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.