Corps de l’article

1. Introduction

In this paper[1] we explore the influence of grammatical proscription on morphosyntactic change (or lack thereof) in two Franco-American[2] communities. Our data was gathered through sociolinguistic interviews with sixty-nine French-English bilinguals including thirty-five informants from Southbridge, Massachusetts and thirty-four from Woonsocket, Rhode Island between June of 2002 and August of 2003.

French is endangered everywhere that it is spoken in the United States. If Franco-American were to follow the path of a great many other language varieties in decline, we would expect rapid linguistic change (e.g., structural simplification) to occur as the shift from French to English advances in New England (Rottet 2001; Russo and Roberts 1999; Mougeon and Béniak 1991; King 1989; Dorian 1981, 1989).

In the French verbal system, 17 intransitive verbs call for être as the auxiliary in compound tenses such as the passé composé; otherwise avoir is used. Since from a functional perspective only one auxiliary is needed, structural simplification would be expected to encourage the spread of avoir and its eventual employ as the only auxiliary verb in the Franco-American variety. Because the use of avoir in cases where the standard language calls for être also represents a usage that is linguistically proscribed, an examination of its distribution in Franco-American provides the opportunity to consider whether linguistic proscription can slow or even stop qualitative changes which might otherwise add to the stigmatization of this variety.[3]

We begin with a brief historical overview of French-Canadian immigration to Southbridge and Woonsocket followed by an explanation of the sociolinguistic field methods we used to gather our data. Next, a discussion of previous research on auxiliary variation in North American French provides the basis for the selection of the linguistic and sociolinguistic variables that we include in our multivariate analysis. Finally, a detailed presentation of the results of the analysis and a discussion of our findings in the light of what has been reported elsewhere lead us to conclude that Franco-American French is not undergoing the rapid morphosyntactic change predicted by its current situation. Our results suggest that the Franco-American communities of Southbridge and Woonsocket stand out as a unique case of language shift.

2. The communities

The Franco-American communities of Southbridge, Massachusetts and Woonsocket, Rhode Island have much in common historically. The francophone immigration of the nineteenth and early twentieth centuries drew relatively homogenous populations from the same geographical region of Canada to both locations over roughly the same time period. The first French Canadian family to settle in Woonsocket came as early as 1814 or 1815 (Bonier 1920: 95). Southbridge, which was incorporated in 1816, saw the arrival of French Canadian immigrants in 1832 (Gatineau 1919: 15; Brown 1980: 26). In both locations, these Franco-Americans were joined by friends and family who filled the towns to work in bustling factories. In each community, the majority of families came not only from the same rural region of Quebec, but from many of the very same small towns (Brault 1986). In the 1830’s for example, immigrants from the town of St. Ours in the Richelieu valley settled in high numbers in both locations (Brault 1986: 56).

In 1900, Franco-Americans already made up 60% of the populations of both Southbridge and Woonsocket (Brault 1986: 54-55). Their numbers continued to grow until the 1930’s when the Depression and changing immigration laws ended the arrival of new Francophones (Roby 2000, 1990; Brault 1986). By that time, Southbridge and Woonsocket had become two of New England’s oldest and most well known Franco-American centers.

As the Franco-Americans of Southbridge and Woonsocket profited from the social mobility made possible by their majority status and a booming American economy in the nineteenth and early twentieth centuries, the French speaking populations exhibited a great deal of social and economic stratification in both locations. In fact, a number of prominent Franco-Americans held public office and were therefore quite visible in both communities (Gatineau 1919; Brown 1980; Bonier 1920).

Franco-Americans were successful in creating an impressive number of groups including mutual aid societies, social clubs, French language newspapers, and parish organizations. Through the establishment of Franco-American national parishes and bilingual parochial schools, the Catholic Church played a key role in language maintenance and transmission. French was also commonly used in places of work such as factories and mills, where Franco-Americans worked together in large numbers. Complex French language infrastructures therefore provided strong support for the language in both Southbridge and Woonsocket, as was the case in a number of other Franco-American centers throughout New England (Roby 2000; 1990; Brault 1986; Brown 1980; Gatineau 1919; Bonier 1920.

Since the 1930’s, language shift has been advancing in both communities. French is no longer commonly used at work. All but a handful of masses in French, and every one of the bilingual school programs gradually disappeared. Along with the decline of French in work, education, and religion, many of the organizations that had been so important in promoting the language are no longer in existence. Currently, French is used mainly between family members and friends, and even these domains have been greatly penetrated by English. In other words, there is no longer a single domain that is reserved exclusively for French in either community (Fox and Smith 2007, 2005).

3. Fieldwork

By investigating eight communities throughout New England, two teams of researchers lead by Professors Cynthia Fox (University at Albany) and Jane Smith (University of Maine, Orono) have created a mega-corpus of Franco-American French (see Fox and Smith 2007, 2005 for an overview of the project). Southbridge and Woonsocket are two of eight communities targeted in the project.[4] During fieldwork, interviews were conducted with approximately 30 Franco-Americans from each location, 275 in all. Interviews were guided by use of a questionnaire to gather information on topics such as the acquisition, use and transmission of French, and access to francophone culture and media. A translation task (English to French) was also used to elicit structures that are infrequent in conversation. Through the study of this mega-corpus, Fox and Smith (2007, 2005) have confirmed that language shift is more advanced in southern New England than in the north, and that Franco-American communities can vary greatly in terms of language maintenance.

Fieldwork has also shown that despite the many historical similarities between Southbridge and Woonsocket, language shift has progressed to different degrees in the two locations. For example, the shift from French to English is more advanced in Southbridge than in any other community studied and Woonsocket stands out because language maintenance has been higher there than elsewhere in Southern New England (Fox and Smith 2007; 2005; Fox 2007). U.S. census reports from 1990 and 2000 also indicate that when compared to Southbridge, a greater percentage of Woonsocket residents declared French or French Canadian ancestry as well as French use in the home in both years.

Finally, fieldwork has revealed that conservative language attitudes exist in both communities (Fox 2005; Stelling 2007). A number of speakers such as SO-S6 and SO06 for example, voiced concerns about the quality of the local French, often making reference to more prestigious or standardized varieties of the language, such as the French taught in parochial or public schools.[5]

The conservative language attitudes of Franco-Americans from Southbridge and Woonsocket and their exposure to the standardized varieties of the language (through school, travel and media) may make the communities unique when compared to other cases of language shift. The variety may be resistant to certain qualitative simplifications that might be labeled by parents, teachers and other speech monitors in the communities as “bad” French. In the following analysis, we investigate whether linguistic proscription may be preventing avoir from spreading to replace être to become the sole remaining auxiliary verb in the passé composé, a development that has been noted in other communities in the United States (Locke 1949; Marie-Francia and Marie Anaïs 1958; Papen and Rottet 1997; Russo and Roberts 1999).

4. Previous research on the passé composé in North American French

Auxiliary variation in the passé composé has been studied in New England, Montreal, Ontario, and Louisiana. Some of the factors that researchers have found to have a significant influence on variation are: age, gender, socioeconomic status, degree of language restriction, verb frequency, the presence of a transitive homonym, and the presence of an adjectival homonym. Each of these is therefore included in the multivariate analysis.

In New England, Locke (1949) found that avoir is the standard auxiliary for intransitive verbs of motion (e.g., Il a parti à cinq heures ‘He left at five o’clock’) whereas Sisters Marie-Francia and Marie Anaïs (1958) reported that the use of avoir with these same verbs was specifically associated with the working class. A possible explanation for their contrasting findings is that Locke worked with a single informant from Maine and Sisters Marie-Francia and Marie Anaïs examined a variety of Franco-American newspapers representing three social classes. Decades later, Russo and Roberts (1999) found that there was no social meaning associated with the non-standard use of avoir by twenty-two informants from Vermont and that its spread in their speech was promoted by main verbs which occur infrequently in conversation (e.g., revenir, descendre, arrêter) and by main verbs which have a transitive homonym (e.g., Je suis rentréà cinq heures ‘I came home at five o’clock’ vs J’ai rentré les foins ‘I brought in the hay.’)

Sankoff and Thibault (1980) reported that variability carried social meaning in Montreal. Use of être was specifically associated with higher socioeconomic status and use of avoir was especially frequent among men. Choice of auxiliary also depended on the exposure to and attitudes toward Standard French of a given speaker. Among the linguistic factors considered, they found that verbs with an adjectival homonym (il a parti à cinq heures ‘he left at 5’ vs il est déjà parti ‘he is already gone’ (state)), a transitive homonym, and verbs of motion were the most susceptible to the spread of avoir. Their general conclusion was that a change in progress in favor of avoir had been blocked by pressure to conform to Standard French.

Canale et al. (1978) reported that several linguistic factors influenced auxiliary variation in Ontario. High frequency verbs and verbs which were morphologically complex (e.g., revenir vs venir) showed resistance to the spread of avoir while verbs with a transitive homonym were more likely to be used with this auxiliary. Finally, the authors stated that they could not fully account for variation and that there was at least one unknown factor at work. When Béniak and Mougeon (1989) and Nadasdi (2005) analyzed auxiliary variation, they each found that language restriction played a significant role. In both studies, highly restricted users of French showed a preference for avoir whereas unrestricted speakers did not.

In a study of Cajun French, Papen and Rottet (1997) showed that when être is used, the speaker is referring to the resulting state of an action while when avoir is used, reference is made to the action itself. Use of être was limited to a very small number of verbs and was found only among the oldest group of speakers in their corpus. Picone and Valdman (2005) echoed the conclusions that être was used only by older speakers and with a very limited set of verbs. They also found that avoir may be used with all verbs, including intransitives and pronominals. Given the very similar observations found in the two studies, it is evident that the invariable use of avoir is well on its way to replacing être in Cajun French, a variety which finds itself today in a stage of language shift closely resembling the situation in New England.

5. The multivariate analysis

Concordances were made using Simple Concordance Program, version 4.08.[6] Before extracting instances of the passé composé from the transcriptions, all lines spoken by the interviewer were eliminated. Interviews with multiple informants were then divided into multiple documents, one for each speaker.

Concordances were created from the conversation data only, as the translation task did not involve systematic use of the past tense of any of the seventeen intransitive verbs studied here. All lines containing past participles of verbs that require être as an auxiliary in Standard French were retained. These past participles include allé, entré, passé, monté, tombé, arrivé, né, venu, sorti, retourné, descendu, resté, parti, and mort. Derivatives of these verbs containing the prefixes de- and re- such as rentré, devenu and revenu were also included in the concordances. Only intransitive uses of these verbs were retained, since their transitive counterparts were conjugated exclusively with avoir.

After concordances were completed, token files were created by coding each instance of the passé composé for characteristics of the speaker (age, gender, community of residence, socioeconomic status, number of generations in the U.S., frequency of French use, and schooling in French) and the linguistic context itself (grammatical person and number, the existence of a transitive homonym, the possibility of an adjectival use of the past participle, and frequency of the main verb). A total of 1,194 tokens were included in the study. These tokens were produced by 57 informants, 25 from Southbridge and 32 from Woonsocket.[7] Our results show that when taken together, Franco-Americans from Southbridge and Woonsocket use avoir with a frequency of 34%.

The binomial multivariate analysis was performed using GOLDVARB 2001.[8] The goal of this analysis was to measure the contributions of each of the social and linguistic factors on the appearance of avoir with the seventeen intransitive verbs mentioned above. Although the data set included six other verbs that are morphologically derived from the original set, these were not retained in the multivariate analysis. The main reason for this is that none were used more than once by informants in conjunction with the passé composé, and thus did not show variation in this corpus. These are repasser, remonter, retomber, ressortir, redescendre, and repartir.

Results from the multivariate analysis were significant at the level of p=0.015. As shown in Table 1, six of the seven external constraints were found to have a significant effect on auxiliary variation. These were frequency of use of French, age, gender, number of generations in the U.S., socioeconomic status, and schooling in French. Only community of residence was not found to be significant. The one and only linguistic constraint found to be significant was the presence of a transitive homonym. Though the data suggests that verb frequency may also play a role in variation (which we will see in subsection 5.2), verb frequency was not chosen as a significant factor in the multivariate analysis.[9] Grammatical person and number and the possibility of an adjectival use were not found to be statistically significant.

Table 1

Contribution of significant factors to the use of avoir (p=0.015)

Contribution of significant factors to the use of avoir (p=0.015)

-> Voir la liste des tableaux

5.1 Significant factors on the use of avoir

Informants from all eight communities studied in the Franco-American French project (including Southbridge and Woonsocket) were divided into five groups based on their self reports on the frequency with which they use French today (see Fox and Smith 2005, 2007). These categories (never, rarely sometimes, often, and daily) were included in the analysis because of the impact of language restriction on choice of auxiliary that has been observed in Franco-Ontarian French when a similar five point scale was used to create an index of language restriction (Nadasdi 2005; Béniak and Mougeon 1989).

It is important to note however, that the categories used in the present analysis come from self-reports on frequency of French use only and do not include an index based on functional language use domains. Studies of Franco-Ontarian French have taken into account 1) a variety of settings (home, community and school) 2) different interlocutors (parents, siblings and peers) and 3) directionality of the interaction (informant > interlocutor or interlocutor > informant) (see Mougeon and Béniak 1991: 71-73). This approach was deemed inappropriate for Franco-American French for several reasons. First, none of the speakers in this corpus are in school. Second, in both communities, French has lost its place as a communicative tool outside the home. Third, the parents of the vast majority of speakers are deceased. Finally, the directionality of use of French is quite individualized from one speaker to the next and is difficult to ascertain in some cases as it was not specifically addressed in the questionnaire.

When frequency of use of French is considered, it appears that infrequent use of French does not favor the spread of avoir as we would expect if language shift were encouraging such a change. In fact, daily use of French is the only factor in this group which favours use of avoir as an auxiliary verb.[10] As shown in Table 1, while use of avoir is prominent among daily users (54%, factor value = 0.616), those who have fewer occasions to speak French are more likely to use être.

In the present analysis, the age groups were 1) below 60 years of age, 2) 60 to 69 years, 3) 70 to 79 years, and 4) 80 years of age or above. Although age was not found to be a significant factor in auxiliary choice in Vermont French, it has been found to be indicative of a change in progress (the spread of avoir) in both Montreal French and Franco-Ontarian French (Russo and Roberts 1999: 80, table 7).

Results of the present analysis show some possible (though not conclusive) evidence of a change in progress. The younger speakers in the corpus (those below the age of seventy) are the most likely to use avoir, where speakers in their seventies, eighties and nineties have a higher probability of using the more conservative être. However, we can see that any possible change in progress has been slowed or even stopped if we examine the younger speakers in the corpus.

First, there is not a single age group that uses either form exclusively with this set of verbs. A considerable amount of variation can be witnessed both in the youngest (56% avoir vs. 44% être) and oldest (25% avoir vs. 75% être) age groups. Clearly, if there is a change in progress, it has not been so rapid as to produce categorical users of avoir even at the youngest end of the continuum. Furthermore, the factor effects for speakers in their sixties (0.669) and for those below the age of sixty (0.668) are nearly identical. This indicates that a possible change in progress among speakers above the age of seventy has not continued among younger informants.

The role of gender as a mechanism of language change has been shown to differ between “healthy” varieties and those that are undergoing language shift. In stable situations of variation in “healthy” language varieties, women typically use more conservative forms than men (Labov 2001). Gender does not typically influence variation in situations of shift because rapid linguistic changes tend to erase the effect of social factors other than age (King 1989; Russo and Roberts 1999; Dorian 1989, 1981).

Let us remember that gender was found to be significant in the choice of auxiliary in the passé composé in Montreal French, where men “seem to use avoir more than women” (Sankoff and Thibault 1980, 340). Based on this and on their other findings, the authors found auxiliary variation to be stable, and not indicative of a change in progress. Our results confirm that Franco-American women are less likely to use avoir than are Franco-American men. In other words, gender based variation has not been eliminated by language shift. The gender effect on the use of avoir is not typical of a change in progress in favor of avoir, but rather of stable variation.

Number of generations in the United States was included as an external constraint for this analysis for two reasons. First, its status as an immigrant language distinguishes Franco-American French from each of the other four major varieties of North American French. Second, in cases of immigrant languages, shift is said to occur across generations (Dorian 1989). Interestingly, migration status was not found to be a significant factor in Vermont French (Russo and Roberts 1999; Russo 1997). As mentioned above however, the Vermont corpus consisted only of first and second-generation speakers (immigrants and children of two immigrants) whereas the Southbridge and Woonsocket corpora include third and fourth-generation Franco-Americans as well.

Results from the present analysis show that the probability of use of avoir does not increase in a parallel fashion with the number of generations in the United States. First-generation speakers who have emigrated from Quebec are the least likely to use avoir (factor weight = 0.265). However, the situation is all the more interesting when we look at generations two and three.[11] While being a member of generation two increases the probability that avoir will be used, generation three speakers are more likely to use être. It would seem then, that while the children of immigrants are quite susceptible to the spread of avoir, generation three demonstrates a substantial shift back toward the use of être.

Auxiliary choice was also found to correlate highly with socioeconomic status. While working class status favors the use of avoir (factor weight = 0.573), the professional class shows a clear preference for être. This is perhaps our most convincing evidence that this particular variable bears a great deal of social meaning in Franco-American French.

Schooling in French is also a significant factor in choice of auxiliary verb. French as a sole medium of instruction strongly discourages the use of avoir (only 8%, factor weight = 0.194). This represents a sharp contrast with respect to all of the other forms of schooling. Most of the speakers who have had French as a sole medium of instruction had this opportunity while residing in Quebec, where French can be used in all communicative domains. Since Quebec French is viewed as a prestige variety by Franco-Americans (see Bagate et al. 2004), it is likely that the experience of living and going to school there has had a lasting effect on the speech of these informants. Although none of the other factors in the group discourage the use of avoir, informants who have had French only as a “foreign” language use avoir with much lower frequency (18%) than those who have had bilingual schooling (42%-46%). If we consider that both French as a sole medium of instruction and French as a foreign language are less favorable to the use of avoir than the categories involving bilingual schooling, our results lend support to the notion that contact with varieties other than the source dialect discourages use of the linguistically proscribed form. Although our data set is too limited to tease out further information about the effects of schooling on auxiliary variation, it is certainly an area that merits future investigation.[12]

According to Russo (1997), Boulanger (1992) has identified transitive homonyms in Quebec French for eight of the seventeen intransitive verbs studied here. These verbs and their transitive meanings and examples as given in Russo (1997: 32-34) are:

Because transitive verbs are conjugated with avoir, it has been found that analogy may play a role in the spread of this auxiliary to the intransitive verbs listed above. The existence of a transitive homonym has been found to favour the use of avoir in both Ontario and Vermont (Canale et al. 1978; Russo and Roberts 1999).

In the present analysis, the existence of a transitive homonym turned out to be one of the most significant factors in choice of auxiliary. It was also the only linguistic constraint found to have a significant influence on variation. Our results highlight a marked discrepancy between those verbs that have a transitive homonym and those that do not. The existence of a transitive homonym strongly favors the spread of avoir to the intransitive forms (75%, factor weight = 0.862). Verbs forms which can only be used in an intransitive manner on the other hand, do not favor the use of avoir (factor weight = 0.415).

5.2 Factors not selected as significant

Four factors in this analysis were not found to have any significant effect on auxiliary variation. Three of the four linguistic constraints were not selected. These were grammatical person and number, the existence of an adjectival homonym, and verb frequency. In contrast, the only external constraint that was not selected is community of residence.

Although Southbridge speakers used the standard être with greater frequency (73%) than did informants from Woonsocket (65%), this difference was not found to be statistically significant. The fact that community of residence was not selected must be interpreted with care. Woonsocket speakers were interviewed in the home, and often told lengthy uninterrupted stories. The majority of Southbridge interviews, on the other hand, took place outside the home (see Stelling 2008, 70-75). It is possible that the relative formality of Southbridge interviews when compared to those conducted in Woonsocket has masked the effect of language shift within the data set. While we may expect the highest use of avoir to coincide with the community (Southbridge) in which language shift is more advanced, we also may expect the use of être to coincide with the community (also Southbridge) where interviews were more formal in nature. This point will be addressed further in the discussion section.

The first linguistic constraint that was not selected as significant was grammatical person and number. Although none of the authors cited in the pages above have studied the effect of this variable on auxiliary variation, it was included here because there is a relationship between grammatical person and number and the frequency with which verb forms appear in conversation. In a discussion of linguistic change, Martinet (1969) concluded that frequently used forms are more resistant to simplification than infrequent forms (cited in Russo and Roberts: 70). This observation is particularly relevant to verb inflection paradigms because singular forms occur with greater frequency than plural forms and third person forms are more frequent than first and second person forms (Croft 2003: 140-142). Accordingly, grammatical person and number were coded into six categories based on the verb form itself (1st sing., 2nd sing., 3rd sing., 1st pl., 2nd pl., and 3rd pl.).[13] There is little evidence that grammatical person and number have much if anything at all to do with variation in this case. For example, second person forms (singular and plural combined) accounted for only 1% (N= 13) of the data set, and similar percentages of avoir usage were found with first person singular (35%) and third person singular (29%) forms.

Another linguistic constraint not selected as significant in this analysis was whether the past participle of a given verb has the possibility of being used as an adjective. Because adjectives are often preceded by être (and never by avoir) we hypothesized that analogy may lead these verbs to be conjugated with être more frequently than verbs without an adjectival use, as was found to be the case in Ontario (Canale et al. 1978). In the two following examples we can see the past participles of mourir and partir used as adjectives:

The possibility of an adjectival use of the past participle was not selected as significant. Although the adjectival uses themselves were eliminated from the data set, main verbs were coded for whether they can also be used as adjectives. The analysis showed that the percentage of avoir remains nearly identical regardless of whether the past participle may be used as an adjective (66%) or not (67%).

The final factor that was not selected as significant is the frequency of the main verb. Because frequent forms are more resistant to change than infrequent ones (see above), we originally hypothesized that highly frequent verbs may be conjugated most often with être. Following the lead of researchers in Ontario and Vermont (Canale et al. 1978; Russo 1997), we made a distinction between high and low frequency verbs based on frequency indexes given in Juilland et al. (1970).[14]

In the studies of auxiliary variation in Franco-Ontarian and Vermont French, verb frequency was found to be an important factor. In both cases, low frequency verbs encouraged the spread of avoir (Canale et al. 1978; Russo and Roberts 1999). However, the verbs naître and mourir were not included in either study. Naître is a low frequency verb that nonetheless showed an extremely high degree of resistance to the spread of avoir in the present analysis. Avoir was used in only 3% of tokens containing naître. It is important to consider that despite its low frequency in French, naître is not conjugated like other infrequent re- verbs, a fact which suggests the verb acts quite differently with respect to linguistic change or lack thereof. It is also important to note that naître was contained in the passé composé in several questions at the beginning of the vast majority of the interviews. Because the interviewers used être as an auxiliary with this verb, this may have influenced the forms used by informants.

If we eliminate naître and each of the five verbs that produced a very low number of tokens (descendre, monter, passer, entrer and tomber; N < 20) from the data set, it becomes possible to examine the possible relationship between verb frequency and auxiliary choice. Avoir is used with low frequency verbs in the majority of cases. Avoir use with these verbs ranged from 66% (retourner) to 90% (sortir). With high frequency verbs however, avoir use ranged from only 2% (aller) to 68% (partir). It is also important to note that although the verbs monter and descendre are not included in the table because of low token counts (N= 9 and 5 respectively), they are both low frequency verbs which each showed 100% avoir application. All in all, it appears that high-frequency verbs are more resistant to the spread of avoir than infrequent ones, a fact which mirrors the findings of the Vermont and Ontario studies mentioned above.

6. Discussion

The multivariate analysis has shown that although a great deal of variation exists within the communities, language shift does not appear to be causing rapid linguistic change. In fact, Franco-Americans from Southbridge and Woonsocket use avoir with exactly the same frequency (34%) as has been reported for speakers of Quebec French and unrestricted speakers of French in Ontario (Nadasdi 2005). Additionally, generation three speakers from Southbridge and Woonsocket use avoir only 13% of the time. If language shift were accelerating grammatical change, we would expect to find a considerably higher frequency of avoir application within the data set, especially among third generation speakers.

Structural convergence with English would promote the spread of avoir because ‘to have’ is the sole auxiliary verb of the English present perfect. The fact that this development has not progressed in the speech of informants from Southbridge and Woonsocket suggests that convergence with English is not occurring in the passé composé.

In situations of language shift, the expected tendency is the development of a simplified linguistic system with a strong a preference for transparency, or a one-to-one relationship between form and meaning (Bradley and Bradley 2002; Jones 1998; Dorian 1989). Given that two auxiliaries are not functionally necessary, it is in fact quite remarkable that both have been retained, even in Southbridge where the shift to English is quite advanced.

Contrary to what has been reported for Cajun French (see above), where use of être with intransitive verbs of motion among a range of fluent speakers aged 30 and older is found only among the oldest (70 years old and above) both auxiliaries are widely used among informants of all ages in Southbridge and Woonsocket. Furthermore, there are no categorical users of avoir in our corpora and there are virtually no differences observed between speakers in their sixties and those below the age of sixty. Although French is being used less and less in Louisiana and New England alike, its linguistic structure remains considerably more intact in the passé composé in Franco-American than in Cajun French.

Another important finding related to the lack of structural simplification in Southbridge and Woonsocket is that social factors continue to exercise a significant influence on variation. With the exception of community of residence, every social constraint considered was found to be significant in this respect. This also runs counter to the expected pattern whereby accelerated grammatical change during shift effaces the social meaning behind variability. Results from the multivariate analysis validated the association made by Sisters Marie-Francia and Marie-Anaïs (1958) between the use of avoir and working class status. Contrary to the finding of Locke (1949), who stated (based on his observations of one working class informant) that avoir is the standard auxiliary for intransitive verbs of motion, the present analysis indicates that être is still widely used with these verbs by Franco-Americans from Southbridge and Woonsocket, especially among those belonging to the professional class.

We hypothesized in Section 4.2 that community of residence may not have been selected as a significant factor because the effects of formality and language shift could potentially counterbalance one another in the data. The interviews were relatively more formal in Southbridge, where the shift to English is more advanced than in Woonsocket. However, given the results of the analysis, which suggest that language shift is not favoring the spread of avoir, it does not appear to be the case that the effects of formality and language shift are counteracting one another within the data set. Rather, it is more likely that community of residence was not a significant factor because community norms have remained remarkably similar in the two locations (once all other intervening factors are accounted for), despite the differing degrees to which language shift has advanced.

Results from the present study differ from the findings of Russo and Roberts (1999) in Vermont, who reported that the usage of avoir was spreading and that avoir would replace être and become the sole auxiliary of intransitive verbs of motion. However, the Vermont study included only first and second-generation speakers and a total of only four highly restricted and three unrestricted users of French. The remaining fifteen subjects were all labeled as semi-restricted French speakers. The tendency to use être observed among third-generation speakers and among those who do not use French on a daily basis in Southbridge and Woonsocket indicates that these two limitations of the Vermont corpus are most likely responsible for the differing results of the two studies.

When interpreting the results of the present study, it is important to consider that French is not a typical minority language. It is a language of wide international currency. Some informants have had the chance to use their French outside of the communities themselves through travel and through access to French language media sources. Additionally, nearly all of the informants have had some education in French, and many of these speakers have been exposed to other varieties of the language in the classroom.

When taken as a whole, the findings of the present analysis indicate that despite the sociolinguistic situation of language shift, the passé composé shows a high degree of resistance to grammatical simplification in both Southbridge and Woonsocket. These findings echo those of Fox (2005), who stated that the syntax of Woonsocket French appears to be relatively robust and shows resistance to certain radical linguistic changes. Franco-American provides strong evidence that in association with the right set of sociolinguistic factors such as education in the mother tongue, exposure to other varieties of the language, socioeconomic stratification, and linguistic conservatism, linguistic proscription may slow or even block certain linguistic developments during situations of shift.

7. Conclusion

In the pages above, we have seen that linguistic proscription has had an impact on the linguistic consequences of language shift in Southbridge and Woonsocket. Although shift to English is advancing in both locations, the analysis provided no convincing evidence that the passé composé is undergoing structural simplification or convergence with English. If language shift were causing radical structural simplification, we would expect to find a higher rate of avoir application within the data set and that social factors do not influence variation. To the contrary, we have observed that auxiliary variation does hold a great deal of social meaning. Among the social factors considered, only community of residence was not found to be significant, a fact which suggests that community speech norms involving past tense formation are remarkably similar in the two locations once all other factors are accounted for.