Published in The Journal of English Linguistics. This is a revision of a paper presented by the first author at the American Dialect Society meeting in April 1998 in New York City.

Authors: Wendy Ryback-Soucy, Temple University

and

Naomi Nagy, The University of New Hampshire

Abstract. We present data showing the existence of an ethnically marked dialect of English spoken in Manchester, NH. Through examination of the speech of a group of 16 Franco-American English speakers, we illustrate the existence of correlations between phonological variation and social factors that relate to degree of involvement in the Franco-American community. Speakers who are more actively involved in French activities such as social clubs, or who speak French regularly, or who live on the Franco-American side of town are significantly more likely to exhibit French features in their speech (even if they are monolingual English speakers), specifically with regard to the variable use of dental stops in place of interdental fricatives, and deletion of word-initial /h/. This suggests that, while the use of the French language is decreasing, speakers are still marking their French ethnicity in their English speech patterns.

1. INTRODUCTION AND BACKGROUND

Numerous factors can influence diversity in a spoken language, e.g., the languages one is exposed to, the geographic area one is raised in, and family traditions. This research focuses on the first of these factors, examining the French language’s influence on Franco-American English in Manchester, a city in southern New Hampshire (NH).

The questions we explore include: Does the Franco-American community of Manchester have its own French-influenced dialect? What sociolinguistic factors influence this dialect? Do speakers with stronger cultural ties to the Franco-American community produce a more French-sounding English? Can immersion in Franco-American culture have an influence on speakers who are monolingual (in English)? Our hypothesis is that the speakers who are culturally more Franco-American-oriented and/or linguistically more French dominant, will have a greater incidence of French variants in their English language. Correlations between linguistic attitude, social background, and choice of a particular pronunciation variant of a sound have been demonstrated repeatedly (cf. Labov 1972, Nagy et al. 1995).

There has been a strong presence of French in Manchester, NH, since the late 1800's, and therefore, evidence of French influence on that area's dialect seems likely. The 1976 Survey of Income and Education found that 1.9 million Americans reported French language background, making French the fourth largest language minority group in the U.S. (Grosjean 1982:45). NH's population was reported to be 15% French (ibid. 47). Nationwide, approximately 40% of people of French descent were speaking French at least part of the time as of 1976 (ibid. 56). As of the 1995 census, 1,702,000 U.S. residents (nearly 1% of the U.S. population) report speaking French natively (U.S. Bureau of the Census 1995). Speaking French every day is an important link to the Franco-American heritage. The use and/or exposure to French on a daily basis influences the production of American English in this community.

Manchester, NH was chosen for this investigation due to its large population of Franco-Americans. Between the years 1870-1930, there was a large scale immigration of French Canadians to New England because of the explosion of the textile industry and the ensuing need for laborers. By 1900, the population of Manchester, NH, was 40% Franco-American (Brault 1986). For years, the west side of Manchester has been called "Little Canada" because of its density of Franco-Americans. Residents there are very involved in Franco-American culture and community, attending Franco-American churches, holding membership in the Association Canado-Americaine and the American-Canadian Genealogical Society, and attending conferences, meetings, and social events put on by these groups. It is not surprising that, in some individuals, this active ethnic pride affects their speech.

2. THE LINGUISTIC VARIABLES

The two linguistic variables examined are:

(TH) Interdental fricatives. These may be pronounced (1) as [†] and [¶], as in English, or (2) as [t] and [d], because French lacks interdental fricatives, and these stops are the most similar phonemes available from the French inventory. In Manchester, them may be pronounced as [¶´m] or [d´m] and with as [wˆ†] or [wˆ t], the latter in each pair suggesting French influence.

(H) Word-initial /h/. This segment may be either (1) articulated or (2) deleted, producing forms such as [´r] hair, [ænd] hand, and [´dek] headache. The French language lacks /h/ in any environment (Janda & Auger 1992), therefore deletion of this sound indicates French influence.

3. METHODOLOGY

The fieldwork was conducted during the summer of 1997. Social and linguistic information from a group of Franco-American volunteers was obtained via interviews conducted in English. The first author conducted and tape recorded seventeen interviews with speakers contacted via her personal network. This network has been established over several generations during which the first author's husband's family has resided in Manchester. Of the willing participants, sixteen were chosen, eleven females and five males-- this imbalance was due to greater willingness on the part of women to participate. The selection criteria were: the speaker was born and raised in Manchester, the speaker was of Franco-American heritage, and the speaker was willing to participate in a tape-recorded interview. The speakers ranged in age from 35 to 89 years old. Each interview lasted 45 to 90 minutes. The interviews consisted of descriptions of personal experiences and favorite stories, elicited by a series of interview questions. Topics of the interview questions revolved around what it was like growing up in Manchester, degree of involvement in Franco-American activities, attendance at church, schooling and occupation, and how the Franco-American culture extends into their lives today. At the beginning of each interview, all speakers were asked if they considered themselves monolingual or bilingual. The answers were coded as follows: (1) monolingual in English, (2) said they are monolingual in English but the first language spoken in childhood was French, and (3) bilingual in French and English.

After each interview, the tape recording was listened to and the following information was noted: cultural practices during the speakers' childhood, schools and churches attended (French or English language), marriage and family life, customs and traditions passed down through the generations, and when and if French was used during their lifetime. The social factors were catalogued as follows:

Age and gender

Females less than 45 years old

Male less than 45 years old

Female greater than 60 years old

Male greater than 60 years old

Language use (self-reported)

1 = Monolingual in English

2 = Monolingual in English, but L1 was French in childhood

3 = Bilingual in English and French

Use of French in high school, at work, with the family, and at social clubs

0 = no use of French in any of the four categories

1 = use of French in one of the four categories

2 = use of French in two of the four categories

3 = use of French in three of the four categories

4 = use of French in all of the four categories

Neighborhood (the neighborhood one grew up in)

West side of Manchester ( "Little Canada")

East side of Manchester

North side of Manchester

Highest level of education achieved

Elementary

High school

College

Occupational class

Blue collar (e.g., a laborer)

Pink collar (e.g., a secretary)

White collar (e.g., a manager)

Table 1 shows how each speaker was coded for each of the social variables, and provides the percentage of instances in which the speaker produced the standard variant for the two linguistic variables (coding method discussed below).

Table 1: Description of speakers

Speaker	Age	Language use	Use of French	Neighbor-hood	Education Level	Type of occupation	(TH) % fric.	(H) % [h]
Women
Lynn	35	1	0	west	college	pink	100	100
Danielle	42	2	2	east	high school	pink	100	100
Brea	43	3	1	east	high school	pink	100	90
Yvonne	66	2	2	east	high school	pink	100	96
Irene	70	3	4	east	college	pink	98	100
Marie	73	3	2	east	elementary	blue	74	96
Marguerite	75	3	2	west	high school	pink	100	100
Janette	76	3	4	east	elementary	blue	77	80
Sarah	77	3	2	west	elementary	pink	79	62
Muriel	89	3	4	west	college	pink	40	66
Men
Jean-Guy	39	3	1	west	college	pink	100	100
Henri	41	2	3	west	high school	white	88	100
Gerard	63	3	2	east	high school	pink	86	98
Roger	70	1	0	north	elementary	blue	61	94
Rene	76	3	4	west	college	pink	82	98

The first author transcribed a portion of each recording beginning about 15 minutes into the conversation. This was done in order to avoid the initial awkwardness that may be experienced by the speaker because of being tape recorded. Approximately fifteen minutes into the conversation the speaker should feel more comfortable talking with the interviewer and use more vernacular speech, the topic of this linguistic analysis.

Tokens of words containing the two variables (TH) and (H) were coded for the following independent linguistic variables: the phonological context of the segment (preceding and following segments), voicing of the segment (for (TH) only), and the segment’s position in the word (beginning, middle, or end). For (H) another category was added: whether the word containing the segment was an unstressed pronoun or not. Word-initial /h/ is frequently deleted in American English if it appears in an unstressed pronoun ('I saw 'im'.). The independent social variables are those listed in Table 1 above.

Approximately 1600 token were examined for (TH), 100 per speaker. Approximately 800 tokens of (H) were examined, 50 per speaker. A quantitative analysis program, Goldvarb 2.0 for the Macintosh (Rand & Sankoff 1990), was used to perform a binomial regression "one-level") analysis of the coded factors. Significance of the effects is determined by the step-up, step down binomial regression procedure in Goldvarb. In the statistical analyses, the native English-like forms (fricative for (TH) and [h] for (H)) are used as the application values. That is, all percentages and probability weights presented are for production of native English-like forms.

4. RESULTS for (TH)

Overall, there is an 83% rate of fricative production for (TH) in this data sample: 1,070 out of 1,283 tokens of (TH) were produced as fricatives. This translates to an input value of 0.885. Seven of the 16 speakers examined showed categorical use of the fricative variant, indicating a lack of French influence as measured by this variable. Of these speakers, four also exhibit categorical presence for (H), further supporting their status as non-French influenced. Data from these seven speakers have been eliminated from the statistical analyses, but this fact is considered in interpreting the effects of the social variables. Six of the 11 women did not show variation for (TH), while only one of the five men categorically exhibited the fricative variant. However, these results can be interpreted another way, due to the skewed data sample. The five youngest women and the youngest man all categorically produced the fricative variant, so the pattern may be more age-graded than sex-correlated.

The speaker who uses the stop variant most frequently is Muriel. Muriel is an eighty-nine year old bilingual female who grew up on the west side of Manchester and has used French in almost every aspect of her life. She attends French Catholic masses, and in her younger years, was actively involved with French social groups, used French at home and occasionally at work. Her involvement with the French community has clearly influenced her pronunciation of English. This suggests that the non-standard variant is a marker of strong ethnic group membership.

In the following sections, effects of each significant independent variable are discussed. This may be seen as an effort to understand which aspect(s) of Muriel's background are most responsible for her high rate of stop production for (TH). More generally, we are supporting the hypothesis that the use of these non-standard variants is indexed to strength of ethnic affinity.

4.1. Linguistic variables

Although a number of linguistic variables were examined, only the fricative's position in the word proved significant. (TH) is far more likely to appear as a stop in word-final position than elsewhere, i.e. when it is in coda position. See Table 2. Voicing of the fricative and the category of the preceding and following segments did not show significant effects and were, therefore, eliminated from further analyses.

Table 2: Significant linguistic variable for (TH)

Position in word	Number of fricatives	Percent fricatives	Factor weight
Beginning	873/1047	83	0.501
Middle	156/174	90	0.594
End	41/62	66	0.238

4.2. Social variables

A typical age and sex pattern for a non-standard variant is found (Labov 1990). All the younger (<45 years old) women categorically produce the standard variant. The older women (>60 years old) produce far fewer standard forms, with a factor weight of only 0.423. See Table 3. The pattern for men is similar, with younger men producing more standard forms than older men. Younger men, however, produce fewer standard forms than younger women, while there is very little difference between the older men and the older women. This pattern leads us to propose a change in progress away from the stop variant. Given this evidence of a female-led change toward the standard or more generally-used form, we can explain the correlations of (TH) and the other social variables.

Table 3 : Effects of significant social factors on (TH)

Speaker group	# of [†, ¶] / Total	% [†, ¶]	Factor weight
Age and sex
Older men	183/195	94	0.467
Older women	660/790	84	0.423
Younger men	227/298	76	0.812
Younger women	300/300	100	(1.000)
Occupation type
Pink collar	862/988	87	0.644
Blue collar	208/295	71	0.121
Level of education
Post-secondary	316/395	80	0.315
High school	469/495	95	0.534
Elementary	285/393	73	0.647
Neighborhood
West side	481/591	81	0.333
East side	528/592	89	0.670
North	61/100	61	0.480
TOTAL	1070/1283	83

There is a significant correlation with type of occupation. As shown in Table 3 blue-collar workers have a much lower factor weight for (TH), indicating higher use of the non-standard stop variant, compared to pink-collar workers.

The effect of education is initially surprising: those speakers with only an elementary level of education have the highest factor weight, indicating the most standard speech. Those with college education have the lowest weight, indicating least standard speech. However, these results may be skewed by the fact that several speakers were eliminated from the analysis because they exhibited categorical use of fricatives. Two of the four college-educated speakers were eliminated, and five of the seven high school-educated speakers were eliminated, but no elementary-educated speakers were eliminated. This brings the elementary-level speakers in line as the group showing the highest use of stops. However, taking the categorical speakers into account does not account for the fact that the high school-educated speakers having a higher rate of standard forms than the college-educated speakers.

There are several possible explanations. Perhaps the high school-educated speakers feel more pressure to prove themselves "all-American," while the college-educated speakers are secure in their position and do not fear other people's judgments of them as much. In a high-pressure situation (such as an interview with a stranger), the middle group of speakers may exhibit the highest degree of linguistic insecurity-- producing far more standard forms than they do in a casual situation. Such patterns are common (Labov 1972:244). In such cases, the most educated speakers are likely to exhibit more of the non-standard dialect forms, because they are secure in their social position.

Alternatively, it may be the most educated speakers who are producing atypical speech. Might these speakers use a different register when talking outside the home, reducing the French influence in their language? Additionally, the type of college one attended may affect the rate of stopping. Adapting the marché linguistique approach (Sankoff & Laberge 1978), those that graduated from a technical/trade school would most likely be employed doing skilled mechanical work, where one’s speech is not a focal point. Those with a baccalaureate degree may acquire jobs with more chances for public speaking, and therefore may value more standard speech. Speakers who attained a college degree were not differentiated by type of degree. This is something to consider in future studies. There is a great deal of overlap between level of education and type of occupation. The findings for the occupation variable do directly support this marché linguistique analysis: pink collar workers have a much higher factor weight than blue collar workers (See Table 3). To resolve this issue in future analyses, one might select a speaker sample with less interaction between education and occupation.

Even when including social variables such as education, occupation, age, and sex, which typically correlate to degree of non-standard usage in the model, the effect of living in a Franco-American neighborhood is clear: the weight for speakers on the West Side ("Little Canada") is half the weight for speakers on the East Side: people surrounded by other Franco-Americans have a higher rate of stops, indicating the influence of French in the dialect. That is, no matter what the level of education or employment is, being in contact with the Franco-American community is conducive to a higher rate of use of the non-standard variant. Only two West Side speakers categorically produced fricatives, and they are both college-educated, pink collar workers.

Use of French was not found to be significantly correlated to this linguistic variable. Its effect is unclear: even in runs where the weights for the different French use scores differed significantly, they were not ranked in order with the amount of usage of French (See Table 3). Similarly, reported degree of bilingualism do not correlate significantly to the linguistic variable. Comparing these findings with those for neighborhood, it appears that interaction with other Franco-Americans (and therefore, at least some French-speakers) is better correlated to variation for this variable than is actual personal use of French. This remains for further investigation-- perhaps a different scale of language use is necessary, since this scale does not produce scores that correlate well with (H) either.

Figure 1 illustrates that the factors selected as significant (position in word of fricative and the following social variables: neighborhood, education, occupation, and age and sex) account for a large portion of the variation for (TH). All points but one cluster closely about the regression line indicating a good fit between the model and the data. The binomial analysis which provided the factor weights listed in Table 2 and Table 3 has a total c2 of 187.4, a c2/cell of 5.2, and a log-likelihood of -482.

Figure 1: Scattegram and regression line for (TH) analysis

[INSERT FIGURE 1 HERE.]

 

5. Results for (H)

Overall, there is a 90% rate of [h]-production in this data sample: 542 out of 601 tokens of (H) were produced as fricatives, rather than deleted. This translates to an input value of 0.976 in the binomial analyses presented below. Six of the 16 speakers examined showed categorical use of the fricative variant (the standard), indicating a lack of French influence. Of the ten speakers who sometimes delete (H), nine were over the age of 60. Four of the six speakers who categorically produce the standard variant of (H) also categorically produce the standard variant of (TH), showing strong correlation between the two linguistic variables under examination.

The speaker with the most deletions is Sarah. Sarah is a seventy-seven year old bilingual who exhibits a 38% deletion rate. She grew up on the West Side of Manchester, achieved an elementary school education, uses French with her family and friends, and met her husband at a Franco-American social club. Sarah is the only speaker who demonstrated hypercorrection for this variable. When talking about a local high school, she called it [ho'li he'nd½\l haj] "Holy (H)angel High." Her husband, also Franco-American, joked with her about "HHH" during the interview, indicating that this pattern is recognized and stigmatized.

5.1. Linguistic variables

We find the expected correlations to the linguistic independent variables. As shown in Table 4, significant effects of the preceding segment and the type of lexical item are found. (H) is present most when there is a preceding vowel, followed by preceding liquids, then less for a preceding consonant and least when it is phrase-initial (following a pause). Word-initial /h/ is articulated less frequently in unstressed subject pronouns than in other words, as is true in all dialects of English. We do not pursue the effects of linguistic factors further here, but only include them in our model to be sure that we are not presenting a skewed analysis of the social factors.

Table 4: Significant linguistic variables for (H)

Group	# [h] / Total	% [h]	Factor weight
Preceding segment
vowel	271/277	98	0.769
liquid	53/57	93	0.412
consonant	212/258	82	0.245
pause	6/9	67	0.078
Lexical category
all other categories	394/417	94	0.575
unstressed pronoun	148/184	80	0.336

5.2. Social variables

Dividing the speakers into four groups by age and sex, we see the typical pattern for a non-standard (stigmatized) variant (see bottom of Table 5). Younger male speakers produce (H) categorically, as do two of the three younger women. Younger female speakers (represented only by Brea, as all others in this category have categorical presence of (H)), have a much higher weight for this variable than older female speakers, indicating less use of the non-standard form. Older men have the highest weight, showing the most use of the non-standard form. However, the apparent oddity of a much higher factor weight (indicating higher use of the standard variant) for older men than for older women is diminished when the two categorically (H)-producing older women (who were eliminated from the statistical analysis) are taken into account.

Table 5: Effects of social factors on (H)

Group	# [h] / Total	% H	Factor weight
Age and sex
Older men	99/101	98	0.848
Older women	348/400	87	0.336
Younger men	100/100	100	(1.000)
Younger women	95/100	95	0.730
Level of education
College	132/150	88	0.568
High school	291/301	97	0.499
Elementary	119/150	79	0.433
Occupation
White collar	50/50	100	(1.000)
Pink collar	454/501	91	0.557
Blue collar	88/100	88	0.242
Neighborhood
East Side	379/401	95	0.672
West Side	163/200	82	0.192
Use of French
In 2 environments	294/301	98	0.778
In 3 or 4 environments	203/250	81	0.267
In 0 or 1 environment	45/50	90	0.076
Degree of bilingualism
bilingual	444/501	89	0.573
monolingual, French = L1	98/100	98	0.186
TOTAL	542/601	90

A much clearer effect of education is observed for (H) than for (TH). Table 5 shows that the less education one has, the more likely one is to delete (H). Type of occupation is also strongly correlated: white-collar workers categorically produce [h] (and have, therefore, been excluded from the statistical analysis), and pink-collar workers do so more frequently that blue collar workers.

The production of (H) is also indexed to several culturally relevant variables, some of which are directly representative of the degree of "French-ness" of the speaker. Speakers who grew up on the West Side of Manchester are more likely to delete (H), compared with those on the East Side. Living on the West Side with its strong presence of Franco-Americans evidently has an impact on the speakers’ English, with respect to (H).

Similarly, speakers that used French in more contexts (scores of 3 or 4) had more (H) deletions than those that used French in fewer contexts (scores of 2). Surprisingly, those who use French in the fewest contexts (0 or 1 contexts) have the lowest rate of (H). Are they (over-) using this marker in English to compensate for not having a French mother tongue as a cultural marker? Supporting this argument, speakers who have lost their French mother tongue have a much higher rate of usage of the French-like variant than the bilinguals.

The factors which have been determined significant (See Table 4 and Table 5) provide a model with a total c-square of 261.7, and a c-square/cell of 3.6, and a log-likelihood of -120.0. The accuracy of the model can be observed in the scattergram in Figure 2. We see that this model is not quite as good a fit as the model developed for (TH), but there is reasonable clustering about the regression line.

Figure 2: Scattergram and regression line for the (H) model

[INSERT FIGURE 2 HERE.]

 

6. DISCUSSION

We have shown that the French language does have an impact on the language of the Franco-Americans of Manchester, New Hampshire. This French influence is predominantly seen in the older speakers. As predicted, most of the younger speakers do not show signs of the French language’s influence, that is, they do not produce (TH) as a stop or delete word-initial (H). The younger speakers are not linked as tightly to the French community as the older speakers. According to their self-reports, older speakers have been more involved in the Franco-American community. Most of the involvement took place when the speakers were younger because at that time there were more activities available. Such activities include attending Catholic churches where mass is conducted in French, memberships in French social groups such as the Association Canado-Americaine and the American-Canadian Genealogical Society. In recent years involvement in the Franco-American community has declined. The Franco-American festivals that were at one time enthusiastically anticipated are virtually non-existent today.

As an example of this trend, the youngest speaker, a thirty-five year old monolingual female, has the least amount of French influence in her speech; she did not produce any stops or deletions. The rest of the speakers in the "younger" category were closer to the upper age limit of 45 and minimal language variation was noted in their speech. As a whole, the younger generation does not show as much interest in Franco-American traditions as the older generation. Only two out of the five younger speakers consider themselves bilingual, while nine of the twelve older speakers do so. This age-grading corresponds with the proposal that the speakers who are culturally more French-oriented and/or linguistically more French dominant, will have more evidence of French influence in their speech.

All of the speakers involved in Franco-American clubs, who are also all bilingual, produce more stops for (TH) than those Franco-Americans who did not have any social club involvement. During these social events the speakers use French in casual conversation. Active involvement in social groups allows one to make new friends, as these speakers did, and had more instances to use French compared to those who were not involved with the clubs. Similarly, growing up on the west side of Manchester has had an influence on the Franco-Americans’ speech. The "West-Siders" speech sample contains many more instances of the non-standard variants for both (TH) and (H).

Speakers that classified themselves as bilingual produced more stops than the other two groups (monolinguals and monolinguals who have lost their French mother tongue). While the variation for the two dependent linguistic variables did not prove significantly correlated with these self-reported degrees of bilingualism, the relatively high degree of stops produced for (TH) by the middle group shows that the speech of this group does reflect some influence from French, even though these speakers consider themselves monolingual. Among the monolinguals, roughly 20% of the (TH) tokens were produced as stops. All of these tokens come from Roger, an older man who grew up in North Manchester. He reports that mostly English was spoken in his home but there was occasionally French. Of note, Roger has been married to Muriel for seventeen years. Muriel has the highest incidence of stops among the speakers at 60%. Knowing Muriel for twenty years may have had an effect on Roger’s speech. In contrast, all other monolinguals speakers produce only fricatives for (TH).

Gender is correlated to both linguistic variables, and, again, this may relate to the notion of degree of involvement in the French culture. We had expected males to have more contact outside of their home environment than females, and therefore, to show less French influence in their speech. This did not turn out to be the case. However, all of the females speakers interviewed did, in fact, work outside of the home at one time or another. Although the females stayed home a majority of the time and raised the children, it is possible they had more connections with the neighborhood women, who may not have been French (especially on the East Side). This could influence their speech to sound less French-influenced.

To conclude, there was one older bilingual speaker, Marguerite, who did not produce any (TH) stops or (H) deletions in her speech, although her social profile would predict that she do so. She was raised on the West Side, involved with the Franco-American Center, married to a Franco-American, uses French with family and friends (including some of the speakers we interviewed who exhibit a strong influence of French) and yet shows no signs of influence from the French language. This demonstrates the need to analyze linguistic variation across a group of speakers as opposed to simply analyzing a single speaker who may not fit into the general pattern. When these twenty speakers are examined as a group, the stochastic effects of the social variables are clearly evident. An overall pattern of correlation between integration in the French culture and French influenced English is evident. Even though French isn’t spoken much on the streets of Manchester today, its effect is still heard in its French-influenced dialect of English.