Distribution of intermittent episodes of creaky voice in relation to prosodic constituents

Abstract

The aims of this study were to investigate the use of creak in relation to prosodic constituents. Creak is a phonation type (Laver, 1980) that is easily perceived by listeners (Blomgren et al, 1998). Para linguistically, creak signifies mood, emotions and attitude (Gobl and Ní Chasaide, 2003), extra linguistically, it is a social identifier (Wolk et al, 2012; Stuart-Smith, 1999). Linguistically it is phonemic in some languages (Gordon and Ladefoged, 2001) but also manifests at points of prosodic significance such as utterancefinal position (Redi and Shattuck Hufnagel, 2001). In addition to this, creak may indicate unhealthy vocal habits or an early-stage voice problem (Colton et al, 2011; Mathieson, 2001). The rationale for the study, therefore, was to provide normative information regarding the occurrence of this complex set of phenomena, which may be of use in the decision-making process of speech and language therapists. A pre-existing corpus of 15 female, south-east English speakers was prosodically annotated using the ToBI framework (Beckman and Ayers Elam, 1997; Silverman et al, 1992; Beckman and Pierrehumbert, 1986). The data comprised a reading passage of approximately one minute in length. The speakers were between the ages of 19 and 38 years; English was their native language. Tonal categories were established, based on the tones perceived in each word, its divisibility within the speech signal, and their overall frequency. The data were perceptually analysed for creak. The main findings suggest that creak occurs most often in tonal categories containing boundary tones, particularly when low, and in phrase-initial position. A wide range of speaker variation was also attested in the data. Findings support current literature suggesting that creak occurs in a non random distribution.