Contact Linguistics. Chap

is job ke liye
this for
"Many boys were called for this interview but nobody was found suitable for this job."

(84) Life ko face kiijiye with himmat and faith in apane aap.

ACC do courage self
"Face life with courage and faith in self."

Both examples contain a combination of alternation and insertion, and the latter involves both insertion of English elements into a Hindi structure (is job ke liye) and vice-versa (with himmat etc.). Clearly the problem of deciding between insertion and alternation is not restricted to pairs of typologically similar languages. The latter cases, however, do pose a greater problem for deciding language assignment.¹³

Both the "interacting grammars" model of Sankoff & Poplack and the MLF model of Myers-Scotton and her associates would have problems generating these combinations of code switching strategies - the former because it essentially neglects insertion, the latter because it says little about alternation. Clearly though, bilinguals' competence includes both kinds of ability, so the models of code switching should be seen as complementary rather than opposed to each other.
Finally, none of the models of code switching has so far accounted satisfactorily for certain kinds of innovation and restructuring in code switching behavior. Phenomena such as "bare forms" and innovative compound verb constructions as well as certain types of structural convergence (discussed earlier) belong to neither of the languages in contact. Models of bilingual mixture must allow for the creative processes that generate such new structures. Furthermore, the same kinds of innovation can be found in other cases of language contact, thus raising the question whether similar processes and constraints are involved in all these situations. What unifies them is the fact that speakers in all these cases are exploiting both L1 and L2 knowledge to create innovative structures, whether these are new lexical entries, or new structural patterns. Further examples of the creativity of bilinguals will be examined in the following chapter, which deals with new contact languages (so called bilingual mixed or intertwined languages) which have emerged as a result of code switching behaviors.


Summary.

The flexible and varied nature of code switching makes it difficult to characterize the phenomenon in absolute terms. It is clear that there are two broad patterns - one involving alternation of structures from two languages, the other insertion of elements from one language into the morphosyntactic frame of the other. Prototypical examples of the former are alternations between entire sentences, or entire clauses. Examples of the latter include cases in which single morphemes or phrases from language A are embedded into the structure of language B.

Different models of code switching have been proposed to account for these two broad types. Those that attempt to analyze alternational codeswitching propose a grammar that is a combination of the grammars of the languages involved. The code switching grammar of Sankoff & Poplack (1981) is perhaps the best example of this approach. This model appeals to the notion of surface-level syntagmatic equivalence to constrain the kinds of alternation that occur. However, it is clear that this version of the equivalence constraint is inadequate to explain the alternations that do in fact occur across language pairs. A revised version of the constraint which appeals to congruence in categorial status and grammatical function would seem to be better suited to this task.
The dominant model of insertional code switching is the MLF model of Myers-Scotton and her associates. In this approach, code-switched utterances are seen as generated by a single grammar - that of the ML. To some extent, the grammar of the EL plays a role in determining what EL morphemes or structures can be substituted for their ML counterparts. The EL grammar appears to be more fully activated in switches involving phrases or islands. As Myers-Scotton (1993b:36) notes, there is a "switching of procedures" from those of the ML to those of the EL in cases of EL island formation. In this model too, the notion of congruence between EL and ML elements is an essential aspect of the constraints on switching. Insertional code switching is highly sensitive to the subcategorization requirements of ML heads, whether these are verbs, prepositions, determiners or other functional heads. As long as these ML subcategorization frames are preserved, single EL morphemes that match their ML counterparts in semantics and categorial status can be freely substituted for the latter. In fact, even functional items like prepositions, complementizers etc. can be switched, provided their subcategorization properties match across the languages involved.
In addition, entire EL phrases, which may be maximal or partial projections of N, A etc. can be switched, even though they may differ from equivalent ML phrases in their internal structure, provided once more that the switch does not violate the subcategorization frames of the relevant ML heads. Pandit (1990:43) makes essentially this point when she notes that "code switching must not violate the grammar of the head of the maximal projection within which it takes place." This approach recognizes that switching of phrases is regulated by constraints similar to those that govern single-word switches.
An introduction to Contact Linguistics.
Chapter 6. Bilingual Mixed Languages.


1. Introduction.

The previous chapter focussed on bilingual mixture in situations where the two languages involved are maintained, and the mixed code itself has not achieved autonomy as a distinct language. We saw that code-switching phenomena in such situations constitute a continuum of outcomes ranging from relatively simple types of insertion to more complex types of alternation. Such patterns of mixture may persevere for generations without affecting the autonomy of the languages in contact. This seems to be the case in Brussels, Strasbourg, Nairobi and similar communities.

Closely associated with such code-switching is the emergence of “fossilized” mixed codes among bilinguals caught between two languages and their cultures. For instance, Abdulaziz & Osinde (1997) discuss two relatively codes, Sheng and Engsh, used by youths in Nairobi and other cities in Kenya. Sheng employs a Swahili morphosyntactic frame with lexicon from English, Swahili and other mother tongues and is used primarily by less affluent slum dwellers. Engsh is based on English, with words from Swahili and other ethnic languages and is used by more affluent youths. Other examples include Tsotsitaal and Isicamtho, which emerged in the Black urban townships of South Africa. Tsotsitaal is essentially non-standard Afrikaans with a significant mixture of English words, some Zulu words, and a great deal of slang terms (Slabbert & Myers-Scotton 1996:329). Among the young, especially males, Tsotsitaal has now given way to Isicamtho, which employs a Zulu (Nguni) morphsyntactic frame, with a heavy infusion of words from English and Afrikaans. Like Tsotsitaal, its lexicon includes a heavy component of slang, perhaps reflecting the origins of these vernaculars in the argot of criminals and prison inmates. Tsotsitaal translates literally as “the language of the hoodlums and thieves” (Childs 1997:342).
All of the urban vernaculars mentioned above have the same social function of distinguishing their speakers as social groups with their own distinct identity. They mark their speakers as “urban, hip and sophisticated” (Childs 1997:344). Similar mixed codes also often arise among later generations of immigrants who are undergoing shift from their ancestral language to that of the host community. For instance, Agnihotri (1987:108) describes the emergence of a mixed Panjabi/English code used among Sikh children in Leeds, England. The types of mixture range from use of a few English words in an otherwise Panjabi sentence, to frequent alternations of Panjabi and English involving intricate patterns of mixture. Similarly, many persons of Hispanic descent in the United States employ mixed Spanish/English codes as an everyday vernacular and in-group language. Some of these have names of their own, such as “Tex-Mex” in Texas. Especially interesting is the argot known as Pachuco, also called Tirilí or Trilongo, used by gang-members and others in cities like El Paso, Texas. Like Tex-Mex, it has a Spanish morpho-syntactic base, with massive insertion of lexical items from English as well as underworld slang (Craddock 1981:209).
Although they serve important functions as markers of group identity, these mixed codes have not generally achieved autonomy as distinct languages. Thus, Slabbert & Myers-Scotton (1996:328) note that speakers of Tsotsitaal and Isicamtho do not think of them as unitary varieties that can be distinguished from Afrikaanz, English or Zulu.
In this chapter we will examine cases where bilingual mixture results in a new and autonomous creation - a bilingual mixed or "intertwined" language which is distinct from both of its sources, and usually not mutually intelligible with either. These new creations typically arise in situations of ongoing or completed language shift, though they do not all emerge at the same stage of the shift process, a fact which helps explain the differences in the nature and degree of mixture found in them. The styles of code-switching found among immigrant groups who are acquiring the dominant language of their host community provide some interesting points of comparison with these new creations. Code-mixing patterns among these groups tend to develop in a certain order, or hierarchy, as follows (Backus 1996):

simple insertions > more complex insertions > alternations > insertions in the other direction.

This hierarchy shows some parallels with the continuum of intertwined languages that have been studied to date. On the one hand we find creations like Media Lengua which involve patterns of (massive) single morpheme insertion from the L2 to the L1; at the other extreme we find creations like Anglo-Romani, with lexical insertions in the opposite direction, from L1 (Romani) into the L2 (English). Other creations might be placed at different points along this scale. Myers-Scotton (1999:24) suggests a similar sequencing of bilingual code mixture in situations of shift. According to her, the first stage is characterized by classic cs with the ancestral language as the Matrix Language; then a composite ML develops with the L2 contributing more and more of the grammatical apparatus; finally, there is a shift to the L2, which can in turn act as the ML for code-switching. This scenario finds support in Nishimura's account of Japanese-English code-switching, as discussed in the previous chapter, where fluent bilinguals use either language as the ML. The different patterns of cs are also clearly related to cross-generational differences in the degree of proficiency achieved in the L2, as Bentahila & Davies (1990:454) have shown. In their study, older Moroccans who are equally proficient in Arabic and French show a strong preference for inter-sentential alternation or for cs in which French is the ML, with insertion of Arabic function words. By contrast, younger Moroccan bilinguals who are Arabic-dominant show a strong preference for intra-sentential cs, with Arabic as the ML. If these tendencies are constant across bilingual situations, we would expect them to be reflected in the differences among the newly-created mixed languages that arise at different stages of shift. As we shall see, the facts bear this out.

2. Definition and classification.

There is general consensus that bilingual mixed languages are composites of materials drawn from just two languages. In what are perhaps the prototypical cases, the grammar is derived primarily from one of the languages, and the lexicon primarily from the other. Examples include Anglo-Romani and Media Lengua (a blend of Spanish lexicon and Quechua grammar). However, the neat separation of grammar and vocabulary by respective source language is not found in all cases. There are cases where structural materials are derived from both sources, as in Michif, which combines French NP structure with Cree VP structure. Whatever the nature of the mixture might be, there is agreement that the resulting outcome is a new creation, distinct from either of its source languages.

It is generally easy to identify the sources of the components, particularly in cases where the source languages are typologically quite different, as in Media Lengua. But even in cases where the two languages are genetically related and hence typologically similar (e.g, Chindo, a blend of Malay and Javanese, both Austronesian, spoken by Paranakan Chinese in Indonesia) it is still possible to identify the precise sources of the components (see Dreyfuss & Oka 1979). Thomason (1997c:80) sums up the characteristics of bilingual mixed languages as follows:

15. 
    They evolve or are created in two-language contact situations.
    
16. 
    The setting involves widespread bilingualism on the part of at least one of the two speaker groups.
    
17. 
    In the resulting mixture the language material is easily separated according to the language of origin.
    
18. 
    There is little or no simplification in either component of the mixed language (reflecting the bilingualism of its creators).
    

There is somewhat less agreement on the classification of bilingual mixed languages and on the historical and linguistic processes involved in their creation. As we shall see, the two issues are related.

Thomason (1995) suggests a classification based simultaneously on socio-historical and linguistic (structural) criteria. According to her, there are two broad types of bilingual mixed languages. Type 1 is the language of a "persistent ethnic group" that has resisted total assimilation to a dominant language group, though subject to "overwhelming cultural pressure" (1995:19). As a result, their language undergoes a long process of lexical and structural “borrowing” from the dominant language, resulting in pervasive, massive mixture in all components, and in the extreme cases, the replacement of large portions, or even all, of the inherited grammar. Examples of mixed languages with pervasive mixture though not total structural replacement include Kormakiti Arabic (an Arabic/Greek mixture spoken in Cyprus), Asia Minor Greek (a Greek/Turkish mixture) and perhaps Wutun (a dialect of Chinese heavily influenced by Tibetan). Examples of mixed languages in which grammatical replacement has occurred include Caló (Romani lexicon with Spanish grammar), Anglo-Romani (Romani lexicon with English grammar) and Ma'a (Mixed Cushitic/Bantu lexicon with mostly Bantu grammar).
Thomason's second category includes languages which emerge rapidly, often within a single generation, as the vernaculars of new social groups or communities that want a language of their own to symbolize their separate ethnic identity. Examples include Media Lengua, Michif and Mednyj (Copper Island) Aleut. Structurally, these BML's differ from those in Category 1 in that the latter show "across the board linguistic effects" while the former don't. In other words, languages in Category 1 display evidence of heavy influence from the dominant group's language in all aspects of structure, grammar as well as lexicon. By contrast, languages in Category 2 display a clear compartmentalization of components, or as Thomason & Kaufman (1988:107) put it, "categorial specificity of the structural borrowing". Thus Michif draws its VP structure (almost) entirely from Cree, and its NP structure from French, and both components are more or less intact on all levels (phonology, lexicon, morphology and syntax). Thomason's classification is presented in summary in Table 1.


Table 1. Thomason's (1995) classification of bilingual mixed languages.
Examples Socio-historical characteristics Linguistic processes.


Category 1

Kormatiki Persistent ethnic group resisting Pervasive linguistic influence

Arabic assimilation to dominant group from L2 on all aspects
language. of the ethnic language.
Asia Minor Overwhelming cultural pressure Massive replacement
Greek from dominant language. of L1 grammar.

Ma'a Gradual long- term change. Near-total replacement

& Language shift may or may of L1 grammar.
Anglo-Romani not occur.
Category 2

Media Lengua Newly-emerging social group Clear compartmentalization

& which creates a language of its of components from the
Michif. own as symbol of its identity. two sources. Structural
features specific to
a particular source

Mednyj Aleut Abrupt creation, sometimes Structural components

within a single generation. preserved more or less intact,
with minor simplification or
restructuring.
Bakker (1994) takes issue with Thomason and offers a quite different classification of bilingual mixed languages based, first, on the nature of the contact settings and the groups who created the mixed code, and second, on the nature of the mixtures that result. He challenges the validity of Thomason's Category 1 and her view that some bilingual mixed languages arise as a result of extreme structural borrowing. In his view, a distinction should be made between languages like Kormakiti Arabic and Asia Minor Greek which are the result of extreme structural diffusion from an external source language, and languages like Ma'a and Anglo-Romani. In Bakkers' view, the latter are not the result of massive grammatical replacement in Thomason's sense, but rather arise via a process of "language intertwining," that is, the combination of the lexemes of one language with the grammatical system (phonology, morphology, syntax) of another" (1994:20). As Bakker notes, unlike “borrowing”, the term does not suggest any direction of the process, but emphasizes the act of language creation. Only outcomes that arise in this way qualify as bilingual mixed languages in Bakker's classification. Hence creations like Anglo-Romani and Ma'a are essentially similar in their genesis and make-up to languages in Thomason's Category 2 such as Media Lengua and Michif, though they may differ in terms of their sociohistorical background.
It seems justifiable to distinguish cases of extreme structural borrowing from cases of language intertwining, even though both result in extreme mixture. As we saw in the previous chapter, languages like Asia Minor Greek, Wutun, etc., involve such pervasive mixture within every component of the grammar that it is well nigh impossible to separate the contributions from the source languages into neat discrete categories. Neither the lexicon nor (any component of) the grammar of these languages constitutes a homogeneous entity that derives primarily from one source. This contrasts with cases of language intertwining, where in general, the components are preserved relatively intact, and differ little from their counterparts in the source languages. This remains true even when the lexicon and/or the grammar are blends of material from different sources, as in Michif or Mednyj (Copper Island) Aleut (see section 7.3 below).
Henceforth, therefore, we will follow Bakker in restricting the designation “bilingual mixed” or “intertwined” languages only to those outcomes that satisfy the criteria just mentioned. We will use both of these designations to refer to these outcomes of language contact. Genetically, these mixed languages “cannot be classified as belonging solely to family A or solely to family B” (Bakker & Mous 1994:5). Rather, they are related to both.
But even this definition has its problems. As Bakker & Mous (ibid.) suggest, the genetic classification of bilingual mixed languages "is a matter of degree, and it is not a priori clear what qualifies as a mixed language and what as a case of extreme borrowing." They suggest that the two types can be distinguished according to the extent of the lexical input from the L2 or external source language. First, in cases of extreme borrowing, lexical borrowing never exceeds roughly 45% of the whole lexicon. In bilingual mixed languages, the proportion of "foreign" lexical elements is usually closer to or over 90%. Second, in extreme borrowing, foreign elements affect the core vocabulary only to a limited degree. In bilingual mixed languages, most of the core vocabulary tends to be from the external source language. In general, however, bilingual mixed languages are recognizable only if their history and their sources are known. Certain clues facilitate their identification. For example, the presence of sufficient bound morphology helps to establish that the source of the grammar is different from the source of the lexicon. Source languages are easier to identify when they are typologically distant, though some BML's arise from contact between genetically related languages (eg. KiMwani (2 Bantu languages) and Stedsk (2 Germanic languages) (Bakker & Mous, ibid.).
Bakker’s (1994:24) classification of bilingual mixed languages is given in Table 2. He distinguishes two broad categories according to socio-historical criteria, that is, the social settings and circumstances in which they arose. Category 1 includes languages created by settled ex-nomadic groups who need a secret language for communication among themselves, especially in the presence of strangers. Such languages typically have the lexicon of the original ethnic language and the grammatical system of the "host language" (1994:24). Examples would include Anglo-Romani, Caló (a blend of Spanish grammar and Romani lexicon spoken in Spain, and other Romani-lexicon mixtures (see Boretzky & Igla 1994). Also included is Callahuaya, a mixture of mostly Puquina lexicon and Quechua grammar spoken by itinerant healers in Northwestern Bolivia (Muysken 1997b). A final example is Krekonika, a blend of Greek grammar and Arvanitika vocabulary, used as a secret language by masons of the Peloponnese in an area of Greece where Arvanitika used to be spoken (Konstantinopulos 1983). All of these outcomes seem to have originated in the last stages of shift from the original inherited language to the host language.
Category 2 includes languages which arise in mixed households involving men speaking language A who "invade" the territory of language B and marry local women (ibid.). Examples include Michif, Island Carib, Chindo, etc. Bakker's classification is presented in Table 2.
Table 2. Bakker's (1994:24) classification of "intertwined" languages.

Examples Socio-historical characteristics Linguistic characteristics

Category 1

Anglo-Romani, Created by ex-nomadic groups L2 provides the grammar,

Caló, who need a secret language. L1 provides most of
Callahuaya, the lexicon.
Ma'a, etc.
Arise during shift from L1
to L2
Category 2.

Michif Arise in mixed households Grammar derived from

Island Carib with immigrant men and mothers' language (L1)
Chindo, Petjo, local women lexicon, from fathers' (L2).
Javindo, Ilwana,
etc.
It’s debatable whether Bakker's classification accounts adequately for all known cases of language intertwining. Presumably, Ma'a would fit into Category 1, while Mednyj Aleut would fall into Category 2, at least on socio-historical grounds. Media Lengua might be included in Category 2 on linguistic grounds, but not by the socio-historical criteria (it didn’t arise in mixed households). Note also that languages placed in different categories according to socio-historical criteria may be quite similar in their makeup. For example Island Carib (Category 2) is a mixture of L1 lexicon and L2 grammar, similar to Anglo-Romani or Ma'a (Category 1). It is clear that classifications of languages based on socio-historical criteria do not match up exactly with those based on purely structural criteria. This reflects the fact that intertwined languages do not all conform to some ideal prototype, but rather constitute a varied assortment of outcomes, with different histories and structural characteristics. We will return later to the issues of genesis and structural change as they relate to these creations.
In the following sections, we will examine four well-known outcomes of language intertwining to provide some idea of the variety of ways in which two languages can combine to produce a new creation.


3. Media Lengua.

Media Lengua is among the best known of the BML's and indeed is often taken as a prototypical exemplar of this class of contact vernaculars. The language is a blend of predominantly Quechua grammatical structure and Spanish-derived lexical forms that make up about 90% of the vocabulary (Muysken 1981a:52). The language is spoken in several small towns or village communities in the central Ecuador highlands. It functions as an in-group language among Indian peasants, craftsmen and construction workers, particularly among younger men who work in the nearby capital city Quito in industry and construction. One of the core communities where it is spoken is the town of San Miguel de Salcedo (or simply Salcedo) in the Cotopaxi province of Ecuador. Muysken (1997a:374) notes that Media Lengua is the native language of younger adults and most children in the communities nearest to Salcedo. These villages are situated both geographically and culturally between the world of the urban centers in the valley and the Indian world of the mountain slopes (ibid.). Here Spanish is used for communication with the non-Indian world, Quechua for contact with the traditional Quechua-speaking mountain communities, and Media Lengua for everyday interaction within the villages. Both culturally and linguistically, Media Lengua is "half-way between Quechua and Spanish" (Muysken 1981a).

The language appears to have come into being quite recently, perhaps between 1920 and 1940. Muysken explains its genesis as due to the fact that "acculturated Indians could not identify completely with either the traditional rural Quechua culture or the urban Spanish culture" (1981a:75). Hence they created Media Lengua as a means of expressing their separate group identity. The same motivation appears to lie behind the emergence of intertwined languages in general.
3.1. Structural characteristics.

Media Lengua emerged within the context of the long period of contact between Quechua and the colonial language, Spanish, from the Spanish colonization of Ecuador around 1540 to the 20th century. This contact led, as would be expected, to mutual lexical borrowing between the languages and to some degree of structural change in the second language varieties of Spanish acquired by native speakers of Quechua. It is clear, however, that Media Lengua is neither Quechua with heavy Spanish borrowing, nor Spanish subjected to substratum influence from Quechua. The language is a quite distinct creation, unintelligible to speakers of either of its source languages. The following are examples of Media Lengua sentences from Muysken (1981a:68-69) that illustrate its unique blend of features. Spanish-derived items are italicized.

(1) a. ML Unu fabur-ta pidi-nga-bu bini-xu-ni.

one favor-ACC ask-NOM-BEN come-PROG-1sg.
"I come to ask a favor." "I come to ask a favor."

b. Q Shuk fabur-ta maña-nga-bu shamu-xu-ni

one favor-ACC ask-NOM-BEN come-PROG-1sg

c. Sp. Vengo para pedir un favor.

I-come for ask-INF a favor.

(2) a. ML No sabi-ni-chu Xwan bini-shka-da

NEG know-1sg-NEG John come-NOM-ACC
"I don't know that John has come"

b. Q. Mana yacha-ni-chu Xwan shamu-shka-da.

NEG know-1sg-NEG John come-NOM-ACC

c. Sp. No sé que Juan ha venido.

NEG I-know that John has come.

As can be seen in these examples, Quechua grammar is preserved intact in Media Lengua, as reflected in the SOV word order and the rich array of verbal morphology, including affixes that mark topics, objects, person/tense, negation, subordination and so on. On the other hand, the stems to which such suffixes are attached are Spanish-derived. However, it would be an oversimplification to say that Media Lengua consists of purely Quechua grammar with exclusively Spanish-derived stems. There are several respects in which the Quechua grammar has been modified, mostly under Spanish influence. Spanish-derived items include not just content items like nouns, verbs and adjectives etc, but function items like prepositions, conjunctions as well as other closed-class items such as personal pronouns and question words. In addition, some stems are preserved from Quechua. On the whole, though, it is clear that Media Lengua is the result of the incorporation of Spanish phonological shapes into a Quechua morphosyntactic frame - a process that involved both adaptation of the Spanish elements to Quechua structure and some structural change in the Quechua grammar under Spanish influence.

3.2. Adaptation of Spanish items to Quechua structure.

Muysken (1981a:54-60) discusses several respects in which Spanish-derived items are adapted to Quechua patterns in phonology, morphology and morpho-syntax. First, Spanish sounds are replaced by their closest perceived Quechua counterparts - a process of "phone substitution" common in both borrowing and second language learning situations. The following are some examples of changes in vowels:

Spanish ----> Quechua Examples.

e i decir > dizi "say"

o u poder > pudi "can, be able"
ie [ye] i bien > bin "well"; siete > siti "seven"

With regard to consonants, Spanish [v] changes to [b] as in breve > brebe "quickly." The voiced consonants [b, d, g] which are allophones of [p, t, k] respectively in Quechua, occur freely in Spanish-derived stems. Spanish intervocalic [s] becomes [z] in Media Lengua, as in hacer > azi "do, make". Spanish initial [fw] becomes [xw], as in fuerte > xwerte "strong."

The agglutinative morphology of Quechua allows suffixes to be attached to stems relatively freely, without morphophonemic alternations. Stems are easily isolatable from their suffixes, hence easily substitutable by corresponding Spanish forms. This makes for easy incorporation of both nouns and verbs, and explains why Media Lengua has been able to replace most Quechua open-class items with equivalent Spanish forms. For the most part, this process has had no impact on Quechua morphology. Spanish stems behave in all respects like Quechua stems and are fully integrated into the Quechua morphosyntactic frame.
Spanish forms are adapted in other ways. For instance, adverbs often undergo reduplication, as in the following example from Muysken (1997a:384):

(3) Yo-ga bin-bin tixi-y-da pudi-ni

I-TOP well-well weave-INF-ACC can-1sg.
"I can weave very well."

This appears to be an innovation in Media Lengua, since there is no equivalent pattern of reduplication in Quechua. Spanish forms have also undergone regularization and simplification of the sort found in informal second language learning. For instance Spanish irregular verbs are regularized, eliminating the morphological variation found in Spanish. Thus dar "give" is always realized as da in ML, whereas in Spanish it is realized in the present as doy (1st sing.) das (2nd sing.) da (3rd sing), damos (1st pl), dais (2nd pl.) and dan (3rd. pl.), and also has different realizations in the past. Other verbs that have been simplified in this way include ir > i "go", ver > vi "see"; hacer > azi "do, make"; and saber > sabi "know." These stems remain invariant, with Quechua suffixes expressing differences in tense, number, person etc. Other changes in Spanish forms include what Muysken calls "freezing", ie, the blending or fusion of separate Spanish words into a single Media Lengua word, as in the following examples:

(4) Spanish Media Lengua Meaning.

no hay núway "There is no.."

aún no aúnu "not yet"

Another instance of simplification of Spanish forms is to be found in the pronominal system. Muysken (1997a:394) offers the following comparison of personal pronouns in the three languages:

Table 3. Pronouns in Media Lengua and its source languages.


Person Quechua Media Lengua Spanish (Nom/Oblique)

1sg: ñuka yo/ami + case yo/me/mi “I/me”

2sg: kan bos tu/te/tí "you (intimate)"
vos/te "you (familiar)"
usted/le "you (polite)"

3sg. pay el él/le "he/him"

ella/le "she/her"

1pl. ñukunchi nustru nos(otros) "we/us"

2pl kan-guna bos-kuna ustedes/les "you (pl.)"

3pl. pay-guna el-kuna ellos/les "they/them (masc.)"

ellas/les "they/them (fem.)"

Note how Media Lengua has eliminated all of the morphological irregularity in the Spanish pronouns, reducing each person to a single form in both singular and plural. Some of the Media Lengua pronouns are direct imports from Spanish (yo, bos, etc.) while others are combinations of Spanish forms and Quechua affixes (bos-kuna, el-kuna, etc.).

There have also been some interesting kinds of adaptation in Media Lengua derivational morphology, involving a compromise between Spanish and Quechua strategies. Quechua tends to use derivational suffixes to create new verbs from a small set of roots, whereas Spanish tends to have separate verbal roots for the same concepts. The Quechua compositional forms are replaced in Media Lengua either by Spanish simplex forms, or by hybrids containing a Spanish root and the Quechua derivational suffix. Muysken (1981a:59) offers the following illustration of the way concepts associated with the Quechua verb riku- "see" and its derived forms are represented in Media Lengua.

(5) Meaning Quechua Spanish Media Lengua

"see" riku-ø ver bi-

"appear" riku-ri se ver bi-ri-n

asomar asoma-
aparecer parisi-

"show" riku-chi hacer ver bi-chi

mostrar mustra(-chi)

"stare" riku-ra espiar bi-ra

chapar "spy' chapa(-ra)

Muysken (1997a:388-93) discusses other examples of adaptation and compromise in semantics and structure in the Media Lengua lexicon.

Despite the massive incorporation of Spanish forms into Media Lengua, we find little evidence of the incorporation of Spanish morphology. The few exceptions include the gerundive suffix -ndu < -ndo, as in trabajando < trabajar "work"; the dimunitive suffix -itu/ita as in muchachito/a < muchacho/a "boy/girl", and the past participle -do as in trabajado < trabajar "work". The latter two occur in Quechua as well. The dimunitive suffix is of course a derivational affix, and can be seen as a borrowing in both Quechua and Media Lengua. Borrowing of derivational affixes is by no means rare or suprising. The past participial suffix -do appears only in Spanish adjectival forms (e.g., cansado "tired" < cansar "to tire") that were adopted as wholes. Hence this is not a case of borrowing of a productive inflectional morpheme. The gerundive suffix -ndo, on the other hand, appears to be a true structural borrowing, though its function in Media Lengua is modelled after that of the Quechua nomalizing suffixes -sha and -kpi which are used as "same subject" and "different subject" subordinators respectively in Quechua embedded clauses. Examples of the former use in Quechua and Media Lengua are as follows (Muysken 1981a:73-74):

(6) a. Q. Chaya-sha-mi miku-sha

arrive-NOM-AFFIRM eat-1stFUT
"I'll eat when I arrive."

b. ML el-kuna-ga asi nustru abla-ri-k-ta-s uye-ndu-ga aprendi-n.

3p-PLU-TOP thus we talk-REFL-NOM-ACC hear-NOM-TOP learn-3p
"They learn it when they hear what we speak."

There is also some incorporation of free Spanish function morphemes into ML. These include prepositions like de "of", entre "between, among", por "for" and others, as well as conjunctions like pero "but." These are exceptions to the general rule that only those lexical categories found in Quechua are to be found in Media Lengua. Quechua has only postpositions, and its conjunctions are clitics. Hence these borrowings do not follow the usual pattern of relexification, but rather introduce alternative case-marking and conjoining strategies on a limited scale. The Quechua case-marking and conjoining systems are preserved in the vast majority of cases.

Finally, Media Lengua syntax is overwhelmingly Quechua in character, with relatively little impact from Spanish. Like Quechua, Media Lengua is head-final, except for the few cases of prepositional borrowing mentioned above, and some instances of VXP order apparently due to Spanish influence. In some cases we also find Noun+Adjective order (as in Spanish) rather than the Adjective+Noun order of Quechua. Another minor change is found in comparative structures where the Quechua serial verb yalli "surpass" is relexified with Spanish gana "win", but the latter is uninflected, unlike the former. There has also been a slight change in embedded questions, where we find regular inflections on the verb instead of the Quechua strategy of nominalization. Media Lengua has also adopted a few Spanish complementizers (e.g., ki < que "that") and subordinators like porque "because" and aunke < aun que "even if", thus introducing a COMP-initial strategy as opposed to the COMP-final strategy found in Quechua. In general, however, Media Lengua preserves the Quechua system of nominal and adverbial marking of verbs as its subordination strategy in embedded clauses. The structural borrowings from Spanish remain marginal to the overwhelmingly Quechua character of Media Lengua syntax.


3.3. Processes of change: the Relexification Hypothesis.

We might now consider what processes of contact-induced change were responsible for the creation of this language. Muysken (1981a:61) proposed that Media Lengua arose via a process of "relexification - the process of vocabulary substitution in which the only information adopted from the target language in the lexical entry is the phonological representation." He distinguishes this process from "translexification", that is, "the process of vocabulary substitution in which, in addition to the phonological representation, all other levels of information are adopted from the target language as well" (ibid.). He notes that the boundaries between these two processes are quite fluid, and there is a continuum of possible results between the two. Lefebvre (1996:234-35) suggests that relexification involves two steps - first, the copying of the lexical entries of an established lexicon (in this case the L1 of the creators) and second, the replacement of the original lexical entry's phonetic representation by a phonetic string drawn from another language. The process is illustrated in Figure 1 (I have adapted Muysken's (1981a:61) and LeFebvre's (1996:235) illustrations somewhat).

Figure 1. The process of relexification.