1. Introduction
Bilingualism can be an individual affair, as when someone decides to learn a foreign language; alternatively, it may occur within a small group of individuals defined in terms of work or family ties; finally, it may be societal in scope, resulting, as Escobar (1978:33) puts it, from el contacto de grupos étnicos que coexisten y compiten, en un régimen en el cual involuntariamente se deviene bilingüe para sobrevivir (‘contact between ethnic groups that coexist and compete, in a regime in which involuntarily one becomes bilingual in order to survive’). It is bilingualism of this third type that constitutes the main theme of this section.

Many bilingual speakers will be more proficient in one language than another (note that the term ‘bilingual’ in linguistics applies to any person who can communicate, however rudimentarily, in two languages). Accordingly, we may draw a distinction between the primary and secondary languages, the first of which is often, though not always, a speaker’s ‘mother tongue’ or first-learned language. In terms of the secondary language, societal bilingualism is normally characterized by a bilingual continuum; that is, a spectrum of speech types that may range from an almost unrestricted or standard usage to the lowest possible levels of oral proficiency. The phenomenon is illustrated, for example, by the varieties of imperfectly learned Spanish employed in the Andes by native speakers of Aymara and Quechua. Escobar (1978) refers to those varieties collectively as the ‘interlecto’ or interlanguage, noting that a speaker’s level of proficiency se identifica por correlación con el juego de ciertas variables, a saber: a) la escala de castellanización, b) el lapso de escolaridad, c) el tiempo de exposición al castellano y d) la tasa de frecuencia de su uso (‘is conditioned by certain factors, viz. (a) the degree of Hispanization, (b) educational level, (c) length of exposure to Spanish, and (d) frequency of use’).

Although individuals may advance up the bilingual continuum as their linguistic competence in the secondary language develops, certain of the non-standard features they exhibit at one or more phases in their linguistic development may be widespread and enduring in the community as a whole. These socially generalized features frequently conform to determinate patterns, as the same solutions are reached time and again to the problems involved in using two distinct systems. A survey follows, illustrating the most common of these solutions in the phonological and grammatical areas, of some of the most notable bilingual Spanish phenomena. After this, a brief account is given of code-switching.

2. Bilingualism and phonology
The two most characteristic types of phonological interference phenomena are the under-differentiation of phonemes and phone substitution (Weinreich 1953:18–19). The first of these consists in the elimination in the secondary language of a phonemic distinction as a consequence of this distinction’s being absent in the primary language. A good illustration is provided by the vowel system in Peruvian interlanguage Spanish. Among speakers whom Escobar (1978) classifies as ‘bilingües iniciales’ (as opposed to ‘bilingües avanzados’), the three-way height contrast (low ~ mid ~ high) of mainstream Spanish is replaced by a binary opposition that replicates the structure of vocalism in Aymara. Thus where monolingual Spanish exhibits five vowel phonemes, /a, e, i, o, u/, some varieties of interlanguage Spanish have just three, which conventionally are represented as /a, i, u/. The front and back phonemes /i/ and /u/ may be realized high (as in [misa] mesa ‘table’ and [rutu] roto ‘broken’) or as mid (as in [tʃekas] chicas ‘girls’ and [ɸɾota] fruta ‘fruit’) but no clear linguistic rule seems to condition either type of articulation. The merging of the front pair /e/ and /i/ has a knock-on effect on the diphthongs [je] and [ej], whose components constitute, for the speakers in question, allophones of the same phoneme. The solution consists generally in replacing both diphthongs by [e:] or by [i]; e.g. [pe:ðɾas] piedras ‘stones’, [pe:ne] peine ‘comb’, [emolinte] emoliente ‘emollient’.

Phone substitution arises when the same phoneme occurs in the two contact languages but is realized differently in each of them. A tendency among bilinguals may be to replace the realization associated with the phoneme in the secondary language with its realization in the primary language. For example, both English and Spanish have the voiceless stops /p, t, k/, which are often aspirated in English – [pʰ, tʰ, kʰ] – but never in Spanish. One common feature of the Spanish used by English-dominant bilinguals is the replacement of the standard non-aspirated allophones with the aspirated sounds, as in [pʰatʰatʰa] patata ‘potato’. Conversely, in the English spoken by Spanish-dominant bilinguals the aspirated stops may be replaced by their non-aspirated counterparts, a tendency that is often exploited in cinematic representations of Latino or Chicano accents.

A rarer phenomenon is the introduction into the secondary system of a wholly foreign phoneme. The presence of /ʎ/ in Peruvian interlanguage Spanish is perhaps a case in point, as the palatal lateral is absent from mainstream varieties of Latin American Spanish but present in Aymara and at least some dialects of Quechua.

3. Bilingualism and grammar
Turning now to grammar, the strategies employed by bilinguals in this area include simplification, overgeneralization, the development of periphrastic structures in place of inflected forms, and transfer (compare Silva-Corvalán 1995:9-10). As will become evident, bilingualism rarely leads to the introduction of wholly foreign structures into the grammar of the secondary language. Instead, existing forms are assigned new meanings or the incidence is reduced of a given structure, in some cases leading to its complete abandonment by later generations.

3.1. Simplification
One feature of Andean interlanguage Spanish that might be characterized as tending towards simplification is the non-realization of object pronouns, an omission that enables a speaker to avoid observing rules concerning placement and gender/number marking (examples from Stratford 1989:116, and Escobar 1990:89):

(1) Aquí están los medicamentos. ¿Cómo has traído?

(2) A veces dejó su quacker ya preparado; en la mañana calientan y toman.

Null pronouns are of course common in standard Spanish, but only when the reference is indefinite:

(3) Necesitamos bombillas; a ver si compramos esta tarde.

The same bilingual varieties exhibit a strong tendency towards using count nouns as mass nouns, as in La señora vende huevo ‘The woman sells egg<s>’, Lleva piedra ‘He/She carries stone<s> (Escobar 1978:94) and Vende vaca ‘He/She sells cow<s>’ (Hardman de Bautista 1982:150). Again, the phenomenon is not uncommon in informal standard Spanish, but what is distinctive in the Andean case is its greater frequency. As with the null pronouns, the net effect seems to be a simplification of the grammar, given that mass nouns are not inflected for number and do not require a determiner.

3.2. Overgeneralization
The first illustration of overgeneralization concerns clitic se. In all varieties of Spanish including the standard, this may be used with certain verbs in an emphatic type of construction – e.g. Aquel hombre se bebió un litro de cerveza ‘That man drank a litre of beer’ – where it has neither a syntactic role nor a place in the lexical form of the verb. In Andean interlanguage varieties, however, this use of se may be extended to verbs with which the clitic cannot normally be construed in standard or other mainstream varieties of Spanish (examples from Solís 1988:196):

(4) Con los dos (quechua y castellano) lo que yo me bromeo, me hablo.

(5) Cuando ella también se hace casar sus hijos.

(6) Mi opinión se es esto.

A phenomenon that is observable in Mexican, Central American and Andean interlanguage Spanish provides a second illustration of overgeneralization. The co-occurence of a 3rd person direct object clitic with a lexical direct object, as in Lo vi a tu tío ‘I saw your uncle’, is not uncommon in informal mainstream Spanish, as long as the lexical object is a definite NP (note that this is a distinct phenomenon from clitic left dislocation, in which an object clitic resumes a topicalized object, as in A María no la contrataron ‘They didn't hire María’). In interlanguage Spanish, however, a direct object clitic (frequently invariant lo) can co-occur with a lexical direct object that is an indefinite NP, as in Lo trae un chiquihuite ‘He/she brings a basket’. Moreover, the same varieties exhibit the use of direct object clitics with intransitive verbs, as in Lo llegó ‘He/she arived’ and Dicen que lo nació en Belén ‘They say he was born in Bethlehem’.

The possibility of syntactic transfer cannot be discounted, but this is unlikely, as transfer tends to rely on speakers identifying a specific element in the secondary language with another in the primary language, and none of the relevant indigenous languages (e.g. Nahuatl and Quechua) have any element whose function or sentence position consistently mirrors interlanguage clitic usage. The phenomenon is more plausibly treated as a case of overgeneralizing the pattern lo (or la) + verb, rather as students of Spanish frequently generalize falsely from Lo siento ‘I’m sorry’ to *Lo siento que + subordinate clause.

3.3. Periphrastic constructions
A case in point is the displacement of the synthetic preterite by the analytic perfect in bilingual Andean Spanish. Klee & Ocampo (1995), for instance, show that among Quechua-dominant bilinguals in the Peruvian town of Calca (50 kms from Cuzco) the perfect frequently occurs in extended discourse structures, where the preterite is strongly preferred in standard Spanish. For example (p. 62):

Un [sic] tía que […] una vez que ha llevado comida a los obreros que trabajaban en el arriendo de mi abuelita … una culebra se lo había envuelto al pie […] y, la chica se ha asustado … y, por sí sola, la culebra se ha bajado.

Mendoza (1991) and Stratford (1991) document the phenomenon in La Paz Spanish and Altiplano Spanish (Lake Titicaca area) respectively.

3.4. Transfer
This, the final grammatical strategy to be considered here, is one of the most interesting and controversial areas of bilingualism, with some scholars taking the view that social factors, such as the intensity of intercultural contact, are the primary determinants of the degree of transfer between languages and others arguing that transfer only occurs in those areas of the system in which a degree of structural parity already exists between the two languages. The Spanish data point strongly to the possibility of fairly significant grammatical transfer, but always in ways that do not represent a radical departure from mainstream usage.

An interesting example is the increase in frequency of the estar copula at the expense of ser that has been observed in bilingual Los Angeles Spanish. What seems to be happening is that the twin-copula system of Spanish is being eroded under the influence of English, in which there is only one copula. Two uses of ser appear to be vulnerable.

The first concerns gradable adjectives, such as grande ‘big/large’ and delgado ‘thin/slim’. With ser, these are always syncategorematic, in that they must be construed, explicitly or implicitly, with another term. Enrique es grande, for example, means that Enrique is a big man, say, rather than that he is big in some absolute sense. When used with estar, on the other hand – in standard Spanish, at least – gradable adjectives are fully categorematic (the contrary of syncategorematic). Thus X está delgado says that X is thinner than normal, and this is a property that in principle is independent of any secondary categorization of X as a man, woman or anything else. Mexican American bilinguals in Los Angeles, particularly those for whom English is the dominant language, are starting to use estar not just for what we are calling the categorematic use of gradable adjectives, but also for the syncategorematic use (examples from Silva-Corvalán 1989:188-198):

(7) Tengo una casa; está grande mi casa, muy grande.

(8) Mi hermano está grande. No está delgado ni chiquito. Se ve fuerte.

Both of the sentences above could in theory illustrate the categorematic use of the adjectives in question; that is, (7) might have been uttered just after an extension was added to the speaker’s house and (8) might be prompted by a growth spurt on the part of the speaker’s brother. But, apparently, the full context rules out those (unlikely) interpretations.

The second encroachment of estar into the territory of ser consists in the use of the former with predicative nouns, as in

(9) Es imposible que me obedezcan mis hijos, ya están jovencitos.

Syntactically there is no reason why jovencitos in (9) should not be treated as an adjective. But such an analysis is ruled out by the meaning of the sentence. What is meant, obviously enough, is that the children have become youths or teenagers, and not that they have become young.

Ordering patterns, to change the example, are often prime candidates for interlinguistic transfer. However, given that the major syntactic constituents in Spanish can be ordered with quite a degree of freedom, transfer of this sort is often reflected in an increase in the incidence of a grammatical but marked sequence, rather than in the introduction of a wholly new model. A case in point is the placement of lexical objects before verbs in the speech of Quechua-dominant bilingual speakers in Calca, near Cuzco in Peru (Ocampo & Klee 1995), as in Demasiado [sic] travesuras hacía ‘He was doing too many pranks’ and Cicatriz tengo ‘I have a scar’. This sequence of constituents is not ungrammatical in standard Spanish, but it is normally reserved for focalizing the direct object. Bilingual speakers in Calca have, apparently, extended this word order pattern to cases in which no focalization is intended.

A similar case relates to the word order associated with intransitive verbs occurring with indefinite subjects, as in Llegaron muchos chicos. From the discourse–pragmatic point of view, such constructions have a partly presentational role, as they serve among other things to introduce new participants in the events described. The unmarked word order in this case is verb–subject in Spanish but subject–verb in English. In some varieties of bilingual Spanish in the USA, the English word order appears to have displaced the unmarked Spanish word order, as in (10) below (from Silva-Corvalán 1989:173), which sounds odd to the native Spanish ear but is not actually ungrammatical:

(10) Nos agarramos y, y, un montón de policías vinieron.

Perhaps the most striking example of interlinguistic transfer – a clear case, in fact, of grammatical convergence – concerns the encoding of data source or the grounds the speaker has for asserting a proposition. This is regularly indicated in sentences in Jaqi languages (including Aymara), as well as in Quechua, although not to the same extent. (According to Hardman de Bautista (1982:152), data-source marking in Quechua largely reflects borrowing from Jaqi in the first millennium.)

There are at least three grades of data source in Aymara: personal knowledge, knowledge through language, and non-personal knowledge (the examples that follow are from Hardman de Bautista 1982:152–153). Assertions based on personal knowledge are unmarked, in that only the assertoric affix w(a) is present, as in Mama Marsilaw t’ant’ manq’i ‘Ms. Marcela ate bread’, where the speaker has direct sensory experience of the event described. Knowledge through language is indicated through the use of siw, the 3rd person singular pres. tense of saña ‘to say’, as an independent particle, although again the assertoric affix w(a) may be present: Mama Marsilaw t’ant’ manq’i siw ‘Ms. Marcela ate bread (so I’m told)’. And the prototypical exponent of non-personal knowledge, in the 3rd person, is the verbal affix tayna, as in Mama Marsilax t’ant’ manq’atayna ‘(Apparently) Ms. Marcela ate bread’. This category of data source also includes information that comes as a surprise to the speaker: Pitañ yatitayna ‘(So) they knew how to smoke!’. Comparable evidential distinctions are made also in Quechua, especially using the suffix -sqa.

This tripartite system of data source marking is replicated in interlanguage Spanish and also, as a consequence of linguistic transfer, in popular dialects of monolingual Andean Spanish through a redeployment of existing resources. More specifically, both the pluperfect tense (había hecho) and the verb form dice ‘he/she says’, the latter recycled as a generally clause-final particle, come to have an evidential or data source-indicating meaning. The following examples of pluperfect usage are from an article about graffiti in a public lavatory that appeared in the Cuzco weekly paper El Tiempo (18th to 24th September 1994):

(11) El jefe había tenido pelos en su sentadera. (i.e. ‘has hairs …’

(12) El Justo Condorhuamán había sido hijo de Pablo Muchotrigo. (i.e. ‘is the son of …’)

Like the exponents of non-personal knowledge in Aymara and Quechua, the Spanish pluperfect in the varieties in question is used also to express surprise, a feature of usage that seems to have seeped into the monolingual speech of even educated Limeños:

(13) ¡Ahhh! ¡Vivo habías sido! (i.e. ‘Ah! So you’re alive!’)

(14) ¡Habían sabido fumar! (i.e. ‘Ah! So they knew how to smoke!’)

Evidential dice, which is frequent in the interlanguage, has perhaps a greater currency in monolingual Andean Spanish than the evidential pluperfect. This presumably is due to the fact that the dice phenomenon represents a smaller departure from mainstream Spanish. Hardman de Bautista (1982:154) gives the following examples:

(15) Marcela está enferma dice.

(16) Estaba enojada dice.

A variant form of this construction, involving the complementizer que, arises when dice occurs clause-internally:

(17) Shumaya dice que comió pan.

Note that the impersonal meaning of dice endures and so, unlike in standard Spanish, Shumaya cannot be interpreted as the subject of dice.

4. Code-Switching
To conclude this section, some remarks are required concerning code-switching, or alternating between two languages (or dialects) within the same stretch of discourse, as in I was ... I got to thinking vacilando el punto you know and No creo que son fifty-dollar suede ones (Gumperz 1982:133, Poplack 1980:584).

Code-switching does not occur at random but seems to be favoured in certain types of context, such as in direct quotations or when an assertion is reiterated:

Y un día, cuando la conocí, me dice, ‘Oh, my daughter used to dance at Knott’s Berry Farm.’ Y digo yo pa’ mí: ‘A poco es la muchacha que retraté yo.’ (Silva-Corvalán 1989:180)

The three old one [sic] spoke nothing but Spanish, nothing but Spanish. No hablaban inglés. (Gumperz 1982:133)

Furthermore, bilingual speakers themselves will often attribute a given code-switch to the fact that a determinate linguistic resource in one language is more attuned to the matter at hand than the corresponding unit or structure in the other language.

Code-switching among fluent bilinguals appears to be subject to two constraints (Poplack 1980). First, it never takes place within words (or between units that are not free morphemes) unless phonological adaptation has occurred. Consider the following three sentences (from Silva-Corvalán 1989:181):

(18) Mi hermano está watching the game.

(19) Mi hermano está huachando the game.

(20) ?Mi hermano está watchando the game.

Sentence (18) represents a possible utterance, because the switch to Spanish does not occur within a word. The switch is word-internal in (19) but the sentence is again likely to be uttered because the root has been Hispanized: [watʃ] rather than [wɒtʃ]. Sentence (20), on the other hand, would not occur naturally, as the switch is word-internal and the root has not been adjusted.

The Miller Chill advert shown to the right (produced by Y&R Chicago and released in 2007) deliberately flouts the constraint just alluded to, by combining beer and -veza (from cerveza) into a single word.

Secondly, fluent bilinguals only switch languages at those points at which the syntax in one language matches the syntax in the other. For example, a switch to Spanish would not be made within either ‘told him’ or ‘would bring it’ in (21) below, because the corresponding strings in Spanish (le dije) and (lo/la trajera) involve procliticization and so do not replicate the syntax of their English counterparts:

(21) I told him that so that he would bring it fast.

Similarly, given that the adjective tends to follow the noun in Spanish – a rule that admits of no exceptions in comparative constructions involving más or menos – but precedes it in English, a sentence such as (22) below is apparently ruled out, while (23) is quite legitimate (example from Silva-Corvalán 1989:182):

(22) ?Juan va a comprar una casa bigger in the valley.

(23) Juan va a comprar una bigger house in the valley.

Code-switching among weak bilinguals, in contrast, often represents a strategy for overcoming deficiencies in linguistic proficiency and so is less subject to the constraint that structural parity be preserved.

References
Escobar, Alberto. 1978. Variaciones sociolingüísticas del castellano en el Perú. Lima: Instituto de Estudios Peruanos.
Escobar, Ana María. 1990. Los bilingües y el castellano en el Perú. Lima: Instituto de Estudios Peruanos.
Gumperz, J.J. 1982. Discourse Strategies. Cambridge: Cambridge University Press.
Hardman de Bautista, Martha. 1982. ‘The mutual influence of Spanish and the Andean Languages.’ Word 33:143–57.
Klee, Carol A. & Ocampo, Alicia M. 1995. ‘The expression of past reference in Spanish narratives of Spanish-Quechua bilingual speakers’, in Spanish in four continents: studies in language contact and bilingualism, ed. Carmen Silva-Corvalán. Washington DC: Georgetown University Press.
Klee, Carol. A. & Ramos García, L. (eds.) 1991. Sociolinguistics of the Spanish-speaking world. Tempe: Bilingual Review Press.
Mendoza, José G. 1991. ‘Aproximación morfosintáctica al castellano paceño’, pp. 207-229 in Sociolinguistics of the Spanish-speaking world, ed. Carol Klee and L. Ramos García. Tempe: Bilingual Review Press.
Ocampo, Alicia M. & Klee, Carol A. 1995. ‘Spanish OV/VO word-order variation in Spanish-Quechua bilingual speakers’, in Spanish in four continents: studies in language contact and bilingualism, ed. Carmen Silva-Corvalán. Washington DC: Georgetown University Press.
Poplack, Shana. 1980. ‘Sometimes I’ll start a sentence in English Y TERMINO EN ESPAÑOL: Toward a typology of code-switching.’ Linguistics 18:561–618.
Silva-Corvalán, Carmen. 1989. Sociolingüística: Teoría y análisis. Madrid: Alhambra.
Silva-Corvalán, Carmen. 1995. Spanish in four continents: studies in language contact and bilingualism. Washington DC: Georgetown University Press.
Solís, Gustavo. 1988. ‘La categoría de voz media en quechua y su reflejo en el español andino’, pps. 191–199 in Pesquisas en lingüística andina, ed. Luis Enrique López. Puno/Lima: Universidad Nacional del Altiplano.
Stratford, Billie Dale. 1989. Structure and use of Altiplano Spanish. Ph. D. dissertation, University of Florida.
Stratford, Billie Dale. 1991. ‘Tense in Altiplano Spanish’, pp. 163-81 in Sociolinguistics of the Spanish-speaking world, ed. Carol Klee and L. Ramos García. Tempe: Bilingual Review Press.
Weinrich, Uriel. 1953. Languages in Contact. The Hague: Mouton.