Modeling the Emergence of Lexicon in Homesign Systems

Homesigners are deaf children who are not exposed to a conventional language model, yet spontaneously develop their own gestural communication systems. Drawing on the seminal work of Susan Goldin-Meadow, this thesis employs a multi-agent computational model simulating parent–child interaction in a referential game to investigate the mechanisms that enable the emergence of a stable lexicon under conditions of extremely impoverished and inconsistent input. In comprehension, it compares the respective roles of iconicity and co-occurrence statistics across three developmental conditions: deaf children exposed to sign language, hearing children acquiring spoken language, and homesigners. In production, it examines the contributions of iconicity, corrective feedback, and internal self-supervision across the latter two conditions.