Variation in the formant of ethno-regional varieties in Nigerian English vowels

1.1 Ecology of NigE

Nigeria is known for its diversity and dynamic interplay between indigenous languages and sociocultural factors shaped by regions. Nigeria is home to over 500 languages, and the distribution of these languages reflects the country’s complex sociolinguistic landscape, shaped by geography, migration, and historical interactions (Lewis et al. 2009). Each region has a dominant language(s) within its ecology. For example, in the west, Yoruba is spoken predominantly, but also by westerners who reside in the east, north, and other parts of the country due to migration and inter-ethnic business, marriages, and education (Adewale 2005, Olajide 2014, Amrevurayire and Ojeh 2016, Ajani and Fakunle 2018). The immediate regions where westerners live contribute to their vowel variations. For this study, the speakers’ region (as a place of birth and residence) is expected to play an important role in shaping how WNigE vowels are realized. In regions such as the north and east of Nigeria, where Yoruba speakers coexist with speakers of Hausa (in the north) and Igbo (in the east), WNigE vowel spaces may exhibit adaptation influenced by dominant languages in the north and east (Puig-Mayenco et al. 2018). For instance, the formant trajectories might change differently between the westerners who are born, raised, and residing in the west and the westerners in the northern and eastern regions with dominant other languages (e.g., Hausa and Igbo) (Nearey 2013, Feimster Holt and Ellis 2018).

1.2 Present direction

Our study describes the variance in formants characterized by high-low WNigE vowels (TRAP∼BATH, KIT∼FLEECE, and FOOT∼GOOSE) produced by WNigE speakers in the east, north, and west regions of Nigeria. The dynamic variations in the formants F1 and F2 were examined using VL, TL, FV, and GAMM. Content words were included in both citation and passage reading. This study investigates whether employing two- or three-point measures yields differing results in formant transitions and whether these measures better describe the variability of WNigE high-low vowels. Additionally, the potential of VL, TL, FV, and GAMM is explored in estimating the extent of formant changes over time.

A key focus of this study is analyzing WNigE vowels in both passage and word list styles. These vowels were selected for two main reasons. First, they are among the most debated monophthongs regarding whether vowel contrast exists in WNigE. According to Jowitt 2018, Oladipupo and Akinola 2022, these vowels do not exhibit phonetic contrast. However, they are widely used across WNigE, making them ideal for cross-dialectal comparisons. Examining these vowels provides insight into how WNigE speakers may differ from speakers in other regions, particularly in terms of accent variation and indigenous language influences. Second, these vowels cover a broad spectrum of the vowel space offering a comprehensive basis for phonetic analysis.

The selection of these vowel pairs is further motivated by their phonetic status within NigE sub-varieties, as identified in previous studies (Isiaka 2019, Jowitt 2018), which suggest that regional variation plays a significant role in vowel restructuring. Rather than comparing these contrasts to RP, this study focuses on how they emerge within NigE itself. The choice to prioritize vowel height (F1) over frontness/backness (F2) is based on VISC research (Fox and Jacewicz 2009, Hillenbrand 2013), which indicates that F1 transitions are more dynamic and informative for vowel contrast. Specifically, Jowitt 2018 has observed vowel height distinctions in NigE, making F1 an important acoustic parameter for this study. Given that this study employs formant trajectory analysis, vowel height (F1) provides a more robust dimension for capturing regional and sociolinguistic variation in NigE vowels.

Analyzing these vowels in both word list and passage reading provides insight into both the stability and variability of vowel production in WNigE speakers. Word list reading tends to overestimate phonetic distinctions, whereas passage reading reflects more naturalistic speech patterns, allowing for a comprehensive examination of vowel variation (Heinrich and Smakman 2015). Examining the same vowels across these two reading styles provide insights into whether sociolinguistic and linguistic factors contribute to differences in vowel production.

1.3 Research hypotheses

Hypothesis 1

Vowels TRAP∼BATH, KIT∼FLEECE, and FOOT∼GOOSE produced by WNigE speakers from the west, north, and east exhibit significant differences in the amount and direction of spectral change, as measured by VL, TL, FV, and formant trajectories. These differences are expected to reflect through social and socio-biological factors.

Hypothesis 2

The reading style (word list items vs passage) influences the formant transitions of WNigE speakers’ vowels TRAP∼BATH, KIT∼FLEECE, and FOOT∼GOOSE, with word list expected to show more stable and distinct formant patterns than passage reading. This difference is hypothesized to be more pronounced/distinguished in vowels from WNigE speakers living in the west than those from the north and east.

Hypothesis 3

Several point formant measurements yield differing results in describing the vowel formant transitions of WNigE speakers’ vowels TRAP∼BATH, KIT∼FLEECE, and FOOT∼GOOSE (Wells 1982). Specifically, nine-point measures are hypothesized to provide a more detailed and better representation of the dynamic variability in formant changes, particularly in better capturing the formant shape, change, and direction with GAMM. This offers a better understanding of vowel identity and contrast in WNigE.

The predictions indicate that WNigE vowels vary based on the speakers’ regions and linguistic styles. WNigE speakers from the north are expected to show greater F1 formant variability and higher F1 trajectories moving in the opposite direction with speakers from the west. For the second research question, previous studies (Fox and Jacewicz 2009, Jacewicz and Fox 2012) show that vowels in citation/word lists are more peripheral and exhibit greater spectral variation than vowels produced during passage reading. Consistent with these findings, vowels in word item lists are anticipated to display more stable formant movement than passage. Finally, addressing the third research question, three-point measurements are expected to capture linear, nonlinear, and curved formant trajectories more effectively than two-point measures.

2.1 Participants

Seventy-five (75) WNigE speakers participated in the study: 25 each from east, north, and west (age range = 18–42 years). These speakers have a minimum of university education to satisfy the educated variety of WNigE (Udofot 2004). The speakers resided specifically in Ogbomoso, Oyo state (west), Nsukka, Enugu state (east), and Jos, Plateau state (north) and they were operationalized as place of birth (POB). Ogbomoso in Oyo state is Yoruba’s ancestral home in western Nigeria. Nsukka, Enugu state, is home to Igbo speakers, and Plateau is home to Hausa speakers; however, there are westerners domiciled in Hausa and Igbo communities. In addition, in our study, the westerners in the east speak the eastern dominant language (Igbo) simultaneously with Yoruba, and those in the north are fluent in Yoruba and Hausa. We expect to see the effect of Igbo and Hausa on vowel production to differentiate them from westerners who speak Yoruba and reside in Yoruba dominated environments. The metadata retrieved attests that WNigE speakers in the east and north use Yoruba, Hausa, and English; however, the use of these three different languages depends on context and the language that the hearers understand (Adegbija 2000). The case is analogous to WNigE speakers in the east. For the WNigE speakers in the west, they use Yoruba and English depending on the context and the language that the hearers speak. All the speakers can speak and write in Yoruba, though we did not test their proficiency; however, we spoke (central dialect of) Yoruba at different times during the data collection.

Also, as seen in the sociolinguistic questionnaire administered to all the participants during the reading task, the 75 WNigE speakers had comparable education, socioeconomic status, and birthplace (all born in three major regions of Nigeria). They were either degree or post-graduate holders with English as a lingua franca.

2.2 Speech materials and procedures

The data collection instrument consists of two sections: the first section contains a passage with 171 words, while the second section includes a 40-word item list. Both passage and citation have TRAP∼BATH, KIT∼FLEECE, and FOOT∼GOOSE in content words as target words. These target vowels have varied consonants: alveolar, dental, labial, palatal, velar, and labiodental in the preceding and following phonological environment. WNigE speakers in the three regions were individually recorded in a noise-minimized environment in Ogbomoso (west), Nsukka (east), and Jos (north). Geographically, the west is 663.42 km from the East and 456.36 km from the north, respectively (Natural Earth n.d.).

The speakers were recorded for the two tasks after securing their consent. First, they read the passage and then the word item lists. This approach does not make the speakers aware of the target words. Both tasks were presented to the participants in written form. Both the text and the audio files were forced-aligned automatically in WebMAUS and the alignment was manually adjusted where necessary in Praat (Boersma and Weenink 2030). After this stage, a trained degree holder in English was recruited to cross-check the adjustment, and the interrater reliability agreement was 90%. The 10% differences were resolved before the formant extraction. A modified Praat script by Riebold 2013 was used to extract F1, F2, and F3 at 10% intervals for nine points. A total of 3,375 tokens were extracted for the target (content) words. By regions, there were 1,141 tokens for WNigE speakers in the west, 1,065 tokens from the WNigE speakers in the north, and 1,169 for WNigE speakers from the east (Table 1).

2.3 Vowel measurement

3.1 Data analysis and interpretation

Direction and amount of formant movement and direction of high-low vowels (TRAP∼BATH, KIT∼FLEECE, FOOT∼GOOSE) of WNigE speakers across POB and reading styles were examined. Figure 1 compares the group mean formant trajectories for high-low vowels for WNigE. Based on visual inspection, WNigE high-low vowels showed different formant movements. However, Figure 1 further reveals that the patterns of formant movement varied across the POB. Differences in formant direction and/or magnitude were observed among WNigE speakers, and the effect of POB contributed to their formant transitions, as shown in Figure 2. Figure 3 further reveals how the vowels varied across reading styles.

Figure 1

Bark normalized F1 × F2 values averaged across all participants for the TL, direction and elbow (straight vs elbowed) of WNigE high-low vowels across 20, 50, and 80% of vowel durations.

Figure 2

Bark normalized F1 × F2 values for the TL, direction, and elbow (straight vs elbowed) of WNigE high-low vowels across 20, 50, and 80% of vowel durations across the regions: east, north, and west. The F1 × F2 trajectory levels and directions show different levels across time points.

Figure 3

F1 × F2 high-low WNigE vowels for the trajectory composite for different speech styles.

First, Figure 1 presents the formant trajectories for six vowels (TRAP, BATH, KIT, FLEECE, FOOT, and GOOSE) in WNigE across vowel duration. The first formant (F1) and the second formant (F2) are plotted on the y-axis, with the frequency in Hz indicated on the left and timepoints on the x-axis.

Observations in Figure 1:

TRAP: F1 shows a downward trend, indicating a decrease in formant level as the vowel progresses, while F2 remains relatively stable, suggesting a consistent tongue position in terms of height.

BATH: F1 shows a slight decrease over time, while F2 increases slightly, particularly at the 80% time point, reflecting a shift in tongue position toward the front of the mouth.

KIT: F1 is stable with a slight increase at the final timepoint, and F2 shows a slight decrease, showing a shift in the phonetic retraction.

FLEECE: Both F1 and F2 show minimal changes, revealing that this formant level stays consistent regarding tongue height and backness throughout its production.

FOOT: F2 decreases slightly, while F1 formant elbow is still flat, suggesting a minimal phonetic retraction.

GOOSE: Formant levels F1 and F2 for this vowel remain stable across all time points, showing a steady articulation with slight variation in tongue position.

Conversely, formant elbows reveal that F1 remains relatively stable (flat/steady) for most vowels, while F2 shows slightly different formant levels, particularly in vowels like BATH and FOOT. This suggests that while the vertical tongue position (related to F1) remains consistent during vowel articulation, there is some horizontal movement (related to F2), indicating subtle shifts in tongue position toward the front or back of the mouth. The stability of F1 and F2 in vowels like FLEECE and GOOSE points to a more uniform articulation, while vowels like BATH and TRAP show more dynamic changes, particularly in F2, reflecting regional or stylistic differences in pronunciation within WNigE. This analysis provides a clear overview of how vowel articulation varies across different vowels in WNigE, with implications for understanding the phonetic characteristics of this English variety.

Meanwhile, formant trajectories for these vowels in WNigE across different regions (east, north, and west) revealed different levels of variations. The formant frequencies F1 and F2 are plotted on the y-axis, while the time points (20, 50, and 80 Bark) are shown on the x-axis, indicating the changes in the formant frequencies over time for each vowel production.

For the WNigE speakers in the east, as shown in Figure 2, vowels like TRAP and FLEECE show relatively stable F1 trajectories, while F2 shows slight variation, particularly in TRAP, where F2 decreases over time. For vowels like FOOT and GOOSE, F2 remains higher across the time points, indicating a higher tongue position during pronunciation. The BATH vowel shows a slight decrease in F1, while KIT displays a stable F1 and a slight rise in F2. Comparatively, WNigE in the northern region showed more varied formant trajectories. For instance, TRAP has a rising F2, while BATH shows a stable F2 trajectory with a slight decrease in F1. KIT and FLEECE have relatively stable trajectories, with F2 showing minimal change. FOOT exhibited a more pronounced dip in F2, indicating a lower tongue position during this vowel’s production. GOOSE shows a consistent F2 with a slight rise at 80%. Meanwhile, for WNigE speakers in the western region, TRAP and BATH show similar trends to the north, with a slight decrease in F1 and stable F2 trajectories. KIT and FLEECE have relatively flat formant trajectories, indicating a minimal tongue height or position change over these vowels. FOOT and GOOSE in the west show relatively stable F1 and F2 trajectories, with a slight rise in F2 for GOOSE toward the end of the vowel duration.

Importantly, across all regions, F1 generally shows less variation compared to F2, indicating that the vertical position of the tongue (related to F1) remains relatively consistent during vowel production, while the horizontal position (related to F2) shows more variation, particularly in vowels like TRAP, FOOT, and GOOSE. The differences in formant trajectories across regions suggest that place of birth influenced vowel pronunciation, with subtle variations in how these vowels are articulated in WNigE. This influence could be due to regional accents or dialects within the WNigE accent, which affect how vowels are produced.

In Figure 3, the formant elbows revealed trajectories for the reading style: citation/word list and passage. The formant trajectories, represented by lines, show the formant frequencies during the vowel production.

The formant movements revealed that the TRAP vowel slightly increased F1, with F2 remaining relatively stable across both reading styles. For BATH, F1 shows a slight decrease, while F2 remains steady, with a minimal increase in citation style at the 80% time point. Similarly, trajectories for KIT were relatively flat, with F2 slightly decreasing in word item list. FLEECE displayed a stable F1, with a slight decrease in F2. FOOT also showed a minimal rise in F1 and an upward trend in F2 within citation style. GOOSE has a stable F1, with F2 slightly increasing at the 80% time point.

The influence of reading style on these vowels is evident in Figure 3. However, for most vowels, the formant trajectories do not show significant differences between the word item list and passage reading styles, suggesting that the formant frequencies are relatively consistent across these conditions. The stability in the passage style compared to the citation style indicates more uniform articulation during connected speech. Importantly, vowels like FLEECE and GOOSE exhibit higher F2 positions, which suggest a higher tongue position closer to the roof of the mouth during production. Figure 3 shows the effect of style on vowel productions, offering valuable insights into the acoustic characteristics of these vowels.

3.2 FV

The analysis of FV across WNigE vowels revealed distinct patterns influenced by social and linguistic factors (Figure 4). Using MANOVA, the study analyzed F1 and F2 for FOOT∼GOOSE, TRAP∼BATH, and KIT∼FLEECE, age category, syllable sequence significantly influenced FV for both F1 and F2 (F1: Pillai = 0.020, F(4, 2,194) = 5.566, p < 0.001; F2: Pillai = 0.012, F(4, 2,194) = 3.416, p = 0.009), while other predictors such as gender and education were not significant (p > 0.05). For TRAP∼BATH, there was no statistically significant relationship with POB, gender, or education for F1 or F2 (p > 0.05).

Figure 4

FV for F1_onglide, F1_offglide, F2_onglide, and F2_offglide for the six vowel lexical sets: TRAP, BATH, KIT, FLEECE, FOOT, and GOOSE.

However, FLEECE revealed a significant main effect of vowels and syllable sequence on FV in F1 (vowel: Pillai = 0.004, F(2, 1,482) = 3.320, p = 0.036; syllable sequence: Pillai = 0.009, F(2, 2,946) = 3.5427, p = 0.01), indicating variation in the onglide and offglide velocities across these vowels, while other predictors (e.g., age) remained non-significant (p > 0.05). These results suggest that FV varies by some linguistic and social factors, for example, education is not found to affect FV.

3.3 VL

The separate repeated-measures ANOVAs were conducted for each vowel. The ANOVA on the VL values for KIT and FLEECE revealed a non-significant effect (F(1, 1,440) = 2.742, p = 0.097). Meanwhile, these vowels (KIT and FLEECE) are significantly different across POB (F(2, 1,440) = 21.229, p < 8.2 × 10⁻¹⁰, η ² = 0.03) and reading style (F(1, 1,440) = 4.511, p < 0.05, η ² = 0.005). The two-way interaction, vowels × reading style did not reveal any significant effect, (F(1, 13) = 2.908, p = 0.088, η ² = 0.01). However, three-way interaction, POB × reading style × education (F(1, 1,440) = 5.404, p < 0.05, η ² = 0.006). The three-way interaction with age category also revealed a significant main effect of age in the phonemic differentiation between KIT and FLEECE (F(1, 1,440) = 4.849, p < 0.05). The estimate value revealed that FLEECE (β = 0.30) is longer than KIT. The approximate effect size of this difference is Cohen’s d ≈ 0.35, which reflects a small-to-moderate effect, contrary to earlier claims (Akinjobi 2006, Jowitt 2018) that the WNigE does not have a phonemic distinction between KIT and FLEECE. Their arguments were premised on single-time vowel measurement without harnessing several point measurements or VL values.

Similarly, the ANOVA on the VL values for TRAP and BATH revealed a non-significant effect (F(1, 554) = 0.817, p = 0.3665, η ² = 0.001). Meanwhile, these vowels (TRAP and BATH) significantly differ across POB (F(2, 554) = 4.248, p = 0.01, η ² = 0.02). The two-way interaction, POB × age category, vowels × reading style, POB × education, and reading style × education, consistently revealed a significant effect (p < 0.05). The estimate revealed that BATH is 0.11 longer than TRAP. Though this relationship was not statistically significant, the estimate of 0.11 revealed that TRAP and BATH are not the same in VL. This result contradicts several investigations (Akinjobi 2006, Jowitt 2018) that claim that the WNigE accent does not have a phonemic distinction between TRAP and BATH. Their arguments consistently established that TRAP and BATH are the same vowels in WNigE.

The average VL for FOOT and GOOSE are 1.692 and 2.073, respectively (Figure 5). The ANOVA results for the VL comparison between the vowels FOOT and GOOSE, considering factors such as POB, gender, age category, reading style, and education, provide several insights. The main effect of vowels is not statistically significant (p = 0.09), but education is statistically significant (p < 0.01).

Figure 5

Inter-quartile range with mean VL values for WNigE high-low vowels with mean VL values for east, north, and west in citation/word list and passage reading styles.

However, most interactions between vowels and other factors (such as POB, gender, age, and reading style) are insignificant (p > 0.05). This suggests that the effect of these factors on VL does not vary significantly between ‘FOOT’ and ‘GOOSE.’

3.4 TL

As it was done for the VL measure, a repeated-measures ANOVA was conducted with the TL. The ANOVA results for FOOT and GOOSE across POB, gender, age, reading style, and education reveal a significant effect. POB has a significant main effect (F(2, 1,057) = 6.774, p < 0.00119, η ² = 0.02). The estimate revealed that both north and west have higher TL than east. Notably, the WNigE in the north region (β = 0.85) and the west region (β = 0.80) are significantly higher than the eastern region, indicating that WNigE speakers from the northern and western regions have longer TL compared to those from the east (F(2, 3,372) = 42.17, p < 2.2 × 10⁻¹⁶, η ² = 0.05). The statistically significant relationship supports the differences (north = p < 2 × 10⁻¹⁶; west = p < 1.49 × 10⁻¹⁴). Similarly, age shows a significant main effect (p = 0.032, η ² = 0.01). These results indicate that regional and age-related factors play an important role in the production of VL for these vowels. Also, the consonant sequence (CVC and CV) significantly affects TL. Specifically, the closed syllable has a shorter TL than the open syllable (β = −0.60, SE = 0.11, p < 4.10 × 10⁻⁰⁷, d = 0.65).

Also, there was a significant effect in the interaction of WNigE vowels and speaker’s age (F(1, 1,057) = 5.279, p < 0.05, η ² = 0.01) and the three-way interaction between POB, age category, and reading style (F(1, 1,057) = 5.635, p < 0.01, η ² = 0.02). This indicates that the TL for FOOT and GOOSE is influenced by a combination of POB, age, and reading style. Additionally, the interaction between vowels (FOOT*GOOSE), reading style, and education is also significant (F(1, 1,057) = 4.021, p < 0.01, η ² = 0.015), suggesting that educational background further modulates the influence of reading style on vowel articulation. These results show the importance of considering multiple sociolinguistic factors when analyzing vowel production, as they collectively explain the variation observed in the acoustic properties of vowels. Similarly, the ANOVA results for TRAP and BATH reveal significant effects of POB and interactions on the vowel TL.

Also, between TRAP and BATH, POB shows a significant effect (F(2, 1,061) = 6.766, p < 0.001, η ² = 0.02), indicating that the region where the speaker is from significantly influences the TL of TRAP and BATH contrast. The interaction between these vowels and age category is also significant (F = 5.279, p < 0.05, η ² = 0.01). Similarly, the age of the speakers influences the TLs of the TRAP and BATH vowels differently (F(2, 1,061) = 3.449, p < 0.05). Moreover, the three-way interaction shows that vowel type, age category, and reading style statistically influence TL (F-statistic = 5.635, p < 0.01). The syllable type similarly affects TL (F(2, 1,061) = 13.215, p < 1.951 × 10⁻⁰⁶, η ² = 0.03). However, the interactions between gender, reading style, education, vowels (particularly TRAP and BATH), POB, and age category were not statistically significant (p > 0.05).

The ANOVA results for KIT and FLEECE reveal significant effects of the POB and age category on TL and notable interactions between these factors and the vowels. Specifically, the effect of POB is significant (F(2, 1,061) = 6.766, p < 0.001, η ² = 0.02), indicating that the region of birth influences the TL of these vowels. Additionally, age category significantly impacts TL (F(2, 1,061) = 3.449, p = 0.05, η ² = 0.01), suggesting that the age group of the speaker also plays a role in how these vowels are articulated. However, gender, reading style, and education alone do not show significant main effects, indicating that they are not statistically significant on TL (p > 0.05). The analysis of TL by syllable sequence reveals significant effects. While there is no significant difference between no onset in VC and CVC sequences, the contrast between CV and CVC shows a statistically significant relationship (KIT∼FLEECE [F(2, 1,082) = 4.082, p <0.01, η ² = 0.02]; TRAP∼BATH and FOOT∼GOOSE [F(2, 1,082) = 4.082, p = 0.017, η ² = 0.01]). These findings suggest that the phonological environment influences TL.

The interaction between vowels and age category (F(1, 1,061) = 5.347, p < 0.05, η ² = 0.01) and the three-way interaction between POB, age category, and reading style (F(1, 1,061) = 7.828, p < 0.01, η ² = 0.02) are both statistically significant. These results suggest that the influence of POB and age category on TL differs depending on whether the vowel is KIT or FLEECE. The effect is further phonetically conditioned by reading style. This indicates that the production of KIT and FLEECE vowels is statistically significant across the region, age, reading style, and education.

3.4.1 Formant trajectories with GAMM

Data from each vowel production were examined with GAMM. We expect significant differences in vowel space by social and linguistic factors. In the original model, the results indicate that F1 values are statistically significant for vowel realization. For example, the model for F1 shows significant differences in formant levels between the reference vowel (BATH) and other vowels (Figure 6). The intercept (β = 6.89, p < 0.001) represents the average F1 value for BATH. Relative to BATH, all other vowels except TRAP show significantly lower F1 values, with FLEECE (β = −2.19, p < 0.001, d = 0.95), FOOT (β = −2.14, p < 0.001, d = 0.93), GOOSE (β = −2.24, p < 0.001, d = 0.98), and KIT exhibiting lower F1 levels (β = −2.15, p < 0.001, d = 0.94). TRAP has a slightly higher F1, indicating a marginally lower tongue position than BATH (β = 0.08, p < 0.05, d = 0.04). Smooth terms reveal that the F1 trajectory over time differs significantly for BATH (p < 0.001), FLEECE (p < 0.01), GOOSE (p < 0.001), and TRAP (p < 0.001) with R ² = 31.6%.

Figure 6

Predicted average trajectories for POB vowels. Each panel shows the predicted values from GAMM fitted to vowel production, with separate trajectories for the three regions.

Similarly, POB influenced F1 for the speakers from the west compared to the north (β = −0.05, p < 0.05, d = 0.10). Gender differences revealed that male speakers had significantly lower F1 values than female speakers (β = −0.22, p < 0.001, d = 0.40). The smooth terms indicated significant temporal variations for most vowels, with timepoint effects for BATH (p < 0.001) and GOOSE (p < 0.001) standing out as particularly notable contributors to F1 dynamics (R ² = 32%).

Similarly, the F2 model highlights significant differences in tongue frontness across vowels, with the intercept representing the average F2 for BATH (β = 11.47, p < 0.001), FLEECE (β = 1.97, p < 0.001, d = 0.85), GOOSE (β = 0.64, p < 0.001, d = 0.28), and KIT (β = 1.86, p < 0.001, d = 0.80) have significantly higher F2 values compared to BATH. This shows a greater tongue frontness for other vowels except for BATH (Figure 6). However, the relationship of the frontness was not significant for FOOT (β = 0.10, p = 0.103, d = 0.04) and TRAP (β = −0.03, p = 0.549, d = −0.01). Smooth terms reveal significant temporal variation in F2 for FLEECE (p < 0.001), GOOSE (p < 0.01), and KIT (p < 0.001), while BATH shows no significant variation over time (p = 0.889). The model explains the variance in F2 at 21.3%, which indicates that tongue backness varies across vowels but to a lesser extent than F1 dynamics. In addition, POB strongly influenced F2 values, with speakers from the north (β = −0.39, p < 0.001, −0.34) and west (β = −0.53, p < 0.001, d = −0.45) showing significant decreases compared to the speakers from the East. Gender differences in F2 were statistically insignificant (β = −0.01, p = 0.661, d = −0.005). The smooth terms highlighted significant temporal variation for FLEECE (p < 0.001), KIT (p < 0.001), and GOOSE (p = 0.003), indicating dynamic changes in F2 trajectories over time (R ² = 23.3%).

The formant trajectories of vowels differ in distinct ways across the east, north, and west regions. In the east, KIT and FLEECE exhibit close and phonetic frontness, while GOOSE and FOOT are farther back (or retracted), with TRAP and BATH showing minimal vertical variation. In the North, the trajectories for GOOSE and FOOT show more phonetic overlap and less distinction than speakers in the East, while KIT and FLEECE maintain their frontness slightly closer (near merger). TRAP and BATH in the north display more vertical divergence, indicating greater variation in vowel height. In the west, KIT and FLEECE remain fronted but are more phonetically distinct than among the speakers in the north. At the same time, GOOSE and FOOT are phonetically distinct and exhibit different trajectories. TRAP and BATH show closer vertical alignment than in the north, suggesting more compression in their formant spaces. These results are consistent with Nagamine (2024) and Feimster Holt and Ellis (2018) that region affects formant trajectories.