Did a Poet with Donkey Ears Write the Oldest Anthem in the World? Ideological Implications of the Computational Attribution of the Dutch National Anthem to Petrus Dathenus

Introduction
The Wilhelmus has been the official national anthem of the Kingdom of the Netherlands since 1932. The song carries a wider relevance that extends well beyond the Low Countries. According to the authoritative Guinness Book of Records, the Wilhelmus is the national anthem with the oldest music in the world: we are able to date the tune and text to the years 15681572 during the Dutch Revolt, a key episode in the history of the Early Low Countries. Moreover, in the song's fifteen couplets, an anonymous poet has immortalized a dramatic internal monologue of William

the Silent, Prince of Orange (1533 - 1584), a well-

known figure who has played a decisive role in the political history of Europe (Van Stipriaan, 2007).

In the earliest sources, the Wilhelmus has invariably survived anonymously, in print collections of rebel songs (the so-called geuzenliederen or ‘beggar songs') that date back to the Spanish Occupation in the Low Countries (De Bruin, 1998). Only some of these songs are attributed to known authors; the majority, including the Wilhelmus, are not. Apart from the supposed date of composition (1568-1572), there are few historical facts that could help the attribution. Although the Wilhelmus does not explicitly choose sides in contemporary religious conflicts, circumstantial evidence strongly suggests that the text was written by an author of Flemish or Dutch descent, who was living in a German refugee community at the time, perhaps in the vicinity of Heidelberg, because of a number of striking intertextual connections to other songs that were composed in that area.

Ever since its creation in the late sixteenth century, the attribution of the song has not ceased to puzzle scholars as well as other inhabitants of the Low Countries. Only decades after the song's composition, there seems to have been considerable confusion already: in various sources, we find widely divergent attributions of the song to a number of famous authors, such as Marnix of Saint Aldegonde (the mayor of Antwerp, during the city's famous Fall in 1585) or the religious author and philosopher Dirck Coornhert. Many other candidate authors would be suggested in the next centuries, the credibility of which could vary strongly. In the public opinion, Marnix has long remained the most popular candidate, although scholars have never reached any definitive agreement on the issue. As late as 1996, for instance, an entire doctoral thesis was devoted to the authorship of the Wilhelmus. In this thesis, Maljaars predominantly argued that Marnix could not have been the author, relying on traditional evidence: the results of the close reading of the Wilhelmus, and comparison between the Wilhelmus and other texts by the presumed author(s).

In 2016 an interdisciplinary team of scholars has tackled this age-old issue from a new perspective: sty-lometry. For most of the candidate authors which have been suggested for the Wilhelmus, we have available relatively sizable oeuvres of lyrical poems or even highly similar songs. The comparison of the Wilhelmus to those reference oeuvres, using state of the art stylo-metric methodologies, should allow us to estimate the relative distance from the anthem to each candidate author (authorship attribution) and verify their authorship (authorship verification). Many issues, however, make this comparison far from trivial: the texts are short (the Wilhelmus only counts 500 words), we only know younger, potentially corrupted versions of the texts and rarely have autographs, the spelling of the material is highly unstable etc. We have tried to tackle the latter issue through part-of-speech tagging and lemmatizing the texts (Kestemont et al., 2016b): instead of performing measurements on the original surface forms, we would restrict our analyses to the most frequent tag-lemma pairs (MFTLPs), which normalize the spelling of tokens.

In this paper, we will report several authorship experiments, using both the attribution and the verification setup (Kestemont et al., 2016a), in which we have compared the Wilhelmus to a representative set of contemporary authors, among which the main candidate authors as well as some background authors that merely served as ‘distractors' or ‘imposters'. We include a small selection of these below. Surprisingly, these experiments without exception pointed towards an obscure, vilified author who has never even been mentioned as a candidate author: Petrus Dathenus (ca. 1531-1588). The first series of plots are rather naive Principal Components analyses (300 MFTLPs) which each confront textual samples by two candidate authors and the Wilhelmus (in white). In these binary comparisons, the Wilhelmus is attributed to Dathenus without exception. The same goes for the verification experiment (which runs entirely parallel to the experiments run on the Caesarian corpus in Kestemont, et al. 2016a): when compared to both target and imposter authors, the Wilhelmus is significantly closer to Dathenus's texts than to any other candidate author from this period for which we have texts available.

Fig. 1: Naive 2-dimensional PCA plots in which textual samples by two authors (Datheen vs Marnix; Datheen vs Heere) are confronted, including the Wilhelmus.

Fig. 2: Cluster map for the verification results obtained for the Wilhelmus and a number of highly relevant candidate

authors (Kestemont et al. 2016a).

Fig. 3: A late eighteenth century caricatural depiction of Petrus Dathenus with donkey ears, to symbolize his alleged poetical ignorance.

Dathenus is primarily known as the author of a complete Dutch adaptation of the Psalms, which became extremely influential in the second half of the sixteenth (and which is in fact still sung today in some reformed communities). His contemporaries considered him a great and dangerous orator because of his convincing way with words. Nowadays, Datheen has the reputation of being a very poor poet. To what does he owe this bad reputation? Our present-day image of the man goes back to the late eighteenth century when the pressure grew to have his Psalm adaptation replaced by a more modern one in churches. In order to increase the pressure on Dathenus's Psalms, people started mocking the poet through the dissemination of caricatures in which the man would even be depicted with donkey ears (see Fig. 3). It is striking how strongly our present-day view of Dathenus is still determined by the highly anachronistic eighteenth image of this author, instead of that of the respected and influential individual he was known to be in his own time.

In this paper, we will re-assess the sparse, historical evidence that is available for Petrus Dathenus and show that he is, in fact, an unusually strong authorial candidate for the Wilhelmus. Here, we limit our discussion to a single new fact that recently emerged. The Wilhelmus is a so-called contrafact: the song has been composed by writing a new set of lyrics for an already existing melody, a very common practice in early modern song culture. The original melody which was used

for the Wilhelmus was a French song: O la folle entreprise du Prince de Conde. Musicologist have been able to pinpoint when this song was created: it must have been composed (as a Protestant song) during the Siege of Chartres in 1568. The tune must have been introduced in the Low Countries via the Wilhelmus and was not known beforehand. Therefore, it has always puzzled scholars how the Wilhelmus author might have been exposed to this French tune. Intriguingly, it turns out that Dathenus must have been present at the Siege of Chartres as a field preacher on the protestant side. Thus, although he has never made it to the official candidate list, Dathenus is in fact the only candidate, who not only has the right stylistic profile, but of whom we also argue that he was directly exposed to the base tune of the Wilhelmus.

In our paper, we will not go as far as to claim that the neglect of Petrus Dathenus as a potential candidate author for the national anthem of the Netherlands has been an ideological ‘cover up operation'. We will discuss, however, the anachronistic biases and prejudices which so far have prevented the identification of Petrus Dathenus as a potential candidate author. From the point of Digital Humanities, it is important to stress that we base this research on a bold computational attribution to an author who, at first sight, seems a highly unlikely candidate; a human expert would never even have dared to think of this attribution. Nevertheless, exactly because machines do not carry the same set of preconceptions as humans, the application of stylometry is able to induce serendipity in humanities research and open up new perspectives.

Bibliography
De Bruin, M. (1998). 'Het Wilhelmus tijdens de Republiek'.

In: Volkskundig Bulletin 24, p. 16-22.

Maljaars, A. (1996). Het Wilhelmus: auteurschap, datering

en strekking: een kritische toetsing en nieuwe interpre-tatie. Kampen.

Kestemont, M., Stover, J., Koppel, M., Karsdorp, K. and

Daelemans, W. (2016a) ‘Authenticating the writings of

Julius Caesar'. In: Expert Systems with Applications 63: pp. 86-96.

Kestemont, M.; De Pauw, G.; Van Nie, R. and Daelemans, W. (2016b). Lemmatization for variation-rich languages using deep learning. In: Digital Scholarship in the Humanities, advanced access:

http://dx.doi.org/10.1093/llc/fqw034.

Ruys, T. (1919). Petrus Datheen, Amsterdam.

Van Stipriaan, R. (2007). ‘Words at War. The Early Years of William of Orange's Propaganda.' In: Early Modern History 11, p. 331-349.