A dynamic neural field model of leaky prosody: proof of concept

Recent work has shown that lexical items come to take on the phonetic characteristics of prosodic environments in which they are typically produced, a phenomenon referred as "leaky prosody". Focusing pitch patterns Mandarin, we show leaky prosody can be derived from flat (i.e., non-transformational, non-optimizing) model speech production. Formalized using Dynamic Field Theory, our model, lexical, phonological, and inputs each exert forces Neural representing pitch. Notably, exerted by these reflect surface distributions large corpus spontaneous speech. Our simulations showed derives short timescale effect prominence production well longer prosody. By updating based form, words consistently produced high/low positions those environments.