The retraction of /s/, particularly in triple consonant clusters, is the most recent star amongst the vowel-space-crazed research in sociophonetics. It has stimulated debates on phonological processes (Shapiro 1995, Lawrence 2000, Rutter 2011, Baker et. al. 2011) and social patterns (Durian 2007, Gylfadottir 2015, Hinrichs et. al. 2015). While there is some indication for change in apparent time (Gylfadottir 2015, Hinrichs et. al. 2015) in the US, other linguistic and social factors are relatively understudied.
In contrast to previous corpus studies, the current study generated a larger number of tokens per speaker with greater variation in canonical styles. Guised as a memory study, subjects performed picture naming, memory naming, a reading passage and narration. Audio files were partially transcribed and force-aligned (FAVE, Rosenfelder et al. 2011). The retraction rate was calculated by z-scoring each speaker’s Center-of-Gravity measurements for /s/ and /ʃ/ and rating their performance of /s/ in str-clusters in comparison to this range (CoGs measured in Praat, Boersma & Weenink 2012).
If the main predictor for retraction were ease of articulation, variation in style is to be expected due to the cognitive differences and speech rate in the activities performed by the participants, hierarchically structured as follows:
lowest expected retraction-rate → highest expected retraction-rate
reading - naming - remembered naming - narration - free interview
formal → informal
Interestingly, street in the reading task shows a remarkable rate of retraction (CoG measurements as low as 3.5kHz, a typical /ʃ/ CoG (Johnson2003)), which is stylistically the most formal task. In contrast, in the same task striped and stretch show lower rates. Similarly, restaurant presents with a slightly higher retraction rate in the picture-naming-task than straw.
Contrasting this performance, participants name attention paid to speech as predictor of retraction in post-experimental verbal reports. The described retraction in the reading task thus indicates a subconscious lexical diffusion by frequency, a characteristic of gradual sound change (Bybee 2010). I will discuss to what extend this retraction reshapes the assimilation-based approaches to linguistic factors and, combined with the significant social predictors I found, paints a much clearer picture of the complex status of /s/-retraction.
