Patterns of Adaptation in Child-Directed and Child Speech in the Emergence of Hebrew Verbs.

Children approach verb learning in ways that are specific to their native language, given the differential typological organization of verb morphology and lexical semantics. Parent-child interaction is the arena where children's socio-cognitive abilities enable them to track predictive relationships between tokens and extract linguistic generalizations from patterns and regularities in the ambient language. The current study examines how the system of Hebrew verbs develops as a network over time in early childhood, and the dynamic role of input-output adaptation in the network's increasing complexity. Focus is on the morphological components of Hebrew verbs in a dense corpus of two parent-child dyads in natural interaction between the ages 1;8-2;2. The 91-hour corpus contained 371,547 word tokens, 62,824 verb tokens, and 1,410 verb types (lemmas) in CDS and CS together. Network analysis was employed to explore the changing distributions and emergent systematicity of the relations between verb roots and verb patterns. Taking the Semitic root and pattern morphological constructs to represent linked nodes in a network, findings show that children's networks change with age in terms of node degree and node centrality, representing linkage level and construct importance respectively; and in terms of network density, as representing network growth potential. We put forward three main hypotheses followed by findings concerning (i) changes in verb usage through development, (ii) CS adaptation, and (iii) CDS adaptation: First, we show that children go through punctuated development, expressed by their using individual constructs for short periods of time, whereas parents' patterns of usage are more coherent. Second, regarding CS adaptation within a dynamic network system relative to time and CDS, we conclude that children are attuned to their immediate experience consisting of current CDS usage as well as previous usage in the immediate past. Finally, we show that parents (unintentionally) adapt to their children's language knowledge in three ways: First, by relating to their children's current usage. Second, by expanding on previous experience, building upon the usage their children have already been exposed to. And third, we show that when parents experience a limited network in the speech of their children, they provide them with more opportunities to expand their system in future interactions.