The article presents an empirical study on the laryngeal pattern of English obstruents in ‘voiceless’ obstruent clusters (VOCs) in light of Szigetvári’s (2020) *FortisFortis (*FF) phonotactic claim, which posits that adjacent fortis obstruents are prohibited within morphemes due to articulatory constraints. Using acoustic data from our target words: absolute, rhapsody, and knapsack gathered via YouGlish (n.d.), we analyze the correlates of aspiration, segment durations, various vowel ratios, and voicing facts in Praat (Boersma and Weenink 2025), in order to test the validity of the *FF hypothesis. Rigorous statistical analysis reveals that the VOC of rhapsody patterns with that of absolute, indicating phonological /bs/, rather than /ps/. Our findings provide support for the *FF constraint and suggest that English transcription conventions sometimes obscure phonological reality.
Varieties of English differ in (i) the amount of aspiration/voicing in plosives, traditionally measured in Voice Onset Time (VOT), and (ii) the degree of vowel clipping before fortis consonants (the Voicing Effect). The two seem to correlate with each other, with a possible functional explanation, but they also interact with the phonological status of vowel length within the dialect. To eliminate this latter effect, we investigate two varieties, Pākehā and Māori New Zealand English, which share the vowel system. We measure vowel duration and plosive VOT in Praat, conduct statistical analyses using the lme4 package of R and show that correlations between the two features exist independently.
American English is well-known for optional /t d/ deletion in clusters, as in best [bes] man, kept [khep] going or whipped [wɪp] butter. However, research shows that even in the absence of an acoustic trace of /t d/, articulatory gestures related to the alveolar constriction are still there, although they may be of a smaller magnitude, which questions the completeness of deletion from a phonological standpoint. This study uses motion capture to look at /t d/ trapped in between consonants vs a neutral context and explore the interplay of acoustics with articulation. The results show significant differences between sequences with deleted and retained coronals in both lips and acoustic measurements. Some gradient effects and weakening strategies are also discussed, pointing to the need for a more wholesome approach to the study of (apparent) sound deletion in the context of overall articulatory strategies.
/tr/ onset clusters in Polish and English show different phonetic realizations. Contra previous accounts locating the cause for this in a different gestural timing or a different phonological representation, we propose the two languages employ different perceptual cues necessary for the perception of consonants in language-specific clusters: in English, voiceless plosives are followed only by sonorants in onsets, and the sonorant release sufficiently cues the plosive, while in Polish, with its much larger cluster inventory, an audible release of a plosive followed by another plosive is necessary to identify the place of articulation of the first plosive. A formalization of the mapping between /tr/ and its language-specific cues is provided and employed for native and non-native perception. The resulting predictions are compared to existing perceptual studies.
There is a general consensus in the phonological literature that the English affricates /tʃ/ and /dʒ/ are unitary segments rather than clusters of a stop and a fricative. However, for /tʃ/, not all the evidence is consistent with this approach. While /tʃ/ patterns as a single segment in linguistic performance, some authors on British pronunciation have pointed out that glottalization targets /tʃ/ as if it were a cluster (Higginbottom 1964; Gimson and Cruttenden 2001). This paper proposes that the conflicting evidence can be resolved under a stratal model (Bermúdez-Otero 2018a). We argue that /tʃ/, in the relevant dialects, is a cluster at the stem level but surfaces as a monosegmental affricate in the course of the derivation. The behaviour of /dʒ/, on the other hand, suggests that it is a single segment at all levels of representation.
