A Case Study in Analysing Digitised Archive Data: Authors at the Comédie-Italienne, 1760–70 1

The use of ‘big data’ in the humanities is becoming increasingly prevalent, as researchers try to exploit the capacities of modern computing to provide new perspectives on and access to more traditional forms of documentation. In this vein, two major projects currently seek to digitise and analyse the financial records of the major Parisian theatres in the eighteenth century, creating searchable databases that promise to bring the information to worldwide audiences and allow entirely new types of analysis. This article uses a small-scale study of a subset of the Comedie-Italienne’s administrative records in the period 1760-70 to explore the potential benefits and challenges of such enterprises. It makes new data on audience, takings and authorial payments for this decade freely available to readers, and examines the process of capturing and rationalising data drawn from early modern records. To demonstrate the potential of the larger projects, this subset of data is then used both to reconsider the commercial status of the Comedie-Italienne with respect to its French counterpart, and to provide the first account of the finances of the authors who wrote for the Italian troupe. Finally, this article argues that despite the many and varied benefits of digitisation and computational manipulation of archive data, in order for such projects to make a serious contribution to scholarship, it is vital that the material, contingent, human element of original records is not lost.

records of the theatre, capturing and analysing a large proportion of the financial data for this decade. My aim was to use these relatively underexplored records to provide sorely needed concrete context for contem porary accounts of dramatic authorship, extracting information on audiences, theatrical takings, salaries, career patterns, genre trends and payments to build a picture of the authorial experience at one of the major theatres of the period. In the absence of a large research team, the sophistication of the techniques I could use was necessarily limited; however, the process of curating a small dataset and using it to inform a broader project led me to encounter in microcosm many of the problems and questions that must be tackled by the larger schemes. This article uses this case study to explore the potential of the vast historico-theatrical enterprises outlined above, to consider the challenges and possible pitfalls of such undertakings in the digital humanities, and to demonstrate how even a fraction of the information in the registers, combined with modern computing, can transform our understanding of eighteenth-century French theatre.

Studying Early Modern Theatre History
Scholarly interest in the administrative aspects of early modern French theatre is not new. The socio-historical dimension of theatrical life in late seventeenth-and eighteenth-century Paris has been widely studied both in its own right, and for the light it can shed on contemporary literary production. In the 1940s and '50s, Henry Carrington Lancaster published two volumes of data on audiences, takings and actor/author payment drawn from the 1680-1774 Comédie-Française registers, which have remained the standard reference ever since. Clarence Brenner performed a similar service for the Comédie-Italienne a decade later, although his transcriptions of the Italian registers from 1716 to 1793 included only audiences and takings. Whilst theatre and literary historians in the second half of the twentieth century relied on these two sources (Alasseur; Lagrave; Lough; Rougemont), the turn of the millennium brought with it a new generation of researchers who returned to the administrative documents of the two theatres in order to re-evaluate the institutional context in which actors, audiences and authors moved in eighteenth-century Paris. Gregory Brown on authors at the Comédie-Française (A Field of Honor; 'Literary Sociability'), Mark Darlow and Solveig Serre on the administration of the Opéra, Jeffrey Ravel on theatre audiences, David Trott on unofficial theatres, Françoise Rubellin and Isabelle Martin on the Foire and Andrea Fabiano, Emanuele de Luca and Silvia Spanu on the repertoire of the Comédie-Italienne have contributed to creating a panorama of the contemporary context that brings new life both to our understanding of 'the business of theater' (as Clay's 2013 study would have it), and to readings of the texts produced in the period. 3 Two of these scholars were part of the team that first recognised the potential of combining this renewed interest in theatrical administration with the growing discipline of digital humanities. From 2000, Ravel and Trott worked with Barry Russell to produce the Calendrier électronique des spectacles sous l'ancien régime et sous la révolution (CESAR), initially funded by the Arts and Humanities Research Council (AHRC). The project integrated personal and institutional databases created by researchers across the world, along with digitised versions of contemporary reference books, treatises and police reports. The resulting site allows users to search for play titles, authors, troupes, locations, dates and publications, and thus track both individuals and their outputs across the eighteenth century. Whilst this is an invaluable resource, the project -which has only been sporadically updated since the bulk of the work was completed in 2008 -does not allow for any more sophisticated output than individual search results, nor does it currently have the capacity to integrate different sorts of data, such as the financial elements of theatre administration. In the context of an academic community that was developing increasingly ambitious and sophisticated tools in the digital humanities, this unrealised potential encouraged Ravel to launch a new project to digitise the available registers of the Comédie-Française far more comprehensively.
In fact, with over a century of administrative records available (from the theatre's creation in 1680 to its reconfiguration under the Revolution), these registers provide a perfect example of the sort of raw 'big data' that could benefit from digitising, and from the resulting ability to spot patterns and themes that are not visible on an 'analogue' reading (Smith). In the context of the Comédie-Française, this might mean being able to map the types of ticket sold against specific titles, study the effect of political or historical events on the popularity of certain genres or authors or on general ticket sales, or consider the financial success of new works as compared with the traditional repertoire of Molière, Racine and Corneille (Lipshin et al). The variety of possible data makes a digital version of the registers a useful tool for researchers in a range of disciplines -history, sociology, literature, and combinations of the above -who could all use different elements of the information in different ways. Such a broadly attractive project seems doubly alluring given that it depends on the digitisation of financial records: an apparently regular, consistent form of document, which should lend itself well to the logical tagging of individual components on which digitisation relies (Schöch).
My work on the Comédie-Italienne registers was motivated by a similar sense of the potential of relatively large-scale electronic data to add to our understanding of theatre history. Despite the recent renewal of interest in archival studies of Ancien Régime theatre, and particularly of the hithertoneglected Comédie-Italienne, its administrative workings have remained relatively under-studied in comparison with those of its more famous French rival, with its registers largely mined only for their performance information. 4 Moreover, whilst Brown's important studies have brought to light a wealth of information on dramatic authors at the Comédie-Française (A Field of Honor; 'Literary Sociability'), the status of Comédie-Italienne authors has attracted far less interest, despite the overlap in these two constituencies that makes fully understanding one without the other nigh on impossible. I set out to redress the balance both by gathering administrative data that would allow more equitable comparisons with the Comédie-Française, and -more specifically -by examining in closer detail the financial status of authors at the Italian theatre, in order to integrate this understanding into existing narratives of dramatic authorship in eighteenth-century Paris.
This pilot study, composed of a small dataset digitised for very specific purposes, does not by any means represent the capacities of the fully searchable and manipulable databases complete with visualisations that are envisaged by the two major projects. 5 Nonetheless, it can provide a glimpse of the potential advantages and problems of their proposed complete digitisation of a larger dataset. Lisa Spiro suggests that most digital humanities projects remain in a 'permanent beta phase', constantly responding to the findings of their users, whilst Daniel Cohen refers to the process of 'iterating towards perfection'. This study, then, is one such iteration, located somewhere between Brenner's pen-and-paper calculations and the projected online database: an examination of the process of compiling and using such a dataset which can provide useful observations towards the development of the larger projects, as well as a contribution, through the publication of elements of the data themselves, to the growing body of online information 4 Lagrave uses Brenner's data to compare audiences in the two Comédies up to 1750, but the only recent examination of audience data for the latter part of the century was conducted by Paola Ranzini, whose preliminary article 'Il pubblico parigino di Carlo Goldoni' sets out the potential of using such data to understand Goldoni's audiences in Paris. 5 The CFRP will not include information on expenditure, and therefore some of the calculations on authorial payments I make later in this article will not be possible for the French troupe even when the project is completed.
relating to the Comédie-Italienne, even as the central digitisation project is still in development. 6 The Administration of the Comédie-Italienne Three factors might account for the lack of studies on the Comédie-Italienne and its authors as compared with their French counterparts. The first is the sense of a hierarchy among the Parisian theatres that was present even in the eighteenth century. The Comédie-Française was linked to the royal court, performed the noble genres of tragedy and high comedy and was the guardian of the cultural heritage of the plays of Racine, Molière and Corneille. The Opéra held a similarly high status. The Comédie-Italienne, on the other hand, whilst also patronised by the king, was associated with farce and frivolity, and the reputedly low morals of the itinerant foreign troupes with whom it had originated. Though the seventeenth-century Italian troupe performed at court as frequently as its Italian counterpart, its banishment in the final decade of the century -for a play that allegedly attacked the king's secret wife -seems to have reinforced its more lowly position in the following century. Modern scholarship, particularly outside the Italian-speaking world, has until relatively recently reproduced this hierarchy, focusing the majority of its attention on the supposedly more 'prestigious' theatres, and tending to dismiss the merely 'popular' Italian troupe and its position in contemporary cultural life (Mangini,(35)(36)(37)(38)(39)(40)(41)(42)(43)Lagrave,Jomaron,. 7 Secondly, the Comédie-Italienne began life in 1664 as a typical commedia dell'arte troupe, producing performances based on skeleton outlines of plots, known as canevas, around which the actors would improvise according to the traditions of their mask-based characters (Fitzpatrick). The traces left of these productions are necessarily sparse, so there is little capacity for textual study of this output. 8 Moreover, the compositional traditions of this type of theatre 6 See for example 'Les Savoirs des acteurs italiens', the work of a group based at the Institut de recherche en musicologie (IREMUS). 7 Virginia Scott's work on the seventeenth-century Comédie-Italienne is among the first serious attempts to address the Parisian troupe. Since the 1990s, the work of (among others) Andrea Gherardi), whilst in 1762 its amalgamation with the musical theatre known as the Opéra-Comique brought not only authors but also composers to its theatre at the Hôtel de Bourgogne. However, the Comédie-Italienne is still often thought of as an authorless troupe, and has thus never been the subject of a study on authorship. And finally, unlike the Comédie-Française, the Comédie-Italienne does not have a dedicated archive; instead, material is spread between the Archives Nationales and the Bibliothèque de l'Opéra. Poor organisation on the part of the troupe at the time, compounded by the losses suffered in the 1887 fire at the Opéra, means information is limited, partial and sometimes mathematically inaccurate; all reasons to explain a lack of studies dealing with the administration of this troupe. The records that do remain are a series of documents in the collection relating to the Maison du Roi in the Archives Nationales, 9 and the TH/OC series at the Bibliothèque-Musée de l'Opéra, which predominantly contains the troupe's registers. 10 The information in this article is based on the 1760-70 registers, from which I extracted the basic financial and attendance data, repertoire and actor lists, and information on authorial payments. This document is the largest available source of information on the finances of the troupe. Each day, the managing actors (semainiers) and Linguet the accountant recorded the plays performed (usually between two and four a day), the actors involved, the tickets sold, total takings and some small daily costs (contributions to the church, candles, guards and dancers). In addition, at the end of each month, a two-to six-page spread gives monthly totals and records in more detail both regular payments such as actors' salaries and the quart des pauvres (a charitable contribution levied on all theatres), and expenditure on materials, publicity, musicians and authors. A further account at the end of each year largely reiterates the same information, and provides a final figure showing the overall profit or loss. The remainder of this article takes the information gathered from these documents in two sections: first, the daily performance Italian theatre in the 1760s, whilst Spada and Gambelli performed a similar service for the seventeenth century. 9 ANF.O 1 .848-54. Some of these documents are reproduced by Campardon, although often with incorrect codes for the modern system. 10 These are the records transcribed by Brenner (The Théâtre-Italien), however not only are his transcriptions partial, but there are also a number of errors, particularly in his indications of which plays are respectively French or Italian, and in his marking of the premieres of certain titles. See below for further discussion of these errors. BO.TH/ OC.178 is transcribed by Spanu in 'Le Répertoire et la dramaturgie'. data that allow me to situate the theatre in terms of takings and attendance with respect to its French rival, and secondly, the records of author payments that give a glimpse of how the individuals who wrote for the Italian troupe might have integrated the theatre into their complex careers. For each section, I examine the process of collecting, recording and analysing the relevant data, before giving examples of this data in use.

Audiences and Takings
Recording the daily performance data is relatively straightforward, given its regular presentation. A single record per day can capture the information according to its original divisions on the page. 11 On some days there are odd extra notes, recording a box hired by an aristocrat or an advance payment made. These are largely irrelevant to my calculations, though for a more complete understanding of the running of the troupe they would have to be tagged in detail. The most complex element of this data capture process is correctly identifying play titles. They are generally given in abbreviated form, some of which could relate to either a play or a homonymous ballet, and some of which -since titles are often reused over time -could refer to more than one version of a play. The short period of my study largely avoids any confusion on the latter count, since repeated titles within a few months of one another can generally be assumed to refer to the same version of a play, and the premiere dates of different versions of a title can act as confirmation of this fact. The potential confusion between ballets and plays of the same title is also relatively straightforward to resolve: the registers usually note 'ballet', and the corresponding play and ballet usually appear on a programme together. Where this is not the case, the ordering of titles acts as a further clue: where ballets or divertissements are performed (and the two terms seem to be used interchangeably), they are always the final title listed. 12 My primary aim in transcribing the numerical components of 3,160 days' worth of this data was to analyse the popularity and financial success of the Comédie-Italienne with respect to its supposedly much more prestigious and successful French counterpart, particularly through an analysis of daily takings (recettes) and attendance. The data for recettes were the least complex element to record and manipulate. In the original records, ticket sales are divided according to ticket type, with one figure recording the number of tickets sold in a particular category, and a second figure giving the total takings this represents (see fig. A). The auditorium of the Hôtel de Bourgogne was split into three levels of balconies (premières, secondes and troisièmes) and the parterre (standing places). Individual seats purchased for the balconies were either in serried ranks or in loges (boxes): the latter could also be hired wholesale for an evening or even a whole season (BO.TH/OC.1). For the majority of the decade, the registers list total individual places sold in the premières, secondes, troisièmes and parterre, plus the total number of loges hired outright for that evening -in one season, loges are further broken down by location and capacity. Most days give the total daily recette, a figure that is also listed in the end-of-month accounts. 13 Two different pricing schemes were in operation. Whilst for the majority of the time the pricing structure was almost identical to that applied at the Comédie-Française, on Tuesdays and Fridays, known as the petits jours, the Comédie-Italienne troupe was forbidden from performing music, and instead put on traditional Italian commedia titles, to which a lower price was applied (fig. B). 14 The accuracy of the transcribed takings figures can be verified with a number of checks that digital recording renders straightforward, and which could be automatically built into any more complex system. Multiplying the number of tickets sold in a category by the price of that ticket type confirms the total takings achieved in each category, 15 whilst the sum of all recorded takings in different categories can be compared with the transcribed overall total, which is also repeated in the month-end accounts. There are one or two occasions on which the sum of the different components does not match the given total, implying an error of calculation or recording by the original compilers of the register. 16 Such minor inaccuracies are an inevitable feature 13 All figures to the nearest livre. The original data is given in livres, sols and deniers (20 sols = 1 livre; 12 deniers = 1 sol). Calculations were made by converting payments into decimals, then back again (rounding where necessary) once the operation had been completed. 14 Tuesday and Friday were not only the most popular days for the Comédie-Française, but also the performance days of the Opéra, who had permitted the Italians to perform comic opera only on the basis that they would not attempt to compete. 15 For the parterre tickets no such multiplication is possible, as they are priced at just 1 livre each and no separate number of tickets sold is provided. Similarly, for the loges, for which a variety of price categories were available, the inclusion of all these categories within a single total prevents this form of verification. 16 I am confident in my own transcriptions as, having identified all inconsistencies between the two totals, I returned to the original documents and double-checked possibly anomalous figures. Where errors on the part of the compilers were clear and verifiable (e.g. where two figures in the 'total' had been transposed) I corrected the data. In general, where the difference between my added total and the recorded total of accounts produced by hand (Poirson,26), and will have to be accounted for in the larger digitisation project; however, the difference is usually so small (one or two livres) that it makes no significant difference to monthly or annual takings calculated by adding these daily totals. 17 It is more complex to achieve accurate audience figures. The free tickets handed out to authors and actors as partial payment for their work were rarely recorded, and loges hired for a full season do not appear in daily accounts; thus, the presence of a certain proportion of audience members cannot be quantified. The only figures that can be calculated with any accuracy relate to tickets bought daily on the door. The numbers of all individual and parterre tickets sold can be straightforwardly added to give the number of audience members present in these areas. However, the daily loges hire figures, representing multiple people within one price unit, complicate matters, as do the suppléments that appear from time to time. These elements present the challenge of extrapolating some form of audience figures from composite data, a challenge that will also be faced by the CETHEFI project if its database is to allow audience analysis on the level of individual attendees.
My system for extrapolating audience numbers relies on educated approximation. For the single season in which the number of each type of loges hired is specified (1763-64), I assume the maximum occupancy: Premières loges (48l) = 8 places Premières loges (24l) = 4 places Secondes loges (24l) = 8 places Secondes loges (12l) = 4 places Troisièmes loges (16l) = 8 places Petites loges (18l) = 3 places Loges grillées (36l) = 8 places For the remaining nine seasons, where the types of loges are not specified, was less than 5, I used my total, but for the handful of larger errors that were clearly not miswrites (for example a discrepancy of 104 on 17 Aug 1761) I used the original totals, assuming some ticket income was simply not recorded. Brenner's figures are also revealed to be fairly accurate. In one or two examples, he has clearly either copied the figures for the previous day, or transposed two digits. He also does not take .5 livres into account in totals, and relies on the totals as recorded, meaning that any error in the original calculation is preserved in his transcription (The Théâtre-Italien). I have approximated based on the rough average cost of a single loge place. This average is clearly imprecise, since some types of loge are hired far more frequently than others, but it is the only way of reaching any rough figure for the number of occupants with the available information. The average cost per person calculated in this manner is: 48/8 + 24/4 +24/8 +12/4 +16/8 +18/3 +36/8 = 4.57 7 In order to simplify the calculation, and since it is unlikely that all loges were full at all times, I have taken the figure of 5l per person in a loge. Therefore the takings figures for loges where no type is specified have been divided by five to obtain a rough guess at the number of occupants. 18 A similar problem is presented by figures listed as suppléments. This phrase appears sporadically in the daily accounts, and appears to refer to extra attendances for some reason not recorded in the standard totals. Most occurrences simply read 'suppléments = x livres', and thus provide no information about the type of ticket or the numbers of audience members concerned. However, a series of entries in January 1764 give suppléments in the following format: '38 du théâtre = 228, 9 du parterre', '14 du théâtre = 84, 13 du parterre', '50 du théâtre = 300, 2 du parterre'. Here is it clear that 'suppléments du théâtre' -by far the highest proportion of listed supplément totals -are charged at 6l each, whilst the parterre suppléments are, like other parterre tickets, charged at 1l each. Since a high proportion of the supplément figures for which no further information is given are divisible by six, I have used this divisor to estimate how many audience members these figures represent. 19 Though these approximations become less significant on the large, decade-18 Using this approximation results in loges figures that range from 4.2%-6% of audience totals across the decade (excluding the 1760-61 season when loges figures are given on just one day), with a slight trend upwards over time. For the one season for which more accurate loges attendance figures are given (1763-64), they constitute 5.3% of the total: my approximations therefore appear to be satisfactory. 19 Brenner's system for dealing with loges and suppléments is difficult to discern. He does not explain where his audience figures come from, and any attempt to discover a constant divisor for loges is impossible. For example, for the first day of the 1766-67 season, his audience figure is 506. Adding the figures given in the registers for this day gives 468 individual tickets + 36l in suppléments + 118l in loges. If he had adopted my approach to suppléments this would give 474 + 118l in loges, meaning that he had extrapolated an attendance of 32 in the loges: an average of 3.6l per person. Yet for the following day, when he gives a final audience figure of 1,139, the total before the loges calculation is 1,031, meaning he assumed 108 loges attendees paying a total of 465l: an average of 4.3l each. The third day average calculated in this way is 6.8l (The Théâtre-Italien, 292). It may be that Brenner had access to some more detailed description of loges attendance of which I am unaware, but in the absence of such a document, I prefer to use my own figures, in the knowledge of their inaccuracies or approximations.
long scale at which I am generally working, it is important to keep in mind that audience figures quoted in the remainder of the article are informed best guesses, rather than definitive sums. A sample completed spreadsheet for the 1763-64 season is available. 20

Using the Data
Traditional accounts of eighteenth-century theatre painted the Comédie-Italienne as symbolically inferior to its French counterpart, an image that affected both how contemporary authors interacted with the troupe and how modern critics responded to its participants and outputs. However, even the relatively basic data outlined above provide ample evidence to demonstrate that symbolic and commercial success were very different things.
Beginning with annual audience figures, with Comédie-Française figures drawn from the work of Lough based on Lancaster, we find that on-the-door audience numbers for the two Comédies follow a very similar three-year cycle of highs and lows across the decade, with peaks in the 1763-64 and 1766-67 seasons and low points in 1765-66 and 1768-69 ( fig. C). The theatres were clearly broadly affected by the same trends in theatre attendance. Moreover, though both troupes saw their attendance generally increase over the century, the Comédie-Française attracted fewer spectators on the door for the majority of the period. 21 Even if we account for a higher number of performances by the Italian troupe (3,160 across the decade, compared with 2,920 by the French troupe) we find that their average attendance is still marginally higher than that of the French (585 people per performance for the Comédie-Italienne, 568 for the Comédie-Française). Most significantly, there is a meteoric rise in interest at the Comédie-Italienne over the first few years of the decade, in comparison with a virtually unmoving figure for the French troupe. This corresponds with a state-sponsored project to renew the Italian theatre through a new repertoire, an amalgamation with the Opéra-Comique and the employment of the Italian dramatist Carlo Goldoni: a project that obviously had a measure of success (Fabiano,Goodman).
The annual recettes on the door for these two theatres across the 1760s 20 https://docs.google.com/spreadsheets/d/1Z8JPyT4VVO7XJxEJspVQg6muJ3ieByiTs4efHRi jock/edit?pli=1#gid=0 21 Audience figures for loges subscriptions are unavailable for either theatre, but it is known that at the Comédie-Française they grew in importance across the decade until they made up 30% of takings; this may well account for the lower sales on the door at the Comédie-Française. However, see fig. D for a comparison between Comédie-Italienne recettes and Comédie-Française recettes including loges subscriptions, which gives an indication of the parity between the theatres even when Comédie-Française loges are included. unsurprisingly follow the same cyclical pattern as the audience data ( fig. D).
And as with the audience figures, the Comédie-Italienne totals are consistently the largest from 1761-62 onwards; an even more impressive achievement considering that lower prices were imposed two days a week at the Hôtel de Bourgogne. Based on these figures, the Italians earn an average 71l per performance more than the French troupe across the decade. The Comédie-Française seasonal loges hire figures, for which there is no equivalent data in the Italian registers, add an average of 34,011l a year to the French troupe's totals. However, even taking these figures into account, the Italians' annual recette still surpasses that of the French on three occasions across the decade, and comes very close in a fourth season ( fig. D). 22 By comparison, figures for Opéra takings are not only consistently lower, but do not peak and trough in the same years, implying that the two Comédies had a much closer relationship in terms of their audience and popularity ( fig. E).
Perhaps most revealingly, the average amount paid per audience member (on the door) is almost identical for the two theatres, and remains consistent in all season-by-season averages, at 2.2l for the Comédie-Française and 2.3l for the Comédie-Italienne ( fig. F). Given the almost identical pricing structure in the two theatres, these figures imply that the proportion of different ticket types sold by the Comédies-Française and Italienne was more or less the same. 23 In other words, the same types of people frequented both. 24 That the theatres shared an audience is not all that surprising: the mid-eighteenth century had seen an explosion in the popularity of parody, a genre that supposes a shared audience, since it requires at least some knowledge of the original (Luca, 'La Circulation'). Lagrave had come to a similar conclusion for the first half of the century based purely on audience figures (205), and Ravel has exposed the social similarities of the parterre in the period. More signifi-22 Since we do not have any consistent records of subscriptions to Comédie-Italienne loges, we cannot know how much more the Italians earned from this source; however, the ad hoc recording of a handful of hired loges each month implies that they were less significant in the Italian theatre. 23 These low averages indicate that a high proportion of tickets were sold to the parterre. This is particularly true for the Comédie-Française who, unlike the Italians, did not have lower-priced days to bring down their average. 24 It is difficult to predict how loges subscriptions would have affected this average, since they are so rarely recorded for the Comédie-Italienne. Given that the Italians had lowerpriced seats on two days of the week, the fact that their average matches the French figure suggests that they sold more higher-priced tickets than the French, a supposition that would tally with the fact that hired loges at the Italian theatre were more frequently paid on the door, at least in the early part of the decade. If the Italians were selling more expensive tickets on the door, then even adding the French loges subscriptions into the calculation would still allow us to state that a similar number of similarly priced tickets were sold in both contexts.
cant, and not explored by earlier accounts, is the parallel level of investment: laughter, sentimental tears and tragic fear, it appears, are worth the same. It is also possible to conduct revealing comparisons between the two theatres on a more microscopic level. Though such comparisons do not necessarily require digitised data, when made in conjunction with the larger scale observations outlined above they become more significant. There is far more of a fluctuation in day-to-day totals than in the overall yearly figures. The opening and closing days of each season demonstrate just how closely matched the two troupes were. Some seasons open with a storming success for the French troupe (2 April 1761 saw 3,750l of tickets sold at the Comédie-Française, and only 959l taken at the Hôtel de Bourgogne); 25 sometimes the reverse is true (the 1765-66 season closed with 2,874l for the French and 3,472l for the Italians); 26 and more often than not the figures are relatively equal (notably the close of the 1768-69 season, where the troupes took 3,103l and 3,102l respectively). 27 We could make specific comparisons for the premieres of popular pieces, the first or last performances of famous actors, and even political events, in order to map how theatre attendance was affected by specific events. But even the handful of instances identified here confirm that though each theatre had the capacity to outdo the other, neither consistently succeeded in doing so. The largest daily recette either of the two troupes could expect was on a similar scale; they really were competitors, just as they had been earlier in the century (Lagrave,. These comparisons, when taken together, provide concrete evidence to reposition the Comédie-Italienne as an important player in the cultural life of 1760s Paris. Though it may not have had the same symbolic capital as its French cousin, by all quantifiable measures it occupied the same commercial space. It was certainly not the inferior institution that the lack of interest accorded to it by some earlier studies might suggest. Contemporary accounts of the popularity of the Italian troupe -predominantly from individuals associated with the Comédie-Française, who bemoan the audience's lack of taste (Lagrave,(197)(198)(199)(200)(201)(202)(203) -are symptomatic of a much deeper condition of commercial parity, which continued and strengthened even as the Comédie-Française increased its audiences from 1750 onwards (Lough,185). This is significant for readings of the contemporary theatrical field as a whole, but it also creates vital context for my study of authorship, proving -for example 25  -that writing for the Comédie-Italienne of the 1760s provided authors with access to audiences that were as large and socially varied as those at the French theatre. In the context of the burgeoning literary and theatrical marketplace of the late eighteenth century (Turnovsky), such considerations were increasingly important.
There are many more questions that could be asked of these data, which might make similarly illuminating contributions to different studies in different disciplines, and which would benefit even further from their digitised form. We could tag the days on which different pricing structures were applied (not always corresponding to the standard Tuesday/Friday pattern), and analyse how audience make-up altered on these days. This would show, for example, that in the 1763-64 season, on the twenty-five Tuesdays and Fridays where the higher, standard prices were exceptionally imposed, not only did the average ticket price match the annual average at 2.3l per person (as compared with the petits jours average of 1.96l), but the audience was on average nearly twice the size of a normal petits jours crowd, despite fourteen of these occasions being Italian-only programmes. An interested sociologist of the theatre might postulate, perhaps, that the troupe was particularly good at identifying the plays for which raised prices would not put off a crowd, or that audiences were persuaded to change their normal attendance days by pricing changes that implied an increase in quality.
In a different study we could track the success of titles across time, noting perhaps that the Ballet des lavandières, which premiered on the first day of the 1763-64 season, was intensively performed until the start of June (seventeen performances in less than two months), then dropped to an average of two or three performances per month, experiencing small peaks in December and at the very end of the season, before disappearing from the repertoire for the remainder of the decade. This pattern forms a strong contrast to that for the comic opera Le Bûcheron, which, following an initial period of intensive performance, dropped to a relatively regular but gradually decreasing number of performances per season (see fig. G). A literary scholar might track other plays and ballets in this way to analyse if these were standard patterns for these genres, or affected by more contingent external influences; they may also attempt to calculate the relative worth of different types of performance (French, Italian, musical) on the hybrid programmes. 28 Moving the focus away from repertoire, an historian may wish to consider how money spent on candles related to the length of specific plays, the complexity of the décor, or the month of performance (Hawcroft; Clarke, 'L'Éclairage'), to search for correlations between the amount paid to guards and the number of spectators in the rowdier sections of the theatre, or to analyse the payment of the charitable quart des pauvres or the organisation of performances given entirely for the benefit of one or several members of the troupe. Whilst some of these calculations would be feasible even without digitised data, the accuracy and speed of calculating totals and averages, as well as searching across large datasets, is greatly enhanced by holding the information electronically, and the larger projects will provide even greater possibilities.

Authorial Payment Records
For all their slight inaccuracies, partly illegible writing and anomalies, the daily record pages are in fact by far the most regular part of the registers. The question of how precisely to capture and present data becomes far more complex when we enter into the murkier waters of the month-end and year-end accounts. These sprawling pages of figures contain recurring notes on routine payments for lighting, posters, musicians and so forth. They also list the daily totals, the monthly actor salaries, and total income and expenditure. This more routine accounting (which is nonetheless presented in a fairly haphazard fashion) is accompanied by a numbered list of one-off payments such as advance payments to actors and authors; specific costumes; scrawled notes on loges hired or authorial payment rights conceded. This mine of information is incredibly complex to systematize in any fashion, and will be the most challenging element of the digitisation project. For the purposes of my study it was the authorial payment data that were of particular interest: I wanted to investigate who was paid, how much, how frequently, and what for, in order to analyse the role the Comédie-Italienne might play in an authorial career.
The authors paid by the Comédie-Italienne were not the actors who composed the brief, Italian scenarios that formed their traditional output. Rather, they were the authors of a handful of text-only French plays, and of the comic operas imported into the theatre when it amalgamated with the Opéra-Comique in 1762. The latter also required the presence of a composer, who is termed an 'auteur de la musique' by the registers. The records of payments made to authors are generally listed in the routine payments section of the month-end accounts ( fig. H). 29 Each entry refers to a different 29 The Comédie-Française kept meticulously detailed records of authorial contracts and the payments made for each run of plays. At the Comédie-Italienne, no such documents have yet been uncovered. Earlier in the century, when author payments were less frequent, they were more commonly noted on the day a play was performed -this is play, and individual authors are rarely named, with the formulae 'à l'auteur de', 'à celui de', 'aux auteurs de' or 'à ceux de' being preferred. Often, the authors of words and music are listed separately, and in such situations they are consistently paid the same sum, thus apparently enjoying equal status. When the plural is employed, we can assume it refers to these two sorts of 'author'. On occasion, one specifically appears without the other: this is usually indicative of a case when one of the two has given up his rights to ongoing payment in return for a one-off sum, which itself may or may not also be recorded. The amount paid to one or both authors is expressed either based on a flat rate per performance (16l 13s 4d, 33l 6s 8d or 66l 13s 4d) or as a fraction (1/9, 1/12 or 1/18). In almost all cases, the final sum paid for each play that month is recorded. Thus, for example, the entries for August 1762 read: aux auteurs des paroles et de la musique de Sancho pour treize représentations de ce mois le douzième ci: 655l 5d à l'auteur des paroles de Soeurs rivales pour onze représentations à 16l 13s 4d chaque ci: 183l 6s 8d idem à celui de la musique de la susdite pièce ci: 183l 6s 8d à l'auteur de la Jeune Grecque pour une représentation donnée dans ce mois, le neuvième ci: 43l 6d à l'auteur d'Annette et Lubin pour une représentation donnée dans ce mois le dix-huitième ci: 16l 3s 8d à l'auteur du Gondolier vénitien pour la musique, une représentation à 9l. (BO.TH/OC.44) My pilot database compiles this information on authorial payments from 1760 to 1770 (available below). 30 The 725 entries record direct payments for plays; twenty-four payments of other types (advances or one-off sums in return for ceding the rights to a play) have been excluded from my overall calculations, since the shorthand nature of the records means it is impossible to tell how and when advances have been repaid, or how they relate to regular payments. Furthermore, such payments can also be recorded in areas of the registers that I did not systematically transcribe. In any case, since similar records appear for actors, musicians and other personnel, the payment of advances is clearly not specifically related to the status of author. Therefore, the figures I quote pertain only to the routine part d'auteur or droits payments (both terms used in the period). Whilst the clarity of the final spreadsheet might give the impression of regularity, it is important to acknowledge the limitations that result from the complexities and ambiguities of the raw information. The the case for payments made to Marivaux in the 1730s. 30 https://docs.google.com/spreadsheets/d/1Z8JPyT4VVO7XJxEJspVQg6muJ3ieByiTs4efHRi jock/edit?pli=1#gid=0 decisions I made in both recording and analysing the data had to take account of these limitations, and my attempts to render the data usable in spite of unavoidable approximations can provide a useful model for how any digitisation project might proceed in its analysis of all elements of these complex accounts.
The first, almost insoluble problem is that of the errors or inconsistencies already present in the registers. Sometimes, the number of performances of a specific title recorded as paid does not correspond to the number recorded as actually having been performed in the daily listings. Often, the accounts note that performances from previous months are included in a particular payment, having been omitted in the first instance. On these occasions I have re-established payments in their correct timeframe, where necessary dividing the total sum in proportion to the number of performances in each month. 31 Where no number of performances is given, I have inserted the number of performances according to the registers. 32 There are six occasions on which the registers explain a disparity between the numbers of paid and actual performances by the note that 'X number of performances were "au-dessous des règles"', and therefore not liable for payment. This refers to a practice called the chute, which was well established too at the Comédie-Française: once a play began to earn less than a specified sum (below 600l in summer and 1,000l in winter), it was classified as nulle, and was no longer subject to authorial payment. 33 On four occasions where this note is found, the number of nulles performances of the title in that month does indeed match the number of omitted performances. Of the forty-six remaining entries in which the performance figures in the registers and the payment records do not match, another four are underpayments in which the disparity can be accounted for by identifying nulles performances (although the notes identifying them as such never appear after early 1760), or free performances for charitable means. Some titles can be seen to become entirely nulles: L'Écossaise has four out of five performances discounted in September 1760, and thereafter passes into the list of unpaid titles. Other titles are unpaid for one month and return to the payment lists in subsequent months: here is it impossible to know if I am dealing with omissions or entire 31 Cf. five performances of Rose et Colas paid in July 1764, two of which actually belong to June. I have divided the total sum proportionally between the two months, which may not reflect the exact takings upon which the authors' 1/18 was calculated. 32 Cf. payments made for Rose et Colas in December 1764, for which no performance figure is given. I have therefore assumed the payment is being made for the two performances listed in the registers on 22 December and 31 December. 33 Cf. the note on the payment for La Rentrée des théâtres in April 1760, when only five out of the seven performances are paid, 'les deux autres étant au-dessous des règles'. months that were exempt from payment. 34 If such performances were paid and simply omitted, they would obviously have increased the annual totals earned by their authors. The remaining occasions on which the number of performances paid does not match the number that took place must be attributed to incorrect records, or the existence of parallel, unrecorded payment systems. I have not attempted to correct any errors aside from those noted by the registers' compilers, since any changes would be based on speculation. 35 The accuracy of the rest of the information allows me to gain at the very least an overview of the situation, especially since I am largely working with totals over the ten-year period, so individual performances or specific monthly totals have a less significant effect.
A second major drawback is the ambiguity of so much of the registers' content. Since authors are not named in the payment lists, I have supplemented the raw data with authorial attributions drawn from a variety of modern and contemporary sources. 36 I have tried to be cautious, in the hope of producing under-rather than over-estimates of an individual's earnings. First, the overlap and reuse of titles described in the earlier part of this article meant some detective work was required to ascertain which authors were likely to have received payment for particular versions of a play. When a series of authors are known to have reworked a single piece over time, I have assumed the most recent author is the beneficiary, whilst in cases of multiple simultaneous authorship I have attempted to identify the principal contributor. 37 In the handful of cases where there is serious ambiguity, I have 34 There are certain occasions on which we can assume with a reasonable amount of certainty that information has been omitted concerning repayment for earlier performances: in February 1761, for example, a payment is made for a performance of La Nouvelle troupe that does not take place. However, in November 1760 only four out of the five performances of this title had been paid: earlier on in the decade omitted payments are often made in subsequent months, so we can assume this accounts for the disparity. not included these works in calculations for individual author payments. 38 When the accounts state 'à l'auteur' but two authors (of words and music) are known, I have assumed the payment is made to the author of the words 39 whilst, following the practice of giving equal payments to both contributors seen elsewhere in the register, any plural formula ('à ceux') is assumed to refer to an equal split between the two. 40 For two titles I have been unable to establish authors. Quand est-ce qu'on me marie was attributed to Voltaire, 41 but he denied its authorship in a letter of May 1762 (XXXVI, 488-91), whilst Les Braconniers appears to be an anonymous ballet, which would make the payment of its 'authors' problematic (Wild & Charlton,170). The records become more and more standardised over the decade, with the final few years consistently paying 'les auteurs', giving the sum as a fraction, and displaying fewer inconsistencies.
Whilst these questions of inaccuracy and ambiguity will pose a challenge to those planning the larger digitisation project, a final set of limitations in my data result only from my own restricted time and resources, and should therefore be overcome by the vast scale of CIRESFI. By confining myself to 38 One such case is that of Le Dépit généreux. This play was originally by Anseaume; however, d'Origny records a dispute over authorship between Anseaume and Quétant (II, 5). Since there is no further evidence to suggest which of them was paid, I have excluded this title from my calculations. More complex are the cases of Annette et Lubin and Les Caquets, written by the family groupings of Favart and Mme Favart, and Antoine and Mme Riccoboni respectively (of the former, Favart himself states 'c'est une pièce que ma femme a faite avec son teinturier', but follows this with descriptions of his own contribution (I, 233)). Unlike for other plays of multiple authorship, payments are made explicitly to both authors: for Annette, two of the first three payments are 'à Mme Favart' (March and April 1762), and one later payment is explicitly to M. Favart (December 1762), and for Les Caquets one of the eleven payments is to M. Riccoboni (April 1762) and another to Mme (October 1762). Rather than excluding these two high-earning plays from the totals entirely and thus doing their authors out of significant sums, I have divided the remaining authorial payments equally between the two authors on each occasion to give a very rough guess at the droits they received. 39 The Correspondance littéraire recounts a case when the audience called for the 'author', and the troupe assumed this referred to the author of the words, who stepped forward (Grimm and Diderot et al.,VIII,352). 40 In May 1762 the payment records read: 'À l'auteur des paroles de l'Isle des fous pour 2 fois 33l 6s 8d: 66l 13s 4d. A celui de la musique idem: 66l 13s 4d', whilst August 1762 contains: 'Aux auteurs des Paroles et de la musique de Sancho pour treize représentations de ce mois le douzième: 655l 5d.' March 1763, on the other hand, contains the more ambiguous but more common statement: 'Aux deux auteurs du Bûcheron, comédie en un acte melée d'ariettes, pour 14 représentations compris celle du mois de février. La douzième pour les deux: 2553l 5s 10d', followed by a whole series of payments beginning simply 'à ceux'. 41 It was even printed in Condorcet's edition of his Théâtre in the late eighteenth century (I, 307-24). See Quérard,X,312. the day-to-day records, overall totals and the standard information pertaining to authors in monthly accounts, I have been unable to analyse how one-off payments, lump sums, loans and concessions of droits interacted, except in a couple of specific cases. More significantly, the lack of comprehensive financial information for every aspect of the administration means I cannot establish the precise provenance of ratios for authorial payment, even given my knowledge of the formula by which in theory they were calculated. This sort of large-scale calculation, over longer periods than a single decade, will be an invaluable product of the digitisation project.

Mapping Authorial Payment
Despite the limitations set out above, it is still possible to give a general account of author payment at the Comédie-Italienne in the 1760s that has up until now been lacking. One investigative pathway regards the plays themselves: the genres and specific titles that were most lucrative for their authors and composers, and the payment patterns they produced. A comparison between the full registers and the list of payments reveals -as we might expect -that a large number of plays are not included in the lists of titles paid in this period. The largest group here is the traditional Italian canevas, which were clearly considered as collaborative productions by salaried troupe members, and were therefore not subject to extra payment. A second group is the French comedies drawn from the older repertoire of the Comédie, including several titles by Marivaux. Ballets and pantomimes are also never listed, barring the mysterious anomaly, Les Braconniers. The rest of the unpaid titles consist of opéras-comiques or other musical productions whose authors have died or ceded their rights, or which are old enough to be giving only nulles performances. This leaves the new and the successful: ninety-two plays that have recently been created or that have stood the test of time. The titles that are paid are largely opéras-comiques or other types of play with music, and by ranking the payments received by certain titles we can map the styles and subjects that were most lucrative for their creators. The play that earned its authors the most money over the decade was the three-act comic opera Le Roi et le fermier, with 24,441l split between Sedaine (words) and Monsigny (music) for 165 paid performances from December 1762 to the very end of the period studied. 42 This also places it in eighth position for the highest average sum per (paid) performance, suggesting a relatively consistent popularity across this long run, which is confirmed by the continued high payments even after several years. The audacious choice to represent a monarch on the comic stage may have contributed to the ongoing popularity of this title. Furthermore, Collé's La Partie de chasse, based on the same plot, was circulating in private theatres in the early 1760s, but was prevented from appearing on the royal stage in 1764 by order of the censor. The frisson created by this link may well have kept the interest alive in Sedaine's title throughout the decade and beyond (Brown,A Field of Honor,. But this was also clearly a lucrative writing partnership: Sedaine and Monsigny's Le Déserteur was performed forty times in a single season (1769-70), averaging 173l per performance, 86l per author.
Whilst the two plays cited above are both well paid and continuously performed, the overall correlation between number of performances and size of average payment per performance is not particularly strong. The eighteen plays that are paid just once or twice in this decade often attract unusually large sums, perhaps representing an initial burst of interest that was disappointed. 43 The joint second-highest per-performance payment is for the short pastorale style piece La Bergère des Alpes, which earned Marmontel and Kohaut 173l between them for each of six performances in February and March 1766, but after this was no longer performed. At the other extreme, the one-act Deux chasseurs et la laitière is in fifty-second position, with Anseaume and Duni paid an average of 62l between them over its long run of 109 successful performances from July 1763 to June 1767. 44 In this case, relatively low but consistent popularity was more lucrative for the authors.
These examples demonstrate how the payment data can provide a perspective on the success or failure of different plays that is not visible from a simple analysis of the basic registers. However, a second path for investigation, far more directly relevant to my own study of the authorial career, is to sort this data by author, analysing how much individuals earned per season and per performance. This examination of the financial benefits offered by Comédie-Italienne authorship allows us to consider first, from the perspective of the authors themselves, how far material concerns may have been a factor in their choice to work for the Italian troupe, and secondly, from the perspective of the theatre, the monetary value of authors as a commodity.
The highest-earning author at the Comédie-Italienne in this period was Charles Favart, who had been a 7,000l shareholder in the Opéra-Comique 43 A single January 1766 performance of the mysterious Les Braconniers is the highest earning 'average' at 199l. 44 The lowest consistent per-performance payment is for the music of Le Gondolier vénitien, for which Rigghieri is paid just 9l a time, followed by Nanette et Lucas, whose author and composer Framery and Herbain must share the average 18l their work is paid over eight performances.
before the amalgamation with the Italian troupe (Fabiano,61). A total of 485 paid performances of his plays over the decade earned him 24,552l ( fig. I), whilst for the 1761-62 season and from 1769 onwards, he was also earning a pension from the Comédie-Italienne. 45 Second and third places in the overall rankings are occupied by Sedaine (21,113l) and Anseaume (11,789l). 46 The most prolific writers of opéras-comiques, these are the only three authors to be paid across the whole decade (Anseaume for all ten seasons, Favart and Sedaine for nine each), and it is therefore unsurprising that they have particularly high total earnings. But an analysis of season-by-season salaries reveals that the level of payment, even for these clear favourites, is by no means consistent. Favart's annual totals range from 331l to 6,700l -a striking figure compared with the guaranteed 4,000l authorial payment he had enjoyed at the Opéra-Comique -whilst Sedaine's are largely around the 1,000-2,000l mark but vary widely ( fig. J). Even the best-paid authors could not rely on any real consistency of payment from the Comédie-Italienne: it is no wonder that the annual pensions paid to a handful of the most prolific authors were so desirable, providing as they did at least a modicum of financial stability. Though the most successful authors faced inconsistent payment, they did at least receive some relatively substantial sums across the decade: Favart and Sedaine account for eleven out of the twelve highest annual payments made to authors across all ten seasons. At the other end of the spectrum, seventeen authors are paid in just one year, indicating only a fleeting involvement with the Comédie-Italienne. Annual payments of over 2,000l appear in only thirteen out of 101 cases, whilst over half of the total annual payments recorded are lower than 500l. Not only did the majority of authors write inconsistently for the Italian troupe, but when they were paid for this writing, their remuneration was not particularly high. Moreover, this inconsistency is also visible on the level of individual performances. Since the authors were largely paid based on a fraction of the takings, their payment was proportional to the success of their play; or rather, to the whims of the Parisian public from one week to the next. Favart's per-play payments therefore range from a measly 4l 16s 10d average for two September 1761 performances of La Nouvelle troupe to 208l 2s for each of ten performances of Soliman II in April 1761. 47 45 The average daily salary for a mason or a carpenter in the period 1750-75 was around 18s. Their 250 working days a year had to fund an average annual rent of 140l, as well as clothing, heating and food: bread was 2s a pound, and the average person consumed three pounds of bread a day (Avenel,III,150 and 233;I,289). 46 Anseaume was also paid up to 100l a month as prompt for much of this period, adding 1,200l a year to this total. 47 Payments for Soliman II (also known as Les Sultannes) are ambiguous, since they are apparently made to only one 'auteur', implying Favart received the whole total despite the involvement of other authors. The smallest of the payments cited here seems incredibly Clearly, it was possible for very few of the authors described here to rely financially solely on the Comédie-Italienne, but that was not the contemporary expectation. Dramatic authors in the period not only generally worked for several theatres at once, but also frequently had other sources of income: many were journalists or secretaries to the aristocracy, and even the prolific Sedaine continued to work as a stonemason throughout his career (Poirson,195). Though Favart's 2000l annual average, supplemented at times by a pension, might well have constituted a living wage, for the majority of these authors their Comédie-Italienne payments were one among many different income streams, and it is impossible to reconstitute these in any comprehensive sense. 48 More straightforward is to consider authorial payments as they represent the value of the author to the theatre. The Comédie-Italienne was far from inconsistent in its spending on authors: season totals spent on recorded droits payments for words and music, excluding an anomalously low 1760-61 payment before the Opéra-Comique merger, all hover within 2,000l of the 17,585l annual average; 9,636l for payments to authors of words only. This latter figure represents just 2 per cent of annual recettes, and is equivalent to the amount spent on a single principal actor in the only year for which there is a full record of troupe parts (1769-70, 9641l/part). But the wide range in the number and size of payments made to individual authors indicates that whilst it was relatively straightforward for an author to have one play performed at the Comédie-Italienne, and thus procure a small share of this annual budget, the acquisition of a regular position as author was more problematic. Either authors simply did not try (or did not have the talent) to become the next Favart, or the theatre's apparent desire to cultivate authorial loyalty, for example by offering medals to authors who provided multiple successful plays, was merely for show. 49 low as a fraction of total takings: Favart was ostensibly paid 1/18 for the two September 1761 performances, for which the recettes were 541l and 625l -the former slightly lower than the chute level, the latter slightly higher. There is no way of knowing how these recettes translated into the tiny payments received by Favart that month: possibilities include the repayment of debts for the purchase of tickets before the payment of his droits, or particularly high expenses for the performance which were deducted before payments were made. Unfortunately this ambiguity is a function of the way the Italians recorded their author payments. 48 Martial Poirson's prosopography divides authors according to whether they had other occupations/sources of income, however he does not attempt to reconstruct their incomes (541-53). 49 A 'Mémoire présenté par la Comédie-Italienne' (1761-62) (AN.O 1 *.851.78 r -79 r ) sets out the proposal that 'il soit donné à la fin de chaque année par [l]es Comédiens Italiens à l'auteur qui se sera le plus distingué par ses pièces propres pour leur théâtre, une médaille d'or de la valeur de quatre cents livres environ'.
Average per-performance payments across the decade, rough though such figures are, are even more revealing ( fig. K). Of the thirty-six authors that are paid, twenty are paid on average between 30l and 60l per performance for their total 1760-70 output. The highest earners on this scale are De Moissy, with a 74l average over eight performances, and Desfontaines with a 64l average over fourteen performances, 50 whilst Favart receives an average of 51l. However, the picture changes considerably when we examine non-paid performances, an operation only made possible by combining the author payment data with the daily performance information discussed in the earlier part of this article. In Favart's case, over 1,100 performances of his various titles were given across the decade, more than double the 485 for which he was paid. Adjusting the calculations to take this into account, we find that even including 1,800l or so of pension, his actual per-performance rate over the ten years was only somewhere in the region of 17l, 15l without the pension. Sedaine's average payment drops from 55l per performance to 31l when 294 unpaid performances are taken into account.
These calculations play into the heated contemporary discussion regarding the practice of the chute by providing concrete proof of just how lucrative older plays, for which there was no obligation to pay the author, could be for the Comédie-Italienne. Had Favart's non-paid performances been remunerated at the same rate as those for which he did receive payment, the troupe would have faced an extortionate bill. Where the author was still alive, non-payment represented a less successful performance, and thus a lower recette for the troupe too. But for a troupe very concerned with its finances, there was clearly a difficult balance to strike between new titles, so crucial to attract an audience, and popular older plays, where the troupe would keep all the takings no matter what the recette achieved. A regular accusation made against the Comédie-Française actors was that they intentionally incited audience discontent in order to make a play appear unsuccessful and thereby enter into the 'free' repertoire: in the context of the calculations set out above, this becomes quite comprehensible. 51 Authors did have some capacity to play a similar game, for they could decide to opt out of regular payments in favour of a single one-off sum. Given the uncertainty described above, it is little wonder that some authors 50 This excludes two text-only plays of problematic authorship, whose one-off payments to single authors give anomalously high results. Quand est-ce qu'on me marie, attributed to Voltaire but vehemently denied by him, and L'Heureux événement, attributed variously to Montbaron, Le Blanc and Le Bret (Favart,II,174). It also excludes Parmentier's Faux lord, which was paid for one performance but never performed. 51 In 1772 the dramatist Jean-François Cailhava suggested this problem be solved by subjecting old plays to the part d'auteur, to be divided between actors of merit and the government (33-34). took this option, which could not only considerably augment their overall earnings, but was often the financially sound longer term choice. In May 1766, Poinsinet ceded his payment rights for Tom Jones, which had earned him and the composer Philidor an average of 69l each for each of twenty-five performances since February of the previous year, for the sum of 2,000l. 52 In the next three months (June-August) Tom Jones was performed only six times, earning Philidor (now paid half of the 1/9 fraction the two authors had been paid) an average of 39l per performance. At this point Philidor, presumably aware of the steep drop in his earnings, also ceded his rights for 2,100l in September 1766. The rest of the decade saw another twenty-nine performances of Tom Jones. Had the authors been paid at the same level as the first run of performances, they would have earned 1,991l each over the rest of the decade. At the level of the June and July performances however, this payment would have been 1,120l, and given the sharp decline in interest in just two months, it can be assumed that takings would have dropped until the title had become completely nulle. Their respective decisions to give up their regular rights payments for a flat sum seem very canny indeed. The analysis outlined here reveals the precarious nature of dramatic authorship as a profession, the impossibility of surviving on authorial droits alone, and the delicate balance operated within the Comédie-Italienne between attracting audiences with new plays, and earning money through old ones. It further indicates that no Comédie-Italienne authorial career can be examined in isolation: instead, we must be aware of the adaptability and polyvalence of the authors that wrote for the Italian troupe. Their eclecticism means that any study of their careers -and any comprehensive understanding of the theatrical administration of the period -must necessarily draw in data on all of the main theatres; a step that is clearly unfeasible in the bounds of this short study. Though authorial data may not currently be included in the CFRP, the absolute interdependence of the different theatres on various levels underlines why the parallel digitisation projects for the two Comédies are so valuable, and why they must be in dialogue with one another and with researchers across the field in order to ensure their outputs are directly comparable.
However, although such analysis -both on this small scale, and in any future, grander incarnation -adds much to our understanding of the Comédie-Italienne and its authors, and contributes to the growing picture of theatre administration in the period, it is not enough to look at the raw figures alone. This is very clear if we attempt to identify the place the Comédie-Italienne might have occupied in the authorial career trajectory. For example, extant accounts for Voltaire's authorial droits at the Comédie-Française in the period show a per-performance average that was far higher than that of the best-paid authors at the Italian theatre (from 141l to 205l), whilst de Belloy, the author of the runaway 1765 success Le Siège de Calais, earned 241l per performance for its first run at the national theatre, dropping to 181l by the fourth run twelve years later. Both figures imply that the payment scale at the French theatre was superior to that at the Comédie-Italienne, even if de Belloy's payment in just eight out of sixteen seasons indicates that the danger of inconsistency was equally present. This basic financial comparison might encourage us to imagine that the Comédie-Française was the obvious first choice for any author even from a financial perspective; that the commercial parity between the theatres, sketched in the earlier section, had no impact upon authors, who instead existed in a world shaped purely by the traditional hierarchies with which we began.
But raw payment figures are only one facet of the question. Moving away from the big data to consider contemporary regulations, we discover, for example, that the Comédie-Italienne offered a quicker acceptance procedure than its French counterpart. Producing three times as many new titles a year as the French troupe, and operating less stringent regulations for new plays entering the repertoire, it presented authors with the possibility of faster payments, and immediate presence on a Parisian stage. 53 And authors seemed to be aware of this difference. Marmontel -historian, Comédie-Française author and academician -continued to offer his plays to the Comédie-Italienne well after becoming a recognised member of the Parisian literary elite, suggesting that he wanted to exploit the financial rewards the Italian troupe offered more readily than its French counterpart. To compound their financial attractiveness to authors, from 1774 the Italians abolished the chute, instead instating a policy of paying droits whenever the takings for a given performance exceeded a specified total (AN.O 1 .848.11). And there were also far less quantifiable issues at stake: as well as the symbolic capital that association with different theatres could provide, there were the trends in dramatic theory and practice that brought different styles in and out of favour (opéracomique, the drame), and the personal and political alliances that could make or break a career. The financial data add a vital layer of further nuance to our 53 In the first half of the 1760s an average of 30.4 new plays were created per season at the Comédie-Italienne, compared with an absolute maximum of twelve per season at the Comédie-Française. Whilst both theatres required authors to present their plays to the troupe for acceptance or rejection, at the Comédie-Française the delay between presenting a work and having it performed often stretched to years (Brown,A Field of Honor,(83)(84)(85)(86). The Italians had a far quicker turnaround, in part due to their lack of a permanent repertoire of favourites such as that owned by the Comédie-Française.
understanding of what it meant to be an author in eighteenth-century Paris, but it must be combined with close reading of more traditional sources.

* * *
Capturing and manipulating the performance and payment data provides an objective measure of both the status of the Comédie-Italienne and how it valued its authors, painting a picture that reliance on contemporary written accounts may obscure. This practical understanding of the theatrical field is immensely valuable in its own right, for even with the sorts of approximations I have had to make here, it can allow scholars to analyse the role and value of different sorts of artistic production in the period, consider the theatre's relationship to political power, examine the pragmatic aspects of an authorial career, and study the practicalities of contemporary theatrical administration. But it can also contribute to more classical literary studies, by enhancing our understanding of the context within which particular plays were produced, and providing evidence that might help to explain, for example, why an author decided to write in a specific style at a specific moment, why a play that was rejected by one theatre was immensely successful at another, or why a title that initially appeared to have suffered public disapproval in fact ran for many years. Whilst the potential of such studies has long been recognised, the digitisation projects offer an unprecedented opportunity not only for a large-scale examination of these questions and many more that would be impossible to answer without these tools, but for the opening up of these resources -previously accessible only to those able to make long, meticulous archive trips -to a vast audience of researchers in a variety of disciplines across the world. It is, however, important to strike a note of caution. The ability to manipulate the information quickly and flexibly and to make calculations instantaneously relies on data being stored in an accessible and coherent system, which is not always obvious from a first reading of the sources. The vagaries of eighteenth-century recordkeeping -the inconsistencies, the lack of standard recording formats, the information scribbled on the backs of printed forms, and the erasures -complicate attempts to systematise these documents and record them in digital form (Tomasek). Indeed, these complications may well explain why the Comédie-Française project has not -for the moment, at least -attempted to include details of expenditure in its digitised collection. The decisions I have outlined here regarding approximations, the inclusion of external information such as authorial attributions, and what archival data can be made to show given their limitations, will all be faced on a much larger scale by the more ambitious digitisation projects. And if these issues do not surface, it will be important to ask why, and consider what might have been overlooked. It is therefore vital that the projects' creators identify from the start how information will be used, and canvas possible research questions from researchers working on a range of timeframes and topics, in order to optimize the tools that are provided, tag and allocate information in the most helpful possible way, ensure that allowance is made for altering concerns across the century, and avoid a quick descent into obsolescence (Smith). Whilst these challenges are well recognised by the projects' directors (CIRESFI,(9)(10)(11), this short study has revealed both the realities of the specific problems that may be encountered, and how useful conclusions can still be drawn from the data if carefully managed.
Digital humanities projects have frequently grappled with the question of providing access to the original documents alongside digitised data, and many of the largest projects (including the CFRP and the longer-standing ARTFL digital Encyclopédie at the University of Chicago) have decided to link scans of the originals to their digitised counterparts. My work on this small subset of the Comédie-Italienne registers confirms the necessity of this approach. However enticing big data might be, the allure of the 'big' and of quasimagical computational capacities must not cause us to lose sight of what is actually in the figures and documents themselves. The digital humanities, by definition, deal with information that has a contingent, human element, and it is precise the aleatory asides, where the data provided deviate from what might be considered a 'standard' or 'logical' format, that reveal the preoccupations of the original users; preoccupations that could be lost by imposing a retrospective order (Tomasek). 54 Having linked author names and play titles to appear in a helpful search, we might forget that these author names were never in fact present in the original registers; that, therefore, the troupe placed a relatively low value on who exactly was writing for them. We might not see, either, that the author payments are listed alongside payments for posters and costumes; that the authors were part of the furniture rather than deserving of a place onstage beside the actors. We might ignore the multiple different titles under which plays were recorded by different hands in the registers across the century, and the supplementary information that was given or left out. And we might miss the one-off note in March 1765 that records a payment made to the Italian author Carlo Goldoni for his return to Italy -a return that, in the event, never in fact took place.
This, for me, is the greatest challenge of these vast and laudable digitisation projects: treading the fine line between rendering the numerical data accessible, standardised and streamlined, so that illuminating calculations 54 Compare also the recent 'material turn' in digital humanities, Kirschenbaum & Werner, can be made at the click of a button, and retaining its revealing inconsistencies and haphazardness, so that we can still reach across the centuries and try to understand the concerns of the people who compiled it. Curating my own data on a small scale, without a vast team and with unsophisticated tools, forced me to tread this line by taking direct responsibility all the way from the original document to the graphs and charts I reproduce here. The users of big data digitisation projects risk losing this direct relationship with the sources, and it will be up to the projects' compilers to ensure that their human element does not vanish. But alongside this, we should continue to encourage individuals to curate smaller-scale collections and to make them publically and transparently accessible, thereby combining the power of big data with the humanity of traditional methods. It is in this spirit that I make my own work and methodology available, in the hope that it will contribute to such an ongoing, collaborative project.      6 0 --6 1 6 1 --6 2 6 2 --6 3 6 3 --6 4 6 4 --6 5 6 5 --6 6 6 6 --6 7 6 7 --6 8 6 8 --6 9 6 9 --7 0  6 0 --6 1 6 1 --6 2 6 2 --6 3 6 3 --6 4 6 4 --6 5 6 5 --6 6 6 6 --6 7 6 7 --6 8 6 8 --6 9 6 9 --7 0