2.1 The social geography of Lille

The 2004–2005 data are derived from the same corpus as was used for the investigation of code-crossing (Pooley and Mostefai-Hampshire, Reference Pooley and Mostefai-Hampshire2012), as well as variation in the use of French (Pooley, Reference Pooley, Beeching, Armstrong and Gadet2009) in the urban heartlands of Lille, i.e. the central areas of Lille-Métropole and the north-eastern corner around Roubaix, Tourcoing and Wattrelos (Figure 1). The same areas, and in two cases the same schools, were also investigated during the 1995–1999 fieldwork, which was designed to explore firstly, variation in the use of French and secondly, one other part of the subjects’ linguistic repertoire, Picard. From 1998 onwards, it was decided to expand the geographical coverage beyond the urban heartlands into periurban (Bois-Grenier, Cysoing, Sailly-Lez-Lannoy) and rural areas (Cappelle-en-Pévèle, Hantay) within the whole of Lille-Métropole (see Pooley, Reference Pooley2004, Chapter 5 for a full discussion) (Figure 1) (Table 1), but only to carry out the section of the interview which focused on Picard (Figure 1, Table 1).

Table 1.

Ethnicity and geographical distribution of informants in 1995–1999 fieldwork locations (* indicates location also investigated in 2004–2005)

MF = Metropolitan (European) French parentage; OTHER = at least one parent not of MF background

Figure 1.

Location of fieldsites within Lille-Métropole (1995–1999 and 2004–2005)

As Table 1 indicates, in the 1995–1999 study 172 school students were interviewed, fairly evenly divided with regard to gender (82 boys, 90 girls: 48-52%) (compared to 57 subjects, 32 boys, 25 girls: 56-44% in 2004–2005). Most urban informantsFootnote ⁸ (and all those interviewed in the 2004–2005 study) were in V^e or IV^e (13–15)Footnote ⁹ whereas all respondents from the outlying periurban and rural areas were of primary age.Footnote ¹⁰ Here geographical location was allowed to take precedence over age, since the secondary-age sections were located in urban areas, and many of the outlying communes offered only primary education. Location of school attendance clearly cross-cut with other potentially important factors: such as social class, ethnicity and rootedness (or intergenerational continuity of residence). Most of the subjects of secondary age were from SEGPA (Section d'Education Générale et Professionnelle Adaptée) sections where pupils deemed unable to cope with the mainstream curriculum, whether for lack of academic ability, troubled family lives or recent immigration, were placed so that they could be given additional support in basic subjects such as French and mathematics as well as the opportunity to acquire more practical skills. These latter parts of the curriculum were often taught in situations of marked gender segregation with girls being oriented towards such activities as laundry and kitchen work and boys towards, for instance, painting and building. From a social-class perspective, the SEGPA respondents may all be placed into Category 4 (the lowest) of the Ministry of Education's scale of favourability to educational success http.www.education.gouv.fr/ival Footnote ¹¹ and, according to Chauvel's (Reference Chauvel2005: 28) social categorisation, be shared between the working class and underclass.

Table 2.

Ethnicity and geographical distribution of informants in 2004–2005 fieldwork locations

MF = Metropolitan French parentage; OTHER = at least one parent not of MF background

Informants from mainstream school sections and those in periurban and rural areas came from a much wider social spectrum (Categories 1–4 according the scale of favourability to educational success or in Chauvel's categorisation covering working-class, median-class and middle-class). Location of school attendance also correlated strongly with ethnicity. All rural and all but two periurban respondents were of European or Metropolitan FrenchFootnote ¹² (henceforth MF) background, i.e. were born of parents of European descent who were themselves born and brought up in Metropolitan France and testified to no sociolinguistically significant factors, such as use of a language other than French at home or residence outside France. In the 1995–1999 data, 24/86 of subjects (28%) from the urban heartlands had at least one parent born outside France with a first language other than French. Two thirds of these (16/24) were of Maghrebian extraction. The proportions in the later study are broadly similar: 17/57 respondents (29.8%) had at least one non-MF parent with around two thirds of those being of Maghrebian background (65%). With regard to intergenerational continuity within the commune of residence, this proved to be greater among MF subjects than their non-MF counterparts in urban areas. In periurban areas (and here only MF subjects are concerned), mobility was higher than in both urban and rural areas and usually caused disruption of local intergenerational continuity. Two locations, the SEGPAs in La Madeleine and Lys-Lez-Lannoy, feature in both studies and thus permit some degree of direct comparison. The school in Lys-Lez-Lannoy is attended by large numbers of pupils from two recognised social-problem areas, officially designated as zones urbaines sensibles (ZUS) in Roubaix, Le Pile and Trois-Points (marked P and T respectively in Figure 1).

In the 2004–2005 fieldwork, another ZUS was chosen for investigation, Lille-Sud (marked LS in Figure 1), rather than Rouges-Barres (a working-class area in the generally well-to-do town of Marcq-en-Barœul that fills much of the space between Lille and Roubaix-Tourcoing), where the then incumbent head declined my request to undertake a follow-up to the initial 1995 study (reported in Pooley, Reference Pooley1996; Reference Pooley2000, with regard to French and Pooley, Reference Pooley2004, with regard to both French and Picard). The third school location, studied in both sets of fieldwork, was in La Madeleine, a central commune which developed in the industrial boom of the late 19th century, but not classified as a ZUS. During the 2004–2005 investigation, in Lille-Sud and La Madeleine, every member of the V^e and IV^e classes was interviewed and in Lys-Lez-Lannoy, every member of the IV^e.

2.2 Verbal repertoire, social variables and linguistic perceptions

Investigation of other aspects of the informants’ verbal repertoires clearly demonstrated that these correspond to various aspects of their social identity. Firstly, variable use of the regionally marked French variants, back a as in [ɑ] in ça and open o in checked syllables as in [ʀɔz] rose, have been shown to correlate strongly with differences of gender and ethnicity, e.g. back a is most used by MF boys. In the Rouges-Barres data (recorded in 1995), use of back a also correlated strongly with a Regional Loyalty Index (Appendix 2) based on three questions concerned with attitudes to the Nord region (1) and its people (2) and the subjects’ choice of an ideal region of residence in adulthood (3). Use of the same variant was also found to correlate in a later study (Pooley, Reference Pooley, Beeching, Armstrong and Gadet2009) with the adolescents’ popularity among their peers as measured by an Index of Popularity in Class which was based on the number of mentions each individual received from the other members of the class when they were asked to name their best friends.

Secondly, previous results from the same interviews (Pooley and Mostefai-Hampshire, Reference Pooley and Mostefai-Hampshire2012) indicated that MF boys, and to some degree those from Roubaix-Tourcoing rather than the central areas, showed greater knowledge of Arabic than girls of the same ethnic and residential background. More importantly, however, it was again friendship factors, in this case MF boys having Maghrebian friends both in and out of school, which correlated with significant differences in competence in Arabic. Although all the social factors so far mentioned were tested with regard to results of the Picard Skills Tests, only ethnicity proved to be significant (see Table 6). This should not be taken to imply any observed spontaneous use among peers, indicated by differences corresponding to gender and network membership. The 1995–1999 data, in particular, provided the opportunity to investigate a range of other possible factors, including age, social class and area of school attendance/residence. Perhaps more important (and this applies to all respondents) is family background, and in particular whether subjects heard the language spoken in their extended family circle, given that, as already observed, no informant had inherited Picard as a first language from her/his parents. Other indices used with both corpora give some perspective on how subjects perceive Picard: firstly, the Cultural Value Index gives a numerical value to the perception of the distinctiveness of Picard as a separate language and the desirability of knowing it (Appendix 3); secondly, respondents were also asked to assess their competence in Picard, with regard to understanding and speaking (Professed Competence Index, Appendix 4) and whether, given the opportunity, they would take classes to learn the language. Respondents were also asked whether they had any contact with Picard through various media, such as television, radio, the printed page or live performances, testing whether the cultural events described in the introduction had had any discernible effect. Where appropriate, χ² tests were used to evaluate statistical significance, but subjects with relatively high or particularly low test scores were profiled individually in the hope that common factors would become discernible.

3.1. The Picard skills test

As already mentioned, respondents’ knowledge of Picard was assessed by means of testing using a three-part procedure, called the Picard Skills Test, as follows: a) ‘translation’ of ten Picard items into French; b) the ‘translation’ of ten French words into PicardFootnote ¹³ and c) a Speaking Test in which subjects were invited to say anything at all in the language.

Results of part a) are reported in Table 3. In this section, respondents were asked whether they could show understanding of ten Picard items, whether by translation or other means. Over the two field studies, four categories of items were used: a) items recognisable through regular sound correspondences with French, e.g. kien ‘chien’ (Picard [k] for French [ʃ]. b) Picard items not used in regional varieties of French such as quer ‘chercher, tomber’ and whose form is not predictable from regular sound correspondences with French; c) items which may be used in French without modification such as matte ‘fatigué’ and d) in the 1995–1999 study two (standard) French items deemed to be more frequent in regional and dialectal speech. These were replaced in the later round of fieldwork.

Table 3.

Categories of distinctiveness of items used in Picard Skills Test (Part a, Picard-French) (Items verified in Legrand (Reference Legrand1853); Vermesse (Reference Vermesse1867); Viez (Reference Viez1910); Carton and Poulet (Reference Carton and Poulet1991; Reference Carton and Poulet2006); Blanquaert and Blanquaert (Reference Blanquaert and Blanquaert1998)

Percentages indicate that in 1995–99 x% of 172 answers and in 2004–05 x% of 57 answers were deemed adequate or acceptable.

For part b) of the Picard Skills Test, respondents were asked to give Picard equivalents of ten French items. Given the items chosen this was arguably as much a test of awareness of sound correspondences as knowledge of (clearly differentiated) vocabulary (Table 4).

Table 4.

The Picard Skills Test Part b) French-Picard

Percentages indicate that in 1995–99 x% of 172 answers and in 2004–05 x% of 57 answers were deemed adequate or acceptable.

In the 2004–2005 questionnaire, five such variables were tested [s]-[ʃ] i.e. Picard [ʃa] or [ʃɑ] for French ça and [iʃi] for ici; []-[] [mʒe] manger, [mtø] menteur; [ʃ]-[k] [vak] vache [kval] cheval; [wa]-[o] [bo] bois n, [votyʀ] voiture and [ʒ] -[ɡ] [ɡaʀd] jardin ^. In the 1995–1999 field study, awareness of metathesis in the re- prefix in Picard was tested with the item revenir [aʀvniʀ] leaving tabac [tuvak], carette (alternative for voiture) and maintenant [aʃtœʀ] or [astœʀ] as differentiated items. Parts a) and b) were graded on the basis of one point for each acceptable answer, bearing in mind that some latitude was allowed in the scoring. For manger and menteur the points were awarded for appropriate use of the []-[] correspondence making responses such as [mʒ] or [mʃ] and [mtøʀ] or [mtœʀ] acceptable, as well as the more marked r-less form. While the [ʃ]-[k] correspondence was illustrated by vak and kval, a number of forms may be used for the latter item, such as [ɡva], [ɡval], [ɡvɑ], [kva] and [kvɑ], all of which were accepted. For voiture, carette is the most authentic answer, whereas for bois mere use of back a [bwɑ] was deemed to be a French form. Three items ici, vache and jardin were introduced in place of revenir, blanchi and maintenant. The omission of revenir meant that awareness of the metathesis of the re- prefix was not tested ([aʀvniʀ]). The item blanchi caused considerable difficulty since the presence of [] in the French form could give rise to responses such as the ‘incorrect’ [blʃi] or the hyperdialectal [blki]. The form [aʃtœʀ] or [astœʀ] for maintenant although quite frequently noted by observers of Oïl varieties proved to be virtually unknown to the subjects questioned.

In Part c) (the Speaking Test), informants are asked to say something (literally anything) in patois and therefore have the possibility of displaying additional knowledge of lexis and phonology but also grammar without the constraint of trying to recall instantaneously certain set items. Acceptable answers to part c) were awarded three, five or ten points out of a possible ten. Three points were awarded for isolated lexical items, five for grammatically linked sequences of up to five morphemes in Picard and ten for longer sequences (See Appendix 5 for examples). Respondents were thus awarded a total score out of 30 which was converted to a percentage.

4.1 The Variety Labelling Test

As noted in the introductory section, subjects were asked to listen to a series of short recorded texts and classify them according to a typology offering four possibilities: a) French; b) mixture (mixed code); c) patois; d) other (in French autre chose). In the first round of fieldwork carried out between 1995 and 1999, respondents were given six such texts to listen to, five of which could be claimed to be Picard and the sixth containing a clear French-Picard switch. All the speakers on the recordings were from the Nord–Pas-de-Calais, except for one from the Belgian Hainaut. Four of the six extracts were retained for the significantly augmented version of the test administered in 2004 and 2005. The six Picard recordings were increased to thirteen, covering most of the traditional picardophone area (Figure 2), as set out in Table 11.

Table 11.

Density of Picard features in Picard extracts (Extracts marked with an asterisk were also used in 1995–1999)

Figure 2.

The Picardophone area with indication of geographical origin of extracts in Variety Labelling Test

The major sources were Debrie (Reference Debrie1983) and Carton (Reference Carton1972) augmented by two clips collected from patois association meetings and one of my own recordings. Table 11 lists the 13 extracts used in 2004–2005 graded into three sets (low-medium-high) according to the density of Picard features measured by occurrences of any of the 19 features listed in Table 12. The clips being between 59 and 126 words long, the measure of density of Picard features was derived by counting the number of occurrences of any of the nineteen items and calculating the equivalent rate per 100 words. For instance, in the case of the Tourcoing extract, 22 occurrences of any the features were noted in a sequence of 73 words, giving a rate of 28 features per 100 words. In addition to the increased number of examples, two distractors, short clips from teaching materials in Romanian (Deletant, Reference Deletant1983) and Basque (taken from London Metropolitan teaching materials) were included to give genuine choices for the ‘autre chose’ option. At the other end of the spectrum, the speaker in the Roubaix extract is speaking French but quoting a past conversation in which she used some Picard features (hence density score of 8 features per 100 words).

Table 12.

List of Picard features used for evaluating 2004–2005 data (* = also used in 1995–1999)

The Picard extracts selected for the test were spoken by people from all five picardophone départements plus the three largest towns of Lille-Métropole and two locations in Belgium, exemplifying different degrees of picardité. The use of 19 features constitutes a considerable increase compared to the eight used for 1995–1999 (Pooley, Reference Pooley2004: 444), partly warranted by the greater geographical coverage of the more recent study. On the basis of the measure of density, the clips clustered fairly readily into three sets, which have been labelled as follows: a) low density (up 8 tokens per 100 words); b) medium density (26–36 tokens per 100 words); c) high density (47 or more tokens per 100 words). A secondary measure of dialectal density based on occurrences of morphological and lexical features was also calculated but this did not yield as consistent results as the measure of density based on the 19 features.

4.2 How the test corpus was perceived

While distinctiveness of grammar and lexis are the traditional criteria for distinguishing dialects, the subjects’ responses suggest quite strongly that only the latter is of consequence in their perception. At least two explanations may be put forward for the absence of grammatical considerations in the metalinguistic comments. Firstly, as Léonard (Reference Léonard and Eloy1998) observes with regard to another Oïl variety, the subjects may well believe that Picard (or Poitevin) has no grammar in the sense of one that is written down and can be studied, as is the case for ‘une vraie langue’ such as French; secondly, grammatical features may be manifested through a single sound difference, resulting in a different word, such that [avo] is perceived, for instance, as a different word from avait, although the alternation depends on a single sound segment. When asked how they recognised patois, subjects often had recourse to explanations in terms of words but more frequently they referred to sounds (and occasionally letters).

As regards lexical differences, opinions were varied: ‘Ils n'emploient pas les mêmes mots que nous en français’, ‘Ils disent des mots qui ressemblent au français, mais c’est pas pareil en fait’ or ‘Ils disent les mêmes mots, mais ils ont un accent’. Many subjects referred to differences of pronunciation from French, sometimes expressed in moderate terms such as: ‘Eux quand ils parlent français ils ont un petit accent’ or conveying an air of strangeness ‘une drôle de voix’, or even some of the oft-quoted perceptions regarding Picard or Nordiste pronunciation such as ‘un accent grave’ or ‘ils écrasent les mots’, both expressions often used to refer to the backing of a, which in some clips was velarised and markedly rounded (e.g. Amiénois). The comments also contain a strong thread of disapproval, which may be expressed either as distorting words: ‘Ça s'entend, ils déforment les mots’, ‘ils ont un accent et ils déforment presque tous les mots’ or even of speaking like foreigners trying to communicate in French. The language exemplified by the clips is often perceived as somehow incorrect and unlike the informants’ own way of speaking, which, in their own perception, emerges as some kind of reference: ‘ils parlent pas comme nous, comme les Français’ or ‘ils parlent pas comme tout le monde’ or ‘ils parlent pas “normal”’ or ‘j’aime pas cette façon de parler’. Only a small minority of respondents recognised the clips for what they were by relating them to the way that people of a certain generation or from a certain area speak or by quoting something in patois.

As Table 13 indicates, it is the two foreign languages, Basque and Romanian, which are perceived as ‘other’ with the greatest degree of consensus, but the Picard of the Belgian Hainaut, judged by over 90% of informants in the 1995–1999 study as ‘other’, still achieves a level of agreement far above any other variety spoken in picardophone territory (75%). Curiously, the other Belgian variety was also widely perceived (38.6%), albeit without a real consensus, as ‘other’. The only other majority verdict in the later study was that 59.6% of subjects deemed the Roubaix variety to be French. The results are thus largely characterised by relative lack of certainty, in several cases little better than purely chance guessing, particularly with regard to mixed code and patois, since no extract achieved a majority verdict with these labels.

Table 13.

Informants’ classification of extracts

Taking the categorisations that respondents were asked to use and those derived by calculating the density of Picard features, such that French corresponds to low density, mixed code to medium density and patois to high density, the perceptions presented in Table 13 may be compared to the linguistic classification of Table 12. This is shown in Table 14. The results, unlike those of Kerswill (Reference Kerswill, Preston and Long2002), are marked largely by a lack of correspondence between the two evaluations, with only 6/15 clips being placed in the corresponding slots, including the two foreign languages. Of the four ‘Picard’ clips on which there is agreement, three are clearly towards the French end of the dialectality spectrum (Roubaix, Beauvaisis, Santerre) leaving only Vermandois in the middle range. There seems to be no clear pattern in how medium-density clips are perceived, even if morphological and lexical criteria are taken into account. One of the most striking results is the perceived otherness of the Belgian varieties, although on the basis of the linguistic features they do not fall into the high-density category. In the case of the Hainaut extract, this finding is consistent across the two data sets (Tables 13 and 15) although perhaps attenuated in the later study by the presence of the more obviously ‘other’ varieties. The perception of the Mouscron clip is all the more surprising given the proximity of Mouscron to Lille and the well-documented similarity of the traditional varieties (cf. Legrand, Reference Legrand1853; Blanquaert and Blanquaert, Reference Blanquaert and Blanquaert1998; Maes, Reference Maes1942).

Table 14.

Comparison of categorisations according to linguistic criteria (Table 12) and informant perception (Table 13)

Table 15.

Comparison of perceptions of four extracts used in 1995–1999 and 2004–2005

Separating the Variety Labelling Tests results of subjects considered to be high (50%+) or low (6.7% or less) in the Picard Skills Tests failed to reveal significant differences nor indeed any sign of a possible patterning closer to the linguistic analysis.

4.3 Indicators of the real-time trends

With regard to subject perception, the degree of uncertainty manifested in 2004–2005 is considerably greater than in the 1995–1999 study, as indicated by Table 15, where the clear majority verdicts regarding the Libercourt, Saint-Polois and Cambrésis varieties have dropped to most frequent but minority answers in the later study. Apart from the Hainaut clip (‘other’) only the Roubaix extract (French) receives a clear majority verdict (Table 13).

In the light of the comments quoted earlier about strange forms of French and the recurrent theme of condemnation and indeed puzzlement pervading them, these results seem to add credence to the hypothesis that Picard is continuing to recede from the perceptual landscape. The filter of French seems to be stronger, hence the greater frequency of responses along the lines of: ‘Il y a du français et puis autre chose’. Arguably it has also become increasingly problematical to state what this ‘autre chose’ might consist of.

Secondly, the proportion of subjects labelling the varieties used in both studies according to the most frequently chosen category categorising the four varieties is markedly lower. This discrepancy in perception may be due in part to the higher proportion of MF subjects (85%) in 1995–1999 compared to 70% in 2004–2005, although the latter proved less inclined to give a majority verdict. One of the clearest findings of the earlier study was that non-MF subjects were significantly more likely than their MF counterparts to construe these varieties as French (Pooley, Reference Pooley2004a: 438ff). Whether the 2004–2005 produced a similar result is tested in the tabulation presented in Table 16, which shows respondents’ categorisation of ten of the clips. For all ten recordings, MF subjects offered the answer ‘patois’ more frequently than their non-MF classmates, whereas the latter responded more frequently with French in every case except that of Libercourt, where scores for French were virtually equal. Moreover, their greater use of the ‘mixed code’ response (as in the case of Libercourt) also hints at a stronger tendency to interpret the data through the filter of French and to avoid the problematical choice between ‘patois’ and ‘other’.

Table 16.

Comparison of perceptions of ten extracts (2004–2005) by subjects of Metropolitan French (MF) and other ethnic (non-MF) backgrounds

Comparing the perceptions of the varieties used in both data sets, Tables 16 and 17 lend a degree of confirmation to the finding that non-MF subjects are more likely to interpret the data as French, whereas MF subjects seem less confident about labelling varieties as patois.

Table 17.

Results of Variety Labelling Test for MF and non-MF subjects for clips used in both studies