Many of the Latin manuscripts thought to come from al-Andalus were surveyed by Manuel Díaz y Díaz, who carefully cataloged them in an important study.18 The liturgical examples among them are largely fragmentary and contain only isolated liturgical texts or materials related to liturgical practice. One example is a book list (ESCR.II.18) that includes antiphoners, prayer books and a martyrology. Díaz y Díaz argued that the list came from Córdoba and reflects the types of liturgical books that would have been available in religious institutions, although he thinks they ended up in a private library.19 One of the books in this list has since been identified as the Azagra codex (BN29), a tenth-century compilation of poetry partly copied in al-Andalus.20 It features a number of hymns, including one for John the Baptist by Cyprian, a ninth-century archpriest of Córdoba.21 Another more recently-discovered fragment (Coimbra) from an eleventh-century liber misticus has been considered a product of Toledo before the conquest of the city by Alfonso VI in 1085, because of the type of musical notation it features (see below).22 The marginal markings in Bibles identified as ‘southern’, such as BU31, indicate their division for liturgical readings, although palaeographical study would be necessary to establish when this marginalia was added.23 In general, more thorough criteria for determining southern provenance that goes beyond palaeography-of both text and music-is necessary to more firmly place these manuscripts in al-Andalus, since many of the features long defined as ‘southern’ have been identified in northern sources.24

Another attempt to assess the Andalusi liturgical tradition has been done through analysis of northern manuscripts thought to have been copied from southern models, but this approach must also be revisited. An important example is L8, the mid-tenth-century richly notated and illuminated antiphoner (chant book) from the León region, and one of the most complete sources for the Old Hispanic musical tradition.25 Certain features in this manuscript have suggested that it had a southern model, including the obits of southern bishops in the calendar and liturgical material that would only have been relevant in Toledo, e.g. a procession to the tomb of Leocadia, located in Toledo.26 But although isolated rituals may have been reproduced in L8, this does not mean the entire manuscript was copied from a southern model.27 The tenth-century liber misticus BL45 from San Millan de la Cogolla is likewise thought by some to have been based on a southern model, due to the presence of liturgies for saints from Andalusi centres (e.g. Pelagius and Zoylus, both saints linked to Córdoba).28 Since these saints only receive proper (unique) masses and hymns in BL45 and not proper offices, the idea of a southern model can only go so far. We might expect that such important saints would have been celebrated with proper offices in Córdoba, leading us to wonder if the isolated and partial materials in BL45 could have been copied from a Cordoban model.29 A final example, BN56, is a one-folio fragment of an antiphoner from San Zoilo de Carrión in Plasencia, which some think would have been modeled on an exemplar from San Zoilo in Córdoba, but based only on the fact that the monks of Plasencia are thought to have come from Córdoba.30 As this last example shows, such southern attributions are influenced by the idea—possibly legendary, in some cases—that many monasteries in the northern kingdoms welcomed monks from the south throughout the tenth and eleventh centuries.31 We must be wary, however, of extrapolating too much on the supposed ‘southern elements’ of northern manuscripts.

A series of manuscripts associated with the former Visigothic capital of Toledo has been treated as an important source for understanding the liturgical rite practiced in al-Andalus, but once again, in ways that have recently undergone reassessment. The Toledan manuscripts have been linked to that city in part because they were kept there (largely in the Cathedral Library - see Table 1), and in part because they share certain palaeographical features and mostly use ‘horizontal notation’, a type of musical script different from the so-called ‘vertical notation’ found in manuscripts from León and Navarre (Figs. 1 and 2).32 Originally thought to have been produced between the ninth and eleventh centuries, the Toledan manuscripts were re-dated by Anscari Mundó to the eleventh through fourteenth centuries. 33 This challenged the former understanding that the manuscripts were witnesses to the rite as practiced in the city under Islamic rule. They are now thought to reflect practice at several parish churches in Toledo, the make-up of whose congregations is unclear. Originally thought to have been populated by native Toledan Christians who were permitted to continue practicing their rite when the rest of the kingdom changed to the Franco-Roman tradition, the six ‘Mozarabic parishes’ are now claimed to have been founded in the mid-twelfth century by Christian refugees fleeing persecution from the Almoravids and Almohads, who established their domain over the south of the peninsula. There is no evidence of the existence of these parishes prior to 1156.34 If this is true, the Toledan manuscripts associated with the Mozarabic parishes could be seen as transmitting the form of rite that was practiced in the southern-most part of the Iberian Peninsula, and which may have continued to be practiced in those areas as long as they remained under Muslim rule.

There are reasons to wonder if the Toledo manuscripts reflect the rite brought by Andalusis who settled in Toledo, however, and if so, in what ways. These manuscripts bear witness to two different versions of the Old Hispanic rite: Traditions A and B. The basis for the distinction is found in differences in notation (musical script), chant melodies, the structure of the liturgical day, the name given to particular genres of chant and prayer, and the choice of texts for any given feast day.35 Both traditions are present in manuscripts associated with Toledo, which has led to debate about what each tradition represents. For some, Tradition A reflects the practice of Christians in the northern kingdoms of the Peninsula and also Toledo itself, which continued at parishes such as Santa Eulalia, where several of these manuscripts were kept. The few Tradition B manuscripts that survive, and which are linked to the parish church of Santas Justa and Rufina, would be a parallel tradition that was brought to Toledo after the twelfth century by Andalusi refugees, particularly from Bética.36 Other scholars see Tradition B as a later evolution of Tradition A, since the Toledan Tradition A manuscripts date largely to the eleventh to thirteenth centuries whereas the Tradition B manuscripts are from the thirteenth and fourteenth centuries.37 Challenging this last theory, the discovery of a different notational system in Tradition B manuscripts suggests that it did not just ‘evolve’ out of Tradition A.38 Whether Tradition B is a southern Andalusi tradition and Tradition A is from Toledo is difficult to determine due to the scarcity and fragmentary nature of the manuscript sources.39 But if so, it would suggest that there were different variations of the Old Hispanic rite celebrated within al-Andalus itself. Table 2 includes a summary of the possible explanations for the differences we see between this small cluster of manuscripts. Further close palaeographical analysis of both text and notation will surely allow a better understanding of the relationship between the manuscripts, their liturgical contents and whose practice they might represent.40

One potential key to unraveling the Tradition A vs. B puzzle is CEA, an Arabic translation of the Gospels followed by a list on ff. 92r-95v that designates which part of the Gospels was read on each of a series of feast days (capitulare evangeliorum). The importance of this manuscript lies in the fact that it is the only surviving liturgical book that has been clearly linked to al-Andalus and it is in Arabic, an aspect to which we return below. The Gospel reading was one of three Biblical readings in the Old Hispanic mass, the other two being from the Old Testament and from the Pauline Epistles.41 CEA provides for these mass readings by listing, in the order of the liturgical calendar, the chapters that should be used, starting with the Sundays in Advent through to the Quotidian Sundays, and including a number of saints’ feasts (on which see more below in the section on calendars).42 For example, for the feast of Saint Eulalia, the list states that the reading should be taken from Matthew, chapter 25.43 The manuscript has been dated to the fourteenth century and is from North Africa, but a colophon indicates it was copied from a model made in 1145.44 Philippe Roisse thinks the twelfth-century model was moved to the Maghreb in the wake of the Almohad conquest of al-Andalus. He also believes that the twelfth-century text was based on an even earlier model, possibly from the tenth century, something that has since been supported by Juan Pedro Monferrer-Sala.45 This first model is associated with Seville because there are unusually two feast days dedicated to the saints Justa and Rufina, on 17 and 18 July respectively; all other Iberian calendars place the sister saints together on 17 July, suggesting CEA was from their patron city, Seville.46 One of our only liturgical witnesses is therefore attributable to an Andalusi Christian context, but only with some reservations given its late date, layers of composition and copying, and its distance from the place of production of its alleged model-not to mention that it contains only a tiny element of what would have been a much larger order of service.

Since CEA is attributed to an Andalusi centre-Seville-we might imagine that if there were such a thing as a distinct Andalusi tradition (i.e. Tradition B), CEA would reflect it. And indeed, Roisse, in his evaluation of the manuscript, noted the close correspondence in the choice of Gospel readings in CEA with those found in the Tradition B manuscripts.47 This reading is nevertheless compromised by Roisse’s choice to use later attempts to reconstruct the Old Hispanic rite (the Neomozarabic Breviarium Gothicum and Missale Mixtum) in his comparison, and the fact that the surviving manuscripts only contain a limited part of the liturgical year; this makes comparison difficult.48 The picture we do gain does suggest some overlap in the readings preserved in Tradition B and CEA, especially in Lent which, in some cases, differs considerably from the Tradition A manuscripts.49 This is not consistent, however, and sometimes CEA matches neither Tradition A nor Tradition B.50 We should not expect complete liturgical uniformity in manuscripts from different centres, but it is difficult on the basis of this comparison to draw a clear conclusion that the Toledan manuscripts associated with Tradition B and CEA together reflect one tradition that is fundamentally different to Tradition A. This is not to say that there was no distinction, but the fragmentary nature of the Tradition B manuscripts and the fact that CEA provides us with only one element of the liturgical repertoire (the Gospel reading) limits our ability to evaluate the problem of the two traditions on the basis of this evidence.

The so-called Cordoban Calendar is the only other extant source, besides the CEA and the above-mentioned fragmentary martyrology, that allows a possible glimpse into the sanctorale celebrated by Christians in al-Andalus, but the source raises many questions. The analysis that follows will address the textual variation within the different manuscript witnesses, the text layers, the comparison with other calendars and liturgical texts, the orientation of Andalusi Christians towards other communities and the terminology used in the text, all of which are leading us to rethink important assumptions made about this well-known text.

The Cordoban Calendar is an Arabic almanac supposedly written around 961 or 973 at the court of the Umayyad Caliphate.88 Early scholars, such as Reinhart Dozy, Francisco Javier Simonet, Évarist Lévi-Provencal and Charles Pellat emphasized the value of this source for understanding Cordoban society in the tenth century.89 Specifically Simonet and Marius Férotin considered the liturgical part in the surviving manuscripts valuable as a reflection of the sanctorale of Andalusi Christians. They ascribed the liturgical calendar to a Christian author, the bishop Rabi ibn Zeid, whom they identified as Recemundo, a bishop of Elvira.90 The sections that include knowledge about the astro-meteorological system of dividing the year (anwāʾ) used by the ancient Arab Bedouins, they claim would have been written by ʿArīb ibn Saʿīd.91 There have been significant questions raised around these assumptions, however, with Ann Christys convincingly showing that there is no solid evidence to identify Rabi ibn Zeid with Recemundo.92

Until now, it has often been assumed that two manuscripts constitute what became known as the Cordoban Calendar. These are the two manuscripts used in the edition produced by Dozy: a Judeo-Arabic version (CCA) and a copy of a Latin translation, probably made in the twelfth century by Gerard of Cremona (CCL).93 Unfortunately, references are sometimes made to the Cordoban Calendar when actually what is meant is the CCL.94 There is, however, now generally more awareness of the importance of treating the two manuscripts separately.95 Not only do these two manuscripts (CCA and CCL) differ significantly, they are also not the only manuscripts we have. Seventeen other manuscripts—of which seven we have newly discovered and present here for the first time—have been identified as versions of the Cordoban Calendar due to a clear overlap in phrasing and structure of the calendar, as well as references to Córdoba and al-Andalus.96 Despite their textual connection, these manuscripts all differ with regard to the liturgical feasts and commemorations they include, with some manuscripts containing very meager evidence of a Christian sanctorale or no liturgical content at all.97 The variety in content of the different text versions and our lack of understanding about text transmission make it difficult to establish what the original would have been. We should see the manuscripts as witnesses to a tradition or a ‘living text’, rather than as one source of which we will be able to recover the urtext. Without a clear urtext, the Cordoban Calendar’s usefulness for determining tenth-century Andalusi Christian practice is open to question.

It is clear that some of the calendar entries in certain manuscripts are the result of interpolation at a later stage within the text tradition. Martínez Gázquez has established that some of the Latin copies (Vic167, BN6036) were influenced by the Catalonian tradition (which followed the Franco-Roman rite), whose elements were interpolated after the abolishment of the Old Hispanic rite.98 This becomes especially clear through the inclusion of saints who died in the twelfth and thirteenth centuries.99 Problems arise too when taking at face value the topographical information contained in CCL that relates to religious practice in Córdoba.100 In this version, we find abundant information as to where saints were buried, where feast days were celebrated in the city, even mentioning the exact churches and where they were located.101 For example, the text reports that the tomb of Pelagius, a tenth-century martyr, is in the Church of Tarsīl. It is difficult, however, to imagine that the veneration of Pelagius was possible in Córdoba during the reign of caliph al-Hakam II or that an entry about Pelagius was recorded in a tenth-century calendar composed for al-Hakam II, since Pelagius supposedly died at the hands of al-Hakam II’s father, Abd al-Rahman III. This suggests that like the additions made to other versions in the examples mentioned above, Pelagius was added to CCL or its model and was not in the tenth-century ‘original’.

CCL is often cited as evidence for the veneration of individuals from the ninth-century ‘martyr movement’, but can it actually be read as such?102 Caution is needed here as well. The individuals involved (Paul Alvarus, Esperaindeo, and the neomartyrs Perfectus, Adulphus, John and possibly Emila) are reported to have their tombs in the city of Córdoba and we might read their inclusion in CCL as evidence of them being celebrated in the liturgy.103 First, we do not know if the burial of a saint necessarily meant veneration. Second, not only is there some discrepancy between CCL and Eulogius’ martyr accounts, the CCL remains the only source from al-Andalus that attests to the celebration of these individuals.104 CCL could therefore have been supplemented later with this information to serve the needs of the Mozarabic parishes in Toledo, where the text was most probably translated by Gerard of Cremona. Those communities might have wanted to reconstruct what they believed was the religious practice of Andalusi Christians in Córdoba-but was perhaps not actually celebrated there.105 Taking into account possible later additions to CCL should make us more wary of approaching CCL as a straightforward reflection of Christian practice in al-Andalus. Whereas some entries in CCL might prove insightful for the sanctorale and whereabouts of Christian practice in Córdoba, we should be careful of others.106

Despite the caution required, some of the topographical entries in CCL can be viewed with less skepticism because they are also supported by the Arabic manuscript versions. Examples of this are entries found in CCA: “Christopher [who] is honored in the neighborhood of Munyat ʿAǧab, in the other part of Córdoba, beyond the river where the lepers are,”107 and “the feast of Eulalia, killed in Barcelona, who is called a martyr. There is a monastery dedicated to her in the plains of Córdoba where they [the Christians] gather.”108 Of all manuscripts, CCL has the most extensive descriptions of Córdoba’s topography, but many Arabic manuscripts contain descriptions of places of worship where the patron saints of Córdoba were celebrated: Acisclus, Zoilus and the ‘three crowns’ (Faustus, Januarius and Martial).109 In certain cases the (newly found) Arabic manuscripts prove valuable, as they contain more topographical details than CCL. In Risala, together with a couple of newly discovered manuscripts, we find more details regarding places of worship than CCL provides, for example: “the feast of Julian, killed in Antioch, who is called a martyr. They gather in the monastery balanshish, known as the white monastery in the mountains of Córdoba.”110

One of the ways to deal with the diversity in the manuscripts of the Cordoban Calendar is to compare their sanctorales to calendars from the northern kingdoms. In a quantitative study, we were able to demonstrate certain patterns and trends that allow us to draw some general conclusions as to the possible date of production of the texts. In a comparison with Old-Hispanic calendars from the North, we see that liturgical calendars from the eleventh century contain more saints’ feasts on average than liturgical calendars from the tenth century, which is in line with a trend generally observed that calendars grow over the centuries.111 The liturgical content of manuscripts of CCA (89 feasts112) and P2521 (80) corresponds more closely to tenth-century Old-Hispanic calendars such as SM (94) and Albelda (94), than CCL (124) does. Vic167 (202), PA45 (151) and ÖNB2462 (250 +) would correspond to even later centuries.113 Such comparative studies can thus help in revealing general patterns that might be used to justify a tenth-century date for the contents of the Cordoban Calendar as found in CCA and P2521, whereas CCL fits better into the average of eleventh-century calendars.114

Comparison of the manuscripts of the Cordoban Calendar to CEA, the only other existing witness to the sanctorale in al-Andalus, is less fruitful however. CEA contains feasts that were important enough to have proper masses (and mass readings). There were other saints in a sanctorale, who would not get proper masses; this explains why CEA contains almost three times fewer saints’ feasts than CCA. The sanctorale in CEA nevertheless bears a close resemblance to the eighth-century OV, which contains the proper offices for saints and is therefore more closely comparable. Both contain a predominance of apostolic and Iberian saints that tells us about the continuing importance of biblical and local saints in Iberia from the Visigothic period to the period under Islamic rule.

The Cordoban Calendar can also elucidate the question of the orientation of Andalusi Christians, when it comes to the liturgical order of celebrating certain feasts. Some feasts are included more than once on different dates in the same manuscript, one according to the Old-Hispanic rite as we know it from the North, and one according to Roman or Eastern Christian rites. For example, in most versions of the Cordoban Calendar, the Annunciation is recorded both on 18 December (corresponding to the Old Hispanic rite as practiced in the Northern Kingdoms) and on or around 25 March (corresponding to the Roman or Eastern Christian rites).115 This double occurrence stands out when compared to the CEA, OV and northern liturgical calendars, in which the Annunciation is only recorded on 18 December. In the Latin translation of Gerard of Cremona, the feast is included on 21 March but only as ‘a Christian feast’, perhaps as recognition that this was the ‘wrong date’ according to the Old Hispanic calendar.116 In the same manuscript, the feast is recorded on 18 December with the topographical information as to where this feast is celebrated in Córdoba.117 This detailed information suggests that the feast could have been celebrated in Córdoba in December, corresponding to the date on which we find it in CEA. The fact that the feast is in many versions also recorded in March could be the result of copying from an Eastern Christian calendar, as will be further outlined below.

Another feast with a distinctive date in the Old-Hispanic rite (8 January), as opposed to the Roman and the Eastern Christian rites (29 December), is Holy Innocents. Whereas Holy Innocents is recorded in the OV and northern liturgical sources on 8 January only, it is found in all versions of the Cordoban Calendar and CEA on 29 December, except for CCL, which has Holy Innocents on both 29 December and 8 January. The formulation of the feast in December in CCL is: ‘the killing of the Children in the city of Bethlehem by the hand of King Herod, when the message that Christ was born reached him. He thought he could kill Christ by killing them all.’ This is very clearly a translation of what is recorded in some Arabic versions, meaning that it was more likely included in the Arabic original.118 The entry for 8 January is not found in any of the Arabic versions, however. Considering the inclusion on 29 December in CEA as well, it is possible that Holy Innocents was actually practiced on 29 December in al-Andalus. CCL may have added the feast on 8 January at a later stage to reflect what was known about the Old Hispanic calendar in the north. Consequently, as opposed to the Annunciation, it could be that Holy Innocents was practiced in al-Andalus on a different date than in the north and that its Old Hispanic date in CCL is the result of a retrospective correction from the twelfth or thirteenth century.

Another question related to the orientation of Andalusi Christians concerns the celebration of feasts characteristic of the Eastern Christian, and, more specifically, Nestorian tradition. These include entries listing the ‘Feast of the Roses’, the ‘Feast of the Wax’, the ‘Feast of the Church of Mary’, the ‘Beginning of the year’ and the ‘Fast of Elijah’.119 These feasts and events are also found in chapters on the Nestorian calendar in al-Bīrūnī’s Chronology of Ancient Nations.120 We might expect the occurrence of such entries in the calendar to be the result of copying from a Nestorian calendar.121 Forcada suggests that the Arabic calendar kitāb al-azmina of the Nestorian physician al-Māsawayh (d. ca. 857) was possibly known by ʿArīb ibn Saʿīd, although a direct relation with the kitāb al-azmina has not been established.122 However, while there is evidence of Nestorian influence on the Arabic Christian literature of al-Andalus, and even the presence of Nestorians before and after the Arab conquest in al-Andalus, there is no evidence which suggests the celebration of Nestorian feasts in al-Andalus.123

One final aspect that is important for the interpretation of the Cordoban Calendar as a testimony of Christian practice is the use of terminology and theological details in describing feasts and practices. The wording in the Arabic versions of the Cordoban Calendar suggests a certain distance from the Christian content. Especially in comparison with CEA, a book supposedly produced and used by a Christian community, particular choices in the use of terminology in the Cordoban Calendar stand out. Whereas CEA uses Arabizations of Latin terms (e.g., šantūrum for the plural of saint, diyāqan and arḥdiyāqan for deacon and archdeacon), the Arabic versions of the Cordoban Calendar use proper Arabic terms for the same and other words (e.g., šammās for deacon) and mostly leaves out the title ‘saint’. Moreover, CEA refers to biblical figures with their Arabicized Latin names (e.g., Marīya for Mary and Yasūʿa for Jesus) or their typical biblical attributions (Jesus the Lord, John the Baptist and Mary the Virgin, mother of the Lord).124 The Arabic versions of the Cordoban Calendar, on the other hand, use the Arabic/Quranic names to refer to Mary, Jesus and John the Baptist (Maryam, Aʿīsā and Yaḥyā ibn Zakariyā) and cite attributes that are generally more emphasized in an Islamic context, as for instance ‘Mother of Jesus’ for Mary, and ‘the Messiah’ for Jesus. Additionally, the attribute ‘Baptist’ is not mentioned with respect to John, while his role as prophet is emphasized.125 A comparison with the terminology applied in other sources, for example an eleventh-century Arabic canon law collection, confirms to some extent that the terms applied in CEA were more specifically used in texts produced and used by the Christian community. As a result, we can imagine that the Cordoban Calendar, or at least the Arabic manuscripts of the text, were produced in an Islamic context.126

The difference in perspective is also visible in some theological details used to refer to biblical events that are presented as historical truth in the CEA, whereas the Arabic manuscripts suggest a certain distance towards the described practice. This is evident in the case of “the finding of the cross” (wujūd al-ṣalīb) in CEA, versus “the feast of the Cross, which they call Santacruz. On this day the Christians claim [yazʿamūna] to have found the abandoned cross of Jesus in Jerusalem” in the Arabic versions of the Cordoban Calendar.127 Another example is found in the description on 15 August on which the “Christians celebrate the feast of which they say [yaqulūna] it is Mary’s Assumption”, an event which in CEA is just “Assumption of Mary.” Such differences suggest that the authors or copyists of these two sources did not share a religious background, since there are different Christian and Islamic traditions surrounding Jesus’ and Mary’s deaths; according to the Quran, Jesus was not crucified and Mary was not assumed into heaven. The linguistic and theological differences in how feasts and figures are named and described in the different sources tell us something about the community of production and use of this source, suggesting that whoever compiled CCA was, unlike the compiler of CEA and CCL, not Christian. This rethinking of the text calls us to question more broadly the purpose of the compilation, its audience and its relationship to the Christian practice it is describing.