TL;DR
This study investigates the phenomenon of /s/-retraction (where the 's' sound in clusters like "str" shifts towards a "sh" sound) in Trinidadian English (TrinE) and American English (AmE). It reveals that this sound change is present in TrinE, particularly in the /str/ cluster, and is more pronounced in younger speakers, suggesting it's an active sound change, though potentially influenced by TrinE's unique phonological features like /t/-affrication rather than solely by American English influence.
What Problem Does This Paper Address?
The paper addresses a gap in sociophonetic research concerning sound changes in "New Englishes" – nativized varieties of English spoken in postcolonial territories. While the retraction of /s/ in clusters like /str/ has been well-documented in British, Australian, and American English, its presence and conditioning factors in New Englishes, specifically Trinidadian English, remained unexplored. The authors aim to determine if this sound change exists in TrinE, how it's influenced by phonetic and social factors, and how it compares to its manifestation in American English, especially given the significant cross-varietal linguistic influence from AmE in the Caribbean.
Methods
The study employed a large-scale comparative acoustic analysis of sibilant sounds. Key methods include:
- Acoustic Measurement: Center of Gravity (CoG) measurements were used to quantify the degree of sibilant retraction. CoG is a reliable metric for measuring the spectral distribution of a sound, with higher values indicating more forward articulation and lower values indicating more retracted articulation.
- Data Processing: Recordings were automatically force-aligned using FAVE (Forced Alignment and Vocalization Environment) and its recalibrated version for TrinE. Spectral CoG measurements were extracted from sibilant tokens using a Praat script.
- Outlier and Normalization: Tokens with short durations (<40ms) were excluded. Tukey's method was used to remove outliers on a per-speaker and item basis. To control for physiological variations, each token's CoG was normalized using Lobanov z-transformation relative to the speaker's sibilant space.
- Sociolinguistic and Linguistic Coding: Tokens were coded for various factors, including place of articulation of preceding sounds, whether /sC(r)/ clusters occurred across word boundaries, and speaker demographics (age, gender, ethnicity).
- Statistical Analysis: Mixed-effects linear regression models were used to analyze the data, with manual variable selection employed to achieve model parsimony.
Datasets
The study analyzed sibilants from two speech corpora:
- Trinidadian English (TrinE) Corpus: Recordings from 100 speakers (35 teachers, 65 upper-level students) from 12 secondary schools across Trinidad.
- American English (AmE) Corpus: Interview recordings from 81 speakers, mostly born in Texas and currently residing in Austin.
A total of 23,602 z-scored CoG measurements from 181 speakers were analyzed. The paper does not introduce a new benchmark dataset but rather utilizes existing corpora for a comparative analysis.
Key Insights and Ideas
- /str/-Retraction is Present in TrinE: The study confirms that /s/-retraction, specifically in the /str/ cluster, is a significant feature of Trinidadian English, with the sibilant being acoustically retracted towards [].
- Phonetic Conditioning is Key in TrinE: The retraction of /str/ in TrinE is primarily conditioned by its position within a word. Clusters occurring across word boundaries are acoustically higher (less retracted) than those at the beginning of a word.
- Social Factors Less Influential in TrinE: Unlike in some other varieties, social factors (style, ethnicity, age) did not show statistically significant effects on /str/-retraction in the TrinE dataset, suggesting a relatively homogenous pattern across different social groups.
- TrinE vs. AmE Sibilant Spaces: While the overall distance between /s/ and /ʃ/ (sh) is similar in both TrinE and AmE, sibilants in TrinE generally have lower acoustic values (are more retracted) than their AmE counterparts.
- Potential Role of /t/-Affrication in TrinE: The authors hypothesize that TrinE's unique feature of /t/-affrication (where /t/ sounds like "ts" or "ch") might be a catalyst for the pronounced /str/-retraction observed, potentially influencing the sibilant more strongly than in AmE. This suggests the sound change in TrinE might be driven by internal phonological processes rather than solely by external AmE influence.
- Age as a Factor (with caveats): While younger speakers in both varieties show a tendency towards greater /str/-retraction, this effect was not statistically significant in TrinE for all age groups, leading the authors to question if /str/-retraction is a current sound change in TrinE or a stable sociophonetic feature.
Key Results
- In Trinidadian English, the /str/ cluster is considerably retracted towards [].
- Other /sC(r)/ clusters in TrinE are generally not retracted and acoustically similar to singleton /s/.
- Acoustic lowering of /str/ in TrinE is significantly larger than in American English.
- In TrinE, /str/ clusters across word boundaries are acoustically higher (less retracted) than word-initial ones.
- While age differences were observed, they were not statistically significant overall in TrinE, suggesting a stable sociophonetic feature rather than a clear sound change in progress driven by age.
Limitations and Future Work
- Statistical Significance of Age in TrinE: The lack of statistically significant age effects in TrinE leads the authors to question whether /str/-retraction is a current sound change or a stable sociophonetic feature in this variety. Further research is needed to clarify this.
- Anecdotal Evidence for /t/-Affrication: The hypothesis that /t/-affrication in TrinE contributes to /str/-retraction is based on anecdotal evidence. More conclusive research is required to confirm this link.
- Cross-Varietal Comparisons: While this study compares TrinE and AmE, the authors emphasize the need for broader cross-varietal investigations to understand the productivity and universality of this pronunciation pattern across different English varieties.
- Explaining Differences: Further research is needed to explore other linguistic patterns in English varieties, such as /t/-affrication, to fully explain differences in the status of /str/-retraction as a sound change.
Summary updated