Connecting Phonological Encoding To Articulation- Is Cascading Required? A Computational Investigation

Current psychological models of word production (e.g. Dell, 1986; Levelt, Roelofs, & Meyer, 1999) only detail how we plan the phonological content of words, and not how we artic-ulate them. To understand how these models may be extended, we must determine how information flows from phonological encoding to articulation. For example, does activation cas-cade from unselected phonological representations? So far, the clearest support for cascading at this interface has come from the finding that erroneously produced phonemes exhibit char-acteristics of the intended phoneme (Goldrick & Blumstein, 2006). In this paper however, we use computational implemen-tations to demonstrate that both cascading and non-cascading models can account for this result. An extension of model be-haviour analysis to other speech error phenomena additionally shows that models based on the classic spreading activation ac-count of word production (Dell, 1986) experience problems in replicating some key aspects of human error patterns