This paper investigates patterning of /l/ in Newfoundland English. Using acoustic and ultrasound methods, the reported displacement of the traditional Irish pattern of word-final light /l/ is assessed. Acoustic results show darker /l/'s in word-final position in both phrases and compounds. Although the standard allophonic pattern is widespread in Newfoundland English, dialectal variation arising from early settlement patterns continues to influence speech patterns with less distinction between initial and final /l/ in Irish-settled areas. Men show relatively less distinction between initial and final /l/, consistent with sociolinguistic patterns in which men retain local variants. Last, light /l/ in final position may be resurfacing among younger speakers. Ultrasound imaging also shows variable rates of distinction between word-final and initial /l/, but without significant main effects of region or gender. Articulatory analysis reveals a small effect of age, with older speakers being less likely to have significant differences in articulation across positions. An interaction between region and gender shows males from an Irish-settled community are less likely to employ distinct lingual shapes across positions. While some articulatory findings complement the acoustic results, it is suggested that differences between these domains result from lateralization or other aspects of articulation not captured in ultrasound imaging.

Using experimental methods, this paper investigates acoustic and articulatory properties of /l/ in Newfoundland English. In contrast with standard North American English, which has a dark [ɫ] in coda position and relatively light [l] in onsets, Irish-settled areas of Newfoundland are reported to exhibit light /l/ in both word-initial and word-final positions (Paddock and Memorial University of Newfoundland, Department of Linguistics, 1982). This is consistent with the realization of /l/ in Irish English (Jones, 1956). Since the earliest studies of /l/ in Newfoundland (Atkinson, 1982; Bartlett, 1977; Paddock and Memorial University of Newfoundland, Department of Linguistics, 1982), major socio-economic changes have led to suppression of many traditional dialect features (Clarke, 2012). Patterning of /l/, however, has not been included in studies of recent linguistic changes in Newfoundland (e.g., Clarke, 2012; Van Herk et al., 2007). This investigation will shed light on the extent to which light /l/ in coda persists among Newfoundland dialects and how variation in /l/-darkness is distributed according to the sociolinguistic variables of age and gender.

Using both acoustic and ultrasound methods, we investigate whether the allophonic pattern associated with mainland Canada is displacing the traditional Irish /l/ pattern in Newfoundland. We find significant effects of gender, age, and dialect zone on acoustic measures and a significant effect of age on the articulatory measures. The articulatory measures also showed an interaction effect of region and gender but no main effects for either of these social factors. This project contributes to literature showing that sociolinguistic and dialectal variation can be elicited using instrumental techniques in laboratory settings (e.g., Lawson et al., 2013; Lawson et al., 2008; Scobbie and Pouplier, 2010).

/l/ is a lateral resonant consonant realized with both a raising gesture of the tongue tip and a retraction and lowering gesture of the tongue body. The relative timing and magnitude of these gestures varies substantially according to linguistic contexts as well as across distinct languages and dialects. Different realizations of /l/ are traditionally classified as one of two categories, dark [ɫ] and light [l]. In terms of articulation, dark [ɫ] has greater tongue root retraction and tongue body lowering than light [l]. The variants also differ in terms of gestural timing with the tongue body gesture of dark [ɫ] preceding the tongue tip gesture and simultaneous gestures in light [l] (Sproat and Fujimura, 1993). Light [l] is characterized acoustically by a higher F2 and a lower F1, while dark [ɫ] has a relatively low F2 and high F1.

The realization and distribution of dark and light variants differs between dialects of English and across languages. Although the majority of literature on dialectal variation in /l/ is descriptive (e.g., Hughes et al., 2005; Jones, 1956; Wells, 1982) articulatory and acoustic studies of dark and light /l/ have been carried out, both in the context of English dialect variation (Scobbie and Wrench, 2003; Turton, 2014; van Hofwegen, 2009) and crosslinguistically (e.g., Recasens, 2012). Sociolinguistic studies have shown evidence of changes in progress (e.g., van Hofwegen, 2010) as well as effects of ethnicity and gender. The acoustic and articulatory study of Bradford English of Kirkham and Wormald (2015) found British Asian speakers to have lighter /l/'s than British Anglo speakers. The study of Morris (2013, p. 205) of Welsh/English bilinguals in North Wales found women to have lighter /l/'s than men in both English and Welsh and also found women to have a greater difference in /l/ darkness between the two languages with significantly lighter onset /l/'s in Welsh than in English.

Many dialects of English, including standard North American English, are described as having dark /l/ in codas and light /l/ in onsets. Linguistic factors which contribute to /l/-darkness are more complex than the coda/onset distinction, however. Coarticulatory effects result in lighter /l/'s before front vowels than before back vowels (e.g., Lehiste, 1964; Recasens, 2012). Morphological constituency affects /l/-darkness with darker /l/'s in morpheme-final position relative to morpheme-initial position (Lee-Kim et al., 2013). Variation in the realization of /l/ also correlates with rime duration with darker /l/'s occurring in longer rimes (e.g., Sproat and Fujimura, 1993; Yuan and Liberman, 2011). Thus, in addition to syllabic position, relevant linguistic factors which affect /l/-darkness include adjacent vowel quality, morphological constituency, and rime duration.

Although dark and light /l/ are often referred to as distinct allophonic categories (e.g., Halle and Mohanan, 1985), instrumental work has argued for a continuum of /l/ darkness with relative degree of darkness (Sproat and Fujimura, 1993) conditioned by factors such as rime duration and vocalic context. Previous work on Newfoundland English (Clarke, 1984) has used an impressionistic methodology in which each token of /l/ is categorized as either light or dark, implicitly assuming a dual-category analysis of dark and light /l/. Although we acknowledge ongoing debate concerning whether the difference between dark and light /l/ should be treated as categorical or gradient (e.g., Turton, 2014), our study uses a gradient acoustic measure to determine degrees of /l/-darkness and a categorical articulatory metric to examine differences in the realization of /l/ across positions in a word.

Newfoundland and Labrador is the most easterly province of Canada and consists of two geographically distinct regions, the island of Newfoundland and the mainland portion of the province, Labrador. Newfoundland has been described, linguistically, as “the most homogeneous province in Canada” (Clarke et al., 1999). This claim refers to the 98% of the population who speak only English. Yet, due to historical patterns of migration and settlement, geography, and rapid socioeconomic change, English in Newfoundland is far from monolithic. Rather, a number of regional varieties exist in the province, which has prompted the claim that Newfoundland shows “the greatest regional diversity to be found anywhere in North America” (Clarke and Hiscock, 2012). In this section, we provide a short overview of this diversity to set the context for the current study.

Dialectal diversity in the province is shaped by patterns of European settlement with the majority of settlers coming from southwest England and southeast Ireland to participate in the fishing industry (Clarke, 2010). Economic dependence on fishing also affected the geography of settlement patterns, with settled areas including geographically dispersed coastal communities who had little contact with each other.

For the island portion of the province, five dialect zones and three transitional areas that sit between them were identified by Paddock and Memorial University of Newfoundland, Department of Linguistics (1982) (see Fig. 1). The major distinctions are known as the English-North (D1), Irishized Avalon Peninsula (D2), English-South (D3), Southern-West Coast (D4), and Northern-West Coast (D5). St. John's (TA1), Placentia Bay (TA2), and Corner Brook (TA3) are considered areas of dialect contact, giving rise to varieties with a hybrid of features.

FIG. 1.

Dialect zones and transitional areas of Newfoundland, plotted using SimpleMappr (Shorthouse, 2010).

FIG. 1.

Dialect zones and transitional areas of Newfoundland, plotted using SimpleMappr (Shorthouse, 2010).

Close modal

The Avalon Peninsula (TA1 and D2 in the figure) is the most densely populated area of the province. It is also home to the province's capital city, St. John's, where the majority of data for this study was collected. A majority of speakers reported on here are from the Avalon. In 1961, just under half the population of the island lived on the Avalon. The population of the Avalon has increased steadily since this time with the majority of the province's approximately 527 000 residents now living on the peninsula. Throughout this period, St. John's has retained approximately three-quarters of the Avalon's population (Statistics Canada, 2016).

The main cultural division on the Avalon is between descendants of English and Irish settlers. Census data in 1961 showed that 75% of the population claimed English ancestry, while only 16% declared themselves of Irish origin (Government of Newfoundland and Labrador, 1970). By 2011, this levelled out to 43% English and 21% Irish (Statistics Canada, Newfoundland and Labrador, 2013). Although the peninsula is comprised largely of these two cultural groups, Seary et al. (1968) identified five distinct regions marked by pronunciation differences. These are identified on the map in Fig. 2.

FIG. 2.

Dialect zones of the Avalon Peninsula, 1 = Northern Shore Line, 2 = Bay Roberts, 3 = St. John's, 4 = Southern Shore Line, 5 = Placentia, plotted using SimpleMappr (Shorthouse, 2010).

FIG. 2.

Dialect zones of the Avalon Peninsula, 1 = Northern Shore Line, 2 = Bay Roberts, 3 = St. John's, 4 = Southern Shore Line, 5 = Placentia, plotted using SimpleMappr (Shorthouse, 2010).

Close modal

With respect to reported allophonic variation of /l/, the dialect zones of the Avalon can be described as follows. The Southern Shore Line extends along the southeastern region of the Avalon. This region is known as the Irish Loop and has been noted to exhibit the absence of any allophonic variation with clear /l/ in all positions, as is the Irish pattern. Seary et al. (1968) note that clear /l/ is “one of the prominent special qualities which identify the Southern Shoreline dialect.” However, Placentia, which is considered distinct from the Southern Shore Line, also has a history of Irish settlement and also exhibits clear /l/ in codas. Although observations of clear /l/ are not restricted to solely Irish settled areas, Paddock notes “clear pronunciations… were brought to Newfoundland almost exclusively by non-English settlers (i.e., Irish, Scots, and French)” (Paddock and Memorial University of Newfoundland, Department of Linguistics, 1982, p. 5). In contrast to Placentia and the Southern Shore Line, the Northern Shore Line was settled primarily by English settlers. As a result, the dialect exhibits dark /l/ in final positions, as a syllabic consonant, and next to low and back vowels. Paddock also reports the presence of a vocalized variant, that occurs in coda clusters (e.g., salt). In addition, a small minority of clear /l/ variants can be found, mostly next to high front vowels. English-settled areas are not as unified as the Southern Shore Line and include regions along the southern, western, and northern coasts. Speakers of the Bay Roberts Dialect exhibit dark /l/ in codas in most cases, but variants can be clear next to high front vowels. At the time of the study of Seary et al. (1968), St. John's showed a strong presence of clear /l/ in all positions, especially noticeable in the Catholic population (Seary et al., 1968, p. 71). St. John's is a transitional area with both Irish and English influences. The allophonic pattern with dark /l/ in codas, also identified in St. John's, is attributed mainly to Northern Shore Line dialect speakers and others who moved to the city from elsewhere.

More recently, Clarke (2012) identifies a diachronic change in the St. John's system. Based on the Sociolinguistic Survey of St. John's English (Clarke, 1984), Clarke shows light /l/ to be the majority variant for those born in the 1920s or earlier, including both “Irish-origin and non-Irish-origin speakers” (Clarke, 2012, p. 510). In the 1980s, though, dark word-final /l/ made up the community norm, with women using significantly less word-final light /l/ compared to men (Clarke, 2012, p. 507). Clarke identifies word-final light /l/, previously a norm among St. John's speakers of various backgrounds, as a feature that has become more narrowly associated with working-class speakers of Irish ancestry (Clarke, 2012, p. 509).

While Clarke's (2012) work is based on data collected in the 1980s, a recent study of dark and light /l/ in St. John's (Pierson, 2016) confirms that an allophonic distinction between dark and light /l/ is the contemporary community norm in the city. Pierson (2016) uses acoustic measures of /l/ tokens collected in sociolinguistic interviews and shows that women have a greater distinction between initial and final /l/'s than do men. Although Clarke (2012) and Pierson (2016) provide impressionistic and acoustic data, respectively, on the patterning of /l/ in St. John's, no recent study, or instrumental study of any kind, has investigated the realization of /l/ in other areas of the province. The study reported below seeks to add articulatory data to the sources of knowledge on /l/ allophony in Newfoundland, as well as data from speakers outside of the capital city.

Participants (n = 46) were recruited using posters, social media, and the friend-of-a-friend method. Two of these participants are excluded from the data reported here. One was from Labrador and we chose to use only speakers from the island of Newfoundland. The other excluded speaker did only one repetition of items, rather than the two repetitions established as the procedure for the rest of participants. Data from this participant was excluded for this reason.

Table I, below, shows the distribution of the 44 participants included in our analysis by dialect zone and gender. Their ages ranged from 18 to 65, with a mean age of 37.5 and standard deviation of 14.27 years.

TABLE I.

Speaker breakdown by gender and dialect zone, mean age in parentheses.

Dialect zone
GenderSt. John'sNorthern Shore LineSouthern Shore Line
Female 7 (45) 12 (37) 5 (41) 
Male 9 (35) 4 (36) 7 (38) 
Total N 16 16 12 
Dialect zone
GenderSt. John'sNorthern Shore LineSouthern Shore Line
Female 7 (45) 12 (37) 5 (41) 
Male 9 (35) 4 (36) 7 (38) 
Total N 16 16 12 

Each of the participants was born and raised in Newfoundland and learned a local variety of English as their first language. The majority of speakers are from the Avalon Peninsula and were categorized straightforwardly according to the St. John's, Northern Shore Line, and Southern Shore Line dialect zones of Seary et al. (1968). One speaker was from Bay Roberts and was classified with the Northern Shore Line group as both areas have primarily English influences and are reported to have word-final dark /l/'s. Five speakers were from various locations outside the Avalon Peninsula, three from the Corner Brook transitional zone and two from the English North. These are all areas reported to have an allophonically conditioned light/dark distinction and we grouped these speakers with our Northern Shore Line speakers. All but seven speakers lived in St. John's at the time of the recording and performed the experiment in a sound attenuated room at the Speech Sciences laboratory at Memorial University. The others were recorded in a location more convenient for them (i.e., at their home or place of employment). Consequently, this group did not participate in the ultrasound component of the study. Each participant filled out a demographic survey and informed consent form before starting the experiment, which lasted approximately 35 min, and was paid $10 at the end of the session.

Participants who were recorded in the laboratory were seated in an ear, nose and throat chair with an adjustable headrest, which comfortably supported and minimized head movement. A Sonosite Titan C11/8–5 MHz ultrasound transducer was secured below the speaker's chin with a Manfrotto 244 variable Magic Arm. While the speaker read the stimuli, ultrasound video of their lingual activity was digitized using a Canopus ADVC-55 and saved as an AVI file using Adobe Premiere Elements video editing software. Audio recordings were made using a Logitech H390 USB recording headset with a sampling rate of 44 100 Hz.

To elicit productions of /l/ in word-initial and word-final positions, participants were asked to read target forms within short carrier sentences (“please say ___ again”) presented one-by-one on a computer monitor using matlab. Elicitation materials were from previous work (Mackenzie et al., 2014) and included intervocalic /l/'s preceding and following a word boundary. Five items contained examples of word-initial and word-final /l/ in phrases (e.g., stealassets vs seelapses).1 Six items included word-initial and final /l/ in compounds (e.g., coilamp vs toylamp). For each comparison, the vowel preceding and following the /l/ was the same in each condition (e.g., word-final vs word-initial). Some items included proper names or unusual word sequences and participants were also shown a context sentence in order to clarify the meaning of each target sequence. A list of stimuli is included in the  Appendix. In addition to target items, participants produced an equal number of fillers. All items, including fillers, were repeated once, yielding two repetitions of each utterance.

In order to carry out the acoustic analysis, phoneme boundaries in each recorded utterance were automatically identified and aligned using prosodylab-aligner (Gorman et al., 2011). F1 and F2 measurements were taken at five proportional points across the duration of the /l/ using praat (Boersma and Weenink, 2014). Tokens with F1 values above 1000 Hz and F2 values above 3000 Hz were assumed to be praat errors and were excluded from the analysis. The remaining F1 and F2 values were normalized using the Lobanov z-score transform (Lobanov, 1971) in norm (Kendall and Thomas, 2009) to facilitate a comparison of darkness measures across speakers. Based on these z-score values, /l/-darkness was calculated as the difference between z-score normalized F2 and normalized F1. In order to minimize coarticulatory effects of adjacent vowels, and following previous work on English /l/ (e.g., Sproat and Fujimura, 1993; Huffman, 1997; van Hofwegen 2010), formant values from the temporal midpoint were used to calculate our measure of /l/-darkness. Dark /l/'s generally have a higher F1 and lower F2 relative to light /l/'s. The zF2-zF1 value is a measure of overall darkness with higher zF2-zF1 values being interpreted as lighter /l/'s (Sproat and Fujimura, 1993). Statistical analysis of darkness scores was carried out using mixed model regression with darkness as the dependent variable, condition and social factors as fixed effects, and random intercepts for participant and item.

Because some speakers did not participate in the ultrasound study and poor image quality was obtained for others, contours from only 28 of the 44 participants are examined in this analysis. Gesturally, /l/ is a particularly complex consonant, involving at least three coordinated, dynamically unfolding events. This required us to decide on a measurement that represents the production of /l/-for this report, we focused on lowering and retraction differences that have been previously observed for initial and final contexts (e.g., Sproat and Fujimura, 1993).

Following previous articulatory studies of /l/-darkness (e.g., Sproat and Fujimura, 1993; Lee-Kim et al., 2013), we selected the frame corresponding to the point at which the tongue body was most retracted for analysis. The most retracted frame was selected by visual inspection and consisted of the frame with the tongue furthest back in the leftmost third of the image. This frame was extracted and saved as a JPEG image. Using EdgeTrak (Li et al., 2005) this image was modeled and converted to a series of 50 x-y coordinates that allowed us to quantify the overall shape of the tongue in a 2D Cartesian space, with height represented on the y axis and retraction on the x axis.2 It was posited that final /l/'s would exhibit more retracted or lower tongue body positions compared to initial /l/'s. An example of this is found in the upper image of Fig. 3, where a participant's lingual shape during the production of a word-final /l/ is both lower and more retracted than its word-initial counterpart. This contrasts with the image below in which there is virtually no difference between them. Note that the back of the tongue is on the left side of the images.

FIG. 3.

Illustration of distinct tongue shapes (above) and non-distinct tongue shapes (below). In the figure on the left, word-final /l/ is both lower and more retracted than initial /l/.

FIG. 3.

Illustration of distinct tongue shapes (above) and non-distinct tongue shapes (below). In the figure on the left, word-final /l/ is both lower and more retracted than initial /l/.

Close modal

Since our overall goal was to determine how these gestures operate in the production of light and dark /l/ in the speech communities examined here, the x-y contour data, or splines (shown in Fig. 3) were then statistically examined to assess the degree of similarity in the lingual shapes in word-final and word-initial positions using the smoothing spline analysis of variance (SSANOVA) (Davidson, 2006) function in the gss package (Gu, 2014) in r (R Core Team, 2016). All items in the experiment were repeated twice and the two splines for each condition, word-final and word-initial, were averaged for each item to form a pairwise comparison. For example, both splines from the two productions of the word-final condition “teal eaves” were averaged as were the splines for “tea leaves,” the corresponding word-initial condition for that item. Splines were not averaged across items as differences between items in terms of adjacent vowel quality are expected to significantly affect tongue shape, resulting in non-comparable splines (e.g., word-final /l/ in “teal eaves” was not averaged with word-final /l/ in “steal assets”).

The SSANOVA function calculates 95% confidence intervals which are represented by the width of the splines as in Fig. 3. Visual inspection of the SSANOVA splines was carried out for each pairwise comparison in order to determine if the word-final /l/ exhibited either a more retracted tongue dorsum or a more lowered tongue body relative to the word-initial /l/. If the splines contained non-overlapping confidence intervals with the word-final spline being more retracted in the left-most third of the image or with the word-final spline being lower in the central third, the comparison was coded as distinct. If confidence intervals overlapped throughout the spline, the comparison was coded as indistinct. We took this approach in order to compare across speakers since variation in vocal tract sizes and “probe-cranium orientations” (Scobbie et al., 2012, p. 117) preclude any straightforward comparison of lingual shapes across the sample. In total, we looked at 249 pairwise comparisons of word-final and word-initial /l/ in phrases and compounds from speakers who imaged well and utterances that were also included in the acoustic analysis.

FIG. 4.

(Color online) Boxplots showing median zF2-zF1 values for all speakers in word-final and initial position in phrases and compounds.

FIG. 4.

(Color online) Boxplots showing median zF2-zF1 values for all speakers in word-final and initial position in phrases and compounds.

Close modal

Next, to examine the influence of age, gender, and dialect zone on the realization of lingual distinctions, we submitted the binary outcome, distinct or indistinct, of all 249 pairwise comparisons to a mixed-effects logistic regression test using the glmer function of the lme4 (Bates et al., 2015) package in r. The model included lingual distinction as described above (distinct or indistinct) as the dependent variable, region, gender, and age as fixed factors, and participant and item as random effects.

1. Word-initial versus word-final /l/

Taken as a group, our speakers show the standard North American pattern, with significantly darker /l/'s in word-final position than in word-initial position, in both compounds and phrases. Median zF2-zF1 values for the midpoint of /l/ across conditions in phrases and compounds are shown in Fig. 4. The mean zF2-zF1 value in final position in phrases is −1.62 with a standard deviation of 0.84 versus a mean of −0.44 and a standard deviation of 0.85 in initial position. In compounds, the mean in final position is −1.1 with a standard deviation of 0.91. The mean zF2-zF1 for compounds in initial position is 0.05 with a standard deviation of 1.05. Median zF2-zF1 values for the midpoint of /l/ across conditions in phrases are shown in Fig. 4.

There is substantial variation among speakers, however, with some speakers failing to show any difference in darkness between initial and final position. For example, the left side of Fig. 5 shows the box plot for speaker 48, a 46-yr-old man from the Southern Shore Line. This speaker has no distinction in zF2-zF1 values in initial and final position (Welch two sample t-test, p = 0.37). This contrasts with the pattern seen in the plot on the right-hand side. This is the plot for speaker 10, a 55-yr-old woman from our Northern Shore Line group. This speaker has a substantial difference in darkness values with no overlap between word-initial and word-final productions (Welch two sample test, p < 0.001).

FIG. 5.

(Color online) Boxplots showing median zF2-zF1 scores in initial and final position, speakers 48 and 10.

FIG. 5.

(Color online) Boxplots showing median zF2-zF1 scores in initial and final position, speakers 48 and 10.

Close modal

A look at the data broken down by age and gender suggests that social factors play a role in structuring the variation we find between individuals. Figure 6 shows data points by age and gender for word-final and word-initial /l/'s including data from both phrases and compounds. The trend lines for males and females show that men have lighter /l/'s than women, only in word-final position, that is, in the darkening environment in the standard North American allophonic pattern. This figure also suggests a greater distinction between final and initial /l/'s for older speakers than for younger speakers, with trend lines ascending, towards higher zF2-zF1 values, for older speakers in initial position but descending, toward lower zF2-zF1 values and, hence, darker /l/'s, in final position.

FIG. 6.

(Color online) zF2-zF1 values in initial and final position by age and gender, phrases, and compounds.

FIG. 6.

(Color online) zF2-zF1 values in initial and final position by age and gender, phrases, and compounds.

Close modal

In order to investigate the role of social factors in conditioning differences between speakers, we used the lme4 package in r (R Core Team, 2016) to perform a linear mixed effects analysis of the relationship between l-darkness and experimental condition, word-final versus word-initial. In addition to condition, we included gender, region, age (centered), and whether the word boundary occurs in a phrase or a compound as fixed effects. We also included interaction terms for all social factors with condition. The model included random intercepts for subjects and items. A summary of the results of the model is given in Table II. Reference level for condition is word-final. Reference level for structure type, phrase or compound, is compound. Reference levels for gender and region are female and Northern Shore Line.

TABLE II.

Results of linear mixed effects model with interaction terms for Condition and Sex, Condition and Region, and Condition and Age, Phrase vs Compound as a fixed effect and random intercepts for Item and Participant. Model: [darkness ∼ (condition gender) + (condition age) + (condition region) + phrase vs compound + (1|participant) + (1|item)].

Fixed effectsEstimateStandard errort value
(Intercept) −1.322594 0.175357 −7.542 
Condition, word-initial (e.g. see lapses) 1.519380 0.066288 22.921 
Sex, male 0.212311 0.082081 2.587 
Age −0.005316 0.002808 −1.893 
Region, St. John's 0.101415 0.094381 1.075 
Region, Southern Shore 0.294883 0.101424 2.907 
Phrase vs Compound, phrase −0.474233 0.241778 −1.961 
Sex: Condition, male: word-initial −0.397578 0.081188 −4.897 
Age: Condition, word-initial 0.017334 0.002782 6.230 
Region: Condition, St. John's, word-initial −0.143647 0.093790 −1.532 
Region: Condition, Southern Shore, initial −0.435336 0.100405 −4.336 
Fixed effectsEstimateStandard errort value
(Intercept) −1.322594 0.175357 −7.542 
Condition, word-initial (e.g. see lapses) 1.519380 0.066288 22.921 
Sex, male 0.212311 0.082081 2.587 
Age −0.005316 0.002808 −1.893 
Region, St. John's 0.101415 0.094381 1.075 
Region, Southern Shore 0.294883 0.101424 2.907 
Phrase vs Compound, phrase −0.474233 0.241778 −1.961 
Sex: Condition, male: word-initial −0.397578 0.081188 −4.897 
Age: Condition, word-initial 0.017334 0.002782 6.230 
Region: Condition, St. John's, word-initial −0.143647 0.093790 −1.532 
Region: Condition, Southern Shore, initial −0.435336 0.100405 −4.336 

As expected, the effect of condition shows /l/'s to be lighter in initial position relative to /l/'s in word-final position. There was no interaction of condition and structure type, phrase vs compound, but there was a main effect of structure type with phrases having darker word-final /l/'s (χ2 = 3.8963, p < 0.05). Likelihood ratio tests showed a significant effect of the interaction of condition and gender on darkness (χ2 = 23.909, p < 0.001) with men having lower normalized F2-F1 values, and hence darker /l/'s, in word-initial position relative to women. In conjunction with the sex simple effect, which shows men to have lighter /l/'s than women in final position, these results show that men have a smaller distinction between word-initial and word-final /l/. With respect to age, F2-F1 values lower with age in word-final position, indicating darker /l/'s. In initial position, we find the reverse, with older speakers having lighter word-initial /l/'s. Likelihood ratio tests showed this interaction effect of condition and age on darkness to be significant (χ2 = 38.495, p < 0.001). The interaction of region and condition is also significant (χ2 = 19.416, p < 0.001) with speakers from the Southern Shore Line having darker initial /l/'s and lighter final /l/'s relative to our default group of speakers from the Northern Shore Line and outside of the Avalon. Although speakers from St. John's also have darker initial and lighter final /l/'s than our default group, t values in Table II suggest that only the Southern Shore level of region differs significantly from our default group of Northern Shore speakers. The effects of age and region on /l/ darkness are illustrated in Fig. 7, which shows data points for all speakers by region and age.

FIG. 7.

(Color online) zF2-zF1 values in initial and final position by age and region, phrases, and compounds.

FIG. 7.

(Color online) zF2-zF1 values in initial and final position by age and region, phrases, and compounds.

Close modal

2. Discussion: Acoustic results

The acoustic results show a significant difference between /l/'s in initial and final position with word-final /l/'s being darker in both phrases and compounds. Patterns of individual speakers vary widely with some speakers failing to show an initial-final distinction. In the compound and phrase data, we find a significant interaction of gender and condition with men having lighter /l/'s in word-final position, the canonical darkening environment, and darker /l/'s in word-initial position, relative to women. Although men as a whole do show a distinction between word-initial and word-final /l/, this distinction is less than what we find among the female speakers in our study. This gender difference is consistent with sociolinguistic patterns in which men retain more local variants, such as light /l/ in final position, while women adopt more standard patterns. This gender effect has been observed both in general patterns of language change (e.g., Labov, 2001) and specifically with respect to traditional features of Newfoundland dialects (e.g., Van Herk et al., 2007).

With respect to region, we find that the heavily Irish-influenced Southern Shore Line speakers have lighter final /l/'s relative to the reference group of speakers from the Northern Shore and outside of the Avalon. The St. John's group also has higher zF2-zF1, hence lighter /l/'s, in word-final position relative to Northern shore speakers but this effect was not significant. Darkness values for the Southern Shore speakers differ more substantially from the Northern Shore group with Southern Shore speakers having higher zF2-zF1 values by 0.29 in word-final position. The effect of region shows that, although the standard pattern of allophonic variation is widespread in Newfoundland English, the dialectal variation arising from early settlement patterns continues to influence speech patterns.

Our findings with respect to gender and region are consistent with regional dialect patterns in Newfoundland more generally as well as with extensive literature on the role of gender in shaping patterns of language change (e.g., Eckert and McConnell-Ginet, 1999; Holmes and Meyerhoff, 2003). These findings also show that this dialectal variation can be studied, even through the highly formal method of elicitation in which speakers read stimuli while wearing a headset and, for most of our subjects, having an ultrasound probe secured under their chins. Although the most vernacular register may be missed with this methodology, our findings support previous work such as Lawson et al. (2008); Scobbie and Wrench (2003); and Turton (2014) that demonstrate that social variation can be successfully studied using laboratory methods.

In addition to gender and region, age also had a significant interaction with condition with increased age having a small lowering effect on /l/'s in final position and a small raising effect on /l/'s in initial position. This effect may be counter to expectations that older speakers are more likely to retain traditional patterns (e.g., Labov, 2001). However, Dubois and Horvath's (1999) study of language change in Cajun English demonstrates that stigmatized variants avoided by middle generations may be “recycled” by younger speakers. Previous work has shown that certain features associated with Newfoundland English have undergone such a periodic pattern in which features which have decreased in use among older speakers have been readopted by younger speakers (e.g., Van Herk and Childs, 2015). Our data suggests that light /l/ in final position may also be undergoing such a process in Newfoundland English.

1. Word-initial versus word-final /l/, ultrasound data

As outlined in Sec. IV, the articulatory analysis consisted of two steps. The first was to compare productions of /l/ across initial and final positions for each speaker. Once the comparisons were determined to be either similar or different using the SSANOVA, we submitted the frequency of distinctions by speaker to a logistic model in r. Based on these results, we first note that of all 249 initial and final comparisons, 78% exhibited significantly different lingual shapes, defined here as either lowered or retracted. Nine speakers showed a categorical distinction, with every pairwise comparison coded as distinct. These speakers were balanced across genders: 5 females and 4 males. They were also found throughout each dialect region: 4 from St. John's, 3 from the Northern Shore Line group, and 2 from the Southern Shore Line. The rest of the sample exhibits lowering or retraction in final position at variable rates (mean = 76%, standard deviation = 22) with the lowest rate of distinctions being 25%. This shows that the main identifiable articulatory gestures, tongue lowering and retraction, are variable processes for most of our speakers. We then want to determine if this variation is governed by any of the demographic factors of speakers in our data set (i.e., age, gender, dialect region).

By virtue of their Irish influence, we predicted lingually similar articulations (lower rates of distinction) in the speech of participants from the Southern Shore Line. In contrast, we expected Northern Shore Line speakers to distinguish between initial and final /l/ more consistently (higher rates of distinction), with St. John's reporting a mixture of articulation patterns. These predictions are not borne out by the data. Rather, as Table III shows, St. John's has the highest rate of retracted and lowered distinctions across positions whereas the Northern Shore Line and Southern Shore Line are relatively similar. These frequency results also show that males exhibited a higher rate of distinctions between initial and final articulations. Results of the binomial logistic regression analysis (see Table IV), however, did not show significant differences between genders, nor between the Northern Shore Line and Southern Shore Line. Nonetheless, two significant patterns were found. First, there is a significant interaction between gender and region with males from the Southern Shore Line being less likely to use a lowered or retracted tongue shape to distinguish between initial and final /l/ relative to males from our default group of speakers from the Northern Shore and outside of the Avalon. Second, we found a small but significant effect of age with older speakers being more likely to exhibit distinctions in tongue gestures across initial and final position.

TABLE III.

Frequency of significantly different tongue contours (i.e., either retracted or lowered) across initial and final positions by Region and Gender. Distinctions determined by visual inspection of SS-ANOVA plots. Standard deviation of individual frequency rates in parentheses. Number of participants in brackets.

Dialect zone
GenderSt. John'sNorthern Shore LineSouthern Shore LineOverall
Female 71% (30) [4] 70% (23) [9] 87% (23) [3] 74% [16] 
Male 91% (8) [7] 80% (15) [3] 53% (19) [2] 82% [12] 
Total 84% [11] 73% [12] 74% [5] 78% [28] 
Dialect zone
GenderSt. John'sNorthern Shore LineSouthern Shore LineOverall
Female 71% (30) [4] 70% (23) [9] 87% (23) [3] 74% [16] 
Male 91% (8) [7] 80% (15) [3] 53% (19) [2] 82% [12] 
Total 84% [11] 73% [12] 74% [5] 78% [28] 
TABLE IV.

Results of binomial logistic regression.

EstimateStandard errorz valuep
(Intercept) −0.55396 0.65678 −0.843 0.399 
Region     
Southern Shore Line 1.28108 0.71611 1.789 0.0736 
St. John's −0.07603 0.56323 −0.135 0.8926 
Sex     
Male 0. 59193 0. 67967 0. 871 0.3838 
Age 0.04319 0.01801 2.398 0.0165 
Gender: Region     
Male, Southern Shore Line −2.80998 1.11566 −2.519 0.0118 
Male, St. John's 1.02407 0.95174 1.076 0.2819 
EstimateStandard errorz valuep
(Intercept) −0.55396 0.65678 −0.843 0.399 
Region     
Southern Shore Line 1.28108 0.71611 1.789 0.0736 
St. John's −0.07603 0.56323 −0.135 0.8926 
Sex     
Male 0. 59193 0. 67967 0. 871 0.3838 
Age 0.04319 0.01801 2.398 0.0165 
Gender: Region     
Male, Southern Shore Line −2.80998 1.11566 −2.519 0.0118 
Male, St. John's 1.02407 0.95174 1.076 0.2819 

2. Discussion: Articulatory results

Like the acoustic patterns reported in Sec. V A, the articulatory analysis found significant differences between /l/'s in initial and final position. The main articulatory variable investigated here, lingual distinction, was found at variable rates both across and within speakers. Some speakers showed statistically significant differences between all initial and final comparisons though no speaker was categorically similar. This indicates that all speakers articulated a distinction in /l/-darkness between initial and final position at least in some instances. These results also suggest that the local, Irish-based light /l/ in final position is produced variably and, despite characterizations of the dialect, no speaker in this study employs the traditional form all the time.

It is not clear when the light /l/ pattern became variable for these speakers but our analyses found a significant effect due to age whereby older speakers were slightly more likely to have initial and final differences. This means that the younger members of our sample appear to use the traditional pattern more often than the older speakers. The other significant finding of the logistic regression analysis concerned an interaction between gender and region: males from the Southern Shore Line communities were less likely to distinguish initial /l/ from final /l/.

The fact that articulatory distinctions are more frequent with older speakers suggests that the traditional Irish pattern of light /l/ in final position is undergoing recycling (e.g., Dubois and Horvath, 1999) in which a salient regional feature is rejected by one generation of speakers and readopted by a later generation. The interaction effect of region and gender shows a lower rate of distinctions in the production of initial and final /l/ among men from the Southern Shore line relative to elsewhere. The Southern Shore is home to a larger concentration of speakers of Irish heritage and we interpret the interaction of region and gender as a gender-based continuation of traditional forms among Southern Shore speakers. If both older men from Irish-settled areas and younger speakers more generally are the highest users of the traditional form, then we might expect a division in the social meaning of /l/. Older speakers are retaining a traditional pattern, and a traditional gendered difference in the realization of that pattern, while younger ones, after a generational lag, are linguistically marking their Newfoundland identity. Van Herk and Childs (2015) and Childs and Van Herk (2014) have argued for just such a pattern for other salient features of Newfoundland English. They show that distinctive local features, such as verbal s-marking (e.g., I loves it), follow a gendered pattern for older rural speakers, with men using higher rates of the traditional feature relative to women, whereas younger speakers use these features as positive markers of local identity which are, in fact, most frequent among younger, urban, females.

While we present these results as a positive contribution involving the use of experimental methodologies in the study of dialect variation, it must be noted that we also found substantial differences between the articulatory and acoustic datasets. We turn now to discuss the overall results and address issues in relating these domains under a unified view of /l/-allophony in Newfoundland English.

This paper examined both acoustic and articulatory measures of /l/ in data collected from a variety of Newfoundland English speakers. The patterning of /l/-allophony was previously reported to be involved in a change in progress in Newfoundland, with the allophonic pattern displacing the Irish pattern of light /l/ in all positions (Clarke, 2012). However, with the exception of recent work on St. John's English (Pierson, 2016), evidence for this change has not been examined using instrumental procedures.

One of the questions this study was designed to answer is whether the allophonic pattern associated with both mainland Canada and English-origin varieties of Newfoundland English is displacing the traditional dialect feature of light /l/ in final position, particularly associated with Irish-settled regions of the province. This question has a complex answer. First, both acoustic and articulatory analyses of /l/-darkness reveal that, although differences between word-initial and final /l/ are quantitatively robust when our speakers are considered as a group, the distinction is not present for all individuals. Acoustic measures of darkness are affected by region with smaller differences between word-initial and word-final /l/ in the Irish-settled Southern Shore Line than among our Northern Shore speakers. This pattern is supported in the articulatory data with respect to the interaction of region and gender and the lower rates of articulatory distinctions exhibited by male speakers from the Southern Shore. Therefore, the reported spread of mainland Canadian features has not resulted in dialect leveling, with the standard allophonic pattern adopted by all speakers. Linguistic variation with roots in early settlement patterns persists in contemporary speech. Furthermore, overall differences between word-final and word-initial /l/ are shaped by social factors of age and, in the acoustic domain, gender. Men have a smaller distinction in zF2-zF1 between initial and final /l/ than women, suggesting a less standard pattern. The light /l/ pattern is associated with working-class speech and the gender distinction found in our acoustic data is consistent with other attested patterns of sociolinguistic variation in which men retain traditional dialect variants at greater rates than women (Labov, 2001). The effect of age is less expected, with older speakers showing somewhat greater acoustic and articulatory distinctions between word-final and initial /l/. Childs and Van Herk (2014) and Van Herk and Childs (2015) show that some salient and stigmatized features of Newfoundland English have undergone a pattern of recycling in which features previously shown to be in decline are readopted by younger speakers. These authors attribute this pattern to an increase in local pride and affiliation. Our findings with respect to /l/ suggest that the pattern of word-final light /l/ may also be participating in such a change.

While some social effects are observed in our ultrasound analysis of differences in lingual shape across word-initial and final positions, the articulatory results raise some points for further discussion. We looked at two distinctions in lingual shape as signifying distinctions between dark and light /l/: lowering and retraction of the tongue body. Based on this measure, speakers showed initial-final distinctions at variable rates. This particular articulatory measure shows a small but significant effect of age as well as an interaction effect of region and gender. However, we found no main effect of gender or region. This is particularly notable given the significant effects of gender and region in our acoustic results.

One obvious possible explanation for these differences is that the acoustic and articulatory data were collected from different participants. Forty-four participants are included in the acoustic data and a subset of 28 are included in the articulatory data. In order to explore this possibility, we ran the mixed effects models used in our acoustic analysis on data from the subset of speakers included in the articulatory data. Although the magnitude of effects differed from that found among the whole set of 44 speakers, the directionality and significance of effects was unchanged. Even when only the participants included in the articulatory data are considered, Southern Shore speakers show a smaller difference between initial and final /l/ than speakers from the Northern Shore and men show a smaller difference than women. The interaction of region and condition is significant (χ2 = 37.567, p < 0.001), as is the interaction of gender and condition (χ2 = 39.123, p < 0.001) and the interaction of age and condition (χ2 = 36.221, p < 0.001). These results are summarized in Table V.

TABLE V.

Results of linear mixed effects model, phrases and compounds, 28 speakers included in articulatory analysis, with interaction terms for Condition and Sex, Condition and Region, and Condition and Age, and random intercepts for Item and Participant. Model: [darkness ∼ (condition gender) + (condition age) + (condition region) + (1|participant) + (1|item)].

Fixed effectsEstimateStandard errort value
(Intercept) −1.514482 0.155289 −9.753 
Condition, word-initial (e.g. see lapses) 1.558316 0.073967 21.068 
Sex, male 0.318490 0.103399 3.080 
Age −0.005008 0.003963 −1.264 
Region, St. John's 0.012862 0.113643 0.113 
Region, Southern Shore 0.422356 0.135610 3.115 
Sex: Condition, male: word-initial −0.615935 0.097856 −6.294 
Age: Condition, word-initial 0.022903 0.003782 6.056 
Region: Condition, St. John's, word-initial −0.188932 0.108309 −1.744 
Region: Condition, Southern Shore, word-initial −0.772459 0.126454 −6.109 
Fixed effectsEstimateStandard errort value
(Intercept) −1.514482 0.155289 −9.753 
Condition, word-initial (e.g. see lapses) 1.558316 0.073967 21.068 
Sex, male 0.318490 0.103399 3.080 
Age −0.005008 0.003963 −1.264 
Region, St. John's 0.012862 0.113643 0.113 
Region, Southern Shore 0.422356 0.135610 3.115 
Sex: Condition, male: word-initial −0.615935 0.097856 −6.294 
Age: Condition, word-initial 0.022903 0.003782 6.056 
Region: Condition, St. John's, word-initial −0.188932 0.108309 −1.744 
Region: Condition, Southern Shore, word-initial −0.772459 0.126454 −6.109 

Another possible explanation could follow from the different types of measures used in the acoustic and articulatory data. In the acoustic data, formant values are averaged across items and groups of speakers. In the articulatory data, differences in the size and shape of individual vocal tracts preclude averaging data across speakers. Instead, splines for final and initial /l/ were compared for each item individually for each speaker. In order to examine individual behavior in the acoustic data for those speakers included in the articulatory study, we performed t-tests on initial versus final acoustic darkness measures for each item for each speaker. Items with two repetitions of each condition were compared using a paired-samples t-test. Where an acoustic measurement did not have a match in initial or final position, no test was carried out. The frequency of acoustic distinctions for each speaker was then tallied and compared to the frequency of distinct articulations, as determined by the SSANOVA. These frequency comparisons are found in Table VI. This comparison, which matched items that had both articulatory and acoustic measurements, included 198 comparisons. Overall, lingual distinctions were found at a rate of 76% whereas only 23% exhibited statistically significant acoustic differences. A Pearson Correlation test was then used to compare speakers' rates of articulatory distinctions as determined by comparison of SSANOVAs (mean = 76) against individual speaker distinctions in the acoustic data as determined by the paired t-tests (mean = 16.68). The test did not show a significant correlation between the frequency of initial and final distinctions as determined by these articulatory and acoustic measures (r = − 0.17, p = 0.387).

TABLE VI.

Frequencies of significant distinctions by speaker based on SSANOVA (articulatory data) and paired samples t-tests (acoustic data). N is the number of pairwise comparisons included in both the articulatory and acoustic comparisons.

SpeakerN =% of articulatory distinctions% acoustic distinctions
50 17 
10 90 10 
25 25 
43 29 
10 50 75 
11 40 20 
12 57 14 
16 63 25 
17 86 14 
18 25 
20 100 
21 10 100 60 
23 88 
24 67 
25 10 90 30 
26 100 14 
28 100 13 
29 100 
31 100 
32 100 
33 71 
34 10 90 
36 100 25 
37 86 
43 75 
46 100 50 
47 88 13 
48 44 33 
SpeakerN =% of articulatory distinctions% acoustic distinctions
50 17 
10 90 10 
25 25 
43 29 
10 50 75 
11 40 20 
12 57 14 
16 63 25 
17 86 14 
18 25 
20 100 
21 10 100 60 
23 88 
24 67 
25 10 90 30 
26 100 14 
28 100 13 
29 100 
31 100 
32 100 
33 71 
34 10 90 
36 100 25 
37 86 
43 75 
46 100 50 
47 88 13 
48 44 33 

An additional possible explanation for the articulatory-acoustic mismatch could be found in the timing of the measurements. As explained above, our tongue contour measurement was based on the point in the production of /l/ when retraction was at its maximum. On the other hand, the acoustic measure of darkness was determined at the vowel midpoint. The potential temporal mismatch, then, raises the possibility that speakers may realize a dark/light distinction at different points in the articulation of the /l/, contributing to differences in the articulatory and acoustic measures.

Differences between articulatory and acoustic measures of /l/-darkness, however, are not limited to this study. Turton (2014) reports on acoustic and ultrasound measures of /l/'s elicited in a wide range of phonological contexts from speakers of a variety of dialects. For one speaker, Turton found significant distinctions in acoustic measures of /l/-darkness which are “in no way reflected in the average tongue splines” (Turton, 2014, p. 120) whereas, for other speakers, she found distinctions in the articulatory data “which the acoustics miss altogether” (Turton, 2014, p. 218). Consistent with these reports, other work has suggested that there is a non-linear (indirect) relationship between underlying lingual articulations and acoustics measurements of F2. Ying et al. (2012, p. 112) provide evidence that challenges the direct relationship between F2 lowering and tongue body retraction, positing that lowering of F2 found in dark /l/ may be “attributed more directly to some aspect of tongue lateralization.” In their study of EPG and MRI data drawn from intentionally produced light and dark /l/ by four phonetically trained speakers of American English, Narayanan et al. also suggest that acoustic differences between light and dark /l/ likely result from both “back cavity and lateral channel effects” (Narayanan et al., 1997, p. 1074).

The study of Narayanan et al. also showed significant articulatory variation between individuals. For two speakers, light /l/'s “exhibited significantly more front-region contacts” in the EPG data (Narayanan et al., 1997, p. 1072) while the other two speakers showed comparable contact across /l/ types. Significant individual variation is also found in the production of /r/, which, like /l/ involves coordination of multiple articulatory gestures (e.g., Westbury et al., 1998). Differences in articulation of /r/ are identified with significant social (Lawson et al., 2011) and dialectal variation (Delattre and Freeman, 1968).

Based on these observations, we take the variable rates of articulatory distinctions between initial and final /l/ in our data to reflect both variation in the allophonic patterning of dark and light /l/ across speakers, as well as variation in how the dark and light distinction is articulated. Some aspects of these articulatory distinctions are captured by the ultrasound data analyzed here whereas other potential articulatory distinctions, such as degree of lateralization, are not. Such variation may nonetheless affect formant structure and contribute to the mismatch between the acoustic and articulatory results. These findings speak to the value of research such as Mielke (2015), in which a variety of articulatory measures are taken and exploratory analysis is used to investigate which articulatory measures correlate with acoustic distinctions, in the case of Mielke (2015) formant values of rhotic and non-rhotic vowels. With respect to dark and light /l/, to our knowledge, no studies have systematically examined the relationship between lateralization and darkening which leaves open an area for further research.

We therefore conclude, positively and cautiously. Concerning the former, this study has provided further confirmation that sociolinguistic variation can be elicited using instrumental techniques and suggests that understanding of the inherent variability of sociophonetic variation can be obtained through studying acoustic and articulatory data. However, we also recognize that for this endeavour to provide novel insights into the study of linguistic variation and change will require further examination of the relationship between the underlying articulatory strategies and their overt acoustic properties.

This research was supported by the Social Sciences and Humanities Research Council of Canada Grant #430-2013-1001 “Allophony in Newfoundland English: Production, perception, and variation.” We also thank the anonymous reviewers as well as Michael Wagner, Meghan Clayards, Lisa Davidson, and Rosanna Pierson for input and contributions to this manuscript.

List of stimuli, all items from Mackenzie et al. (2014).

Word-final vs word-initial, phrases:

  • see lapses – steal assets,

  • woo lasses – fool asses,

  • grey land – frail ant,

  • free label – feel able,

  • pay labs – pale abs.

Word-final vs word-initial, compounds:

  • hay lifter – hail impact,

  • day louse – mail out,

  • jay loaner – jail owner,

  • toy lamp – coil amp,

  • tea leaf – teal eave,

  • tow limit – goal inning.

1

We originally elicited data from six items comparing word-initial and final /l/'s in phrases. The word-final condition of one item failed to be presented to all participants and both conditions of this item were excluded from the analysis.

2

Recent work (Mielke, 2015; Heyne and Derrick, 2015) has argued for converting spline data to polar coordinates. This method has advantages, particularly with respect to comparisons of tongue tip and root configurations. However, the diagnostics of /l/-darkness rely largely on tongue dorsum and tongue body shape and use of x-y coordinates in Cartesian space has been successfully used to compare tongue gestures of /l/ across conditions in previous work (e.g., Lee-Kim et al., 2013). We plan to adopt use of polar coordinates in future work.

1.
Atkinson
,
M.
(
1982
). “
A preliminary report on a study of the acoustic effects of variants of /l/ on preceding vowels
,” in
Papers from the 6th Annual Meeting of the Atlantic Provinces Association
, pp.
1
8
.
2.
Bartlett
,
J.
(
1977
). “
A spectrographic study of post vocalic /l/ and its allophones in Newfoundland English
,” M.Phil. thesis,
Memorial University of Newfoundland
,
St. John's, Newfoundland, Canada
.
3.
Bates
,
D.
,
Maechler
,
M.
,
Bolker
,
B.
, and
Walker
,
S.
(
2015
). “
Fitting linear mixed-effects models using lme4
,”
J. Stat. Softw.
67
(
1
),
1
48
.
4.
Boersma
,
P.
, and
Weenink
,
D.
(
2014
). “
Praat: Doing Phonetics by Computer
,” version 5.3.32, www.praat.org (Last viewed March 1, 2016).
5.
Childs
,
B.
, and
Van Herk
,
G.
(
2014
). “
Work that –s!: Drag queens, gender, identity, and traditional Newfoundland English
,”
J. Socioling.
18
(
5
),
634
657
.
6.
Clarke
,
S.
(
1984
). “
Sociolinguistic variation in a small urban context: The St. John's survey
,” in
Papers from the Fifth International Conference on Methods in Dialectology
, edited by
H. J.
Warkentyne
,
Victoria
,
British Columbia
, pp.
143
153
.
7.
Clarke
,
S.
(
2010
).
Newfoundland and Labrador English
(
Edinburgh University Press
,
Edinburgh
), pp.
1
212
.
8.
Clarke
,
S.
(
2012
). “
Phonetic change in Newfoundland English
,”
World Englishes
31
(
4
),
503
518
.
9.
Clarke
,
S.
, and
Hiscock
,
P.
(
2012
). “
The Dialect Atlas of Newfoundland and Labrador
,” http://www.dialectatlas.mun.ca (Last viewed July 7, 2016).
10.
Clarke
,
S.
,
Paddock
,
H.
, and
MacKenzie
,
M.
(
1999
). “
Language
,” Newfoundland and Labrador Heritage Web Site, http://www.heritage.nf.ca/articles/society/language.php (Last viewed July 7, 2016).
11.
Davidson
,
L.
(
2006
). “
Comparing tongue shapes from ultrasound imaging using smoothing spline analysis of variance
,”
J. Acoust. Soc. Am.
120
,
407
415
.
12.
Delattre
,
P.
, and
Freeman
,
D. C.
(
1968
). “
A dialect study of American R's by x-ray motion picture
,”
Linguistics
6
(
44
),
29
68
.
13.
Dubois
,
S.
, and
Horvath
,
B.
(
1999
). “
When the music changes, you change too: Gender and language change in Cajun English
,”
Lang. Var. Change
11
(
3
),
287
313
.
14.
Eckert
,
P.
, and
McConnell-Ginet
,
S.
(
1999
). “
New generalizations and explanations in language and gender research
,”
Lang. Soc.
28
(
2
),
185
201
.
15.
Gorman
,
K.
,
Howell
,
J.
, and
Wagner
,
M.
(
2011
). “
Prosodylab-Aligner: A tool for forced alignment of laboratory speech
,” in
Proceedings of Acoustics Week in Canada'11
, Vol.
39
(
3
),
192
193
.
16.
Government of Newfoundland and Labrador
(
1970
). “
Historical statistics of Newfoundland and Labrador
,” Vol.
I
, 1970, http://www.stats.gov.nl.ca/Publications/Historical/PDF/Full.pdf (Last viewed July 7, 2016).
17.
Gu
,
C.
(
2014
). “
Smoothing spline ANOVA models: R package gss
,”
J. Stat. Softw.
58
(
5
),
1
25
.
18.
Halle
,
M.
, and
Mohanan
,
K. P.
(
1985
). “
Segmental phonology of modern English
,”
Ling. Inq.
16
(
1
),
57
116
.
19.
Heyne
,
M.
, and
Derrick
,
D.
(
2015
). “
Using a radial ultrasound probe's virtual origin to compute midsagittal smoothing splines in polar coordinate
,”
J. Acoust. Soc. Am.
138
(
6
),
EL509
EL514
.
20.
Holmes
,
J.
, and
Meyerhoff
,
M.
(
2003
).
The Handbook of Language and Gender
(
Blackwell
,
Oxford
), pp.
1
759
.
21.
Huffman
,
M.
(
1997
). “
Phonetic variation in intervocalic onset /l/'s in English
,”
J. Phon.
25
(
2
),
115
141
.
22.
Hughes
,
A.
,
Trudgill
,
P.
, and
Watt
,
D.
(
2005
).
English Accents and Dialects: An Introduction to Social and Regional Varieties of English in the British Isles
, 4th ed. (
Hodder Arnold
,
New York
), pp.
1
224
.
23.
Jones
,
D.
(
1956
).
The Pronunciation of English
(
Cambridge University Press
,
Cambridge
), pp.
1
223
.
24.
Kendall
,
T.
, and
Thomas
,
E.
(
2009
). “
Vowels: Vowel manipulation, normalization, and plotting in R
,” R package, version 1.0-3.
25.
Kirkham
,
S.
, and
Wormald
,
J.
(
2015
). “
Acoustic and articulatory variation in British Asian English liquids
,” in
Proceedings of the 18th International Congress of Phonetic Sciences
, The Scottish Consortium for ICPhS 2015, University of Glasgow, Glasgow, United Kingdom, pp.
1
5
.
26.
Labov
,
W.
(
2001
).
Social Factors, Vol. 2 of Principles of Linguistic Change
(
Blackwell
,
Cambridge, MA
), pp.
1
592
.
27.
Lawson
,
E.
,
Scobbie
,
J.
, and
Stuart-Smith
,
J.
(
2011
). “
The social stratification of tongue shape for postvocalic /r/ in Scottish English
,”
J. Socioling.
15
(
2
),
256
268
.
28.
Lawson
,
E.
,
Scobbie
,
J.
, and
Stuart-Smith
,
J.
(
2013
). “
Bunched /r/ promotes vowel merger to schwar: An ultrasound tongue-imaging study of Scottish sociophonetic variation
,”
J. Phon.
41
(
3-4
),
198
210
.
29.
Lawson
,
E.
,
Stuart-Smith
,
J.
, and
Scobbie
,
J.
(
2008
). “
Articulatory insights into language variation and change: Preliminary findings from an ultrasound study of derhoticization in Scottish English
,”
Univ. Penn. Work. Pap. Ling.
14
(
2
),
102
110
, available at http://repository.upenn.edu/pwpl/vol14/iss2/13.
30.
Lee-Kim
,
S.-I.
,
Davidson
,
L.
, and
Hwang
,
S.
(
2013
). “
Morphological effects on the articulation of English intervocalic /l/
,”
Lab. Phonol.
4
(
2
),
475
511
.
31.
Lehiste
,
I.
(
1964
). “
Acoustical characteristics of selected English consonants
,”
Indiana Res. Center Anthropol. Folklore Ling.
34
,
10
50
.
32.
Li
,
M.
,
Kambhamettu
,
C.
, and
Stone
,
M.
(
2005
). “
Automatic contour tracking in ultrasound images
,”
Clin. Ling. Phon.
19
(
6
),
545
554
.
33.
Lobanov
,
B.
(
1971
). “
Classification of Russian vowels spoken by different listeners
,”
J. Acoust. Soc. Am.
49
,
606
608
.
34.
Mackenzie
,
S.
,
Olson
,
E.
,
Clayards
,
M.
, and
Wagner
,
M.
(
2014
). “
Allophonic variation in English /l/: Perception, production, and segmentation
,” presented at the
Annual Meeting of the Canadian Linguistics Association
, Brock University, St. Catharines, Ontario, Canada.
35.
Mielke
,
J.
(
2015
). “
An ultrasound study of Canadian French rhotic vowels with polar smoothing spline comparison
,”
J. Acoust. Soc. Am.
137
,
2858
2869
.
36.
Morris
,
J.
(
2013
). “
Sociolinguistic variation and regional minority language bilingualism: An investigation of Welsh-English bilinguals in North Wales
,” Ph.D. thesis,
University of Manchester
,
Manchester, United Kingdom
.
37.
Narayanan
,
S. S.
,
Alwan
,
A. A.
, and
Haker
,
K.
(
1997
). “
Toward articulatory-acoustic models for liquid approximants based on MRI and EPG data. Part I. The laterals
,”
J. Acoust. Soc. Am.
101
(
2
),
1064
1077
.
38.
Paddock
,
H. J.
, and
Memorial University of Newfoundland, Department of Linguistics.
(
1982
).
Languages in Newfoundland and Labrador
(
Department of Linguistics, Memorial University of Newfoundland, St. John's
,
Newfoundland, Canada
).
39.
Pierson
,
R.
(
2016
). “
The current state of /l/ allophony in St. John's English
,” Master's thesis,
Memorial University of Newfoundland, St. John's
,
Newfoundland, Canada
.
40.
R Core Team (
2016
). R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria, https://www.R-project.org/ (Last viewed July 7, 2016).
41.
Recasens
,
D.
(
2012
). “
A cross-language acoustic study of initial and final allophones of /l/
,”
Speech Commun.
54
,
368
383
.
42.
Scobbie
,
J. M.
,
Lawson
,
E.
, and
Stuart-Smith
,
J.
(
2012
). “
Back to front: A socially-stratified ultrasound tongue imaging study of Scottish English u
,”
Riv. Ling.
24
(
1
),
103
148
.
43.
Scobbie
,
J. M.
, and
Pouplier
,
M.
(
2010
). “
The role of syllable structure in external sandhi: An EPG study of vocalisation and retraction in word-final English /l/
,”
J. Phon.
38
,
240
259
.
44.
Scobbie
,
J.
, and
Wrench
,
A.
(
2003
). “
An articulatory investigation of word final /l/ and /l/-sandhi in three dialects of English
,” in
Proceedings of the 15th International Congress of the Phonetic Sciences
, pp.
1871
1874
.
45.
Seary
,
E. R.
,
Story
,
G. M.
, and
Kirwin
,
W. J.
(
1968
). “
The Avalon Peninsula of Newfoundland: An ethno-linguistic study
,” Queen's Printer, pp.
1
115
.
46.
Shorthouse
,
D. P.
(
2010
). “
SimpleMappr, an online tool to produce publication-quality point maps
,” http://www.simplemappr.net (Last viewed July 7, 2016).
47.
Sproat
,
R.
, and
Fujimura
,
O.
(
1993
). “
Allophonic variation in English /l/ and its implications for phonetic implementation
,”
J. Phon.
21
,
291
311
.
48.
Statistics Canada, Newfoundland and Labrador (
2013
). “
National household survey (NHS) profile
,” 2011 National Household Survey, Statistics Canada Catalogue no. 99-004-XWE, Ottawa, http://www12.statcan.gc.ca/nhs-enm/2011/dp-pd/prof/index.cfm?Lang=E (Last viewed July 7, 2016).
49.
Statistics Canada (
2016
). “
Newfoundland population residing on the Avalon Peninsula
,” Demography Division (1 June 2016) http://www.stats.gov.nl.ca/statistics/population/PDF/Population_Estimates_CDCMA.pdf (Last viewed June 1, 2016).
50.
Turton
,
D.
(
2014
). “
Variation in English /l/: Synchronic reflections of the life cycle of phonological processes
,” Ph.D. thesis,
University of Manchester, Manchester
,
United Kingdom
.
51.
Van Herk
,
G.
, and
Childs
,
B.
(
2015
). “
Active retirees: The persistence of obsolescent features
,” in
Linguistic Variation: Confronting Fact and Theory
, edited by
N.
Dion
,
A.
Lapierre
, and
R.
Torres Cacoullos
(
Routledge
,
New York)
, pp.
191
207
.
52.
Van Herk
,
G.
,
Childs
,
B.
, and
Thorburn
,
J.
(
2007
). “
Identity marking and affiliation in an urbanizing Newfoundland community
,” in
Papers from the 31st Annual Meeting of the Atlantic Provinces Linguistic Association
, pp.
85
94
.
53.
van Hofwegen
,
J.
(
2009
). “
Cross-generational change in /l/ in Chicano English
,”
English World-Wide
30
(
3
),
302
325
.
54.
van Hofwegen
,
J.
(
2010
). “
Apparent-time evolution of /l/ in one African American community
,”
Lang. Var. Change
22
(
3
),
373
396
.
55.
Wells
,
J. C.
(
1982
).
Accents of English: The British Isles
(
Cambridge University Press
,
Cambridge
), pp.
279
466
.
56.
Westbury
,
J. R.
,
Hashi
,
M.
, and
Lindstrom
,
M. J.
(
1998
). “
Differences among speakers in lingual articulation for American English /
ɹ
/
,”
Speech Commun.
26
(
3
),
203
226
.
57.
Ying
,
J.
,
Shaw
,
J.
,
Kroos
,
C.
, and
Best
,
C. T.
(
2012
). “
Relations between acoustic and articulatory measurements of /l/
,” in
Proceedings of the 14th Australasian International Conference on Speech Science and Technology
,
3–6 December
,
Macquarie University, Sydney, New South Wales, Australia
, p.
109
.
58.
Yuan
,
J.
, and
Liberman
,
M.
(
2011
). “
/l/ variation in American English: A corpus approach
,”
J. Speech Sci.
1
,
35
46
.