A comparison of speech errors elicited by sentences and alternating repetitive tongue twisters

Sound-level errors collected by ear from continuous communicative speech have been interpreted as mis-selections of planning elements, which are then produced fluently without residue of the original target (Lashley 1957, Fromkin 1972, Garrett 1975, Shattuck-Hufnagel 1982). In contrast, articulatory measures of tongue twister errors reveal gestural intrusions: target and intrusion elements are co-produced, sometimes resulting in a gestural error which is imperceptible to listeners (Pouplier 2003, Goldstein et al. 2007; see also Mowrey and MacKay 1970). Is this apparent difference due to structure and processing differences between the two utterance types, i.e., sentences (e.g., The top cop saw a cop top) vs alternating repetitive word lists (e.g., top cop top cop top cop) generally produced with quasi-periodic timing? Or, do articulatory measures simply capture the nature of sound-level errors more accurately? We elicited errors using both types of stimuli in the same experimental session; perceptual and acoustic analyses show that sentences provoke more apparent whole-segment substitutions (e.g., /tap/ for /kap/), while alternating repetitive lists provoke more errors with two onset bursts (e.g., /tkap/), resembling gestural intrusions. This suggests that there may be more than one mechanism underlying spoken errors, and that different materials may engage these mechanisms to different degrees.