phenomenon in a rural region of northwestern Vermont in which the dialect is ... Vermont. Glottalization, as will be discussed below, has been widely reported.
AS OLD BECOMES NEW: GLOTTALIZATION IN VERMONT JULIE ROBERTS University of Vermont
abstract: Glottalization is a well-researched variable most often noted in Great Britain and Ireland but rarely in North America. The current study examines this phenomenon in a rural region of northwestern Vermont in which the dialect is popularly thought to be dying out. Forty-seven Vermonters, aged 3–80 years, were interviewed and recorded. Glottalized tokens of /t/ were coded perceptually and classified according to position in word and phonological environment. Social factors comprised age and sex. Results revealed that, unlike some vowel features historically associated with the region, glottalization appears to be a robust feature of Vermont speech. Younger speakers, particularly the adolescents, showed the highest rate of glottalization. However, differences in the patterning of the feature in different age groups begs the question of whether the glottalization in the speech of the younger speakers is, in fact, the same feature as that heard in the oldest speakers in the study or a new, less local variant. Finally, the importance of combining accounts of linguistic behavior to allow for better interpretation of data is discussed. In this study, phonetic, historical, and sociocultural information were all utilized in the interpretation of the findings, as none presented a convincing interpretation alone.
The purpose of the current study is to explore two very different issues
through the examination of one phonological dialect feature, glottalization of /t/, as in Milton [mIl?@n] or put off [pU? Af]. The first of these issues is the existence and possible progression of glottalization in northwestern Vermont. Glottalization, as will be discussed below, has been widely reported in English dialects outside of North America but rarely within the continent. Second, glottalization in northwestern Vermont will be used to examine the obsolescence of the area’s rural dialect; it is an apparently robust feature in an otherwise leveling speech variety. Northern Vermont is a subsection of Western New England. It has traditionally been a farming region. The largest town in the area is Swanton, with a population of 6,203 (U.S. Census Bureau 2000). Until relatively recently, the region was fairly isolated from outside influence. Within the last few decades, however, improvements in highways as well as the emergence of large agricultural businesses, often replacing family farms, have led to increasing immigration into the area. Although its population is still not as mixed as Burlington, 40 miles to the south, Swanton is more diverse than American Speech, Vol. 81, No. 3, Fall 2006 doi 10.1215/00031283-2006-016 Copyright 2006 by the American Dialect Society
227
228
american speech 81.3 (2006)
in the past (Meeks 1986). Consequently, the local dialect can be seen as potentially endangered, as has been noted in press reports for Vermont in general (Zind 2002; MacQuarrie 2004; O’Connor 2004).
BACKGROUND Glottalization is a variant of /t/ in which oral closure is either replaced by [?], called glottaling or glottal replacement, or in which the [t] and [?] are produced more or less simultaneously, sometimes labeled glottal reinforcement or preglottalization. Glottalization has been widely reported in Great Britain, but there is virtually no research on this feature in North America. The lack of relevant research, however, does not necessarily mean that no forms of glottalization exist. Rather, glottal reinforcement of /t/ appears to be so common as to go unnoticed, particularly as most research on phonological variation focuses on vowels rather than consonants. In general American English, /t/ is often reinforced by /?/ in word-final and word-medial positions following vowels and liquids. These productions were frequently noted in the Linguistic Atlas of New England (LANE; Kurath 1939b), for example. Glottaling, however, is not known to be a feature of American English but is the primary focus of this article. Glottalization was first reported in western Scotland in the 1860s but soon after was noted in many areas of Great Britain (Andrésen 1968; cited in Milroy et al. 1994). Milroy, Milroy, and Hartley (1994) note that because glottalization has been found in Ulster Scots, it may have been present in Scotland before the Ulster plantations brought Scottish immigrants to Ireland in the early 1600s. The possibility of this early appearance of glottalization in Ulster Scots is relevant to the present study because Vermont was the location of a secondary settlement of the Ulster Scots population in New England in the 1700s. Although the roots of this feature in Vermont cannot be traced with any certainty, the Ulster Scots and English settlements may have influenced speech in the area. Glottalization has become a very useful and complex variable in the study of British English. Earlier studies focused on its social constraints and prestige level. The earliest studies, Trudgill (1974) in Norwich, Reid (1978) and Romaine (1978) in Edinburgh, and Macaulay (1977) in Glasgow, revealed that glottalization is associated with low prestige. This evaluation is supported by social class and stylistic differentiation; gender differentiation was found to be less important in these early studies. More recent work, however, has demonstrated that the glottal stop is even more versatile than previously shown. Explorations of this variable have been
As Old Becomes New: Glottalization in Ver mont
229
used to build theory in the areas of language change, dialect leveling, social networking, language acquisition, and language and gender. For example, in her work on language change in Cardiff, Wales, Mees (1987) notes that it is important to study glottal replacement separately from glottal reinforcement in order to understand their social evaluation. When so studied, she found both gender and social class differences in the evaluation of the features. Milroy et al. (1994) note that glottalization in Tyneside seems to be both stigmatized and spreading, with younger female speakers spearheading the change. They demonstrate that these seemingly conflicting findings can indicate a change in prevailing assumptions about language and gender. As female speakers appear to lead the change toward glottaling, yet reject it in more careful styles, the usual explanation of women gravitating toward a prestige norm does not hold. Milroy et al. suggest that rather than following the prestige pattern (in this case, exemplified by male speakers), the females confer prestige on this feature. Kingsmore’s (1995) detailed study of glottalization among the Ulster Scots in Coleraine, Northern Ireland, is particularly relevant to the current work. Her findings both complement and contrast with other work on glottalization in the United Kingdom in which glottalization is not an innovative urban phenomenon among the Ulster Scots, as it was first reported to have been in Scotland (Wells 1982), but rather the conservative variant, associated with rural speech. In contrast, the [Q] variant, or tap, is the advancing form, found by Kingsmore to be led by working-class male speakers who have contact with workers in Belfast. Both [Q] and [?] appear to be increasing in Coleraine, the change toward [Q] led by men and the increase in [?] led by women. Kingsmore attributes these findings in part to differing concepts of prestige, masculinity, and femininity, with men showing a preference for variants associated with urban, working-class norms and women preferring variants associated with “homestead values”: family, church, and neighborhood (191). Social evaluation and gender, however, are just two of the issues associated with glottalization studies. Milroy and colleagues have also found glottalization useful in the study of alternation between local and supralocal norms (Milroy, Milroy, and Hartley 1994; Milroy et al. 1994; Docherty et al. 1997). They note that in many accounts of British English, standard English is often assumed to cause leveling of local dialects, but the situation can be portrayed more accurately if gradations of “local-ness” can be differentiated (Milroy et al. 1994, 2). In fact, they interpret the glottal replacement in Tyneside as an emergence of a nonlocal variant—one that is also increasing in the standard dialect as well as in urban dialects (Wells 1982).
230
american speech 81.3 (2006)
Other studies provide a contrast between phonetic interpretations of glottalization and those based on variation theory. Interpretations of glottalization as a lenition process, such as the Harris and Kaye (1990) study of New York and London speech, argue via phonetically based accounts of this phenomenon that loss or rearrangement of elements occurs under particular phonological conditions and can result in the formation of a glottal stop, as in London, or flap, as in New York. In other words, when conditions are met in such environments as intervocalic position, word final position, or /t/ following /r/, vocalized /l/, or nasal, then lenition occurs. Dialect variation is accounted for in the choice of one variant over another, but not in the occurrence or nonoccurrence of lenition. Variationist accounts of glottalization point out that reliance on large pools of data results in problems for lenition explanations. For example, rather than a continuum of least-to-most-weakened variants across speakers, social variation results in different evaluations and usage patterns across speakers, depending on social class, age, and gender, as noted above in the study by Milroy et al. (1994). In addition, Docherty et al. (1997) found that glottalization occurs in their Tyneside data in environments where it should not, according to lenition accounts. For example, lenition accounts claim that glottalization can take place in ambisyllabic stops (e.g., bottle, city) but not in syllable-initial stops (e.g., return). Docherty et al. found evidence that in Tyneside English, glottalization may take place in both. In summary, lenition accounts can capture general and common patterns but are not at all, as the studies assert, categorical. Finally, Foulkes, Docherty, and Watt (1999) examined the acquisition of glottalization in young children in Tyneside. They found that even their youngest speakers were making good progress in the acquisition of complicated glottalization patterns. Children as young as two years of age showed evidence of preaspirated glottalization (an unusual pattern found in Tyneside English in which the alveolar gesture occurs before the laryngeal stop) for word-final /t/, a finding that runs counter to the assumption that children favor rules of simplification. This preglottalization pattern is dominant in the speech of young women, who are generally responsible for child care in this community, indicating that children again may be participating in sound change by showing preference for female-led changes, learned in early childhood from their caretakers. I chose glottal stop replacement of /t/ as the variable to be explored in the current study because of its presence in the speech in all ages of Vermonters and its prominence in answer to the question “What is Vermont speech like?” Although participants did not mention the words “glottal stop,” they frequently referred to “t-dropping” or “not saying the ends of words.” As it
As Old Becomes New: Glottalization in Ver mont
231
appears to be present in the speech of children as well as adults, glottalization was of interest for two additional reasons: (1) if the variable were present in the speech of children equally or to a greater extent than in adults, it might suggest that glottalization is a strengthening or emerging variable in a potentially moribund dialect, and (2) the presence of glottal stop in child speech would indicate a venue in which to study children’s participation in language variation and change in an endangered dialect. dialect obsolescence. The model of language or dialect obsolescence most relevant to northwestern Vermont speech is that of dialect dissipation (Wolfram 2002). In this case the primary causes are generally economic in nature, having to do with outside businesses threatening smaller local ones, which in turn may influence the number and types of contacts speakers may have with those from outside their communities. Vermont currently fits the prototype for this type of change. Family farms, particularly dairy farms, which at one time dominated the economy of Vermont, are facing increasingly difficult times financially. Larger agricultural businesses are taking over these farms, and larger national chains are similarly replacing locally owned retail and service businesses (Meeks 1986; Vermont Business Roundtable, Jan. 1993). Finally, tourism has become increasingly important to the state’s economy, particularly in the autumn when tourists arrive to view the leaves as they turn color, and in the winter when skiing is a primary attraction (Meeks 1986; Department of Community Development and Applied Economics, Oct. 2000). These changes have resulted in many more outsiders (“flatlanders” in local parlance) moving into the area, as well as the exiting of some locals to obtain education or employment (Meeks 1986; Vermont Business Roundtable, Jan. 1993; Vermont Department of Employment and Training, 2003). As has been noted in other rural areas, this increased contact between speakers of many dialect areas may result in a dissipation of the local dialect (Callary 1975; Wolfram and Schilling-Estes 1995; Labov 2001). Features that have been noted in dialect obsolescence include the reduction of phonological distinctions, the loss of marked phonological features, and increased individual variation (Wolfram 2002). The latter two are of particular interest to the current study of glottalization in northwestern Vermont. Glottalization appears to be a marked variant in Vermont: speakers often respond to questions asking them to describe Vermont speech by noting that Vermonters often “drop their t s” and that those who use Vermont speech tend to be farmers, those who are not educated, or older male speakers. The frequency of the above remarks suggests that glottalization is noticed by speakers and stigmatized. One might predict, therefore, that
232
american speech 81.3 (2006)
glottalization would be a feature destined for obsolescence, particularly if it is, unlike in the United Kingdom, a localized variant. The issue of increased variability in moribund dialects and languages is also of interest in the current study. Findings on this topic are not consistent, but a number of studies have shown that in dying varieties individual variation may increase dramatically and without the systematicity generally noted in studies of sociolinguistic variation (Dorian 1994). Cook (1989), however, notes that what looks to be an increase in idiosyncratic variation may instead be simplification across a large number of speakers at different rates. Wolfram (2002) also notes that observed increases in variability may be frequency differences. Wolfram and Schilling-Estes (1999, 119) discuss the process of dialect death and note that the disappearance of a dialect does not necessarily mean a change toward standard English. In point of fact, the features may change greatly while preserving dialect boundaries. Glottalization could provide a look at this process if it proves to be an increasing, rather than a decreasing, feature of northern New England speech. Although there have been no previous studies of glottalization in Vermont, it appears likely that it is not a new phenomenon there. Although not discussed specifically by Kurath, there are occasional instances of glottal replacement in Vermont in the maps of the LANE (e.g., [mAUn?n] in mountain laurel, vol. 2, part 1, areas 254, 276, 286).1 In addition, as previously noted and borne out by current research, glottalization is found in all ages of speakers in Vermont and is associated by participants with older Vermonters. It may also be productive to examine leveling in Vermont speech as a process of diffusion by accommodation, as discussed by Trudgill (1986). He notes that accommodation may be the best way to explain the specific ways in which dialect features travel from larger to smaller cities, then to rural areas, essentially the process seen in leveling from outside influences: generally, urban features begin to replace local features in previously isolated areas. Trudgill outlines several factors that appear to encourage or inhibit linguistic accommodation, two of which are most relevant to the current study: salience and naturalness. Salience includes the presence of overt stigma, as well as involvement in linguistic change. However, another point to bear in mind is that too much salience has also been associated with loss of a feature (Trudgill 1986). Naturalness, particularly in the form of lenition accounts (see Harris and Kaye 1990), has already been associated with glottalization and will be further examined here. Both of these factors contribute to a “fixed route” of accommodation, which means that some features are more likely to be transferred than others, and, similarly, some features are more vulnerable to leveling processes.
As Old Becomes New: Glottalization in Ver mont
233
settlement of vermont. According to Meeks (1986), Vermont is largely a “Yankee state” in spite of the earliest inhabitants being indigenous groups—primarily Abenaki—the French, and some Dutch settlers. In 1760, after the end of the French and British hostilities, Vermont settlement by British-descended groups began. The composition of counties of Vermont by particular groups is challenging to discern, and many sources of such information disagree. One reason for the confusion is the proprietorship system developed in the early settlement history of Vermont (Klyza and Trombulak 1999). Under this system, land was granted to settlers on the condition that they occupy the land themselves. The system deteriorated over time, however, and speculators acquired land that they never lived on. For example, according to Klyza and Trombulak, in Windsor, Vermont, only 3 of the 59 proprietors granted land in 1761 ever settled there. Most settlers came from Connecticut, fewer from western Massachusetts and New Hampshire. Several Vermont settlements appear relevant to sociolinguistic study. As Meeks pointed out, they are primarily “Yankee,” but what Meeks does not distinguish is the fact that there appear to be at least two language and cultural varieties present in that British mix: English and Scotch-Irish (Kurath 1939a, 78, 88, 100; 1971, 18; Ford 1941, 221–48). English settlers came primarily via Connecticut and New Hampshire out of the Massachusetts Bay Colony, and the Scotch-Irish in the mid- to late 1700s from Worcester and western Massachusetts (Kurath 1939a, 88–89; Ford 1941, 228–29). By the time of the completion of the Erie Canal in 1825, the settlement of Vermont had slowed as many who were looking for frontier land went west, rather than north. However, it is also important to note that the settlers who were in Vermont tended to stay there. Of those who had arrived in Vermont by 1775, 70% remained in the same town in 1791 and 95% remained in the state (Klyza and Trombulak 1999, 55). These facts illustrate the relative isolation of the area, so important in maintaining a robust rural dialect.
METHOD The speakers for the current study comprised 47 participants, all of whom resided in Swanton, a small town in northern Vermont, and its surrounding area. They ranged in age from 3 to 80. As shown in table 1, speakers were divided into age groups as follows. Preschoolers comprised 3- and 4-year-old children from two local preschool programs. Kindergartners were 5 years old and from the local public elementary school. Fourth-graders comprised 9-year-old students from the middle school. The high school students ranged in age from 15 to 18. The parents’ ages ranged from early 20s to early 40s.
234
american speech 81.3 (2006) table 1 Age and Sex of Speakers Age Group Preschool Kindergarten Fourth grade High school Parents Older adults total
Female 5 4 3 3 10 2 27
Male 6 3 3 3 4 1 20
Total 11 7 6 6 14 3 47
These speakers were recruited through their children via follow-up telephone calls. Finally, three older speakers were recruited as a sort of baseline data source for this rural dialect. They lived at a local nursing home. As is typical in nursing homes, women far outnumbered men. This group consisted of two women and one man. At the time of the arrival of the Europeans, northwestern Vermont was inhabited by the Abenaki people. Many of the current residents claim Abenaki ancestry. It was found during the data collection phase, however, that although the Abenaki retain a cultural presence in the community, their linguistic assimilation, including the loss of their language in the 1950s, has made ethnicity impossible to utilize as an independent variable in the current study. Further studies, including examinations of vowels, are needed to explore the issue of ethnic identity further. Other ethnic backgrounds represented by the speakers and in the area more generally were English or Scottish and French Canadian. As with the Abenaki, many of the speakers identified themselves as some combination of the above (if they identified at all), and no relevant linguistic differences were noted among them. procedures. The children participated in one to four interview sessions lasting approximately 45 minutes each. The number of interviews depended upon the amount of speech gathered in each session. Older children and adults participated in one or two interview sessions, depending on scheduling constraints. The interview sessions were tape-recorded using a Sony Digital Audio Tape Corder (DAT) TCD-D8 with a Sony ECM-T140 lavalier microphone. All tokens of words containing a /t/ in a word-medial or word-final position were coded impressionistically as to the variant of /t/ produced as well as to the independent variables. Only those word-medial /t/s that were syllable-final or did not precede a stressed syllable (as in return) were included in the data. As the distinction between glottal replacement and glottal reinforcement is a
As Old Becomes New: Glottalization in Ver mont
235
crucial one in the study, these variants were coded conservatively. That is, any token that was in any way ambiguous was coded as the more typical variant in American English, glottal reinforcement. Goldvarb 2001 (Robinson, Lawrence, and Tagliamonte 2001), a probability-generating multivariate analysis application program for use with a Windows-based personal computer, was applied. It is a two-step statistical procedure: logical regression followed by a post hoc step-up/step-down comparison. Its purpose is to show the relative contributions of several independent factors to the overall variability of the dependent variable. Multivariate analysis is preferable to univariate analysis in cases, such as this one, when the data consist of spontaneous speech and, hence, are not evenly distributed across the independent variables (Cedergren and Sankoff 1974; Sankoff and Labov 1979). variables. Dependent Variable. As previously noted, the dependent variable in the present study is the use of glottal stop as a replacement for word-medial (syllable-final) or word-final /t/. Five possible variants were coded: glottal replacement, preglottalization, flap, aspirated /t/, and deletion. However, for the purposes of this examination, the glottal stop replacement will be the focus, as the other variants, as they appear in Vermont speech, are unremarkable in American English generally and are not stigmatized either locally or more generally in the United States. The [?] application variant was contrasted with the combined preglottalized, flapped, aspirated, and deleted variants for the primary part of the study. However, because of some interaction between flap and replacement in some phonological environments to be discussed below, Goldvarb runs with flap were also completed separately. These will be discussed with the phonological environment results. Independent Variables. Linguistic Factors. Three sets of linguistic independent variables were examined. The first was position of the /t/ in the word: medial or final. Glottalization does not occur in initial position. Although wordmedial /t/ occurs less frequently as a whole in the corpus than word-final /t/ and its glottalization is produced only in specific phonological environments, information gathered during the interviews indicated that it is a more stigmatized position, in that when speakers gave an example of “Vermont speech,” it was almost invariably a word-medial glottalized example, such as Milton or mountain. The second linguistic factor studied was the preceding segment. Possibilities coded included vowel, obstruent, liquid, glide, or nasal. The third factor was the type of segment that followed /t/, which included the same possibilities as preceding segment with the addition of pause. It will be important to note for the interpretation of this study that following segment implies the
236
american speech 81.3 (2006)
presence of a word boundary plus an additional segment or, if the token is at the end of an utterance and there is an easily discernible break before the speaker continues, a pause. Social Factors. Two social factors were investigated in the current project. The first was age. These included preschool (3–4 years old), kindergarten (5 years old), fourth grade (9 years old), high school (15–18 years old), parents (early 20s through early 40s), and older speakers (70–80 years old). Speaker sex was also included in this analysis. There has been much discussion about the inadequacy of dichotomous sex coding as a helpful strategy for interpreting language variation data (Eckert 1989; Eckert and McConnell-Ginet 1992; Bucholtz 1999; Milroy 2002). However, speaker sex was included in this examination for two reasons. First, it will supply a basis for further interpretation. Second, there have been findings of gender differences and conflicting accounts of their significance in many previous studies of glottal stop replacement of /t/ (cf. Mees 1987; Milroy et al. 1994; Kingsmore 1995).
RESULTS The results of the analysis of the effects of linguistic and social factors will be discussed in the following section. The results, comprising raw numbers, percentages, and probabilities are listed in table 2. linguistic factors. There were three linguistic factors examined in this study: position in the word, preceding segment, and following segment. Of these, only the first did not have a significant impact on glottalization. That is, there was no significant difference in the probability of glottal replacement of /t/ between word-medial /t/ and word-final /t/. This finding was somewhat surprising as the overall impression in listening to Vermont speakers is that word-final glottalization is much more frequent than word-medial glottalization, as well as less stigmatized, according to speaker report. In addition, the possible environments for word-medial glottalization are very few. This process occurs in Vermont speech only when the /t/ is preceded by a vowel, nasal, or liquid and followed by a syllabic nasal, generally /n/. This is a far more restricted environment than has been noted in British English by other researchers in this area (e.g., Trudgill 1974; Mees 1987; Milroy et al. 1994; Kingsmore 1995; Foulkes, Docherty, and Watt 1999). The distribution of the current data suggests that, unlike British English, Vermont speech allows word-medial glottalization only in the environment in which glottal reinforcement of /t/ (or preglottalization) is permitted and alveolar flap is
As Old Becomes New: Glottalization in Ver mont
237
table 2 Variable Rule Analysis Results for Glottalization in Vermont Input probability = 0.139 App./Total % Prob. Linguistic Factors Position in word Medial Final Preceding segment Vowel Consonant Following segment Pause Nasal/Glide Nasal Glide Obstruent Vowel/Liquid Vowel Liquid Social Factors Gender Male Female Age Group 3–5-year-olds 9-year-olds/Older 9-year-olds Older 15–18-year-olds Parents total
χ2
p
