Ph.D. Dissertation

MORE THAN FRICATION:

COMPENSATORY CUES OF /S/ REDUCTION IN VOICED STOP CONSONANTS IN WESTERN ANDALUSIAN SPANISH

The primary aim of this dissertation is to examine the fricativization of voiced stops as a compensatory process of coda /s/ lenition from both a productive and a perceptual perspective in Western Andalusian Spanish (WAS). It revolves around two main research questions with two distinct goals: first, to provide a linguistic analysis of the phonetic realization of /s/ + voiced stops /b d g/ in WAS; and second, to examine how phonemic /s/ + voiced stops /b d g/ are perceived in WAS.

In order to address the first research question, speech samples were collected from five different speakers from a rural town near Seville, Spain, using a semi-guided sentence completion task. The task was based on minimal pairs of the type “una bota” /u.na.bo.ta/ → [u.na.β̞o.ta] - versus “unas botas” /u.nas.bo.tas/ → [u.na.’vo.ta]. We hypothesize that the singleton fricative allophones [v ð ɣ] will results from the interplay of two phonological processes in WAS, the debuccalization of /s/ and the spirantization of /b d g/ in onset position. Speech samples were acoustically analyzed according to different factors: degree of friction, duration of the segment, intensity, zero-crossings, and spectral moments. Additionally, the production of the Compensatory Voiced Fricatives (CVFs) was also analyzed in terms of different linguistic factors, such as phonological context (preceding and following vowel), tonicity of the syllable, and position of the fricative segment within the syllable. Results from this task show that CVFs exhibit a greater degree of friction (z-crossings and observable energy in higher frequencies) and a longer duration than their approximant counterparts. However, the spectral moments did not yield significant differences in the characterization of the CVFs with respect to the approximant productions of /b d g/. The linguistic factors under consideration also did not seem to have an effect on the productivity of the FLCs.

In order to address the second research question, three different perceptual tasks were administered to 21 listeners from WAS. The first was an AX discrimination task, aimed at addressing the perceptibility of the CVF acoustic nuances among WAS speakers. The results for this task showed that WAS listeners can effectively discriminate between approximant [β̞ ð̞ ɣ̞] vs. fricative [v ð ɣ] productions in isolated speech. The second task was an Identification task that sought to investigate whether WAS listeners can associate the fricative singletons [v ð ɣ] with plurality, hence implying a phonological categorization as allophonic variants of /b d g/, and thus compensating for the lenited /s/. Results for this second task indicated that listeners were indeed able to associate the approximant productions with the singular token representations and the fricative productions with the plural token representations. Finally, the role of segmental duration was assessed in a third perceptual task using a Visual Analog Slider. Five different modifications were made to the duration of the CVFs based on the production data obtained in the first experiment. Results showed that fricative segments of shorter duration were paired with lexical items in the singular form, implying an association of these segments with an intervocalic approximant production [β̞ ð̞ ɣ̞]. In contrast, fricative segments with longer duration were identified with plural lexical items, implying a phonological association of [v ð ɣ] with /s/ + /b d g/, despite /s/ not being physically realized in the stimuli.