Abstract
This article aims to explore the psycholinguistic processing issues, in terms of the type of transfer that they exemplify, needed to account for in the emergence of two mixed languages and a mixed register with a Quechua structure: Media Lengua (Ecuador) and Kallawaya (Bolivia), both relexified varieties within the Quechua language family, and bilingual mixed songs in Peru, waynos. The two issues that require most attention are (1) the mental status of roots vs affixes in the transfer process; (2) the possibility of manipulating lexical access in transfer. The languages and the register share a number of structural features, but are sociolinguistically totally different. In Media Lengua the lexicon comes from a ‘new’ language (Spanish), and in Kallawaya from an ‘old’ language (Puquina). Media Lengua is an informal community language, while Kallawaya is a ritual healing language only used by male adults. Waynos are a very popular musical genre in large parts of the southern Andes in Peru. The root/suffix asymmetries in the mixed languages are confronted with the mirror phenomenon of Spanish suffixes that occur in Quechua, to help us further understand the processing issues involved.
1. Introduction
In the large area in the South American Andes where members of the Quechua language family are spoken, several interesting contact varieties have emerged. In the north, in Ecuador, these comprise varieties of Media Lengua, where Spanish roots are inserted into Quechua morphosyntactic and lexical frames. In the centre, particularly in southern Peru, intense mixing is apparent in a specific register, the bilingual songs named waynos. In an area in the south, we find Kallawaya, a ritual healing language used only by male adults, in which lexical roots mostly from an ancestral language (Puquina) are introduced. Throughout the area, Spanish lexical elements, but also suffixes, have been transferred into Quechua.
Primarily drawing upon, expanding and synthesizing my own work in this area, I compare and contrast these varieties and discuss their relevance for the transfer debate. In Table 1 a number of features of these varieties are presented and contrasted, including their status with respect to the two distinctions introduced in the work of Grosjean (this issue): dynamic interference vs static transfer, and monolingual vs bilingual language mode. It also briefly characterizes them in terms of the MAT (morphemic matter) vs PAT (structural patterns) distinction introduced in the work of Sakel (2007), a distinction which goes back at least to Heath’s (1978) distinction between direct (MAT) and indirect (PAT) diffusion, but probably even to earlier sources. It is apparent that there are both similarities and important differences between these varieties.
Features of the three mixed varieties discussed
The root/affix asymmetries referred to in the title of this article are typologically and areally specific: roots are single initial elements that either require another suffix (verbs) or do not, in Quechua and surrounding Andean languages. They do not involve the kinds of units that one finds in, for instance, a lot of the French loanwords in Native Canadian languages, where nouns often get borrowed with an article attached.
2. Relexification
Relexification is a general term used for massive lexicon replacement in a language, and has its roots in Creole studies. The concept has been applied in two areas in this subdiscipline. First of all, it has been used to explain the intricate interactions between African and European language features in the Caribbean Creole languages. Thus, some researchers that take the substratist position in Creole studies have embraced some version of the notion of relexification.
Second, it has been used to justify the assumption of West African Portuguese Pidgin as the common substrate of the Caribbean Creole languages, explaining many of their common features. This is referred to as the monogenetic position. In these different areas, it often has a somewhat different meaning, however.
In the African feature debate, one of the early exponents was Adam (1883), who proposed the term hybridologie linguistique. Since then, there have been numerous more informal attempts to suggest that many patterns in Caribbean Creoles resulted essentially from European word shapes coupled with African meanings and patterns. The most explicit and extreme defendant of this position has been Lefebvre (1998), who argued that Fongbe was relexified with French word shapes to produce Haitian Creole. A cursory survey of the literature will reveal that Lefebvre’s very strong claims have aroused much debate. Many Creolists will allow more modest amounts of relexification as one of the constitutive processes in Creole genesis.
The debate surrounding West African Portuguese Pidgin went under the label monogenesis vs polygenesis. In line with the monogenetic position, but in a less extreme version, was Hesseling’s (1933/1979) paper on Papiamentu influence on Negerhollands, the Dutch lexifier Creole language formerly spoken on the Virgin Islands. Hesseling assumed that there had been a group of Papiamentu speakers present in the early stages of the genesis of Negerhollands, who later shifted to a variety of the Dutch Creole. Not only did they leave lexical MAT+PAT influences (as kabay for horse; cf. Sp. caballo, Port. cabalho). They also were responsible for some PAT meaning configurations, as the use of the Dutch word form wil to denote both the wish to do something and the desire for an object or person (cf. Sp. querer, Port. quer, Pap. ker with both meanings). In Dutch itself, wil can only mean the wish for some action or state. Thus, ik wil jou means ‘I want (to have) you’ but not ‘I love you’, while the Negerhollands equivalent has both meanings, as does Spanish yo te quiero. Or course, these meanings are not unconnected, but you can want someone on your team or for a particular position without loving that person.
In Voorhoeve (1973) it is argued that the original Creole slave population of Surinam was Portuguese Pidgin rather than English Pidgin speaking (due to the prominent role of Portuguese Pidgin in the Atlantic slave trade), but that their language was progressively relexified towards English under the influence of the English plantation owners. The Saramaccan maroons escaped into the jungle before their language was fully relexified, which accounts for the high number of Portuguese elements in their language, according to Voorhoeve. In this account, both MAT and PAT are involved in relexification.
These arguments are shown schematically in Table 2.
Use of the notion of relexification in Creole genesis by different authors
3. Media Lengua
In the Andes of Ecuador several cases of mixed Spanish–Quechua languages have been documented (Muysken, 1979, 1981, 1988, 1997a; Gómez Rendón, 2006, 2008), which are often labelled as Media Lengua (other terms are Chaupi lengua, Chaupi shimi [both: ‘half language’], Utilla ingiru [‘little Quechua’], Chaupi quichua [‘half Quechua’], Quichuañol [‘Quechu-anish’] and Chapu shimi [‘mix language’]). The best studied cases are the Media Lengua of Salcedo, Cotopaxi province (studied by Muysken) in the centre, and the Media Lengua of San Pablo, Imbabura province (studied by Gómez Rendón) in the north, while Muysken has also documented varieties near Cañar, Cañar province and Saraguro, Loja province in the south. These varieties span a large part of the Ecuadorian Interandean corridor, and as far as we know are unrelated. I cite examples here from Gómez Rendón’s valuable monograph (2008), since the data in my own publications are more easily accessible and have already been frequently cited.
Basically, Media Lengua is a form of Quechua in which the large majority of the roots have been replaced with Spanish elements. These elements retain their basic lexical properties, although they are partially adapted phonologically. Thus we have examples such as the following, from a narrative (Gómez Rendón, 2008, p. 85; glosses adapted):
(1) ai-manda lexo-ta bi-kpi-ka
there- uno blanko asienda kaza-mi asoma-ri-xu-shka-n-ga
one white hacienda house-
wagra dueño-ka alla-man-mi contento i-shka
cow owner- pero el-ka akorda-ri-shpa-wan anda-xu-shka
but 3- patron-ka solta-wa-nga-chu
ima-shi kuanto-ta-shi
boss- kobra-wa-nga
yuya-shpa anda-xu-shka
charge-
‘Then, while he saw it far away, a white hacienda house became visible, and the owner of the cow walked towards it happily, but thinking by himself he walked along: “will the boss let me go, what will it be, how much will he charge me?”, thinking those things he was walking along.’
The Quechua elements are italicized throughout this article. Notice that in (1) the large majority of root elements are from Spanish. The exceptions are wagra ‘cow’, ima ‘what’, and yuya- ‘think’. Wagra may be a term locally used in Spanish as well, ima is part of a fixed expression ima-shi ‘what will it be?’, and yuya- is a genuine counterexample to the claim that all roots are from Spanish. Gómez Rendón (2008) notes that Imbabura Media Lengua contains more Quechua elements than the Salcedo variant. Muysken (2010) argues that the most prominent apparent counterexample in the Salcedo Media Lengua data, the Quechua copula ka- ‘be’, is actually a clitic in the relevant Quechua variety, and hence is expected not to be relexified.
Are just Spanish MAT items, outward morphological shapes, imported, or is underlying semantic PAT material brought in as well? For many words, this is hard to establish. Thus the meaning of Quechua puñu- ‘sleep’ and Spanish dormir ‘sleep’ is not sufficiently different in their semantic range to decide whether Media Lengua durmi- is just a MAT or also a PAT transfer. However, for other words, this is easier. In Table 3 I have tried to establish, on the basis of the discussion and examples presented in Gómez Rendón (2008), which Media Lengua verbs show evidence of Quechua meaning (PAT) transfer. A number of verbs fall into this category.
Relexification (PAT transfer) operant in Imbabura Media Lengua verbs
imp. = impersonal; caus. = causative.
Different varieties of Media Lengua show different degrees of transfer of Quechua semantic distinctions (see Table 4). Saraguro is the most Quechua-like version. The distinction between
Media Lengua personal pronouns in the different varieties
4. Bilingual songs, the wayno
In Peru and Bolivia there is the popular genre of the wayno, bilingual popular songs performed by small bands and transmitted through radio and cassette or CD. These songs are sung at festive occasions but also at dances and in bars. Very frequently a combination of both Quechua and Spanish occurs in these waynos. Large collections of these waynos have been produced in print. A typical example of a wayno is given in Escobar and Escobar (1981, p. 256) (Quechua elements italicized):
(2) Pobre sicuan-eñ-ita, [Poor girl from Sicuani] poor Sicuan- ¿a qué habrás venido, [wherefore have you come?] to what have.
Kay
runa-h
this man-
cry-
Kay runa-h
this man- ¿ grieve-
Mama-y-mi ni-wa-ra-n [My mother told me] mother-1-
ama
mama-y-mi ni-wa-ra-n [my mother told me] mother-1-
ama
….
The bilingual element comes in by various means. In the song above two means are illustrated: code-switching (the switch from initial Spanish to subsequent Quechua), and the one of concern here in this article, bilingual doubling.
The poetic effect in this genre of songs is reached for a large part with this technique of parallelism or doubling. Phrases are repeated, but often with a slight lexical modification. Thus we have a number of semantically roughly equivalent lexical pairs in subsequent lines, at least in the universe of the song (indicated in bold in (2) above):
(3) wasi ‘house’ waqa- ‘cry’ ri-pu- ‘go away’
llahta ‘town’ llaki- ‘grieve’ pasa- ‘pass (by)’
Since it is often difficult to find a semantic equivalent in the same language, as in the first two pairs in (3), often equivalents from Spanish are taken, as in the third pair. This is by itself not remarkable, since Spanish words can easily borrowed into Quechua. However, five features stand out in doubling in bilingual songs:
It is an extremely frequent phenomenon;
It involves particularly verbs rather than nouns, while ordinarily nouns are borrowed with much more frequency (although Spanish verb borrowing is not impossible in Quechua);
It involves basic vocabulary as well, not just more marginal vocabulary;
It frequently involves verbs that are never borrowed in ordinary discourse, as can be established from corpus studies of spoken Cuzco Quechua;
The verb occurs with all the relevant Quechua suffixes, as illustrated in (4), taken from the last line in (2):
(4) Ama pasa-y-rah-chu
‘Don’t leave from here yet’
Typical verb doublings found in waynos include:
(5) sabi- ‘know’ yacha- ‘know’ bulta- ‘return’ kuti- ‘return’ pasa- ‘pass’ ri-pu- ‘go away’ tuma- ‘drink’ uxya- ‘drink’ tupa- ‘meet’ tinku- ‘meet’
I assume these forms to be conventionalized doublets, which can be freely entered into the wayno for doubling purposes, and then receive the full range of Quechua affixes.
5. Kallawaya
In a very different speech genre, something similar to both Media Lengua and the wayno verb doubling is found: the Kallawaya ritual language of the professional healers of the Charazani region north of La Paz in Bolivia. Compare the paired examples in (6) and (7):
(6) a. Q’ari-s, warmi-s, alkalde-tah ri-n-ku. ( man- b. Laja-kuna, atasi-kuna, alkalde-tah isna-n-ku. ( man- ‘The men, the women and the mayor went.’ (Stark, 1972, p. 216) (7) a. Ri-pu-nki mana willa-ku-spa. ( go- b. Isna-pu-nki u uri-ku-spa. ( go- ‘You went away without telling.’ (Oblitas Poblete, 1968, p. 34)
The forms in (6a) and (7a) represent the ordinary speech of the speech community, while the forms in (6b) and (7b) the special ritual language. The non-italic forms in (6b) replace the Quechua equivalents in (6a), while the Quechua morphology and grammar is maintained (in addition there is a Spanish loan, alkalde ‘mayor’, in both language samples of (6); it need not concern us here). As far as we can establish many of the lexical roots of Kallawaya are of Puquina origin, but there may also be other languages involved, such as Leko, Tacana, Moseten and Uru. Finally, a number of words may simply be neologisms; striking is the avoidance of loans from Spanish, in contrast with all the languages of the area, and certainly with Quechua.
Just like in the case of Media Lengua, there is replacement of Quechua roots with elements from another language, and these elements are partially adapted phonologically to Quechua. In addition, there words are affixed with the standard Quechua affixes, for the most part. For all intents and purposes, contemporary Kallawaya is a form of Quechua with roots from another language. It resembles the wayno songs in that through lexical replacement different registers are created: in the case of Kallawaya this is the ritual register, in the case of the wayno songs this is the ‘doubling’ register.
6. Issues of genesis and processing
In this section, I briefly comment on issues of genesis and processing with respect to Media Lengua, mixed bilingual songs and Kallawaya. In terms of scenarios of genesis, three primary scenarios come to the fore: creation, shift and borrowing.
6.1 Creation
It is quite possible that in all three cases, processes of conscious creation have played a role. Media Lengua may have emerged out of a language game in the early decades of the 20th century, when so far almost monolingual Quechua-speaking construction workers from rural villages suddenly found themselves working in the rapidly expanding capital of Quito. Both the expansion of the capital and the mobility of the workforce were the result of the construction of a railroad connection to the coastal port of Guayaquil. This language game may then have become conventionalized in the communities the migrant construction workers were from.
There has been no study so far of the history of popular music in the southern Andes, but there is no doubt that the bilingual songs were the result of a process of semi-conscious creation, triggered by the requirements of the process of semantic doubling in Quechua poetry and facilitated by the widespread bilingualism in the area.
The origins of Kallawaya remain obscure, but it is clear that in the more contemporary forms of Kallawaya usage, creative processes linked to the ritual practices play an important role. Kallawaya usage is highly performative in nature, as far as we know.
6.2 Shift
Media Lengua may be interpreted as a linguistic phenomenon that accompanies the overall shift in many rural highland communities in Ecuador from Quechua to Spanish; indeed, in all cases where forms of Media Lengua have emerged we find shift occurring as well. However, several observations speak against a strong intrinsic link between Media Lengua formation as such and shift. First, in many Andean communities there has been language shift without the creation of Media Lengua. Second, the genesis of Media Lengua took place at a time when there was no shift yet to Spanish in the relevant communities. There is no intrinsic link between Media Lengua formation and shift, particularly also when we take other mixed languages into account.
The use of bilingual songs in the southern Andes is indicative of widespread bilingualism, but not necessarily of shift. Rather, the continued presence of Spanish and Quechua in these songs suggests a form of possibly stable diglossia.
In the case of Kallawaya, there has been shift originally, but in this case away from the lexifier language (Puquina) to the structure language (Quechua). The resulting ritual language, however, is more like a case of counter-shift or U-turn.
6.3 Borrowing
Again, the relation with borrowing is quite complex. In the areas where Media Lengua is spoken there is also extensive borrowing, and the way Spanish borrowed forms are adapted to Quechua is exactly like the way relexified items are adapted. However, borrowing is quantitatively restricted to about maximally 40 per cent of the root tokens in the local varieties of Quechua (Stark & Muysken, 1977), while in the case of Media Lengua we have almost 100 per cent of the root tokens. Qualitatively, borrowing is mostly restricted to non-basic vocabulary and the distinction basic/non-basic is irrelevant in the case of Media Lengua.
In the varieties where waynos are sung there is also widespread borrowing (although more limited than in Ecuador), but the pattern of bilingual verb doublings involved verbs that are never borrowed.
Regarding Kallawaya, it is clear that the systematic use of Puquina and non-Quechua other words in the ritual language is very different from what we may find elsewhere in the region; there have been reports of some unusual specialized vocabulary, but the Quechua of the area is overwhelmingly non-Puquina influenced, as far as can be gathered from the materials published so far.
6.4 Language mode and dynamicity
As to language mode, the picture is different for the three varieties at hand.
It is clear that in the original invention stage of Media Lengua the two languages must have been present in the mind of the speaker in order for her or him to be able to relexify; however, as the Media Lengua stabilized, relexifications became conventionalized, and there was no need for the activation of either language. In fact, there are speakers of Media Lengua without good knowledge of Quechua (vocabulary), although the initial creators of Media Lengua surely were highly proficient speakers of Quechua, incipient bilinguals in Spanish. For MAT transfer as in relexification to occur, there is no need for very deep knowledge of the second language. Words are adapted phonologically, nativized, although some properties of the original lexemes are retained (see below).
The producers of and many of the listeners to bilingual songs are bilinguals, and indeed the confrontation of the two languages is part and parcel of the aesthetic pleasure that these songs provide. However, many of the actual verb doublets are highly conventionalized.
The present-day speakers of Kallawaya who are left do not know Puquina. The Puquina words are simply part of a lexicon of non-Quechua words that they can use in speaking Kallawaya, while they are also able to use the Quechua lexicon in speaking Quechua.
6.5 Matter and pattern
The final issue that concerns us here is that of the transfer of matter vs pattern. How do we account for (a) the lexicon/grammar split in these three varieties and (b) the affix/root split?
What the mixed language data clearly show is that manipulating access to a lexicon separate from the one that is conventionally attached to the grammar is clearly a possibility in this case, in special but not exceptional circumstances. While speakers need to know the matrix grammar and phonology well, the transferred lexicon is possibly only incompletely known. The phonology of the mixed languages involved shows mixed features (Van Gijn, 2009). Vowel distinctions ([mid]/[high] in the case of Media Lengua, [long]/[short] in the case of Kallawaya) for instance are closer to those of the donor language, while phonotactic patterns are more like those of the recipient language. For this reason, the phonology does not provide us with much of a clue here.
The question remains, however, why Quechua as a language has allowed these processes of relexification or massive lexical transfer. Part of the answer may lie in its history as an imperial language which was adopted in many parts of the Andes as a second language during and even after Inca rule, and in the bilingual sociolinguistic context within which it is spoken. This cannot be the whole story, however, since other languages in the world also have this character of imperial languages and do not show relexification to the same extent, if at all.
What facilitates the separation of the Quechua lexicon from its grammar is that most of the burden of interacting with the actual grammatical system in Quechua lies in the affixes rather than in the roots. Most grammatical work is done by suffixes, not by the lexical roots themselves:
Roots belong to two word classes, nouns (with a subclass of adjectives) and verbs. Elements such as pronouns, quantifiers, adverbs and conjunctions are subclasses of the noun class.
Some elements are both nouns and verbs.
All nouns can occur as bare forms in the language, most often they carry some case marking, person marking, topic marking, evidentials, etc. Verbs can never occur as bare forms.
Thus taking roots from another language does not have a major impact on grammatical processing, which mostly interacts directly with the affixes. Furthermore, roots and affixes are clearly distinct from the perspective of lexical processing.
Roots are always initial in Quechua: there are only suffixes, no prefixes or infixes. Furthermore, there are different phonotactic constraints for roots and affixes, and phonological rules such as voicing, vowel raising and contraction in central Ecuadorian Quechua clearly distinguish between affixes and roots. Most roots have two syllables, affixes one. Roots have more types of sounds and are phonologically more complex. Affixes have a much more abstract meaning than roots. The token frequency of affixes is much higher than that of roots.
These differences are coupled with the agglutinative morphology characterizing Quechua. Altogether, the grammatical and morphophonological properties of Quechua on the whole are propitious to a process of relexification involving the roots of the language, and not the affixes. The affixes function as separate units and theoretically could be relexified by themselves.
Indeed we find some cases of Spanish suffixes in Media Lengua, including -ndu ‘adverbial subordination’, and -du ‘resultative nominalization’. However, only the form -ndu appears to be productively used, and is also the suffix which does not occur frequently as a borrowing in other Quechua varieties. Spanish suffixes in Quechua is the subject of the next section.
7. Spanish suffixes in Quechua
In Table 5 (summarized from Muysken, in press) the different Spanish suffixes are listed, loosely ranked in terms of their grammatical status and productivity, that occur in varieties of Media Lengua and Quechua.
Suffixes borrowed or relexified from Spanish in different varieties of Quechua and Media Lengua
We can conclude that there is a wide variety of Spanish suffixes that have been adopted into different Quechua varieties. Broadly speaking, they fall into four categories:
I. Suffixes that replace a Quechua suffix, often in the verbal paradigm:
(8) -kpi and –shpa > -ndo ‘adverbial subordinator’ (only in Media Lengua) -sqa, -ska, -shka > -do ‘resultative nominalizer’ -q, -k > -dor, -dero ‘agentive’ ?-pacha > -hora ‘temporal subordinator’
In some varieties these suffixes are only partially productive and limited to the lexical domain, but this requires more study. The element -hora may replace the suffix -pacha ‘time, world, since’, but in most Quechua varieties this suffix is not used grammatically as -hora is in Inga in Colombia.
An example of the use of -ndo or -ndu in Salcedo Media Lengua is:
(9) ahi-da-ga abi-n, piru tarde-ya-ndu-ga gana-u-nga-y
there- ‘It is there, but when it gets late he will be winning.’ Media Lengua, Ecuador (Muysken, 1997a, p. 386)
This -ndu replaces the different subject subordination marker -kpi here. Cases of -dor, which replaces the Quechua agentive marker -q (Peru) or -k (Ecuador), are:
(10) Chay runa-ka macha-dor-mi
that man- ‘That man is a drunkard.’ Ecuador (Ross, 1960, pp. 51–52) (11) sementerio-ma apa-dor ka-rka-kuna
cemetery-to take- ‘They used to take [us] to the cemetery.’ Inga, Colombia (Levinsohn, 1976, p. 106)
II. A range of diminutive suffixes that only partly come in the place of Quechua suffixes, but also derive some gender properties from the donor language Spanish (Cochabamba Quechua; Urioste, 1964):
(12) *-itu after Quechua words that end in /u/, but also partially sensitive to (particularly natural) Spanish masculine gender
punqu punq-itu ‘little door’ *-ita after Quechua words that end in /a/, but also partially sensitive to (particularly natural) Spanish feminine gender
uma um-ita ‘little head’ *-situ after Quechua words that end in /i/
rumi rumi-situ ‘little stone’
We also find the form -ditu occasionally as a diminutive or a characterizer.
III. A range of characterizing and affective suffixes often loosely modelled on Spanish suffixes but without clear Quechua models:
(13) -ilu/-ila diminutive, affective -lun characterizer -likido characterizer -nyintu characterizer -iru characterizer
Some selected examples:
(14) siki-lu ass- ‘with a big ass’ Santiago del Estero (Bravo, 1985, p. 296) (15) wacha-chi-lun give.birth- ‘midwife’ Ecuador (Muysken, 1977) (16) rumi-likido stone-liquid ‘like stone’ Lamas Peru (Taylor, 1975, p. 54)
Interesting is the fact that these suffixes appear to be characteristic of two closely related, affective semantic domains in nominal morphology: affective and characterizing. Seifart (2009) stresses the tendency towards specialization within a single domain as a feature of morphological borrowing.
IV. A final category is the Spanish plural suffix -s, which is used almost obligatorily with Quechua nouns ending in a vowel (the vast majority) in Bolivian Quechua.
(17) warmi-s ‘women’ algu-s ‘dogs’ Bolivia (Urioste, 1964)
It is rare if not non-existent in other varieties of Quechua. The origin and spread of this use of -s merits further historical study.
Altogether the range of Spanish suffixes and their spread across a number of varieties of Quechua is striking, although further comparative work on similar situations involving other languages, such as Nahuatl (Karttunen & Lockart, 1976; Field, 2002) will be needed to be sure of this. In any case, these findings tend to support the observations made in section 7 about the special separate status of affixes in Quechua.
8. Conclusions
This article has ranged over different Andean territories, from Colombia and Ecuador to Argentina, and has explored the psycholinguistic transfer types needed to account for in the emergence of two mixed languages and a mixed register with a Quechua structure: Media Lengua (Ecuador) and Kallawaya (Bolivia), both relexified varieties within the Quechua language family, and bilingual mixed songs in Peru, waynos. The two issues that require most attention are (a) the mental status of roots vs affixes in the transfer process and (b) the possibility of manipulating lexical access in transfer. The languages and the register share a number of structural features, but were found to be sociolinguistically totally different. In Media Lengua the lexicon comes from a ‘new’ language (Spanish), and in Kallawaya from an ‘old’ language (Puquina). Media Lengua is an informal community language, while Kallawaya is a ritual healing language only used by male adults. In waynos there is evidence of relatively balanced bilingualism. The mixed languages are confronted with the mirror phenomenon of Spanish suffixes that occur in Quechua, to help us further understand the processing issues involved.
With respect to the mental status of roots vs affixes in the transfer process, we can conclude that affixes in Quechua are fairly autonomous, and separate from roots. With respect to the possibility of manipulating lexical access in transfer, we can conclude that roots, but not affixes, can easily be manipulated. Presumably, affixes carry the grammatical processing load by themselves, freeing roots for being transferred, either from an ancestral community language, as in the case of Kallawaya, or from a dominant postcolonial language, as in the case of Media Lengua. In mixed songs, as well, we see the freedom of manipulation with respect to roots, though not affixes.
Footnotes
Appendix: Glosses used
Acknowledgements
I am grateful for the detailed comments from a reviewer for the journal and from the editors.
This research was published with the support of the Academy Chair of the Royal Netherlands Academy of Sciences (KNAW).