Code-switching and dialogism: Verbal practices among Catalan Jews in the Middle Ages

From a strict linguistic viewpoint, code-switching intertwines with a diverse range of language contact phenomena, from strict interference to several kinds of language mixture. Code-switching has also been addressed as an interactional phenomenon in everyday talk, an approach that implies a synchronic perspective. In this article, however, data are drawn from the records of communicative practices left behind by Catalan Jewish communities of the 14th and 15th centuries. These communities lived under well-defined cultural, political, and social conditions and displayed a rather complex linguistic repertoire of both linguistic resources and verbal genres. I analyze two of these verbal genres, which themselves must be viewed in the context of a broader Hispano-Arabic cultural tradition; they draw on a heteroglot background in which Semitic and Romance languages merged. In this analysis of the functions that code-switching played in these verbal practices, a contrast emerges between the use of code-switching and lexical borrowing (or alternation vs. insertional types of code-switching) in both verbal genres. This has implications for a much debated issue – the alleged existence of a medieval Catalan Jewish language – and challenges the idea that forms of linguistic practice must always be reduced to a bounded code.

sal grammatical constraints on the switching points in utterances (or in their abstract correlate, the sentence), or toward the description of the socio-communicative functions performed by code-switching itself in particular speech communities and speech events. In a further step, the articulation of pragmatic or anthropological explanations for the presence or absence of code-switching in local communities has been attempted (Heller 1988). Both the contrast and the relevance of the opposition between local and global contexts in socio-cultural, socio-economic, and socio-political terms have been emphasized, and the idea of a local economy of language resources in speech communities has increasingly taken shape (Gal 1988). Thinking in terms of language resources instead of language systems has proven to be an important change in the search for a social foundation for dynamic linguistic facts, as I hope will become clear below in my consideration of medieval Catalan Jewish communities and their verbal genres and practices.
Code-switching is defined as the use of two languages or two codes in the same communicative event between speakers who are bilingual to a certain degree. As it is usually understood, then, code-switching implies the recognition of previously coexisting codes and their interplay within particular speech events in speech communities. The data under analysis show that this common premise needs to be qualified, even if one admits that, more often than not, the aforementioned characterization includes very different kinds of linguistic and sociolinguistic phenomena.
Insofar as human communication -as can be observed in everyday experience -involves linguistic, interactional, and sociocultural components, the issue of codes becomes highly relevant for synchronic linguistic, pragmatic, and anthropological analyses. It should come as no surprise, then, that concepts and analyses elaborated from differing perspectives prove contentious. If, in addition to synchronic data and analyses, we are concerned with the way codes and codeswitching have manifested themselves through history, in periods other than our own, then we must also rely on other disciplines, such as philology, historical linguistics, and social history. These may lead us to look at things in a new lighteither consolidating our current concepts and analyses or leading to realization of the need for a new conceptualization -but eventually they may also increase contention. The important point, however, is not so much how many alternative analyses are made available to our discipline-driven way of looking at data and interpreting them, but rather to what extent differing viewpoints can mutually enrich one another through cooperatively crossing established academic borders. If an integrated view emerges from this move, perhaps it will be seen as a further step toward development of a precise methodology in the field of historical sociolinguistics, or a subfield within it.
Because code-switching is a phenomenon that is both inherent to bilingual conversation -and thus to the oral use of language -and has particularly drawn the attention of ethnographers of communication and interactional sociolinguists, it is understandable that the synchronic approach should have prevailed. In this essay, however, I present some instances of code-switching drawn from recorded historical data from the Middle Ages, and I advance several explanations that take into account the special nature of historic texts. If the search for diachrony within synchrony may improve the picture we obtain from both -according to Jakobson's repeated suggestion, to Weinreich, Herzog & Labov's (1968) early formulation, and to Labov's systematic research -then the projection of synchrony onto diachrony may also reveal the latter's secrets. 1

C O D E -S W I T C H I N G I N W E D D I N G S O N G S A N D O T H E R V E R B A L G E N R E S I N 1 4 T H -1 5 T H C E N T U R Y C A T A L A N J E W I S H C O M M U N I T I E S : T H E D A T A
Twenty-five years ago, Riera 1974 edited a collection of poetic texts composed in Catalan Jewish communities in the 14-15th century. He did so with all the caution required by such aljamiat materials, 2 given the restricted evidence and the ambiguities of the correspondence between the Hebrew and Latin writing systems. These texts were not numerous, but they were sufficient to furnish a systematic picture of the phenomenon addressed here. They are instances of two clearly different verbal genres, which we may call "wedding songs" and "festive songs." The collection includes four wedding songs and one festive song ( piyyut naeh) in Hebrew aljamiat writing, that is, in Catalan Romance written in Hebrew script. This phenomenon was not uncommon in the Iberian Peninsula, 3 where Christians, Muslims, and Jews lived together for so long, maintaining significant cultural interrelations -albeit not free of conflict -until the Jews were banished from the Spanish kingdoms by the Catholic monarchs in 1492, the same year the kingdom of Granada, the last Muslim stronghold in Spain, fell into their power; the "Moriscos," the Muslims who had converted to Christianity and remained in Christian kingdoms, were expelled in 1609.
I reproduce below fragments of two of these songs, following Riera's reading, edition, interpretation and glosses, with slight modifications. 4 I offer a more or less free translation into English opposite the original text. The Catalan verses of the text appear in roman characters, the Hebrew in italics. The first is an example of the wedding song verbal genre: Shir nashir, Let us sing a song. nashir la-kalla shir^hadash Let us sing a new song for the bride, shir nashir.
Let us sing a song.
Dona, vejats el marit sovín; Woman, you should see your husband often; serviu-lo bé, millor (o quin) take care of him as well as you can, do your best. digats si és gran o poc o nin: Whether he be old or young or a child, tell him: "Dodi li ben shaday yalin." "My beloved is lodged between my breasts." Shir^hadash A new song.
Si altra dona demana per si Should he wish to marry another woman vos passau la mà (bell bací); pretty yourself, saludats: "Sènyer, ací, ací and greet him: "Sir, come here, Serviu-lo ben (a bon) pensir Take care of him the best you can, per tal que (als fets) ho puga servir for him to be able to serve you the best Digats-li si . l vets abnujós ni cosir: [at the same time If you see him annoyed or worried, tell him: "'Uri, 'uri, dabri shir." "Awake, awake, strike up the chant!" Shir^hadash. 5 A new song.
The second example is a fragment from a "festive song," following the same restrictions and conventions as in the previous case: Al te^hallel passà qui primer donà He infringed the precept: "Degrade not thy daughter" sa filla al zaqen, qui la{n féu zona.
who first gave her to an old man, who made a harlot of her. El zaqen se'n va a colgar al rosh ha-mita.
The old man goes to bed and falls asleep La ne'ara lo desperta amb gran gevura.
The young girl awakens him with great vigor. Lo zaqen li{n diu: "(Que n'és) tu sota The old man tells her: "How crazy you are! She'er we-kesut n'hauràs, mas no pas 'ona." You will have food and clothing, but not love." 6 La ne'ara li.n diu: " Perdut n'haveu el moa^h! The young girl tells him: "You have no brains! No n'haveu virtut ni punt de koa^h! You have not the slightest virtue or strength! Tot lo vostre feit és un bel rua^h! "All you keep doing is but utter futility!" Mas io prec en Déu que en breu ne sia almana." But I pray to God that I become a widow soon." My intention in the following pages is to concentrate on both a formal and a socio-cultural analysis of these two instances of secular verbal genres among late Medieval Catalan Jewish communities. These genres share, among other things, the fact that they are constructed by means of a systematic use of code-switching, even though the type of code-switching each manifests is different. One of the properties of code-switching, at least as it was initially observed and studied, was that it appeared to be an ingroup form of talking (i.e., it was produced in communication between members of the same ethnolinguistic group, rather than between members of different ethnolinguistic groups; see Gumperz 1982). This is also the case of the data considered here. Later studies did not necessarily obey this restriction. This has allowed a surprising increase in the amount of observed cases of code-switching; however, in my view, it has also contributed to the erasure of one of the features that rendered code-switching most intriguing as a communicative phenomenon.
Our previous basic definition of code-switching in the introduction was rather general and language-centered, and the data just presented, though they have reached us from written sources, were once intended as oral texts enunciated through singing and performed in public. Of course, an analysis of these texts is possible without reference to the social agents that created, reproduced, and transformed them -and, most important, made sense of them. But in order truly to understand these texts, we need to situate them in their socio-cultural and situational context. For the understanding of an utterance is a function of at least its truth-conditional value, its situational context, and the broader socio-cultural context in which it is grounded. Insofar as this situational context is built up not only by the participants' individual cognitive structures, including their presuppositional background, but also by their social interaction in specific settings in the real world, this understanding is always a socio-cultural construct. 7 Bearing in mind this socio-cultural and interactive construal of meaning in socially situated utterances, the first problem for the appropriate interpretation of code-switching in old texts is precisely the loss of the situational and sociocultural context in which the communicative events were anchored, or at least our partial knowledge of it. To be sure, part of this context may be recovered from the texts themselves, from independent historical sources, and, finally, from the projection of comparable contemporary phenomena, the imposition of synchrony onto linguistic and communicative diachrony. This goal -the recovery of meaning -resembles the philological pursuit: besides the correct reading and careful editing of an historical text, the philologist's ultimate aim is the interpretation of the meaning that this text entailed for those who produced it and those to whom it was addressed. However, the fresh interpretation that speakers can provide us in ethnographic research is not available when we are seeking the meaning of historical texts, nor can we avail ourselves of the fieldworker's direct observation of verbal behavior. 8 To begin this analysis, therefore, we need some knowledge of the way Catalan Jewish communities lived in medieval times, particularly toward the end of that period. Unfortunately, we cannot ascertain unequivocally which local community was the source of these texts. 9 However, information is available on the Jewish communities in the states of the Argonese crown -Aragon, Catalonia, the Balearic Islands, and Valencia -all differing from one another and differing as a whole from Jewish communities in other Christian kingdoms in Spain. 10 In medieval times, the Jewish communities of Catalonia moved to the towns at an early stage and increasingly adopted an urban lifestyle, engaging in urban socio-economic activities and occupations (moneylenders, traders, officials, translators, physicians, weavers, jewelers and other crafts). In the early days, the Jews were spread all over the city, but they soon tended to congregate in a certain quarter. This Jewish quarter was usually established near the walls or the castle, under the direct protection of the king or the feudal baron. The relationship between the lord and the Jews was based on an interchange of services: protection in return for financial aid, political counsel, and medical care. The Jewish quarter was not a ghetto, for its occupants were free to move throughout the town. It has been described as a small town within a town (Cantera 1998:150). The aljama, the institutional organization of the Jewish community, comprised religious, political, juridical, socio-economic, and charitable institutions; the urban space occupied by the aljama was termed call in Catalonia (Romano 1979). 11 The aljama had its own laws and regimentation (Hebrew taqqanot), and the Jews had to learn to live under a complex system of jurisdictions -Jewish and Gentile, civilian and religious. The institution of the aljama was created at the end of the 12th century or the beginning of the 13th. From an institutional point of view, the evolution of the aljama must be viewed from the perspective of a double, inner0outer dynamic. The tension between oligarchic and democratic forms of government within the aljama was latent through the ages and manifested itself in the form of religious movements that usually were more prone to castigate the wealthy nobles than to organize the commoners on the lower end of the scale; but the need to preserve the community from external threats gave it internal cohesion at critical moments. 12 The family was the basic social unit; the members of a family were strongly bound together much as the members of the whole community were. The main events of family life were of utmost importance and were regimented by detailed religious and liturgical principles, but also by traditional folklore and customs. The birth of a child -particularly the first-born -the betrothal and wedding of a bride and groom, or the death of a loved one were events associated with public and0or family rituals. The Hebrew family and society were patriarchal; the main roles assigned to women were that of wife and mother, and reproduction of the lineage; her daily tasks were looking after her husband and children and housekeeping. A woman's social status in medieval Jewish society was reflected by her lower legal standing: she was subordinate to the authority of her father, husband, or male family head and had no economic autonomy. Moreover, she was free from the positive religious precepts (mizwot) and did not count for the constitution of a minyan, the minimum number of ten individuals required for communal worship (only males over 13 years of age counted). As a consequence, women were unschooled both in literacy and in the Hebrew language. 13 Nonetheless, the mother was the object of affection and respect within the family and was portrayed as the mistress of the house (Cantera 1998:82-7). One of the verbal genres under analysis reflects both certain secular wedding rituals and the role that a married woman was expected to play. In a way, one can see this genre not just as a reflection of a prevailing social pattern but also as a type of discourse committed to reproduction of this pattern. 14 A thematic link between this wedding song and the festive song is the fact that both concern marriage. The injunction made in the latter not only casts light on the cliché of the ill-married woman (treated in more detail below), but also bears on the legitimacy of concubinage. The mystical-ascetic movement represented by the Kabbalists, along with ethical-social tendencies in the 13th century, induced a transformation in the rabbis' thinking on this point. An influential Catalan rabbi, Jonah Girondi, was the first to promulgate the prohibition of concubinage, in the following terms (the emphasis is mine): Degrade not thy daughter by making a harlot of her, lest the land fall into harlotry, and the land become full of lewdness (Lev. 19.29). The rabbis explain (Sanhedrin 76a) that therein one is warned not to permit his daughter to enter into sexual relations not sanctioned by wedlock . . . for concubinage, without formal wedlock, was permitted only to the king, whose authority effectively restrained others from having converse with his concubine, so that the king's relation to her was tantamount to marriage. Beyond this royal privilege the rabbis decreed that relations even with one's bride are forbidden until the marriage benedictions have been pronounced. Intercourse with a slave woman is a capital sin . . . for the sinner defiled the holiness of God by loving and possessing "the daughter of an alien god" (Malachi 2.11). His alien offspring will be a snare to him and a reminder of his sin. (quoted in Baer 1961:256).
A comparison of this text with the festive song exemplifies how traditional religious material could be recycled again and again for different purposes as the occasion demanded, whether for a festive song or a rabbinical dictum. As we assume a socio-cultural and contextual construal of meaning, this material might be interpreted in several ways, depending not only on its moral and ritual range but also on its efficacy as an authorized voice (see below) and its function as a means of legitimizing and "naturalizing" the discourse of power.
The reference appears once more in a similar context, a responsum 15 by another influential religious authority, Rabbi Solomon Ibn Adret of Barcelona, who reinforces Jonah Girondi's position (emphasis mine): First of all, I must say that this thing is prohibited and vile besides. The prohibition is clearly biblical, for such action violates the biblical injunctions, "There shall be no harlot among the daughters of Israel" (Deut. 23. 18), and "Degrade not thy daughter by making her a harlot" (Lev. 19. 29). It is also a vile practice, for this reason. Before the Revelation of the Torah a man might meet a woman in the street, and if they pleased each other he would have carnal relations with her, and then he would either keep her as his wife or else pay her the price of her favors and go his way. But the Torah has consecrated marriage and established the legal and ritual procedures for such a union. In many places there is even a communal statute which provides that a marriage ceremony can take place only in the presence of ten men, and that the bride's father and mother, or other relatives of hers, and the hazzan of the kahal must also be present. Certainly marriage by seduction is utterly forbidden, and it is all the more reprehensible when done by a man who already has a wife; for bigamous marriage, even when properly contracted, was banned by Rabbenu Gershom, of blessed memory, except that Maimonides was inclined to permit polygamy, as practiced by the Moslems. (quoted in Baer 1961:435) 16 Before proceeding to analysis of the poetic texts, I shall consider the cultural tradition from which these genres were borrowed and evolved, as well as their historic, linguistic, sociolinguistic, and cultural background. This analysis will take us toward issues such as (i) the different kinds of sequential adjacency of co-occurring languages in these Romance-Hebrew compositions vs. their Anda- lusian Arabic models -a juxtaposition of literary vs. vernacular language being implied, at least in the latter case; (ii) the contribution of code-switching to the construction of the poetic framework as regards its poetic, stylistic, rhetorical, and discursive functions; (iii) the typology of code-switching that can be drawn from these texts; (iv) the superposition of code-switching and intertextuality, as well as the superposition of code-switching and polyphony; (v) the juxtaposition of the profane and the sacred in one verbal genre, and the cultural ambiguity that derives from this fact in terms of deauthorizing0actualizing an authoritative voice; and (vi) the dynamic relationship between types of code-switching and the emergence of fused lects or mixed varieties. Finally, I discuss a relevant topic occasionally raised in connection with the type of language merging that appears in festive songs; (vii) whether this merging should be understood as an instance of an old Catalan Jewish language. After debating some of the misconceptions associated with this topic, I advance my own conceptual and methodological proposal for dealing with it. This proposal aims to provide a sound socially oriented basis for the phenomenon studied, and it takes as a starting point the view that we are dealing with verbal resources (rather than language systems) and attempts to uncover their allocation in the language economy of a particularly complex community.

A N A L Y S I S O F T H E D A T A : T R A D I T I O N A L L I N K S , I N N E R S T R U C T U R E A N D C O N T E X T U A L I Z I N G P R A C T I C E S
The discovery of muwashshahs and zajals, two innovative strophic forms in Arabic poetry, originating in Al-Andalus -the name given to the territory under Moorish rule in the Iberian Peninsula in medieval times -apparently without any connection with the pre-Islamic tradition, along with the mid-20th century decipherment of Mozarabic kharjas by S. M. Stern (1948Stern ( , 1949) uncovered a systematic use of code-switching in traditional poetic compositions. The first documented muwashshahs date from the beginning of the 11th century, although it has been suggested that they may have been invented in the early 10th century and orally transmitted since then. These compositions came to be imitated later on by the local Hebrew poets, and to be collected either in aljamiat -Hebrew written in the Arabic alphabet -or Hebrew writing. This Hebrew muwashshah is termed shir ezor or, ezor pizmon. 17 Scholarly debate concerning these Arabic compositions centered basically on (i) their formal and thematic definition, (ii) their origins, and, intertwined with these two, (iii) the issue of a possible relation with Medieval Romance literature, particularly that of the troubadours. The muwashshah -a girdle poem -is composed of a prelude of rhyming verses, for instance AA, followed by a series of strophes where each is made up of two elements: first, a sequence of monorhyme verses (e.g., bbb), and second, a refrain that resumes the initial rhyme. Thus, their form, spelled out according to a specific rhyme pattern and its conventional representation, could be represented as AA bbbaa. The zajal differs from the mu-washshah in that the refrain reproduces just half the rhyme pattern of the prelude: AA bbba. However, at least for the Hebrew counterpart of muwashshah, evidence from recorded sources suggests that the prelude was repeated after each strophe in song or collective recitation by the audience, and so the complete form of muwashshah would follow an AA bbbaa AA rhyme pattern. Otherwise, the muwashshah follows the model of classical Arabic poetry in terms of theme, language, and style, whereas the zajal, though also a courtly genre, is written in colloquial Arabic and deals with topics inappropriate for the muwashshah genre. Finally, a constant feature of the muwashshah, in contrast to the zajal, is the presence of a closure or concluding refrain in colloquial Arabic or Romancethat is, in one of the two local vernacular language varieties. This closure, in Arabic termed kharja or markaz, often represents the voice of a maiden lamenting the absence of her beloved.
As far as origins are concerned, once the possibility that these two strophic forms came from the Oriental tradition was excluded, the debate centered around two problems: determining whether there was a possible genetic relationship between the two, and exploring whether they arose under the influence of some autochthonous Romance form. In order to progress from conjecture to actual evidence, it was necessary to find concrete links, whether these were formal or thematic or both. The most objective datum in this respect was the existence of the Mozarabic kharjas. The style of the kharjas and the fact that they are more similar to the Galician-Portuguese cantigas d'amigo 18 than to anything else known in classical Arabic poetry suggests the existence of a similar oral traditional genre in Romance. This Romance genre may have had a pattern similar to that of the muwashshah and could thus have given birth to the latter. Up to this point, there was relative agreement among scholars. However, the third issue, the question of the influence of Hispano-Arabic poetry on Romance poetry, proved more controversial. The debate is between proponents of the theory of the Hispano-Arabic origin of Provençal and Galician-Portuguese troubadour poetry -and, by extension, most later European poetry -and those scholars who were always rather skeptical on this point. However, this is not the occasion to examine these claims in detail.
Nevertheless, the linguistic interest of these compositions is great, for they reflect the linguistic and sociolinguistic complexity of the society inhabiting Al-Andalus in the Middle Ages. They also reveal the consequences of the symbiosis of Semitic and Romance languages spoken by Mozarabic people and also adopted by the Muslims, whether of Arabic or Berber origin, and by the Jews. Islamic invaders, whether Arabic or Berber, were much less numerous than the native inhabitants of the peninsula. The latter spoke a Romance dialect known as Mozarabic, which they did not give up even when they became islamized and learned Arabic. The islamized natives (called muwallads in Arabic) and the newcomers merged to a certain extent into a single community of human bonds and social interests, and most of the invaders learned to speak the Romance language. This means that Al-Andalus Muslims were bilingual or multilingual (some spoke Berber as well). Five languages or language varieties were widely used. Colloquial Arabic was spoken by Muslims and Jews, either autochthonous or allochthonous, as well as by most Christians. The Romance vernacular was spoken by the Mozarabs, as well as by Muslims and Jews as a second language. Generally, the use of these vernaculars was independent of people's religious affiliation. In contrast, literary languages were distributed according to religious affiliation. Classical Arabic, also used as a courtly and formal language, was the only literary language of the Muslims. Eventually it would also be used by Christians and, in a rather more everyday way, by Jews, who wrote it in Hebrew script within their community. The liturgical and literary language of the Christians was Latin, although they could also use Classical Arabic. Finally, the Jews, who had both Arabic and Romance as vernaculars and used both the Arabic and the Hebrew writing systems, used only Hebrew as a religious and poetic language.
To borrow from Bakhtin 1981, and in accord with recent developments in linguistic anthropology, we can term this sociolinguistic situation one of polyglossia and heteroglossia. That is, the situation was characterized by both multilingualism (coexistence of several languages in the same social space) and by the coexistence of codes that maintained a dynamic relation with one another. These codes need not have coincided with any well-established language from the viewpoint of linguistics and0or local politics; rather, they were made up of various verbal resources of different linguistic origins. They were partially shared by the social groups and their members and unequally distributed among them, and they expressed their users' social position, cohesion, and cleavages. At the same time, the dynamic relationship among these codes prevents us from seeing the existing heteroglossia as merely the result of juxtaposition. On the contrary, the codes intersected and criss-crossed, and they were created, maintained, and given up by virtue of their interrelations in the activities of social groups and in interactions between members of different social groups, resulting in a peculiar kind of sociolinguistic stratification. 19 There were great differences between the social situation in 10th-11th century Al-Andalus and that of 14th-15th century Catalonia. These differences notwithstanding, there are certain undeniable similarities between the verbal art forms reviewed above and the Catalan Jews' compositions edited by Riera, although the latter belong to a later age and were by then expressed in the Romance vernacular -Catalan, in this case -in spite of their poetic character. With regard to the use of code-switching as a resource in verbal expression, particularly but not only in artistic verbal expression, two aspects should be emphasized. The first is the contrast between the older Hispano-Arabic compositions and these songs of the Catalan Jews, insofar as the sequential relationship between the languages involved is inverted in the latter: indeed, in them the bulk of the composition is in a Romance language, but the refrain and the verse closing the stanza are in a Semitic language, Hebrew. Second, there is a significant contrast between the wedding songs and the festive poetry with respect to the kind of code-switching that appears in them. That contrast between the two Judaic verbal genres also involves other elements -formal, thematic, and stylistic. 20 The wedding songs are true epithalamia. Their traditional genre is attested in Jewish communities throughout the Diaspora.According to historical records, their performance involved a concerted action by members of the community. Once the liturgical part of the marriage ceremony was over, the new spouses sat down on thrones opposite each other. Wedding songs were addressed to either the bride or the groom, to the accompaniment of a lute, flute, or brass horn, while young and old danced around. 21 The songs offered to the young couple were either taken from a traditional repertoire or else composed for the occasion, usually based on the melody of well-known local folk songs. The tenor of these wedding songs could be either grave and moralizing or mischievous, even lewd. The participants in the festive event were, of course, members of the local Jewish community, but they may be thought of as a community of practice (Eckert & McConnell-Ginet 1992, 1999Holmes & Meyerhoff 1999): "an aggregate of people who come together around mutual engagement in an endeavor" (Eckert & McConnell-Ginet 1992:464). 22 The participants in the wedding feast probably fulfilled the three criteria of a community of practice in Wenger's terms (cited in Holmes & Meyerhoff 1999:175): (i) mutual engagement; (ii) a joint negotiated enterprise; and (iii) a shared repertoire of negotiable resources accumulated over time. Wedding songs are part of these shared traditional resources. At the same time, these songs are constructed out of shared available verbal resources: here, from Catalan grammar and lexicon, Hebrew formulas (shir nashir, shir^hadash), biblical lines, strophic patterns of Hispano-Arabic origin, rhymes common to Catalan and Hebrew, code-switching, etc. In participating in the social event, people engaged in a common pursuit and activated shared values, beliefs, and ideologies. They made sense of themselves and others through their participation in the event and their contribution to it (Eckert & McConnell-Ginet 1999:186). Membership in the Jewish community might be externally constructed -i.e., assigned from the outside; however, membership in the community of practice was internally constructed. For participants, enacting the performance that created the community of practice was also a way of enhancing their membership in the Jewish community.
Let us now turn to formal analysis of our Catalan poetic texts. They are composed of seven, eleven, or fifteen monorhyme stanzas, specifically quartets, of octosyllabic masculine-ending lines, the last of which is in Hebrew and reproduces a biblical line, in a more or less successful echoing of the meaning and syntax of the three previous verses in Catalan. The prelude-refrain, also in Hebrew, is the same for all the compositions, except that the word referring to the groom (^hatan) is substituted for the word referring to the bride (kalla), as appropriate. It typically contributes to the arrangement of discourse by artfully framing it and breaking it into patterned pieces.
Typologically, the instances of code-switching in this text belong to one of the two types distinguished by Auer 1998b: alternation code-switching. Scholars have outlined various typologies of code-switching from either a formal or a functional point of view. I mention here the one advanced by Muysken 1994 and retained by Auer 1998b. In particular, Auer -who takes a dynamic and participantoriented approach to bilingual speech -distinguishes between the "alternation" type and the "insertional" type of code-switching. In the former, "a return after the switch into the previous language is not predictable"; in the latter, this is not the case, but "a content word . . . is inserted into a surrounding passage in the other language" (313-14). The first type usually takes place at a syntactic clause boundary and is used as a rhetorical0stylistic device; the latter usually affects small constituents (a word or a ready-made expression) without threatening the language of interaction. Catalan0Hebrew code-switching in wedding songs fulfills at least the formal features that Auer attributes to alternation code-switching, even though in this case -countering the first term of Auer's description -the switch back to Catalan from Hebrew is predictable from the poetic or discursive pattern of the song, once this pattern is grasped. 23 Moreover, the function of code-switching in this case seems fairly clear: formally, it accomplishes a demarcation of the poetic text, specifically of the strophic blocks (or stanzas) in which the discourse is structured and framed, as well as at least a double indexical function: while it points toward the text's poetic character, it also indicates its ethnic affiliation. Furthermore, this code-switching is a demonstration of intertextuality, polyphony, and dialogism with respect to the communicative domain within which it has been created and exists, and by virtue of its presence, this verbal text composed in a "modern" language becomes rooted in a very old tradition, so that it can be seen as a consequence of the revealed word of ancient times. Intertextuality is present, then, because the last line in each stanza refers to a verse from the scriptures; 24 polyphony, because of the sequential adjacency of distinct voices -the moralizing human voice addressing the bride, and the divine lover's revealed voice whose words are put in the bride's mouth; and dialogism, because these voices interact, the latter introducing the former while the former responds to the latter, even appropriates it, and, in so doing, transforms it and makes it sound a little ambiguous, half reproducing its power, half de-authorizing it. The authoritative voice 25 -as in the case of sacred language -demands the speakers' acknowledgment, their unconditional allegiance. It comes from a distant past; it is hierarchically higher than any contemporary voice; and its authority has already been acknowledged in the past. It is the word of forebears, and as such, it is a prior discourse. It is also a hieratic discourse, to be taken and preserved as an undivided whole that cannot be manipulated. Hence, it cannot be contextualized in another's discourse; that is, it cannot be smoothly framed by means of a gradual transition -it may be either cited or recited, but not used to construct one's own discourse. Nonetheless, we are con-fronted with the fact that segments of the authoritative traditional voice are actually used to construct the wedding song (the use of the term "used" as opposed to merely "mentioned" is intentional). By this very fact, these lines are tainted with cultural ambiguity. All this notwithstanding, the use of Hebrew in the context of Romance presupposes and indexes the local Jewish community and the ritual character of the communicative event (the wedding feast), while its drawing not only on the Hebrew language but also on the holy texts of the religious tradition confers legitimacy and efficacy on these songs, as they are anchored in the source from which all legitimacy and authority emanate.
In contrast, what we find in the one "festive song" of the collection edited by Riera is very different. Formally, this song could be deemed a zajal, since it follows the metrical pattern AA, AAAA, BBBA, CCCA, . . . , AA. The number of syllables per line varies from eight to twelve; the only constant element is the four word-stresses, which give birth to diverse rhythmic patterns that match the melody. The topic of the composition is the traditional one known in Catalan as la malmaridada, ("the ill-married woman"), or, to put it in emic terms, the infringement of the biblical injunction al te^hallel (Lev. 19, 29: "Degrade not thy daughter by making a harlot of her"). In this case, however, the code-switching used is very different from that of the wedding songs. Here the Catalan base language of the text appears merged with seven Hebrew expressions and 19 different Hebrew words. The code-switching is of the insertional type (Auer 1998b). By the very nature of the data, it may not be easy immediately to distinguish between codeswitching and lexical borrowing, although the numerous cases of rhyme-words must certainly be treated as instances of code-switching; these are almost all Hebrew and at the same time constitute most of the Hebrew material in the composition. 26 Again, the presence of Hebrew words in this poetically prominent position shows us both its demarcative -line-demarcative, in this case -and indexical function. Auer's predictability condition for insertional code-switching fits here, since a switch back to Catalan can generally be expected after a Hebrew rhyme-word. 27 This is also true of the other instances of code-switching or borrowing in the song. However, this does not seem to respond to rhetorical factors such as verse patterns or composition, but neither does it respond, except in some sporadic cases, to specific cultural concepts in the Jewish tradition. Together, these features raise another question which will be explored below: whether there existed a medieval Catalan Jewish language. 28 Riera (1974:9) considers this linguistic merging as providing neither "originality nor an increase in aesthetic value," even though he concedes that the author has been original in being "able to find enough Hebrew words to complete the rhymes." Now, it is a well-established fact that rhyme, so closely related to sound patterns in itself and having both clear demarcative and cohesive functions, also has a bearing on grammatical structure. In addition, it is not strange for rhyme to acquire a lexical import (compare the cases of derivative rhymes, equivocal rhymes, the sestina's rhyme-words, and others). Moreover, if the rhyme incorporates code-switching, then this represents an increased value -if not in aesthetic terms, always a matter of taste, at least in terms of descriptive poetics. Furthermore, in addition to the cited demarcative, cohesive, and indexical functions of rhyme-words, the presence of other borrowed material leads us to think that a differential code, independent of poetic genre, might have been emerging; in other words, these data would point to the existence of what Riera refers to as "a particular dialect of Catalan calls." This issue deserves separate comment, as it has led to varied interpretations and even downright confusion well beyond the field of Catalan medieval studies.
Before proceeding with this discussion, let us turn once more to our festive song. If the "Catalan Jewish dialect" really existed, then we would have to question whether this piece could be seen as containing code-switching at all, or whether the Hebrew elements in rhyme-position were accomplishing a demarcative function within a merged code. 29 Were that the case, we would be in a position to ask ourselves whether there is any point in maintaining the assumption that "codeswitching implies the recognition of previously coexisting codes and their interplay within particular speech events in speech communities," as stated above. This is undoubtedly a fair characterization in terms of grammar and lexicon from the analyst's point of view, but it need not be so from an interactional, participantoriented point of view. In dealing with conversational code-switching, it has been observed that codes may be achieved through a negotiation of communicative interaction; that is, codes may emerge as an outcome of on-line bilingual speech, and not just as a precondition for code-switching communication to take place (Auer 1998:15). This line of thought leads naturally to the recognition of "mixed codes," code-switching side by side with other types of merging phenomena. 30 Comparison of the two genres permits us to draw some conclusions regarding controversial issues in the field of codes and code-switching. The first one has been tentatively enunciated. In addition, as suggested in note 23, the predictability vs. unpredictability debate and the efforts to distinguish between code-switching and borrowing turn out to be crucially dependent, in texts of these types, on the poetic framing function that patterns of verbal art assign to switching. This should be viewed as a new instance of the subordination of linguistic facts to the "dominant" -in Jakobson's early term -in poetic texts. 31 Although grammatical constraints on code-switching are not a concern here, it should be noted, as a descriptive remark, that lexical borrowing and0or insertional code-switching in our text is a choice that selectively involves arguments, not predicates (nouns or nounlike categories). Furthermore, these code-switched elements combine with Catalan specifiers, modifiers, and prepositions. 32 Both these facts may be linked to the choice of matrix language. However, I shall not derive theoretical implications from them. Indeed, it would appear difficult to argue for a process of grammaticalization taking place, unlike in the case of Hebrew0English pragmatic markers observed by Maschler 1998 and commented on by Auer 1998a; rather it would seem more a process of relexification.
Riera states that "there are clues enough . . . to believe that these poetic compositions contain more than one feature of this dialect that until now has not been studied by anybody." This view is difficult to refute, 33 though it would appear to rely on a priori belief, albeit founded on rational clues. In any case, the issue is not at all a trivial one: Can we state empirically that Jews in the medieval Catalan calls spoke a specific language different from the Catalan spoken by their Christian contemporaries? This matter has been a source of controversy among specialists in medieval Catalan Jewry. 34

A N A L Y S I S O F D A T A A N D R E L E VA N C E T O A N A L L E G E D C A T A L A N J E W I S H L A N G U A G E
To begin with, recall that major scholars of Judaism (J. A. Fishman and C. Rabin, among others), have been at the forefront of study of the several vernaculars spoken in the Jewish Diaspora throughout history, and they have engaged in propounding the linguistic phenomena that they labelled as the Jewish languages -for example, the Yiddish and the Sephardic languages and their varieties (in the latter case, we may mention Ladino and Judezmo). 35 In approaching the study of such languages, attention must be paid to the roles played by such phenomena as mixing, borrowing, calque, and other forms of syncretism. 36 Mixing and calque, for instance, were determinant in the emergence of Ladino, which should not be confused with the Spanish Jewish vernacular, Judezmo. In fact, Ladino is a "calque language" born out of the process of wordfor-word translation of holy and liturgical texts, 37 in such a way that it may be characterized as Hebrew -eventually Aramaic -syntax implemented with a Romance lexicon, or, more figuratively, "Hebrew dressed up as Spanish" (Séphiha 1991:165). 38 Its origins can be traced back to the 13th century. Judezmo, by contrast, is a spoken language, a conservative form of Castilian that was maintained by the Sephardic Jewish communities in the Near East, the Balkans, and North Africa after their expulsion from Spain; we find it through the 16th century, and it appears to have been clearly individuated by the first quarter of the 17th century (Séphiha 1974:170). Since then, the interplay and reciprocal influence between these languages may have tended to "vernacularize" Ladino and "Ladinize" Judezmo. In addition, the hazardous and migratory history of Jewish communities, as well as the internal religious and cultural movements within Judaism, is often reflected in the evolution of these Jewish languages. 39 Study of these languages, then, could prove fertile terrain for study, in that the kind of linguistic phenomena being dealt with can always be acknowledged and controlled. For the Catalan area, side by side with Catalan texts containing sparse Hebrew loanwords -like those in the festive song dealt with herethere exist many texts by Catalan Jews in Hebrew replete with borrowings from Medieval Catalan, as shown by E. Feliu's translations and studies (e.g., 1988-89), among others. Calque translations also belong to this local tradition (see, e.g., Riera 1971-75).
In opposition to Séphiha's picture, other scholars give support to the view that both Ladino and Judezmo antedate expulsion. Wexler 1988 argures that "the existence of a colloquial judaized language is a prerequisite for a Jewish calque language", and also that "the obsolescence of the former entails the obsolescence of the judaized calque language" (1988:2). According to this view, a Jewish calque language (Ladino) rooted in the 13th century would imply a vernacular language of equal, if not greater, antiquity. Now, if Ladino actually came into being through scriptural practices, it had no previous independent oral existence.
Wexler's point, however, is not only advocating for a peninsular, prior to 1492, origin of a Judeo-Castilian language, as he terms it, but (i) asserting "the prior existence of a Judeo-Catalan language that derived from Judeo-Latin", (ii) positing "a Judeo-Arabic substratum in Judeo-Castilian", and (iii) dating "the switch from Judeo-Arabic to Judeo-Castilian from as early as the 11th century in some parts of the Peninsula" (Wexler 1988:10-11). A rather complex nonlinear evolution of Judeo-Ibero-Romance languages emerges from his picture, in which both genetic relationships from a Judeo-Latin language to Judeo-Romance languages and language-shift from some of these to Iberian Judeo-Arabic, and back again, are taken into account. Wexler discriminates six chronological stages in this development (1988:14-16).
Several misconceptions have spread in this field, but these affect the study of Catalan Judaism neither exclusively nor mainly. Some of them arose from calque translations or, even more absurdly, from aljamiat Romance texts being taken for instances of a supposed Jewish vernacular (for an example, see Magdalena 1993). Within the specific field of Sephardic studies, there is further confusion between the Sephardic language and a possible variety that may have been spoken by Castilian Jews, or even those of other peninsular kingdoms, before the expulsion. 40 The latter misunderstanding runs contrary to not only linguistic but also historical data, and it falls into anachronism when interpreting the territorial range to which the term Sepharad has been applied in several historical periods. 41 Even today, in a recent review of the Romance Jewish languages, Sala 1998 distinguishes between those that, like Sephardic, have a "well-defined" structure, and those whose structure "one has not been able to define," such as the hypothetical Catalan Jewish language and the Provençal Jewish language. In the same vein, Sala distinguishes between those languages that exist beyond the borders of their base language, such as Judezmo, and those that potentially could exist within these borders; the latter would be true of the Catalan and Provençal Jewish languages. Finally, he distinguishes between those that have survived and are still spoken, albeit steadily fading (Sala 1970), such as Judezmo, and those that are already extinct, if they ever existed, such as the rest of the Romance Jewish languages. Sala is clearly uneasy in dealing with this matter in general and with Catalan Jewish in particular, at least insofar he attempts to treat it as an autonomous, self-contained linguistic system (Sala's first distinction between "welldefined" and "ill-defined" languages speaks for itself ). This may be due to the lack of documentary data, to the nature of these data, or, as Riera suggests, to the lack of studies, but it also may be due to the nonexistence of a specific language in the sense usually understood by linguists.
At this point, I suggest a shift of perspective and propose to adopt a rather anthropological point of view, rather than centering the debate on the linguistic notion of a language -or dialectal variety -as an autonomous, self-contained, rule-governed system. The hypothesis is that in medieval society, as in any other, there were social agents who, in everyday life and for everyday purposes, interacted in accordance with certain social behavioral patterns. These included particular communicative practices which tended toward the elaboration of socially differentiated codes by means of which certain connotations of individuals' ethnic, religious, or social affiliation were conveyed. These agents were making use of the linguistic resources at the disposal of the community; for instance, it seems that the pejorative vocabulary borrowed heavily from the Hebrew lexicon (Sala 1998), 42 while certain "Christianisms" and Arabisms were avoided (Séphiha 1991). This usage was taking place in a specific sociocultural context, ranging over a wealth of communicative events, and also in a specific social context with its own order and conflicts, institutions and groups, established traditions and intruding dynamics. These socially situated settings engendered heteroglossia, as constituted by the dynamic centrifugal concurrence of interrelated social codes. One such code is to be thought of as a concrete sociolinguistic belief system that defines a distinct identity for itself within a surrounding heteroglot context. This code is not to be understood as a well-defined unitary language system -indeed, these systems, as far as they existed, were part of the heteroglot context -but, in Bakhtin's wording, it "is pregnant with possibilities of further dialectological individuation: it is a potential dialect, its embryo not fully formed." To summarize: "Language in its historical life, in its heteroglot development, is full of such potential dialects . . . ; some fail to develop, some die off, but others blossom into authentic languages" (Bakhtin 1981:356).
It is my contention that from this heteroglot picture there may have emerged, among others, a code that would match the ill-defined notion we are concerned with. This might also lead to a more realistic and socially well-founded approach to the hypothetical existence of "a dialect of Catalan calls." One way of thinking of this expression's denotatum takes it to be a definitively crystallized language entity, side by side with Catalan or Hebrew; a somewhat different view sees it as a dynamic co-occurrence of linguistic resources, where some form of "organic hybridization" takes place. 43 The transition we have seen above from the alternation type of code-switching to the insertional type may be more than a mere coincidence. After all, such a transition plays a progressive role in the hypoth-esized continuum from code-switching to language mixing, and then to fused lects, as argued by Auer 1998b in his dynamic approach to a typology of bilingual speech. Now, the analysis should proceed to examine whether and to what extent it would have been possible for a case of linguistic syncretism to appear 44 as a consequence of a particular symbolic usage of the available expressive resources, as well as a particular management (i.e., the adoption of certain ways of speaking and the rejection of others) of the linguistic variability that this syncretic continuum offered to the speakers, individually and as members of groups and subgroups within their community and the society at large, in order to adapt to changing communicative circumstances (to a range of well-defined events and settings, or to the dynamics of social change). If all this makes any sense, we could then move on to consider whether the use of some of these codes may have been part of a strategy of neutralization of socio-cultural differences or identity conflicts, rather than a mere construction of specific verbal genres. 45 It is obviously difficult to draw conclusions about this sort of spoken usage from written records, and even more so when we are dealing with communities in which scriptural practices had reached a level of considerable importance and autonomy. 46 Nevertheless, the research strategy just suggested would likely prove more profitable and effective than the cul-de-sac to which the expression "a dialect of the Catalan calls" leads, in its literal linguistic sense -the existence of an autonomous self-contained system (provided, of course, that we do not confuse it with the calque language appearing in texts that are translations, in a phenomenon parallel to Ladino, a confusion against which Riera warns us; see note 37). Furthermore, this new approach is applicable not only to verbal practices or the creation of oral codes, which may or may not include cases of language syncretism, but also to scriptural practices and the codes derived from them, which embrace such phenomena as aljamiat writing, calque translations, lexical borrowing from Romance in Hebrew texts and from Hebrew in Romance texts, 47 as well as code-switching with multiple nonexcluding aims -rhetorical, stylistic, discourse-building, and indexing both the historical communicative domain and the speakers'awareness of belonging to an ethnic, social, and religious community.
In strictly linguistic terms, then, we can but share the position expressed by Feliu when he states: Then, the Jews who inhabited the main towns in Catalonia in the High Middle Ages undoubtedly spoke the language of this country, and they continued to do so when later on they spread all over what is now modern Catalonia, and this was the language they spoke when they were banished from the country. This had always been their everyday language. In all likelihood, they also spoke other languages, just as people did both then and now: Arabic -those hailing from Muslim territories, and Aragonese, Castilian or Provençal, when circumstances and relations required. (1998-99:91; my translation from Catalan) 48 By implication, then, Catalan Jews did not speak, in their everyday life, a language separate from that spoken by their Christian counterparts. 49 Feliu's argument, however, does not end here. Instead, it follows naturally from his previous statement that Catalan Jews "never had Hebrew as their own spoken language, although in exceptional cases they might recur to it (to communicate with Jews from other countries -just as the Christians did with Latin -or to avoid being understood by Christians in certain well defined situations, etc.)." That is, Feliu's first statement does not imply that Hebrew was not used at all, either in oral or written form.
As for written language, Feliu states: Although this written form of the Hebrew language is usually called Medieval Hebrew, this is not at all a new established linguistic system, comparable in some way to Biblical or Mishnaic Hebrew: rather it is the result of the use that different authors in many countries and in areas of both Muslim and Christian culture made of a form of Hebrew at that time, conditioned as they were by the sociological, political and linguistic environment within which they lived. (1998-99:91-2).
A similar opinion is expounded by Rosén (1995:55ff ). 50 This is not to deny that "the Jews in Catalonia mastered this written Hebrew very early and there were authors writing in Hebrew since the middle of the 11th century," or that "sometimes, however, even when the mastery of Hebrew as a language of culture was complete, if they thought a Hebrew word would not be immediately understood by the reader, they resorted to the vehicular language (written in Hebrew script), and this vehicular language was always, as should be obvious, Catalan" (Feliu 1998-99:92). Whatever the reasons, this systematic recourse to the lexicon of the vernacular when writing in Hebrew is a well-established fact -as is recourse to Hebrew when using Catalan, as we have seen. This phenomenon is not rare in situations of language contact, but it is a clue to the permissiveness and permeability of the use of the available linguistic resources. In this respect, however, it should come as no surprise that different authors or groups in the same community had different attitudes and verbal practices, just as they had different ideological attitudes with respect to the understanding and identity of a common Jewish tradition. 51 Undoubtedly, Feliu clings to a traditional, linguistically oriented approach to this issue, or at least his answer implies a traditional linguistic framework, even if it includes clear references to the broader setting of historical, political, and societal processes. It is from this strictly linguistic stance that I am inclined to agree with him. However, I have argued in this essay for a rather different approach which takes as its basic tenet that research should center on the exploration of three problem areas: (i) The community's available verbal resources, rather than self-contained linguistic systems -these systems or their correlates forming part of these resources. 52 (ii) The verbal practices of speakers and social groups and communities in socially situated and well-defined speech events. (iii) The ways of speaking used by speakers in their interactions, leading to certain systems of verbal genres and codes that convey particular social meaning and express the speakers' reciprocal solidarity or conflict and their membership (inclusion or exclusion), and that give birth to a socially conditioned and heteroglot stratification of language. These codes should not exclude forms of linguistic syncretism.

C O N C L U S I O N
Code-switching appears intertwined with a range of interlinguistic phenomena in language contact, from strict interference to several kinds of language mixture. Code-switching has been studied primarily in everyday talk and in social interaction between members of a particular speech community, which implies a synchronic approach. My aim has been to approach a number of cases of codeswitching as they appear in the records of communicative practices left behind by members of past communities, living within well-defined cultural, political, and social conditions and displaying a rather complex linguistic repertoire in terms of both linguistic resources and verbal genres. My research has focused on the multilingual, multiethnic, religiously diverse Iberian Peninsula of medieval times, and particularly on Catalan Jewish communities in the 14th-15th centuries. Wedding songs and festive songs have been presented against a broader context of Hispano-Arabic cultural tradition from which their forms had been borrowed and re-elaborated, and also in the context of the linguistic merging of Semitic and Romance languages and speech varieties. This was followed by analysis of the poetic, discursive, and indexical function that code-switching -whether of the alternation or insertional type -plays in them. The contrast between the code-switching practices (alternation vs. insertional) and forms of lexical borrowing found in the two verbal genres led to consideration of another crucial issue for an ethnography of speaking of these communities: whether or not there existed a medieval Catalan Jewish language.
I am rather skeptical with regard to this linguistic entity for a number of reasons: the very small corpus of data available, their nature, and their potential confusion with non-vernacular linguistic, or rather scriptural, practices such as calque-translations, in these and other Jewish communities. I strongly suspect, however, that there did exist ethnically and socially marked codes in a complex heteroglot context in medieval times in Catalonia. My proposal, then, is to approach the issue by focusing on the linguistic resources available to the community, on the verbal practices developed by its members in definite socio-cultural and situational contexts, and on the resulting ways of speaking and the creation of dynamic and eventually syncretic codes that expressed social position, solidarity, and cleavages -inclusions and exclusions -of social agents in their everyday communicative interaction.
N O T E S *Albeit retrospectively, I first and foremost thank J. Riera and E. Feliu for their valuable effort to make medieval verbal art and other linguistic material of Catalan Jews accessible to the Catalan reader. I especially thank E. Feliu for his comments and suggestions on an earlier draft of this article which he was kind enough to read. I am also indebted to P. Auer, M. Heller, J. Hill, M. Nahir, and K. Woolard for discussion and suggestions on a more advanced version. It goes without saying that responsibility for its final form and content is my own, as is responsibility for its shortcomings. This work has benefited from financial aid from the Dirección General de Investigación Científica y Tecnológica, research project number PB96-1155. 1 Jakobson's seminal idea can be traced through his work from at least 1927 (e.g., his study on the historical phonology of Russian, his statement of the principles of diachronic phonology, and the Retrospective Overviews in his Selected Writings). See also Jakobson &Halle 1956, andWeinreich, Herzog &Labov 1968, which was a first attempt at the construction of an empirical foundation for linguistic change. See especially Labov 1974 for an appraisal of the research strategy applied here.
2 Aljamia denotes texts in Romance written in Arabic or Hebrew script, and also Arabic texts in Hebrew script. It could thus be said that the use of aljamia -or aljamiat writing -is a kind of cultural hybridization of linguistic0scriptural practices.
3 See below for reference to Arabic0Romance and Hebrew0Arabic forms of aljamiat writing. 4 The texts of three of these songs had been previously edited and analyzed by Lazar 1971, with a translation of the Hebrew passages and some philological comments. I keep to Riera's edition, aimed at Catalan readers, although I adapt the Hebrew passages to English conventions. My translation departs from his glosses at some points. 5 The last lines in each stanza correspond to the following lines of the following biblical books, respectively: Song of Sol. 1,13;Prov. 7,16;Judg. 5,12. 6 Food, clothing and sexual relations were the three areas in which a husband was obliged to provide for his wife. 7 For a critique of formal theories of meaning in North American linguistics and a defense of a socio-cultural and context-dependent conception of meaning, see Hanks 1996. 8 This serious handicap has not been an insurmountable obstacle in the writing of ethnographies of communication for groups of the past. Bauman's (1983) study of the early Quaker communities in 17th century England is a case in point. Of course, even direct observation and interpretation are never unproblematic. 9 Riera 1974 suggests as plausible the communities of Girona or Perpinyà in northern Catalonia. 10 A main source for the study of medieval Jewish communities in Spain is Baer 1961-66. Cantera 1998 is also useful for everyday life in these communities. 11 Aljama, from Arabic algamáa, means "the community," whose Christian equivalent at the time was universitas (later, ajuntament). Call comes from Latin callis, corresponding to Spanish calle. On the internal organization of the community and its conflicts, see Baer (1961:212-36).
12 See Baer 1961, passim. 13 A quotation from a 16th-century manual de inquisidores illustrates the point: "Women were not obliged to perform public worship, and they never gathered with men [for prayer], nor were one hundred women worth more than an unweaned infant for the purpose of accomplishing the number. Pious women had their room near the Synagogue and a rabbi prayed in Romance for them the prayers uttered in Hebrew in the Synagogue. There was no obligation of number among them." (quoted in Cantera 1998:84, my translation). 14 A thematic comparison of our wedding song addressed to the bride with a wedding song addressed to the bridegroom would illustrate this point. I will not pursue this line of inquiry here.

C O D E -S W I T C H I N G A N D D I A L O G I S M
Language in Society 30:3 (2001) 15 A "reply" to a question put forward by another rabbi, community or a member of the community on interpretation or application of a religious or legal norm. Responsa were a characteristic rabbinic genre and were collected in books of "questions and answers." 16 The hazzan, or "singer," was a servant in the synagogue. 17 A deep-rooted preeminence of the religious tradition in the Hebrew world prevented the rise of secular poetry among the Eastern communities. Conditions were more propitious, however, in the Jewish communities under Muslim rule in Al-Andalus, insofar as these were not so subject to traditional patterns and grew prosperous in an environment of religious tolerance and great cultural and ethnic diversity, as has been described (Baer 1961:18-21). The influence of refined Arabic literary and cultural models was of the utmost importance for innovation in verbal diction, prosodic and metrical patterns, rhyme, strophic poetry, etc., and an increase in the range of verbal genres and topics. An important social factor in the development of secular Hebrew poetry was the role played by Jewish noblemen. The relationship between the noble and the poet was one of patronage and protection against conservative religious forces. This institution began to decline in the mid-12th century in the Muslim kingdoms, but the model was then transferred to Christian territories. 18 Cantigas is the traditional name for old Galician-Portuguese poetic compositions that were intended to be sung. These compositions might be lyrical or narrative in character. Usually three main genres are distinguished: the so-called cantigas d'amor ("love songs"), cantigas d'amigo ("malefriend songs"), and cantigas d'escárnio e maldizer ("mocking and slanderous songs"). Both the cantigas d'amor and cantigas d'amigo are love songs; they differ, however, because the former express a man's love for his beloved, while the latter express a woman's loneliness because of the absence of her beloved and her longing for him. The relevant point is that in the cantigas d'amor, it is a man who speaks, while in the cantigas d'amigo, it is a woman. These three types do not exhaust the catalogue of genres in Galician-Portuguese medieval poetry. 19 I use "stratification" in a rather lax sense, in line with Bakhtin, as opposed to the technical meaning assigned to it in Labovian sociolinguistics. As far as the use of the notion "heteroglossia" is concerned, my source is its current use in linguistic anthropology. It should be pointed out, however, that in line with Bakhtin's goals, a distinction should be borne in mind between heteroglossia as a phenomenon of everyday languages and language usage in historical speech communities, and its intentional representation in a novel for artistic purposes, which is the primary aim of Bakhtin's reflections. 20 In keeping with the aim expressed in the preceding note, in Bakhtin's view heteroglossia and its components apply more to prose writing than to poetry. He states clearly that "heteroglossia or even a foreign language is completely shut out of a poetic work," although he concedes that "a certain latitude for heteroglossia exists only in the 'low' poetic genres, in the satiric and comic genres and others" (1981:286-87). Indeed, both verbal genres in our Jewish songs may be considered as belonging to this category. I think, however, that Bakhtin mainly has in mind written authorial poetry rather than traditional poetic verbal art, with its diversity of oral genres, or else his ideas should be qualified from the point of view of our current knowledge of what traditional oral poetry is and how it comes to be "crystallized" (Nagy's term, 1996:108-9). Insofar as these compositions -not only epic songsmight well be performed while composing and composed while performing (Nagy 1990(Nagy , 1996, they necessarily implied a dialogical relationship of potentially heteroglot voices and consciousness through time and space. 21 Dancing in public was often forbidden by the rabbis within the communities. Wedding and bridal feasts, however, were particularly appropriate occasions for group dance. 22 The community of practice differs from the traditional community primarily "because it is defined simultaneously by its membership and by the practice in which this membership engages" (Eckert & McConnell-Ginet 1992:464). 23 This remark, however, is not intended to invalidate Auer's description. Auer's aim in the first place is to account for conversational code-switching, not the kind of textual phenomena we are concerned with. In fact, one of the oldest cases of alternating code-switching I am aware of is 6thcentury Greek 0Latin antiphonal singing. It is said that, in order to prevent the congregation gossiping during the liturgy, Caesarius of Arles made them participate in it by learning and singing hymns, alternating a versicle in Greek with one in Latin. This was in all likelihood alternation code-switching, even though the switching was absolutely predictable. Obviously, this is not always the case in poetic texts; rather, it depends on code-switching's acquiring an inherent discourse-patterning quality. The analysis of the "festive song" will illustrate the assertion. 24 In a rather broad interpretation of the term, it is not only the code-switched Hebrew passages that show specific intertextuality, but the whole composition could be deemed intertextual in a generic sense, insofar as its form evolved from Hispano-Arabic muwashshah. 25 Cf. Bakhtin (1981: 342-43). 26 Out of 30 lines, only four employ a Romance rhyme-word, one of these being the first one, reproduced here. For an alternative view, however, see below.
27 See note 23. Additionally, the switch back is not always completed, and the rhyme-word is not always code-switched either (see also note 26). 28 At least, they have been taken as posing such an issue by Riera 1974, and will provide us with an opportunity for discussion. 29 This point was suggested by Peter Auer (personal communication). It implies -or induces -a particular interpretation of the issue in hand. 30 Illustrative are the arguments by Álvarez-Caccamo 1998, as is the evidence he provides on Spanish0Galician, and that offered by Moyer 1998 on Yanito, or a Spanish0English code-switching pattern in Gibraltar. 31 "The 'dominant' may be defined as the focusing component of a work of art: it rules, determines, and transforms the remaining components. It is the dominant which guarantees the integrity of the structure" (Jakobson 1981:751). 32 DiSciullo, Muysken & Singh 1986, seeking to account for formal constraints on code-switching in terms of dependency rather than linearity, in a GB framework, provided for the switch between a specifier and its head by means of the neutralizing character of the former. Similar arguments appear in Nortier 1995 and are criticized by Alvarez-Cáccamo 1998:35. 33 The forcefulness of the final part, in which the author regrets (or at least acknowledges) the lack of study prevailing on this issue, contrasts with the prudent wisdom of the initial part, in which the very existence of the dialect is not at all taken for granted. 34 I remember causing a heated controversy by daring even to discuss the issue. Nonetheless, the passion with which a prominent member of the Barcelona Jewish community and a scholar of Catalan Jewish heritage argued for the existence of a Catalan Jewish language was not proportionate to the evidence at hand. I can only assume that their attitude was more a function of current Catalan Jewish feeling and self-perception than an effort to reconstruct the historical past. 35 See some arguments for the need to distinguish these modalities in Séphiha 1974, reiterated in Séphiha 1991 These same phenomena have also played a significant role in the elaboration of Modern Hebrew. Cf., e.g., Fishman 1987. 37 See, e.g., Séphiha 1974Séphiha , 1991, and the references to his own bibliography there. As Riera advises in relation to the Catalan area (1971-75:54-55), it is also important to distinguish between aljamiat texts and translations from Hebrew, which are often literal and word-for-word. The language of these translations is not the reflection of a spoken dialect but a written and often recited variety which takes on the appearance of a mixed language owing to its origin. 38 Unlike "mixed codes," a "calque language" necessarily involves both two previously existing languages and the emergence of a new code. 39 Besides Fishman's and Rabin's work, among many others, see, as a contribution from the Catalan area, Berthelot 1993. 40 See note 41. For a clarification of the issue, see, among others, Riaño 1993. 41 See a clarification of the matter in Feliu 1998Feliu -1999, who cites rich documentation. 42 As a case in point, see our "festive song." 43 As Bakhtin would have it, referring to defined types of (unconscious and impersonal) utterance language mixing in historical everyday languages, as distinct from intentional and individualized artistic utterance language mixing "with a represented language and a representing language," aiming at creating an "image of a language" (a key concept in his conceptual framework). (Cf. Bakhtin 1981:358-62). 44 To put it in the terms of J. & K. Hill, who borrow the term from Kuryłowicz (1964:40), who, working only within the framework of internal linguistic analysis, understood the term as "the suppression of a relevant opposition in certain conditions." In adopting and reinterpreting Kuryłowicz's term, Hill & Hill try to avoid the ambiguity, problems, and connotations that the use of the term "mixed language" has traditionally provoked. As anthropologists and linguists studying a determined language contact setting, they apply the notion to a dynamic situation characterized by a selective use of the ver-bal resources in the dual continuum of the concurrent languages (Nahuatl and Spanish, in their case) and by the processes of relexification, syntactic convergence0divergence, hybridization0differentiation, semantic or syntactic calque, and others that give birth to particular ways of speaking (at the same time that they exclude other ways of speaking) within a given society and a socio-cultural setting determined by local relations as well as by the global context into which this society is inserted. Without falling into mechanical retrospective projections, I suggest that Hill & Hill's conclusions may serve to illuminate the situation we are concerned with here, bearing in mind the caution that the case demands. 45 However, the fact of centering sociolinguistic research on the topic of conflict has all too often induced scholars to think of neutralization as the suppression of one of the relevant terms of an opposition, as expounded by linguists (see Kuryłowicz's definition in the preceding note), rather than the coexistence of both or suppression of the relevance of the opposition itself. This dichotomous either0or thinking at the level of structure -not just reminiscent of but identified with Saussurean paradigmatic relations -may occasionally be reinterpreted as a both0and feature at the level of discourse and social interaction practices. What is implied in this reflection is a case of the exploitation of simultaneity in the prototypically sequential nature of speech as it appears to our current understanding. For the deployment of similar ideas, see Woolard 1999, who emphasizes the influence of Bakhtin's writings on anthropological thinking in this respect. 46 As far as our texts are concerned, an added difficulty should be noted in this respect: the fact that they are poetic texts means their contribution to our knowledge of everyday social life heteroglossia is rather limited and indirect. This holds true irrespective of Bakhtin's views on poetic genres (cf. note 12). 47 All references are to texts in Hebrew script, whether aljamiat (when the language is Catalan) or not (when the language is Hebrew). The songs considered above pertain to the former and the borrowed words are Hebrew, as we have seen. A different phenomenon is represented by the Hebrew words in Roman script found in Christian documents (e.g., notarial, royal, ecclesiastic, etc.) concerning such issues as purchase and sale contracts, inventories, legacies, legal contentions, and referring to Jewish items and concepts (the months of the Jewish religious calendar, liturgical feasts, names of the biblical books, concepts from Talmudic law, names of the institutions of the aljamas or Jewish quarters, with their assemblies and synagogues, and so forth). 48 "Hence, Catalan Jews in the 10th century might have been bilingual in Arabic and Romance, with Catalan as their primary language." (Wexler 1988:6). 49 It may be relevant here to remember what has been said about women's illiteracy and lack of familiarity with the Hebrew language; see note 13. 50 "Medieval Hebrew is not one language." (Rosén 1995:56). But he immediately adds: "We would not go so far as the students of Medieval Latin go and say that there was no Medieval Hebrew but 'several Medieval Hebrews'." (Rosén 1995:56). This notwithstanding, Rosén understands "Medieval Hebrew" as a Traditionssprache and traces a parallelism between Medieval Hebrew and Medieval Latin, in that both were languages of a sacred tradition, "which those who were concerned made an effort to transfer to secular use"; both were languages of a written tradition "which those who were dedicated to their maintenance desired to create afresh in order to rend them suitable for oral usage"; both were languages "with reference to which an endeavour at modernization was made in order to assure them a relatively wide diffusion in living use" (Rosén 1995:59). However, his proposed program of research aims at accounting for a pluralism of shapes of Medieval Hebrew -not only its Arabicized or its Ashkenazi forms, "but also for a Germanized, Francized, Italianized Hebrew language, to mention only the more important branches of Jewish literature." (Rosén 1995:58). Obviously, these forms of Hebrew are not to be confused with Jewish languages. 51 See, for instance, Feliu 1991Feliu , 1998 apropos of the grammarian Profiat Duran (second half of the 14th century) and his description of the doctrinal groups of his time: the Talmudists, the rationalists -those who philosophized following the Greeks -and the Kabbalists, among others who did not share the same final aim. 52 Of course, I am not denying the validity of defining linguistic systems as a way of improving our knowledge of linguistic reality, provided we bear in mind that these are but theoretical constructs or abstractions partially to encompass this reality. It should come as no surprise that code-switching cannot be accommodated within available well-defined linguistic systems as linguists have understood them. Likewise, neither should we be surprised that code-switching and other forms of linguistic mixture and syncretism are part of both language reality and social reality, if these may be differentiated for analytical purposes. Our unitary systematic concept of language is, more often than