Jump to ContentJump to Main Navigation
That Men Would Praise the LordThe Reformation in Nimes, 1530-1570$

Allan Tulchin

Print publication date: 2010

Print ISBN-13: 9780199736522

Published to Oxford Scholarship Online: September 2010

DOI: 10.1093/acprof:oso/9780199736522.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 30 March 2017

(p.203) Appendix B

(p.203) Appendix B

Constructing the Notarial Database

That Men Would Praise the Lord
Oxford University Press

(p.203) Appendix B

Constructing the Notarial Database

Notarial records of wills and marriages are crucial to this study. They open an essential window onto people’s lives, and fortunately, they do so at critical, solemn points in the life cycle. At the same time, they give us only certain information, and the window they provide is necessarily somewhat skewed and partial.

In order to make the records useful, they had to be transformed into machine-readable form. Readers who wish to understand the basis for the statistics used in the body of this book are entitled to understand how the database was created; they may also be interested in the details of its creation if they intend to use databases in their own research. This appendix will cover five main issues: (1) bias, (2) coding last names, (3) occupational information, (4) social status, and (5) network analysis issues.


The database of notarial contracts used in this book is quite large. It consists of every will and marriage contract which survives for Nîmes from 1550 to 1562 Old Style, that is, March 25, 1550, to March 24, 1563. This totals about 1,100 marriage contracts and just under 700 wills. For these contracts, I recorded the principals (husband, wife, testator); the fathers (for marriages); and all witnesses. The result was about 13,000 references to named individuals, the largest database of its kind known to me. The data should therefore be particularly reliable, although no sixteenth-century statistics can be considered absolutely trustworthy. Nonetheless, there are two biases in the sample that are important in some contexts. First, Jean and Jacques Ursi, two notaries who were early converts to Protestantism, account for about half of the surviving acts. The second is that the Ursis’ acts are not evenly distributed across the database: instead, wills from other notaries survive in large (p.204) quantities for the late 1550s and 1560s, while the Ursis’ wills and marriage contracts are the overwhelming majority for the early period (1550–1555). Figure A.1 illustrates how the records of different notaries survive in different quantities.

Appendix BConstructing the Notarial Database

Figure A.1. Surviving acts by notary, 1550–1563. Source: ADG, Series IIE.

As noted in Chapter 3, the major consequence of this skewing is that Ursi and non-Ursi acts must be separated when looking at the evolution of religious formulas. In a chart of the use of the phrase “Holy Mother Church,” for example, looking at the entire database, use of the phrase would appear to have declined in the early 1550s, as Ursi clients stopped using it, and then seem to have increased again in the mid-1550s, because non-Ursi acts that continued to use it become better represented in the database. There is also another pitfall in compiling data on religious formulas in notarial acts. In draft versions of notarial registers, which sometimes survive, the formulas are frequently omitted. Sometimes, what is apparently an unbroken series of registers is actually a mix of draft and final copies: presumably, at some point, a clerk threw out all the draft registers except for those years when the final copy had disappeared, and he inserted the draft copy at that point. This can wreak havoc, if historians assume that the absence of Catholic formulas is a sign of incipient Protestantism. Fortunately, for Nîmes, there are few surviving drafts (they come from the notaries Rançon Alirand and Jean Lansard); they were included in the database, because it would have made little difference to the statistics had they been excluded.1

Coding Last Names

For certain purposes, however, many references did have to be excluded. Unlike a database constructed from a tax roll, it is perfectly possible for people to appear more than once in a database compiled from notarial acts: in fact, some people appear more (p.205) than twenty times. Multiple references to individuals are both a blessing and a curse. They allow us to get a much better fix on individuals, their professions, friends, and origins. Sixteenth-century documents are much less rigorously exact than those of later centuries, and scribes frequently omitted professions, villages of origin, and so forth from documents. Furthermore, multiple references can help to clarify poor handwriting or a water-damaged page. But it is frequently difficult to be certain that two references are indeed to the same person. All sixteenth-century spelling is highly inconsistent. First names are nonetheless extremely reliable, because the repertoire of first names is relatively small, while there is a much larger repertoire of last names, which makes them very difficult to transcribe. Scribes also clearly had trouble with unusual last names: in a largely illiterate age, they could hardly turn to the person and ask, “Would you spell that, please?” Professions are more reliable, when included, but they frequently are not, especially for witnesses. Furthermore, people could have multiple professions or change professions: a laborer decides to rent an inn, an artisan becomes a town consul for a year, or a rich lawyer buys a property and then becomes entitled to be called seigneur.

Multiple references must be found, however, and aggregated, because even the simplest statistical tests are invalid otherwise. For example, if 2.39 percent of the references in the database are to advocates and doctors of law, that hardly means that 2.39 percent of the individuals in the database, or in the town, were lawyers. In general, more prominent people tend to be referred to more frequently, so not compensating for multiple references would bias the sample. It would also make the analysis of family ties and factions completely impossible.

Appendix BConstructing the Notarial Database

Figure A.2. Coding last names. Source: ADG, Series IIE.

Unfortunately, establishing that two references are to the same person is not entirely straightforward. Consider the set of references in figure A.2. It is quite clear that these all refer to the same person, although his last name is spelled three different ways. Paleographic difficulties and water damage (and typos in later data entry, of course) only compound an already difficult situation. But, in this case, we have the advantage of an unusual first name (he may have been the only man in Nîmes to bear it) and a precise occupation, seigneur de Sauvinhargues. It is by contrast difficult to determine whether two references to “Jean Dumas, travailleur” are to the same person.

In general, it has been assumed that, if two documents mention the same name and occupation, they refer to the same person, unless there is clear evidence to the contrary: this certainly introduces error, but to assume the contrary would be worse and would render the task impossible. Furthermore, multiple references have been exploited whenever possible to fill in missing data in other references: had a reference (p.206) to “Hermingard Falcon” been found that did not mention that he was seigneur de Sauvinhargues, it would have unquestionably been marked as the same person. A reference to a name without an occupation was automatically assumed to be the same person unless there were two persons with the same name and different occupations, in which case it was left indeterminate unless there were other clear factors to make a judgment possible.

In order to identify persons and eliminate differences in spelling, an identification system based on the Soundex system was created. Soundex, used widely in the United States, on Illinois driver’s licenses, for example, is a system for assigning code numbers to last names, by giving numbers to letters of the alphabet. The initial letter is retained, double letters are eliminated, and consonants are given numbers in clusters: B, V, P, and F are all given the code “1,” M and N are “5,” and so forth. Only the first three applicable consonants are coded. “Hermingard,” for example, would have the code H655, or H-r-m-n. If the name is short, blank spaces are filled in with zeros: “Jack” would be J200. Unfortunately, Soundex could not be applied to the database without some modifications. First of all, consider the case of our Hermingard: is his name “Faucon” or “de Faucon”? Because particles appear and disappear in last names throughout the records with no rhyme or reason, they were eliminated in the coding, except in certain cases: Dumas, for example, and Dupont, although frequently written “du Mas” and “du Pont” in the documents, never appeared as “Mas” or “Pont.” Articles (“le” and “la”) were more consistent and were generally retained. Women in this period did not bear their husbands’ last names, but rather a feminine form of their fathers’: Mme. Perrière would be the married daughter of M. Perrier, not his wife, for example. These endings were eliminated, so that they could be properly linked to their male relatives. Finally, because of the difference between French and English pronunciations, final, unsounded consonants were not coded since, not being pronounced, they were frequently omitted. “Jacques,” for example, would be coded like “Jack,” as mentioned above: J200, since the “s” is not pronounced. (Nasalized “n” was pronounced and therefore always retained for coding.)2

Each person received an individual identification number consisting of four parts: a Soundex version of their last and first names; a code for their occupation, grouped into larger categories (numbers between 500 and 599 were used for building, metal, and wood-related trades, for example); and two final digits used to distinguish homonyms, thus normally zero. In this system, Hermingard Falcon’s ID code would be F‑425‑H655‑100–00, that is, F for the first letter of his family name, then 425 for the consonants l-c-n, H655 for his given name, and the occupational code “100” for high status. The final zeros indicate that there is no one else in the database with the same name. The next-to-last zero would be changed to a 1 if there were another family of a different last name but with the same Soundex code, while the last digit was used to distinguish people of the same first and last names and occupations, usually fathers and sons. Although Soundex did an excellent job of eliminating spelling differences, occasionally different last names would be given the same code, or the same person was initially given different codes: the whole database had to be inspected (more than once) by hand, or rather, by mouse.

(p.207) (p.208) Occupational Information

Once Soundex numbers were created for the 13,000 entries in the database, identical persons noted, and multiple references consolidated, a statistical profile of mid-sixteenth-century Nîmes could be compiled. The results are in figure A.3. Persons who were described with different but related occupations were usually assigned to a general category: someone described equally often as “yeoman” and “husbandman” was coded for “agriculture,” for example. In order to convert these detailed figures into the summary information used in the text, the classification system in figure A.4 was used.

Appendix BConstructing the Notarial Database

Figure A.3. Percentage of men in each occupation. Note: Figures may not add up to 100 due to rounding. Source: ADG, Series IIE.

Appendix BConstructing the Notarial Database

Figure A.4. General categories for coding occupations in the notarial database.

Unfortunately, despite the efforts taken to collate information from every reference, the occupations of fully one-quarter of the adult male population could not be determined (only a handful of women were recorded as having occupations). Since even prominent persons are listed in the records without occupational data, it would be unwise to assume that the unknowns represent exclusively low-status occupations like travailleur and laboureur (translated as “husbandman” and “yeoman”). For this reason, it has been assumed in the discussion below that the percentages “excluding unknowns” are the best representation of the distribution of occupations.

Some groups are undoubtedly underrepresented. Servants were certainly more numerous than the database would indicate, for example, although servants perhaps should naturally occur less frequently in a chart of men’s, and not women’s, occupations. The destitute are essentially absent, though some very poor people do appear: some women had dowries in the single digits (£1.5 is the lowest, for the marriage of (p.209) Catherine Fabresse to Jean Michel, travailleur), and one bridegroom, Nadal Clemens, did not even know his father’s first name.3

By contrast, high-status occupations are almost certainly overrepresented: it is plausible that Nîmes, whose présidial was such an important institution, should have a large percentage of lawyers (2.71 percent), but could 6.46 percent of the population really have been merchants? One possibility is that master artisans were recorded as merchants, as some records suggest. Unfortunately, it is difficult to estimate the extent of the practice. Other possible sources of systematic error were investigated but could not be substantiated. For example, was there status inflation, of the sort that promotes a husbandman (travailleur) to a yeoman (laboureur) on his wedding day? This appears not to have been the case: variation in descriptions seems to have been quite random. As a result, some of the categories have to be read in groups. Weavers and wool carders were occasionally confused, likewise drapers and hosiers (drapiers and chaussetiers), while carpenters and furniture makers (charpentiers and fûtiers) were regularly commingled, as were the various branches of the leather trade. As a result, it is perhaps more accurate, as well as easier on the eye, to group occupations into general categories, as is done in the text. The overall percentages given in the text are necessarily imperfect, but they probably represent about as accurate a count as we are likely to get from the sources, recognizing the bias in favor of elites and the undercounting of the poorest (which, after all, still exists, in the United States at least). The one additional difficulty concerns not the results, but their classification: how does one determine what is really a high-status occupation? Determining status is a difficult, interesting question, to which we will now turn.

Social Status

Traditionally, historians have tended to assume, with considerable justice, that status is largely a matter of money. Exceptions are easy to locate: some professions are considered more honorable than their mere earnings may indicate, a point in which contemporary professors of history may well take some comfort. Generally, honor may (p.210) derive from lineage or education, but money certainly helps, and after all, it has long been customary for the rich but uneducated to remedy their defects in their children. In Nîmes’s notarial records, evidence of net wealth is difficult to come by, and estimates of income virtually impossible. Nor is even elementary evidence of education, such as the ability to sign one’s name, possible before the very late sixteenth century. Wills only occasionally give legacies of named amounts, usually designating instead a couple of small legacies and then leaving a single heir or a group of persons as residuary legatees.

Dowries, therefore, are one of the few good ways to get a sense of relative wealth and social status. Even this figure, however, is not completely reliable: some marriage contracts do not give an exact figure for the dowry, but instead give the augment dotal, which was probably treated as a percentage of the dowry, either one-half or one-third. Such contracts, concentrated at the lower end of the social scale, amount to nearly a third of all cases. Of course, it would be kinder to historians had the percentage been fixed: as it is, it is impossible to use the augment dotal to ascertain the dowry figure precisely. For our purposes, where the dowry was unspecified, the augment has been used, multiplied by two, to approximate the implied dowry of the wife. This assumes that the husband and wife come from similar social backgrounds. But even if one husband or wife makes a misalliance, the average dowry for wives of bakers, for example, should give a good notion of the rank of bakers within the wealth hierarchy. The plausibility of the results obtained suggests that this system of assessing dowries does provide a good means of ascertaining social status.

It is actually possible to be yet more precise and perhaps to quantify some of the more intangible elements of social status by using dowries. Witnesses to marriages tend to represent important people in the wedding: close relatives of the wealthy, servant women’s masters, fellow carders for wool carders, and so forth. A cursory examination shows what should not surprise anyone, namely, that witnesses to marriage contracts vary in prestige according to the social status of the persons getting married. Thus, the field can be usefully widened: the dowry figure can be applied not just to the husband in a marriage contract, but to all the witnesses to that contract. This has the additional benefit of helping to get a good fix on professions where the sample is small and may be unrepresentative. Only six marriage contracts survive in which the grooms are surgeons, for example, but surgeons appear forty times if occasions are included in which they are witnesses.

At this point, it should be remembered, the dowry figure is not being used as an actual sum of money but as a device to clarify social ranking, so that priests’ social rank is given by the average dowry of marriages at which they appear, although as celibates they should not have been getting married themselves. This being the Reformation, however, three priests did get married, and it is striking that the average dowry for those three priests was only £50, while it was over £250 when they were witnesses. Here is a good example of the advantages of using witnesses as part of the sample: priests who were getting married were clearly less respectable than priests in general, and that is why the average dowry figure shifted so markedly. Even ex-priests from wealthy backgrounds could not obtain heiresses in marriage.

(p.211) The data from both methods are worth considering since, as in the case of priests, the differences between them can be quite revealing. However, because the inclusion of witnesses results in a larger sample, a number of small occupations appear in that sample that do not appear when only husbands are considered. For that reason, the results will be presented separately, ranked from highest dowry to lowest, rather than combined in one table. It should also be noted that, for a variety of reasons, the average dowry is slightly higher when all cases are considered, although high-status occupations tend to fall and low-status ones to rise. All money has been converted to livres tournois and decimalized (1 sou = .05 livres).

Some results in these figures should not surprise historians of early modern France. For example, lawyers rank above even wealthy merchants and bourgeois, while farmers and agricultural workers rank at the bottom of the scale. In figure A.6, a plausible division would be to mark dowries above £250 as wealthy; £125–250 as an upper-middling group; £50–125 as middling; and below £50 as poor (for figure A.5, it would probably be more reasonable to set figures of £200+, £100–200, £50–100, and below £50, respectively). Thus, professionals, officials, and bourgeois (retired merchants) would rank as rich, while some of the more prestigious occupations like merchants, notaries, other men of the law like clerks and praticiens, and apothecaries would be in the upper-middling group.

The middle would consist heavily of those in the cloth trades, with workers in leather goods in the upper half and hatters and wool carders at the lower end. A group of high-status cloth trades, grouped together as “cloth making” (code 400) because there are too few of each of them separately, belong here, too. They include drapers, mercers, embroiderers, and velvet makers. Also prominent in the upper half of the middling group were yeomen, independent farmers who were eligible to be fourth consul. In the lower reaches of the middle group were the food trades: bakers, butchers, innkeepers. The poor consisted of those in the building trades (masons), carters, and above all, poor agricultural laborers.

Certain occupational groups have quite different averages in the two figures. Servants, for example, rank as poor in figure A.5, while relatively high in the broad sample. In this case, figure A.5 almost certainly gives a better picture: servants tend to appear as witnesses only in the marriages of the wealthy families that employed them, thus distorting the average. The lower figures for their own dowries more accurately reflect their real position. By contrast, students, who were mostly from privileged backgrounds, do quite well in figure A.6, as they should, but quite poorly when they themselves are getting married. This is almost certainly because “student” is a life-cycle-bound occupation. Masons can be fifteen years old or fifty, but students are usually fairly young, probably between fifteen and twenty. As witnesses to weddings, they frequently appear with their relatively prominent families, while students who are marrying young appear not to be marrying well. But the sample is also small: only two students married in the period examined. A few other occupations with surprising figures, like college professors, are also probably errors due to small sample size. Similarly, the “military” category consists largely of poor harquebus men and crossbow men, but two groups of army captains are witnesses to weddings of the high aristocracy, with dowries of £3,500 in one case and £500 in the other, thus biasing the results. (p.212)

Appendix BConstructing the Notarial Database
Appendix BConstructing the Notarial Database

Figure A.6. Average dowries of marriages attended. Source: ADG, Series IIE.

The foot soldiers considered separately have an average dowry of £41, ranking them in the higher ranks of the poor.

Dowry figures have another important use: they can give us an assessment of the relative social status of the clientele of the various notaries whose registers survive. This is particularly important because, as mentioned earlier, the records of the Ursi family of notaries are unusually Protestant in their language early on, and many future prominent leaders of the Protestant party appear there more often than anywhere else. Are the Ursis’ clients different in any way? One simple method to evaluate this is to consider the average dowry by notary (see figure A.7).

The figures make it reasonably clear that notaries’ clienteles varied dramatically in their social status, with Lansard’s, Sabatier’s, and Duchamp’s clients solidly tending toward the well-to-do, while the Ursis’ tended toward the lower middle. The average husband in a contract written by the Ursis married a woman whose dowry was comparable to the wife of an average innkeeper. Interestingly, the dowries in Ursi marriage contracts are not lower because the husbands in them come from different professions than average, but rather because they were poorer members of each profession. Obviously, the percentages are not exactly the same, so, for example, 6.04 percent of the husbands in Jacques Ursi’s registers are in the legal professions, compared to 8.93 percent of the whole sample, and 41.41 percent of them are in agriculture, compared to 34.57 percent overall. Nonetheless, the percentage of artisans, for example, is just 0.5 percent lower. In one sense, this is not surprising, since Jacques Ursi’s registers, in particular, comprise such an important part of the surviving records: but that only shows how significant the difference in average dowry is. The early Protestant milieu is in some ways sharply different from the general milieu of the Ursis’ registers: the percentage of artisans among early Protestants is higher, and the percentage of those in agriculture is much lower. Still, there was a distinct overlap, and the two groups, Protestants and Ursi clients, both contained fewer lawyers and officials than the average.

The result of using the marriages attended for averaging was that 11,000 of 13,000 references could have a social status number assigned to them. This still left 2,000 references, unfortunately, and therefore a significant number of people’s status could not be specified. This problem was solved through interpolation: if an individual consistently appeared with high-status people, that individual was assumed also to be of high status. Once I calculated the average dowry of marriages attended for each person in the (p.214)

Appendix BConstructing the Notarial Database

Figure A.7. Average dowries by notary. Source: ADG, Series IIE.

(p.215) database, it became possible to assign each act (will or marriage) an average social status based on the persons participating in it. Acts with high-status people, whether wills or marriages, have high average social status figures; wills of humble tradesmen and agricultural workers, or their widows, rarely do. Because this technique gives a status to every act, not just marriage contracts, it was possible to calculate the average social status, not of marriages attended, but of acts attended, and to assign a social status to all but a handful of references in the database. Interpolating in this way does introduce some errors: one high-status person appearing at a humble marriage can drive the average for that act sharply upward. But if people appear in several acts, that effect tends to dissipate. Interpolation does have one further effect: widening the sample tends to make averages converge toward the center. Agricultural laborers will still be at the bottom of the heap, and lawyers and officeholders at the top, but their average social status numbers will be a bit closer to each other. I call this number “social status by event.”

Network Analysis

Properly coding the database and sorting out the names and occupations laid the groundwork to perform network analysis. The crucial steps that had been taken were that each person now had a unique identifier, and so did each event, that is, each will and marriage contract. Two individuals who attended the same event could be considered to be related in some way. Pairs of husbands and wives could be assembled, for example. There were difficulties, however. There is a rigid distinction in network analysis between “actors” and “events.” This distinction poses a certain problem in doing a network analysis on the elite of sixteenth-century Nîmes, because many ties could not be (p.216) represented by events in the database. Consider the case of a married couple, Anne Durande and Bertrand Favier. Their marriage is the event that ties them together. What is the tie between Anne and her brother Charles Durand, or between Bertrand and Charles? No event—no will or marriage, in the context of the database—joins them. Nor can one consider the sister to be the connecting line between Charles and Bertrand: she is an actor, not an event. The solution was to do the network analysis on the basis of families, not of individuals. Thus, when Anne marries Bertrand, that creates a tie between the Durand and the Favier families, not between the individuals.

While using families solves an important problem, it creates others. The first is a problem of definition: what is a family? It is traditional, for example, to look at a marriage and say that, when Catherine Richière married Pierre Vallete,4 the marriage created a tie between the Richier and the Vallete families. But to Catherine, Pierre, and their descendants, it may have seemed more a process of establishing a new family, which may not have coincided with the Vallete family name. It is equally traditional to assume that such ties, especially among the elite, represented the desires of the families rather than those of the couple, but that may not be valid in every case. There is the additional complication that the database covers only a narrow band of time, 1550–1562/63, while marriage ties from years or decades before may have created a sense of family connection. Finally, family is also a porous term: although a computer might prefer that a person be either a member of a family or not, the experience of family can be, of course, quite different. Distant cousins, for example, may or may not be considered truly “family.”

In general, because it would be extremely complicated to do otherwise, I have used traditional, family-name-based definitions of families, which has involved ignoring certain indications of other sorts of family ties. Information gleaned from wills, for example, was used to ensure that persons with the same surnames were indeed from the same family, but similar information about marriage ties in earlier decades was not used to create links between those families, because the references were too scattered and incomplete. Albert Puech, a local historian who was active in the late nineteenth century, compiled extensive genealogical information about over 250 important families, (p.217) which I also used to assess whether a similar surname really corresponded to a family tie.5 In general, I assigned people to the same family only if the relationship was reasonably close: parents, children, grandparents and grandchildren, siblings, uncles and aunts, nieces and nephews, and occasionally first cousins when the parties themselves stressed the relationship. I tended to be suspicious of assuming a family tie based on the same surnames when the name was a common one, and I was somewhat more lenient with unusual family names. Because of this use of surnames, it might be more appropriate to refer to “lineages” rather than “families.”

Another consideration when using families as the unit of analysis is that some important information relates not to families but to individuals. One might want to know whether weavers, say, tended to be friends with each other, for instance by being witnesses at each others’ wills and marriage contracts. But while, for example, Étienne Serre may have been a weaver, one could hardly call the whole Serre family weavers, since other family members did other things, like woolcarding. (And how does one define the female Serres’ occupations?) Most important of all, if Antoine Moleri is a heretic, can we make the assumption that everyone in the Moleri family is also heretical? In this last case, the answer has to be yes, because otherwise it would be impossible to compare Protestant to non-Protestant families.

Unfortunately, if family can be a problematic term, so can elite. Elite can be defined by reference to power obtained through holding office, through wealth, through education, or through connections to other members of the elite. Defining the cutoff point between elite and non-elite can also be problematic: some families may be more elite than others. When Ann Guggenheim examined Nîmes’s elites, she looked at about seventy persons, generally officeholders and men of law.6 Defining elite status by occupation has certain distinct advantages: occupation is usually easy to determine. But it somewhat arbitrarily excludes some wealthy artisans, for example, who can be shown to have played an active role in town affairs even if most artisans did not. Instead of using occupation to determine status, this book has used the dowry measures discussed above: while wealth is not precisely equivalent to status, it is at least a less crude measure than occupation. In any case, because we will be able to examine a much larger number of families, the overwhelming majority of lawyers and officeholders can be included.

Once I calculated a “social status by event” number for each individual, I selected a smaller group for closer analysis, specifically men with a social status number of 100 or greater and women with a social status number of 80 or greater. The choice of 100 to define the elite was deliberately generous: remember that, using the occupational averages as a guide, I suggested that persons with a social status number above 250 ranked as wealthy, 125–250 corresponding to an upper-middling group, 50–125 indicated the middling, and below 50 was poor. To ensure that a social status number was not just a fluke, however, I required that there be at least four references to each man (since there are many fewer references to women, I did not apply that restriction to them). Very few prominent persons, as judged from town and church records, did not appear at least four times. It would be difficult, in any case, to assign individuals to different factions based on only two or three references to them.

After identifying all probable relatives of the initial sample, I obtained slightly more than 300 families. For technical reasons,7 an upper limit of 250 was desirable; I therefore (p.218) chose to include only families that met at least one of the following criteria: either an average social status of at least 140 or at least nine references. (Again, it is important to remember that this represents an extremely generous conception of what elite means.) Therefore, only families who were both relatively poor and had few references in the database were excluded.

Appendix BConstructing the Notarial Database

Figure A.8. Multiple mentions of family members. Source: ADG, Series IIE.

It is perhaps worth noting that social status information was originally collected for individuals: in order to convert it to rank families, it must be weighted. Consider the Borrellon family, which includes two brothers, one of whom has two children (see figure A.8). The two brothers, Arnaud and Jacques the travailleur, appear much more often in the database than do Jacques the ropemaker and Anne, who are Arnaud’s two children. How should the family’s social status be calculated? One could add the four social status numbers, 103.26, 101.57, 101.84, and 220.75, divide by four, and use that as the average. But in order to weight the family correctly, one must instead take into account how often each member of the family appears in the database.

The use of the two criteria ensured that only poorer families who also appeared relatively infrequently were excluded. Families that contained a notary, a priest, or a Protestant minister were automatically included (clergy because of the nature of this study, notaries in order to ascertain whether faction allegiances could have affected the sample under study). This created a sample of manageable size, with just under 3,000 references to 249 families. Each family was then given a unique identifier, usually the first eight letters of its last name, except for a few cases where there was more than one family with the same name, in which case a number (“Arnaud1” and “Arnaud2,” for example) was added. Once all the elite families were identified, families with at least one Protestant member could be compared to non-Protestant families.

Once we have identified a group, if we can determine a way to measure connections between its members, we can measure its cohesiveness, analyze its factions, and so forth. Fortunately, the notarial records in the database were ideal for this, because (p.219) wills and marriage contracts are, by definition, documents that link individuals in webs of family connections: either one generation to another or one family to another. Furthermore, it is well known that witnesses to these acts were commonly friends of the parties, as indeed the case of the Rozel and Barrière families (see Chapter 6) shows. Because each individual was likely to get married only once or twice, while he might be a witness to friends’ marriages and wills fifteen or twenty times, in the end witnessing proved to be a much more powerful tool for measuring family connections.

The mechanics of measuring family connections were reasonably straightforward. Each family had a unique identifier, and so did each event (will or marriage). The unique identifier for each act was the volume and page number where it is located in the archives. Each record therefore could be linked in two directions: to other references to the same family and to other families who were present at the same event. The 3,000 references to members of elite families were divided, therefore, into two groups: (1) principals, that is, husbands, wives, and testators; and (2) witnesses. Each witness family was presumed to have a tie to each principal family if they appeared at the same event. The database program (Microsoft Access) was then programmed to create sets of pairs of families linked by their co-attendance. These pairs were then exported to a spreadsheet program (Microsoft Excel), which created a pivot table, where the sets of pairs were converted into a matrix, where all the families were listed at the top and left-hand edges, and a number was put at their intersection points (1 for one tie, 2 for two, etc.). It should also be noted that the mathematics of network analysis requires symmetrical matrices for many important functions and that the sets of pairs of principals and witnesses are not symmetrical, because the two lists from which the pairs were drawn, principals and witnesses, are by no means identical. To deal with the problem, each pair was entered twice, that is, in both possible orders, before being exported to the matrix. At that point, network analysis became possible, using any standard program; in this case, I used UCINET IV, which created factions. Some basic statistics about the factions are summarized in figure A.9. Social status numbers above 250 (that is, based on average dowries of over 250) indicate that all of the factions were wealthy on average, but of course that means that many of the members were only upper middling in status.

Appendix BConstructing the Notarial Database

Figure A.9. Basic information about Nîmes’ elite factions. Source: ADG, Series IIE.



(1) . Thus, Colette Sardinoux, in her M.A. thesis, “Les premières traces de la Réforme à Anduze,” was surprised to note that Protestant wills tended to clump in certain registers, not realizing that she was looking at drafts (brouillards). Usually, brouillards will be noted in the printed inventory (Gouron, Répertoire numérique des Archives Départementales du Gard), but not always. I would like to thank Professor Gabriel Audisio for warning me of this probem. For Lansard and Alirand, see Gouron, Répertoire numérique des Archives Départementales du Gard, 330, 334.

(2) . On the issues involved in coding last names, see Harvey and Press, Databases in Historical Research, 228–31.

(3) . ADG, IIE1 248, fol. 293; IIE36 322, fol. 336.

(4) . ADG, IIE1 233, fol. 196, 18 September 1551.

(5) . Puech, “Débuts” and Une ville.

(6) . Guggenheim, “Calvinism,” 2. She states that the elite was composed of “less than thirty-five active policy-makers at any one time and about thirty more persons serving in advisory capacities.” In her table of the religious sympathies of officeholders in 1560 (221), she includes thirty-one persons; in an appendix (343–44), she adds thirty-nine more.

(7) . Family ties had to be codified into a matrix for export to UCINET using a pivot table, which creates a square matrix. For this purpose, I used Microsoft Excel, which permits an infinite number of rows, but only 256 columns, with the practical effect of limiting the matrix to 256 x 256 families.