Liverpool English is spoken in Liverpool and surrounding areas of Merseyside in North West England (see Figure 1). Commonly known as ‘Scouse’, Liverpool English is one of the most widely known accents in the United Kingdom (Leach et al. Reference Leach, Watson and Gnevsheva2016; Montgomery Reference Montgomery2007). Numerous negative stereotypes exist about Liverpool and its inhabitants in the UK (e.g. Coupland and Bishop Reference Coupland and Bishop2007; Montgomery Reference Montgomery2007; Trudgill Reference Trudgill1986). In popular media, ‘Scousers’ are commonly portrayed as ‘urban, working-class, and male’ (Cooper and Lampropoulou Reference Cooper and Lampropoulou2021, 109). The stereotypical ‘Scouser’ is perceived to be a thieving, violent ‘scally’ (i.e. a jobless miscreant) and compulsive complainer (Belchem Reference Belchem2006; Crowley Reference Crowley2012; Sangster Reference Sangster2002). Within the UK, negative beliefs about Liverpool are intimately tied to perceptions of the accent. Previous studies have found British listeners to judge the accent as coarse and ill-educated (Trudgill Reference Trudgill1986), socially unattractive and low in prestige (Coupland and Bishop Reference Coupland and Bishop2007; Montgomery Reference Montgomery2007; Sharma et al. Reference Sharma, Levon and Ye2021). Despite this, within Liverpool, the Scouse accent holds covert prestige and is considered to be a ‘principal marker of local identity and allegiance’ (Grey Reference Grey, Grant and Grey2007, 189).
Figure 1. Map of the UK with Liverpool highlighted in the light grey and the Wirral in dark grey. Liverpool and the Wirral are separated by the river Mersey. The shaded areas within the square represent the boroughs which make up Merseyside; Liverpool and the Wirral are both part of Merseyside.
In contrast, it is ‘generally accepted, within Liverpool at least, that people who are born and/or live on the other side of the river Mersey cannot be considered Scouse’ (Peschier Reference Peschier2018, 92). The people of Liverpool frequently refer negatively to those who live ‘over the water’, on the Wirral, as ‘plastic Scousers’ (Boland Reference Boland2010, 6) or ‘plazzy’ Scousers (Wilkinson Reference Wilkinson2017): ‘plastic; false; imitation; second-rate’ (Crowley Reference Crowley2017, 181).
Wirral speakers, on the other hand, have been found to have ‘no self-identifying label’ (Cooper and Lampropoulou Reference Cooper and Lampropoulou2021, 119). Instead, for those from the Wirral, there seems to be a ‘consistent denial of the Scouse identity’ (Cooper and Lampropoulou Reference Cooper and Lampropoulou2021, 119). Byrne (Reference Byrne, de Velde, Hilton and Knooihuizen2021) observed that ‘Wirral students were highly cognisant of defining themselves in relation to Liverpool and Scouse, with a Wirral dialect or identity being secondary to this’ (p. 176). She further noted that Wirral speakers were acutely aware of ‘what a Scouse identity might be, but it was not one that they related to themselves’ (p. 171). Wirral identity then seems to be defined in terms of not being Scouse.
Although Wirral speakers do not directly identify as Scousers themselves, their speech has been shown to draw selectively upon elements of the Scouse feature pool. For example, despite fronted nurse being a salient feature of Liverpool English (Watson and Clark Reference Watson and Clark2013), acoustic analysis has shown that middle-class female Wirral adolescents display a similar degree of fronting to Liverpool females. The similarity in usage between Wirral and Liverpool female adolescents, could be taken to indicate ‘a linguistic and social shift towards aspects of Scouse’ (Byrne Reference Byrne, de Velde, Hilton and Knooihuizen2021, 176) and ‘pride at being identified as belonging to the Merseyside geographic region’ (p.171). More broadly, Wirral speakers have been observed to draw upon a covertly prestigious Scouse repertoire, making use of ‘a linguistic resource [ . . . ] to symbolise covert prestige and informality, rather than to define themselves as “Scousers”’ (p. 177).
Given reports that Wirral speakers tend not to identify with a Scouse identity, and that Scouse has a negative image in the UK, Scouse has, perhaps surprisingly, been reported to have ‘ousted the traditional dialect’ of the Wirral (Knowles Reference Knowles1973, 14). Although Scouse has been defined as the ‘form of speech that is characteristic of Liverpool, Birkenhead and adjacent areas of urban Merseyside’ (Watson Reference Watson2007, 110), there are reasons for supposing there exist ‘different kinds of Merseyside Englishes’ (Grant Reference Grant, Grant and Grey2007, 141) tied to distinct identities within the region, not least due to reports that Wirral speakers do not identify directly with the Scouse label.
Given the discrepancy between the observations that Scouse has replaced the Wirral accent (Knowles Reference Knowles1973) on the one hand, and, on the other, that Scouse is rejected by speakers on the Wirral (Byrne Reference Byrne, de Velde, Hilton and Knooihuizen2021; Cooper and Lampropoulou Reference Cooper and Lampropoulou2021), we decided to undertake an investigation into Wirral and Liverpool adolescent speech to examine the extent to which, if at all, these varieties differed from one another and whether there was variation within the two groups.
In line with recent studies into Wirral speech, we targeted an adolescent population on the basis that previous sociolinguistic research has shown adolescence to be a critical site for the study of language variation (e.g. Eckert Reference Eckert1988; Reference Eckert1989; Oxbury and de Leeuw Reference Oxbury and de Leeuw2020; Tagliamonte and D'Arcy Reference Tagliamonte and D'Arcy2009). Adolescents are ‘significant bearers of change’ whose ‘distinct social identity means that they are willing to modify their speech’ (Kerswill Reference Kerswill1996, 198). On this basis, teenage speakers might be expected to display the greatest level of language innovation in relation to the expression of identity, so we considered them to be the most interesting age group to examine.
Our focus was on the lateral in this study because little is known about this variable in Merseyside speech, relative to others, such as the merger between nurse and square vowels (e.g. Watson and Clark Reference Watson, Clark and Hickey2017), and the extensive lenition of plosives (e.g. Knowles Reference Knowles1973; Watson Reference Watson2006, Reference Watson2007). Another reason we chose to study laterals was due to the existence of multiple articulatory strategies, which make this an exciting variable to study using both acoustic and articulatory methods. In particular, the study was driven by Turton's (Reference Turton2014) observation that, for laterals, there were ‘findings in the articulatory data collected in this study which the acoustics miss altogether’ (p. 218). This raised the possibility that the variable could be subject to covert articulatory variation.
Covert articulatory variation refers to the use of ‘different articulatory strategies to produce sounds that are not easily distinguished acoustically or auditorily’ (Smith et al. Reference Smith, Mielke, Magloughlin and Wilbanks2019, 5). Ultrasound tongue imaging (UTI) offers a means of showing tongue movements not evident acoustically. For example, UTI was applied by Lawson et al. (Reference Lawson, Stuart–Smith and Rodger2019), who investigated the articulatory strategies used in the production of the goose vowel by monolinguals from England, Ireland, and Scotland. They found that whilst the vowel did not differ in terms of F2 across varieties, it did differ in terms of the articulatory configurations used to bring about the same acoustic result. For Anglo and Hiberno-English speakers, a more advanced tongue position occurred, whilst Scottish speakers demonstrated a more retracted tongue back.
In terms of covert articulatory variation, /r/ has been the focus of numerous investigations, whilst /l/ is much less explored in this regard. Interestingly, however, articulatory differences in terms of velarisation versus pharyngealisation as lateral darkening mechanisms have not been found to be evident acoustically (e.g. Turton Reference Turton2014; Charles and Lulich Reference Charles and Lulich2018), suggesting a parallel with /r/. There are various potential realisations of the rhotic /r/ in English (Delattre and Freeman Reference Delattre and Freeman1968), which are potentially indistinguishable to listeners (Twist et al. Reference Twist, Baker, Mielke and Archangeli2007). For the rhotic, articulatory variation can be ‘entirely idiolectal’ (Smith et al. Reference Smith, Mielke, Magloughlin and Wilbanks2019, 5), as has been reported in the use of bunched versus retroflex variants of /r/ in American English (e.g. Magloughlin Reference Magloughlin2016; Mielke et al. Reference Mielke, Adam Baker and Archangeli2016). However, socially meaningful variation has also been observed in the use of differing articulatory strategies despite the apparent absence of acoustic evidence. For example, Lawson et al. (Reference Lawson, Scobbie and Stuart–Smith2011; Reference Lawson, Scobbie and Stuart–Smith2013) observed that middle-class Scottish speakers favour bunched articulations compared to working-class speakers who favoured tip-up articulations. In such cases, speakers appeared to select specific tongue configurations on the basis of social group membership, rather than on a speaker-by-speaker basis. Like /r/, /l/ is an articulatorily complex liquid, raising the possibility that covert articulatory variation could relate to social group membership for the lateral too. Therefore, we considered the application of UTI to potentially reveal variation in adolescent Liverpool and Wirral speech which might not be evident through acoustic analysis alone.
If covert articulatory variation were to be observed, interesting questions would emerge as to how and why such variation might have taken place. In studies on rhotics, it has been observed that covert articulatory variation can be recovered from coarticulatory effects and visual cues (Lawson et al. Reference Lawson, Stuart–Smith and Scobbie2014), as well as less commonly considered acoustic metrics (Johnson Reference Johnson2011; Johnson and Bakst Reference Johnson and Bakst2016; Lawson et al. Reference Lawson, Stuart–Smith and Scobbie2018). It is also worth considering that speech is a multi-modal phenomenon. Numerous studies have found listeners to be sensitive to visual information and not just auditory information in isolation (McGuire and Babel Reference McGuire and Babel2012; McGurk and McDonald Reference McGurk and MacDonald1976; Ménard et al. Reference Ménard, Dupont, Baum and Aubin2009; Traunmüller and Öhrström Reference Traunmüller and Öhrström2007), providing speakers with a means of detecting seemingly covert articulatory differences.
Acoustic and articulatory characteristics of lateral approximants in Liverpool, the Wirral and surrounding areas
In articulatory terms, both forms of the lateral involve the formation of ‘lateral cavities [ . . . ] on one or both sides of the constriction, allowing air to flow past the constriction’ (Carter Reference Carter2002, 76). Both allophones also involve the raising of the tongue tip towards the alveolar ridge. However, the tongue tip gesture is reduced for [ɫ] relative to [l], and the dark variant tends to show a lowering and retraction of the tongue dorsum in contrast to the clear form which typically shows a raised and fronted tongue body (Kirkham et al. Reference Kirkham, Turton and Leemann2020; Recasens Reference Recasens2012; Sproat and Fujimura Reference Sproat and Fujimura1993; Turton Reference Turton2017). It has also been reported that the dark lateral is typically more pharyngealised in Standard British English varieties (Turton Reference Turton2014, Reference Turton2017), involving ‘postdorsum retraction towards the uvular zone or the upper pharyngeal wall’ (Recasens Reference Recasens2012, 368).
In terms of acoustics, lateral darkness is commonly examined in terms of the relationship between F1 and F2: the higher the F1 and lower the F2 values, the darker the lateral is taken to be (Carter Reference Carter, Local, Ogden and Temple2003; Carter and Local Reference Carter and Local2007; de Leeuw et al. Reference De Leeuw, Tusha and Schmid2018; Sproat and Fujimura, Reference Sproat and Fujimura1993). Both velarisation and pharyngealisation result in lower F2 and increased F1 (Recasens Reference Recasens2004). One consequence of this is that it is not possible to discern using just F1 and F2 values whether dark laterals are produced using velarisation or pharyngealisation (Charles and Lulich Reference Charles and Lulich2018; Turton Reference Turton2014). In this study, ultrasound tongue imaging will be used to help distinguish which, if any, articulatory strategy is used in the production of dark laterals in Liverpool and Wirral adolescents.
Articulatory research has shown that speakers from different dialect regions across the United Kingdom and United States differ in the degree of differentiation shown between laterals occurring in various phonological contexts, and in the strategies used to produce lateral darkening (Turton Reference Turton2014, Reference Turton2017). Turton (Reference Turton2014) observed acoustic results that were ‘in no way reflected in the average tongue splines’ (p. 120) when examining the production of /l/. When looking at ultrasound data, she found articulatory distinctions that the acoustics appeared to ‘miss altogether’ (p. 218). Of particular note, she found that a Liverpool speaker displayed velarised dark lateral tokens, not found in Standard British English. This velarisation was ‘not reflected in any of the acoustic measurements’ (ibid.), indicating covert articulatory variation.
Regarding lateral production in Liverpool English, early observations were inconclusive: Jones (Reference Jones1956) stated that the lateral is light in both onset and coda positions, perhaps as a result of influence from Hiberno-English dialects. Historically, laterals in Irish varieties of English were described as clear in onset and coda positions (e.g. Hughes et al. Reference Hughes, Trudgill and Watt2012; Wells Reference Wells1982). In contrast, Knowles (Reference Knowles1973) observed that Scouse /l/ could ‘sound perversely “dark” initially and “light” after a vowel’ (p. 106), being ‘velarized’ (p. 256) in all positions.
More recent work, however, suggests that Liverpool English does, in fact, display a categorical distinction between clear and dark /l/ on the basis of syllable position. Turton (Reference Turton2014) observed that a Liverpool male speaker produced dark /l/ with velarisation rather than tongue root retraction. This was not found to be the case for other British speakers. Kirkham et al. (Reference Kirkham, Nance, Littlewood, Lightfoot and Groarke2019, Reference Kirkham, Turton and Leemann2020) similarly observed that speakers from Liverpool uphold a robust distinction between onset and coda laterals. However, they noted that the difference between initial and final productions was smaller for Liverpool speakers than speakers of some other British varieties. In this regard, Liverpool speakers seem to pattern like speakers of other Northern dialects in demonstrating smaller initial-final contrasts.
Regarding Wirral speech, one fleeting reference exists in the literature, with Barry (Reference Barry1958) stating that ‘the lateral is always thin in quality’ (p. 15) for speakers in Neston, an area located in the southwest of the Wirral peninsula. As Orton (Reference Orton1933) outlines, ‘thin’ refers to a formation in which ‘the back of the tongue is not raised towards the soft palate’ (p. 7), i.e. not velarised. Therefore, Barry's account appears to echo that of Jones (Reference Jones1956) regarding laterals in Liverpool English, suggesting that the lateral is light in both onset and coda positions. Newbrook (Reference Newbrook1986, Reference Newbrook, Foulkes and Docherty1999), who has provided the most systematic coverage of Wirral speech to date, did not include laterals when investigating this variety.
It must be highlighted at this juncture, however, that this study does not draw upon sufficient quantitative data to establish whether the inhabitants of Liverpool and the Wirral possess conscious knowledge surrounding variation in lateral production, including lateral darkening strategies, However, the absence of overt commentary on lateral production in Merseyside in existing literature suggests that the variable may function as a first order indexical (Silverstein, Reference Silverstein2003), with little to no salience.
One may question the extent to which variation in lateral production, be it acoustic or articulatory, may be open to use for social purposes. In the context of Liverpool speech, Juskan (Reference Juskan2018) previously noted style-shifting for happy, despite the variable being taken to be entirely below speakers’ conscious awareness. Similarly, velar nasal plus (the presence of [g] after [ŋ] in <ing>), taken to be a Labovian indicator, was also subject to style-shifting. This raises the possibility that lateral production may function as a useful sociophonetic variable, even if it might operate below the level of conscious awareness for Merseyside inhabitants.
Research questions
Based on the above discussion, this study sought to address the following research questions:
RQ1: Do adolescent speakers from the Wirral and Liverpool differ in their production of laterals in onset versus coda position? For example, are speakers from Liverpool more likely to produce a clear-dark dichotomy than speakers from the Wirral?
RQ2: Do adolescent speakers from the Wirral and Liverpool differ in how they produce dark laterals (if produced at all), as observed through ultrasound analysis? For example, are adolescents from Liverpool more likely to employ velarisation as a darkening articulatory strategy over pharyngealisation?
Although we were most interested in uncovering the broad differences between Wirral and Liverpool adolescents, we collected data from females and males, as will be discussed shortly. On this basis, we additionally included sex an independent variable given the common claim that female speakers orient towards prestigious forms, as shown in auditory, acoustic and articulatory studies (e.g. Labov Reference Labov2001; Lewis et al. Reference Lewis, Mehrabi and de Leeuw2019; Trudgill Reference Trudgill1972). Given that the pharyngealised coda /l/ appears to form the standard within the UK (Turton Reference Turton2014), we anticipated that female speakers might display a preference for pharyngealisation as a darkening strategy. Although we had initially intended to include social class as a variable in this study, the variable ultimately could not be incorporated due to model convergence issues.
Methodology
Participants
Simultaneous audio and ultrasound data were collected and analysed for 16 participants, all 17–18 years of age. Nine participants were recruited from a mixed comprehensive school in Allerton, South Liverpool; seven Wirral participants were recruited from across two single-sex grammar schools. Participant information was collected through a background questionnaire. In this study, participants were not questioned about their gender identities, but rather their sex. It was felt that asking students about sensitive aspects relating to identity within a high school setting could be considered inappropriate.
Six of the Liverpool participants described themselves as female and three as males (see Table 1). With the exception of LM2, who identified as mixed race, all participants classified themselves as white British. Of the Liverpool participants, all were born in Liverpool and had only ever lived in South Liverpool. South Liverpool tends to be described as more middle class than the traditionally working-class north side of the city. As such, the accent is generally considered to be comparatively ‘weaker’ (Juskan Reference Juskan2018).
Table 1. Liverpool participant demographic information
Four female participants and three males were recruited from the Wirral (see Table 2). All identified as white British. Although WF3 was born in Liverpool, she had spent most of her childhood in a town located on the Wirral peninsula. Similarly, WM3, though born in Chester, had grown up on the Wirral. WM1 had spent four years in Germany (although in monolingual English environments), two years in Northern Ireland and three years in Cambridge, growing up on military bases, before moving to the Wirral at the beginning of high school (aged 11/12) and residing there for approximately 5 years when testing took place. Impressionistically, this speaker sounded as if he were from the area and his results did not appear dissimilar to those of the other Wirral males, so they were included.
Table 2. Wirral participant demographic information
Acoustic and articulatory data collection
All participants were tested in their respective high schools. Informants from the Liverpool and Wirral girls’ schools were tested in office spaces. Those from the Wirral boys’ school were tested in a classroom as these were the spaces made available to the first author. It was hoped that the school setting would lead to greater ecological validity than a linguistics lab would have. However, recent studies have indicated that the school setting itself may have had unintended consequences. Lampropoulou and Cooper (Reference Lampropoulou and Cooper2021) noted that in the predominantly middle-class setting of Wirral grammar schools, there was a ‘grammar school pressure’ from students and teachers, pushing students to aim for standard English and ‘speak in a normalised way’ (p. 11). A similar idea was noted by Byrne (Reference Byrne, de Velde, Hilton and Knooihuizen2021), who argued that Wirral speakers were more likely to style-shift in formal settings than Liverpool ones, believing themselves to speak ‘dead posh in school’ (p. 174). Given the relative formality of the school setting, it is possible that both Wirral and Liverpool speakers may have attempted to utilise a greater number of standard forms.
Prior to testing, participants were fitted with a head stabilisation set (Articulate Instruments Ltd 2008) to ensure ultrasound probe stability and subsequently enable accurate measurements (Scobbie et al. Reference Scobbie, Wrench, van der Linden, Sock, Fuchs and Laprie2008). Before recording for the experimental task began, the words ‘bite plate’ were displayed on a laptop screen. Each participant was asked to press their tongue upwards, firmly against the underside of an acrylic bite plate (modelled on that used by Scobbie et al. Reference Scobbie, Lawson, Cowen, Cleland and Wrench2011). The bite plate image was recorded using the equipment and settings described below. The xy coordinates were offset in relation to the bite plate trace. This was used to provide a degree of standardisation between speakers in terms of the orientation of tongue contours.
During the experimental task, participants each read a wordlist containing 170 real words, presented in five-word blocks. Each block of stimuli was presented for eight seconds on a laptop screen using Articulate Assistant Advanced V217.05 (Articulate Instruments Ltd 2018), hereafter referred to as ‘AAA’. Ultrasound data were collected using a microconvex ultrasound probe, with a frequency set to 5MHz, a depth 90mm, field of view of 100% and dynamic range of 68dB. There were an of 84 frames per second. Audio data were obtained using a FocusRite Scarlett Solo preamplifier and Audio-Technica AT2020 Cardioid Condenser microphone. Audio was sampled at a rate of 22,050 Hz.
Each speaker read through the wordlist twice, producing a total of 340 words. Fifteen monosyllabic CVC words containing onset /l/ and fifteen containing coda /l/ were included in the initial wordlist. Coda laterals preceded by /u/ were excluded from the final analyses for all speakers on the basis that the boundary between /u/ and coda /l/ could not be reliably ascertained. The words included in the final analysis are shown in Table 3 below. As each participant read the wordlist twice, this yielded 30 tokens of onset /l/ and 20 of coda /l/ per participant (total = 800). For the final analyses, 56 tokens were removed, resulting in 442 tokens of onset /l/ and 302 coda /l/.
Table 3. Target words from wordlist
Acoustic analysis of laterals in Liverpool and Wirral adolescents
Stimuli preparation
Praat textgrids were imported into ELAN version 5.2 (2018). Auto-alignment was conducted using FAVE align version 1.2.2 (Rosenfelder et al. Reference Rosenfelder, Fruehwald, Evanini, Seyfarth, Gorman, Prichard and Yuan2014) but manually corrected where necessary. A high-pass filter of 20Hz was used to ensure that low-frequency background noise was removed, thereby facilitating zero-cross marking. For onset productions, the lateral was deemed to begin when regular periodic voicing began in the waveform (see Figure 2 for an example). The appearance of a voice bar and the onset of a visible formant structure in the spectrogram was also used to help determine the onset of the lateral. The endpoint of the laterals in onset position was determined by increased wave height and intensity, indicating that the following vowel had begun. The first zero-crossing of the vowel was used to mark the endpoint of onset laterals.
Figure 2. Onset production of /l/ in the word ‘leg’, produced by WM2. The segmented lateral is represented through the capital ‘L’ in the top tier of the textgrid.
As has been noted by Kirkham (Reference Kirkham2017), ‘[t]he idea of a steady state for liquids is an approximation’ (p. 21). Nevertheless, in order to maintain consistency with existing literature (Carter and Local Reference Carter and Local2007; Kirkham Reference Kirkham2017; Kirkham et al. Reference Kirkham, Nance, Littlewood, Lightfoot and Groarke2019), coda laterals were deemed to begin where relatively stable F2 formant trajectories could be observed (see Figure 2). The endpoint of coda laterals was determined by a drop in intensity and the cessation of regular periodic waveforms.
Formant measurements
A Praat script was used to label midpoints in a point tier and extract F1 and F2 values at the lateral midpoint (see Figures 2 & 3). For male speakers, the maximum formant value was set to 5000Hz, whilst for females this number was set to 5500Hz. Formant frequencies were checked for each lateral in Praat. Frequencies for F1 and F2 were hand corrected to remove the need for outlier removal and prevent any further reduction in the number of usable tokens. First and second formant frequencies were obtained since higher F1 and lower F2 frequencies indicate darker laterals, whilst lower F1 and higher F2 values suggest clearer laterals (Carter Reference Carter, Local, Ogden and Temple2003; Carter and Local Reference Carter and Local2007; de Leeuw et al. Reference De Leeuw, Tusha and Schmid2018; Sproat and Fujimura Reference Sproat and Fujimura1993). As F2–F1 provides a degree of normalisation (Kirkham et al. Reference Kirkham, Nance, Littlewood, Lightfoot and Groarke2019; Turton Reference Turton2014; Turton and Baranowksi Reference Turton and Baranowski2021), no further normalisation was conducted.
Figure 3. Coda production of /l/ in the word ‘mill’, produced by WF1. The segmented lateral is represented through the capital ‘L’ in the top tier of the textgrid.
In line with RQ1, we wanted to find out whether adolescents from both the Wirral and Liverpool produced a higher F2–F1 in onset position and a lower F2–F1 in coda position, or whether the groups differed from one another in the presence versus absence of a clear-dark dichotomy.
Articulatory analysis of laterals in Liverpool and Wirral adolescents
Annotations from Praat were imported into AAA. Splines were automatically fitted and then manually corrected. Measurements were extracted at the lateral midpoint, as in the acoustic analysis. In total, 42 Cartesian coordinates of the tongue were extracted per ultrasound spline. However, ultrasound images for participant LF6 could not be used for the articulatory analyses, as the probe shifted from its initial mid-sagittal positioning during the course of the wordlist task, meaning that the tongue was not accurately captured for this speaker.
Therefore, of the 744 files used in the acoustic analyses, 638 of these tokens were used in the ultrasound analysis: 109 tokens were disregarded due to synchronisation errors, poor imaging, and the removal of data for participant LF6. As such, ultrasound data for 375 onset lateral tokens and 263 coda lateral tokens were collected for use in the ultrasound analyses.
Statistical analysis
All statistics were undertaken in R (R Core Team Reference Core Team2020). For the acoustic analysis, linear mixed effects models were produced using the lmerTest package (Kuznetsova et al. Reference Kuznetsova, Brockhoff and Christensen2017) to examine group differences. Model selection was undertaken using the step function. For fixed effects p values were calculated on the basis of Satterthwaite's approximation, whilst for random effects, p values were calculated on the basis of likelihood ratio tests. Post-hoc pairwise comparisons were conducted on the stepped-down models using the emmeans package (Lenth Reference Lenth2020). In each case the Kenward–Roger degrees of freedom method was used.
Following Westerberg (Reference Westerberg2020) and Coretta (Reference Coretta2020), GAMM plots were used to visualise tongue curves for each individual speaker separately, meaning that group level comparisons were not undertaken as in the acoustic analysis. In this case, models were constructed using the rticulate package (Coretta, Reference Coretta2021a). The polar_gam function was used, which first converts the Cartesian coordinates into polar ones and fits smooths using these coordinates (Coretta Reference Coretta2020). Polar coordinates were used as they can more accurately map tongue contours (Heyne and Derrick Reference Heyne and Derrick2015). Random effects were incorporated using factor smooths for word and flanking vowel.Footnote 1
Coordinates were then automatically converted back to the Cartesian coordinate system for display purposes, using the plot_polar_smooths function. In order to ascertain where tongue contours significantly differed, a method advocated by Sóskuthy (Reference Sóskuthy2017) was used, whereby difference smooths were plotted. In this case, plotting was done using the plot_difference function within the tidymv package (Coretta Reference Coretta2021b). A difference smooth is a curved line displaying the difference between two contours. Positive values indicate one curve to be significantly higher than another, whilst negative values demonstrate the reverse. Specifically, at those points along which the confidence interval for the difference smooth does not touch 0, the difference between contours was found to be significant. Although difference smooths are provided, analysis of the ultrasound data is principally based upon visual inspection of the GAMM output.
Therefore, through the acoustic analysis both inter- and intraspeaker inferential comparisons were possible, whilst for the articulatory analysis only intraspeaker inferential comparisons were possible, although interspeaker comparisons are descriptively reported.
Acoustic results
F2–F1 results
In order to investigate RQ1, and establish whether adolescents from the Wirral and Liverpool differed in their production of laterals in onset versus coda positions, F2–F1 was analysed as the dependent variable. The final model for the data (shown in Table 4) included fixed effects for position (onset versus coda), location (Wirral versus Liverpool), sex (female versus male) and an interaction term for location and sex. Random intercepts were included for flanking vowel (i.e. /ɪ/ or
Table 4. Type III Analysis of variance table for the fixed effects tested in the linear mixed effects model for F2–F1 (n = 744)
/ɛ/). By participant random slopes and intercepts were included for position. Position (F(1, 15.302) = 79.269, p < .001), location (F(1, 11.906) = 6.618, p = .025), sex (F(1, 11.905) = 10.864, p = .006), and the interaction between location and sex (F(1, 11.905) = 10.990, p = .006) were found to be significant. Onset and coda laterals were realised distinctly by all speaker groups and both Wirral and Liverpool adolescents exhibited clearer production in onset, relative to coda position. Therefore, with regards to RQ1, it was indeed the case that adolescents from both the Wirral and Liverpool produced a clear-dark dichotomy as observed through the acoustic analysis of F2–F1.
To further explore the interaction between location and sex, post-hoc pairwise comparisons were conducted using emmeans. Considering the contrast between Liverpool and Wirral adolescents as a function of sex (Table 5), a significant difference was found for males (t-ratio = 3.474, p = .005), with Liverpool males showing higher F2–F1 (i.e. clearer) values than Wirral males (i.e. darker) overall. Liverpool and Wirral females were not found to differ significantly in terms of F2–F1 values (t-ratio = -0.554, p = .590).
Table 5. Pairwise comparisons of estimated marginal means for F2–F1 on the basis of location as a function of sex. Differences between estimated marginal means for each contrast are displayed in the ‘Estimate’ column, alongside standard errors (‘SE’), degrees of freedom (‘DF’), t-ratios and p-values
A second set of comparisons (Table 6) considered the contrast between males and females as a function of location. Liverpool females were not found to differ from Liverpool males with regards to F2–F1 values (t-ratio = -0.013, p = .990), whereas Wirral females differed from Wirral males (t-ratio = 4.117, p = .001). In this case, Wirral females displayed significantly higher (i.e. clearer) F2–F1 results than Wirral males (i.e. darker).
Table 6. Pairwise comparisons of estimated marginal means for F2–F1 on the basis of sex as a function of location. Differences between estimated marginal means for each contrast are displayed in the ‘Estimate’ column, alongside standard errors (‘SE’), degrees of freedom (‘DF’), t-ratios and p-values
Figure 4 shows that onset and coda results for Wirral males were focused lower in F2–F1 space than they were for the other three groups.
Figure 4. Differences between F2–F1 for Liverpool and Wirral females (left) and males (right). Mean values and confidence intervals, taken from the emmeans output, are displayed alongside observed values, which are represented as points.
In summary, with regards to RQ1, acoustic analysis of F2–F1 revealed that adolescents from both the Wirral and Liverpool produced a clear-dark dichotomy, but that the laterals of Wirral males were darker overall.
Ultrasound results
To examine RQ1 and establish whether adolescents from the Wirral and Liverpool differed in their production of laterals in onset versus coda position, tongue contours were first examined using ggplot2 (Wickham Reference Wickham2016) in R. The images below show mean tongue splines for onset and coda productions for all adolescents. Raw Cartesian values were used for the generation of these splines. Along the x-axis one can see the anterior portion of the tongue to the left of each image and the root to the right. Along the y-axis, one can see tongue body height: the higher up the y-axis, the greater the height.
In Figure 5, it appears that all adolescents differentiated between onset and coda /l/ to some extent, although this was least clearly the case for Wirral males WM1 and WM2. Despite the presence of interspeaker variation (Figure 5), one can observe that the front of the tongue was typically lower for coda than onset productions, as one would anticipate with a darker production. Furthermore, the back of the tongue was typically more retracted for coda than onset /l/, revealing a darker lateral production in coda position. In general, these results align with the F2–F1 data which revealed an overarching clear-dark dichotomy, but with Wirral males producing darker variants in both positions. Just as Wirral males appeared to show less of a distinction between onset and coda lateral productions than other adolescents in terms of F2–F1, they also appeared to show less of an articulatory distinction between the laterals.
Figure 5. Ultrasound spline images for onset (grey) and coda (black) lateral productions. Splines are shown for all adolescents, excluding LF6.
GAMM analyses
In order to investigate RQ2 and establish whether adolescents from the Wirral and Liverpool differed in how they produced dark laterals, the results from the GAMM analyses are displayed in Figure 6. Looking at the difference smooths, all speakers were found to produce statistically significant distinctions differences between onset and coda laterals (as shown by the areas highlighted in red). However, the difference smooths are not particularly informative when it comes to examining the differences between velarisation and pharyngealisation. Therefore, in this section, the results are interpreted visually (as done elsewhere, e.g. Mousikou et al. Reference Mousikou, Strycharczuk, Turk and Scobbie2021). Liverpool males, LM1, LM2 and LM3 appeared to display greater preference for velarised, rather than pharyngealised, laterals in coda position. More specifically, the Liverpool males all showed clear raising of the dorsum towards the velum. Of note, however, the same can also be seen in the case of two of the Wirral males: WM2 and WM3, again indicating velarisation during coda lateral production.
Figure 6. GAMM plots of tongue contours, with difference smooths displayed underneath, for laterals produced by the male adolescents. The difference smooth shows the height of the onset lateral relative to the coda one.
Figure 7 below more clearly illustrates the process of velarisation, focusing on speaker WM3. In the image, the posterior dorsum can be seen to display greater elevation for the coda lateral relative to the onset one. It is this elevation of the dorsum which indicates that the coda lateral is velarised. Although speaker WM2 displayed a small degree of root retraction for coda productions, not apparent for the other adolescents, there was nevertheless prominent raising in the posterior region of the dorsum, suggesting that velarisation served as the coda darkening mechanism.
Figure 7. GAMM plot of tongue contours for lateral productions by Wirral male speaker WM3.
In contrast to the male participants, female adolescents displayed a greater preference for retraction as a darkening strategy for coda /l/, i.e. pharyngealisation (see Figure 8). In relation to the male adolescents, the female adolescents displayed lower degrees of velar raising in coda than onset productions. This is illustrated most clearly in Figure 9 below, depicting the GAMM results for Wirral female WF3. In this example, it can be observed that the posterior region of the dorsum was lower in coda position than onset and instead extensive root retraction was evident.
Figure 8. GAMM plots of tongue contours, with difference smooths displayed underneath, for laterals produced by the female adolescents. The difference smooth shows the height of the onset lateral relative to the coda one.
Figure 9. GAMM plot of tongue for lateral productions by Wirral female speaker WF3.
Therefore, in summary for RQ2, the GAMM analysis showed that males displayed a preference for velarisation as a darkening mechanism whilst females showed a preference for pharyngealisation and that this was the case for both Liverpool and Wirral adolescents.
Discussion
Our first research question sought to address whether adolescent speakers from the Wirral and Liverpool differed in the production of onset versus coda laterals. We wanted to establish whether Liverpool or Wirral speakers were more likely to produce a clear-dark dichotomy. According to both acoustic (F2–F1) and articulatory results, both Liverpool and Wirral teenagers tended to differentiate between lighter onset laterals and darker coda laterals. However, exploration of the F2–F1 results suggested that Wirral males displayed less of a contrast than other adolescents and opted for darker lateral productions in both positions. This could also be seen in the GAMM results, where Wirral males appeared to produce less of an articulatory distinction between onset and coda laterals than other adolescents, overall. On the basis that Wirral males appeared to exhibit greater darkening, it appears that Scouse has not entirely ‘ousted’ (Knowles Reference Knowles1973, 14) the Wirral dialect. If Scouse had ‘ousted’ the Wirral variety, we would have expected no difference between the two varieties. Our finding that Wirral males performed differently than Liverpool males aligns with a growing body of work highlighting subtle distinctions between the two accents (Byrne Reference Byrne, de Velde, Hilton and Knooihuizen2021; Cooper and Lampropoulou Reference Cooper and Lampropoulou2021; Lampropoulou and Cooper Reference Lampropoulou and Cooper2021; Newbrook Reference Newbrook, Foulkes and Docherty1999).
Given that numerous Northern dialects possess a small contrast between initial and coda laterals (e.g. Carter and Local Reference Carter and Local2007; Kirkham et al. Reference Kirkham, Turton and Leemann2020; Turton Reference Turton2014), Wirral males could potentially be tapping into a feature of Northern English. The greater similarity between onset and coda variants for Wirral males could then be taken to represent their greater alignment with the North more broadly. It is possible that Liverpool speakers have a Northern identity which comes secondary to a Liverpool one, whereas for Wirral males a Northern identity is more central (see Juskan Reference Juskan2018 for a discussion of the secondary role played by a Northern identity for younger Liverpool speakers). Interview data would be beneficial to help uncover the extent to which this is true.
Our second research question concerned whether Wirral and Liverpool adolescents differed in the articulatory strategies used to bring about dark /l/. The ultrasound results demonstrated that both velarised and pharyngealised coda laterals were produced by speakers from the Wirral and Liverpool. However, for male speakers from both Liverpool and the Wirral, the dorsum was more commonly raised towards the velum in coda than in onset position, whereas female adolescents typically opted for pharyngealisation as a darkening strategy. The latter darkening mechanism appears to be more common nationally (see Turton Reference Turton2014).
This is not the first study to find such a distinction between male and female speakers. In an MRI and EPG study of /l/, Narayanan et al. (Reference Narayanan, Alwan and Haker1997) found that two out of four tested American English speakers produced dark /l/ with a significantly raised tongue body in the velar region, whilst the other two displayed greater tongue backing. The two speakers displaying velarisation were male, whilst the two that displayed pharyngealisation were female. This raises the possibility that the articulatory differences observed in that study and the present one reflect sex-based variation in the size and shape of the vocal tract. However, female speakers can and do utilise velarisation as a lateral darkening strategy (as seen in Lawson et al. Reference Lawson, Stuart–Smith and Scobbie2018, for example), suggesting that physiological differences may not entirely explain lateral darkening strategies. Indeed, in the current study, velarisation was present in two Wirral female speakers: WF1 and WF4, for example. Therefore, it seems that physiological, sex-based differences may not have been the source of the different articulatory strategies observed.
In this study, differences between velarisation and pharyngealisation did not surface at the acoustic level, consistent with the findings of Turton (Reference Turton2014) and Charles and Lulich (Reference Charles and Lulich2018). However, that is not to say that there are no acoustic means by which hearers might be able to discern distinctions between the two darkening mechanisms. It may simply be the case that the commonly used metrics of F1 and F2 are insufficient to determine the differences. For example, it has been observed that covert articulatory variation may be uncovered by looking at less commonly studied acoustic variables like F3 and F4 (Johnson Reference Johnson2011; Johnson and Bakst Reference Johnson and Bakst2016; Lawson et al. Reference Lawson, Stuart–Smith and Scobbie2018).
An additional possibility is that speakers are attending to more than just the acoustic signal, and it is this which is providing them with a means by which to differentiate lateral darkening strategies. Knowles (Reference Knowles1973) observed that Liverpool speech is generally velarised; that is to say, velarisation is a broader property of the variety, rather than a property inherent to laterals. For male speakers, it may not be that the velarisation of coda /l/, specifically, constituted an articulatory target, but rather that velarisation was consistent with a broader general articulatory setting.
Interestingly, Knowles (Reference Knowles1973, Reference Knowles and Trudgill1978) drew attention to a potential link between velarisation and a close-jaw external articulatory setting. Biomechanically, jaw lowering leads to the retraction of the tongue root towards the pharyngeal wall (Mooshammer et al. Reference Mooshammer, Hoole and Geumann2007), i.e. pharyngealisation. In contrast, velar consonants have been observed to show less jaw opening than pharyngeals (Elgendy Reference Elgendy1999), suggesting that there may be some physiological underpinning to the relationship suggested by Knowles. The co-occurrence of velarisation with a close-jaw setting offers something that listeners can visually attend to. Even if speakers may not be able to hear or observe velarisation versus pharyngealisation, they might nevertheless visually observe differences in external articulatory setting.
A close-jawed articulatory setting has previously been considered ‘the object of social affect’ (Labov Reference Labov1963, 307). Labov noted that ‘linguistic variables’ can be ‘variously affected by the overall tendency towards a favored articulatory posture’ (p. 308), suggesting that jaw settings themselves can operate as a socially meaningful articulatory targets. Similarly, Pratt and D'Onofrio (Reference Pratt and D'Onofrio2017) noted that ‘imbuing of jaw setting with social meaning entails the assignment of social meaning to the jaw setting's acoustic results’ (p. 308). Rather than having an impact on acoustic results in this case, it is possible that a close-jaw setting may have been used to bring about an articulatory difference in terms of lateral darkening, without a clear acoustic correlate.
Regarding the potential meaning of such a setting, it is worth considering the opposite pattern: open-jaw setting. As discussed in Levon and Holmes–Elliott (Reference Levon and Holmes–Elliott2024), the presence of lax, lowered and immobile jaw settings has been linked with elites in both the UK and US. In the UK context, in particular, the use of an open-jaw setting has been associated with upper middle-class and upper-class speech, exemplified in parodic performances such as the ‘Gap Yah’ series of sketches performed by Matt Lacey (Levon & Holmes–Elliott Reference Levon and Holmes–Elliott2024). Given that Liverpool speech is generally characterised as ‘urban, working-class, and male’ (Cooper and Lampropoulou Reference Cooper and Lampropoulou2021, 109), it could be that a close-jaw setting may be linked in some fashion to covertly desirable working-class and masculine attributes. If articulatory setting and velarisation are linked in the manor suggested by Knowles, this could potentially explain the use of velarisation by Liverpool males.
However, this would not explain why Liverpool females did not also opt for velarisation, or why Wirral males did. Although it seems likely that differences in lateral darkening strategies, and potentially differences in articulatory setting, could be below speakers’ level of conscious awareness, it is nevertheless possible that attention to speech resulted in broader shifts in production. For instance, Juskan (Reference Juskan2018) observed style-shifting for a variable below speaker's conscious awareness in Liverpool speech. He linked artificial speech contexts to greater attention to speech, which, in turn, lead participants to ‘focus on how they were articulating’ (p. 211). Given the relatively unnatural character of wordlist reading tasks, undoubtedly confounded by the artifice of ultrasound tongue imaging, it seems plausible that speakers in this study may also have been focussing on how they were articulating, even for variables they may not have consciously been aware of.
Liverpool speakers have been noted to style shift, ‘avoiding drawing upon a Scouse repertoire in more formal situations where they wish to present themselves positively’ (Byrne Reference Byrne, de Velde, Hilton and Knooihuizen2021, 175). Moreover, Merseyside inhabitants have been shown to be aware that Scousers can be perceived as ‘uneducated’ (Byrne Reference Byrne, de Velde, Hilton and Knooihuizen2021, 172) and ‘thick’ (Juskan Reference Juskan2018, 190). Despite identifying as Scouse, some speakers, particularly Liverpool females, may have wished to avoid the associations held towards the stereotypical Scouser. Although there are problems in drawing links between sex, class and prestige (James Reference James, Bergvall, Bing and Freed1996; Talbot Reference Talbot2019), it is commonly claimed that female speakers typically orient towards overtly prestigious forms whereas males do not (e.g. Labov Reference Labov2001; Trudgill Reference Trudgill1972). This may have resulted in Liverpool females utilising more ‘standard’ articulatory patterns when attending to their speech, resulting in changes to articulatory setting. This, in turn, may have resulted in the production of more ‘standard’ pharyngealised dark /l/.
In contrast, males, regardless of class, have been found to utilise non-standard speech production patterns on the basis that they hold covert prestige (Trudgill Reference Trudgill1972). Males have been observed to draw upon ‘specific linguistic strategies that are positively associated with canonically working-class male traits such as strength, physicality and toughness’ (Lawson Reference Lawson2020, 419). In this case, whether directly or indirectly, the use of velarisation may be linked to a broader desire to tap into such ‘desirable masculine attributes’ (Trudgill Reference Trudgill1972, 183) associated with Scouse, overruling the desire to avoid some of Scouse's more negative associations. The similarity in lateral darkening strategies exhibited by Liverpool and Wirral males may have resulted from the shared use of a ‘Scouse’ articulatory setting.
As noted by Juskan (Reference Juskan2018), the non-salience of a variable enables it to be used ‘sub-consciously’, providing speakers with a means to express ‘a regional identity without noticeably deviating from their local accent’ (p. 219). In this case, Wirral male speakers may have utilised an articulatory feature, be it velarisation or wider articulatory setting, on the basis of a Merseyside identity. At the same time, the acoustic results of employing a ‘Scouse’ articulatory configuration, rather than more ‘standard’ one, are the same, meaning that Wirral males would not be deviating from their local accent, or even broader Northern norms. In fact, the potential non-salience of the feature may be precisely why velarisation as a darkening mechanism may have been open to utilisation in the first place. For instance, Cooper and Lampropoulou (Reference Cooper and Lampropoulou2021) previously observed that adolescent Wirral females selectively drew upon the Scouse repertoire, but nevertheless tended to avoid salient features of Scouse.
Unfortunately, this study does not incorporate jaw data, nor does it examine speakers’ awareness of broader articulatory setting. It could also be interesting to consider parodic performances of Scouse, such as those by the comedian Harry Enfield, where it is clear that the Scouse persona is linked to a cluster of physical gestures, from the movements of the arms to the positioning of the jaw and lips. Such a study could highlight the benefits of considering more than just what is happening inside the mouth during Merseyside speech production.
Due to the small sample size, the results presented here must be considered with caution. However, the findings suggest potentially fruitful topics for further investigation. In particular, it may prove worthwhile to examine articulatory setting and its relationship with lateral darkening mechanisms. Future studies may benefit from considering a wider variety of articulatory parameters, incorporating additional equipment, such as lip and jaw tracking (as in studies into articulatory setting, e.g. Wilson and Gick Reference Wilson and Gick2014). The social interpretation of results presented here is speculative. Unfortunately, insufficient data was collected to provide a nuanced exploration of the relationship between gender, class and lateral production. Future studies would benefit from incorporating richer qualitative data, considering social factors directly in relation to Merseyside speakers’ awareness of lateral production and articulatory setting.
Conclusion
This study investigated lateral production in Merseyside speech. It was found that that Wirral and Liverpool adolescents differentiated between onset and coda laterals, overall, although Wirral males tended to produce darker laterals in both onset and coda positions. Differences in darkening strategies were also observed: in both Liverpool and Wirral adolescent males, velarisation was generally preferred, whilst pharyngealisation was preferred by most, but not all, females. These findings highlight the benefits of considering articulatory information in order to enhance our understanding of covert variation.
Dr SCOTT LEWIS is a postdoctoral researcher at the University of Lausanne. He holds a PhD from Queen Mary University of London. During his PhD research, Scott analysed variation in Merseyside speech, drawing on acoustic and articulatory methods. He is currently working alongside Professor Esther de Leeuw on a project entitled ‘Mother Tongues’. In this project, ultrasound tongue imaging is being used to better understand language acquisition in infants and small chidren. Email: [email protected]
ESTHER DE LEEUW is a Professeure Associée at the University of Lausanne. She specializes in the production, processing, representation, development and acquisition of speech sounds in the context of bilingualism and multilingualism. Before moving to Switzerland with her family, she was a Reader in Linguistics at Queen Mary University of London where she founded and directed the Phonetics Laboratory. Currently, she is leading an SNSF funded project called Mother Tongues to investigate tongue movements in bilingual babies using ultrasound technology. Email: [email protected]
Open access
