Contact Linguistics. Chap

Exercise: Compare the kinds of language mixture found in Media Lengua and Michif with those found in the code mixing of Japanese/English bilinguals as described by Nishimura (1986; or 1997, 95-114). What similarities and differences do you find? See the discussion of Nishimura’s data in Chapter 5.


4.8. Mechanisms and processes in the genesis of Michif.

Scholars have debated whether Michif is the product of code-switching, structural borrowing, relexification, or some other process such as "language intertwining". None of these terms seems to account adequately for the genesis of the language. The problem seems to reside in a failure to distinguish the actual mechanisms of change involved from their manifestations as different kinds of bilingual mixture.

For instance, the fact that Michif is reminiscent of some kinds of code switching simply implies that its creation involved processes similar to those that produce code switching behavior. As we saw in the previous chapter, code switching itself is not a “mechanism” but rather a form of behavior in which certain strategies of mixture come into play. There is a similar problem with the concept of “language intertwining”, mentioned earlier, which Bakker (1994:24) defines as a distinct process in its own right. But this label just describes the outcome, not the processes involved, and does not account for the very different patterns of structural and lexical mixture that are found in different intertwined languages.
The notion of “relexification” also seems ill-suited to explain the mechanisms of Michif formation. The term is more appropriate for the process of single morpheme insertion into a maintained ML frame that we saw in Media Lengua. However, there are parallels between Michif and other patterns found in code-switching. The incorporation of French NP’s and PP’s into the Cree morpho-syntactic frame of Michif is similar to constituent or EL island insertion. The main difference is that Michif restricts this process to French NP’s (and some prepositional phrases), and of course applies it consistently, at least for NP’s. But even this account is somewhat simplistic, since French NP structure is not imported intact, but modified so as to conform to Cree morphosyntax and semantics.
There is also the question why this particular pattern of insertion, so different from that in Media Lengua, is found in Michif. What motivates the incorporation of French elements into Cree NP structure, rather than its VP structure? Bakker (1997:224ff) suggests that the typological relationship between French and Cree can explain this. Whereas French is inflectional, Cree is a polysynthetic language with simple noun derivation and highly complex verbal structure. Since neither Cree nor French noun morphology is very complex, and since noun stems can be easily isolated in both, this facilitates the substitution of Cree noun stems by their French counterparts. The latter can therefore take Cree nominal affixes.
As we saw in chapter 5, the primary constraints on this kind of constituent substitution are first, there must be congruence of function between the relevant constituents, and second, the subcategorization requirements of the relevant ML heads must be met. So, while Michif may not appear identical to the products of intra-sentential code switching. the mechanisms and constraints involved in both kinds of mixture appear to be the same.
By contrast, Cree verbs are highly complex in structure, consisting only of bound grammatical elements. This applies even to verb roots, which cannot be isolated easily. The following example from Bakker (1997:228) illustrates:

(18) nôhtê-wâp-am-ito-w-ak

want-look-AN-REC-3-PL
“They want to see each other.”
(NB. AN = Animate; REC = Reciprocal)

It is difficult to separate stems from affixes here. Even apparent roots like –wâp– never occur in isolation. Hence it is difficult to identify a specific stem that might be substituted with a French verb (Bakker 1997:229).

Here is another example from Bakker, (p. 230), illustrating a complex verbal form with initial, medial and final elements.

(19) mihkw-âpisk-isw-êw

INIT-MED-FINAL-INFL
red-stone-by.heat-3-->3
“He makes the stone red hot”
(NB. 3--->3’ = 3^rd p. subject, 3^rd p object)

Here again, even the stems themselves are combinations of bound forms that are not easily substitutable by French forms. The very different typological structure of the Cree VP would therefore constitute a strong barrier to the incorporation of French elements. This explanation is supported by the fact that similar constraints seem to operate in other cases of contact between Algonquian languages similar to Cree and European languages like English and French. Bakker (1997:181f) discusses instances of Cree/English and Naskapi/English code mixing from Saskatchewan and Montréal (Quebec) respectively. He notes that they all display a similar pattern of insertion of English nouns, noun phrases and prepositional phrases into the Algonquian frame, with the VP left intact. This is also true of the pattern of Montagnais/French code mixing found in the community of Betsiamites, Quebec. In this case, however, French verbs are also incorporated, along with French nouns, prepositions, etc. The verbs are inserted as infinitival forms that combine with a helping verb -tut- “do”, which bears all necessary inflections. This strategy is quite common in borrowing and code switching behavior, as we saw in chapter 5. Its occurrence in Montagnais/French code mixing suggests that the strong structural constraints on the incorporation of verbs that we mentioned earlier can be overridden.

In general, all these examples suggest that contact involving different pairs of languages with similar typological distance will lead to similar patterns of mixture (Bakker 1997:186). Presumably this is because the same constraints apply in all cases.
Finally, there is need to account for the modifications in French NP structure as well as the intrusion of some French structural apparatus into the Cree syntactic frame of Michif. The processes involved in the first case seem quite similar to those that characterize the integration of elements introduced into a recipient language from a foreign source. Similarly, the incorporation of French function morphemes such as prepositions, copulas, modals etc., has parallels in both “borrowing” and code-switching behavior. All of these French elements are free forms that could easily be accomodated within the Cree grammar. This is in keeping with the view that Michif was created by Cree/French bilinguals who inserted French elements into a maintained Cree grammatical frame.


5. Creations associated with language shift.

The cases we have examined so far all involve maintenance of basic L1 structure, with incorporation of features from an external source. But several bilingual mixed languages appear to have arisen after a process of shift in which an original ethnic language is replaced by a host language. The newly acquired host language is then used as the grammar-source or matrix language into which (mostly) lexical material is incorporated from the original ethnic language, thus creating a new intertwined language. Many of the languages that arose in this way were created by nomadic groups who used them either as secret in-group languages or as symbols of their separate ethnic status. As noted earlier, they include Anglo-Romani and other mixed Romani varieties such as Caló and Armenian Romani as spoken by various Roma or Gypsy groups, Ma'a or Mbugu, Krekonika and Callahuaya. Other intertwined languages have been documented for nomadic groups in Afghanistan, Ireland and Scotland (Hancock 1984a), India, the Middle East (Kenrick 1976-77) and elsewhere.

The genesis of these shift-induced mixed languages has also been the subject of some disagreement. Some researchers (e.g., Thomason 1995) attribute their formation to massive grammatical borrowing or replacement. Others, (e.g., Boretzky 1985) argue that they arose when the ethnic group shifted to the language of the host community but retained most of the vocabulary of their original language. Before we decide on the merits of either position, let us examine one such outcome, the language known as “Ma’a” or “Inner Mbugu.”


6. The case of Ma'a.

Ma'a is spoken in several communities in the Usambara mountains of Northeastern Tanzania by groups who migrated to the region several hundred years ago. Almost all of its grammar is Bantu, and closely similar to that of Pare, a neighboring language spoken by immigrants from the Pare mountains. Roughly half of its lexicon comes from (mostly) Southern Cushitic, while the rest is drawn from Bantu languages. The group refers to itself as the Ma'a - a term which is also used for the mixed language they speak. Outsiders refer to them as the Mbugu, which is also the name of the Bantu language they speak. Ma'a, or in-group (inner) Mbugu, is identical to Mbugu except for the lexicon and some minor structural differences.

6.1. Historical background.

Oral tradition has it that the Mbugu were originally from Lukipya (present-day Lackipya in Kenya). To escape harassment by the Masai, they apparently migrated to the northern Pare mountains, then to the south Pare mountains, from where many later moved to the Usambara mountains. It is generally agreed that the Mbugu originally spoke some variety of (Southern?) Cushitic, and preserved it for a long time before shifting to Pare (or in some cases to Shambaa, the dominant language of the Usambara region). Those who call themselves Ma'a apparently resisted assimilation longest, and created a mixed language as a sign of their resistance and their autonomy as a distinct ethnic group. The language is considered incomprehensible by their neighbors.

6.2. Structural features.


6.2.1. Lexicon.

Roughly half of Ma’s lexicon is Cushitic, according to Ehret (pc cited in Thomason 1997b:475). The basic vocabulary is primarily Cushitic, and much of its cultural vocabulary also comes from that source. The rest of the vocabulary is drawn from different Bantu languages such as Pare, Shambaa and Swahili, as well as from Maasai of the Eastern Nilotic family, and includes many verbs, some body parts and cultural words. In certain semantic fields such as color, the lexical semantics of Bantu-derived words is Cushitic. This particular pattern of intertwining of basic and cultural vocabulary from two primary sources is probably unique among bilingual mixed languages.

6.2.2. Grammar.

The combination of Bantu grammar and Cushitic vocabulary typical of Ma'a is illustrated in the following example from Mous (1994:176). Non-Bantu items are in italics.

(18) hé-ló i-?azé i-wé áa-sé va-ma?a na va-sitá

16-have 5-day 5-one 1:PAST-call 2-Mbugu with 2-Shambaa

na va-?ariyé vá-so vá-zé-m-hand-í-ya ma-gerú ku?u

with 2-Pare 2:SUBJ-go 2:SUBJ-IT-01-plant-APL-F 6-bananas his.

"On a certain day, he called the Mbugu, Shambaa and Pare people to go

and plant his banana trees."

(Note: APL = applicative; IT = "Itive"; F = final vowel).

It seems clear from this example that Ma'a is for the most part structurally identical to Mbugu, which in turn is practically identical to Pare. The exceptions are minor, and generally marginal to the grammar. In most cases, these non-Bantu structural elements derive from Cushitic. For instance, the phonemic inventories of Ma'a and Mbugu are identical except for a few non-Bantu phonemes in the former. These include voiceless lateral fricative /hl/, the glottal stop /?/, the voiceless velar fricative /x/ and the complex sound /n¡x/ (voiceless nasal + voiceless velar fricative). As might be expected, these are found only in Ma'a words of non-Bantu origin (Mous 1994:197).

Ma'a nouns and verbs have all of the inflectional and derivational morphology of Pare, again with minor differences. For example, many Cushitic-derived nouns lack a noun-class prefix, while both Ma'a and Mbugu have a noun class (14:2) which Pare lacks. Some Ma'a verbs ending in -a fail to inflect the final vowel, unlike Mbugu and Pare.
Some of the closed classes of Ma’a show more significant non-Bantu influence. There are three non-Bantu demonstratives that show no agreement with their nouns, unlike those in Mbugu. The precise source of these forms is unclear. Ma'a also has different personal and possessive pronouns than Mbugu. Moreover, possessives fail to show agreement in Ma'a as they do in Mbugu. One or two pronominal forms show some slight resemblance to forms in Cushitic languages, but in most cases a specific source language cannot be established. Note however that despite the formal differences, both demonstratives and personal pronouns in Ma'a show the same semantic distinctions as those in Mbugu, differing somewhat in that respect from Cushitic languages.
Ma'a syntax generally follows Bantu patterns, with SVO and noun-adjective order, as well as prepositions, where Cushitic has dominant SOV and adjective-noun order and postpositions. Interestingly, some Ma'a prepositions (e.g., he "to" and na "from") derive from Cushitic lexical items. In cases where Bantu and Cushitic are typologically similar, Ma'a may employ constructions from both. This is true of genitive and copula constructions, for example. The copula is obligatory in Ma'a, following Cushitic and unlike Bantu, which frequently omits this marker. Other Ma'a features due to Cushitic influence include its large class of adjectives (Bantu has few of these) and the use of a verb -lo "have" to express possession. For the latter, Bantu uses a pronoun-connective combination (e.g., a-na "he-with") in present contexts, and a form "to be with" for other tenses/aspects (Thomason 1983a:214-16).
Finally, several derivational suffixes in Ma’a are of Cushitic origin. Some of these are apparently still productive, such as causative –ti (attached to verbs) and “amplicative” -∫a (attached to both nouns and verbs) (Thomason 1983a:214). Others have ceased to be productive, while a few survive as fossilized forms. At the same time, however, many Bantu affixes have become part of the Ma’a derivational system (ibid.).