Predicting and classifying Polish singleton and geminate consonants at various speech tempos using logarithmic duration

According to Amano et al. [2021, JASA, 150(3), 1830–1843], a linear function with logarithmic durations fairly effectively predicts and classifies Japanese singleton and geminate consonants at various speaking rates. In the current study, we tested whether the same methodology can separate singletons from geminates at changing speaking rates in other languages with gemination, such as Polish. To this end, we investigated the production of 12 Polish singleton/geminate pairs in a carrier sentence by 22 native speakers of Polish. There were seven different speaking rates expressed in BPM, ranging from 16 to 70 BPM. The speakers were required to durationally fit their productions to the metronome visualizations presented on the screen. The regression and discriminant analyses showed that a linear function with two variables—the logarithmic duration of the consonant's closure and the logarithmic average syllable duration—effectively predicts the distribution of singletons and geminates in Polish (R2 = 0.507 and R2 = 0.663, respectively). The results show that a combination of a logarithmic duration of closure or frication and an average logarithmic duration of syllable is an invariant acoustic feature of Polish singletons and geminates against speaking rate variation, and that the linear function with these logarithmic durations can effectively discriminate between singletons and geminates in Polish. [Work supported by the National Science Centre Poland, Grant No. OPUS 2023/51/B/HS2/01429.]