Constituting the Corpus, Database, and Network Visualisations of the Querelle

The corpus is principally composed of French-language publications, published between 1762 and 1789 inclusive. It also includes twelve publications that appeared between 1750 and 1761, which fulfil the criteria for inclusion (given below), and which are referred to by Querelle texts. The corpus includes both stand-alone publications and articles from three periodicals: the Année littéraire, the Mémoires pour l’histoire des sciences et des beaux arts (commonly known as the Mémoires de Trévoux or Journal de Trévoux) and the Mercure de France. The reasons for selecting these periodicals are outlined below.

All texts in the corpus do at least one (and often more) of the following:

1.         state an opinion about collège teaching practices;

2.         present plans for the reform of collège-level, public literary teaching;

3.         engage with a text that does (1) and/or (2), and specifically with its ideas about literary education.

To my knowledge, the corpus includes only first publications of texts that first appeared during the years of the Querelle. Finally, it includes thirteen interventions by authorities, made between 1762 and 1789, which do not themselves fulfil the criteria above, but which encourage public engagement with the dispute and are regularly cited in Querelle texts. These ‘catalysts’ include academic concours questions and arrêts de parlement pertaining to the collèges.

I used a number of resources to identify Querelle texts. Four scholarly texts were particularly helpful for their rich bibliographical work: Ferdinand Buisson’s two-volume Dictionnaire de pédagogie et d’instruction [p.256] primaire (1887-1888); Marguerite Figeac-Monthus’s Les Enfants de l’Emile? L’Effervescence éducative de la France au tournant des XVIIIe et XIXe siècles (2015); Marcel Grandière’s L’Idéal pédagogique en France au dix-huitième siècle (1998); and Martine Sonnet’s L’Education des filles au temps des Lumières (1987). Also helpful was the series of sixty-five republished eighteenth-century texts La Réforme de l’enseignement au siècle des Lumières, edited by Dominique Julia (1979). I supplemented these sources with research using digital library catalogues, in particular those of the Bibliothèque nationale de France, British Library and Bodleian Library. Digitised editions of periodicals were also particularly helpful.^[1]

Periodicals

To delimit the scope of enquiry, I considered three periodicals – the Mercure de France, Année littéraire and Journal de Trévoux – to show how titles with different political and editorial stances engaged with the Querelle. The Année littéraire’s articles conform to Fréron’s polemical, anti-philosophe stance.^[2] The Mercure de France addresses a wide public of worldly and literary readers, and generally seeks not to offend.^[3] The Journal de Trévoux, meanwhile, is the mouthpiece through which the Jesuit Order seeks to combat ‘heretical’ thought, be that Jansenism or philosophie, while attempting to avoid excessive controversy.

Articles from periodicals represent 25 per cent of the texts in the Querelle des collèges. One may wonder whether this percentage is artificially high because of the ready availability and searchability of the three periodicals consulted, all now available online. However, the participation of periodicals is not necessarily overrepresented, for two main reasons. First, digital cataloguing is also now remarkably complete for non-periodical publications. Having cross-referenced multiple digital-library and archive catalogues (for [p.257] collections in France, the US and the UK), as well as Google Books and scholarly bibliographies of stand-alone publications, I estimate the place of periodical and non-periodical publications in the corpus to be roughly proportional to their participation in the Querelle. There are undoubtedly articles and stand-alone publications that have been missed, but my method has sought to avoid favouring one type of publication over the other. Secondly, insofar as there are just three periodicals included in the corpus – of the many dozens of eighteenth-century French periodicals – the true figure for periodical participation in the Querelle is probably higher still. We can assume that periodicals counted for more than 25 per cent of this dispute: quite how much more, though, is a question for another study.

It is also worth noting that the publication history of the three periodicals considered is complex. Their titles changed over time, often with the appointment of a new editor or new owner of the privilège; such changes often also brought shifts in periodicity, thematic focus or political stance. The corpus only includes articles published in these periodicals for as long as they retained their title. This means that Année littéraire and Mercure de France articles are included across the duration of the Querelle (1762-1789), while those of the Journal de Trévoux are included only from 1762 to December 1767. In December 1767, some years after the Jesuit expulsion, the owner of the privilège (Didot) and the journal’s new editor (Jean-Louis Aubert) renamed it the Journal des Beaux-Arts et des Sciences.^[4]

It is difficult to make precise author attributions for articles in these periodicals, due to the common practice of leaving articles unsigned. Typically, the only signed texts are ones submitted by readers, and even these often adopt pseudonyms. Some details can nonetheless be ascertained. We know that, until his death in 1776, Année littéraire articles were written or edited by Elie-Catherine Fréron. From 1776 until 1790, the privilège was transferred to Elie-Catherine’s widow, [p.258] and the journal was initially directed during this period by his son, Louis-Marie Stanislas Fréron (1754-1802). As for the Mercure, from 1750 to 1789 the journal had over half a dozen editors, and an even greater number of regular collaborators. And from 1745 to 1762, the Journal de Trévoux was directed by the Jesuit priest Guillaume-François Berthier (1704-1782), who was replaced in April 1762 (following the Jesuit expulsion) by a succession of editors.

For the purpose of the corpus and network visualisations, all articles not attributed to a precise author (or authors) are classified simply as being by the journal. For example, the review article ‘Lettres sur l’éducation, par M. Pesselier’, in the Mercure de France of October 1762, is classified as being by ‘Mercure de France’. In cases where a journal article is signed, I record both the individual author and periodical title in the corpus. However, for the purpose of the network visualisations (in which nodes are coloured according to actor type), the individual and not the periodical is retained as the author.

Unpublished Works

There are two unpublished texts in the corpus, which nevertheless deserve a place in the Querelle. The first is Barletti de Saint-Paul’s Institutions nécessaires, discussed earlier, which gave rise to polemical Querelle articles and pamphlets in 1764 and 1765.^[5] The second is the ‘Lettre circulaire de MM. les agens-généraux du clergé de France’, dated 8 November 1780, which enquired about the state of the collèges held by the Gallican Church.^[6] Both the Lettre and Barletti’s Institutions are publicised – and parts of them are published – by works in the dispute. Insofar as these works are thus crucial points of reference, and catalysts for further Querelle texts, they have a place in the corpus.

Order and Dating

The corpus lists Querelle texts in approximately chronological order of appearance, so as to highlight features such as temporally tight dialogical chains of ‘text’ and ‘response’. This corpus should be taken neither as complete nor as authoritative, but rather as a working corpus that indicates the scope, scale and range of actors of the Querelle. I hope future scholars may be able to add to or amend it, [p.259] particularly with regards to the dating of texts, which often proved difficult due to missing information. Where a date is given in square brackets, a letter identifies what that date corresponds to:

Works that I have been unable to date any more precisely than by year are listed alphabetically at the end of each year. The order in which texts are listed matches the order in which they appear in the dynamic network visualisation of the Querelle. The only exception is for the texts first published between 1750 and 1761, and later drawn into the Querelle. In the corpus these are listed chronologically by date of publication. However, in the dynamic network visualisations, they are only implicated in the Querelle when they are first referred to by a text in the Querelle proper: in other words, they appear when they are pulled into the Querelle.

Database

The static and dynamic network visualisations of the Querelle are based on a digital database of the corpus. Besides the basic bibliographical information in the corpus, the database records additional information about each text, and about the other Querelle texts to which it refers. This information is gleaned through close reading of each work. The database thus enables the Querelle to be analysed quantitatively, but this quantitative data is itself derived from detailed qualitative reading. [p.260]

The database records each time one Querelle actor refers to another. It also categorises the nature of this reference, by assigning to it a number. The criteria for categorisation are described below:

[p.261] When describing the Querelle in the preceding chapters, unless otherwise stated, ‘negative reference’ refers to any reference classed as -2 or -1, while ‘positive reference’ includes those classified as 2 or 1.

It is possible to consider the Querelle either as a network of people (who sometimes wrote more than one Querelle text, or who refer to one another without specifying a particular text) or as a network of texts, which are themselves the actors of the debate. The database records information about both people and texts, so that the Querelle can be analysed from either perspective.

Network Visualisations

Using the statistical computing software R and the statnet suite of packages, specifically networkDynamic and ndtv, social network scholar Marc Sarazin and I used the quantitative information in the database of the Querelle to produce static and dynamic network visualisations of the references made between actors in the debate.^[7] Some of these visualisations represent the Querelle as a network of texts, while others display it as a network of people. All but two of the texts in the Querelle are published. ‘People’, meanwhile, includes individual authors, as well as institutions or collectives publishing under one name (for example, a journal such as the Mercure de France, an académie or a specific regional parlement. In both types of network visualisation, actors (‘nodes’) are represented as coloured circles:

In the dynamic visualisations, it is possible to click on each node to view its short label, even while the visualisation is playing. Once selected, the node is highlighted, allowing users to ‘track’ it over the course of the Querelle. The references made between actors (in network terms, the ‘ties’ or ‘edges’ between nodes) are directed; in other words, they are displayed as lines with arrows at the end, which point from Actor A (referring actor) to Actor B (actor referred to). These arrows [p.262] are colour-coded to represent the nature of the reference, as listed in the table above.

Unless stated otherwise, node size is weighted according to the number of ties that the node sends and receives. Put simply, nodes grow when an actor makes or receives references. The rate of this growth is determined according to a logarithmic function, meaning that nodes grow at a marginally diminishing rate with each tie sent or received.^[8] This prevents nodes that receive many ties (for example, Rousseau’s Emile) from occupying inordinately large areas of the visualisation, and conversely it prevents nodes that receive very few ties from being so small as to be almost invisible. This logarithmic function means, however, that node size should not be interpreted as proportional to the number of ties sent or received.

The layout of all but one visualisation (in other words, the spatial positioning of each node vis-à-vis other nodes) is computed using the Kamada-Kawai algorithm (see Figures 3, 4, 6 and 7).^[9] Briefly, this widely used algorithm attempts to minimise the distance between nodes that are either directly connected or connected through other nodes. The closer nodes are in terms of connections (for instance, if they are directly connected as opposed to being connected through three degrees of separation), the closer the algorithm places them in the visualisation. This algorithm struggled to clearly display the static network of texts, due to the presence of texts (i.e. nodes) that were not connected to the main component of the network.^[10] These outlying nodes made the nodes in the main component bunch very closely together. This particular visualisation (Figure 5) therefore employed the Fruchterman-Reingold algorithm, which similarly places nodes that are connected close to each other but, unlike the Kamada-Kawai [p.263] algorithm, also repulses nodes from each other to give a less ‘bunched’ appearance.^[11]

All visualisations omit isolates (texts or people that neither make nor receive references), so as to improve the visibility of the structure of the main network. These texts or people are nevertheless part of the corpus of the Querelle, and are included in any statistics on the Querelle. In the visualisations of texts, 95 per cent of all those in the Querelle are represented (there are just 12 isolates omitted, of a total of 218 texts). In the visualisations of people, 92 per cent of all those in the Querelle are represented (10 of 126 people are isolates).

As well as omitting isolates, the dynamic visualisation of the network of texts also omits minor components of the network, again to facilitate visibility of the way in which the main component evolves over time. These minor components are small, and together comprise sixteen texts. In total, then, also taking into account the 12 isolates omitted, the dynamic visualisation of the network of texts shows 190 Querelle texts of the full 218.