PROJECT SUMMARY Children learn language in the context of complex and multidimensional social interactions. Affective information is a prominent yet, underexplored feature of children's early learning context despite its effects on infants' attention and memory systems.1–3 Understanding how children’s affective environments shape learning is an important public health issue, as approximately a third of mothers in the US experience depression in the first three years of a child’s life.4 The proposed research aims to fill this knowledge gap by capturing theoretically important dimensions of affect (such as valence and arousal5) in children’s real-world early environments, with the ultimate translational goal of understanding the genesis of variation in language and well-being. Central to this goal will be evaluating how children adapt their use of affective information in families with depressed caregivers. We will quantify the affective context of learning events using multiple methods, including experimental studies and descriptive computer vision and machine learning analyses of a large dataset of egocentric-view videos of children’s home environment. In Aim 1, we will ask whether the structure of affective information in caregiver input predicts the age at which children produce different types of words (Aim 1a) and how children adapt to the affective cues of caregivers with depression symptoms (Aim 1b). We will use automatically coded affective features from videos across three modalities (facial, vocal, and semantic) to predict the age at which children produce different words depending on the severity of their caregivers’ depression symptoms. In Aim 2, we build on the correlational, naturalistic evidence from Aim 1 by taking an experimental approach, allowing us to uncover causal links between affect and word learning. Specifically, in three novel word-learning studies, we will disentangle which dimensions of affect (e.g., valence and arousal) are most impactful. Together, this research will provide insight into the affective mechanisms that promote word learning in child-caregiver interactions, with the ultimate goal of improving resilience and language development in households with atypical affective dynamics. Through these aims, the applicant will build on her background in affective development and gain training in computational methods and theoretical knowledge in understanding clinical variation in affect across households. The sponsor, Dr. Michael Frank, has deep expertise in early learning and extensive mentorship experience in developmental and computational methods. The co-sponsor, Dr. Ian Gotlib, is an expert in the study of caregiver depression. The unique training environment at Stanford University provides the resources necessary to successfully complete the proposed work, as well as additional opportunities for mentorship and professional development. In sum, the proposed project and training plan will support the applican...