PHONOLOGICAL TO ARTICULATORY CASCADE 1 Running Head: PHONOLOGICAL TO ARTICULATORY CASCADE Cascading Activation From Phonological Planning to Articulatory Processes: Evidence from Tongue Twisters

Research into spoken word production has often focused on the interaction of lexical selection processes and phonological planning. Less attention has been given to the relationship between phonological planning and articulatory processes. The current study considers evidence from the tongue twister paradigm to investigate such potential interactions. Acoustic analyses of various parameters of obstruents voicing in tongue twister productions show that errors induced in tongue twisters leave acoustic “traces” of the intended target. For example, the voice-onset time of “k”[g] error tokens had a mean VOT that was longer than correctly produced “g”[g] tokens, reflecting a trace of the voiceless [k] target. This effect is attributed to the cascade of partially-activated phonological representations of the target consonant into articulatory processes. Consistent with this account, a post-hoc analysis revealed an additional influence of cascading activation from word-level processes; traces of the target were reduced in word outcomes relative to nonword outcomes. Finally, extension of these analyses to a set of secondary cues to obstruent voicing showed that non-local cues are not influenced by tongue twister production errors. PHONOLOGICAL TO ARTICULATORY CASCADE 3 Cascading Activation From Phonological Planning to Articulatory Processes: Evidence from Tongue Twisters Most theories of single-word production assume three general stages of post-semantic processing. We refer to the first stage as lexical selection; it involves the selection of a word to express a nonverbal concept (e.g., selecting CALF to express “young bovine”). This lexical representation serves as input to phonological planning processes which specify the appropriate sound sequence for this word (e.g., monosyllable, onset: /k/, nucleus:/Q/, coda: /f/). Finally, articulatory implementation processes execute the specified sound sequence (e.g., for onset /k/, elevating the tongue body to create a closure on the soft palate while abducting the vocal folds). Although theories may differ in how each process is implemented, most share these broad distinctions between lexical, phonological, and articulatory processes (e.g., Dell & O’Seaghdha [1991] distinguish word, phonological, and phonetic/articulatory representations). In the current paper, we focus on these distinctions in order to explore how these different types of processes interact. We use the term interaction to refer to the degree to and manner in which two processing stages (i.e., two processes or two distinct groups of processes) influence one another (see Rapp & Goldrick [2000] for further discussion). At one end of the spectrum of interactivity, processes have limited influence over one another. A processing stage generates a single representation on the basis of its input; this single output representation is then transmitted to subsequent processing stages. For example, in such a system, lexical selection processes generate a single word representation (e.g., CALF); this single word is then transmitted to phonological planning processes (generating /kQf/). We refer to this type of theory as a discrete