Commonly referred to as pinjam “changing tones”, tones in Cantonese are subject to morphologically conditioned changes that mark inflectional and derivational morphology. Through an Optimality Theoretic (OT) analysis (McCarthy & Prince, 1995; Prince & Smolensky, 2008), this study demonstrates that these tone changes are instances of grammatical tone. Two main types of grammatical tones that differ in their underlying representation of tonal morphemes are identified, but their surface realizations are derivable from the same phonological constraints. The results challenge the notion that grammatical tones are only common among languages in Africa and the Americas (e.g. Rolle, 2018).

As inhabitants of a former British colony in East Asia, Hongkongers have developed myriad ways to construct their linguistic identities. Some identify themselves as speakers of ‘Hong Kong English’ (HKE), the local variety of English, others as speakers of non-local Inner Circle varieties like ‘British’ or ‘American English’, which are often seen as more prestigious than HKE, or even as speakers of no specific variety at all (Groves, 2011; Hansen Edwards, 2019, 2023). However, one may question whether their self-identification faithfully reflects their language production. Do English speakers in Hong Kong really speak what they claim to speak? Through a phonetic analysis of sociolinguistic interviews and word list reading tasks from 39 fluent English speakers from Hong Kong (22 women, 17 men, aged 18-58), this study examines the relationship between vowel production and identification with HKE. The target sociolinguistic variables are the merger of high front vowels FLEECE and KIT, and the merger of low back vowels LOT and THOUGHT. Both have been described as characteristic features of HKE (Deterding et al., 2008; Hung, 2000, 2012), but they are either absent or realized with a different phonetic target in Southern British English and American English, the two varieties with the most overt prestige in Hong Kong (Chan, 2016). 16,310 FLEECE tokens, 15,160 KIT tokens, 8,046 LOT tokens, and 3,836 THOUGHT tokens from the interviews, and 566 FLEECE tokens, 579 KIT tokens, 570 LOT tokens, and 576 THOUGHT tokens from the word list were analyzed. The Lobanov-normalized F1 and F2 measurements were tested for statistical significance using linear mixed effect regression models (Kuznetsova et al., 2017) and Pillai scores derived from multivariate analysis of variance (MANOVA) in R (Hay et al., 2006; Nycz & Hall-Lew, 2013).

The results indicate a partial alignment between linguistic identity and vowel production, which are presented in Figure 1 and 2. Speakers who fully identify with HKE exhibit a significantly greater degree of merger between the two high front vowels than speakers with partial or no affiliation to the variety, but no such correlation can be found for the low back vowels. Speakers regardless of their linguistic identity exhibit a high degree of merger between LOT and THOUGHT by realizing both as a mid back vowel, which is consistent with previous descriptions of HKE. Thus, the high front merger is used by English speakers in Hong Kong to index HKE identity but not the low back merger. Despite the stigmatized status of HKE, not all of its phonological features are avoided by English speakers in Hong Kong who do not identify with it. This study provides insights on how identity construction impacts pronunciation in the ‘Outer Circle’ of World Englishes, former British and American colonies where indigenized English varieties and non-local prestige varieties coexist in the linguistic landscape (Kachru, 1985).

English, a global lingua franca, is spoken by millions of people worldwide as an L2. In Outer Circle countries like Nigeria and the Philippines, it is caused by British and American colonization, whereas in Expanding Circle countries like the Netherlands and Japan, it is the result of globalization of English but without direct colonial influence (Kachru, 1986). Despite the ubiquity of L2 English varieties, there is yet to be a unified model that explores their sociophonetic variation, especially one that incorporates theories of both second language acquisition and variationist sociolinguistics. Thus, I propose a new model for analyzing variation in L2 English varieties that considers both the developmental factor of language proficiency and the sociolinguistic factor of norm orientation.

As predicted by models of L2 phonology acquisition (Best & Tyler, 2007; Flege, 1995), L2 speakers are influenced by their L1 phonology, causing a transfer of L1 phonological features into their L2. If they take the norm of L1 English speakers (e.g., British English) as their target of acquisition, proficiency should mediate inter-speaker variation, with greater convergence to L1 varieties with increasing proficiency. While the phonology of L2 speakers is almost never identical to that of L1 speakers due to developmental constraints (Flege et al., 1999; Piske et al., 2001), the speech production of speakers with higher proficiency may be closer to L1 norms than speakers with lower proficiency due to lesser degree of L1 transfer.

Then, as demonstrated by Schneider’s Dynamic Model (2003, 2007), L2 English-speaking communities may undergo endonormative stabilization in which a local standard of English replaces L1 varieties to be the norm. In other words, L2 English speakers do not always treat L1 varieties as their goal of acquisition. Depending on how fully developed this local norm is, the relation between proficiency and similarity to L1 norms may differ. In societies where the L1 norm persists, there would be a clearer trend of convergence to L1 varieties with increasing proficiency. In societies where a local norm exists however, one may observe less convergence, and deviance from L1 norms may persist among speakers with higher proficiency as L1 varieties are not their goal of acquisition.

To test the validity of this model, I will examine vowel production of L2 English speakers in societies with (e.g., the Philippines (Borlongan, 2016)) and without (e.g., the Netherlands (Edwards, 2014)) a local norm of English. Speakers of different proficiency levels will participate in a sociolinguistic interview and an English proficiency test. To minimize bias against non-L1 varieties of English, a new assessment method will be developed and employed based on the guidelines of Canagarajah (2006). Formant measurements of all monophthongs and diphthongs will be extracted from the interviews, and generalized additive mixed models will to be used to test the effects of proficiency and norm orientation on the degree of similarity of formant trajectories to Received Pronunciation and General American English, the two L1 models commonly used in English education worldwide (Dziubalska-Kołaczyk & Przedlacka, 2008).

This study examines the tonal alternation of the proximal demonstrative suffix –na in Dagaare, a Mabia language from the Niger-Congo family spoken in Ghana and Burkina Faso. Data was collected from a 27-year-old native speaker of the Nandome dialect of Dagaare. All elicitation sessions were recorded at a quiet location at Georgetown University. A corpus of phrases with the proximal demonstrative in different phonological environments was elicited. The tones were determined with reference to the F0 measurements in Praat (Boersma & Weenink, 2022).

The demonstrative construction in Dagaare is found to have the basic structure of [definite article prefix + noun root + demonstrative suffix]. The proximal demonstrative suffix –na is analyzed as toneless underlyingly, but it receives a surface tone opposite to the last underlying tone in the noun root to satisfy the Obligatory Contour Principle (OCP). When the last underlying tone of the noun is H, e.g., /à-dɔ́ɔ́-na/ ‘This man’, the demonstrative surfaces as a L tone -nà, e.g. [à-dɔ́ɔ́-nà] ‘This man’. It also applies when the last TBU of the noun root is underlyingly toneless as in /à-bíe-na/ ‘This child’, which surfaces as [à-bíè-nà] ‘This child’ with both the last TBU of the noun root and the demonstrative suffix receiving a L tone. On the contrary, when the last underlying tone of the noun is L, e.g., /à-nàà-na/ ‘This chief’, the demonstrative surfaces as a H tone -ná, e.g. [à-nàà-ná] ‘This chief’. However, there are exceptions in which the noun root alternates in tone but not the demonstrative suffix. For instance, in /à-bàà-na/ ‘This river’, the tone of the root changes from L to H while -na receives a L tone, causing the phrase to surface as [à-báá-nà] instead of *[à-bàà-ná].

I account for these patterns using Optimality Theory (OT) (McCarthy & Prince, 1995; Prince & Smolensky, 2008) with proposed constraints and rankings as *TONELESS >> IDENT(tone) >> OCP >> DEP(tone) and *TONELESS >> DEP(association) >> DEP(tone). *TONELESS is undominated because all TBUs in Dagaare must have tones. IDENT(tone) dominates OCP because the consecutive sequences of tones from the lexicon are preserved in the output. OCP dominates DEP(tone) because underlyingly toneless TBUs receive surface tones through insertion of tones that are opposite to those of the adjacent TBUs. Since tones that are not present in the input can be inserted in the output, the constraint DEP(tone) is the lowest ranked. Also, DEP(association) disallows association between an underlying tone and a TBU that is not present in the input, causing toneless TBUs to receive surface tones through tone insertion. Then, lexically-associated co-phonologies of individual noun roots are responsible for the exceptions in tonal alternation. The noun bàà ‘river’ in /à-bàà-na/ ‘This river’ has a OT constraint ranking of *TONELESS >> OCP >> IDENT(tone) >> DEP(tone) instead of the default *TONELESS >> IDENT(tone) >> OCP >> DEP(tone). Since the rankings of OCP and IDENT(tone) are flipped, the tone of the noun root alternates.

This study examines whether diachronic change in vowel production occurs in Hong Kong English (HKE), an indigenized variety of English, and whether it can be attributed to contact with other English varieties. The production of FLEECE and KIT vowels, which are realized differently in HKE and Inner Circle Englishes, were analyzed. In HKE, the KIT vowel becomes tensed and merges with the FLEECE vowel, i.e., /ɪ/-tensing, while no such change occurs in any Inner Circle Englishes (Hung, 2000; Wells, 1982). 39 fluent Cantonese-English bilinguals from Hong Kong (22 females, 17 males, aged 18-58) participated in a sociolinguistic interview and a work list reading task in English. Each participant provided information regarding their age, gender, educational level, language orientation, attitudes towards HKE, and contact with other varieties of English, and they recited a list of 155 words, including 30 monosyllabic words with FLEECE and KIT vowels. The formants of all vowels from the word list were extracted at the one-third point of vowel duration and normalized using the Lobanov method (Lobanov, 1971). Mixed-effect regression models were used to examine whether the normalized F1 and F2 of FLEECE and KIT vowels varied with each social variable. The degree of overlap between the two vowels was assessed using Pillai scores derived from by-speaker MANOVAs fits (Nycz & Hall-Lew, 2013). Spearman’s rank correlation coefficients were performed to test the significance of correlation between the Pillai scores and the social variables, as well as to examine whether the social variables co-variated with each other.

Younger speakers showed a significantly lesser degree of /ɪ/-tensing than older speakers, which suggests a diachronic reversal of /ɪ/-tensing in HKE. On top of that, unlike previous descriptions, /ɪ/-tensing was found to occur only before non-velar consonants but not before velars. However, older and younger speakers did not differ in their level of contact with non-local speakers, implying that the reversal is unlikely to be caused by direct social contact. Instead, other sources like media consumption, prescriptive language instruction, and exonormative ideals of English might be responsible for the change.

Conventionally, lexical tone is seen as a marked feature and is assumed to be easily eliminated when languages with different prosodic systems come into contact (McWhorter, 1998; Salmons, 1992; Trudgill, 2010). However, recent literature on language contact in Africa and the African diaspora suggests that non-tonal European languages may undergo tonogenesis as a result of contact with tonal African languages (Good, 2004; Guri, 2012; Gussenhoven, 2014; Steien & Yakpo, 2020; Yakpo, 2018), For example, Central African French developed a {H, L} two-tone system due to contact with Sango, in which content words and function words were assigned H tone and L tone respectively (Guri, 2012). Thus, tonogenesis may occur not only due to language-internal phonetic factors, but also contact between different word-prosodic systems.

To examine whether tonogenesis could be a possible outcome of prosodic contact between tonal substrates and non-tonal superstrates in other regions of the world, this study examines the pitch production of monosyllabic content and function words in Hong Kong English (HKE), a contact variety of English with tonal Cantonese substrate. Data were extracted from a larger sociolinguistic study on HKE with 39 speakers aged between 18-58 (22 females, 17 males). As shown in Table 1, 20 monosyllabic words from a word list reading task were selected for analysis, which include 9 pairs of content and function words that are segmentally identical in Standard English. Two more function words with no content word homonyms were included as additional data. Words were embedded in the carrier phrase “Say _ again” and presented in pseudo-random order. Since the target words were in focused position in the carrier phrase, any difference in F0 between content and function words must be lexical rather than intonational. The F0 was measured at the midpoint of the sonorant portion of the target words using Praat (Boersma & Weenink, 2022). The measurements were then converted to semitones and z-score normalized.

The results of mixed effect regression modelling indicate that the majority of HKE speakers showed a difference in F0 between content words and function words, which content words were produced with a significantly higher F0 than function words even when their prosodic position remained constant. As shown in Figure 1, the difference was relatively consistent among function words as they all had a significantly lower F0 than content words with the exception of modal verbs. Furthermore, the sociolinguistic factors of speaker gender and attitudes towards HKE were found to condition F0 production. Female speakers and speakers with positive or neutral attitudes towards HKE were more likely to show the F0 distinction than male speakers and speakers with negative attitudes.

The findings of this study support Yiu (2014) and Wee (2016)’s analyses of HKE as a tone language with a contrastive {H, Ø} two-tone system. I propose that among monosyllabic words, all content words are assigned H tone while function words are variably assigned H or Ø tone depending on their parts of speech. Prepositions, auxiliaries, determiners and conjunctions are assigned Ø tone while modal verbs are assigned H tone. Ø tone surfaces as M tone in non-phrase-final position, creating tonal minimal pairs like four /H/ and for /Ø/ as shown in Figure 2.

Given the similarities in word-level prosody between HKE and the various contact languages in Africa, I conclude that tonogenesis is indeed possible when non-tonal languages are introduced to ecologies in which tone languages dominate, as in Sub-Saharan Africa and East Asia. In addition, as a linguistic innovation, the lexical tone system bears sociolinguistic significance for speakers of HKE as it serves to index a positive stance towards HKE and an affiliation towards HKE instead of Standard English.

Among younger speakers of Hong Kong Cantonese (HKC), an allophonic split in the sibilant affricates has emerged, with alveolar /ts, tsʰ/ appearing as palatalized [tʃ, tʃʰ] before round vowels (Chan 2007, Liu 2010). Some have attributed this change to language contact with English (Chan 2007, Lan 2017), but that explanation does not fully account for the new allophonic distribution. The palatalized allophones do not appear before unround vowels, nor has the change affected the sibilant fricative /s/. The current study examines the underlying phonetic motivation for this change through a dynamic articulatory analysis of tongue and lip movement.

12 native speakers of HKC (6 males, 6 females, born 1956-2001) participated in the experiment. Each participant recited a list of 126 disyllabic words containing all attested combinations of the target onsets /s, ts, tsʰ/ with the vowels /i, y, u, ɛ, œ, ɔ, a, ɐ/, as well as 44 filler words. Target sibilants appeared in the onset of the first syllable, which had either Tone 1 (high-level) or Tone 3 (mid-level). Words were embedded in the carrier phrase [ŋɔː˩˧ waː˨] ___ [jɐt˥ tsʰiː˧] “I say ___ once” and presented in pseudo-random order. Synchronized audio, lip video, and ultrasound tongue images were recorded in Articulate Assistant Advanced (AAA). Ultrasound data were captured at 84 frames per second (fps) and lip video at 60 fps, with the ultrasound transducer and side-view camera held in place with a stabilizing headset. The positions of the tongue and lips were measured at each frame within the duration of the target sibilants. Ultrasound tongue contours were analyzed using polar smoothing spline ANOVA (SS-ANOVA, Mielke 2015). Degree of lip rounding was calculated by measuring the distance from the corner of the lips to a line intersecting with reference marks on the upper and lower lips. Acoustic spectral peak frequency was measured using a 10ms Hamming window centered around the corresponding ultrasound frames.

Anticipatory lip rounding was observed for all three sibilants when they appeared before round vowels. The timing and degree of lip rounding did not differ significantly between the three sibilants, with a gradual increase in lip rounding occurring during the first third of each sibilant, all three of which had similar overall durations. This increase in rounding, however, has differing acoustic consequences for the affricates /ts, tsʰ/ as opposed to the fricative /s/. For /s/, the increase in rounding occurs during frication and is therefore audibly recoverable. For /ts, tsʰ/, the lips are fully round prior to the release of the closure, thus the frication phase is influenced by anticipatory lip rounding to a greater extent than that of the fricative.

Because lip rounding and tongue backing have similar effects on the acoustic output, alveolar affricates with coarticulatory rounding are acoustically similar to post-alveolar affricates, providing a basis for articulatory re-interpretation. Results of the ultrasound analysis reveal that some younger speakers have indeed re-interpreted anticipatory lip rounding as palatalization, producing /ts, tsʰ/ as [tʃ, tʃʰ] before round vowels. Apparent-time comparison shows that older speakers have one invariant alveolar place of articulation for the affricates while younger speakers show an allophonic split based on vowel rounding: alveolar before unround vowels and post-alveolar before round vowels.

This study demonstrates that ambiguous acoustic signals, which can be produced by multiple articulatory configurations, opens the possibility for articulatory re-interpretation of alveolar affricates in HKC. This study also illustrates the importance of employing articulatory measurement tools in the study of sound change, as the interplay between acoustic and articulatory phonetics cannot be observed through the use of acoustic measurement alone.

The allophonic system of coronal affricates /ts/ and /tsʰ/ in Hong Kong Cantonese (HKC) has undergone restructuring in recent decades with the emergence of palatalized allophones (Yeung 1980, Chan 2007, Liu 2010, Lan 2017). However, there is debate surrounding the conditioning environment for palatalization, as well as how the pattern is stratified by age and gender. Moreover, previous studies relied on impressionistic coding rather than acoustic measurement. This study combines acoustic and articulatory analysis to investigate this change from a novel perspective.

Two methodological approaches were employed: apparent-time acoustic analysis of spontaneous speech and articulatory experiment with ultrasound tongue imaging. First, tokens were collected from 24 speakers of HKC (12 males, 12 females, born 1945-1999) in an RTHK TV programme. Spectra of the affricate frication were analyzed in Praat for spectral peak frequency, an indicator of place of articulation (Stevens 1998, Jongman et al. 2000). Mixed-effects linear regression modelling indicates that palatalization is conditioned by roundedness of the following vowel, and that degree of palatalization varies by age and gender. Affricate peak frequency is higher before unrounded vowels and lower before rounded vowels, with this difference being greater among younger vs. older speakers and among female vs. male speakers.

The second experiment disambiguates the contributions of tongue position and lip configuration to the observed acoustic differences. The tongue movements of 12 speakers of HKC (6 males, 6 females, born 1956-2001) were recorded using ultrasound tongue imaging. The results indicate variation in lingual articulation for coronal affricates before rounded and unrounded vowels by age and gender. Older males have one invariant tongue position, younger females show an allophonic split based on vowel rounding: alveolar before unrounded vowels and post-alveolar before rounded vowels, whereas older females and younger males use an in-between pattern. For the more conservative older male speakers, the acoustic difference between affricates followed by rounded vs. unrounded vowels is caused by anticipatory lip rounding, while for the more innovative younger female speakers, such difference is further enhanced through palatalization. Overall, these results support the hypothesis of an ongoing restructuring of affricate allophones in HKC.