Introduction

1 Aim of this book

The present study seeks to uncover everything that all the proper names in the Chanson de Roland (from now on Rol.) can tell us about three fundamental aspects of the text: its structure(s), its world-imaging content (i.e., its relationship with the world, even where it does not intend to be "representational reality" in Auerbach's sense) and the extent to which it is the outcome of a development process.

Setting such a goal today is certainly a challenge: it entails a considerable amount of work, and there are two reasons for that.

2 Inventory of names in the Song of Roland

First, the Rol. has an exceptionally rich selection of names: Segre's index of names has twelve and a half pages containing 407 lemmata1 (of which 47 are only references to variants, e.g., from Bramidonie to Bramimunde, but even these require some explanation); the lemmata are spread, according to Duggan's Index Verborum, over 1823 occurrences.2 This means that within the total of 4002 verses, on average, a name occurs in almost every second verse, and a new name in every 10th verse. The two smallest categories are the 10 named horses and 7 named weapons. The named individuals are somewhat different: if we subtract the biblical characters, the saints, the 'heathen' gods and also Homer, Vergil and Turold - 24 altogether - there are 116, that is to say 57 on the Christian side and 59 on the non-Christian side.3 The explanation for this almost perfect quantitative parity on the two sides in the inventory of names (not in the numerical strength of the armies!) lies largely in the fact that the poet constructs the Battle of Roncevaux almost completely (and to a lesser extent the Baligant battle) as a succession of single combats. Furthermore, in his depiction of the two sides, even in the council scenes and other similar scenes, he deliberately provides an almost equal amount of detail, albeit with opposite moral signposting. This reveals not only the poet's basic narrative approach - he is the omniscient author - but also the work's most significant structural principle, which is parallelism. If we count as geographical names all the feudal and homeland information, e.g. (Engeler de) Guascuigne, (Margariz de) Sibilie as well as the corresponding personal forms such as li Guascuinz (Engeler) and the adjectives that go with them such as (li Sarrazins) espans, (palie) alexandrin, the total comes to 201. This is almost twice as many as the named individuals (as defined above), which gives us an early indication that the geographical names are not to be dismissed as being of secondary importance. Indeed, they should be investigated, not only in terms of their real geographical content and their potential symbolic meanings, but also in terms of the contribution that they make to the structure of the work.4

3 Outline of previous research

Secondly, this extensive of corpus of names (even if we start by excluding most of the variants) has been available to researchers since 1837, indeed since Francisque Michel's editio princeps of the Rol. based on the Oxford Ms.; this makes the Rol. one of the oldest topics of all in French medieval studies. The year 1869 can be considered as the date when a methodical study of its names truly began: this was when, even before the launch of Romania, Gaston Paris published his first, almost four-page essay on the geography of the Rol., quickly followed by a second essay in 1873, on the 'heathen' people. Since then, a huge quantity of research material has emerged. For many years, it was mostly about the geographical names; only a few scholars were interested in the personal names, including especially Rajna (1886 to 1897), Tavernier (1903 and passim until 1914-1917a and b) and Boissonnade (1923), although the last two were only partially successful in their efforts. There were a few heated discussions over the course of these many decades, and they unexpectedly came to a head when Lejeune (1950b) brought the names Olivier and Rollant, especially in their paired appearances, into the spotlight. Thanks to Menéndez Pidal (1960) the controversy between (neo-) traditionalists and individualists - to use the terminology of that time - expanded further into a debate about the basic principles of these two different approaches, which medievalists felt obliged to consider as antagonistic rather than as complementary.

Shortly after 1970, however, both the geographical and the personal names began to attract less attention, and this trend has continued until the present time; the only major exception being de Mandach's last book (1993), and even this is not primarily conceived of as an onomastic study. Unfortunately, we cannot say that the topic has been "exhausted" in a way that shows consensus on the key issues or on most of the points of detail. On the contrary, the situation is more like the kind of exhaustion that would be called burn-out in sociomedical circles.

How can we explain this decline in interest? First of all, the big controversy between traditionalists and individualists did not produce a winner. It simply faded away, and in the judgment of most observers it ended in a non liquet. Secondly, a new major problem emerged, through Noyer-Weidner's (1968, 1969, 1971 and again 1979) emphasis on the symbolic elements in the geographical names of the work, especially in the catalogue of 'heathen' peoples, and his sharp criticism of their literal geographical interpretation. And thirdly, an even bigger non liquet burden fell on the huge number of isolated problems that are to be found in epic onomastics as a whole, not just since the period immediately before 1970, but for many decades before that. There are often three or more mutually exclusive explanations for each name, and it is not possible to regard all but one as disproven. In extreme cases there can be many more: in the relevant sections of this study below, I have identified nine for Durendal, and 23 for Tervagan. In the 1970s, this must all have led to the impression that everything examinable about the prehistory and onomastics of the Rol. had been examined but had produced a plethora of possible answers, or at least did not add up to any bigger picture, and therefore in the end, had contributed little to our appreciation of the work: the knowable seemed not worth knowing, multa, sed non multum. Any new researcher who reaches a conclusion like this will surely find another field to specialise in.

And yet this conclusion is wrong, and we could almost say: grotesquely wrong. The whole of this book is a wager on the opposing position. What is knowable and worth knowing about these names has not by a long chalk been ascertained. Almost every name in the Rol. offers novel and interesting aspects, a few names bring real discoveries, and all in all, a whole world opens up in these names. Even more significantly, if for every isolated problem, including the geographical ones, we compare all of the suggested answers, using the usual technical criteria of our discipline, then in a clear majority of cases one answer stands out as far more probable than the others, so that we can with a good conscience call it the right one. In a few cases there may be two competing answers, but almost never more than that. Incidentally, in this endeavour, Segre's stemma holds up magnificently. Moreover, these isolated findings in no way amount to a zero-sum game, but they show that the surviving text of the Rol. is great literature not least in its choice of names and in the way they are used: its names are an important element of the work's structure, they open up aspects of the poet's lived experience and at the same time, they are the key to the previous history of the work.

4 Structure of this study

Admittedly - and this is unfortunately the crux of the problem - we cannot nowadays acquire this kind of knowledge through awareness of methodology or literary sensibility, and certainly not through a new theory of the epic. What we need to do, is collect and then work through much larger bodies of material than most philologists are happy to consider: any conclusions will require a much broader and more thorough underpinning.

The study proceeds, in terms of the order of presentation, from the far to the near: from the Orient with its 'heathen' peoples and their overarching structure over North Africa and Spain, to the Frankish realm and the Franks. Because of the above-mentioned almost-parity in numbers between the two sides in the inventory of names, it is easy to identify two large complexes: the representation of the non-Christian (A) and that of the Christian (C) side. A slimmer mid-section (B) consists of the smaller categories of weapon names, the provenance of the textiles, and the names of horses: there is no reason to allocate these to the two main sides because this would only separate items that are comparable with each other. I anticipate that this A-B-C order will give the reader a more compact overview, than would be possible with the reverse ordering. This means, however, that the work begins with the tricky, and thus far fundamentally misunderstood, catalogue of heathen peoples. The...