SRPP: Variation in Naturalistic Speech: The curious case of /ia/ in Portuguese

According to historical and phonological accounts, the vowel sequence /ia/ is produced as a sequence of two full vowels (i.e., hiatus) in Portuguese, while most Romance languages prefer to produce /ia/ as a semi-vowel /j/ followed by a full vowel /a/ (i.e., diphthong). This description is supported by data recorded in laboratory conditions, but there are good reasons to expect that the cross-Romance typology may break down in fluent, spontaneous speech which is subject to significant reduction processes. Using large corpora of radio and TV show recordings, we investigate whether /ia/ is indeed maintained as hiatus in naturalistic speech in Portuguese and whether factors such as proximity of /ia/ to the stressed syllable and position within the word influence its realization. For over 6800 tokens of /ia/, we measured F1, F2, and the duration of the vowel sequence, submitted the signals to Functional Principal Components Analysis, and ran LMERs on the resulting Principal Components scores. One of the main findings is that the acoustic realization of /ia/ in Portuguese ranges from hiatus-like formant configurations to almost monophthongal realizations, mostly depending on /ia/’s proximity to the stressed syllable. While the amount of variation is considerable and likely higher than in laboratory speech, a preliminary cross-linguistic analysis shows that Portuguese, nevertheless, seems to prefer hiatus-like over diphthongal productions of /ia/ in comparison to Italian, Spanish, and Romanian. We will discuss the findings with respect to the benefits and challenges of working with large corpora of naturalistic speech and of using a dynamic vs. a static analysis technique.