are based on the proposal that the bias to attend shape young children demonstrate when learning names for novel solid objects is the developmental product of the early noun vocabulary. Thirteen studies will test specific predictions concerning the relation between young children's vocabularies, category knowledge, and the development of word learning biases. In so doing these studies will further our understanding of the mechanisms that support young children's smart word learning abilities. The present research has potential implications for developmental delays in language learning and specifically for children who evidence delays in learning words. Previous research in this area suggests that children given intensive exposure to sets of words designed to create a precocious word learning bias not only develop an early bias but show acceleration in their subsequent acquisition of new words outside the laboratory. This suggests the possibility of interventions to help children with delays in word learning. Before such interventions can be developed, however, much more needs to be known about the basic processes that support the development of word learning biases. Thus, the present research is designed to lay the foundation for the application of [unreadable] what is known about the development of word learning biases to children with delays in word learning. [unreadable] [unreadable]

DESCRIPTION (provided by applicant): Word learning is a complex phenomenon because it is tied to many different behaviors. It also involves many different perceptual and conceptual systems and is extended in time. Although complex, achieving a richer understanding of early word learning is a fundamental goal given that vocabulary development is correlated with later cognitive functioning and with processes that have a pervasive impact on general cognitive abilities such as executive function. Moreover, deficits in early word learning have a profound effect on cognitive functioning in atypical populations including children with specific language impairment. Given the complexity of word learning, a central challenge has been to establish empirical paradigms that effectively reveal the processes of word learning and to develop new theories that uncover the mechanisms that move word learning forward. Our previous work suggests word learning occurs in a cascade of individual decisions about word meaning in the moment. These decisions build on each other to shape subsequent decisions and, over multiple timescales, create developmental change. The goal of this grant is to develop and test a unified model of word learning that captures processes at both the second-to-second and developmental timescales and provides a process-based account of how individual behaviors accumulate to create development. The research plan builds and tests this model. Specific Aim 1 creates a unified model of word learning behaviors that extends beyond our prior work on noun generalization to include processes of comprehension, production, referent selection, and generalization from multiple exemplars. Specific Aim 2 adds a more complete account of object-word interactions and the development of word learning biases. Specific Aim 3 adds a memory process, enabling the model to learn a lexicon over multiple timescales and develop word learning biases. The end result of this work will be models of individual developmental trajectories that integrate word learning processes over multiple timescales-a necessary step towards intervention in cases of atypical development such as SLI, or prematurity. The integration of processes over multiple timescales and at the level of individual participants is an issue few models have addressed directly. Furthermore, the model we propose brings together work on object processing in visual cognition with work on early word learning an is related to a larger program of neurally-grounded modeling work that integrates across multiple visual, spatial, motor and language systems. Thus this model will be foundational to a full understanding of the multiple systems involved in word learning and vocabulary development, and to the interaction of these processes with cognition more generally. PUBLIC HEALTH RELEVANCE: The goal of this grant is to develop and test a unified model of word learning that captures both children's individual uses of words, and how these individual behaviors accumulate to create development. Achieving this goal will have broad implications for our understanding of child development because early vocabulary development is correlated with later cognitive functioning and with processes that have a pervasive impact on general cognitive abilities. The end result of this work will be models of individual children's development-a necessary step towards intervention in cases of atypical development such as late talkers, SLI, or prematurity.

DESCRIPTION (provided by applicant): Word learning is complex because it is tied to many different behaviors (looking, pointing), subserved by multiple perceptual and cognitive processes (attention, memory, categorization), and because it is extended in time. Thus, explanations of early word learning must bring together processes of visual attention, visual looking and learning, processes for tracking of what is where visual binding of what is where, and processes for the formation and updating of word-object links across contexts. And, critically, such explanations must be dynamic - able to capture how these processes work together in as children behave in the moment and how they change over learning and development. The goal of this grant is to test the first formal theory of early word learning that integrates visual object processing and word learning across multiple timescales. We build a formal computational model based on two successful extant models: the model of early word learning we developed in the last grant period, and a model of how infants visually explore their environment. The end product will be a theory that speaks to how individual children coordinate attention, visual memory, and word learning, and how these systems co-evolve over development. In four specific aims we propose 20 experiments to test model predictions, developmental hypotheses and generate data that speak to critical debates in the field. Specific Aim 1 examines the influence of words on visual exploration in space. Specific Aim 2 probes visual dynamics in tasks where children must track how often words and objects occur together. Specific Aim 3 cross-situational learning tasks examine the continuity in processes of attention and word learning in looking and reaching tasks. Finally, Specific Aim 4 extends our novel discovery from the last grant period-that children can use consistency in space to learn words-to category learning, and examine the influence of developmental changes in spatial cognitive development on word learning. The 15 proposed simulations qualitatively capture targeted findings from the literature and data from our experiments, and do so using the same parameters for related studies. The resultant theory will be poised to make extensive connections to other aspects of development including processes of preferential looking, working memory, and executive function-processes we have argued are critical for understanding how children's developing word knowledge influences in-the-moment learning. Moreover, the level at which we seek to understand these processes-the level of individual children-is a necessary step towards intervention in cases of atypical development such as SLI or prematurity55. Indeed, because our theory targets individual differences very early in word learning, the present work opens the door to future efforts using the theory to predict which infants are at risk for language delays and to use our theoretical model as a clinical tool to test candidate interventions.