The phonological characteristics of French in Bamako, Mali. A sociolinguistic approach

Com pan y chapter 4 he phonological characteristics of French in Bamako, Mali Chantal Lyche and Ingse Skattum Introduction jam 1. ins Pub University of Oslo lish ing A sociolinguistic approach rec ted pro ofs - J ohn Ben In this chapter we will examine the variety of French spoken in Mali’s capital Bamako. hough French in Mali is no-one’s irst language (L), it is the country’s oicial language and main language of instruction. his status as second language (L2) (in the sense Cuq 99 gives this term),2 distinguishes it from other foreign languages (cf. the dichotomy français langue seconde (FLS) / français langue étrangère (FLE)). Bamako is, like most African capitals, an ethnic and linguistic melting pot (Calvet 994), where the oicial language coexists with a number of regional and local languages. Our corpus relects this multilingualism, with ive Ls represented. An examination of the phonological characteristics of French in Bamako must therefore take into consideration the possible impact of L on our informants’ French pronunciation. Certain extralinguistic factors, which are particularly important given the sociolinguistic context of our study (see Skattum forthc.), will be taken into account as well. Special attention will be given to 1. We are grateful to Guri Bordal and Doug Walker for their comments on an earlier version of the paper. Un cor . “Le français langue seconde […] se distingue des autres langues étrangères éventuellement présentes sur ces aires par ses valeurs statutaires, soit juridiquement soit socialement, soit les deux et par le degré d’appropriation que la communauté qui l’utilise s’est octroyé ou revendique. Cette communauté est bi-ou plurilingue. La plupart de ses membres le sont aussi et le français joue dans leur développement psychologique, cognitif et informatif, conjointement avec une ou plusieurs autres langues, un rôle privilégié” (Cuq 99: 39). Com pan y Chantal Lyche and Ingse Skattum lish ing deviations from le français de référence (FR) (see Chapter ) that correspond to regional or pan African tendencies. We will irst give a brief description of the country (Section 2), followed by an outline of the sociolinguistic situation (Section 3) and the methods of data collection (Section 4). he phonemic inventory is studied in Section 5, followed in Sections 6 and 7 by an analysis of schwa and liaison respectively. Our conclusion (Section 8) will return to some of the issues sketched above. Pub . Mali rec ted pro ofs - J ohn Ben jam ins Mali is a landlocked country bordering the Ivory Coast and Guinea to the south, Senegal to the west, Mauritania and Algeria to the north, and Niger and Burkina Faso to the east. It is one of Africa’s largest countries, but with only 3.76% of arable land, the Sahara desert occupying the north and the semi-desert Sahel the center. Most of the approximately 3 million inhabitants live in the Sudanese savannah in the south and along the two main rivers, the Niger and the Senegal. he urban population (35.9%) mainly lives in the capital Bamako, in the southern part of the country, but also in regional cities like Segu, Sikasso and Koutiala in the south, Kayes in the west, Mopti and Djenne in the center, and Timbuktu and Gao in the north. he country was colonized by France between 880-95 and was proclaimed an independent republic on September 22, 960. It changed its name from Soudan français (given by the French in 892) to Mali (West Africa’s most prestigious medieval empire). Since the 992 elections, it has been considered one of the most democratic states south of the Sahara. Mali is, however, also one of the poorest countries in the world, coming 60th of 69 countries in the Human Development Indicator (HDI) of the United Nations Development Program (UNDP), which takes into account standard of living, life expectancy and literacy. Statistics show that the Gross Domestic Product (GDP) per capita is $649, life expectancy 48.8 years and the literacy rate 26% (igures drawn from L’état de l’Afrique 2011). 3. French, Bambara and national languages in Mali cor Like most former French and Belgian colonies, Mali kept the colonial language as its oicial language ater independence. However, because of its landlocked situation, Mali’s contact with the French colonizers and their language was later (end of the 9th century) than in the coastal states (from the mid 7th century). Un 74 Com pan y he phonological characteristics of French in Bamako, Mali rec ted pro ofs - J ohn Ben jam ins Pub lish ing he French were also less numerous in the hinterland than on the coast. Mali is, in fact, the least “francophone” state south of the Sahara according to Rossillon (995). In his Atlas de la langue française, he estimates that Mali has 5% of “real” and 5% of “potential” speakers, the total (0%) placing Mali as the very last of the 7 francophone African countries. Even this low igure may be too optimistic, based as it is on education statistics (the level of competence being hard to ascertain otherwise): six years of schooling or more gives a “real” speaker while a minimum of two years gives a “potential” speaker. It is clear, however, that six years is no guarantee of “real” competence (see Skattum 2004; Boutin, Gess & Gueye this vol.). It is also well known that this competence is easily lost for lack of practice (Dumestre 994b: 3). An important reason for the modest use of French in Mali is the existence of an indigenous lingua franca (Bambara), conining French to the formal sphere. his is a characteristic Mali shares with a handful of other African states (Skattum 997: 79–80), inter alia the two represented in this volume, Senegal (Wolof) and the Central African Republic (Sango). In Mali, Bambara is the L of around 40% and an interethnic means of communication of another 40% of the population. Mali has a multiethnic and multilingual past that further motivates the low corpus and high status of the French language (in Chaudenson’s sense of the term).3 he medieval empires Ghana, Mali and Gao, uniting diferent ethnic groups from the eighth to the sixteenth centuries, fostered pride in traditional culture and laid the ground for a general acceptance of multilingualism (relected in the 992 Constitution, giving all indigenous languages equal rights) (see Skattum 2008). his may explain the reluctance to recognize the majority language as the oicial language alongside French: Bambara is given no oicial privilege, though in real life it is steadily progressing in terms of geographical extension and usage domains, at the expense of the other indigenous languages (Dumestre 2003). he positive attitude towards multilingualism is also expressed through the prominence given to bilingual education (Skattum 997, 2000). Mali is a pioneer in this ield among the francophone sub-Saharan countries,4 having introduced  indigenous languages as means of instruction alongside French in primary cor 3. “‘[S]tatus’ (et non ‘statut’) désigne le ‘statut’ (langue oicielle, unique ou non, nationale, etc.), les emplois (oiciel, administratif, juridique, etc.) et les fonctions du français. Sous la rubrique ‘corpus’ sont regroupés quatre ensembles : les modes d’appropriation, la véhicularisation et/ou vernacularisation, les types de compétences, les productions et consommation langagières” (Chaudenson 99, cited in Chaudenson et al. 993: 9). Un 4. African language literacy is much more developed in the “anglophone” African countries, as mother tongue education in the irst years of primary school was introduced already under colonial rule (see Brock-Utne & Skattum 2009). 75 Com pan y Chantal Lyche and Ingse Skattum J ohn Ben jam ins Pub lish ing school. However, this education seems lately to be regressing: introduced in 32.6% of primary schools in 2005–2006 and programmed to develop further, only 2.% of the total number of primary schools were bilingual French / a national language in 2008–2009 (see Skattum 200b). French thus continues to dominate as the language of instruction, indicating that it may hold its position as the language of prestige and social promotion despite its limited use among the population. However, a look at some igures illustrates the problems facing Mali in efforts to raise the level of literacy and thus French language proiciency among its inhabitants. In 990 the GER (Gross Enrollment Rate)5 for primary school was the lowest in francophone West Africa, and it remained the lowest in 2003–2004 (Banque Mondiale 2007: 26–27). In 2008–2009 it had improved (82% according to the Cellule de Planiication et de Statistiques)6 but as the GER includes repeating students but does not count dropouts (both high), the real picture is better shown by the retention rates (enrollment at the superior level) from primary to higher education. In 2003–2004, these rates dropped from 69.0% (7–2 years), to 35% (3–5 years), to 0% (6–8 years, including technical and professional education plus teachers’ colleges). In higher education (where age varies so that the GER is calculated relative to 00,000 inhabitants), there were only 286 students per 00,000 inhabitants (Banque Mondiale 2007: 26). Besides Bambara, there are about 20 indigenous languages in Mali (Canut & Dumestre 993: 220),7 a moderate number in the African context.8 3 of these have been given the status of “national language”, ten of them as early as 967: Bambara (bamanankan),9 Bomu (bwamu, bobo), Bozo, Dogon, Fulfulde (peul), Mamara (minyanka), Syenara (sénoufo), Songhay, Soninke and Tamachek, to which were added in 996, nearly 30 years later, Hassaniyya (maure), Maninka rec ted pro ofs - 5. he World Bank deines GER as “Les efectifs scolarisés par rapport aux efectifs scolarisables pour chaque niveau d’enseignement, calculé sur l’âge (7–2 ans; 3–5 ans; 6–8 ans)” (Banque Mondiale 2007: 26). 6. Personal communication, 02/7/200, from Youssouf Haïdara of the Direction nationale de l’Education de base. 7. he Summer Institute of Linguistics (SIL) counts 50 indigenous languages in Mali (Ethnologue 2005: 4). he diference is due to diverging methods of distinguishing language from dialect. cor . he Ivory Coast, for example, is a “medium” case with around 60 languages, while Cameroon has around 85, and the Democratic Republic of Congo around 220 indigenous languages (Skattum 997: 79; see also Brock-Utne & Skattum 2009). . Names and spelling vary; we use the most common English designations and give French or local names in parentheses where there may be doubt whether we are speaking of the same language. Un 76 Com pan y he phonological characteristics of French in Bamako, Mali rec ted pro ofs - J ohn Ben jam ins Pub lish ing (malinké) and Xassonke. his status implies their codiication (oicially recognized alphabet and orthography) as well as a certain role in the public sphere, particularly in the media and the educational sector. hese 3 languages belong to three language families: the Niger-Congo, the Nilo-Saharan and the Afroasiatic.0 he Niger-Congo family is the most important in Mali as well as on the continent. In Mali, it is represented by the Mande, Atlantic and Gur groups. he Manding cluster (Bambara, Maninka, Diula and Xassonke) dominates within the Mande group, which also includes two West Mande languages: Soninke and Bozo. he Atlantic group is represented by Fulfulde, and the Gur group includes Syenara, Mamara and Bomu (dialectal variation is important in this group and internal classiication varies, cf. Dombrowsky 994: 2; see also Section 5., Note 4 below). Songhay (whose classiication is still debated (cf. e.g. Nicolaï 989)) is the only Nilo-Saharan language, while the Afroasiatic family is represented by the Berber language Tamachek of the Tuareg people and the Semitic language Hassaniyya of the Moores. he Dogon language is still unclassiied (for details, see Skattum 2008). he position of the local languages varies greatly according to their vehicularity and their demographic, social and political weight. While Bambara is the national lingua franca, there are three vehicular languages at the regional level: Fulfulde (center), Soninke (west) and Songhay (north). Tamachek was, for political reasons, one of the irst four languages to be codiied and to be introduced as a language of instruction, though it has fewer speakers than e.g. Soninke. Some languages, like Syenara, Mamara, Xassonke and Maninka, are “invaded” (through language mixing), and progressively supplanted by Bambara (Dumestre 994b: 8). 4. PFC in Mali cor he PFC survey in Mali was carried out in collaboration with the CFA (Contemporary French in Africa and the Indian Ocean: usage, varieties and structures) project. CFA adapted the PFC protocol (Durand & Lyche 2003) to the multilingual context (Boutin, Lyche & Prignitz 2007) and to the wider CFA goals, including syntactic and sociolinguistic analysis (Dister et al. 2008; Lyche & Skattum 200b). 47 informants were recorded. hey were chosen according to four parameters, in the Un 10. For a historical overview and discussion of the classiication, see e.g. Heine & Nurse (2000, 2004). 11. http://www.hf.uio.no/ikos/english/research/projects/cfa/index.html 77 Com pan y Chantal Lyche and Ingse Skattum cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing following order of importance: () level of instruction; (2) age; (3) L; (4) gender – age and gender being standard PFC parameters, but showing no impact here. he level of instruction is the most important of these criteria, since French in “francophone” Africa is mainly – and in Mali nearly exclusively – learned at school. We have deined this level according to the diplomas delivered, which in Mali are: () CEP (Certiicat d’études primaires) = six years of schooling; (2) DEF (Diplôme d’études fondamental) = nine years; (3) Bac / Bac+ = 2 years / + eventually higher education. Levels of education not corresponding to these categories (interrupted schooling, professional education, diplomas no longer in use…) have been standardized to it into these three categories. Multilingual competence is particularly relevant in Africa, where on the one hand people frequently speak three or more languages (see Table  below), and on the other hand claim that they can detect a person’s L through their French. We therefore recorded informants speaking L from ive typologically distinct groups: Bambara (Mande group), Fulfulde (Atlantic group), Syenara (Gur group), all from the Niger-Congo family, and Songhay (Nilo-Saharan family) and Tamachek (Afroasian family). his enables us to test the inluence of L on FLS. In addition to these parameters, two extralinguistic factors turned out to have an impact on the pronunciation and general language competence of our informants: their exposure to French (in family and at work) and their mobility (place of origin, journeys inside Mali or abroad, length of stay in Bamako). hese factors were brought out through a perception test in 2008, where 6 Malian subjects listened to a one-minute extract from the semi-directed interview of 4 Malian speakers, in order to identify their L (Lyche & Skattum 200a). hough not included among the original parameters, this information could be retrieved from the interviews, which inter alia bore on the informants’ life story. Table  shows the proile of the 3 informants chosen for the PFC database: age, gender (F/M), languages spoken (the order corresponding to declared mastery), level of education, exposure to French (as expressed by their profession) and mobility (place of origin, moves within Mali, stays abroad and time spent in Bamako). hey are presented in alphabetical order following the individual part of their PFC code (maaaw, maabd, etc.). We shall see that the most striking phenomenon in the pronunciation of French among these informants is the great variation. his is due to the sociodemographic variables deined above. But we have also been able to identify the inluence of L in some areas. his is doubly interesting, as very little has been Un 7 Com pan y he phonological characteristics of French in Bamako, Mali Table 1. Proile of the 3 informants Level of education Profession Mobility maaaw (53) F TA, FR, BA CEP Housewife Born in Kidal region (north), nomad family, 4 moves in Mali, 26 years in Bamako, frequent visits to Kidal maabd (46) M BA, FR CEP (+2, no diploma) Oice boy Born in Bamako, always lived in Bamako maabh (24) M SO, TA, FR, BA, RU, EN Bac+ (+, socio- Student anthropology) lish ing PFC code L1 (in (age) bold), gender L2, L3… Pub Born in Gao region (north), 6 months in Bamako DEF (CEP+4, dressmaking) Dressmaker Born in Gao region (north), 4 years refugee in Burkina Faso, many years in Bamako maafc (50) F FU / FR, BA, SO, EN Bac+ (DEA + NGO senior Born in Djenne city (center) of Fulani 2 years, French executive family, 2 years in Bamako age 7–9 studies) (French spoken in family), France 0 years, 4 years in Bamako maaic (26) M FU, BA, FR, EN, AR Bac+ (master I, Student linguistics) maajs (5) F SE, BA, FR Bac (DEF+4, accounting) Maant (57) F SE, BA, FR DEF (+2, kindergarten teacher diploma) Kindergarten teacher Born in Sikasso region (south), 5 moves in Mali, 26 years in Bamako maasd (68) M BA, FU, FR, SO, MO, SO Bac Executive, retired Born in Djenne region (center) of Fulani family, 6 moves in Mali, one year in USA, 44 years in Bamako, considers he has several Ls but mastery of Bambara is best maash (22) M SO, FR, BA Bac+ (+2, Student French studies) Born in Gao region (north), one year in Bamako maass (62) F BA, FR, EN, SP Bac+ (+2 medi- Medical cal secretary secretary +3 special secretary) Born in Burkina Faso (at that time a province of the Ivory Coast), 5 years in France, 30 years in Bamako Accountant Born in Sikasso region (south), DEF in French boarding school in Sikasso city, 8 moves in Mali, many years in Bamako J rec ted pro ofs - cor Un Born in Mopti region (center), senior high school in Bamako, master (4 years) in Algeria, master I in France/ Netherlands ohn Ben jam ins maabm TA, FR, (30) F BA, SO 7 Com pan y Chantal Lyche and Ingse Skattum Table 1. (continued) Level of education Profession Mobility maatc (62) M BA, FR CEP Bus driver, retired Born in Kati near Bamako, always lived in Bamako maazw (36) F TA, SO, FR, BA DEF (9 years, without diploma) Musician Born in Timbuktu region (north), nomad family, many years in Bamako, international career, extensive travelling lish ing PFC code L1 (in (age) bold), gender L2, L3… Pub Languages: AR = Arabic, BA = Bambara, DO = Dogon, EN = English, FR = French, FU = Fulfulde, MO = Mossi, RU = Russian, SO = Songhay, SP = Spanish, SY = Syenara, TA = Tamachek rec ted pro ofs - J ohn Ben jam ins done on accents in FLS (Lyche & Skattum 200a), while interference2 is a central issue in studies of language contact. Today, the impact of extralinguistic factors is generally accepted, but there is debate as to the respective roles of intra- and interlinguistic motivations in language variation (see Skattum 200a). While some favor intrasystemic rather than intersystemic motivations for language change in French (Chaudenson et al. 993; Gadet & Jones 2008), others point to the inluence of local languages, like A. Quefélec (2008: 73): “Les normes locales, relativement permissives, subissent l’inluence des langues en contact et des véhiculaires africains dominants”. Intersystemic variation in French is well documented in Africa at the lexical level, but less at the syntactic (Skattum 200c)3, and very little at the phonological level (Woehrling & Boula de Mareüil 2006). It is, however, well established that the phonological level is even more “permissive” to local inluence than the two others (see Sankof 2002). We have also found common features in French pronunciation for some of the informants having diferent Ls, pointing to the existence of regional varieties of French – which is natural, since neighboring languages may well inluence one another. And inally, some characteristics are shared by other varieties of French in Africa, thus sustaining the hypothesis of a pan African French (Chaudenson et al. 993). cor 1. We use the term in the sense of a learner’s unconscious transfer of features from L to L2, and not in the sense of a learner’s strategy, which can be positive if the mother tongue structure matches that of the target language, but negative if it does not (Hamers 997). 13. hough Ploog (2008: 25) refers to some studies that focus on diferent aspects of interference in syntax. Un 0 Com pan y he phonological characteristics of French in Bamako, Mali Pub lish ing hough we look here at French in one country, Mali (more speciically the capital Bamako), we agree with Gadet & Jones (2008: 238) that “the very act of considering languages at a ‘national’ level could be criticized as something of an over-simpliication”. In Africa, the national level is particularly ill-suited as a frame, since the colonial powers did not consider ethnic or linguistic communities when tracing the borders. Whether national or regional, these languages nonetheless have an impact when coming into contact with French, which is why the description of our informants’ phonemic inventories will be preceded by a presentation of the inventories of the ive languages chosen for study. 5. he phonemic inventory 5.1 Five local Ls ohn Ben jam ins Taking into consideration all the diferent factors potentially inluencing the French spoken by our speakers, we do not expect to arrive at a single phonemic inventory for all our speakers. Notwithstanding, this section will show the emergence of common features, mostly shared by other varieties of French in Africa. In a context where French is always an L2 (in the sense deined above), and therefore systematically in contact with a plurality of Ls, a brief description of the ive Ls is called for. Un cor rec ted pro ofs - J As stated in Section 4, among 20 or so possible Ls, we retained ive major languages for their geographical coverage and their linguistic diversity: Bambara, Syenara, Fulfulde, Songhay and Tamachek. Bambara and Syenara, both present in the South of Mali, share the property of being tone languages, which is not the case for the three others. Fulfulde is spoken in the center of Mali while Songhay and Tamachek are northern languages. he irst diiculty in attempting a general presentation lies in the large number of dialects co-existing in some of the languages, dialects which may be highly diverse typologically. As a detailed presentation of the ive languages lies beyond the scope of this chapter, we will limit ourselves to some general information while focusing on the segments which the native languages lack in comparison to French, and which might therefore impede a standard acquisition of French phonemes. At this point, it might be useful to present what Clements (2004: 5) proposes as the African prototypical phonological system based on 45 phonological inventories (see Tables 2 and 3). Bambara is a CV language with the prototypical seven oral vowel system: /i, u, e, o, ɛ, ɔ, a/, enriched by seven nasal vowels /ĩ, ũ, ẽ, õ, ɛ̃, ɔ̃, ã/ (Dumestre 1 Table 2. African prototypical phonological system (consonants) Labiodental Dental / alveolar (Alveo-) palatal f t/d s/z n c/ɟ ʃ ɲ p/b stops fricatives nasals m Front Back u o ɔ ins a h ŋ Pub i e ɛ high mid-high mid-low low Laryngeal k/ɡ lish Table 3. African prototypical phonological system (vowels) Velar ing Bilabial Com pan y Chantal Lyche and Ingse Skattum cor rec ted pro ofs - J ohn Ben jam 2003) and by phonemic length. All seven oral vowels can be long, mostly in noninal syllables, vowel length usually stemming from vowel coalescence caused by intervocalic consonant deletion. Its consonant inventory matches that of Table 2 without the fricatives */ʃ, z/ which, however, may appear as allophones of /s/. Plosives /p, t, c, k, d, ɟ/ are restricted to word-initial position unless the syllable is reduplicated. Bambara is a strict CV language, CV accounting for over 99% of the syllables, with less than % of V and VC syllables. he strict CV structure may be broken up by complex onsets due to interconsonantal vowel elision (Cissé 2009). Fulfulde, which shows certain linguistic similarities with Bambara, has long been in contact with that language, resulting primarily in numerous lexical borrowings. he consonant system of Fulfulde includes the same nasal, plosive and fricative series as Bambara with additional implosives and prenasal plosives (Cissé 2009). It has a ive-vowel system /i, u, e, o, a/ with phonemic vowel length, long vowels being allowed anywhere within the word. Although it is a ive-vowel system, mid-open vowels appear when the following syllable contains an open vowel. In its syllabic structure, Fulfulde, a CV language, tolerates a larger number of (C)VC syllables (30%) than Bambara, but it does not allow complex onsets. he two languages difer essentially in their prosody, Bambara being a two-tone language system and Fulfulde having lexical stress. Syenara, like Bambara and Fulfulde, belongs to the Niger-Congo language group. Its phonemic system4 matches Table 2 with a full set of voiced fricatives 14. We refer here to the Supyire dialect spoken in Sikasso, as described by Carlson (994). Bendor-Samuel (97, cited in Dombrowsky 994: 2) classiies Supyire with Mamara and considers this the main group of Syenara (85%), both belonging to the northern branch of Syenara. Un  Com pan y he phonological characteristics of French in Bamako, Mali rec ted pro ofs - J ohn Ben jam ins Pub lish ing and additional nasal vowels /ĩ, ũ, ɛ̃, ɔ̃, ã/ (Carlson 994); all vowels may be short or long, length deriving nearly exclusively from the elision of an intervocalic consonant. Its syllabic structure is overwhelmingly CV, but like Fulfulde, it may allow simple codas word inally (Roulon 968). Prosodically, it is classiied as a tone language, the number of tones varying from two to four depending on the dialect and the linguistic description (Carlson 994; Roulon 968). Songhay and Tamachek share the property of being spoken in the northern part of Mali. Songhay can be subdivided into a number of varieties (Nicolaï 98). We will concentrate here on the variety spoken in the Gao region where our two informants are from. In that variety, Songhay is a ive-vowel system with phonemic length. he consonantal system matches the prototypical consonant system given in Table 2 for occlusives and nasals, with additional prenasalized stops, and includes in the Gao region the three voiceless fricatives /f, s, ʃ/ and two voiced ones /z, ʒ/ (Nicolaï 98). he syllabic structure of the language is CVC, where the coda is usually a sonorant or the voiced bilabial. In varieties of Songhay in close contact with Tamachek, any consonant may constitute a coda. Although some varieties of Songhay may distinguish up to four distinct tones, the Gao variety has ixed stress falling usually on the last syllable or on the penult of a word. When Songhay is in close contact with Tamachek, it tends to adopt a very strong word stress (Nicolaï 98).5 Tamachek, a southern Berber language, exhibits a standard ive vowel pattern /i, u, e, o, ɑ/ with, in addition, two short vowels /æ, ə/. From the description provided by Heath (2005), we assume that the so-called ‘full’ vowels are bimoraic and the short vowels monomoraic. A lowering rule applies in certain environments, and Heath (2005: 35) gives a few examples where /e/ is lowered to [ɛ] and /o/ to [ɔ]. he full series of voiceless and voiced fricatives (excluding /v/) are present in the consonantal system, which includes uvular, pharyngeal and laryngeal consonants. Tamachek possesses a CVC syllabic structure and allows complex constituents. Stress, which is always assigned at the word-level, is either lexical or rule-determined in the case of an unaccented stem. he ive languages share a number of linguistic features: the absence of front rounded vowels and of the voiced fricative /v/; the presence of an apical trill /r/; cor In oicial Malian documents, Syenara and Mamara are oten grouped together, Syenara designating the Sikasso dialect (ibid.). Our two Syenara speakers are both from the Sikasso region. Un 15. “A la diférence de ce qui se passe en songhay oriental les formes du songhay septentrional du groupe nomade sont afectées d’un fort accent d’intensité, qualitativement identique à celui de la (sic) tamacheq” (Nicolaï 98: 234). 3 Com pan y Chantal Lyche and Ingse Skattum he vocalic system lish 5. ing vowel lowering triggered by a rhotic in coda position.6 he languages do difer, however, in their prosodic structures and within their vocalic systems. Songhay and Tamachek are strict ive vowel systems with few mid-low vowels and we can expect their speakers to struggle with the opposition mid-low vs. mid-high in L2 acquisition. Songhay and Tamachek are both CVC languages in contradistinction with the other three and they are both characterized by lexical stress. cor rec ted pro ofs - J ohn Ben jam ins Pub When the native speakers of these diferent African languages learn French, they are faced with the diiculty that the target language possesses a larger number of vowel categories than their own. One can assume that they will easily perceive the categories of their native language, but that they will need to ‘learn’ to perceive the categories that they lack, and in particular /y/ and /ø, œ/. We observe that while most speakers show a reasonable mastery of /y/ in their production, very few, if any, reach a FR native-like proiciency when it comes to /ø, œ/. How can we account for this discrepancy in production? Recall that when a language uses three vowels only, these vowels are /i, u, a/, which ill up the extreme periphery of the vowel space, thus exhibiting a maximum of contrast and minimizing perceptual confusion. Boersma & Hamann (2008: 22) describe this phenomenon as a “primary auditory dispersion efect: categories tend to be located within the auditory space in such a way that they are perceptually maximally distinct”. In all the African languages under consideration, the vocalic space is already fairly crowded in its center with aperture distinctions between [o] and [ɔ], [e] and [ɛ], even though these distinctions are not systematically categorical, but rather allophonic in all but Bambara. herefore, introducing [ø, œ] renders the space even more crowded, hindering both perception and acquisition. In addition, a markedness factor no doubt intervenes in the deviant production of front rounded vowels (assuming that they represent the intended target). Vowel inventories including front rounded vowels are rare among the world’s languages, and /ø, œ/ are the last vowels to be acquired by children, who replace them either by their unrounded mid counterparts /e, ɛ/ or choose to give precedence to the feature [+ round] and produce back rounded vowels (/o, ɔ/) (Andreassen forthc. for Swiss French children). None of our speakers chooses the latter scenario, thus conirming that L2 acquisition activates speciic strategies. 16. Vowel lowering before a rhotic is part of a general vowel lowering process in at least Syenara and Tamachek (Carlson 994; Heath 2005). Un 4 Com pan y he phonological characteristics of French in Bamako, Mali rec ted pro ofs - J ohn Ben jam ins Pub lish ing Regarding high vowels on the other hand, we observe that the upper part of the vowel space is less crowded than for mid vowels, the dispersion between /i/ and /u/ being extreme. FR includes only one series of high vowels and when /y/ is introduced into the system, it competes with /i/ and /u/ but not, for instance, with a lax vowel. We thus expect that /y/ will be easier to acquire as a category than /ø, œ/. In case of confusion, once again, the feature [round] proves to be the culprit, and the speakers opt systematically for the front unrounded vowel /i/ and never for the back round /u/.7 hus, both mid and high vowels conirm that the combination of features that our speakers struggle with is [+front, +round] and that they always resolve the problematic cooccurrence in favor of the former over the latter. Although L2 acquisition triggers diferent strategies from L acquisition, this particular observation comes as no surprise when we consider Jakobson’s (94/97) developmental hierarchy for vowel sounds: irst low /a/, then high /i/, then either back /u/ or mid /e/, then mid-back /o/. he speakers give priority to [+front], not [+back, +round] in the acquisition process. Note moreover that if /y/ may be realized as [u] in certain varieties (as in sucre [sukr], Boutin, Gess & Gueye this vol.), /ø/ is never realized as [o], but may be realized as [e], thus conforming to what is predicted by the hierarchy. he degree of production competence varies from one speaker to the next, but we would like to claim that /y/ and /ø/ are part of all speakers’ inventory even though the distribution of these phonemes might be diferent from FR.8 In our perception test (Lyche & Skattum 200a), the subjects regularly refer to the dificulties encountered by the speakers of Songhay and Tamachek in articulating front rounded vowels. hese comments only partially relect a linguistic reality as basically all speakers at some point merge /e, ø, œ/ while only one (maabh) shows a strong tendency for using /i/ for /y/: dans la rue ([ri]), étude and all its derived words are realized with an /i/ ([etid]). his informant is indeed a young (age 24) speaker of Songhay (a student, very recently arrived in Bamako who was raised by his grandfather, himself a speaker of Songhay and of Tamachek). Speaker maabh alternates between the two vowels according to lexical items: j’ai eu [ʒey], nourriture [nurityr] vs. étudier [etidje]. cor 17. We do not ind any such instance in our data, which does not mean that those realizations do not occur. See Boutin, Gess & Gueye (this vol.) for Senegal, where both strategies are used. It seems, however, that there exists a strong tendency to preserve the frontness of the vowels rather than their roundness, as the deviant realizations of /ø, œ/ conirm (see below). Un 1. he PFC word list does not speciically test the pronunciation of /y/. he PFC text, on the other hand, allows a comparison of the diferent realizations among speakers with words like commune, usine, plus, s’assure, paru and shows that certain items are always pronounced with /y/. 5 Com pan y Chantal Lyche and Ingse Skattum J ohn Ben jam ins Pub lish ing he mid-vowels behave diferently from their FR counterparts, with all our informants exhibiting a certain degree of confusion between /ø, œ, ə/ and /e/. Let us stress that FR /ø, œ, ə/ are only exceptionally realized as open /ɛ/19 even though the category /ɛ/ is present in the vowel inventory, as in Bambara. One informant only, maafc, performs like a native FR speaker in this respect: maafc is a speaker of Fulfulde, a highly educated 50 year-old woman, leader of a non-governmental organization, who resided in France for 0 years. Close to FR too,20 we ind a bambarophone, maasd, an educated 68 year-old male, originally fulaphone. As none of our speakers of Songhay and Tamachek but one (maabm, a 30 year old female with 0 years of schooling) makes clear aperture distinctions between the front rounded vowels (which they readily replace by unrounded vowels), there appears to be a certain link between the L and the robustness of the rounded mid vowel category in the speakers’ phonemic systems. Recall that both Songhay and Tamachek are ive-vowel systems with few open mid allophones. One should keep in mind, however, that Songhay and Tamachek informants either have a maximum of 0 years of schooling, or they moved recently to Bamako and have had reduced exposure to French. It thus seems possible that the higher rate of confusion observed among them can be attributed to factors other than the L. Tamachek includes in its inventory a schwa whose presence does not however facilitate the realization of mid front rounded vowels. It is then obvious that the speakers, when constructing a new vowel inventory, favor phonemic distinctions even though they could make use of a phonetically closer sound, thus following the category proximity principle as deined by LaCharité & Paradis (2005: 226) for loanword adaptation: rec ted pro ofs - Category proximity principle If a given L2 phonological category (phoneme) does not exist in L, this L2 category will be replaced by the closest phonological category in L, even if the L inventory contains acoustically closer sounds. Even though we are here dealing with the acquisition of an L2, we see the same principle at work: the speakers choose the phoneme /e/ over the phonetically closer schwa [ə]. In the word list, all speakers but two (maafc and maasd) pronounce chemise and petit with an unrounded vowel ([ʃemiz] and [peti] respectively). Similarly, when the verb venir in any of its possible forms occurs in the two conversations, cor 1. We noted a pair of such realizations for meurtre (maabm and maabh). 0. As in our perception test, subjects commented on vowel articulation (for ex. “Tamachek speakers have problems with /y, ø/”), we assume that they regard the pronunciation of /y/ as [i] and /ø/ as [e] as deviant from FR. Un 6 Com pan y he phonological characteristics of French in Bamako, Mali cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing it is systematically realized [venir]. As will be developed in Section 6, we have grounds to claim that schwa is restricted to inal syllables (clitics and polysyllables) as vowels are stable in all other positions. his implies that most speakers store in their lexicons chemise, petit, venir, etc., with an unrounded mid vowel in the initial syllable of the words. On the other hand, all speakers round the vowel in jeune or jeûne. In our data, schwa alone (and not FR /ø/ or /œ/) is realized as the front unrounded vowel [e]: parce que [parske], beaucoup de [bokude], pas de temps [padetɑ̃]. he lack of prosodic salience could be at play here, since a schwa syllable in French rarely carries stress. In an unstressed context, the demands for auditory contrast are reduced, favoring articulatory economy, which, in our case, implies producing an unrounded vowel. In addition, certain speakers show a strong tendency for associating a graphic <e> with /e/, while <eu> usually corresponds to a front rounded vowel. maajs for example (whose L is Syenara) rounds the vowel in des jeunets [deʒønɛ] and déjeuner [deʒøne] while she realizes des genêts as [deʒenɛ]. he same speaker of Songhay who experiences diiculties with the /y/-/i/ distinction, maabh, pronounces rauque (FR [ʁoːk]) with a inal [e]: [roke]. He singles himself out as well by producing a mid-open vowel in meurtre [mɛrtr], although the rest of his performance rates as average. It is therefore possible to consider that schwa is realized as [e, ø], a tendency due to a combination of prosodic and orthographic factors, while the phoneme /ø/ is appropriately acquired, barring a few individual idiosyncrasies,2 which sets it apart from its FR equivalent. he distribution of the unrounded mid vowels /e, ɛ/ follows the classical norm: */e/ in closed syllables where /ɛ/ is required, but a standard contrast in open syllables, épée in opposition with épais (/epe/ vs. /epɛ/) for all speakers. One item exhibits a high degree of variation: piquet is pronounced with a mid-low vowel by eight of the 3 speakers. We see here the inluence of the orthography as piquais on the other hand shows a mid-high vowel in accordance with the orthoepic rule associating a mid-low vowel with the ending -ais. In inal closed syllables, the mid vowel /O/22 tends to follow the pattern described for /E/, oten being open in opposition to the preferred norm. From the PFC word list, we have: rauque, FR [ʁoːk], pronounced [rɔk] by seven speakers; paume, FR [poːm], pronounced [pɔm] by ive speakers; and gnôle, FR [njoːl], pronounced [njɔl] by ten speakers. Orthography, and in particular the association in schools of the circumlex accent with an open mid vowel, probably explains Un 1. Among various idiosyncrasies, let us mention isolated instances of hypercorrection: piquet [pikɛ] pronounced [pikø] and fêtard [fɛtar] pronounced [føtar]. . Capital letters are given for underspeciied segments. 7 Com pan y Chantal Lyche and Ingse Skattum he consonantal system J 5.3 ohn Ben jam ins Pub lish ing the large consensus in the pronunciation of gnôle (‘brandy’), a rather rarely used word. he same phenomenon is observed for jeûne which is pronounced [ʒœn] by four speakers in isolation, while most speakers show a high insecurity level in the minimal pairs: either the two words jeune and jeûne are pronounced alike, preferably with a mid-high vowel, or they are inverted. his fact stresses the problems that all speakers face regarding the realization of /ø, œ/ and their distributions. From our observations, we conclude that the phonemic inventory contains the phoneme /ø/, but that the distribution between the two allophones [ø, œ] remains subject to individual variation in spite of a number of (pan) African tendencies, with peuple, creuse for example, always showing a mid-high vowel (Boutin & Turcsan 2009).23 he low functional load and the low frequency rate of the vowel ensure that the phoneme /ø/ remains endangered in most speakers’ inventories. Similar to other African varieties, the French spoken by our Mali informants does not oppose a front and a back /A/, the only low vowel being [a]. he situation for the nasal vowels proves more complex: bambarophones diferentiate easily /œ̃/ and /ɛ̃/, as expected, considering the rich nasal vowel inventory of their L, while the situation varies for the other speakers. We observe here a clear link between the use of the two phonemes and extralinguistic factors like education, mobility, exposure to French, the speakers of Songhay and Tamachek tending to one phoneme only, the preferred vowel being rounded /œ̃/. Generalizing somewhat across speakers, we propose a vowel system composed of nine oral vowels /i, y, u, e, ø, o, ɛ, ɔ, a/ and three nasal vowels /œ̃, ɔ̃, ɑ̃/. cor rec ted pro ofs - he French consonant system is well assimilated by our speakers, difering essentially from FR in the articulation of the rhotic which is either an apical trill or a tap. Apart from the rhotic, we observe minimal interference from the consonantal system of the diferent Ls. Such interference, when it exists, concerns individual speakers and cannot be generalized to all speakers of the same L. We mentioned in Section 5. that Bambara and Fulfulde do not include in their inventories voiced fricatives and the voiceless alveo-palatal fricative /ʃ/24. We then expect speakers to show a certain degree of confusion in the production of these sounds. In fact, only one of our speakers, maabd, a bambarophone, does not possess the alveo-palatal fricatives: benjamin [bɛ̃zamɛ̃] for [bɛ̃ʒamɛ̃], chaussures [sosyr] for 3. According to Boutin & Turcsan (2009: 46), in Ivory Coast French, the mid-low vowel occurs in a syllable closed by /r/ or when the graphic vowel bears a circumlex accent. Un  4. Although both [ʃ] and [z] are quite common as allophones of /s/. Com pan y he phonological characteristics of French in Bamako, Mali cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing [ʃosyr] among other examples. Another bambarophone, maasd, shows a tendency to use the velar glide [w] instead of the fricative [v] (pouvait [puwɛ]). We could attribute those pronunciations to reduced schooling and lack of exposure to French, but we need more data to test this hypothesis. While the two fulaphones, who display excellent FR proiciency, have indeed completed higher university education, thereby supporting our hypothesis, the proile of these two Bambara speakers invalidates it: maabd completed six years of school (CEP) only, but on the other hand, his work environment (the University of Bamako) requires regular use of French; maasd inished his high school education (baccalauréat), but he is much older without current exposure to French. We underlined in Section 5. that Bambara, Fulfulde and Syenara are all more or less strict CV languages in opposition to Songhay and Tamachek. he CV preference triggers for example, the insertion of a vowel to break up a cluster as in trop usually pronounced [toro] by basilectal25 speakers of French with a CV language as L. Our speakers on the other hand, do not show diiculties in the articulation of complex onsets. We searched for a CV preference word-inally among speakers of Bambara, which is a strict CV language. Our data show subtle distinctions that we will exemplify with intact and infect taken from the word list: intact is realized [ɛ̃tak] by two speakers of Syenara and two speakers of Bambara, and as [ɛ̃tat] by one speaker of Songhay; infect is realized [ɛ̃fɛk] by one speaker of Syenara, one speaker of Bambara and three speakers of Tamachek. Out of seven speakers whose L is a CV language, two speakers only (Bambara and Syenara) drop the inal plosive in both words. Diferent tendencies emerge when we consider the clusters in explosion, extraordinaire, ex-mari. he clusters are maintained by all northern speakers except maazw (Tamachek, an artist who travels extensively) who simpliies the cluster in extraordinaire [ɛstraɔrdinɛr], while ive of the CV language speakers simplify the cluster in one or several words: maajs (Syenara) simpliies all clusters and drops inal consonants, but her behavior singles her out. We can conclude that there is a certain amount of interference from the speaker’s L counterbalanced by extralinguistic factors. All speakers, however, tend to devoice stops and fricatives word inally, regardless of their L. In the text for example, all speakers maintain a voiced fricative in chemises en soie ([ʃəmizɑ̃swa]) when the inal consonant is resyllabiied as an onset of the following word within the same prosodic unit, but village, barrage, chaudes, all show devoiced segments. his feature appears to be common to several varieties of African French (see Boutin & Turcsan 2009; Bordal this vol.; Boutin, Gess & Gueye this vol.). Final stops tend to be unreleased (see Boutin, Gess & Gueye this vol.), but the picture here is blurred Un 5. In a linguistic continuum of varieties, the basilect is the furthest removed from the superior pole, called acrolect, with intermediate varieties making up the mesolect (Chaudenson 997).  Com pan y Chantal Lyche and Ingse Skattum cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing by the interference of the L and by the amount of exposure to French: maafc for example, who practices French daily and who lived in France for 0 years, clearly releases all inal stops while devoicing the inal fricatives (des barrages, where the inal expected [ʒ] does not show any voicing bars in the spectrogram). Searching for distinctions between the speakers due to interference from the diferent Ls leads us to consider the behavior of the rhotic. It is well known that French Creoles, all CV languages, eliminate /R/ in coda position (Nikiema 2002), and this particularity carries over to bilingual (Creole-French) speakers when they speak French. Bordal (2006) conclusively shows that in the variety of French spoken in Reunion Island, /R/ exhibits a high degree of instability especially in coda position. /R/ may drop in both unstressed and stressed syllables, which may induce a compensatory lengthening efect on the preceding vowel, or it may reduce to a vocalic appendix of a schwa-like quality. Diferent varieties of French in Africa behave similarly (Boutin & Turcsan 2009; Bordal this vol.) when the speakers’ L is a CV language. Given the divergent syllabic structures of the Ls under consideration in this study, we expected distinct speaker behaviors and in particular, we hypothesized that speakers of Songhay and Tamachek would articulate /R/ in coda position in opposition to speakers of Bambara, Syenara and Fulfulde. his prediction was veriied throughout the data although the /R/-less pronunciation of a few items is generalized: all speakers produce parler, peut-être and parce que without the rhotic ([pale], [pøtɛt], [paskø] respectively). he pronunciation of the last two items corresponds to a FR realization where obstruent + liquid clusters simplify readily in conversations (Laks 977; Wachs 997), and where [paskø] is the most common realization of the conjunction. he verb parler then constitutes the only exception, one that we attribute to frequency of usage. he /R/-less form is widely used in Reunion French as well as in the Central African Republic, and we suspect that it is stored as such in most speakers’ lexicons. Songhay and Tamachek speakers massively maintain the rhotic in all environments, including those where it is highly susceptible to deletion in FR. Final obstruent + liquid clusters are regularly simpliied in FR. Stigmatized at irst, and more advanced in North American varieties of French, this phenomenon proves remarkably active in all registers of FR. he PFC text, read by all our subjects, contains a number of candidates for inal cluster simpliication with words like ministre, centre, etc., which the coding system allows us to retrieve. We searched the database for cluster simpliication among upper class Parisian speakers who can be characterized as representing a conservative variety of French (Lyche & Østby 2009) and found 8 such occurrences. We performed the same search for Bamako, which gave us (for our 3 speakers) 38 instances of simpliication with only two emanating from a Tamachek speaker and none from a Songhay speaker: maazw (an international artist much exposed to French) produces ministre twice Un 0 Com pan y he phonological characteristics of French in Bamako, Mali 6. Schwa rec ted pro ofs - J ohn Ben jam ins Pub lish ing without the /R/.26 maaaw, another Tamachek speaker (a housewife with only CEP and not much exposure to FR), is typical in this respect. In her semi-formal conversation, she regularly used the verb partir, and unlike Syenara, Bambara or Fulfulde speakers, who frequently delete both /R/s, she articulates them distinctly. he word théâtre happens to be recurrent in the same conversation, and she never simpliies the inal cluster. In comparison, maant, a Syenara speaker, deletes most of the rhotics in coda position: garçon [ɡaːsõ], jardinière [ʒaːdinɛə] or [ʒaːdinɛr]. he rate of /R/ presence clearly diferentiates Songhay and Tamachek speakers from their Bambara, Syenara or Fulfulde compatriots. In the perception test carried out by Lyche & Skattum (200a), the subjects performed better at identifying northern speakers, and among the discriminating factors they pointed to a diferent articulation of the rhotic. According to the literature at hand (Carlson 994; Cissé 2009; Dumestre 2003; Heath 2005; Nicolaï 98), all the languages under scrutiny realize the rhotic consonant as a dento-alveolar trill. his particular articulation seems to be the only one used in Songhay and Tamachek, while the other three languages alternate between a dento-alveolar trill, a lap and a uvular approximant. We concur with the subjects in perceiving a more distinct articulation of /r/ in Tamachek and Songhay, a distinction which could arise from a larger number of vibrations. We propose, however, a divergent explanation, partially independent of articulation. We contend that the overwhelming presence of rhotics in the discourse of Tamachek and Songhay speakers singles them out and contributes to their identiication (Lyche & Skattum 200a). In addition, but probably of less consequence, the distinct articulation of the consonant combined with its stability reinforces positive identiication of the speakers’ origins. Further tests and further research are required in order to fully understand the factors at play, but there is no doubt that the behavior of the rhotic constitutes a strong discriminating element among the diferent Ls. cor he stability of schwa characterizes a number of varieties of French spoken in Africa, for example in Senegal or Central African Republic (Boutin, Gess & Gueye, Bordal this vol.). In this regard, our speakers’ behavior conirms previous observations: schwa is restricted to word-inal position (clitics and polysyllables) and to word-internal position, but excluded from word-initial position. Let us irst underline that by schwa we understand a vowel alternating between a full vowel ([ø, œ, Un 6. She is also the only Tamachek speaker to simplify internal clusters in the word list as already mentioned above. 1 Com pan y Chantal Lyche and Ingse Skattum rec ted pro ofs - J ohn Ben jam ins Pub lish ing ə, e]27) and zero. A schwa usually corresponds to the letter <e>. We adopt here a narrow view of schwa in that when a vowel is stable, it can no longer be a schwa (as has been proposed for Midi French, see Durand 995; Eychenne 2006). Based on the results of the PFC coding system (Durand & Lyche 2003), we will comment on the presence/absence of schwa in the diferent positions in the word. Word inally, schwa rarely surfaces, but its frequency in this speciic context surpasses what we can ind in other survey points. Considering exclusively the two conversations, very few schwas occur word inally ater a single consonant (6 schwas for ,093 unrealized schwas), and this number shrinks further to 8 when we exclude all inal schwas followed by the hesitation marker euh, e.g.: là ils viennent euh (maaaw). We presume that in those cases, the vowel might be part of the hesitation marker or that it signals ongoing discourse planning. At the end of a word ater a consonant cluster, however, schwa shows a high degree of stability: 50 absences vs. 68 presences. he absence of schwa occurs when the cluster is reduced to a single consonant (maasd: par exemp(le) [ekzɑ̃p] mon vieux), while the vowel is retained when both consonants are pronounced even if the following word is vowel initial: pas du tout impossible [ɛ̃posiblə] à relever (maash). Without the schwa, the word-inal syllable would include a heavy coda, banned by Bambara, Syenara, Fulfulde: [ekzɑ̃pl], [ɛ̃posibl].28 A highly undesirable structure is repaired either minimally by the simpliication of a coda (but still a marked structure) or by the presence of a inal schwa creating a new syllable and ensuring that the last two syllables are open. his repair strategy concerns mainly speakers whose L is a CV language, while Songhay and particularly Tamachek speakers (as already mentioned in Section 5.3) are much less prone to simplify the clusters, much less in fact than what is observed in FR. Contexts other than word-inal unveil, as expressed in Table 4, a conirmed tendency towards a stabilization of the vowel in the context V(#)Cə, a prime environment for variation in FR (Dell 973). he table shows an extremely high level of stabilization of schwa. In wordinitial position schwa shows a presence rate of greater than 97%. he two occurrences where schwa is dropped word initially are standard cases involving the preix re- and the adjective petit: il est r(e)venu à Bamako (maant); Ils avaient une p(e)tite connaissance (maass). he absence of schwa in these two examples 7. As shown in Section 5, a schwa is oten realized as [e]. cor . Note that the illicit structure is not repaired by a possible enchaînement (forward linking), usually so characteristic of French prosodic structure. For all our speakers, the basic stress unit is built upon a lexical word, and cannot incorporate several such words, as is usually the case in FR where the basic stress unit may include several lexical items. We will see in Section 7 how prosodic structure impacts on liaison as well. Un  Table 4. Schwa distribution in conversations Schwa present Schwa absent 69 359 40 2 58 49 ing Polysyllables initial Clitics Polysyllables medial Com pan y he phonological characteristics of French in Bamako, Mali rec ted pro ofs - J ohn Ben jam ins Pub lish cannot be directly correlated with the education level of the speakers as maass, a medical secretary who has studied ive years in France, shows a diferent proile from maant, a kindergarten teacher with  years of schooling who has always lived in Mali. Among the 69 occurrences of realized schwas we count several instances of petit(e), a large number of the preix re-, indicating a lack of correspondence between the absence of schwa and the frequency (or the internal structure) of an item. his entails that schwa syllables are absent word-initially for the large majority of speakers and that word-initial syllables only include lexically stable vowels. In contradistinction to FR, schwa is massively present in clitics, including je (je pense que). Absence of schwa is limited to a few frequent locutions. he interrogative qu’est-c(e) que and the locution n’est-ce pas are the most likely candidates for the absence of the vowel and we presume that they are both lexicalized as such. In addition, a third (9 out of 58) of the occurrences of absent schwas involves the negation ne. he presence of the particle ne constitutes a pan African tendency29 and requires some attention since a large number of studies (inter alii, Ashby 98; Coveney 996) have shown that the irst part of the negation in conversational French is massively absent. he Vendée data, taken here as an example of FR, provide a case in point. he PFC search engine allows searching for autonomous words (or strings of characters) in the transcription and we count 307 occurrences of pas for 4 occurrences of ne.30 In our data, on the other hand, a similar search extracts 269 occurrences of ne for 572 occurrences of pas, that is about 50% of the occurrences.3 he coded portion of the data shows that, when the negation particle ne is present (37 occurrences), it may be reduced with a reduction rate close to 50% (8 with a vowel and 9 without). A closer cor . Quefélec (2008: 73) notes the “maintien très fréquent (statistiquement largement supérieur à ce qu’on observe en français oral européen) de l’adverbe ne comme signe de la négation verbale : même en situation informelle, cet indice négatif clitique reste présent dans le discours parlé africain, là où il a largement disparu en français hexagonal”. Un 30. Recall that the PFC protocol speciies that any element absent from the signal is not transcribed as in il vient pas, faut qu’il vienne, etc. 31. In this number are included repetitions and locutions like pas encore. 3 Com pan y Chantal Lyche and Ingse Skattum 7. Liaison rec ted pro ofs - J ohn Ben jam ins Pub lish ing examination of the verbs involved does not highlight any tendency; no lexicalization is taking place, for example, in je ne sais pas, je ne peux pas which, considering the frequency of these two expressions, would appear as good candidates for the absence of the vowel. Among clitics, ne maintains an underlying schwa while the vowel may have stabilized elsewhere. Another locus of true variation, as indicated in Table 4, is word-internal position with about a 50% rate of schwa retention. Here, too, numbers alone do not present an accurate picture of the data and can, in fact, be deceptive. Out of the 49 occurrences where schwa is absent, 24 concern parce que and nine the adverb maintenant, which reduces considerably the number of items subject to variation. If parce que never appears with a schwa, thus indicating that it is stored as /parskø/ (or more oten /paskø/) by the speakers, maint(e)nant freely alternates with maintenant and has not (yet) lexicalized without the vowel. It is noteworthy, however, that this particular adverb should provide the prime locus of variation: in all varieties of French its frequency of use ranks extremely high and it is systematically pronounced without schwa except in the most conservative varieties of Midi French. Before concluding this section, we turn briely to the pronunciation of schwa in the text. As expected, the vowel shows even more stability in the reading task. Out of  schwas absent in clitics, seven involve the interrogation qu’est-ce que, and only six occurrences of absent schwas word-internally can be observed. As a comparison, the Paris data from the upper class bourgeoisie, with only one more speaker, total 40 absent schwas in clitics and 35 word-internally in the text. he high level of schwa presence in the Bamako data inds a natural explanation in the way French is acquired. As underlined in Section 3, French in Mali is never an L; it is learned at school and, in spite of the crying lack of teaching material, the relationship to the language remains highly inluenced by the written word. We thus conclude that the scope of schwa is much more restricted than in FR, conined as it is to word-inal syllables where it exhibits a larger degree of variation in polysyllables than in clitics. If a few speakers have internalized a word-medial schwa, the word-initial position admits full vowels only. cor Limited variation, particularly prevalent for schwa among our speakers, characterizes liaison as well. In an analysis of liaison in four survey points in Africa (Abidjan, Bangui, Ouagadougou and Bamako), Bordal & Lyche (2008) showed that categorical liaison was restricted to determiners + nouns, pronouns + verb, and the prepositions en, dans + NP, all other contexts being variable. A number Un 4 Com pan y he phonological characteristics of French in Bamako, Mali of studies have underlined that morphological information is relayed via liaison (Morin & Kaye 982; Durand & Lyche 2008, among others), which strengthens this particular context, as stated in Rosset (9: 283): ing Il faut remarquer que, mis à part les locutions toutes faites […], les mots qui font liaison sont tous terminés en t, s ou n ; t, s, n étant des consonnes de désinences, avaient un rôle morphologique qui leur donnait une plus forte résistance à la tendance phonétique qui les amüissait. cor rec ted pro ofs - J ohn Ben jam ins Pub lish In other words, the reason why the word-inal consonants /t, s, n/ do not delete stems from their morphological roles which have strengthened them and ensured that they are not afected by the general tendency toward inal consonant weakening.32 In our survey, the distribution of the three consonants is as follows: /z/ 22, /n/ 30, /t/ 46. Morphology alone however fails to account for the massive supremacy of /z/ as a liaison consonant. Liaison is maintained when it fulills a morphological role but exclusively within a close prosodic unit, preferably a prosodic word deined as a potential stress-bearing domain. When we consider the 22 occurrences of realized liaisons in /z/, we observe that liaison is indeed systematic between a determiner or a clitic and its head, with only ive cases falling outside these categories. Out of these ive tokens, two involve the preposition dans (dans un rôle de parleur) and three adverbs: two occurrences of très (très intéressant, très exactement), one of pas (c’est pas un p/). None of the other contexts triggers categorical liaison in the diferent tasks. he conversations, for example, show a total of 44 instances where (c’)est does not link to the following word, conirming that the morphological role of the liaison consonant is restricted to the indication of a plural marker. Following polysyllabic words, liaison rarely occurs (8/9 occurrences in the conversations) and all the instances but one (étaient élevées) concern the plural form of an adjective followed by a noun: autres, premières, diférentes, petits. he text further conirms the variability of liaison in contexts other than det/clitics + head and more speciically the absence of standard liaison between an adjective and a noun: if  speakers pronounce a liaison in grand honneur, only nine do so in grand émoi. his highly restricted liaison usage, conining it mostly to categorical liaison, is similar to what has been observed in other African countries (Bordal & Lyche 2008; Bordal this vol.; Boutin, Gess & Gueye this vol.). It relects the situation in another region where French interacts daily with another language, namely Un 3. Note, however, that while /z/ is associated with a plural marker and /t/ with a verbal marker, the role of /n/ as a morphological desinence is less clear. 5 Com pan y Chantal Lyche and Ingse Skattum rec ted pro ofs - J ohn Ben jam ins Pub lish ing Louisiana French (Klingler & Lyche this vol.).33 When reduced to its bare minimum, liaison in its core function bears a prosodic role enhanced by a morphological one.34 What we do not observe in those varieties is a high degree of variation between the diferent styles. While this absence supports the view that Louisiana French exhibits a number of characteristics of attrition (Rottet 2005), reinforced by the lack of literacy in French among speakers, such should not be the case in Mali, where French is mostly the language of formal interactions. Just like schwa can be claimed to stabilize thanks to a spelling inluence, we would expect in our data a higher rate of liaison than what is observed. In fact, our speakers, even the most educated ones, do not present any of the characteristics associated with the inluence of spelling on pronunciation, and in particular, we do not ind a single occurrence of unlinked liaison (liaison sans enchaînement) as described by Encrevé (988). he explanation for this state of afairs probably rests upon a prosodic factor common to African varieties of French and Louisiana French. In none of the diferent Ls of the speakers does stress carry a demarcative function as it does in FR. Recall that French does not have word stress, only phrasal stress, and that liaison has been seen to strengthen the internal cohesion of the constituents within a stress group (Selkirk 974). he prosody of the diferent varieties of French in Africa difers greatly from its FR counterpart and the impact of the L should not be underestimated (Lyche & Skattum 200a). From this discussion, we infer that the prosodic word, as a stress unit in Mali French, integrates a lexical word and its clitics. Liaison is then categorical within the prosodic word. It follows from this deinition that monosyllabic prepositions are not integrated within a prosodic word as posited in earlier studies (Selkirk 974); rather, they constitute an independent prosodic unit. If such is the case, the contrastive behavior of the two monosyllabic prepositions dans and chez calls for an explanation. In our corpus,35 a categorical liaison follows dans but not chez, although both prepositions are monosyllabic. We attribute the diference to a frequency factor: in the PFC database, we count 4860 occurrences of dans for 963 of chez, indicating that dans is much more present in daily interactions and therefore forms with the following lexical item a construction in the sense of Bybee (200). An on-going study among non-literate speakers in Senegal by Boutin (p.c.) 33. he fragility of the system is also stressed by Walker (this vol.) for French in Alberta. cor 34. See in addition the study of Aub-Buscher (962) of the dialect of Ranrupt in eastern France. 35. A similar situation is observed in Abidjan, Ouagadougou and Bangui (Bordal & Lyche 2008). Un 6 Com pan y he phonological characteristics of French in Bamako, Mali conirms the main taxonomy established above, thereby eliminating any impact of orthography on liaison in African varieties of French. . Conclusion jam ins Pub lish ing his chapter has examined the main characteristics of the French spoken in Bamako by 3 speakers of ive diferent Ls. Although the phonemic inventory we posited for these speakers difers little from FR, we noted four salient characteristics: () a speciic distribution of round mid-vowels; (2) the fronting of schwa and its stabilization in word-initial and word-internal positions; (3) the speciic realization of the rhotic; (4) the quasi absence of variable liaison. Taken collectively or individually, these features do not single out a particular variety of French, as identical traits prevail in other varieties of African French (Bordal this vol.; Boutin, Gess & Gueye this vol.; Bordal & Lyche 2008). his could be taken to support the idea of an “inter-African norm”, oten mentioned in the literature, e.g. by Manessy (992: 62): ohn Ben Plus surprenante est l’impression de cohérence que procure l’examen d’ensemble des données36 et qui a conduit certains auteurs à postuler l’existence d’une ‘norme interafricaine’ justiiant la reconnaissance d’une variété régionale de français coextensive à l’Afrique noire francophone. Un cor rec ted pro ofs - J Our study showed, however, that a number of diferences emerged when we took into consideration the L of the speakers. We opposed in particular the two northern languages Songhay and Tamachek to Bambara, Fulfulde and Syenara. he fact that both Songhay and Tamachek are C(C)VC(C) languages distinguishes their speakers from the others, who tend to simplify clusters and in particular to drop the rhotic in coda position. Our speakers of northern languages maintain the rhotic in contradistinction to their compatriots, as well as to speakers of FR. he vowel inventory of the L was also shown to impact the distribution of vowels in French. Finally, although not explored here, the prosody of the L colors the French of our speakers (Lyche & Skattum 200a), a fact deserving further attention. When examining certain extralinguistic factors, however, we see that the impact of the L is leveled out, reminding us that French in Mali is an L2, always subject to development in contact with other languages. In this respect, we also acknowledge the importance of Bamako, the capital, a melting pot where Bambara 36. From seven African countries. 7 Com pan y Chantal Lyche and Ingse Skattum functions as a vehicular language, its regular use smoothing out the speciic features of the other Ls. he urban inluence is stressed by Manessy (994: 5): ing il se crée […] en Afrique des variétés locales, plus exactement nationales, car ce phénomène [‘la systématisation des manières de dire’] est surtout urbain et l’efet de la capitale y joue un grand rôle […] que ce soit sur le plan phonétique, sur le plan lexical ou, au niveau du discours, dans le maniement des ressources intonatives […]. jam ins Pub lish hough we concur with Gadet & Jones (2008: 238) (cited above, see Section 4), that “the very act of considering languages at a ‘national’ level could be criticized as something of an over-simpliication”, we do not see a contradiction in terms in the coexistence of local and pan African features. Together with, for example, the level of education and exposure to French, diatopic diferences across varieties give African French its distinctive color. he French spoken in Bamako is thus one and many, its diversity contributing to the wealth of the French language. Ben References cor rec ted pro ofs - J ohn Andreassen, H. N. forthc. he acquisition of schwa in Swiss French. Phonological variation in grammar during L acquisition. PhD diss., University of Tromsø. Ashby, W. 98. he loss of the negative particle ne in French: A syntactic change in progress. Language 57: 674–687. Aub-Buscher, G. 962. Le parler rural de Ranrupt (Bas-Rhin): essai de dialectologie vosgienne. Paris: Klincksieck. Banque, M. 2007. L’éducation au Mali. Diagnostic pour le renouvellement de la politique éducative en vue d’atteindre les objectifs du Millénaire. Washington D.C. Boersma, P., and S. Hamann. 2008. he evolution of auditory dispersion in bidirectional constraint grammars. Phonology 25: 27–270. Bordal, G. 2006. Traces de la créolisation dans un français régional. Le cas du /R/ à l’île de la Réunion, Master thesis, University of Oslo. www.projet-pfc.net. Bordal, G., and C. Lyche. 2008. La liaison en terre africaine. Paper presented at the conference Phonologie du français contemporain: variation, interfaces, cognition, December –3, 2008. FMSH, Paris. www.projet-pfc.net. Boutin, B. A., C. Lyche, and G. Prignitz. 2007. PFC en terre africaine. Bulletin PFC 7: 297–326. www.projet-pfc.net. Boutin, B. A., and G. Turcsan. 2009. La prononciation du français en Afrique: la Côte d’Ivoire. In Phonologie, variation et accents du français, eds. Jacques Durand, Bernard Laks and Chantal Lyche, 3–52. Paris: Hermès – Lavoisier. Brock-Utne, B., and I. Skattum, eds. 2009. Languages and education in Africa. A comparative and transdisciplinary analysis. Oxford: Symposium Books. Bybee, J. 200. Phonology and language use. Cambridge: Cambridge University Press. Calvet, L.-J. 994. Les voix de la ville. Introduction à la sociolinguistique urbaine. Paris: Payot. Un  Com pan y he phonological characteristics of French in Bamako, Mali Un cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing Canut, C., and G. Dumestre. 993. Français, bambara et langues nationales au Mali. In Le français dans l’espace francophone, tome , eds. Didier de Robillard and Michel Beniamino, 29–228. Paris: Champion. Carlson, R. 994. A grammar of Supyire. Berlin: Mouton de Gruyter. Chaudenson, R. 997. «Acrolecte», «Basilecte», «Mésolecte». In Sociolinguistique. Concepts de base, ed. Marie-Louise Moreau, 9–20, 60, 20. Bruxelles: Mardaga. Chaudenson, R., R. Mougeon, and E. Beniak. 993. Vers une approche panlectale de la variation du français. Paris: Didier Erudition. Cissé, I. 2009. Comparaison de deux langues en contact, le fulfulde et le bambara, dans une perspective typologique: structures phonémiques, syllabiques et lexicales. Mémoire de master, Université Stendhal Grenoble III. Clements, G. N. 2004. Phonologie. In Les langues africaines, eds. Bernd Heine and Derek Nurse, 49–92. Paris: Karthala. Coveney, A. 996. Variability in spoken French: A sociolinguistic study of interrogation and negation. Exeter: Elm Bank Publications. Cuq, J.-P. 99. Le français langue seconde: origines d’une notion et implications didactiques. Paris: Hachette. Dell, F. 973. Les règles et les sons. 2nd ed. 985. Paris: Hermann. Dister, A., F. Gadet, R. Ludwig, C. Lyche, L. Mondada, S. Pfänder, A. Catherine Simon, and I. Skattum. 2008. Deux nouveaux corpus internationaux du français: CIEL-F (Corpus International et Ecologique de la Langue Française) et CFA (Français contemporain en Afrique et dans l’Océan Indien). Revue de linguistique romane 72: 295–34. Dombrowsky, K. 994. La situation sociolinguistique du sud du Mali (pays minyanka). In Stratégies communicatives au Mali: langues régionales, bambara, français, ed. Gérard Dumestre, 3–88. Paris: Agence de Coopération Culturelle et Technique & Didier Erudition. Dumestre, G., ed. 994a. Stratégies communicatives au Mali: langues régionales, bambara, français. Paris: Agence de Coopération Culturelle et Technique & Didier Erudition. Dumestre, G. 994b. Le trinôme langues régionales – bambara – français. In Stratégies communicatives au Mali: langues régionales, bambara, français, ed. Gérard Dumestre, 3–. Paris: Agence de Coopération Culturelle et Technique & Didier Erudition. Dumestre, G. 2003. Grammaire fondamentale du bambara. Paris: Karthala. Durand, J. 995. Alternances vocaliques en français du Midi et phonologie du gouvernement. Lingua 95: 27–50. Durand, J., B. Laks, and C. Lyche, eds. 2009. Phonologie, variation et accents du français. Paris: Hermès – Lavoisier. Durand, J., and C. Lyche. 2003. Le projet ‘Phonologie du Français Contemporain’ (PFC) et sa méthodologie. In Corpus et variation en phonologie du français: méthodes et analyses, eds. D. Elisabeth and J. Durand, 22–276. Toulouse: Presses Universitaires du Mirail. Durand, J., and C. Lyche. 2008. French liaison in the light of corpus data. Journal of French Language Studies 8: 33–66. Encrevé, P. 988. La liaison avec et sans enchaînement. Paris: Seuil. L’état de l’Afrique 2011. 20. Paris: Jeune Afrique, hors série no. 27. Eychenne, J. 2006. Aspects de la phonologie du schwa dans le français contemporain. Optimalité, visibilité prosodique, gradience. hèse de doctorat, Université de Toulouse-Le Mirail. Gadet, F., and M. C. Jones. 2008. Variation, contact and convergence in French spoken outside France. Journal of Language Contact – THEMA 2: 238–248.  Com pan y 100 Chantal Lyche and Ingse Skattum Un cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing Hamers, J. F. 997. Interférence. In Sociolinguistique. Concepts de base, ed. Marie-Louise Moreau, 78–79. Bruxelles: Mardaga. Heath, J. 2005. A grammar of Tamashek (Tuareg of Mali). Berlin & New York: Mouton de Gruyter. Heine, B., and D. Nurse, eds. 2004. Les langues africaines. Paris: Karthala. Orig. title (2000) African languages: An introduction. Cambridge: Cambridge University Press. Jakobson, R. 97. Studies on child language and aphasia. he Hague: Mouton (94, Kindersprache, Aphasie und allgemeine Lautgesetze, Uppsala: Allmqvist & Wiksell.) LaCharité, D., and C. Paradis. 2005. Category preservation and proximity versus phonetic approximation in loanword adaptation. Linguistic Inquiry 36 (2): 223–258. Laks, B. 977. Contribution empirique à l’analyse sociodiférentielle de la chute des /r/ dans les groupes consonantiques inals. Langue française 34: 09–25. Lyche, C., and I. Skattum. 200a. Le rôle de la L dans le français du Mali: une étude perceptive. In Congrès mondial de linguistique française – CMLF 2010, eds. Franck Neveu et al., 93–926. Paris: Institut de Linguistique Française. Lyche, C., and I. Skattum 200b. Le français contemporain en Afrique et dans l’Océan Indien: usages, variétés et structure. Le Français en Afrique 26: 9–20. Lyche, C., and K. A. Østby. 2009. Le français de la haute bourgeoisie parisienne: une variété conservatrice? In Phonologie, variation et accents du français, eds. Jacques Durand, Bernard Laks and Chantal Lyche, 203–230. Paris: Hermès – Lavoisier. Manessy, G. 992. Norme endogène et normes pédagogiques en Afrique noire francophone. In Multilinguisme et développement dans l’espace francophone eds. Daniel Baggioni et al., 43–8. Aix-en-Provence/Paris: Institut d’études créoles et francophones, Université de Provence/Didier Erudition. Manessy, G. 994. Pratique du français en Afrique noire francophone. Langue française 04: –9. Morin, Y. C., and J. Kaye. 982. he syntactic bases for French liaison. Journal of Linguistics 8: 29–330. Nicolaï, R. 98. Les dialectes du songhay. Contribution à l’étude des changements linguistiques. Paris: SELAF. Nicolaï, R. 989. Songhay et mandé. Mandenkan 8: 69–80. Nikiema, E. 2002. Les (parlers) français ont le schwa, les créoles ont l’r. In Créolistique et linguistique, eds. Claudine Bavoux and Didier de Robillard, 77–96. Paris: Antrophos. Ploog, K. 2008. Subversion of language structure in heterogeneous speech communities: he work of discourse and the part of contact. Journal of Language Contact – THEMA 2: 249–273. Quefélec, A. 2008. L’évolution du français en Afrique noire, pistes de recherche. In La francophonie aujourd’hui. Rélexions critiques, eds. Karin Holter and Ingse Skattum, 63–76. Paris: L’Harmattan. Raymond, G. G., and B. F. Grimes, eds. 2005. Ethnologue: Languages of the world. 5th ed. Dallas, Texas: Summer Institute of Linguistics International. Rosset, T. 9. Les origines de prononciation moderne étudiées au 17e siècle d’après les remarques des grammairiens et les textes en patois de la banlieue parisienne. Paris: Armand Colin. Rossillon, P. 995. Atlas de la langue française. Paris: Bordas. Rottet, K. J. 2005. Variation et étiolement en français cadien: perspectives comparées. In Le français en Amérique du nord, eds. Albert Valdman, Julie Auger and Deborah PistonHatlen, 243–260. Québec: Presses de l’Université Laval. Com pan y he phonological characteristics of French in Bamako, Mali Un cor rec ted pro ofs - J ohn Ben jam ins Pub lish ing Roulon, P. 968. Essai de phonologie du tyembara, dialecte sénoufo. Paris: Bulletin 9 de la SELAF. Sankof, G. 2002. Linguistic outcomes of language contact. In he handbook of language variation and change, eds. J. K. Chambers, Peter Trudgill and Natalie Schilling-Estes, 638–668. Oxford: Blackwell. Selkirk, E. O. 974. French liaison and X-notation. Linguistic Inquiry 5: 573–590. Skattum, I. 997. L’éducation bilingue dans un contexte d’oralité et d’exoglossie: théories et réalités du terrain au Mali. Nordic Journal of African Studies 6 (2): 74–06. Skattum, I., ed. 2000. L’école et les langues nationales au Mali. Special issue of Nordic Journal of African Studies 9 (3). Skattum, I. 2004. L’apprentissage du français dans un pays ‘francophone’: le cas du Mali. In Mandé – Manding. Background Reading for Ethnographic Research in the Region South of Bamako, ed. Jan Jansen, 05–7. Leiden: University of Leiden, Department of Cultural Anthropology and Development Sociology. Skattum, I. 2008. Mali: In defence of cultural and linguistic pluralism. In Language & National Identity in Africa, ed. Andrew Simpson, 98–2. Oxford: Oxford University Press. Skattum, I. 200a. Le français parlé du Mali: une variété régionale? In Voix des français: en parlant, en écrivant, eds. Michael Abecassis and Gudrun Ledegen, 433–447. Bern: Peter Lang. Skattum, I. 200b. L’introduction des langues nationales dans le système éducatif au Mali: objectifs et conséquences. Journal of Language Contact – THEMA 3: 247–270. Skattum, I. 200c. Si j’étais riche… Constructions hypothétiques en français parlé au Mali. Le français en Afrique 26: 49–70. Skattum, I. forthc. L’impact des facteurs sociolinguistiques dans l’identiication de la L à travers le français L2: d’un test de perception au Mali. Paper presented at the conference Hétérogénéité et variation: quels objets sociolinguistiques et didactiques aujourd’hui?, 27–29 May 2009, Université Paul Valéry, Montpellier III. Proceedings under preparation, ed. by Martine Dreyfus, Paris, Michel Houdiard. Wachs, S. 997. Le relâchement de la prononciation en français parlé en Ile de France: analyse linguistique et sociolinguistique par génération. hèse de doctorat, Université de Paris X-Nanterre. Woehrling, C., and P. Boula de Mareüil. 2006. Identiication d’accents régionaux en français: perception et analyse. Revue PArole 37: 25–65. 101 Un cor rec ted pro ofs - Joh nB enj am ins Pub lish ing Com pan y