UC Berkeley Phonology Lab Annual Report (2011) VOICE-INITIATING GESTURES IN SPANISH: PRENASALIZATION 1 Maria-Josep Sole 1 and Ronald L. Sprouse 2 Universitat Autonoma de Barcelona, Spain, 2 Department of Linguistics, UC Berkeley mariajosep.sole@uab.cat, ronald@berkeley.edu stop voicing is unlikely to occur utterance-initially without additional maneuvers, simply because the pressure difference is not large enough. This difficulty is aggravated by the fact that after a pause the vocal folds must approximate and be duly tensed and glottal vibration has to be initiated rather than sustained, which requires a greater pressure difference – 3-4cm H 2 O vs 1-2 cm H 2 O. The difficulty to achieve voicing during utterance-initial stops is reflected in phonological patterns: a large number of languages lack (or do not require) actual glottal vibration during initial ‘voiced’ stops (e.g., American English, German); neutralization of the stop voicing contrast utterance-initially (to a voiceless stop) is not uncommon (e.g., Uzbek, Tamil, Cuna, Ewondo [2]); and utterance-initial stops have phonologized certain maneuvers (e.g., prenasalization, implosivization, d-lateralization, retroflexion) which help to preserve glottal vibration [3]. Some muscularly- (passive tissue expansion) and articulatorily-induced changes in supraglottal volume (e.g., larynx lowering) or oral/nasal leakage may reduce oral pressure and create the conditions for voicing initiation and continuation [4]. This study analyzes the occurrence of velum leakage in utterance-initial voiced stops in Spanish and the response of the vocal folds to time changes in oral pressure brought about by such nasal leakage. ABSTRACT This work reports the use of nasal leakage to achieve the pressure difference for voicing in utterance-initial Spanish stops. Multiparametric aerodynamic and acoustic data were analyzed for six Spanish speakers. Two main patterns of prenasalization were identified in postpausal voiced stops: (i) delayed nasal closure relative to the oral closure and (ii) a momentary nasal opening (nasal burst) before phonation onset. Quantitative analysis showed that the time between oral and nasal closure was longer for voiced than devoiced stops, and for phonologically voiced than voiceless stops. Velopharyngeal closure was related to phonation onset such that as voicing was initiated, the velum began to raise. The results suggest that occurrence of velum leakage is related to vocal fold vibration in Spanish utterance-initial stops. Keywords: voicing, nasalization, Spanish. utterance-initial stops, 1. INTRODUCTION This study analyzes the occurrence of velum leakage in utterance-initial voiced stops in Spanish and its relationship to stop voicing. As other Romance languages do, Spanish uses voicing during the closure to cue the voiced-voiceless contrast; therefore, speakers must make the necessary adjustments to achieve vocal fold vibration during the stop closure. It is known that closure voicing during an utterance-initial stop is less likely to occur and is typically shorter than medially. This is because the aerodynamic conditions are less conducive to voicing of utterance-initial stops than of medial stops, where subglottal pressure is high and relatively constant [1]. In utterance-initial stops, subglottal pressure rises above atmospheric pressure in a characteristically linear manner, following a similar time course to the oral pressure increase during the stop constriction. Given that the occurrence of voicing depends to a great extent on the difference between subglottal and oral pressure (and thereby airflow through the glottis), 2. METHOD Simultaneous oral pressure, oral airflow, nasal airflow, and audio signal were obtained for ten Spanish speakers, five French speakers, and six English speakers. Only the results for six Spanish speakers (three female (S1, S4, S5) and one male speaker (S3) of continental Spanish; one female Mexican speaker (S2), and one male Uruguayan speaker (S6)) will be reported here. (These 6 speakers showed the higher rate of prenasalized tokens out of the original 10 Spanish speakers). The subjects were instructed to read the following sentences as if they were a dialogue between A and B in order to obtain two isolated utterances, with