Segmental vs phrase-level creak in Polish: An acoustic analysis

In languages without phonation contrasts, the appearance of creaky voice may not only be a phrase-level phenomenon that may be speaker-dependent but can also have clear phonological origins related to production of single segments. The present study investigates acoustic differences between segmental and phrase-level instances of creaky phonation in Polish. Segmental creak, which in Polish derives from the glottalization of word- or syllable-initial vowels, is compared with phrase-final creak, across ten female speakers. Using VoiceSauce, several acoustic parameters related to fundamental frequency, spectral tilt, spectral noise, and harmonic amplitude were extracted from annotated speech samples. Linear mixed-effects models revealed significantly lower spectral tilt (H1*-A1*, H1*-A2*, H1*-A3*, H4*-H2k*) and higher harmonic amplitudes (A1*, A2*, A3*) for segmental creak. We suggest that the spectral tilt results, which indicate more robust harmonics in the vicinity of vowel formants, may be motivated by a need to produce perceptible vowel quality in a prosodically prominent position.