Francesco Burroni presented a poster titled "Prominence clash induces localized delays in production, not rhythmic readjustments" at the 17th Biennial Conference of the Association for Laboratory Phonology (LapPhon17), held virtually July 6-8, 2020.

Poster Abstract:

In the phonological literature prominence clash is assumed to induce rhythmic readjustments in the first word of a clashing pair. These effects are attributed to stress shift, aka the Rhythm Rule (e.g. Liberman & Prince 1977, Nespor & Vogel 1977, 1989), or to pitch accent deletion with early pitch accent insertion (e.g. Gussenhoven 1991). Phonetic studies of clash, however, have failed to observe the expected correlates of stress shift (e.g. Vogel et al. 1995, Grabe & Warren 1995) and are not fully compatible with pitch accent-based accounts (e.g. Horne 1990 vs Tilsen 2012).

Italian, for example, is often assumed to display clash-driven rhythmic effects (e.g. Krämer 2009), but there is just a single phonetic study, limited to two speakers, that supports this view (Farnetani & Kori 1983). We conducted two acoustic studies with Italian speakers and found no evidence of prominence shift or deletion in measures of duration, F0, or intensity.

To the contrary, we observed an increase of duration in the final vowel/syllable of word 1 in clashing word pairs. We also did not observe any effects on word 2 in clashing pairs. Accordingly, we argue that the main correlate of clash is a localized slowing of speech rate, not prominence shift or deletion.

Methods:

In Experiment 1, 16 speakers of Italian produced 10 x 36 unique trials elicited from visual stimuli (i.e. pictures) representing a three-word noun phrase consisting of a numeral (w0), a target noun with final stress (w1, caffè ‘coffee’, città ‘city’, and colibrì ‘hummingbird’), and a color term with different stress used to manipulate clash (w2).

In Experiment 2, 8 speakers of Italian produced an identical number of similar sequences, with stress varied in w1 (e.g. colibrì ‘hummingbird’ vs. colùbro a type of snake vs. càlibro ‘caliber’). The effects of clash were analyzed with Linear Mixed Effects regressions, with word and speaker as random factors.

Results:

In Experiment 1 we found no significant effects of clash on duration, RMS intensity, or F0 of the initial vowel/syllable of w1. For example [ka] in caffè is not different before néri /vérdi vs bordó /marróni. Surprisingly, durations of the final vowel/syllable of w1 were longer in clash environments, contra the predictions of rhythmic readjustment analyses. For a subset of participants, vowel formants were also more extreme in clash, suggesting hyper-articulation.

Experiment 2 replicated the findings of Exp. 1 and in addition it was found that clash had no effect on the duration, F0, intensity, or formants of the initial vowel of w2. For example, néri after colibrì is not different from néri after colùbri. This shows that clash effects are not manifested on the second word of the clashing pair.

Conclusions:

Rhythmic analyses of prominence clash were not supported by our data: phonetic evidence for prominence shift or prominence deletion was not observed. Instead, the effects of clash on duration and vowel quality in the final syllable of w1 indicate slowing of speech rate and hyper-articulation.

The observation that clash has effects on w1 but not w2 has consequences for analyses based on prosodic boundaries, which can be adapted to generate a variety of predictions. We discuss how the observed effects of prominence clash can be modeled in the framework of Articulatory Phonology (e.g. Gafos 2006, Tilsen 2019).

Francesco Burroni, Sireemas Maspong, Dr. Pittayawat  Pittayaporn, and Pimthip Kochaiyaphum presented a poster titled "Interactions of initial geminates and stress: a case study of Pattani Malay" at the 17th Biennial Conference of the Association for Laboratory Phonology (LapPhon17), held virtually July 6-8, 2020.

Francesco and Siri's poster was co-authored with Dr. P. Pittayaporn (a Cornell Phonetics Lab alumnus) and Pimthip Kochaiyaphum of Chulalongkorn University, Bangkok, Thailand

Paper Abstract:

Pattani Malay is an oft-cited example of a language with moraic initial geminates (Hajek & Goedemans, 2003; Topintzi, 2008), a claim based on an alleged stress shift process.

This paper provides empirical evidence against previous impressionistic descriptions (e.g., Yupho, 1989) that initial geminates cause stress to move from the final to the initial syllable of disyllables. Phonologically speaking, a larger vowel distribution and absence of the neutral /ɨ/ in the final syllable point to prominence on the final syllable regardless of the type of onset.

Phonetically, our acoustic analysis also reveals that words with and without initial geminates display similar acoustic profiles. More specifically, the final syllable always has significantly longer duration in all utterance positions, and significantly higher F0 in phrase-medial positions.

These results indicate that word-level prominence in disyllables in this language is not attracted by initial geminates but remains fixed on the final syllable. Therefore, the claim that Pattani Malay is a language with moraic onsets is not tenable.

Rachel Vogel presented a poster titled "Dissociating prosodic and segmental cues of narrow focus in Nepali" at LabPhon 17, held virtually July 6-8, 2020.  

Poster Abstract:

Cross-linguistically, narrow focus is known to have both segmental and suprasegmental phonetic cues (e.g., Ladd 1980, Avesani et al. 2007, Féry and Kügler 2008, Mücke and Grice 2014, DiCanio et al. 2018). Two ongoing questions in this area are how these cues are related and which prosodic units serve as domains of their realization.

I investigate these issues in Nepali, an Indo-Aryan language with relatively understudied prosodic phonology, looking particularly at whether phonetic cues of narrow focus take as their domain a single prominent syllable or some larger unit. It is demonstrated that the suprasegmental cues tested span the entire word, while the segmental cues are restricted to positions that are otherwise prone to lenition in the absence of narrow focus.

More generally, these findings suggest that different types of acoustic cues of focus can pattern independently, and they need not be restricted to the stressed syllable of a word. Method. The study uses data produced by four native Nepali speakers in an experiment testing 32 real disyllabic words with dental stops /t, tʰ, d, dɦ/ in either word-initial or intervocalic position. Nepali is said to have weight sensitive word-level stress falling within the first two syllables (or first foot) of the word (Acharya 1990); all words tested here are expected to have stress on the first syllable.

The target words were embedded in two types of question-answer dialogues: the target was under narrow focus in the answer of one type, and in a backgrounded position preceding a verb with contrastive focus in the other. The target vowels were measured for F0 (z-scored by speaker) and duration (z-scored by speaker and vowel quality). The prevocalic interval duration (PVI) - used instead of VOT in languages with voiced aspirated stops (Schwarz et al. 2019) - was measured for all the stops. Since deaspiration of aspirated stops has been observed in Nepali as a type of lenition, /th/ and /dɦ/ tokens with PVIs within two standard deviations of the mean of their unaspirated counterparts were coded as deaspirated. Consistent aspiration of phonemically aspirated stops and longer PVI for all stops were taken as a manifestation of relative segmental strength; greater F0 and longer vowel durations were taken to reflect suprasegmental prominence.

Results. Regarding suprasegmental properties, Figure 1 shows that narrow focus significantly enhances vowel duration in both syllables (confirmed with a t-test; p < 0.005). Additionally, in the non-focus condition, the first vowel is significantly longer than the second (p = 0.005). This difference disappears under focus, however, reflecting a relatively greater lengthening effect on the second syllable.

Figure 2 shows that narrow focus significantly raises F0 in both syllables (p < 0.001) and that F0 is higher in the second syllable than in the first in both conditions (p < 0.005). The only significant effects of narrow focus on the segmental properties appeared with the aspirated stops.

Figure 3 shows that in the non-focus condition, /tʰ/ and /dɦ/ have significantly shorter PVIs in medial position than in initial position (p < 0.005), indicating intervocalic, or possibly foot-medial, lenition. Medial /dɦ/ exhibits particularly substantial lenition with a PVI distribution closer to that of /t/ and /d/, and with 37.8% of its tokens deaspirated. As a comparison of

Figures 3 and 4 shows, focus has no effect on initial /tʰ/ and /dɦ/, but lengthens PVI of both intervocalically, such that there is no longer a difference between PVI durations in word-initial and medial positions. There are also no instances of deaspiration under focus. Discussion. In sum, the segmental results show an anti-lenition effect of narrow focus, by which the word-medial stops prone to lenition in the absence of focus (/tʰ/ and /dɦ/) are strengthened. This is concentrated on the unstressed syllable. Suprasegmental effects on the other hand are spread throughout the word, enhancing duration and F0 of both syllables.

These findings indicate that segmental and suprasegmental cues of narrow focus are distributed differently from one another in Nepali, and prominent positions within a word are not necessarily privileged for narrow focus. Rather, it is a larger unit, which seems here to be the full word or a foot, that manifests its effects.

Francesco Burroni and Simone Harmath de Lemos presented a paper titled "A unified account for morphologically governed stress systems in Romance" at the 50th Linguistic Symposium on Romance Languages (LSRL 50), held virtually July 1-3 and July 6-8, 2020

Paper Abstract:

A still debated question in Romance phonology is whether the stress systems of Romance varieties are weight-sensitive. Italian and Brazilian Portuguese are two well-known examples of varieties where both weight-sensitive and weight-insensitive analyses have been put forth in the literature.

In this paper, we show that the assumptions required by weight-sensitive analyses of these languages do not with-stand closer scrutiny and that their empirical coverage is limited. Accordingly, capitalizing on a feature that was already present in some weight-sensitive analyses– underlying accents –, we present a new account based on lexical accents tied to morphemes.

We show that this analysis has both well-motivated assumptions and wider empirical coverage. Further, we show that an analysis of Italian and Brazil-an Portuguese as morphological lexical-accent systems can readily be extended to other Romance varieties as well, in a hitherto unexplored (quasi-)pan-Romance perspective.

Given these shared similarities across Romance varieties, we theorize that a morphological accent system arose in Proto-Romance after the dissolution of the weight-sensitive system of Latin. We then return to our key assumption, the categorical presence/absence of lexical accent on morphemes, and we show that it is supported experimentally by the results of a pilot nonce-word stress assignment study in Italian.

We conclude by sketching out two possible models of the lexical accent systems, one couched in Optimality Theory and one couched in a connectionist model of stress assignment.