1. Phonetic reduction, perceptual illusions, and phonotactic legality Alex McAllister and Matthew CarlsonYLSS, February 18, 2016 I NTRODUCTION : Spanish exhibits a well-known prohibition of word-initial /s/-consonant clusters (henceforth #sC). In repairing these structures, it employs a productive prothesis strategy introducing the vowel /e/. This strategy has been found in both production and perception studies. Production: Where necessary (e.g. in foreign loanwords such as snob  esnob) #sC clusters are productively repaired by adding a prothetic /e/ (Harris, 1983; Hooper, 1976). Perception: This strategy has been linked to perception of an illusory /e/, but not other vowels, preceding #sC sequences (Carlson, et al., 2015; Cuetos, et al., 2011). However, phonetic reduction in otherwise well-formed words, such as reduction or deletion of the initial vowel in Spanish words beginning with #VsC can yield apparent illicit sequences (#sC) in speech. Example: While the legality of the remaining sequence does not appear to be an obstacle to such reduction (Davidson, 2006), there may nonetheless be a relationship between phonetic reduction and the presence of a strongly preferred repair strategy, such as /e/-prothesis. In this study we probe this relationship by asking whether reduction of the initial vowel in Spanish #esC words like espalda ‘back’, which matches the default repair vowel, is more prone to reduction than other initial vowels, such as in aspirina. D ATA AND M ETHODS : 15 native speakers of (10 female, 5 male) of Andalusian Spanish with minimal L2 knowledge given a speeded, single word readaloud task administered via laptop. Each speaker repeated 40 lexical noun items (10 /e/s-consonant, 10 /a/s-consonant, 20 fillers) twice in semi-randomized order in which no token was repeated twice consecutively. Test items were Spanish words split evenly by/e/ and /a/ and contained the following consonant clusters: Ultimately, 578 valid tokens (288 with initial /a/ and 290 with initial /e/) were collected and coded for the duration of the word initial vowel and the following /s/ using PRAAT (Boersma & Weenink, 2015). Vowel onset was set at the beginning of a visible waveform prior to the strong frication indicative of /s/ and offset was taken to be the onset of strong frication or aspiration, or when /s/ was absent, the onset of the following consonant closure A NALYSIS : For the purposes of this study, a logistic mixed effects regression with crossed random effects by Subject and Word, and with the maximal random effects structure justified by the design was implemented. R ESULTS : D ELETION 1.Outright vowel deletion, defined as the absence of any vocalic material preceding the /s/ was uncommon, occurring in 3% of tokens overall. 2.However, initial /e/ was deleted significantly more often (5%) than initial /a/ (0.3%, one token) 3.Current findings support hypothesis that the prothetic repair vowel is more prone to deletion than other vowels in this context R ELATIONSHIP B ETWEEN V OWEL AND / S / DELETION 1.To analyze the relationship between /s/ lenition (common in Andalusian Spanish) and vowel reduction, durations of both the vowel and the fricative were analyzed. 2.The results found below show a negative relationship between the duration of the initial vowel and that of the following /s/. Figure 3: Durations of initial /e/ and /a/ as a function of duration of the following /s/. (Ribbons show 95% confidence intervals). 3.Reduction of the vowel was accompanied by a comparatively longer /s/, and vice versa. 4.Crucially, a robust interaction showed that this relationship held only when the initial vowel was /e/, but duration of /a/ was not related to the duration of following /s/. 5.Interaction remained robust even when tokens with zero duration of one segment or the other were removed. 6.Overall, initial /e/ reduction (the prothetic vowel) was more pronounced than /a/ reduction, but /s/ lenition appears to dampen /e/ reduction: when /s/ was at its minimum, initial /e/ was as long as /a/. D ISCUSSION : 1.Findings provide compelling evidence for a relationship between automatic phonotactic repair strategies and phonetic reduction in speech. 2.It may be that reduction is enabled when the reduced material can be restored via perceptual repair. 3.It may also be that articulatory, frequency, or other properties favoring reduction of the sequences such as /es/ make those sequences good candidates for phonotactic repair, or contribute to perceptual illusions (e.g. Davidson & Shaw, 2012). 4.The present findings thus shed light on how speech perception and production dynamically influence phonological systems. F UTURE D IRECTIONS : 1.Inclusion of other items (i.e. verbs) may provide a larger data set and thus more favorable results. 2.A corpus-based analysis with lower levels of attention paid to speech may result in larger levels of lenition. W ORKS C ITED : Abrahamsson, N. (1999). Vowel Epenthesis of /sC(C)/ Onsets in Spanish/Swedish Interphonology: A Longitudinal Case Study. Language Learning, 49(3), 473–508. Boersma, Paul & Weenink, David (2015). Praat: doing phonetics by computer [Computer program]. Version 6.0.07, retrieved 10 November 2015 from http://www.praat.org/http://www.praat.org/ Carlson, M. T., Goldrick, M., Blasingame, M., & Fink, A. (2015). Navigating conflicting phonotactic constraints in bilingual speech perception. Bilingualism: Language and Cognition, 1–16. Cuetos, F., & Hallé, P. (2011). Perception of prothetic/e/ in # sc utterances: Gating data. Oral communication at the 17th ICPhS. Hong Kong, China. Davidson, L. (2006). Schwa elision in fast speech: segmental deletion or gestural overlap? Phonetica, 63(2-3), 79–112. Davidson, L., & Shaw, J. A. (2012). Sources of illusion in consonant cluster perception. Journal of Phonetics, 40, 234–248. Dupoux, E., Kakehi, K., Hirose, Y., Pallier, C., & Mehler, J. (1999). Epenthetic vowels in Japanese: A perceptual illusion?. Journal of experimental psychology: human perception and performance, 25(6), 1568-1578. Harris, J. W. (1983). Syllable Structure and Stress in Spanish. A Nonlinear Analysis. MIT Press, Cambridge MA. Hooper, J. B. (1976). Word frequency in lexical diffusion and the source of morphophonological change. Current progress in historical linguistics, 96-105.