THE SPANISH OF ANDALUSIA
Perhaps no other dialect zone of Spain has received as much attention--from scholars and in the popular press--as Andalusia. The pronunciation of Andalusian Spanish is so unmistakable as to constitute the most widely-employed dialect stereotype in literature and popular culture. Historical linguists debate the reasons for the drastic differences between Andalusian and Castilian varieties, variously attributing the dialect differentiation to Arab/Mozarab influence, repopulation from northwestern Spain, and linguistic drift. Nearly all theories of the formation of Latin American Spanish stress the heavy Andalusian contribution, most noticeable in the phonetics of Caribbean and coastal (northwestern) South American dialects, but found in more attenuated fashion throughout the Americas. The distinctive Andalusian subculture, at once joyful and mournful, but always proud of its heritage, has done much to promote the notion of andalucismo within Spain. The most extreme position is that andaluz is a regional Ibero-Romance language, similar to Leonese, Aragonese, Galician, or Catalan. Objectively, there is little to recommend this stance, since for all intents and purposes Andalusian is a phonetic accent superimposed on a pan-Castilian grammatical base, with only the expected amount of regional lexical differences. There is not a single grammatical feature (e.g. verb cojugation, use of preposition, syntactic pattern) which separates Andalusian from Castilian. At the vernacular level, Andalusian Spanish contains most of the features of castellano vulgar. The full reality of Andalusian Spanish is, inevitably, much greater than the sum of its parts, and regardless of the indisputable genealogical ties between andaluz and castellano, Andalusian speech deserves study as one of the most striking forms of Peninsular Spanish expression.
The region which is now Andalusia was settled by Greeks as early as 600 B. C., and was conquered by Carthage around 300 B. C. Most of the Iberian Peninsula was conquered by Rome around 200 B. C., although effective Roman occupation did not come until nearly two centuries later. Southern Spain was first known as Hispania Ulterior, later changed to Baetica, corresponding to modern Andalusia. The name of Andalusia derives from the Arabic al-Andalus, bestowed on Moslem Spain by the Moorish conquerors, beginning around 716. Those parts of Spain remaining under Christian control were known as Ishbaniya, the Arabized version of Hispania. Christians were known as rumi or ayami; the latter term, meaning `foreigners,' was eventually applied to Ibero-Romance language written in the Arabic alphabet by reconquered Moslems, under the term aljamía and aljamiado.
The first Moorish capital in Spain was the Caliphate of Córdoba, lasting roughly from 756 to 1010. This marked the high point of Moorish domination and cultural presence in Spain; at its peak, Córdoba had some 500,000 inhabitants, being the largest city in western Europe. When Christians and rebellious Moors sacked Córdoba in 1010, Moorish power shifted to Sevilla, where Moorish control was centralized until 1248. Córdoba was finally defeated in 1236., but its decline had begun much earlier. Between 1247-1248, the Christian armies laid seige to Sevilla, and the city finally fell. This time Moorish power shifted eastward to Granada, where the Kingdom of Granada remained as the last bastion of Arabic control until its final defeat by Fernando and Isabel in 1492.
At first the Moorish presence in Andalusia was limited to the bare minimum needed for political, social, and military control. As more Arabs poured into Spain, a stratified and linguistic diglossic society emerged, in which official functions were carried out in Arabic, while Christians and an increasingly large number of Spanish-born Moors spoke the vernacular Ibero-Romance dialects known collectively as mozárabe. Since many of the Mozarabs were expelled from Spain by late-arriving Moslem troops from North Africa, and since few complete Mozarab texts survive (and those which do are written in Arabic script, making it impossible to determine vowel quality and many consonantal features), the true nature of Mozarabic language in southern Spain cannot be determined. Given the striking differences between Andalusian Spanish, in which Arabic influence was most prolongued, and the dialects of northern Spain, it is tempting to propose a heavy Arabic contribution to Andalusian speech, but the principal phonetic features of Andalusian Spanish appear to derive from a combination of spontaneous evolution and traits evolving from rustic Castilian and Leonese dialects brought in by the reconquest: `... lo que sabemos de la fonética mozárabe tiene poco, o nada, que ver con las características más notables del español de Andalucía' (Narbona et al. 1998: 39).
The nature and extent of Arabic-Romance bilingualism in Al-Andalus is the subject of much uncertainty, and has as a consequence resulted in controversial interpretations of available facts. The discovery of jarchas, poetic texts composed in the Arabic alphabet and which combined strophes in Arabic, Hebrew, and Ibero-Romance, suggest widespread bilingualism at least in the 11th and 12th centuries. By the 13th century, this literary genre had disappeared, although the existence of several Arabic texts with lists or translations of Romance words from the 13th century indicates that knowledge of Ibero-Romance was still common (Narbona et al. 1998: 37). Many scholars believe, however, that by the end of the 13th century, Romance-Arabic bilingualism in Al-Andalus was rapidly receding. Christians expelled to North Africa required translations of the Gospels into Arabic. Since most Christians had been deported from southern Spain prior to the reconquest, the remaining population was uniformly Moslem, and any Romance dialect they may have used was quickly swept away by the conquering Castilian forces: `...ese hipotético romance más o menos residual acabaría sufriendo la misma suerte que sus poseedores a manos de los nuevos señores del territorio, la cual ... no fue otra que la deportación' (Narbona et al. 1998: 38). In Granada, the last holdout of Moslem Spain, intense commercial contact with nearby Castilian-speaking territories resulted in some knowledge of Spanish by the Granada Moors, but it is unlikely that any local Mozarabic dialect survived the reconquest. Galmés de Fuentes (1996) analyzes possible Arabic syntactic influences in medieval Spanish texts.
The Christian reconquest of Andalusia brought with it the ghettoization of the Moslem population, reduced to specific barrios and exercising a minimal linguistic and cultural influence on the remaining population. By the time of the reconquest, many of the Moors themselves spoke Ibero-Romance, so direct Arabic influence on evolving Andalusian Spanish is unlikely under any circumstances: `...nuestros conocimientos actuales de la historia no nos permiten en absoluto, sino que lo desaconsejan, establecer vínculos directos entre el venerable romance mozárabe de Al-Andalus y las formas peculiares que acabó adquiriendo el castellano en Andalucía. No parece que hubiera base humana que sustentara la transmisión de rasgos lingüísticos de origen mozárabe al castellano de los conquistadores, y parece que hubo una verdadera solución de continuidad, un amplio lapso que separó los últimos momentos de vida del romance autóctono del sur de Al-Andalus y la implantación del idioma traído por los castellanos ...' (Narbona et al. 1998: 38).
Unlike what happened in, e.g. Toledo, reconquered in the 11th century, where Arabs, Jews, and Christians returned to a situation of mutual tolerance, Christian reconquest of Andalusia was accompanied by an intense campaign of deportation and concentration, a form of ethnic cleansing. Focusing first on the cities, the Christian forces required Moslems to abandon their homes and lands, and resettle in the remaining Moorish areas of Spain (ultimately reduced to just the kingdom of Granada). Following these mass deportations, the conquering Spaniards frequently brought in Moorish slaves from other provinces, to work as artesans and skilled laborers. The Moslems lived in ghettoes known as morerías or aljamas. The Moors remaining in Christian Spain were known as mudéjares, and became increasingly marginalized members of an overwhelmingly Christian and Castilian society. In rural regions a considerable Moslem population remained, but many were forced to work the lands of Christian conquerors as serfs. This resulted in several uprising of Moorish peasants, and in several regions of Murcia and Andalusia rural Moslems were later expelled to Granada.
Narbona et al. (1998: 41) suggest that the linguistic pecularities of Andalusian Spanish arose precisely in the late Middle Ages, when the Moslem population of Andalusia had been reduced to utter marginality: `hemos de rechazar, pues, con torundidad, cualquier tentación de vincular los procesos lingüísticos andaluces a la contaminacióncon el árabe vulgar andalusí, o a la deturpación del castellano en boca de hablantes arábigos ... esa falta de prestigio se manifiesta en la drástica dismunición' del número de arabismos léxicos que penetran en los siglos XIV y XV en castellano, y continuará en las caricaturas del habla de moriscos en la literatura del Siglo de Oro ...' Moreover, remaining fragments of Mozarabic speech in reconquered Andalusia, for example as appearing in aljamiado and morisco literature of the 15th and 16th centuries, shows tendencies far removed from those prevalent in Andalusia. Only in Granada, where Moors continued speaking Arabic for at least a century following the reconquest, could significant bilingual interaction take place. However, the marginalized status of the Moslem population precluded any significant transference from Arabic to Spanish.
By the 13th century, the Christian settlers in southern Spain began to use the term Andalucía to refer to the southern region, restricting a term which for the Arabs had referred to all of Spain. The majority of the reconquering population came from Castile and León, and the emerging Andalusian Spanish dialect amply reflects the Castilian influence. Tracing Leonese contributions is more difficult, although in Extremadura the Leonese reconquest is linguistically much more evident. Castilla la Vieja, in particular Viscaya, was responsible for the numerically greatest number of Castilian settlers in reconquered Andalusia. Eastern Andalusia, typified by Granada, was largely settled by established Andalusians in the reconquered western region (Narbona et al. 1998: 44-5). Settlers from Extremadura and the two Castiles were also present in large numbers.
There is no single text or event which marks the beginnings of Andalusian Spanish, and the traits--mostly phonetic--which define contemporary Andalusian speech emerged at different times and in different regions. Two large categories of phonetic change, confusion of sibilants as seseo/ceceo and widespread neutralization and loss of syllable- and word-final consonants, are widely attested in Andalusia beginning just past the end of the late medieval period. Already by the end of the 15th century Andalusian speech was regarded as `different' by authors such as Antonio de Nebrija and Juan de Valdés, although few specifics appear in their writings. It appears that the confusion of sibilants was the first feature to become widely associated with Andalusia, although other traits, such as neutralization of syllable-final liquids, probably began to appear even earlier. Narbona et al. (1998: 47-50) summarize the many unsupported qualifications--most of them derogatory--of Andalusians and their speech which appeared towards the end of the 15th century.
In 1579 Damasio de Frías wrote of Andalusian Spanish: `En la Andalucía, pues, no deja de haber sus pronunciaciones, en algunas partes extrañas y muy diversas de las castellanas, como en Jaén, Andújar, y, en general todos los andaluces lo son [diversos] mucho de nosotros, en el sibilo de la s.' (Mondéjar 1991: 40).
Three hundred years later, in 1872 and later in 1900, Juan de Valera described Andalusian pronunciation, particularly that of Córdoba (Mondéjar 1991: 43-5):
La cordobesa, por lo común (y entiéndase que hablo de la jornalera o de la criada, y no de la dama elegante e instruida) aspira la hache...
En la pronunciación dejan un poco que desear las cordobesas. La zeda y la ese se confunden y unimisman en sus bocas, así como la ele, la erre, y la pe. ¿Quién sabe si sería alguna maestra de miga cordobesa la que dijo a sus discípulas: "Niñas, sordao se escribe con ele y precerto con pe.
Mondéjar (1991: 169f.) analyzes a description of Andalusian Spanish offered by the French traveller le Baron Charles Davillier around 1862. Imitating the speech of a mozo crúo, Davillier offers: `Camará, nojotros no necesitamos jeso.' A more detailed imitation of a sevillano:
Puez zeñó ... no crea uzté que la han traido de Pariz ni de Londrez, que tal cual uzte la ve, la hemoz hecho acá en Zeviya ...
Zoy e Zeviya er mas terne;
Gazto la plato rumboso ...
Cojo con zal la naaja [navaja] ...
Tengo una jembra ...¡uy qué jembra!
Es la gloria de Zeviya ...
Speaking of the Andalusian ceceo, Davillier says `le ceceo, espèce de zézeiement qui consiste à prononcer l's comme le c, et à siffler quelque peu en parlant, suffit por trahir dès les ppremières paroles les enfants de l'Andalousie.' Describing the loss of /d/ he remarks:
On peut dire que la lettre D n'existe pas pour eux, car ils ont soin de la retrancher de tos les mots où elle se trouve: c'est ainsi qu'ils prononcent caliá por calidad ... enfaao pour enfadado ... elante e mi pour delante de mí ...
He also notes frequent metathesis: `Les Andalous se plaisent à faire très-fréquemment des inversions dan l'ordre des lettres: c'est ainsi que la Virgen ... devient la Vinge; premitir se dit pour permitir ... et probe pour pobre ...' Finally, Davillier explicitly notes the widespread loss of word-final consonants: `La plupart des [consonnes] sont supprimées à la fin des mots, de sorte que muger ... se prononce mugé; Jerez, Jeré; Cádiz, Caï; licor, licó, et ainsi de suite.'
Ramón de la Cruz also reflected western Andalusia ceceo in his plays, for example (Mondéjar 1991: 176-7):
como ay zan! ¡Bendito zea,
quien te jizo tan discreto
y a mi tan zalvaje! ...
¡Juan, zi zupieras que miedo
me da aqueya nubeciya! ...
Interestingly, even at the end of the 19th century these writers did not see fit to mention what today is the most striking characteristic of Andalusian Spanish, the massive elimination of word-final consonants, and of word-internal preconsonantal /s/. Even 19th century costumbrista writers such as Alvarez Quintero (@) fail to include loss of final consonants in their eye-dialect imitations of Andalusian Spanish. Although consonantal erosion was already generalized in 19th century Andalusia, it appears that higher sociolects still made a conscious effort at retaining these consonants (although in actual speech deletion rates were probably quite high), so that loss of final consonants was not attributed to educated Andalusians, even by writers from other dialect zones. Mondéjar (991: 71) observes that `Todos los sonidos observados por Machado y Alvarez son comunes y generales en la mayor parte de las comarcas andaluzas, pero no todos se dan en todos los estratos sociales. Machado y sus compañeros siempre los hacen--estos y muchos otros--patrimonio del pueblo, del vulgo, de la gente iletrada, inculta.' He theorizes (p. 198) that aspiration of syllable-final /s/ began during the 18th century at the very earliest, since it is not explicitly documented in literary or folkloric texts until the middle of the 19th century
Also writing in the latter part of the 19th century, Antonio Machado did make oblique reference to loss of /s/ in Andalusia (Mondéjar 1991: 66-7):
Así observamos que el andaluz muestra predilección por unas consonantes y aversión decidida hacia otras: gusta mucho de la aspiración de la h y la pronunciación de la s como silbante le fastidia y enoja: transforma en r la l por antipática a su espíritu, y apenas si emplea la d cuando no puede echarla a hombros ajenos.
At another point, Machado is more explicit in his imitation, saying that word-finally the pronunciation of /s/ `fluctúa entre el de la z debilitado y el de la h aspirada' (Mondéjar 1991: 70); he also gives an imitation:
Las lusesita'h que briyan
De noche en er sementerio,
Están disiendo à lo'h vivo's
Que se acuerden de lo'h muerto'h.
In 1847, Serafín Estébanez Calderón, from Málaga, referred to the loss of final consonants in his Escenas andaluzas (Mondéjar 1991: 186):
Manolito Gázquez además del "socunamiento" o eliminación de las finales de todas las palabras y de la transformación continua de las eses en zetas y al contrario, pronunciabia de tal manera las sílabas en que se encuentra la de o la erre, que sustituía estas letras por cierto sonido semejante a la "de".
Many other late 19th century authors described Andalusian language, particularly the lyrics of the cante flamenco (e.g. Schuchardt 1881).
In his Historias para niños sin corazón (1909-1912), Juan Ramón Jiménez offered another imitation of Andalusian speech, in the poem `La carbonerilla quemada' (Mondéjar 1991: 146):
Mare, me jeché arena zobre la quemaúra. Te yamé, te yamé dejde er camino ... ¡Nunca ejtuvo ejto tan zolo! Laj yama me comían, mare, yo te yamaba, y tú nunca benía!
In his Diálogo de la lengua, Juan de Valdés describes Andalusian seseo, although without explicitly attributing this regionally circumscribed trait to Andalusia: `... por hazer dizen haser, y por razón rasón, y por rezío resío.' For Nebrija, who had lived in Sevilla, seseo and ceceo were speech defects rather than regional traits.
Phonetics and phonology
Studies of Andalusian Spanish phonetics and phonology include Carbonero (1982b, 1982c), Lamíquiz (1982), The following features are most commonly associated with Andalusian Spanish, although as will be seen, there is considerable variation within the Andalusian dialect complex. These features are most accurate for `western' Andalusia, that is the provinces of Sevilla, Cádiz, and Huelva, the heartland of the post-reconquest Andalusian Spanish. This is also the region of Andalusia which represented the highest rates of emigration to the Americas; hence the `Andalusian' aspects of Latin American Spanish in reality reflect these western provinces more than being common denominators of all of Andalusia.
(1) Syllable-final /s/ is aspirated or more commonly elided. Word-final /s/ is routinely elided in all but the most formal or artificial speech. There are a few villages in the north of the provinces of Huelva, Córdoba, and Jaén, in which a final sibilant [s] is preserved with some frequency (Narbona et al. 1998: 164-5).
(2) Traditionally western Andalusia does not distinguish /s/ and /θ/; in fact Spanish seseo (the neutralization of the four medieval sibilants /s/, /z/, /ts/, /dz/ to /s/) appears to have originated in southwestern Spain. In reality, most of the rural areas have exhibited more ceceo (neutralization of /s/ and /θ/ to [θ]) although this is a socially stigmatized pronunciation. Currently the urban seseo of western Andalusia is spreading into small towns and rural areas, but there remains considerable variation and sociolinguistic stratification of this pronunciation. Since /s/ and /θ/ are not systematically distinguished, most Andalusians pronounce at least some words with [θ], not always in accordance with Spanish etymology. It is not uncommon for an individual speaker to pronounce a particular word sometimes with [s] and sometimes with [θ].
(3) Word- and phrase-final /n/ is velarized throughout Andalusia. In more advanced pronunciation the velar nasal disappears, leaving only a lightly nasalized vowel. This nasality sometimes disappears, particularly in final unstressed vowels, with the result that verb paradigms can be drastically simplified in vernacular speech (cf. Mondéjar 1970):
(4) The posterior fricative /x/ is a weak aspiration [h] in much of western Andalusia, and alternates with velar and uvular fricatives in the remainder of the region. In Jaén, for example, /x/ tends to receive a strong guttural pronunciation, and the province is jokingly referred to as `la tierra del ronquío' (Carrasco Cantos 1981: 89; Moya Corral 1979: 77-79). Much of eastern Andalusia has a velar fricative [x] (Narbona et al. 1998: 168-9).
(5) Traditional Andalusian Spanish retained many instances of Latin initial /f-/ as [h]: hembra [hembra], harto [harto], etc. Currently this pronunciation is confined to uneducated rustic speech, as well as to the formulaic language of Flamenco culture (e.g. cante jondo).
(6) Word-final and intervocalic /d/ routinely disappears, not only in the participal ending -ado (as occurs throughout most of Spain) but in other combinations as well (Narbona et al. 1998: 177-181). In Flamenco nomenclature, pronunciation without /d/ has become fixed: cantaó(r) `Flamenco singer,' tocaó(r) `Flamenco musician,' tablao `Flamenco dance club.'
(7) Word-internal preconsonantal /l/ and /r/ are frequently neutralized, except in formal speech. Proportionally, [r] tends to predominate as the result (alma > arma, espalda > ehparda). Aspiration of the syllable-final liquid sometimes occurs (carta > [kahta], as does occasional gemination of the following consonant (algo > aggo, puerta > puetta) (Mondéjar 1979: 398-402; 1991: chap. IX).
(8) In word-final position, /l/ and /r/ are frequently lost, except in the most formal speech. This is most noteworthy in verbal infinitives, but this pronunciation in principle affects all words.
(9) It is frequently stated that Andalusians do not possess the phoneme /λ/, and indeed the loss of this phoneme (yeísmo) in Spain is first attested for Andalusia. However, as in the rest of Spain, yeísmo in Andalusia has predominantly been an urban phenomenon, spreading gradually to rural areas. To this day there are small pockets of speakers (even in the overwhelmingly yeísta provinces of Sevilla and Huelva) who distinguish /y/ and /λ/. Mendoza Abreu (1985: 84-5) documents /ll/ for Lepe, Huelva, while Becerra Hiraldo and Vargas Labella (1986: 17-18) describe pockets of /λ/ in the province of Jaén.
(10) Nasalization of vowels not in contact with an etymological nasal consonant sometimes occurs in rural Andalusian speech. For example mejor [mehõŋ]/[mehõ], pie [pie~ŋ]/ [pie~], perejil [pere~hí] (Mendoza Abreu 1985: 36).
(11) In some regions, prevocalic voiceless stops are voiced, and are sometimes pronounced as fricatives: sagudir < sacudir, gabina < cabina, seba < sepa, porgue < porque, la pada < las patas (Becerro Hiraldo and Vargas Labella 1986: 18-19; Salvador 1969; Moya Corral 1979: 53-60).
(12) In a few regions voiceless stops are aspirated, particularly in stressed syllables: pero [phéro], todo [thó o] (Becerro Hiraldo and Vargas Labella 1986: 19; Moya Corral 1979: 60-62; Narbona et al. 1998: 147-8).
(13) The fricative pronunciation of /č/ as is frequent in some parts of Andalusia, particularly in Sevilla (@) and Jaén (Becerro Hiraldo and Vargas Labella 1986: 20; Narbona et al. 1998: 148-9).
(14) In some parts of the province of Jaén, the initial /l/ of definite articles is sometimes lost: a viga < las vigas, de o pobre < de los pobres, o que pasa < lo que pasa (Becerro Hiraldo and Vargas Labella 1986: 20; Moya Corral 1979: 42-3).
(15) Although Andalusian Spanish is noted for the neutralization of preconsonantal /l/ and /r/ (usually favoring [r] as the output), there is some neutralization of intervocalic liquids, a phenomena not typical of other Spanish dialect zones, but once found in early Afro-Hispanic language (Lipski 1986): suero < suelo, clavere < claveles, sare < sale, me fui roca < me fui loca (Becerro Hiraldo and Vargas Labella 1986: 20; Moya Corral 1979: 50-51).
(16) Throughout Andalusia, there is some neutralization of /l/ and /r/ in obstruent + liquid onset clusters; [r] is more usually the result of the neutralization, but [l] occasionally occurs, much as in earlier Afro-Hispanic language (Lipski @): prátano < plátano, prazuela < plazuela, reflán < refrán (Becerro Hiraldo and Vargas Labella 1986: 21; F. Salvador 1978; Moya Corral 1979: 127-8; Narbona et al. 1998: 175). Occasionally in rustic speech, the liquid is lost, much as in early Afro-Hispanic language: made < madre, ombe < hombre, karo < claro, otro < otro, ehtopeao < estropeados, puebo < pueblo, ejepo < ejemplo (Moya Corral 1979: 128).
(17) Syllable-final preconsonantal /r/ undergoes additional modifications, for example in Ubeda, Jaén /r/ > [δ]: cadne < carne, modedno < moderno, gobiedno < gobierno (Becerro Hiraldo and Vargas Labella 1986: 20). Very occasionally, intervocalic /l/ > [ ]: sade~ < salen, cadi < Cáliz, lo dío < los líos, se de < se les (Moya Corral 1979: 51).
(18) In vernacular Andalusian speech, /mb/ sometimes reduces fo [m], /nd/ to [n], and /ng/ to [ŋ] or [n] (Becerro Hiraldo and Vargas Labella 1986: 20; Moya Corral 1979: 88-93): tamié < también, estupeno < estupendo, teno < tengo.
(19) Intervocalic /r/ falls in rustic speech throughout Andalusia: quieo < quiero, tíalo < tíralo, ubiea < hubiera, señoa < señora (Becerro Hiraldo and Vargas Labella 1986: 20; Moya Corral 1979: 48). Loss of intervocalic /l/ is considerably less common: sae < sale, dwe < duele, famía < familia, bae < vale, peota < pelota, piare < pilares, úsua < Ursula, árboe < árboles, etc. (Moya Corral 1979: 45-6). Given widespread loss of word-final /l/, this may lead to a new type of plural formation: arbo/árboe, perá < peral/perae, animá < animal/animae.
(20) The palatal fricative /y/ often acquires a fricative pronunciation, known as y rehilada in Spanish (Moya Corral 1979: 81-84).
(20) The combination /rl/ in infinitive + clitic combinations sometimes emerges as [y] in rustic speech, much as in old Spanish: cantaye < cantarle (Becerro Hiraldo and Vargas Labella 1986: 27).
(21) The initial /l/ of definite articles sometimes assimilates to a preceding nasal in rustic speech: enna [en la] mesa, conna [con la] mano (Narbona et al. 1998: 176).
(1) Western Andalusian Spanish does not use vosotros or the corresponding verb forms; ustedes subsumes all second-person plural functions. This is one of the principal linguistic links between Andalusia (and the Canary Islands) and Latin American Spanish. In rustic speech, the pronoun ustedes may combine with verb forms belonging to the vosotros paradigm (Mendoza Abreu 1985: 103; Narbona 1979: 271): hasta que ustedes no se váyais; ¿a qué hora se vais a ir? The third-person plural clitic pronouns les or se may be replaced by vos [bo] in rustic speech (Mendoza Abreu 1985: 104): os he dicho [boh e dicho], iros fuera [dibo pa fuera]. In the second person plural, the combination se os (often pronounced sos or sus) may be used with reflexive verbs: se os [sos] va a caer gives rise to ¿Cómo sos llamáis? (Rodríguez-Izquierdo 1982a: 132; Becerro Hiraldo and Vargas Labella 1986: 41). In Jaén, where vosotros forms compete with ustedes, the verbal desinence often receives stress in normally unstressed endings such as the imperfect: traiáis < traíais, veniáis < veníais. Even first and third person forms are sometimes affected: habiá < había, traiámos < traíamos (Becerro Hiraldo and Vargas Labella 1986: 27).
(2) Diminutives with intrusive consonants after final stressed vowels (José > Joselito, café > cafelito) are found throughout Andalusia.
(3) There is considerable non-standard gender assignment to certain nouns, as in rustic Spanish worldwide: la calor, la alfiler, etc.
(4) The possessive pronouns suyo, nuestro, etc. are frequently replaced by de él, de usted, de nosotros, etc. (Mendoza Abreu 1985: 105).
(5) More so than in the rest of Spain, imperfect subjunctive usage overwhelmingly prefers -ra endings to the -se paradigms.
(6) Enclitic pronouns are sometimes used with finite verbs in colloquial speech: diríase, trátanse, etc. (Becerro Hiraldo and Vargas Labella 1986: 41).
(7) Analogical forms abound in vernacular Andalusian Spanish, as in rustic speech in the rest of the Spanish-speaking world: frega < friega, entriega < entrega, dijieron < dijeron, pidimos < pedimos, etc.
(8) The widespread loss of word-final consonants often leads to phonological restructuring in vernacular speech, resulting in the loss of canonical plural endings: árbo < árbol, árboles; re < res, reses, etc. (Carrasco Cantos 1981: 99). Non-etymological consonants also appear: quinqué > quinquenes, alfile(r) > alfilele, relo(j) > relore.
There are no unique syntactic characteristics of Andalusian Spanish. In compensation for the massive loss of word-final consonants, there is some increased use of overt subject pronouns which would be redundant in the presence of intact verbal endings (e.g. Mendoza Abreu 1985: 33 for Lepe, Huelva; Alvar 1955: 311; Rodríguez-Izquierdo 1982a: 129), but this is not as noticeable as in Caribbean Spanish dialects. Negative expressions of the sort más nada are sometimes found in Andalusia (Mendoza Abreu 1985: 139; Narbona et al. 1998: 119; Mondéjar 1970: 30; 1991: 163), but not as frequently as in the Canary Islands or the Caribbean. Narbona (1979, 1986) examines several syntactic combinations in Andalusian Spanish, none of which are unique or peculiar to this region. In questions, a second person pronoun may appear in colloquial speech without being attached as an argument of a verb: ¿Verdad tú? ¿No verdad usted que no es así? (Narbona 1979: 264). Also found at times is the use of periphrastic auxiliary verbs with lack of part participle agreement: llevo andado muchos kilómetros; llevamos visto ya tres o cuatro casas (Narbona 1979: 266-7). Very occasionally, double negation is found: `A mí eso no me ... no me gusta, no' (Rodríguez-Izquierdo 1982b: 155; 1985b). Other syntactic studies include Rodríguez Izquierdo 1985a, 1985b; Carbonero 1985; Fuentes 1985; Agudo 1985).
Details of the behavior of syllable- and word-final /s/
Although the statement that syllable- and word-final /s/ in Andalusian Spanish is weak and tends to aspirate and disappear is accurate for the entire region, there are many nuances of this phenomenon, distributed regionally, socially, and idiosyncratically, which merit additional attention.
(1) In general, when word-final /s/ is aspirated, this aspiration remains (usually with resyllabification) in prevocalic environments: los amigos [lohamigo(h)], es esto [ehehto]. As in other Spanish dialect zones, the final /s/ of articles sometimes remains as [s] before word-initial stressed vowels: los otros [losotro(h)], las once [lasonse] (cf. Alvar et al. 1973; Guitart 1981, 1982; Harris 1983: 46-7; Lafford 1982; Lipski 1984; Seklaoui 1989; Terrell 1977, 1979, 1981, 1983 and the references contained therein). However, throughout Andalusia, and in many Latin American dialects which aspirate word-final /s/, word-final prevocalic aspiration is blocked when the following word contains [h] (as a representation of the posterior fricative /x/) in the onset of the immediately following syllable: los hijos [losího(h)], los ajos [losáho(h)], los ojos [losóho(h)]. Blockage of aspiration before a following [h] may even occur before word-initial unstressed vowels, a context in which aspiration of /s/ is otherwise categorical in the same dialects: mis hijitos [misihíto(h)], más ajenos [masahéno(h)]. The blockage of /s/-aspiration before [h] takes one of two forms. The most common is the failure of aspiration to apply, retaining word-final /s/ as [s]. The alternative is complete deletion of word-final /s/, which is otherwise a rare outcome of prevocalic /s/-reduction in these dialects.
The blockage of aspiration only occurs when [h] is the immediately following consonant, with an intervening vowel. If one or more consonants intervenes between the word-final /s/ and [h], /s/-aspiration is not blocked: las orejas [lahoréha(h)], los ángeles [lohánhele(h)]. In the dialects under consideration, blockage of word-final prevocalic /s/-aspiration is general when the first consonant in the following word is onset-initial [h], and occurs only sporadically when the immediately following consonant is syllable-final [h] < /s/: los astros [loháhtro(h)]-[losáhtro(h)] (cf. Torreblanca Espinosa 1976: 58).
The inhibition of final /s/-aspiration has been noticed by Spanish dialectologists, whose comments are instructive in demonstrating that more than an occasional coincidence is at stake. A perusal of some of the accounts provides insight into native speakers' intuitions concerning disharmonic processes.
Carbonero (1982: 33), describing the speech of Seville, notes that `la s se pronuncia con regularidad cuando en la sílaba siguiente aparece el sonido j, para evitar la posible cacofonía que produciría la presencia de dos sonidos similares tan cercanos: no se dice "loh oho" sino los oho `los ojos'; ni "loh iho" sino los iho `los hijos'.' Similarly, Carbonero (1982c: 145) comments that aspiration of word-final prevocalic /s/ occurs except `los contextos en que está cerca otra aspiración procedente de la pronunciación de [x], donde por eufonía permanece la -s final de la palabra y ante vocal de la palabra siguiente. Así [los oho] y no [loh oho].' Vaz de Soto (1981: 79), describing the same dialect, states `... no es raro que reaparezca la "s," sobre todo en pronunciación culta o cuidada ... y siempre, por razones de eufonía, si sigue una "j" en la sílaba siguiente: "los-ojo" ... "mis ijo" ... "dos-ajo".' Torreblanca Espinosa (1976: 58), studying a dialect of southeastern Spain, declares `en el habla de Villena, hay otra posición donde el alófono [s] se conserva al final de palabra: cuando, en la palabra siguiente, aparece una fricativa laríngea, al final de la primera sílaba, o una fricativa faríngea, inicial de la segunda sílaba. Ejemplos: [lasähpä:] "las aspas", [dosóho:] "dos ojos".' Rodríguez Castellano and Palacio (1948: 591-2), commenting on a central Andalusian dialect in which SA is nearly categorical, note that `entre personas campesinas de avanzada edad, no se aspira esta s en palabras que tienen otro sonido aspirado, de cualquier origen que sea ... y se comprende que sea así, pues aunque la tendencia a aspirarse la s es general, en este caso tenía que hacer excepción por un sencillo motivo de disimilación. El que no se diga *loh oho "los ohos" se debe a la misma razón que ha impedido el mantenimiento de la h en la forma etimológica hiho (<filiu).'
Cummins (1974: 72) gives numerous cases of alternation between [s] and [h] for word-final prevocalic /s/ in southwestern Spain; although [s] does occasionally occur when the first consonant following the word-initial vowel is other than [h], /s/-reduction is uniformly blocked by the presence of a following [h]: [seis ihu] seis hijos, etc.
In the Canary Islands, where final /s/-aspiration is categorical in nearly all regional dialects, Catalán (1960) also notes blockage of final SA when two `aspirations' would occur in close succession.
La [s] surge, no sólo en el habla cuidada de la oratoria, sino conversacionalmente siempre que se entrecomilla o subraya una palabra; y desde luego, es aún obligada, en ciertos casos, cuando queda intervocálica por fonética sintáctica ... el progresivo desvinculamiento de la -[h] implosiva respecto al fonema /s/ se patentiza en la tendencia marcada del habla popular a preferir la -[h] ... en la mayor parte de las situaciones en que se convierte en explosiva por fonética sintáctica; sólo en las voces en que hay otra aspirada en la sílaba inmediata triunfa la -[s] por disimilación ... únos-óhoh, grándes-ohoh, ésos-ahnoh ... nos-ábreh la puérta, cómpras-óhah ...; en las voces esdrújulas la disimilación se produce aunque medie una sílaba átona entre las dos aspiraciones: fiéras-águilah, los-útileh de trabáho ... y, en contra-partida, no ocurre la disimilación en las voces agudas, pese a la aparente proximidad de las dos aspiradas: máh atráh, báh atráh ... noh iráh a desir ...
Catalán's description suggests an even longer-range disharmonic effect, but the inhibiting effects which he ascribes to [h] in combinations like los útiles is in reality a manifestation of the well-attested inhibition of /s/-aspiration before stressed vowels in determiner + noun combinations.
The maps of the Atlas lingüístico y etnográfico de Andalucía (Alvar et al. 1973) provide limited confirmation of the blocking of /s/-reduction before [h], despite the fact that only citation forms were used, rather than examples from connected speech. The two relevant test items are los ojos (map 1631) and los árboles (map 1632). In both cases, the final /s/ of a determiner is followed by a stressed word-initial vowel, a context typically favorable to the retention of /s/ as [s] even in /s/-reducing dialects. Indeed the /s/ of los in los árboles emerges as a sibilant in many regions, but among more radical dialects of eastern Andalusia, aspiration to [h] is also frequent. In some eastern Andalusian zones, the final /s/ disappears altogether, laxing the preceding vowel. In the case of los ojos, the final /s/ of los is uniformly a sibilant (except for some cases of total loss in eastern Andalusia). Significantly, the sibilant pronunciation is found even in those areas where final /s/ is aspirated in los árboles. No cases of the contrary distribution are found, i.e. where the first /s/ in los ojos is pronounced as [h] or elided, while the first /s/ in los árboles is pronounced as a sibilant. Given the nature of the sample, these data are not conclusive in themselves, but they confirm the implicational relation described above.
In Latin American Spanish, blocking of /s/-reduction is not widely documented, although careful observation of many dialects reveals configurations comparable to those found in Andalusia. In one of the few explicit descriptions, Castelli and Mosonyi (1986: 117) give data from Venezuelan Spanish: `Las palabras bisílabas que comienzan fonéticamente en vocal acentuada y cuya primera consonante es el fonema /h/, no permiten la aspiración de /s/ final de la palabra anterior.' As examples, the authors cite los ojos, buenos hijos and unos ajos. The restriction of SA blockage to disyllabic words appears to be coincidental; Spanish contains relatively few proparoxytones of the form Vh..., but it is likely that a combination such as los ágiles would also impede final SA in the same dialect.
Another environment not explicitly mentioned in the published literature, but in which blockage of SA is observable to a lesser degree, involves word-final /s/ when an instance of [h] precedes: the final /s/ in cajas especiales, ojos abiertos, coges algo, etc. also resists weakening to [h] in dialects where final SA is generalized. In recorded field interviews by the present writer this trend has been observed in Seville, the Canary Islands, and in several Caribbean dialects, including Cuban, Panamanian, and Puerto Rican.
No quantitative data will be presented here, since the text frequency of the sequence /...hVs#V.../ is too low to be statistically significant. Previous quantitative studies have not focused on this environment, since attention has been limited to the context immediately following word-final /s/ and its effects on SA.
Instead of blocking SA, an alternative route of evolution is followed in some dialects in which /x/ is realized as [h]: word-final prevocalic /s/ simply disappears when /x/ is the immediately following consonant (e.g. Carrasco Cantos 1981: 82, García Martínez 1986, Moya Corral 1979, Zamora Vicente 1943). Equally significant is the behavior of word-final prevocalic /s/ before a following [h] in dialects where SA is categorical, but in which the phonetic realization of intervocalic /x/ is significantly different from that of word-final prevocalic reduced /s/ > [h]. For example, Salvador (1957: 225) finds it noteworthy that in the Andalusian dialect of Cúllar-Baza, reduction of word-final prevocalic /s/ is not blocked by the presence of /x/ in the following syllable `como ocurre en otros lugares.' In the Cúllar-Baza dialect, final /s/ is elided rather than aspirated in this context, so the sequence [...hVh...] is avoided.
The data just surveyed give evidence of systematic, although not exceptionless, blockage of final SA before word-initial (usually stressed) vowels followed by [h], with the latter consonant preferably in intervocalic position. The inhibiting effect of initial stressed vowels on SA is combined with the avoidance of [...hVh...] to provide an especially propitious environment for the observation of consonantal disharmony. Determiner + noun combinations involving commonly-used nouns, e.g. mis hijos, dos ojos, provide frequently recurring syntagms in which avoidance of SA can be implemented. Speakers are likely to internalize such frequent combinations and to consistently realize the article with final [s]. Blockage of final SA in infrequent combinations such as niños ágiles, más ajeno, or tienes ajo is less apt to be noticed, since the first time such a combination is pronounced, a speaker is likely to apply final SA in the normal fashion. If the combination recurs frequently enough to provoke the cumulative malaise which ultimately triggers disharmony, only then might SA be consistent enough as to be noticed. This accounts for the fact that explicit mention of blockage of SA nearly always involves determiner + noun pairs.
There are additional circumstances in which the same disharmony phenomena may be observed. One case occurs in many regional dialects of El Salvador and Honduras, where word-initial /s/ is subject to reduction when following a vowel-final word (Lipski 1983, 1985, 1986). This phenomenon is also found in the vernacular Spanish of northern New Mexico and southern Colorado, and sporadically in vernacular Andalusian Spanish, where it is sometimes referred to as heheo (Narbona et al. 1998: 170-1; Mondéjar 1979: 397-8; Moya Corral 1979: 80-81). Narbona et al. (1998: 170) observe that `... el heheo ofrece todas las características de un proceso de cambio lingüístico embrionario que no ha podido desarrollarse y permanece bajo la forma de variación, esto es, afectando sólo a la pronunciación de determinadas palabras o restringido, todo lo más, al habla de personas muy localizadas, generalmente pertenecientes a estratos socioculturales muy bajos, y casi siempre en registros sumamento descuidados o familiares.' Lipski (1983, 1985a, 1986d, 1987) offers similar observations for the Central American Spanish dialects of El Salvador and Honduras. In practice, initial SA usually occurs in closely-knit syntactic configurations, such as determiner + noun (la semana), adverb + clitic (no se puede), noun + adjective or adjective + noun (parque central, cincuenta centavos), etc. A scan of an extensive corpus of recorded material fails to reveal a single instance of initial SA when the syllable immediately following the /s/ contains [h]: la cejuda, lo sujetaron, una sajadura, etc. Due to the scarcity of configurations allowing this assertion to be tested, the conclusions are suggestive rather than definitive, but the general patterns of initial SA in these Central American dialects fall in line with the behavior of final SA in southern Spanish and Caribbean dialects.
The phenomena described above all involve the presence of a word boundary, the extension of an originally syllable-final SA to intervocalic contexts through the juxtaposition of a vowel-initial word. In the case of word-initial SA in Central American dialects, the syntactic conditioning appears to result from a fleeting misanalysis of the position of the word boundary. If the avoidance of [...hVh...] combinations is the result of OCP-induced consonantal disharmony or similar constraints based on phonological adjacency on some appropriate tier, then there should be no necessary syntactic conditioning. In fact, the constraint in question does hold word-internally; its effects are obscured by the scarcity of opportunities for observation of the disharmony. There are two phenomena which provide corroborative data; one is the blockage of word-internal SA adjacent to [h], and the other is a global morphophonemic constraint against underlying /...hVh.../ combinations.
In the same Central American dialects where word-initial /s/ is weakened, the change /s/ > [h] occasionally affects word-internal intervocalic /s/, particularly before unstressed vowels. In Honduran/Salvadoran Spanish, reduction of word-internal intervocalic /s/ is never as common as in contexts involving /s/ adjacent to a word boundary, but at the vernacular level it is common to hear words like casa, cosa, presidente, etc. pronounced with intervocalic [h]. A scan of the same corpus fails to produce a single case where aspiration of intervocalic /s/ would produce a [...hVh...] sequence: Josefina, pasajero, pajizo, anglosajón, agasajar, masajista, and the like never exhibit the change /s/ > [h]. The low text frequency of probative cases makes the evidence more circumstantial than conclusive, but there is consistency with previous results.
Due to the phonotactic distribution of Spanish /s/ and /x/, when avoidance of [...hVh...] is manifested by blockage of SA, the /s/ in question occupies the first of the two relevant consonantal positions, while the following [h] is a realization of the posterior fricative /x/. For most Spanish dialects, the only way in which [...hVh...] could arise from two instances of underlying /s/ would be in the configuration /...s#VsC/, as in las hostias, los ostiones, estos asnos, etc. Torreblanca (1976)'s citation of SA blockage in los astros and las aspas, and Catalán (1960)'s observation of SA blockage in esos asnos both give avoidance of [...hVh...] as the motivating force. However, syllable-final [h] < /s/ does not routinely trigger SA blockage in any /s/-reducing dialect; otherwise, we would expect frequent retention of /s/ as [s] in hijastro, es esto, eres español, tienes asma, las cajas, tu gesto, etc. In fact, such combinations do not trigger SA blockage with anything approaching the regularity of combinations in which the second [h] is intervocalic, as in más ajo. In the phrases esos asnos and las aspas, the well-documented inhibition of final SA before stressed vowels in determiner + noun groups is the decisive factor; the fact that a syllable-final /s/ is realized as [h] following [...hV] does not usually override an otherwise expected SA. In other words, the true sequence avoided by blockage of SA is [...hVhV...]. The fact that not all instances of [h] are `equal' in blocking SA must be accounted for in any comprehensive analysis of these disharmonic phenomena.
(2) In many Andalusian dialects, when word-final /s/ is lost before a following word-initial vowel, sinalefa or the prosodic linking of the two vowels fails to occur. Thus, la hora [lao-ra] vs. las horas [la-o-ra]. Sometimes a brief glottal stop signals the presence of the deleted consonant (Uruburu Bidaurrázaga 1990: 17 and passim.; Rodríguez Castellano and Palacio 1948: 592). Occasionally, [s] or [ ] is retained when word-final /s/ is followed in close syntactic juncture by a vowel; this is more common when the word-initial vowel is tonic, but also occurs occasionally with atonic initial vowels. Mendoza Abreu (1985: 60) documents los hilos [lo ílo], los alfileres [laharfilére].
(3) Compensatory lengthening of the preceding vowel is a frequent concomitant of loss of word-internal preconsonantal /s/, particularly when the preceding vowel carries the main stress (Hualde 1989a; Uruburu Bidaurrázaga 1990: 45).
(4) In many parts of Andalusia, the combinations /sb/, /sd/, /sg/ emerge as [φ], [θ], and [x], respectively; in other words, the two sounds merge into a single voiceless fricative with the same point of articulation as the original syllable-initial voiced obstruent (Hualde 1989a; Uruburu Bidaurrázaga 1990: 49; Moya Corral 1979: 119; Mendoza Abreu 1985: 64-5).
(5) Although in general word-final plural /s/ in noun phrases is freely lost in Andalusian Spanish, without any necessary compensatory traits either in phonetics or in morphology (Ranson 1992), the first plural /s/ of a noun phrase is sometimes retained as an aspiration [h]: las casas [lah kasa] Uruburu Bidaurrázaga 1990: 68-70 and passim.).
(6) In general throughout Andalusia, women tend to retain more instances of syllable-final /s/ as an aspirate (sometimes even as a sibilant) as opposed to men of the same region and socioeconomic class, who as a group prefer elision of /s/ (Uruburu Bidaurrázaga 1990: 70 and passim.; Salvador 1952; Alvar 1956, 1969; ).
(7) In a small region of Córdoba province, the vowel /a/ in the final syllable of nouns and adjectives is realized as [e] when plural /s/ is lost: las pesetas [le pesete] (Ranson 1992; Alonso 1956; Alvar 1958; Becerro Hiraldo and Vargas Labella 1986: 30). The loss of other word-final consonants produces similar results: hospital [opité], cuántas [kwante], and even más mal [me me]. Ranson (1992) found the phenomenon to be rapidly disappearing, occuring only in the speech of the region's oldest residents. Alonso (1956) describes the geographical area of the `Andalucía de la E' as the triangle formed by the towns of Puente Genil (Córdoba), Estepa (Sevilla), and Alameda (Málaga)
(8) By far the most widely-studied aspect of Andalusian phonology, both descriptively and in theoretical analyses, is the behavior of word-final vowels following the elision of final /s/, /l/ and /r/. In most dialects of Spanish, vowels in closed syllables (ending in consonants) are articulatorily more lax and/or slightly lower than the same vowels in open syllables. This difference is subphonemic and is rarely noticed. In western Andalusia (meaning the provinces of Sevilla, Huelva, Cádiz, and parts of the provinces of Málaga and Córdoba), loss of final consonants is accompanied by tensing of the now word-final vowel. In other words, the vowel quality does not reflect the fact that a final consonant was once present: las casas and más perros sound the same as la casa and más perro. In much of eastern Andalusia, including the provinces of Almería, Granada, most of Jaén (Becerra Hiraldo and Vargas Labella 1986: 13-15; Carrasco Cantos 1981: chap. 1; Moya Corral 1979: 23-30), and parts of Córdoba and Málaga, the vowel laxness remains after the consonant has been lost, and is converted into a phonemic distinction: la casa [la kasa] vs. las casas [lä kasä]; sabe [saße] vs. sabes [saßë]. Psycholinguistic research has shown that eastern Andalusian speakers are capable of distinguish these pairs even in isolated word-reading tasks where no surrounding context is present. Such speakers are sometimes confused by the speech of western Andalusians, who sometimes lax final vowels in a fashion not consistent with the loss of consonants. Similarly, western Andalusians are rarely able to distinguish singular and plural forms in eastern Andalusian Spanish by vowel quality alone.
As an additional facet of Andalusian vowel laxing, in many varieties of eastern Andalusian Spanish, laxing of the final vowel produces vowel harmony, typically affecting vowels up to and including the stressed vowel: álamos [älämö], calabazas [kalabäsä]. The high vowels /i/ and /u/ are normally exempt from vowel laxing, in final syllables and through vowel harmony. In some Andalusian dialects, vowel laxing harmony extends sporadically and variably to pretonic vowels, and even to preceding clitics: me lo decías [mëlödësiä].
Early studies of this vowel laxing focused on its phonological implications, the possible creation of between three and five new vowel phonemes in these Spanish dialects (Alarcos Llorach 1958, 1983; Alonso, Zamora Vicente and Canellada 1950; Cerdà 1984; Contreras 1975-6; López Morales 1984; Llorente Maldonado 1962; Mondéjar 1970: chap. 3; Navarro Tomás 1939; Salvador 1977). There followed a large number of descriptive studies of Andalusian Spanish in general and specific regional and local dialects in particular. At the same time, a search was conducted among other varieties of Spanish, in Spain and Latin America, which might exhibit similar phonological behavior. In Latin America, despite considerable attention by descriptive linguistics and phoneticians, and despite much anecdotal evidence of individual cases of vowel laxing, no convincing case has been made for systematic vowel laxing of the sort found in eastern Andalusia. Within Spain, there is some evidence of compensatory phonemicized vowel laxing in Extremadura (e.g. Salvador Plans 1987a: 26), as well as in neighboring dialects of Murcia and Alicante (e.g. Sempere Martínez 1995: 26; Torreblanca Espinosa 1976: chap. 8). On the other hand, López Morales (1984) finds that most instances of laxed vowels in eastern Andalusian Spanish are accompanied by redundant morphological marking, e.g. subject pronouns or clearly plural articles or adjectives, and calls into doubt the possibility of a true phonemic value for vowel laxing.
Theoretical analysis of Andalusian vowel harmony within non-linear autosegmental or metrical models were commonplace in the late 1970's and early 1980's (e.g. Gómez Asencio 1977; Zubizarreta 1979), and have reappeared in more recent theoretical models (@). Martínez Melgar (1986) offers an experimental phonetic variation of the phenomenon of vowel laxing. Narbona et al. (1998: 142-5) describe a number of distinct configurations of vowel laxing and vowel harmony in eastern and central Andalusia. The parameters are presence of final vowel laxing, presence of leftward laxing harmony, compensatory lengthening of vowels as the result of consonant deletion, and retention of some phonetic vestige of aspiration of /s/. As many as eleven different combinations have been attested, some with clear geographical boundaries, and others occuring only as limited idiolects. Some of the possibilities include:
(1) Alternation of final vowel laxing and final vowel neutralization following loss of consonants, with preference for neutralization. This has been attested for parts of Córdoba province.
(2) Alternation of final vowel laxing and final vowel neutralization following loss of consonants, with preference for laxing. This is attested for several points in Córdoba, Jaén, Granada, Almería, and Málaga provinces.
(3) Final vowel laxing, compensatory lengthening, and sporadic aspiration is found in the province of Jaén, including the capital.
(4) The combination of final vowel laxing, compensatory lengthening, and systematic aspiration is found in the outskirts of the capital of Granada.
(5) Final vowel laxing, with occasional aspiration and sporadic laxing harmony is found throughout most of Andalusia, except for the westernmost areas which do not present systematic vowel laxing.
(6) In the center of Granada province final vowel laxing, laxing harmony, and vestigial aspiration occur predominantly.
(7) Final vowel laxing and compensatory lengthening with no laxing harmony is found sporadically throughout eastern Andalusia.
(8) Systematic final vowel laxing, laxing harmony, and compensatory lengthening is found in isolated areas of Córdoba, Granada, and Jaén provinces.
(9) Final laxing with no laxing harmony predominates in Almería.
(10) Final laxing with only sporadic laxing harmony is found in much of the center of Granada province, in parts of Córdoba province, and has even been attested for a few points in eastern Sevilla province and in Málaga.
(11) Systematic final vowel laxing and laxing harmony is found in southern Córdoba province.
Seseo, ceceo, distinción in Andalusia
Stereotypical Andalusian Spanish does not distinguish /s/ and /θ/, typically preferring /s/ (seseo), but also, at the popular level, realizing all sibilants as [θ] (ceceo). Historical reconstruction suggests western Andalusia as the locus of the merger of /s/ and /θ/, while throughout Latin America, no contemporary dialect distinguishes two sibilants, and apparently such distinction, which may once have existed, was very marginal, largely limited to recent immigrants from regions of Spain where /s/ and /θ/ are distinguished, and was rapidly displaced by the pan-Latin American seseo. A closer look at the regional and social distribution of sibilant phonemes in Andalusian Spanish reveals a much more complex situation, in which true seseo/ceceo, i.e. use of a single sibilant phoneme, is predominantly confined to the western Andalusian provinces of Sevilla, Huelva, Cádiz, and much of Córdoba, while many, perhaps most, speakers in the remaining Andalusian provinces distinguish /s/ and /θ/, although not always with the consistency characteristic of more northern dialects. Mendoza Abreu (1985: 60f.) documents extensive ceceo for a town in central Huelva.
Sawoff (1980) gives anecdotal and some quantitative data on seseo in Sevilla. He gathered his materials from urban and suburban middle-class neighborhoods, a sample more heterogeneous than the Norma Culta and Norma Media samples taken in the same city. In addition to interviews, Sawoff employed a methodology first used by William Labov in New York City, that of approaching strangers in stores and markets and formulating a question that would elicit a particular word: cereza, ciruela, etc. The results of the informal surveys are surprising, in view of the assertions of categorial seseo in Sevilla: between one third and one half of all respondents used [θ] where [s] was expected (but where /θ/ was the etymological source). Ceceo (use of [θ] to instantiate etymological /s/) was also common except among the most educated speakers. Sawoff concludes that seseo is no longer an accurate cover term for sibilant realization in modern Sevilla, if in fact such homogeneity ever prevailed.
Carbonero (1985) also examines Norma Culta, Norma Media, and Norma Popular data for Sevilla. His comparative analysis reveals that whereas 95%-100% of all speakers realize /x/ as [h] and aspirate or eliminate syllable-final /s/, seseo ranges from a low of 74% among educated speakers to a high of 100% in the habla popular. Ceceo, an eminently rural trait, appears only in the habla popular, and only in 19% of the speakers. This relative homogeneity of pronunciation across sociolinguistic groups contrasts with loss of word-final /r/ (23% culta, 41% media, 83% popular), loss of word-final /l/ (20% culta, 36% media, 85% popular), and neutralization of preconsonantal /l/ and /r/ (24% culta, 43% media, 77% popular), which demonstrate considerable variation across social strata.
Uruburu Bidaurrázaga (1990: 125-145) gives variational data on the realization of /s/ and /θ/ among young speakers in the city of Córdoba, capital of the province of the same name. As a province, Córdoba typically does not distinguish the two sibilants, preferring [s] (seseo) except for a few ceceo ([θ]-pronouncing) areas such as Montalbán and Montemayor. The results are strikingly at odds with the long-reputed seseo of Córdoba. Uruburu Bidaurrázaga (1990: 147-163) also conducted attitude surveys among secondary and university students in Córdoba. Some 20% found seseo to be undesirable, compared with 15% with positive opinions, and 45% who found no difference (the remainder of the respondents did not reply to this question). Ceceo on the other hand received only a 9% positive rating, a 30% negative rating, and a 40% indifference level.
Most of the province of Jaén distinguishes /s/ and /θ/, although pockets of seseo and ceceo exist. Speaking of the southwestern part of the province, Becerra Hiraldo and Vargas Labella (1986: 17) note that `El ceceo abunda en las capas populares y se considera más rústico y vulgar que el seseo; incluso en zonas de ceceo, las personas educadas tienden a eliminarlo, bien por el seseo bien por la distinción. El seseo de Andújar es una voluntaria adaptación de la fonética urbana de Sevilla; el de Baeza está en regresión en las capas cultas y jóvenes de la población ...' For Baeza, Carrasco Cantos (1981: 84) observes that `El seseo baezano es un fenómeno sociocultural, se da en las capas sociales de menor cultura ... no obstante, en su comienzo, creemos que tuvo que ser general; si esto fue así, se puede pensar que es un fenómeno que se bate en retirada, aunque todavía tiene toda su vitalidad entre los hablantes que no están sometidos a ningún influjo de la lengua oficial ... el seseo, por tanto, está en relación inversa con el grado de cultura de los hablantes ...' In Baeza, seseo is regarded negatively, unlike in western Andalusia, and children of seseo speakers seek to avoid this pronunciation.
In an early study, Dalbor (1980) provides anecdotal accounts of variation in sibilant pronunciation throughout Andalusia, providing a traveler's account of the considerable regional and idiolectal variation in southern Spain. Narbona et al. (1998: 133-138) offer as many as eight different configurations involving sibilants in Andalusia:
(i) Distinction /s/-/θ/ with the Castilian (apicoalveolar) /ś/. Although such pronunciation is not indigenous to any part of modern Andalusia, it can be heard in the principal cities, due to inward immigration as well as to the effects of mass media and educational norms.
(ii) Distinction /s/-/θ/ with Córdoba /s/ (alveolar, less apical than the Castilian /s/, not as dental as the Sevilla /s/). This pronunciation typifies the capital city of Jaén and is found in the vernacular of Almería province and elsewhere throughout central Andalusia.
(iii) Distinction /s/-/θ/ with Sevilla (dental-alveolar) /s/. This is a transitional pronunciation, found in western Andalusia among groups of middle-class speakers who are acquiring the distinction /s/-/θ/ while maintaining the regional dental-alveolar pronunciation of /s/. This distinction is often heard among middle-class and professional speakers in the capital cities of Huelva, Cádiz, Sevilla (where as many as 30% of the population may practice such distinction), and Málaga.
(iv) Seseo with Córdoba (alveolar) /s/. This pronunciation is found in the capital city of Córdoba and through the southern regions of this province.
(v) Seseo with Sevilla (dental-alveolar/predorsal) /s/. This pronunciation affects some two thirds of the residents of the capital Sevilla, and is found in urban centers in Huelva and Cádiz and to a lesser extent in Málaga. In the capital Granada, this pronunciation is gaining ground.
(vi) Ceceo (neutralization /s/-/θ/ in favor of [θ]). As has been shown above, this pronunciation characterizes much of rural Andalusia, although being sociolinguistically stigmatized and possibly receding in contemporary Andalusia. Overall rates of ceceo are highest in western Andalusia, and are lower in Granada and Almería.
(vii) Confusion of /s/-/θ/ with an intermediate variant. This unstable configuration is found among many speakers in Málaga and Granada provinces, and sporadically throughout all of Andalusia.
(viii) Sporadic mixing and crossover (ceseo/seceo). By definition these are unstable and transitional phenomena, and are found in all social strata throughout Andalusia.
Sociolinguistic identity of Andalusia
The issues of linguistic and cultural identity are nowhere more polarized than in Andalusia. Opinions and attitudes run the gamut, from fierce defenders of andaluz as a separate language and culture to those who minimize linguistic and cultural differences in favor of España, español or castellano. Despite the fact that Andalusian phonetic traits continue to spread throughout Spain (Salvador 1987b: 62, originally written in 1963, states that `el andaluz gana terreno'), attitudes of Andalusians and non-Andalusians alike show a long-standing ambivalence. Within Andalusia, there is a considerable segment of the population which believes that Andalusian Spanish is `poor' speech, while there exist equally ardent defenders of the properness, even superiority, of Andalusian language. For the city of Córdoba, Uruburu Bidaurrázaga (1990: 149-150) asked his respondents to name the language they spoke: 50% responded castellano, 30% called it español, and only 8% responded andaluz. Asked whether Andalusians spoke better, worse, or the same as in other Spanish-speaking regions, 4% responded better, 31% responded worse, and 60% believed that Andalusian Spanish is neither better nor worse than other dialects. In Sevilla, Ropero (1982) reports on similar surveys conducted in the late 1970's. Asked to name their language, 39% of the respondents said andaluz, 22% called it español, 27% called it castellano, 5% español andaluz, and nearly 4% responded castellano con acento andaluz. Asked whether Spanish was spoken better in Madrid or in Sevilla, 38% favored Sevilla, 32% favored Madrid, and 25% felt that the dialects were different but not ranked. To the question `¿Crees que el andaluz es un castellano mal hablado?' nearly 79% responded with a resounding no, while some 22% accepted the negative appraisal. On the other hand, 84% of the respondents confessed to feeling uncomfortable speaking the Andalusian dialect outside of Andalusia, while only 13% indicated no discomfort. The similar question `¿Piensa que el andaluz habla mal?' produced varying answers according to the profession and educational level of the respondents. The negative values range from 4% (university students) to nearly 39% (managers), while the positive ratings varied from 60% (workers) to nearly 92% (liberal arts professors).
Many observers have noted the highly polarized attitudes towards Andalusian Spanish within the region. Although many speakers feel that Andalusian is simply mal hablado, others fiercely defend the regional dialect, to a greater extent and with more fervor than occurs with other dialects of Spanish: `El habla andaluza--al menos, buena parte de sus rasgos característicos--no se reserva ... exclusivamente para la comunicación familiar o privada, sino que se emplea, por toda clase de hablantes, en cualquier situación ... se cree que es debido a que la minoría en cuyas manos ha estado durante mucho tiempo la mayor parte de los bienes y recursos productivos no ha necesitado diferenciarse lingüísticamente de las amplias capas populares. A ello, entre otras causas, podría responder la inclinación de bastantes andaluces a enaltecer su modalidad idiomática, incluso a sentirse orgullosos de ella ...' (Narbona et al. 1998: 23).
As for specific phonological traits associated with Andalusian Spanish, Carbonero (1982c) reports on quantitative data reflecting educated speakers in Sevilla. Aspiration of word-internal preconsonantal /s/ was `accepted' in 95% of the instances, with a 90% level of homogeneity among speakers and a 90% level of confidence (consistency) at the idiolectal level. Aspiration of word-final prevocalic /s/ occurs 70% of the time. Seseo is present in 70% of the data, but with only a 40% inter-speaker consistency rating. Neutralization of preconsonantal /l/ and /r/ is accepted only 27% of the time, while loss of word-final /l/ and /r/ has only a 21% acceptance rate.
Dating the reduction of Spanish final /s/ is the subject of an unresolved debate, with probable dates of inception ranging from the early 16th century or before (e.g. Boyd-Bowman 1975, Frago García 1983, 1984, 1993: 475-488; Lapesa 1980: 387-9, Menéndez Pidal 1962) to the early 19th century (e.g. Alonso 1953: 351). Claims of a very early date are based on isolated words, such as the often-cited Sophonisba > Sofonifa written by Fernando Colón (1488-1539) and highlighted by Menéndez Pidal (1942: 34), and may well be due to scribal error or idiosyncratic developments. Torreblanca (1989b: 284-5) asks why, if at the beginning of the 16th century even educated Andalusians like Diego Colón already weakened /s/, there exist large regions of Latin America where final /s/ is retained as a sibilant. He also points out that no variety of Sephardic Spanish systematically reduces final /s/, thereby pushing the earliest possible dates for widespread loss of /s/ in Andalusia well into the 16th century. Torreblanca finds no firm evidence of /s/-reduction before the early 18th century, in either Andalusia or Latin America.
Some early examples of apparent loss of /s/ or hypercorrect insertion of syllable-final /s/ (indicating widespread elision of /s/ in the same positions) begin in the 14th century, but are so sporadic as to suggest nothing more than the occasional lapse in unguarded speech: lo pechero (Sevilla, 1384), escriuano públicos (Alcalá la Real, 1492), el dicho exámenes (Sevilla, 1485-88). During the 16th and 17th centuries scribal attestations of loss of /s/ increase significantly (Frago García 1993: 475-488), but still constitute sporadic and isolated examples in texts which otherwise maintain all instances of final /s/: para que vo lo digan (1550), fueron sus padrino (1526-9), los engaño (1938), valisdación (1641), etc. By the 18th century, written attestations of aspirated /s/ in Andalusia are more frequent. In the play La infancia de Jesu-Christo (ca. 1784) by Gaspar Fernández of Málaga, we find evidence of the aspiration of word-final prevocalic /s/, through the use of both the grapheme s and the grapheme j (Narbona et al. 1998: 70-71; Mondéjar 1991: 144-5): los jojos, las jorejas, pobres jandrajos, las jarree, mal de jojo, qué jojos.
Turning to Latin America, Jiménez Sabater (1975) presents evidence that /s/-reduction became prevalent in the Dominican Republic at the beginning of the 18th century. Fontanella de Weinberg (1987), tracing the development of Spanish in Buenos Aires, discovered sporadic examples of /s/-reduction prior to the 18th century, but postulates that widespread weakening of /s/ began well into the 18th century. In the Guanacaste region of coastal Costa Rica (originally part of Nicaragua), evidence of systematic loss of final /s/ begins to appear in the middle of the 17th century, and becomes significant by the early 18th century (Quesada Pacheco 1990: 51-2).
Lloyd (1987: 349) arrives at a different conclusion, namely that weakening of /s/ was common at least among the lower social classes of Andalusia (and by extension in at least some American colonies), by the end of the 16th century. Alvarez Nazario (1982: 83-4) also leans in favor of a late 16th or early 17th century origin for /s/-weakening in the Caribbean. Boyd-Bowman (1975) presents examples from the 16th century Caribbean which he interprets as evidence of an early weakening of /s/.
Foremost among the research questions raised by the high degree of /s/-reduction in Andalusian Spanish are the causes for this phenomenon, regionally confined within Spain, although common in the history of other Romance and Indo-European languages. The logical possibilities are several, and speculation has run high on a phonetic issue which literally divides Spain geographically, culturally, and sociolinguistically.
The first possibility is that /s/-weakening is the product of Spanish-Arabic bilingualism in the southern reaches of Al-Andalus. Although initially attractive as a circumstantial hypothesis, this suggestion is easy to invalidate. In Arabic adaptations of Spanish words, the Castilian apicoalveolar /s/ was represented by shin, the Arabic character for /sh/. The later Andalusian dental /s/ was represented by the character @, closer to the Andalusian /s/. No Arabic borrowings of Spanish words lack syllable-final sibilants, and no extant Mozarabic or Morisco text gives the slightest evidence of the eliimination of final /s/.
Narbona et al. (1998: 72) are of the opinion that `la aspiración, y pérdida de -s parece un cambio surgido en un castellano ya formado.' However, there is nothing to suggest that /s/-weakening was carried from Castilian speakers moving southward during the reconquest. This is clear from the extreme resistance of syllable-final /s/ to effacement in most of Castile and León, although pockets of aspiration found in Cantabria and western Salamanca province. Narbona et al. (1998: 73) conclude that `todo parece, pues, indicar que en estas alteraciones de s se trata de un desarrolo paralelo en Andalucía al de otras regiones, con una relativa independencia mutua (es evidente que los contactos entre las gentes de los territorios, casi todos contiguos, que alteran la s implosiva contribuirían a consolidar el fenómeno), y que en Andalucía logró una implantación más homogénea gracias a la tantas veces mencionada autonomía de vida y costumbres que conoció la región en la Baja Edad Medía y siglos posteriores.'
It is equally improbable that the growing trend towards aspiration of syllable-final /s/ e.g. in La Mancha, Toledo, Salamanca, and other Spanish regions outside of Andalusia results from Andalusian migration, since such migration has been all but nonexistent until recent decades, when the principal Andalusian emigration has been to Cataluña and the Basque provinces. Pérez Galdós's Fortunata y Jacinta (1886) described the speech of Fortunata, a native of central Madrid, as `las eses finales se le convertían en jotas sin que ella misma lo notase ni evitarlo pudiese.' Galdós, himself from the /s/-aspirating Canary Islands, was sensitive to this pronunciation, which was already well underway in Madrid more than a century ago, although never coming close to the figures registered for Andalusia.
Walsh (@) and more recently Narbona et al. (1998: 73) have offered a novel explanation for /s/-aspiration in Andalusia. Noting that in many Leonese dialects preconsonantal /s/ was frequently pronounced as  (a phenomenon also common in Sephardic Spanish), and given the notable Leonese contribution to the reconquest of Andalusia, these authors suppose that a palatalized syllable-final /s/ was brought to Andalusia by Leonese speakers. When Spanish /sh/ was backed to /x/ (itself a weak aspiration [h] in Andalusia), syllable-final // was also affected, giving rise to the all-pervasive Andalusian /s/-aspiration. This proposal, while certainly possible, places severe strains on credibility in several respects. Leonse presence was significant mostly in western Andalusia, while aspiration of /s/ covers all of southern Spain. Many Leonese dialects did not palatalize syllable-final /s/ at all. Finally, the remaining cases of the shift // > [x] occurred in syllable-initial (onset) position, while /s/-aspiration is a syllable-final (coda) phenomenon. In phonology, few processes uniformly affect consonants in onset and coda positions; prosodic conditions usually result in different outcomes depending upon syllable position.
Weakening of /s/ in early Afro-Hispanic language
The representation of the speech of African-born black slaves or bozales in Golden Age Spain helps to shed light on the dating of consonantal reduction phenomena. Golden Age habla de negros has been the subject of considerable linguistic research, although the interpretations are still debated. The search is blurred by frequent stereotyping and exaggeration, reflecting a negative attitude toward ethnolinguistically marked varieties of Spanish, and which attributes to all of them a wide range of defects and distortions that are frequently an inaccurate repudiation of this group. In more recent times, the linguistic characteristics attributed to black Spanish speakers, in regionalist literature and folklore, have been simply those of the lower socioeconomic classes, found among the speech habits of more educated individuals, without any objective racial connotations (Lipski 1985b). However, a careful examination of available texts from 16th and 17th century Spain, numbering several dozen plays, entremeses, villancicos, romances and other items, reveals a striking degree of phonetic and morphological consistency, which suggests a reasonably accurate representation of bozal language. Quevedo (1988: 127) once joked in the `Libro de todas las cosas,' that `sabrás guineo [= bozal Spanish: JML] en volviendo las rr ll, y al contrario: como Francisco, Flancico; primo, plimo.' However, most writers gave a more realistic approximation to the speech of at least some Africans. The most frequent stereotyping lay in not in phonetic distortion but in humorous plays on words, such as the frequent cagayero/cagayera for caballero, in the use of onomatopoeia and songs, and in the repetition of stock lines such as `aunque negro(s), gente somo(s).' Sixteenth and seventeenth century Spain, when Andalusian consonantal reduction probably began and when Afro-Hispanic language was first recorded, is therefore a useful place to begin the search.
The conclusion that some attestations of Afro-Hispanic language can serve as mileposts for Spanish consonantal evolution, embodies two fundamental research hypotheses. The first is that the phonological structures attributed to African slaves in Golden Age and colonial Latin American texts are qualitatively accurate: although stereotyping and formulaic repetition is found, the most egregious cases of exaggeration and distortion involve the lexicon and plot lines. The second is that a comparative analysis of these texts will aid in determining the chronology of consonant reduction in other Spanish dialects. The relationship between the literary representation of bozal speech and regional Spanish pronunciation was ultimately bilateral and reciprocal, and this is reflected in the texts. African slaves took as input regional phonetic tendencies, partially altering them to conform to a broad cross-section of West African phonotactics. To the extent that consonantal reduction had already begun in regional Spanish varieties, bozal speech effaced the weakened variants even further, categorically eliminating sounds in contexts where native Spanish speakers performed only partial and variable weakening. Spanish writers gave graphological recognition to bozal pronunciation only when the latter departed significantly from prevailing regional trends, with the possible exception of the very lowest Spanish-speaking social classes, whose speech was also the object of ridicule. The lack of a given weakening process in bozal texts does not necessarily imply that the phenomenon was absent in the speech of Africans, but only that Spanish writers found Africans' speech no different from their own in this feature. The consonantal reductions in question follow well-established historical routes, so a comparison between Spanish and bozal texts can potentially delimit the earliest threshold for the advent of a given reduction process, while subsequent developments in the corresponding regional Spanish dialects allow for a reasonable hypothesis as to when, if ever, bozal and Spanish pronunciation became indistinguishable.
The earliest texts from Portugal and Spain contain no phonetic modification, but concentrate on grammatical oddities such as lack of agreement, use of uninflected infinitives instead of conjugated verbs, substitution of subject pronouns, and drastic syntactic simplification. Also found are African place names, onomatopoeic words which create the flavor of African speech or song, and inappropriate use of Spanish lexical items. Beginning with Sánchez de Badajoz in the first quarter of the 16th century (ca. 1525-30), phonetic modifications began to appear in literary representations of Africanized Spanish, although many texts from the same period contain no such modifications. Lope de Rueda incorporated phonetic deformations more consistently in his plays (ca. 1538-42), and by the end of the 16th century, certain phonetic characteristics had been established in the habla de negros. In the 17th century, Africanized Spanish was used extensively by major and minor writers, particularly in drama. Once `black Spanish' became established in Golden Age theater, the linguistic characteristics move sharply away from pidgin Portuguese, and acquire traits typical of Spanish `foreigner talk,' including considerable phonetic deformation. Judging by the literary examples, the transformation was completed in the last decades of the 16th century, since after the turn of the 17th century, bozal language becomes more consistently `broken Spanish.' This apparent dating may only reflect the solidification of a Spanish literary stereotype, in that Portuguese features may never have been present in significant quantities in Africanized Spanish, or may have disappeared during the first decades of the 16th century. Although documentation of /s/-weakening in pre-19th century Spanish is scarce, literary habla de negros exhibits loss of final /s/ beginning early in the 16th century. Many objections have been raised against accepting bozal cases as valid evidence for the evolution of Andalusian Spanish. Salvador (1981), noting the early loss of /s/ in bozal texts such as those of Lope de Rueda and Góngora, suggests that if loss of final /s/ had already been widespread in Andalusia beginning in the 16th century, these authors would not have attributed the phenomenon only to African slaves. Alvarez Nazario (1971: 84) categorically rejects any connection between Golden Age bozal texts and the development of regional Spanish dialects, claiming that the former `vienen a ser completamente independientes por su origen de las correspondientes evoluciones en el consonantismo del castellano central y meriodional.' Pereda Valdés (1965: 179-80) is of the opinion that `había más inventiva humorística que autenticidad en aquel lenguaje literario deformativo y onomatopéyico.' The opposite point of view is sustained, e.g. by Dunzo (1974: 121): `In an effort to transport local color to the stage, the Spanish playwrights portrayed in a remarkably accurate fashion the speech common to the Blacks of the era.' Deeper investigation into early Afro-Hispanic texts suggests that bozal speech is indeed of use in dating consonantal reduction in southern Spain, although not serving as a simple mirror of contemporary events. For example, loss of /s/ in Golden Age bozal texts demonstrates both internal consistency and compatibility with independently documented Afro-Hispanic language, thus giving to the bozal documents more credibility than suggested by the previous comments.
Loss of final /s/ first appears in bozal Spanish texts in the first decades of the 16th century, beginning with Sánchez de Badajoz and Lope de Rueda. The only consistent case involves the verbal desinence -mos. By the time of Vélez de Guevara's El negro del seraphín (ca. 1643), the final /s/ of second person verb forms is also variably elided (Sánchez 1979). These same texts show very limited instances of /s/-reduction where no morphological conditioning is involved, e.g. in word-internal preconsonantal position, or final lexical /s/. The frequent loss of /s/ in Jesús is probably attributable to Portuguese, or to clipping based on Jesucristo. Sánchez de Badajoz shows a few examples of loss of preconsonantal /s/, as in crito [Cristo], trequilado [trasquilado] and etar [estar] (the unreduced form estar occurs more frequently). These may be scribal errors or idiosyncracies, but their scarcity, in comparison with numerous cases of retained preconsonantal and word-final /s/, renders it unlikely that early bozal Spanish was eliminating syllable-final /s/ in a wholescale fashion.
Elimination of preconsonantal /s/ appears occasionally in a few of Lope de Vega's plays, written in the first two decades of the 17th century. In El santo negro Rosambuco we find vito [visto], riponde [responde], and Franchico [Francisco] (the latter name and the pronunciation without /s/ became a stereotype in bozal literary texts). The form paqua [Pascua] is found in a late 16th century romancerillo, alongside numerous instances of retained preconsonantal /s/. In the late 17th century bozal texts of Sor Juana Inés de la Cruz (actually written in Mexico), loss of preconsonantal /s/ is still very sporadic, with only a handful of cases in her entire Afro-Hispanic corpus: Flasica [Francisca], fieta [fiesta] (alongside fiesa and fiesta), naquete [en aqueste], etc.
In Sor Juana, we find some of the first consistent cases of another example of morphological conditioning of /s/-reduction: loss of plural /s/ in nouns when preceded by a plural article in which /s/ is generally retained: las leina [las reinas], las melcede [las mercedes], lus nenglu [los negros], lo billaco [los bellacos], las paja [las pajas], etc. This configuration, where plural /s/ appears only on the first available position of a NP, is typical of vernacular Brazilian Portuguese (Guy 1981), and is found in many basilectal varieties of Latin American Spanish, particularly those with a strong African connection, in the Dominican Republic, Colombia, Ecuador, and in the Spanish of Equatorial Guinea. In Brazil, highland Ecuador and Equatorial Guinea, syllable-final /s/ is generally not weakened to [h], while in the other dialects morphologically-conditioned retention of /s/ is combined with general loss of word-final /s/.
Other late 17th century bozal texts from Spain begin to show loss of /s/ across all components of the noun phrase, while retaining final lexical /s/ as well as the second person singular verb ending. In a villancico dated 1673 (S-26), we find ¿Lo bajo habemo veniro? ... ¿Lo tiple essá tura junta? Another song, dated 1699 (S-18), contains lines like Reye zamo del Oriente. An anonymous villancico dated 1661 (S-10) provides Hagámole plaça a lo Reye Mago turo lo neglo, e turo lo branco; another dated 1676 (S-6) offers Tlaemo mucho cantare and still another dated 1694 (S-3) contains lines like turu lo Neglico la noche de Nasimienta ha de andal como pimienta. These examples indirectly suggest that reduction of final /s/ in southern Spain and the Caribbean was still not conditioned by purely phonetic factors as late as the beginning of the 18th century.
The Golden Age texts show a striking imbalance in apparent cases of /s/-reduction. The corpus used for the present study reveals more than 440 instances of loss of /s/ in the verbal ending /-mos/, beginning in the early 16th century. This compares to a literal handful of other cases of /s/-loss, none of which occur until well into the 17th century. Among the latter, most involve plural /-s/. Although the exact proportions are irrelevant, it is clear that bozal pronunciation of the verbal ending -mos extended an already weakened pronunciation found in local Spanish dialects.
The earliest Afro-Hispanic texts from Latin America, such as those of Sor Juana, and some songs from Mexico and Peru, properly belong to the Spanish Golden Age tradition, but by the turn of the 18th century letigimately Latin American bozal language was developing. Afro-Mexican texts are of special interest, since many come from interior areas (e.g. Puebla, Oaxaca) where consonantal reduction has never been characteristic, but where contact with evolving dialects of southern Spain was intense in the 17th and 18th centuries. A small corpus of 17th century Afro-Colombian texts also comes from interior highlands where final consonants are not usually reduced. Afro-Peruvian texts are also instructive, since they represent both coastal areas (Lima) where consonantal reduction is characteristic, and the Andean zone (Cuzco) where elimination of final consonants never took root. A few of the 17th century Afro-Peruvian texts (e.g. P-1, P-2) appear to have come from Bolivia, known in colonial times as Alto Peru. In Buenos Aires and Montevideo, Afro-Hispanic texts represent the late 18th/early 19th century period, and provide a probe into a dialect zone where consonantal reduction stabilized at rates between those of Andalusia/Caribbean dialects and the Castilian/Andean zones.
The earliest surviving Latin American bozal texts, from 17th century Mexico, Colombia and highland Peru, show some loss of /s/ in the verbal ending /-mos/, as well as in Jesú/Sesú [Jesús], and various derivations of Francisco/Francisca, a stereotyped name in Afro-Hispanic texts. Hardly any other examples of loss of /s/ are found (although etreya < estrella appears in P-1, from late 17th or early 18th century Bolivia). Bozal texts from 19th century Argentina and Uruguay, as well as Afro-Peruvian texts from the 19th century (all representing coastal areas), show loss of /s/ not only in the desinence /-mos/, but in more general syllable-final environments. The earliest of these texts represent the end of the 18th century, while the latest come from nearly a century later. Of the Spanish bozal texts being considered, none shows reduction of preconsonantal /s/ without reduction in the ending -mos, while the opposite configuration is quite common, characterizing nearly the entire Golden Age corpus. The consistency of the textual data suggests a reasonably accurate transcription of bozal speech, from which it may be concluded that by the end of the 17th century, Afro-Hispanic speech was just beginning to effect wholescale elimination of preconsonantal /s/ and lexical word-final /s/. Andalusian and Caribbean Spanish were obviously not providing a model for elimination of all syllable-final /s/, since the imperfect language acquisition represented by bozal speech invariably reduced syllable structure, and never enhanced it. However, this does not necessarily mean that Andalusian Spanish showed no reduction of /s/ until at least the beginning of the 18th century, only that weakened variants were still perceivable by Africans. It appears that some preconsonantal and word-final /s/ among both Andalusians and Africans was already an aspiration [h], as early as the last decades of the 16th century. In contemporary Spanish dialects where aspiration rather than loss of preconsonantal /s/ is the rule (e.g. Argentina, Paraguay, Uruguay, Chile and much of Central America), native speakers are often unaware of their weakened pronunciation of /s/, asserting that they are actually pronouncing [s]. Literature from these regions rarely represents the aspirated /s/, which is considered unremarkable (cf. Lapesa 1980: 387); only when /s/ disappears altogether is this loss reflected in written form. A diachronic scan of Golden Age bozal texts shows that apparent loss of preconsonantal /s/ is proportionately more common towards the end of this period. Moreover, all instances of /s/-loss discussed so far have been in perceptually weak positions, preconsonantally or word-finally following an unstressed vowel. If Andalusian Spanish were already weakening (but not eliminating) /s/ in these positions, e.g. to an aspiration, Africans, most of whose languages do not contain the distinction between strongly and weakly stressed syllables, might easily fail to perceive any sound at all.
Reduction of Spanish syllable-final liquids embodies so many different phonetic manifestations that it is unlikely that a single chronology accounts for interchange of preconsonantal liquids, interchange of phrase-final liquids, and loss of phrase-final liquids. In each context, evidence is limited and contradictory, allowing an uncomfortably wide margin of error. In Latin America, modification of liquids is not well documented until the 19th century, although extrapolation in such zones as the Caribbean suggests that modification of liquids began substantially earlier. In Latin America, some early attestations of liquid reduction may reflect Peninsular Spanish arrivals rather than local developments.
Lapesa (1980: 385f.) suggests a very early origin of liquid reduction in some regions of Spain: Mozarabic examples from Toledo dating from the 12-13th centuries, and Andalusian examples (generally /r/ > [l]) in the early 16th century. The first examples are isolated and scribal error or idiosyncracy cannot be ruled out. Lapesa's examples involve the shift of /r/ to [l], while in contemporary Andalusia the opposite change of /l/ to [r] is more common in preconsonantal contexts, whereas elision of liquids is the preferred solution word-finally. Sephardic Spanish, often used as a linguistic time capsule from the early 16th century, is of little help, since sporadic interchange of liquids occurs in all Sephardic dialects (involving different words in each case), just as in regional Spanish dialects far removed from Andalusia.
Lloyd (1987: 347-8) analyzes the bulk of 14th and 15th century interchange of liquids as dissimilation rather than neutralization, since most involve words with two or more liquids. By the middle of the 16th century, the accumulation of words exhibiting interchange of /l/ and /r/ as well as loss of final /r/ leads Lloyd to suggest that liquid reduction was common in southern Spain before the end of that century. Adopting a much different stance, based on dialect geography as well as historical documentation, Alonso and Lida (1945) conclude that widespread interchange of syllable-final /l/ and /r/ did not occur until the beginning of the 19th century, particularly in America. In opposition, Alvarez Nazario (1982: 85-6) uses early examples from Puerto Rico to conclude that at least some weakening of liquids was occurring in Caribbean Spanish by the middle of the 16th century. Fontanella de Weinberg (1984, 1987) cites 17th and 18th century examples of interchange of preconsonantal /l/ and /r/ in the Río de la Plata zone, a zone where neutralization of liquids is today conspicuous by its absence. She concludes that by the end of the 18th century, neutralization of liquids was not uncommon among native-born residents of Buenos Aires. This neutralization included not only syllable-final liquids, but also intervocalic and postconsonantal /l/ and /r/. The incompatibility of these findings with contemporary Río Plata Spanish and other South American dialects remains unexplained. Similar data appear in 18th century Costa Rica (Quesada Pacheco 1990: 47-8), particularly in Guanacaste, although today this zone is not characterized by neutralization or loss of liquids in any position.
The softening of the posterior fricative /x/ to an aspiration [h] in Andalusia is similarly difficult to trace, but written attestations such as hentil < gentil, baho < baxo, etc., occur in 16th and 17th century texts (Narbona et al. 1998: 67-8). By the time of the first dictionary of the Real Academia Española (1726), Andalusians and Extremeños were contrasted with Castilians for their aspiration of /x/ and confusion of h, x and j.
The Spanish of Gibraltar
Belonging linguistically to Andalusia, although presenting several peculiarities, is the British colony of Gibraltar, less than 5 km2 of territory on the tip of a peninsula near Algeciras, in Cádiz province (Lipski 1987b; Mondéjar 1991: 165-6; Kramer 1986; Ruíz Fernández 1995, Fierro Cubiella 1997). The peninsula is dominated by the impressive Rock of Gibraltar (el Peñón de Gibraltar), and the city of Gibraltar, on the north shore, is the principal settlement. According to the 1981 census there were 26,479 residents of Gibraltar, of whom 16,640 were natives of Gibraltar, 3,459 were natives of the United Kingdom, 2,694 were natives of Spain, and 2,389 were natives of other nations. In addition there were some 2,265 family members (nearly all natives of the U. K.) accompanying British military personnel in this heavily fortified outpost. Comparing these data with the census of 1860 (Sayer 1862: 458‑9), we see that little has changed in more than a century. In 1860 the total population of Gibraltar was 25,179 inhabitants, and the civilian population was 17,647. Of the later group, 9,802 were natives of Gibraltar, 995 were from the United Kingdom, 1,892 were from Spain, 782 were from Genoa, 240 were `African Jews' [probably from Morocco], and 525 were Portuguese.
In addition to its permament population, Gibraltar houses a large temporary population of Spanish workers, some of whom live in nearby Spanish towns and cross the gate into Gibraltar daily, while others live in the British colony for varying periods of time, due to an Anglo-Spanish accord which allows for such temporary labor.
The British first occupied Gibraltar in 1704, during the War of the Spanish Succession. Felipe V, the French-backed pretender to the crown, was opposed by the Hapburg Carlos III, supported by an alliance of the Netherlands, Austria, and England. Most of the Spanish Gibraltarians fled to nearby San Roque; Italian fishermen in Catalan Bay chose to remain behind, apparently unconcerned with loyalty to one or the other royal pretender. Spain mounted an unsuccessful seige of Gibraltar, and although Felipe V eventually attained the Spanish monarchy, Spain was unable to recover Gibraltar. Eventually, under the conditions of the Treaty of Utrecht (1713) Gibraltar was officially ceded to Great Britain. Over the ensuing centuries, Spain has mounted several diplomatic and military efforts to regain Gibraltar, but all have fallen short of the objective.
Once Gibraltar became a British territory, Sephardic Jews from Morocco, Spain, Portugal, and Italy moved to Gibraltar in large numbers; at one point the population of Gibraltar was nearly half Jewish (Kramer 1986: 14). Throughout its history, British (i.e. British-born) individuals have been a minority in Gibraltar; Jews, Italians, and Spaniards formed the overwhelming majority, and Spanish was always the prevailing language of the territory. By the late 1800's Gibraltar, now under a civilian government, began to receive a sizeable Spanish labor force, beginning a trend which continues to this day. During the Spanish civil was more than 10,000 Spanish refugees poured into Gibraltar. During World War II the British government, fearing a massive Axis attack on Gibraltar, evacuated much of the population to Great Britain and the British West Indies. This forced exile had the effect of solidifying English as the dominant administrative language of the colony. At the same time the marginalized conditions of the refugee camps reinforced emotional ties to Spanish, the home language of many of the temporarily displaced Gibraltarians. Even during the war, Spanish workers continued to enter Gibraltar to work in the port and in other service jobs.
Although English has always been the official administrative language of Gibraltar, Spanish has at times been given more prominence in education than at present. Already in the 19th century Methodist schools had implemented de facto bilingual education; the Anglican schools soon followed suite (Kramer 1986: 33-4). In response, Catholic schools began to include more Spanish-language classes in their curriculum, although still attempting to wean students away from Spanish as a public language. The Irish Christian Brothers in particular encouraged use of Spanish, and even used bilingual readers in lower grades. Spanish continued to be required as a medium of instruction, especially in lower grades, and the governmental educational code of 1880 explicitly allowed for the use of Spanish as a means of explaining English terms and concepts. However, use of Spanish for other purposes was forbidden during school hours.
Matters continued apace until the mass evacuations of World War II; in 1944 the British formed an education committee for Gibraltar, which included in its mission the retention of Spanish through the creation of fully bilingual pupils.
Relations between Spain and the local government of Gibraltar have been strained ever since 1704, the date of the British conquest of this territory, and insistent Spanish claims over Gibraltar have on many occasions restricted free movement between Gibraltar and Spain. Although the majority of Gibraltarians have family ties with Spain, and despite the fact that the work contracts are mutually beneficial to Gibraltar and Spain, there have been time periods in which contact between the two nations was reduced to a minimum, direct travel (only a few minutes by land) had to be by air or sea through third countries, usually Tangier, Morocco. Currently the situation has stabilized in a more favorable configuration.
Within Gibraltar, there appears to be some systematic discrimination against native Gibraltarians as opposed to natives of the United Kingdom, particularly as regards civil service jobs. This results in a sociocultural polarization which underscores the bicultural nature of the colony and the often less than harmonious coexistence. Spain is similarly ambivalent in its treatment of Gibraltar and Gibraltarians. Many Spaniards see Gibraltar as a land of opportunity, due to its traditionally high standard of living (a distinction which has diminished considerably since the entry of Spain into the European Union), and the higher purchasing power of the British pound. In previous decades, Gibraltar also contained a broader array of consumer goods and was governed by a more desirable political system; currently these are moot points for Spain. At the same time, Spaniards often consider Gibraltarians of Spanish descent as lost souls, despised by the British and ignored by Spain. These sentiments have spilled over into the public arena; for example Spanish television once transmitted the statement that `los gibraltareños son ciudadanos de segunda o tercera clase, que no merecen tenerse en cuenta en las discusiones entre Inglaterra y España' (Bruzon 1967: 14). This covert hostility is not lost on Gibraltarians, and in a referendum held in 1967, annexation to Spain received a mere 44 out of the more than 12,000 votes cast.
The majority of Gibraltarians have at least some Spanish parentage, often resulting from the marriage of Gibraltarian or British men with Spanish women. Typical names are composed of a Spanish given name and an English surname; the opposite combination is also frequent, much as found in bilingual areas of the United States and Central America. There are few (perhaps no) native Gibraltarians who speak no Spanish, and in the majority of truly Gibraltarian households Spanish is in daily use, often combined with English. Even natives of the United Kingdom who have resided in Gibraltar for a considerable time use Spanish spontaneously in all but the most formal situations, and their children learn Spanish natively. Despite the recurring political difficulties which have impeded free passage between Spain and Gibraltar, linguistic contacts have never been cut off, and the Spanish dialect of Gibraltar is a variant of western Andalusian Spanish, as spoken around Algeciras and Cádiz. Even natives of Morocco, India, and other nations who have emigrated to Gibraltar learn this regional variety of Spanish.
In Gibraltar, linguistic reality contrasts sharply with official language policies, in which English is the only language acknowledged for all government activities, in newspapers, radio, and television of the colony. The Gibraltar radio station broadcasts a few Spanish-language programs, but these deal only with non-serious topics such as cooking, home decoration, and musical variety. Few native Gibraltarians listen to these programs, preferring instead the more interesting and varied programming of the nearby Algeciras stations. In Gibraltar schools, public and private, English is the sole language of instruction, and the British government has tenaciously resisted proposals for bilingual education programs, apparently in the belief that any official recognition of Spanish as a viable language in Gibraltar will encourage separatist and anexationist feelings among Gibraltarians. That the results of political referenda suggest no such feelings has not been regarded in maintaining the official stance that English is the only language of Gibraltar.
In Gibraltar schools Spanish is taught as an elective subject, and is presented as a foreign language, similar to French and Latin. Neither the textbooks nor the pedagogical approach found in most Spanish classes reflects the fact that the majority of Gibraltarian students speak Spanish natively or quasi-natively. At the same time, English classes presuppose that students are native speakers of English with full fluency in that language, an assumption which is at odds with reality. Naturally these official views have little effect on informal language choice, and Spanish continues to be the language of choice in most non-official domains. At the same time, despite the fact that virtually all children in Gibraltar attend school, the proportion of the Gibraltar population which speaks little or no useful English is surprisingly high. The educational system does indirectly acknowledge the frequently limited English skills of many pupils by organizing English classes according to the skill level of the students, from little or no English through total fluency. Some Gibraltarian households use English frequently, particularly if one of the parents is from the United Kingdom, and when possible many such families send their children to Britain for extended vacations and even to study. Among Gibraltarian children with no British family members and who speak only Spanish as a home language, very few speak English with native ability. Steward (1967: 72), despite demonstrating a sympathetic attitude towards the people of Gibraltar, commented that `the Gibraltarians look and sound so like us that their off‑beat English comes as a shock.' The `off-beat' English is not that of an isolated but native dialect left to run its course in the absence of normative tendencies (such as found, e.g. on the south Atlantic islands of Tristan de Cunha and St. Helena), but rather the second-language approximations of a fundamentally Spanish-speaking population. There are many older Gibraltarians who know no English, and even young residents who have studied English for many years may have little usable ability. My fieldwork in Gibraltar brought me into contact with many bright and motivated students who nonetheless spoke English with great difficulty. It is unlikely that such students can fully benefit from the rigidly monolingual British educational model; this may partially explain the noteworthy preference for natives of the United Kingdom for government jobs in Gibraltar. As late as the 1950's, some 60% of Gibraltarian students could not pass the English language exams administered by the British government (Stewart 1967: 72‑3), which underscores the discrepancy between official language policies and true language usage in Gibraltar.
Spanish, conspicuous by its absence in official domains, is learned in homes and on the street, and through the naturally permeable border with Spain. In Spain, the terms yanitos/llanitos are used to refer to Gibraltarians, possibly deriving from English Johnny or Italian Gianni (Italian immigration to Gibraltar was once heavy, and some elderly residents of Catalan Bay still speak the Genoese dialect).
In the phonological dimension, Gibraltarian Spanish fits neatly in with all western Andalusian traits, as regards velarization of word-final /n/, aspiration and loss of syllable- and word-final /s/, and neutralization and elision of syllable-final /l/ and /r/. Tables 1-3 give quantitative data which situate Gibraltar Spanish as unquestionably a western Andalusian dialect. As in rustic Andalusian Spanish, the level of phonological restructuring is considerable among Gibraltarians, most of whom have not studied Spanish formally and are not aware of originally underlying consonants which have disappeared in speech. Thus the plural of túnel (pronounced as [túne]) is also [tune], the plural of árbol [árßo] is [árßo], etc. This is similar to what has occurred in many rural Andalusian dialects, as well as in parts of Latin America (e.g. Terrell 1982 for the Dominican Republic). As with Andalusians outside of Spain, Gibraltarians have diminished the sociolinguistic differences which within Andalusia separate speakers of varying educational and socioeconomic levels. Given the status of Spanish as an oral, informal code in Gibraltar, Gibraltarians are not compelled to retain final consonants in formal speech, such as sometimes occurs in Andalusia. Elimination of final /s/, /l/, and /r/ is nearly categorical in Gibraltar, whereas within Spain there is some resistance to effacement in more formal styles, particularly as regards word-final liquids. As an anecdote, in the 1980's I recorded an interview in Spanish, broadcast on Spanish radio, between the British governor of Gibraltar and a Spanish radio announcer. The Spaniard, despite his obvious origins in Andalusia, maintained a formal diction and partially suppressed the Andalusian consonant-weakening. The governor, on the other hand, while maintaining impeccable grammar and an appropriately formal vocabulary, eliminated all final consonants without concern, a pronunciation which would be unlikely under the same circumstances within Spain. An exception to this trend is the minimal Spanish-language radio programming on (government-owned) Gibraltar radio; the announcers, occasionally revealing some Andalusian traits, speak with great articulatory precision and their speech is more similar to Castile than to the southern dialects of Spain. When questioned explicitly by this writer as to why they adopted a pronunciation which was so patently at odds with the speech of the surrounding community, the announcers responded that the decision was purely personal, and reflected no official policy. Indeed, there is no official policy whatsoever as regards Spanish-language programming in Gibraltar, and the announcers affirmed that they based their radio styles on visits to Spain and their observations of the speech of fellow radio announcers in Spain.
Gibraltarians do not typically distinguish /s/ and /θ/, and seseo (neutralization in favor of [s]) predominates, unlike the prevailing ceceo in rural areas of neighboring southwestern Andalusia. Kramer (1986: 87) speculates that the tenacious seseo of Gibraltar Spanish may stem in part from the fact that until at least the middle of the 19th century, many Gibraltarians spoke Genoese, which has only /s/. Kramer also asserts that the relative resistance of syllable-final /l/ and /r/ to neutralization may be due to a Genoese carryover, although my own field work suggests a rather high rate of neutralization. Similarly, Kramer (1986: 86-7) claims that velarization of final /n/ is nonexistent in Gibraltar Spanish and that even though final /n/ in velarized in Genoese, this pronunciation is considered vulgar and is avoided in Gibraltar. My own observations reveal a high rate of velarization of final /n/ in Gibraltar.
Lexically, Gibraltar Spanish is described by a small dictionary of regionalisms (Cavilla 1978), which dwells on lexical innovations, many coming from English, but also from Italian/Genoan. The latter dialect has given pavana `seagull,' bucherío `loud noise, racket,' cantín `tin can,' and leveche `southwestern wind.' From English come afolinearse (< fall in), afordar (< afford), chuar (< choose), guardao (< guard house), manolo (< man hole), rólipo (< lollipop). Kramer (1986: 68-75) gives a list of numerous Anglicisms, some frequent, others spurt of the moment translations, in Gibraltar Spanish. Bilingual calques and hybrid combinations are also frequent: estar guilty, hacer nitin (< knitting), dar un ring `call on the telephone.' Finally, loan-translations occur spontaneously, as in other bilingual Spanish-English communities: estar supuesto a (< to be supposed to), dar para atrás (< to give back), venir para atrás (< come back), etc. The expressions with para atrás (pronounced patrá) are also found in many bilingual communities in the United States (Lipski 1985c, 1987c; Otheguy 1993).
It is difficult to directly measure linguistic attitudes in Gibraltar, given the lack of official recognition of Spanish and the natural reluctance of Gibraltarians to challenge estalished authority, but the preliminary results of a survey conducted by a pair of educators (Flores and Ballantine 1983) are illustrative of the larger perspective. First, a questionnaire was administered to 70 primary school teachers, a majority of whom were from Gibraltar. 54% of the teachers believed that bilingualism was beneficial to the overall educational process, and only 6% believed that bilingualism could bring negative consequences. 86% of the teachers felt that the students' native language (usually Spanish) should be used to facilitate the learning of the second language (English), while only 10% maintained the opposing position. As for the proper designation of the unofficial language of Gibraltar, 15% of the teachers used the term `Spanish,' 10% used the term `yanito,' and 70% considered it to be `yanito/Spanish.' In speaking of their students' linguistic abilities, 94% of the teachers indicated that Gibraltarian students usually experience language-related academic difficulties, and 46% felt that these difficulties were widespread. 59% of the teachers believed that it was essential to employ Spanish in the classrooms (despite the official prohibition) to give instrucciones and to convey basic information (60% felt that Spanish was `very useful' in these cases). 52% of the respondents indicated that it was necessary to revert to Spanish in order to adequately teach their respective subjects (for 58% Spanish was `very helpful'); 73% indicated that it was necessary to use Spanish in teaching English (76% found Spanish `very useful'). These figures contrast sharply with a linguistic self-survey included in the 1970 census (this self-survey disappeared from successive census forms). In 1970, of the economically active populatio of Gibraltar, 92.6% responded that they could speak English (89.5% could read English), and 98.5% indicated that they could speak Spanish (95.6% could read Spanish). Of the population above 15 years of age, 81.3% could speak English (78.6% could read English) and 97.5% could speak Spanish (93.0% could read it). It is clear that the notion of `speaking English' reflected on the census forms is not strongly correlated with useable proficiency in the language, and falls far short of native-level fluencye.
The above-mentioned data can be compared with the results of another preliminary survey conducted with a group of secondary school students (Flores and Ballantine 1983). Through the use of verbal stimuli, in the `home' lexical domain, the test indicated that 46.8% of the pupils were Spanish-dominant, and 34.3% were English-dominant. In the `school' lexical domain, 62.% of the students were English-dominant, and 9.4% were Spanish-dominant. When non-verbal stimuli were used and one-word answers were solicited, 59% of the students tested as Spanish-dominant and 50% were English-dominant. When full-sentence responses were requested, 73% of the students were classified as Spanish-dominant, as opposed to 26% English-dominant. It should be recalled that these exams set out to measure general linguistic abilities attained after 7-8 years of intensive formal instruction in English, in the midst of a society which maintains English as its official language, all of which underscores the enormous discrepancies between official educational policies and the linguistic reality of Gibraltar.
Within Spain, there is a general opinion that Gibraltarians speak `poor' Spnaish (Stewart 1967: 72), although few Spaniards classify it as `inadequate.' In reality only Spaniards living in close proximity to Gibraltar and who have frequent contact with Gibraltarians have even the slightest information on Gibraltar Spanish. At one point I conducted informal `matched-guise' or `blind' tests in Spain, playing unidentified tapes of both Gibraltar and Andalusian Spanish speakers to residents of Spain; most Spaniards were unable to correctly identify residents of Gibraltar. On the other hand, if respondents were told prior to listening that the speaker was from Gibraltar, they frequently pointed out `errors' which they attributed to lack of education in Spanish and to interference from English, despite the fact that most of the traits also occur in vernacular western Andalusian Spanish.
Actual observation of language domains in Gibraltar yields results which are somewhat at variance with self-assessment surveys. As noted above, in Gibraltar there is no official policy regarding language choice, except in school instruction, government communications, and government-run broadcasting. In the remaining cases, language choice is determined by a matrix of sociolinguistic variables which characterize Gibraltar society. In the majority of linguistic exchanges which take place in Gibraltar, it is presupposed that all speakers possess the ability to express themselves in either English or Spanish, so that ultimate language choice reflects external factors such as context, style, and social setting. In reality, this working principle must be amended somewhat, since many Gibraltarians are not able to easily maintain a conversation held entirely in English, although their listening comprehension may be high. There are some cases in which Spanish is the preferred opening gambit in Gibraltar, under the assumption that the interlocutor will not be comfortable speaking English. This would be true, for example, with laborers and domestic servants, working-class housewives, and in general in certain neighborhoods of Gibraltar. The majority of predominantly Spanish-speaking Gibraltar households originate from Spanish immigration, and within Gibraltar these families as a general rule occupy a lower socioeconomic position than English-speaking households. Despite this demographic differentiation, it is not possible to base language choice on the physical appearance of a speaker, since many Gibraltarians with obviously Mediterranean/Spanish complexion and features are completely fluent in English, whereas even the most ruddy-cheeked blue-eyed Gibraltarians speak fluent Spanish. This fact notwithstanding, many Gibraltarians themselves claim to be able to predict language preference by looking at the features and complexion of a potential interlocutor. Stewart (1967: 59), a British civil servant who lived for ten years in Gibraltar in the 1950's, remarked that to initiate a conversation with an unfamiliar interlocutor would immediately provoke a response in English, and might well offend a Gibraltarian, who resented being mistaken for a Spaniard or a poorly-educated person. This may well have been the case of the high British official, whose face must surely have been familiar to nearly all residents of the tiny colony, but nowadays Gibraltarians are warmly receptive to visitors who initiate conversations in Spanish. During my own field work I was able to determine this in two different manners. First, I held monolingual conversations--some in English, others in Spanish--with diverse residents of Gibraltar, in no case revealing that I spoke the other language. Conversations held in English were typically somewhat stilted, even in the most casual contexts. Some Gibraltarians, obviously struggling to maintain a high level of English, slipped into what British colonial officials derisively refer to as `Babu English,' that is the erroneous use of erudite and even common words, and seemed unable to fully express their sentiments. Spanish conversations, on the other hand, revealed no reluctance or inability to express nuances and complex ideas, and no conversation held in Spanish appeared strained or artificial. As a second phase, I conducted conversations in which I began in English, and suddenly switched to Spanish after hitting on the `surprising discovery' that Gibraltarians also spoke Spanish. The results of this informal survey were striking, since nearly all interlocutors experienced visible relief, and the tone of the conversations immediately became more animated, punctuated with jokes, anecdotes, friendly profanity, and slang, which almost never occurred in English-language conversations. It was impossible to test the effects of the opposite change, from Spanish to english, since a friendly conversation in Gibraltar would never follow this course, which could actually be regarded as offensive, removing the conversation to a linguistically less comfortable code.
In formal contexts, for example in government offices, bilingual transitions are frequent, although not without occasional communication breakdowns, which induced Stewart (1967: 182) to quip that `everything one says in Gibraltar is half understood, everything said to one is half expressed. Every Gibraltarian you meet is using his second language.' Stewart related incidents in which Gibraltarians proudly refused to admit an insufficient knowledge of English, preferring instead to stumble through tasks inadequately. Obviously the situation has changed in the last half century, since it is now regional Gibraltarian English which defines exchanges in English, regardless of the observable prestige enjoyed by British-born government officials.
My own observations indicate that use of English is nearly categorical in at least the following contexts:
(1) When speaking to an unfamiliar government official, or to any government employee in a public or official setting;
(2) When speaking to any employee or receptionist in a government office when the speaker wishes to maintain a high level of personal dignity;
(3) When speaking to any obviously foreign tourist or visitor who is presumed to not come from Spain.
(4) In general in any situation which might sustain or enhance the prestige and social status of the speaker. This last trait reinforces the pro-British sentiments of most Gibraltarians, to the point that even in official closed-door meetings, at social events, and in public greetings among residents of high social standing, use of English is expected. During my fieldwork I had occasion to attend a public reception, British style, in the governor's mansion. All guests were dressed elegantly, despite the fierce Andalusian summer sun, and since most guests knew one another, greetings and conversations were frequent. As each guest entered the courtyard garden, they exchanged impeccable formal greetings in English. After a few minutes, conversations naturally drifted to Spanish and became more effusive and intimate. However, each time it was necessary to great a newly arriving guest, all responded in English, after which the conversation immediately returned to Spanish. The childeren playing in the garden spoke Spanish to one another, but when a parent had occasion to speak to a child in a loud public voice, English was always used. Finally, as the guests were leaving, as they passed out of the courtyard gate, goodbyes were once more said in formal English.
The majority of religious services in Gibraltar are held in English, and the less common Spanish-language services carry a lower social esteem. Newspapers are published in English and Spnaish, but the Spanish-language papers have small circulation and appear only at irregular intervals. In short, although Spanish is used in the majority of Gibraltar households, in public speech usage is more nuanced, affected by official norms and the social aspirations of the speakers. The need to maintain rigid if informal linguistic protocols gives rise to a transitional period upon leaving a workplace where use of English is expected on the job. During rest periods and as the work day comes to a close, English-only usage shades off into a free code-switching, which in turn gives rise to predominantly Spanish conversations. Telephone conversations are more diverse; when the speaker feels no need to reaffirm social status and presumes that the interlocutor is in a similar situation, Spanish is the usual gambit, with frequent incursions of English. On other occasions English is the preferred medium, and mixed-language conversations are not uncommon, in which one speaker uses Spanish and the other responds in English, being in a situation where use of Spanish would not be appropriate. The social status associated with English rapidly dissipates in non-official settings, especially among individuals who express themselves more easily in Spanish.
There is a striking similarity between the linguistic profile of Gibraltar and those speech communities where a European postcolonial metropolitan language competes with a creole language of inferior social status, often derived from the same European lexifier language (e.g. Jamaica, Guyana, Martinique and Guadeloupe, Cape Verde, and to some extent Haiti) or another European language (e.g. Suriname, Curaçao, Equatorial Guinea, Cameroon, Trinidad. What is unique about Gibraltar is that the yanito Spanish dialect is in no way creolized, nor is it even essentially different than neighboring Andalusian Spanish. In Gibraltar Spanish is not the language of an oppressed group, such as occurs in some parts of the United States, since native Gibraltarians are the descendents of British subjects and Spaniards who emigrated voluntarily to Gibraltar (the original Spanish residents abandoned Gibraltar more than 250 years ago following the British conquest). Despite the sometimes contentious Hispano-British relations centered on Gibraltar, residents are in closer contact with prestige registers of Spanish than with comparable sociolects of English. In terms of their own linguistic production, Gibraltarians come closer to educated norms of Spain than of Great Britain. Despite this fact, the British government has maintained the position that any official recognition of Spanish as anything other than the lingo of immigrants and temporary laborers would nourish separatist tendencies. The end result is a high degree of ambivalence among Gibraltarians, who on the one hand are more fervently pro-British than most residents of the United Kingdom and insist that Gibraltar is an English-speaking colony, and on the other hand maintain Spanish as the primary language of daily intercommunication.
There are also noteworthy parallels between Spanish-English bilingualism in Gibraltar and in the United States. In both regions English is the official language, enjoying higher prestige, and bilingual contacts give rise to a wide range of code-switching phenomenona, semantic and syntactic calques, and mutual influence at all levels. In Gibraltar, however, residents speak both Spanish and English voluntarily and not through coersion, as sometimes occurs among Spanish speakers in the United States, and frequently take the more difficult path of communication in English for purely extralinguistic reasons. In every respect, Gibraltar represents a unique linguistic configuration, different from both Spain and the United Kingdom, and exhibiting both similarities and differences with respect to other bilingual communities throughout the world.
Table 1: Behavior of /s/ in Gibraltar and other Spanish dialects (%)
/sC/ /s#C/ /s##/ /s#V/ /s/#v
[s][h][Ø] [s] [h] [Ø] [s] [h] [Ø] [s] [h] [Ø] [s] [h] [Ø]
1 96 3 0 94 6 4 1 95 78 8 14 4 17 79
2 91 7 0 94 6 9 8 83 23 77 0 0 95 5
0 95 5 0 91 9 5 2 93 69 10 21 1 46 54
0 82 18 0 85 15 1 2 97 0 15 85 2 50 48
2 85 13 0 89 11 2 17 81 75 25 0 0 92 8
94 6 0 69 29 2 82 12 6 92 8 0 96 4 0
11 25 62 29 14 57
C = consonante; # = word boundary; ## = phrase boundary;
V = stressed vowel; v = unstressed vowel
Table 2: word-final /n/ in Gibraltar and other Spanish dialects (%)
[n] [ŋ] [Ø] [n] [ŋ] [Ø]
Gibraltar 2 43 55 54 36 10
Cáceres 0 65 35 21 51 28
Sevilla 2 42 56 40 38 22
Granada 0 77 23 48 35 17
Las Palmas 18 49 33 54 34 12
Madrid 98 0 2 97 0 3
Melilla (pop.) 0 40 60 53 33 13
/n/ ## = phrase final (muy bien)
/n/ #V = word final prevocalic (bien hecho)
The Spanish of Ceuta and Melilla
The Spanish enclaves of Ceuta and Melilla, along the northern coast of Morocco, are esentially linguistic continuations of Algeciras and Málaga, respectively, although the presence of a substantial Arabic-speaking Moroccan population brings first- and second- language varieties of Spanish in contact. Since both enclaves are sustained by Spanish military garrisons, housing soldiers from all parts of Spain, a certain amount of dialect levelling also takes place.
González Las (1990) studies the pronunciation of Spanish in Melilla. This dialect reveals velarization of word- and phrase-final /n/ at rates comparable to those found in Andalusia. The posterior fricative /x/ is more usually a velar fricative [x], with the aspiration [h] being less common. The /s/ is alveolar, and there is some distinction /s/-/ /, although seseo predominates. Syllable- and word-final /s/ are aspirated or elided at high rates, although more educated men retain a sibilant [s] at rates considerably higher than found in Andalusia. Word-final /l/ and /r/ are also lost, but at rates lower than in vernacular Andalusian speech.
Emigration from Andalusia to other areas of Spain, and to industrialized countries in western and northern Europe, has been considerable in recent decades, and transplanted Andalusian speech communities are found in several European cities. The Andalusian immigrants typically represent the lower socioeconomic classes, since economic necessity is the primary factor influencing emigration, and the educational level in Spanish is correspondingly low. As a result, vernacular tendencies confined to the most informal registers and lowest sociolects in Andalusia are frequently extended to embrace all members of expatriate Andalusian communities (much as happens with Gibraltar Spanish, and New York and Dominican Spanish on the U. S. mainland). Children born of Andalusian parents abroad acquire phonologically restructured varieties of Spanish in which variable consonantal modifications have become categorical, and in which underlying representations have been modified to reflect consonantal neutralization and erosion. Narbona et al. (1998: 187) offer examples of uneducated Andalusian Spanish abroad, exemplifying the massive consonantal modifications which have spread unchecked in environments in which formal instruction in Spanish is lacking and little normative Spanish is available in the surrounding environment:
El paladá e otra folma en Epaña que aguí. Kando mi mae ase aguí una paeya faltan la cosa que mete y ayí el pehkao ma freco y etá ma güeno [el paladar es de otra forma en España que aquí. Cuando mi madre hace aquí una paella faltan las cosas que mete y allí el pescado es más fresco y está más bueno]
Po la alemana (a)hín mucha coha rebuelta la ehpañora son mucha behe aín po en una coha (a)hín to de una be (a)hín [pues la (comida) alemana, así muchas cosas revueltas; la española son muchas veces, así pues, en una cosa así, todo de una vez así]
The authors note that `lo que verdaderamente llama la atención en la forma de pronunciar el español de estos hablantes es la concentración e intensidad de tales rasgos, fenómeno que en Andalucía sólo se da de manera excepcional en individuos muy incultos y que tienen poco contacto con su entorno social, cada vez menos numerosos por la profunda modificación de los modos de vida tradicionales ...' (Narbona et al. 1998: 187-8)
From a letter published in a Spanish newspaper in 1986 by a businessman from Algeciras, Cádiz, in response to a letter written in Catalan (Narbona et al. 1998: 17):
Agecira Mare (Cai)
24 d'enero 1986
He recibío zu carta de fesha catorse der corriente me d'Enero.
Nó ha sío una jartá de difisi enterarno de los sucedío, y má o meno eztamo casi orientao.
Lo que toavía no z'entiende der tó e lo de "Que fem aquest estiu?" y ezo no lo podemos conchabá. En cuantito lo zepamos le contestaremo con musho arte.
Eá, zeñores, quedar con Dió.
Table 3: Behavior of /l/ and /r/ in selected Spanish dialects (%)
/r/C /#C r/## /l/C /l/#C /l/## ______________________________________________________________________
[r] [l] [Ø] [r] [l] [Ø] [r] [l] [Ø] [l] [r] [Ø] [l] [r] [Ø] [l] [r] [Ø]
82 0 18 56 1 43 5 0 95 31 29 40 47 17 36 11 0 89
96 0 4 88 2 10 47 0 53 84 0 16 81 1 18 56 0 44
84 2 14 70 0 30 29 0 71 43 38 18 47 18 35 38 0 62
89 1 9 65 1 34 27 1 72 48 31 21 52 10 38 25 0 75
73 12 15 59 17 24 28 8 64 65 23 12 73 2 25 36 7 57
99 0 1 99 0 1 97 0 3 100 0 0 100 0 0 99 0 1
51 0 49 52 0 47
C = consonant; # = word boundary; ## = phrase boundary
 The apparent reluctance of a wide cross-section of Spanish dialects to accept phonetic sequences of the sort [...h...h...], whether the aspiration [h] instantiates the posterior fricative /x/ or the reduction of /s/, is the signature behavior of the Obligatory Contour Principle (OCP), a theoretical postulate accepted by many phonologists (McCarthy 1986; Schein and Steriade 1986; Yip 1988, 1989), which disallows identical adjacent elements in phonological representations. The OCP was originally proposed to cover suprasegmental or autosegmental manifestations such as lexical tone, but there is evidence in some languages that the OCP serves as a quantitative constraint on certain combinations of vowels and consonants, both in underlying representations and as the result of superficial phonetic modifications. In many languages, OCP effects involving sequences of nonadjacent segments do not directly manifest themselves by blocking or triggering rules, but may exist as more subtle morpheme structure constraints whose synchronic existence at times emerges only upon a detailed quantitative cross-section. For example, Mester (1986) verifies the existence of OCP effects and hierarchically arranged feature tiers in Javanese by computing co-occurrence frequencies for a large number of Javanese morphemes. Such morpheme structure constraints, while formally describable by the OCP, may become partially lexicalized across time, admitting lexical exceptions particularly among recent borrowings or neologisms. OCP effects may represent the results of gradual evolution, and occasional exceptions do not necessarily invalidate the postulate of OCP-induced co-occurrence constraints. In scanning the Spanish lexicon, it becomes apparent that for all practical purposes, Spanish lacks morpheme-internal sequences of the type /...xVx.../ or /...hVh.../. The ready availability of combinations in which another consonant intervenes, such as Jorge, ajonjolí, jengibre, etc., demonstrate that consonants block any potential OCP effects triggered by two instances of /x/ separated only by a vowel. The handful of Spanish words containing /...xVx.../ sequences have a vanishingly low frequency in normal speech, are borrowings from non-Romance languages, and often prevail in dialects where /x/ is given an obstruent pronunciation: jején `gnat,' jeja `white wheat,' jijallo `a type of bramble,' jojoto `immature corn [Venezuela],' jojana `mocking tone of voice,' jajá type of bird [Argentina],' Gijón `a city in northern Spain,' Jujuy `a city in northern Argentina,' etc. The reasons for the low frequency of /...xVx.../ combinations in Spanish may originally have been accidental, reflecting the early Romance sources of modern Spanish /x/. The latter sound (which was pronounced [s] or [z] in medieval Spanish) comes from a variety of sources, including the intervocalic cluster /-kl-/ (e.g. oculu > oclu > ojo), the combination /li/ (e.g. filiu > hijo), the combination /-ks-/ (e.g. fixu > fijo), initial /g-/ (e.g. gentem > gente), and several other possibilities. Rarely in Romance did more than one of these proto-sources of /x/ appear in close succession, separated only by a vowel. Aside from the relative scarcity of /...xVx.../, the reluctance of native speakers to freely accept such combinations is somewhat unusual in Spanish, where combinations of identical consonants separated by a single vowel are not uncommon. Certainly /...kVk.../ and /...gVg.../ sequences are not infrequent (although the very term cacofonía suggests that some speakers may find the combinations less than euphonous), so that neither [+back] or [+high] is necessarily implicated in the general avoidance of /...xVx.../ sequences. Similarly, sequences of identical labial, dental, and palatal consonants are evidently not excluded by morpheme-structure constraints of the OCP type, although there exist few phonological processes operating intermorphemically which would further test the possibility of OCP-based co-occurrence restrictions. The existence of word- internal avoidance of [...hVh...] provides additional support for the notion that some type of OCP-induced phenomenon is at stake. The presence of word boundaries is not crucial to the constraint, although combinations involving word-final SA provide the most common test cases. Once expanses larger than a single word are involved, additional prosodic conditioning must be invoked. Given the variability of SA reduction in different environments, the precise domain of the constraint is not clear, but the largest expanse appears to be the clitic group (in the sense of Nespor and Vogel 1986). This accounts for the frequent blockage of SA in determiner + noun and adjective + noun combinations. Occasionally the entire phonological phrase may be at stake, as suggested by Catalán's example of compras hojas, a V + DO sequence.