Felix Kpogo

INVESTIGATING SOUND CHANGE IN TWI VOWEL HARMONY: A SOCIOPHONETIC STUDY OF AGE, GENDER, AND LOCALITY EFFECTS

FELIX KPOGO

Boston University, Graduate School of Arts and Sciences, 2024
Major Professor: Charles B. Chang, PhD Associate Professor of Linguistics

ABSTRACT

This dissertation employs a sociophonetic approach to document sound change in Asante Twi, a language spoken in Ghana. Asante Twi is one major variety of the Akan language. Traditional descriptions of Twi phonology say that [æ] is allophonically derived when /a/ occurs before advanced tongue root (i.e., [+ATR]) vowels /i, u, o/ and palatal consonants. For example, in the word /sika-ni/ ‘rich person’, the syllable-final /a/ is pronounced as [sikæni] before the [+ATR] vowel in the person suffix, {-ni}.

However, a recent study observe an [e] vowel (itself the phonemic [+ATR] counterpart of /ɛ/) in the phonological context where [æ] is expected among urban dwelling Twi speakers (Yankson, 2018). Thus, the traditional descriptions and the recent evidence are at odds since the former does not reflect the way native speakers are currently speaking in the urban area. The dissertation aims to explore how Twi speakers produce and perceive the [æ] vowel, particularly focusing on the language-internal and –external factors constraining its production and perception. The data comprised 105 Asante Twi speakers spanning various age brackets (12 – 75 years) and genders residing in either traditional or urban Twi-speaking communities. Production data were obtained via a picture naming task featuring target items where /a/ preceded a [+ATR] vowel or palatal consonant, alongside items with phonemic [+ATR] /e/ to diagnose the sound change. Acoustic measurements (F1 and F2) at the midpoint of target vowels aided in diagnosing the potential sound change.

Perception data were gathered through a forced-choice identification task to ascertain whether speakers maintained distinct mental representations for [æ] and /e/. The production results indicate a reorganization of [æ] vowel harmony triggers, as well as the raising of the [æ] vowel, leading to its encroachment upon phonemic /e/ vowel in the vowel space. Specifically, traditional speakers predominantly adhere to the traditional harmony pattern, wherein /a/ is primarily articulated as [æ] before /i, u/, and palatal segments (i.e., [+ATR, +high, -low] contexts), except a few who produce an [e]. However, urban speakers produced /a/ as [e] in the same environment. The [+ATR, -high, -low] /o/ context inconsistently triggered ATR harmony in both speaker populations. An analysis of a potential merger between the [+ATR] counterpart of /a/ and the phonemic /e/ showed that traditional speakers, across various age groups, generally produced acoustically distinct vowels. However, among the youngest traditional speakers, a few raised their [æ] productions to align with /e/. In contrast, urban speakers of all generations displayed [æ]-raising and merger with /e/, with the youngest age group showing nearly complete merger. Regarding speaker gender, urban
men were more advanced than women in the [æ]-/e/ merger. Nonetheless, some younger women in the traditional community appeared to converge on the merger observed in the urban community.

Furthermore, the analysis of perceptual data indicated that the perceptual pattern of the merger was influenced by speakers’ own production. Individuals who exhibited merger in production tended to show a similar pattern in perception (i.e., “perception-and-production conflaters”). Other participants showed either production merger but not perception merger (i.e., “production-conflaters”) or a perception merger but not production merger (i.e., “perception-conflaters”/ near-mergers). Interestingly, the study did not find a significant relationship between production and perception, suggesting a potential co-evolving connection between the two modalities that can lead to perception-production alignment or misalignment, in which either perception or production can be more advanced in the sound change than the other modality. Overall, the findings of the dissertation suggest an ongoing sound change in these communities.

This dissertation provides the initial empirical evidence of sound change within the Asante Twi vowel harmony system. It also contributes significantly to the variationist sociolinguistic literature by presenting data from a non-Western
community, challenging some longstanding assumptions on linguistic change from Western societies. Moreover, it enhances language representation in the variations literature by introducing data from an understudied language with unique linguistic features, such as ATR harmony, thereby advancing our understanding of linguistic variation and change. Additionally, it constitutes the most comprehensive collection of sociolinguistic and sociophonetic studies of Asante Twi speakers to date. While it covers a wide range of topics, several questions remain unanswered, and it is anticipated that the findings will inspire numerous future studies in other understudied languages and communities in Africa and beyond.