INVENTING TRADITION AND CONSTRUCTING IDENTITY: THE GENEALOGY OF TJMAR IBN HAFSÜN BETWEEN CHRISTIANITY AND ISLAM

''Utn.ar b. Hafsim, the famous anti-Umayyad rebel in al-Andalus in the ninth century, laid claim at one stage in his career to a long and distinguished ancestry, including several generations of Muslims and four Christian generations. In this article I argue that the ancestry is an invention, invented to serve immediate political needs. There is no reason to suppose it genuine; we have no other example of such a genealogy from the Islamic world and scarcely any from anywhere else; and the genealogy presents other problems. The consequences of this are of some significance: first, understanding the genealogy as an invention enables us to understand the career of Ibn Hafsim himself in a different light, and the better to assess what he was doing (and when) in the course of his long career. Secondly, we are in a position to look very differently at modem interpretations of his career: understanding the genealogy as a forgery means that we have no longer any reason to see Ibn Hafsûn as a descendant of late Visigothic nobility, and hence casts some doubt on the view of his activity as some sort of local Christian political revanchism.

Hafsün himself but also about Simonet and about Spanish writing of a century ago.^ Simonet here, like others, relies on and accepts as true the genealogical information given by the sources.It is worth looking at this genealogy more closely than it has been looked at in the past, I suggest, not only in order to see what it may have to tell us about different styles of writing national, or what is seen as national, history, but also for what it contains of ascertainable historical truth.^ This question is of more than antiquarian interest, for both Acién and Fierro have drawn attention to it as relevant for how we understand the Hafsùnid rebellion.For Acién the genealogical details, and information associated with them, are of importance and value as they tell us that Ibn Hafsün was a member of a group "con ascendientes que los vinculan con las antiguas clases dirigentes", ^ and he sees in such rebels against Islam not only continuators of a local political and social tradition in the peninsula but also, in a physical sense, heirs to the feudal lords of the Visigoths.Ibn Hafsün is a key element in Acién's case, for the genealogy and its associated information demonstrate that Ibn Hafsün had a distinguished ancestry and even that one of his ancestors was a qümis, or local magnate of some sort, in the region of Ronda.^ For Fierro the claim about the descent appears "aceptable", and as to the rank of the ancestor of Ibn Hafsün, she suggests that "en principio no hay razón para poner en duda esta información."^ It seems to me that there is reason to doubt it, and, to the extent that this is of consequence for Aden's case, that it weakens that case.I base my argument here wholly on the genealogy itself, for the second claim, as to the rank of one of the ancestors of Ibn Hafsün, rests entirely on the value we accord to the genealogy: if that falls, then anything based on it must fall too.
"^ For Spanish historiographical attitudes to Islam and the Arabs generally see Monroe, J.T., Islam and the Arabs in Spanish Scholarship (Sixteenth Century to the Present), Leiden (Medieval Iberian Peninsula, Texts and Studies, 9), 1970.

The genealogy: sources
We have several versions of the descent of ''Umar Ibn Hafsùn.These occur in the works of Ibn Idhàrï, Ibn al-Khatîb and Ibn Khaldùn.In addition to these, we have a couple of references in other historical works and in biographical dictionaries, including a version of his descent in the recently published A''lam Malaga.The first three are all more or less of the eighth/fourteenth century: Ibn Idhârï is known to have been writing in Fes in 712/1312-13, while Ibn al-Khatîb (713/1313-776/1375) was a slightly older contemporary of Ibn Khaldùn (732/1332-808/1406).As to the rest, the writer Ibn al-Qütiyya, whom we shall meet below, is of the fourth/tenth century (he died in 367/977) and the biographical dictionary called Alâm Malaga is a product of the seventh/thirteenth century.With the exception of Ibn al-Qùtiyya, however, all of the versions of the background of Ibn Hafsùn that we have depend on the fifth/eleventh century historian Ibn Hayyân (377/987 to 469/1076).
Ibn Idhârî tells us that "according to Hayyân b.Khalaf [Ibn Hayyân], "TJmar b.Hafsùn was one of the great rebels in al-Andalus; his descent was: TJmar b.Hafs, who was known as Hafsùn, b.TJmar b.Ja^' far b.Sh.t.y.m b.Dhubyán b.Farghalùsh b.Idhfíinsh, of the musàlima (converts to Islam) of the dhimmîs, from the Mm of Tâkurunnâ, in the province of Ronda, and Ja^' far b.Sh.t.y.m was the one who converted to Islam, and his descendants spread out (? increased) in Islam."^ If we compare this with our other sources, we get a very similar picture.Thus Ibn al-Khatîb also gives us information about TJmar's background.^ He tells us that he was the son of Hafsùn, the son of TJmar, the son of JaYar, al-Islàmî, the son of K.s.m.s.m, the son of Damyân, the son of F.r.gh.l.w.sh the son of Adhfunsh (= Alfonso).He tells us that according to the "author of the Ta'ríW, i.e., Ibn Hayyân, writing in the fifth/eleventh century, Umar was from Ronda, in the province of Tâkurunnâ; his great-grandfather Ja^' far was a convert to Islam (isldmi) and had moved to Ronda, where he had ^ Ibn IdhM, al-Baym al-Mughrib, ed.G.S. Colin and E. Lévi-Provençal, Leiden, 1948 (repr. Beirut, n.d.), II, 106.The editors note that the form Farghalùsh is vocalised in their source.And they vocalise also the form Idhfíinsh. Hayyàn give his descent (nasab).''^^ We note here that one important generation has fallen out, between Ja^' far and Damyán.And the unusual word al-qümis ('count') has somehow been changed at some point in the transmission into the commoner word al-qiss ("priest", equally plausible, in this context), a change which in Arabic script does not amount to very much.Both of these changes can easily be corrected, as the information in Ibn Khaldün comes from Ibn Hayyán, as does the information in the other sources considered here, and it can thus be checked and controlled from them.
In the work A'iâm Malaga, compiled by Ibn ""Askar and Ibn Khamïs in the seventh/thirteenth century (the earlier of the two writers, Ibn ''Askar, died in 636/1238), we have the following information: ^^ TJmar's family was from Ronda, from where JaYar Musálima (on this last see below).Their ancestor Marghalüsh was a Qümis in Ronda; he [Marghalüsh or ''Umar?It makes little difference, but presumably the reference here is to TJmar] was mentioned by Ibn ^^ The YQxh fakhkhama means to pronounce in an emphatie way; here the meaning cannot be more than a reference to the lengthening of the name, with the ending -un, and Simonet's reference, quoted above, to Ibn Hafsûn's neighbours' changing "su nombre por honor ó énfasis" seems exaggerated, if of a piece with the general tenor of the passage (and his book as a whole).
^^ Abï al-Fayyàd; ^^ and his descent (nasab) is mentioned by Ibn Hayyán".Thus this biographical dictionary, which, though fairly late, antedates our other sources, and, like them, also depends on Ibn Hayyán.

Genealogy and identity in Islam
What these sources show is that Simonet has painted a slightly romanticised picture of "nuestro futuro héroe", as he calls him.But our sources also tell us something else, much less clear.We have here the genealogy of Ibn Hafsùn going back for some seven generations: Umar b.Hafsün b. ""Umar b.Ja' ^far (al-islàmî) b.K.s.m.s.m (or Sh.t.y.m, or Sh.n.t.m) b.Damyán b.F.r.gh.l.w.sh b.Adhfünsh (on the variations in the forms of the names, see below).Such a long recorded ancestry is rare for anyone in the Islamic tradition, as a glance at any biographical dictionary will demonstrate.The reasons are obvious.Even in a society as concerned as those of classical Islam with ancestry and descent, most people cannot boast known ancestors going back for more than three or four generations, in the best cases.The best cases are uncommon, and the great majority of people can name no more than their father and, perhaps, their grandfather.^^ More than this is true.Another glance at the biographical dictionaries demonstrates a second fact: most people, indeed we can probably say all people, in these dictionaries have, or know, or record, or have recorded about themselves, only Muslim ancestors.That is to say: conversion to Islam seems to have operated in such a way that the neo-Muslim was regarded as bom again, washed clean of his pre-Islamic, non-Islamic identity and ancestry.We see this very clearly in the practice, recorded for us in the fifth/eleventh century, of replacing an awkward-sounding non-Arabic name of the real father of ^^ Ibn Abï al-Fayyàd (died in 459/1066) wrote on the history of al-Andalus up to his own day.He was a pupil of the famous heretic al-Talamankï.See Pons Boigues, F., Los Historiadores y Geógrafos Arábigo-Españoles 800-1450 A.D., Madrid, 1898, 138-39, no.105 (with references to entries on him in biographical dictionaries); Alvarez de Morales, C, "Aproximación a la figura de Ibn Abï 1-Fayyád y su obra histórica", Cuadernos de Historia del Islam, 9 (1978-79), 29-127.
^"^ For some useful remarks on the character of the Arab genealogy, see Szombathy, Z., "The Nassâbah: anthropological fieldwork in mediaeval Islam", Islamic Culture, LXXVIII, no. 3, July, 1999, 61-108.a new convert to Islam with the artificial paternal name of ' Abd Allah (literally 'servant of God'), "for we are all God's servants."^^ Even where we might know of the identity of an ancestor of an individual Muslim who was not himself a Muslim (though it is very rarely indeed that we do know this), he is not normally included in actual lists of such ancestors such as are found in the biographical dictionaries.The list of named ancestors stops at the generation before such a link to non-Islamic identity.We can occasionally see this happening, for we can (not very often) identify non-Islamic names at the chronologically earliest stage in some genealogies, and we can assume with a fair degree of certainty that they are the generation of the conversion.Their sons have good Muslim names.(It is striking here that two of our sources, Ibn Idhârï and Ibn al-Khatîb, both of them dependent on Ibn Hayyán, report that the generation of the conversion was that of Ja'^far: if this is true, then we have to assume also a re-naming.The only alternative is to assume that the Christian father of JaYar will have given his son an Arabic, and in the context also Muslim, name: this seems both unlikely and in this specific case a little too serendipitous to be true.Could the man have known when his son was bom that he would convert to Islam?) 1^ Although in some cases it may be that the sons are the converts, common sense suggests that those people with non-Islamic names are themselves the ones who converted, thus putting themselves in a position to give Muslim-sounding names to their children.(It is certain that there must be many other cases where the individual himself or the historical record more generally has, so to speak, sanitised the record, basically by eliminating that first Muslim, but non-Islamic looking, generation from the record.In most cases, individually considered, this cannot now be seen; but taking whole classes of people, such as the bulk of the contents of a particular biographical dictionary, it can be shown to be the case.)This fact was recognised long ago, by Richard Bulliet, and exploited by him over ^^ [Ibn al-'^Attár], Formulario Notarial Hispano-Arabe por el alfaqui y notario cordobés Ibn al-'Àttâr (s.x), éd.Chabneta, P. and Comente, F., Madrid, 1983, 406.^^ Simonet, in the passage quoted above, suggests that JaYar changed his name at his conversion to Islam.We are not told this in any of our sources.It is purely Simonet's invention.We have no reason, far less evidence, to suppose that converts changed their names on conversion.twenty years ago to very exciting effect in his study of the process of conversion to Islam.^^ But it seems to have other consequences too.
If we do not find pre-Islamic identities surviving in the cases of other people, what does it mean when we find recorded not only a pre-Islamic identity, as here, but one reaching back for four generations into the pre-Islamic history of the family, and perhaps also into the pre-Islamic history of the Iberian peninsula?This is not an unusual case, it is worth stressing.It seems to be unique.
It should be noted here that there is an important distinction to be made between actual ancestry, which we all have, and possession of a genealogical list of names attesting to the details of that ancestry.Very few of us, today, in western societies, have that.It is often possible, today, from our sources, to reconstruct the list of the ancestors in the male line (the only relevant ones from his, and his society's, point of view, as markers of his identity) of a medieval Muslim going back for several generations.We can do this because occasionally our sources are very rich and dense, and supply a good deal of detail about certain individuals.We can then put the details together in order to discover what the sources do not tell us explicitly, who was the father of the man in question, and who was his father, and so on backwards.In many cases, however, for medieval Muslims, such an exercise is unnecessary, because we actually have the information, in the form of a list of a man's ancestors, his genealogy.^^ The names of a man's ancestors were part of his name.^^ ^^ Bulliet, R.W., Conversion to Islam in the Medieval Period.An Essay in Quantitative History, Cambridge, MA, 1979.For a discussion of Bulliet's thesis, and a suggestion as to how it may be refined further for the Andalusi case, see the paper of Pénelas, M., 'Some Remarks on Conversion to Islam', presented at the Workshop on Conversion to Islam in the Mediterranean Muslim World, 4-6 September, 1997, in Rome, organised by the European Science Foundation Programme on Individual and Society in the Mediterranean Muslim World, see Pénelas, M., "Some remarks on Conversion to Islam in al-Andalus", Al-Qantara XXIII (2002), 193-200.
^^ That this is not always the case, and that much profit can be derived fi"om reconstructing such genealogies today, are facts which are amply demonstrated in such enterprises as those included in the series Estudios Onomástico-Biográficos de al-Andalus (Madrid and Granada, Consejo Superior de Investigaciones Científicas), in which nearly a dozen volumes have appeared since 1988.
^^ For names generally in Islam see especially Sublet, J., Le Voile du Nom, Essai sur le Nom Propre Arabe, Paris, 1991; and Wild, S.,"Arabische Eigennamen", in Grundriss der arabischen Philologie, I, Sprachwissenschaft, éd.Fischer, W., Wiesbaden, 1982, 152-64, with useful bibliography at 164.This has significance in such a society because identity for Arabs, and hence also for Muslims in medieval Arab-Muslim societies, was expressed in terms of names which reñected, in the first place, descent, in the male line, going back for an indeterminate number of generations.Indeterminate because in theory ancestry could go all the way back to Adam; because memories could not hold a huge amount of information for every case; because a small number of generations was sufficient in the great majority of situations; and despite (? or because of) the fact that a longer genealogy somehow indicated a better genealogy.This applied particularly -in fact for all practical purposes only-to Arab descent.It was the norms of Arab ideas of identity which took over in the world built by the Arabs in the name of Islam.

Genealogy in other societies
If all of this wears a simplistic appearance, it is worth looking at other societies in the area and comparing it with what we find in them.No one remembers the name of the father of Socrates, or of the parents of Hammurabi; although the name of Cicero contains the name of his father, it does so only when considered in a full form which would be almost wholly unrecognisable to anyone who knows him as M. Tullius Cicero.Marcus is his first name, in modem, western terms, Tullius is a sort of family name (whence, perhaps, the old-fashioned English habit of referring to him, half familiarly, as Tully), and Cicero a tribal affiliation, of limited significance in the reality of late republican and early imperial times.His father's name does not appear in his name in this form.And his mother's name is not part of his identity.In the Jewish tradition we know the name of the father of Moses but his grandfather's name survives only because the Bible takes special care, in the case of this peerless individual, to record his entire genealogy back to the patriarchs.But what of such characters as the very successor of Moses, Joshua -son of Nun, but who was his grandfather?If Biblical Jews do have fathers, what can we say of their grandfathers?Jesus is exceptional among Jews, in the extreme, in having a full genealogy, going back to David and beyond, but the logic of his career made that a necessity for the Gospels.And what of the earliest Christians?Who was the father of St Paul? of St Peter?And a little later, the Christians seem rather to resemble ancient Greeks: their identity is in their individual, personal names.When something is known of their descent, it is usually the identity of their parents, rather than that of their father alone, that is known, and that knowledge does not go back further than a single generation.^^ Augustine, to take a well-known example, we know as the son of a pagan father, Patricius, and a Christian mother, St Monica -but this fact is not part of his identity as an individual, it is simply a fact about his backgroxmd.Even in the case of royal dynasties, kings have personal, individual names, and while their descent is clearly important for their succession as rulers, the names of their fathers are not part of their own names.Finally, to bring this closer to home again, if we look at the Christians of Visigothic Spain of the period immediately before the conquest, and at the Christians of al-Andalus itself, we receive the same impression.Our sources for Visigothic Spain and for the Christians of al-Andalus tell us the same as we have seen from other non-Muslim societies: identity was a personal matter; while we might be able on occasion to find out the name of a person's father, and sometimes the names of a person's parents, their names are incidental to a person's identity, not part of it.^i Identity, to return to apparently simplistic statements of the truth, is differently constructed in different contexts.
^° It is in fact very easy to link regular knowledge of the names of both parents, rather than just of one (normally that of the father), as part of a person's identity, with knowledge of his or her descent going back only for one generation: the doubling of the number of characters in each generation, together with the complications attendant on any method for describing such descent, make it clear that a single generation must be enough.(For examples of the awkwardness involved in presenting such data, particularly in pre-modem cultures, see the Jamhara of Ibn Hazm.)In general, it may be added, a society in which a person's identity takes equal account of both sides of a person's ancestry is not excessively concerned with ancestry going back very far (Nazi Germany is an exception here, and it demonstrates, if it does not prove, the rule).
^^ For the Christians of al-Andalus, the evidence is available most conveniently both in Gil (see below,

Genealogies as sources
The names that we find in the genealogies of Muslims can provide precious information of different sorts.Thus, even without the information given by the word al-islàmi, in the case of the genealogy of Ibn Hafsùn, we could tell that Ja^' far was either a convert or the son of one, because of the linguistic (and more broadly cultural) change in the character of the names of his father and himself His father was called something that is realised by Ibn al-Khatîb as K.s.m.s.m, which is clearly to be seen as a corruption of a late Latin name (interpreted by Simonet, perhaps correctly, perhaps not, as a form of the name Septimius).And the earlier generations have names which are as clearly, perhaps even more clearly, Latin, or latinate, in origin: Damyán b.F.r.gh.l.w.sh b.Adhfíinsh.
It could of course be argued, as against this, that the change of names need not of itself demonstrate a change of faith; in the absence of the information about Ja^' far provided by the word aUislaml, it might have been that the change in the linguistic character of the names was not evidence of a change of faith, but simply reflected, several generations after the event, a conversion not in the generation thus indicated, but in the earliest generation recorded here.And indeed, in theory at least, this sounds as if it could be, if unlikely, at least plausible.In the context of our evidence as a whole, however, it falls to the ground.We have not a single example, among all the several thousands of recorded genealogies fi'om al-Andalus (that is to say, not lists of ancestors which can, to greater or lesser degree, be reconstructed, but among those lists which we actually have from the middle ages, for example in the endless entries of our biographical dictionaries, so excellent for al-Andalus) -we have not a single example of a genealogy which records more than a single generation of non-Arabic, non-Islamic names.A Latin or Latin-type name is always the earliest name in a list in which it occurs, and it is always alone, preceded chronologically by nothing and succeeded only by a normal Muslim, Arab name.The exceptions, as will be seen, are so few, in part dubious, and in part of a special type, that they can be disregarded.^^ What follows should not be misunderstood.This is not intended to be a comprehensive collection and study of Latin (together with other, similarly non-Arabic) names in al-Andalus.I am well aware that a thorough examination of other sources, and of other types of source, would throw up additional names of this type.The names as such are not to my purpose here.Here I am concerned with genealogical structures.The little body of material presented here is intended merely as a not unrepresentative sample of what can be obtained, using a large and not unrepresentative collection of material.The broader task of collecting and analysing such names for al-Andalus as a whole is still to be done.
^"^ One of the readers fox Al-Qantara suggests including here also the family of the Banü Martinîl (see Molina, L., "Familias andalusíes", E.O.B.A., II, 1989, 81-2) (= Ibn al-Faradî, nos. 1, 63, 633, 1099, 1112).The form Martín is found, though much later and in a man from outside al-Andalus proper; what we have here, however, is a diminutive, and hence on this ground alone may not necessarily reflect a genuine person's name.5, 11, 13, 14,16,18,19,29,30,39), possibly thirteen (if we allow also nos.17 and 36; but both of these could easily be oriental), ^^ really belong on such a list, in that the names recorded appear definitely to reflect a Latin background.Of the rest, the largest group is that whose Latinity must be regarded as dubious: these are nos.3, 4 (Balakâyash, though Ilyán makes this look doubtful), 12, 24, 25, 28, 32, 33, 34, 37, 38, making a total often or eleven.Then there is the small group (nos.26, 27 and 35), where we have the name Bïbish.It is very tempting to regard this as simply an arabicisation of the Romance Vives -but how early is this name, or any antecedent of it, actually attested?And if it is not attested very early, how likely is it that we should find it in al-Andalus, outside Christian Spain, at the dates in question here?If the identification is attractive, it cannot for all that be regarded as definitive.No, 6, Ibn Abï Qùmis, is not really a name, as Qùmis was a title, and we cannot be sure that in this case it is to be regarded as an Arabic form of a predecessor of Gómez.No. 10, Qizilmán, might be a variant for Qarlumán -but it also might not be.No. 7, Karsalïn, might be Latin, or it might not be.And no.22, y.n.j., could be anything.
This leaves us with a final, quite large, group, that containing the name Lubb (< Lope, fi-om Latin lupus).We find this in ten entries: nos.2, 8, 9, 15, 20, 21, 23, 24, 31, and 32 (nos.3, 24 and 32 also contain the name B.y.t.y.r, which I have classed as of dubious latinity, both because of the presence in it of the letter t and because of the apparent similarity between it and the Arabic root b.y.tr.; and Lubb occurs twice in no.23).The evidence of this name shows two things: first, the name Lubb seems to have undergone a process of acculturation into the Arabic-Muslim onomastic lexicon of al-Andalus (we do not find it elsewhere in the world of Islam) and to have become an accepted name among Muslims there, particularly but not exclusively in areas bordering Christian Spain (vide no.23 on the list, where the name occurs twice -^these are members of the well-known Hùdid family, who were powerful in the north-east of al-Andalus, especially in ^'^ Similarly, Ibn Bashkuwál, no.153, who looks almost like a candidate for inclusion here, I have rejected after some reflection, as more likely to be oriental: he is Abu 'TJmar Ahmad b.Marwán b.Qaysar al-Umawi, Ibn al-y.m.n.M.sh.His nickname, which may well be Latinate, is of course not relevant in this discussion; the name Qaysar, which might be, looks more oriental in form than Iberian.Saragossa, at various periods); and second, this process happened uniquely with this name-there is no other example of a Christian, Latin name which was so accepted (I do not refer here to Latin, or latinate, forms of Biblical names of course, such as Mïkâ'îl, which we find in no.17, which need not, of course, be fi-om al-Andalus).
It is strange that Lubb should be the only such case of onomastic acculturation to Islam in al-Andalus, and there does not seem to be any good explanation for the apparent absence of other similar cases.This exception to the rule enunciated above seems to go some way also towards confirming the correctness of the interpretation which I give to it.The only other exception seems to be no.4, where we have both the name Balakáyash and the name Ilyân.Whether or not we see the latter as some sort of variant of the name Julian, we do appear here to have two generations of what are non-Arabic, and possibly Latin, names in such a genealogy.But one swallow does not, usually, make a summer (and see fiirther below).
It is also not without interest, and perhaps also significance, that the largest group of entries (twenty two) in this list comes from Ibn Bashkuwâl, who wrote in the sixth/twelfth century; the timing of his work coincides very well with a period when such ancestries as are of relevance here were around in great numbers, and were remembered, as it was some four or five generations after the great wave of conversion which we associate with the late tenth and the eleventh centuries.

Christians as ancestors of Andalusi Muslims
We are faced, therefore, with a situation where the names of Christian ancestors do not form part of the accepted lists of ancestors of Muslims in al-Andalus.They are absent for a variety of reasons: partly because they are not relevant, and perhaps also not really wanted, as ancestors of Muslims -not relevant because Islamic descent was what counted for Muslim identity, and not wanted because their presence served as a reminder of an undesirable element in a person's identity; and they are absent also because with the passage of time and generations they simply drop off a list which tends in all such situations to be no longer than three or four, at a maximum five, names.With every new generation, another early generation would tend to be forgotten.A good example of this process is afforded by one of the entries on our list here: this is no.4, where Ibn al-Faradî, no.188 has six generations of ancestors listed, back to someone called Balakayash; this man's father (Ibn al-Faradî, no.564) has five generations listed, back to the same Balakayash.However, the grandfather (Ibn al-Faradî, no.268) has also five generations of his ancestors listed, taking him back to one generation before Balakayash, a man named Ilyán (if this is genuine).It is noteworthy that the first person in this family tree to bear an Arabic-Muslim name, the son of Balakayash, is called ''Abd Allah; we should ask what sort of exception this case really is -could it be that ""Abd Allah here is an example of something like the process described above (text at notes 14-16)?At all events, here we see the gradual dropping of ancestors after four or five generations actually happening.

The ancestry of ^Umar b. Hafsün: questions and answers
In the case of TJmar Ibn Hafsün, however, we are faced with a list of seven (including Umar himself, eight) generations, of which the earliest four are non-Muslim, Christian, a fact which we are also told explicitly, in the form of the word aZ-Zs' /amf attached to the name of Ja^'far.This must represent a problem.Are we to accept this genealogy as genuine, or may we, should we, are we indeed bound to question it?Is it really possible that in this case alone among all the thousands of converts from Christianity to Islam in al-Andalus during the five centuries following the collapse of 711 the previous four pre-Islamic generations of a person's male ancestry have been preserved?Is it possible that, in a pre-Islamic Christian society which set very little store by preserving the memory of a man's male ancestry over several generations, in this single case, of a man whose ancestors were themselves of relatively little importance, such a list should have existed, have survived and then been preserved for a fiirther four generations after the islamisation of one of the descendants?Is it possible that, in a Muslim society which set almost less store by the preservation of the memory of a pre-Islamic identity, four such generations should have been thus remembered?Is it possible in a society which valued and remembered and recorded and preserved the Muslim identities of so many hundreds of local Muslims descended from Christians, like Ibn Hafsün, without recording the names of a single Christian ancestor of a single one of them, that in just one case such a list, with not one but four generations in it, should have existed, been remembered and preserved for historians of a later age to record?And can it be merely coincidence that that single case should have been that of a man who not only was descended from Christians but also rose in rebellion against Muslim rulers and made adroit use of his alleged descent from distinguished Christian ancestors in his rebellion?
We have only to pose these questions to see how difficult they are.We need therefore to examine the evidence in this unique case with particular care.Our list shows four explicitly pre-Islamic generations, complete with non-Islamic, non-Arabic names, all of them Latin, or Latinate, in corrupt form.That they are more or less corrupt need not, of itself, worry, because it is of the nature of non-Arabic words and names, in Arabic script, to become corrupt in transmission.This is all the more true of names which, as being non-Arabic, are completely unknown in the Arabic tradition.It is not surprising that Alfonso, in the form Adhfunsh, should have survived the transmission process virtually intact, as that name in that form (among others) is relatively well-known from our Arabic sources; k.s.m.s.m (or Sh.t.y.m, or Sh.n.t.m), by contrast, is not such a name, and Simonet's recognition in this of the name Septimius may well be acceptable, if only because almost anything, within certain limits, would suit; and it may also be correct, in the sense that it seems to suit the actual letter-forms that we have here better than other interpretations (correct in reñecting an original name in an original source, not of course, or not necessarily, in identifying a real, individual person).But the nature of the corruptions that occur is such that we cannot really be sure in a case like this of the acceptability of a particular interpretation.Nevertheless, corruptions of this sort aside, we are left with the question whether or not we are to regard this impressive and long list as reñecting authentic descent.If it does, then the interpretation of particular letter-forms is relevant to the actual identities of the persons thus named; if it does not, then the meaning of the correctness of particular interpretations of varying letter-forms is clearly very different, for in such a case the persons named never even existed.If it does reflect actual descent, then we still need to ask how the list came into being, and how it was preserved.If the list does not reflect authentic descent, again, then we are faced also with further questions: who invented it?from what materials?for what purpose?when? and what was the reaction to it?how successful was it?This last is particularly important, not least because we find the information in Arabic Muslim sources, and only in them.If we are interested in broader questions like those mentioned at the start of this article, then answers to these questions have clear relevance in any study connected with the career of Ibn Hafsün.

The Christian names in the genealogy of ^Umar b. Hafsün
The names themselves may be worth a second look.The latest of these in time, the father of the first ancestor of TJmar to bear an Arabic-Muslim name, is in Ibn Idhârï and the A^'lam Malaga Sh.t.y.m or Sh.n.t.m.Though these are not identical with K.s.m.s.m., they are clearly no more than variations on each other, and very close, too, to Septimius; the difference between the various Arabic forms is presumably due to the problems of transmission of "awkward" non-Arabic names.^^ Though there does not seem to be any other Septimius known among the Christians of al-Andalus, Septimius is in itself a perfectly good Latin name, and its interpretation as such here by Simonet is eminently plausible.
The next, going backwards in chronological order, is Damyán, or Dabyán or Dhubyân (the differences here are again simply a matter of ^^ See Ibn al-'^Attàr, loc.cit, for this characterisation of such names (he calls them thaqfla and kanha).
(c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://al-qantara.revistas.csic.es a dot or two).Damián is of course the well-known saint of that name, who with Cosmas serves as the patron saint of physicians.He was known in Spain, as we have a reference to a church dedicated to the pair in "a suburb of Toledo", in the life of Ildephonsus attributed to Cixila (or Helladius), which predates the late tenth century.^9 But there does not seem to be any other person bearing that name in our sources.Was the name really part of the Iberian Christian onomastic lexicon?
Next we have Marghalùsh/F.r.gh.l.w.s.The difference in the first letter is basically just a question of a dot, and in the last of three dots.Again, the difference results from ignorance of the original forms among the copyists.Could we have here an attempt at Marcellus, rather than the Frugelo of Simonet?Both names occur in our sources for Christians in al-Andalus.We have no way of choosing between them, but Marcellus certainly seems an attractive possibility.That the form Farghalûsh is vocalised in one of our sources (Ibn Idhârî) need not compel acceptance of that interpretation: the hazards of transmission of non-Arabic forms are so great, and the potential for error so wide-ranging, that no reading of that sort can impose itself without the support of external material.^^ The earliest name, in chronological terms, in all the lists (with the exception of that in the A^'ldm Malaga) is Adhfünsh.This name is ^^ See Cixilanis Vita Ildephonsi, 1, in Gil, J. (éd.), Corpus Scriptorum Muzarabicorum, 2 vols., Madrid, 1973, I, 61, "ecclesia sanctorum Cosme et Damiani, que sita est in suburbio Toletano" (for the dating, see ibid., 59).
^° Maribel Fierro draws my attention to Guichard, P., "Toponymie et histoire de Valence à l'époque musulmane: un chef berbère valencien du Xlème siècle à la conquête de la Sicile?" Primer Congreso de Historia del País Valenciano, vol.II (Universidad de Valencia), 1981, 399-409, where Guichard raises the possibility of a Berber descent for an Andalusî named Farghalûsh who helped the Aghlabids with a fleet during the conquest of Sicily.In this case, however, the man's actual name was Asbagh b.Wakîl, and Farghalûsh was some sort of nickname (al-ma^rüf bi-Farghalüsh), so that this does not have any necessary effect on the argument here.See Ibn Idhárí, Bay an, cited in Amari, M., Biblioteca Arabo-Sicula, Leipzig, 1857, 355-56; see also discussion in Amari, M., Storia del Musulmani di Sicilia, 2"^ edition a cura di C.A. Nallino, Catania, I, 1933, 404, and especially n. 2 (though what Amari says there about the likely provenance of these Muslim sailors/soldiers from Crete is, as it stands, unconvincing.His citation from al-Marrâkushï, ed.R.F.A. Dozy, History of the Almohades, 2"^ edition, Leiden, 1881, 13-14, does not seem to me to have quite as much direct relevance here as he seems to think).Lirola, J., El poder naval de al-Andalus en la época del califato omeya, Granada, 1993, 96, n. 146 (cont.from 95), also notes the matter.
(c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://al-qantara.revistas.csic.esAlfonso.We know this from its occurrence, in reference to other individuals, who certainly existed, in many chronicles and other texts in Arabic from the middle ages.There is even less difficulty about the identification of the name in this case than in the others.But did this man exist?The name occurs commonly in medieval Arabic texts because quite a number of kings in the different medieval Christian states of al-Andalus were called Alfonso.Could it be that the name is used here precisely because of this fact, in order to lend a hint of a halo of royalty to the ancestry, and hence to the identity, of Ibn Hafsùn?I return to this possibility below.

A textual cria with potentially awkward implications
The significant difference between the list in the A'iàm Malaga and those in all the other sources lies in the apparent addition, in the A^'lam Malaga, of an extra generation, with the name Musalima, at the chronologically earliest stage in the genealogy.Musálima looks like an Arab-Muslim name.If it were that, then it would present major problems, for an Arabic name would be in clear contradiction to everything else that we are told, here and elsewhere, about ''Umar's ancestry.It is an Arabic-Islamic name for a person who was not an Arab, at a time before Islam had come to Spain; it is an Arabic-Islamic name at a period which may even, though we cannot judge this exactly, predate the very birth of Islam; it is an Islamic name in a genealogical list followed by four generations of non-Islamic, non-Arabic names, a phenomenon which as we have seen is otherwise quite unknown.And it is an Islamic name at the head of a list of people who are explicitly recorded as not having been Muslims until several generations later.As can be seen from the text of Ibn Idhárí, however, this word is in fact a product of a misapprehension by the compilers of the A^'lâm Malaga, or their proximate source: the text of Ibn Hayyán, on which this material here depends, in fact says simply "one of the musálima (converts to Islam) of the dhimmïs" {min musâlimat al-dhimma; min and bn in Arabic are almost identical).(The Ta 'ríkh Iftitáh al-Andalus, of Ibn al-Qùtiyya, who died in 367/977, at a great age, also mentions Ibn Hafsùn, but without giving a genealogy.However, it does say of him that "his father was one of the converts of the people of the dhimma'' (wa-kana abühu min musàlimat ahí al-dhimmá), which, like the passage just quoted, confirms the non-onomastic character of this word.)^^ So this apparent worry may be discarded.

The original form of the Christian part of the genealogy
We are left therefore with a list which can be realised, on the basis of the forms in our lists taken as a whole, as Septimius b.Damián b.Marcellus/Frugelo b.Alfonso.If the chronologically most recent of the names, Septimius and Damián, are not otherwise attested in the onomastic lexicon of Andalusi Christians furnished by our sources, they are nevertheless perfectly acceptable names, and one is known from a church in Toledo.And as to the other two names, the earliest generations recorded, Marcellus (if it is that) is an excellent Latin name and is attested in our sources, while the alternative to that, Frugelo, is also attested in our sources; and Alfonso, chronologically the first name in the genealogical list, is not only well attested in the sources but is also a common name among Iberian Christian kings, even before the time of TJmar.
11.The genealogy an invention: when and why?Should we see this list as authentic?As has been seen, there are several difficulties involved in such a proceeding.An invention seems more likely.^^ But if it is an invention, then can we date it?and can we explain why it was produced?The Christian section of ^^ Ibn al-Qûtiyya, Ta'rîkh Iftitâh al-Andalus, éd.Ibrahim al-Abyârî, Cairo-Beirut, 1402/1982, 103.* ^^ One of the anonymous readers for Al-Qantara suggests that "Ibn Hafsùn no es un ulema y se le aplican las reglas de éstos para decir que el nasab es inventado".It seems to me that my argument here is different: in general, we do not have genealogies for Christians; genealogies for Muslims (whoever they are) do not seem to include non-Muslims; in this case we have a genealogy, and it does include such characters; this fact in itself must arouse our suspicions.The material drawn from the genealogies of the ''ulamd' merely serves to strengthen the argument offered here; it is not one of the building blocks of it; nor is its background an element that should weaken its relevance here.Quite the contrary.
(c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://al-qantara.revistas.csic.es the genealogy is found in Ibn Hayyân, and it is to be assumed that he would not have invented it.As we have seen, Ibn al-Qütiyya refers to the background of the family, mentioning that the father, not a distant ancestor, was "one of the converts of the dhimmîs."While we need not feel compelled to take Ibn al-Qûtiyya's assertion as to the generation of the conversion quite literally, it does nevertheless offer a striking contrast to the rich detail that we get in the tradition dependent on Ibn Hayyân.It may not be without significance that Ibn al-Qütiyya lived in the fourth/tenth century, closer to the time of the rebel, whereas Ibn Hayyân lived a century later, in the fifth/eleventh century.However, there are other considerations too.It is unlikely in the extreme that a Muslim would have invented such a genealogy.A Christian background seems more than likely.If it is an invention, then its creation must lie among Christians between the time of Ibn Hafsün himself and that of Ibn Hayyân, in the fifth/eleventh century.^^ Ibn Hafsün lived at the end of the third/ninth century, dying, we are told, in 305/917.^"^ The lifetime of the rebel seems the most suitable period for such an invention to have taken place.It is difficult to imagine any good reason why a later Christian would have invented such a genealogy, not to mention the question how it might then have been transmitted to Arabic writers.But a date in his own lifetime would allow for the insertion of the fictive genealogy in later accounts of his career written in Arabic, accounts such as we know to have existed.^^ Consideration of the details of Ibn Hafsün's career does in fact suggest a date when this invention might have taken place: in 286/ 899, Ibn Hafsün is reported to have converted to the Christianity of ^^ In theory one might argue that the invention must have occurred after the time of Ibn Qûtiyya, since in his day we still have what may be the truth as that was known in the early second half of the fourth/tenth century; but this argument does not appear compelling.Ibn Hazm, in the fifth/eleventh century, knew of an entire work devoted to Ibn Hafsün (Ibn Hazm, Risâla fi Fadl al-Andalus, in Maqqarî, Analectes, II, 118), and al-Dabbï, in the sixth/twelfth century (he died in 599/1203), knew of several such works (see his notice of someone who claimed, though without being able to produce a list of the genealogical links, to be a descendant of TJmar b.Hafsün, Bughya, ed.Codera, F., and Ribera, J., Madrid (Bibliotheca Arábico-Hispana, 3), 1885, 393-94, no.1161).This or something similar could easily have included fuller material from an earlier date than Ibn al-Qütiyya.various other Muslim princelings in north Afi*ica even before the rise of the Fatimids.The conversion to Christianity, therefore, which bulks so large and assumes such a portentous air in much modem writing on the man, may well not have seemed so significant a move from his own point of view.If that is the case, however, then we may be able to go fiirther, and locate the invention of the Christian extension to his genealogy more narrowly around the time of the conversion to Christianity, in 286/899.Either way, we have no good reason to think that the Christian element in the genealogy is anything but an invention; the form of the genealogy, as such a list, is wholly alien to the world fi*om which it allegedly draws its information, and is taken whole cloth from the world of Islam; the contents of the Christian section are sufficiently authentic-looking to have been acceptable only to people who think in terms of that Islamic society and pay no attention to the realities which must have underlain the man's actual ancestry; at the same time, and perhaps because of these facts, the list reflects the basic identity of the man as a Muslim descended firom Christians in al-Andalus; and its creation fits well with the strategy reflected in his conversion to Christianity.

Further implications
The creation of such a Christian genealogy matches well, finally, with a fiirther feature of the character and the career of Ibn Hafsün.In her study of Ibn Hafsün, written as a reaction to the work of Manuel Acién, Maribel Fierro asked whether in fact we can be sure that Ibn Hafsün did convert to Christianity.^^ The difficulties involved in deciding whether a person was a Muslim or Christian in that society at that time may have made such identification very difficult and as a consequence facilitated such doubts about a person like Ibn Hafsün.The evidence upon which Fierro relies to illustrate this point, however, does not commend her proposal greatly.As she recognises 38 Fierro, "Cuatro preguntas", 244-46; and Chalmeta, P., "Precisiones acerca de TJmar b.Hafsün", Actas de las lIJornadas de cultura árabe e islámica (1980) herself, her first piece of evidence, concerning the conversion of Qûmis b.Antunyan, is weakened by the fact that there were clearly economic motives involved in the action of those who doubted the man's conversion.In the second case, that of the so-called 'martyrs' of Cordoba, the religious identity of the children of mixed marriages shows, not "cómo los hijos de matrimonios mixtos se movían entre las dos religiones", but rather that there were forms and structures in place in that society which some people tried to ignore, with limited success.What we hear of the families of these "martyrs" suggests much more the style of life of another Iberian group of later centuries, the Marranos, people who had to hide the faith which they were teaching to their children.The third example which she cites shows Hashim b. ""Abd al-'^Aziz killing his prisoners irrespective of whether they claimed to be Muslims or Christians: but, while the story is horrible and suggests a variety of interpretations, it does not, to my mind, prove the point that she is seeking to make, of "indefinición del status religioso".This is all the material which she cites in order to question whether Ibn Hafsün really did convert to Christianity, and to suggest that the question is not so simple, for Christianity, as against Islam, was such a fluid thing that we cannot, somehow, capture its character there at this time.I remain unconvinced by this argument and by the material adduced in its support.And I remain persuaded, with our sources, that Ibn Hafsün did seek to give the impression of having converted to Christianity, and, in consequence, of having returned to the faith of his fathers.However, Fierro draws attention to something else, of much greater import and relevance here: this is the existence and survival of a number of messianic-type stories about Ibn Hafsün in our sources.^^ She points to Ibn Hafsün as being an example of the saviour-character that we find so often in Morocco, and to the stories about him in this role as having the character that we associate with such figures from there.And she stresses what she sees as a set of parallels betwen the figure of Ibn Hafsün as drawn in our sources and that of ""Abd al-Rahman I al-Dakhil.The parallels are impressive and, like the other material examined here, tend to support the broad thesis offered in these lines.What this set of stories, like the Christian ^^ Id., 250-55.These stories occur essentially in Ibn al-Qütiyya, and we should ask why and how they come to do so.

( c )
Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://al-qantara.revistas.csic.es We are told that he changed his name from ""Umar to a Christian one, Samuel (no special reason for the choice of this name jumps out at us), and that his wife, similarly, changed her name to Columba, a name redolent of Christian associations.(Though both names are unmistakably Christian, in their context, they do not seem to radiate any particular Christian message, and there is no obvious way to explain the choice of these names as against any other Christian names by these new converts.)In terms of his career, as a strategic move the conversion was not perhaps well considered: it cost him some Muslim support, and did not lead to any other advantage for him, in the form of practical support from Christians in the north of the peninsula or elsewhere.Levi-Provençal speaks of "l'instabilité de son caractère et ses tendances à l'opportunisme."^^ And indeed, only a short while later, in 297/910, we find him offering his allegiance and support to the newly installed Fatimid rulers in north Africa (they had become rulers there only in the preceding year), a move which points to swings in attitude reflective of hasty opportunism, if not, necessarily, of instability.^^ if we are looking for a time when Ibn Hafsùn might have invented, or commissioned someone else to invent, a fine, long Christian extension to the list of his known (?) Muslim ancestors, then the period of eleven years between these two conversions, the one to Christianity, the other to support of the Fatimids, seems to be the most opportune.At the start of this time, he will have been trying to stress his Christian background, laying emphasis on whatever could benefit his construction of an identity with roots deep in the Christian past of Iberia; in this connection the use of the name Alfonso, and especially at the head of the genealogical tree, would fit in particularly well at this date, for this is precisely the time when a man of that name, Alfonso III, the Great, was on the throne of Asturias (he reigned 866-910).What better compliment to offer him, in hopes of his support, than the use of his name in the construction of a pre-Islamic, Christian identity for Ibn Hafsün himself?By the end of this period Ibn Hafsün's interest had moved on.We find, in fact, that his interest had moved on almost immediately, for he is reported to have explored contacts with (c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://al-qantara.revistas.csic.es(c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://al-qantara.revistas.csic.esAQ, XXIE, 2002