1Among the rules followed by future spouses with regard to their marriage, there is a whole group that can be studied by means of demographic statistics, namely those which determine the time of year chosen for the wedding.

2After describing French customs in the matter, the author examines how they have evolved since the early nineteenth century. Certain practices, such as those linked to the Catholic rules on marriage, are tending to disappear, while others are becoming rapidly more widespread; the absence of weddings in May and September, for example, is a phenomenon dating from the last century. The emergence of such customs in the nineteenth century reveals a little-known aspect of French folklore.

3In a recent article on the relations between sociology and demography, J. Stoetzel revealed, in these very pages, [1] the artificial nature of the boundaries customarily drawn between these two intellectual disciplines, and easily convinced us of the need for these two sciences to provide each other with mutual support. Indeed, while demographic phenomena depend to a certain extent upon biological factors, they also possess a human character which means that acceptation of the rules they follow entails membership of a social group. Hence, a purely biological, genetic or physiological “explanation” of population phenomena is not sufficient; they must also be “understood” and their social meaning identified.

4And there is perhaps no phenomenon more charged with meaning than marriage, and it would not be difficult for a sociologist or historian to find, in the rites surrounding the current celebration of marriage in France, numerous vestiges of ancient customs passed down from the Celts, Romans or others. While it is true that events surrounding birth and death often give rise to similar observations, in marriage there is a circumstance that is not found elsewhere: while our date of birth or death is dictated, in part at least, by nature, we are free to choose our wedding day. This article examines the rides governing this choice in France. We shall see that some of them have an evident religious origin, while others, on the contrary, have no obvious explanation.

5We will not seek to give an opinion here, but will leave it up to sociologists and historians to answer the questions that we raise.

6*

7The information at our disposal consists of monthly marriage statistics published each year by the French statistical office for the entire country and for each département. A quick glance at these statistics immediately reveals that numbers of marriages vary substantially from one month to the next and that the months when weddings are most frequent, for example, are the same each year. Clearly, marriages are seasonal, and the task we have set ourselves is to identify this seasonal pattern. We shall apply a standard method; it involves calculating each year, over a period of 10 years for example, the monthly distribution of a fixed annual number of marriages, generally 12,000. For a given month, the numbers obtained generally vary little from one year to the next, and their average is taken to be characteristic of the month and period studied.

8The set of 12 monthly averages calculated in this way represents the seasonal pattern we are seeking. [2] Graph 1 illustrates the process applied to the whole of France over the period 1927-1938. If we exclude the months of February, March and April, the results obtained are satisfactory and Graph 2 reveals the seasonal pattern of marriages from May to December and January. Over this nine-month period, May and January are the least popular months for marriage, and October the most popular. However, for the months of February, March and April, the curves of Graph 1 show much larger fluctuations than the curves of the other months, and a reason for these differences must be found.

Figure 1

Proportion of marriages registered each month for 12,000 annual marriages (1927-1938)

Proportion of marriages registered each month for 12,000 annual marriages (1927-1938)

Figure 2

Seasonal pattern of monthly change in numbers of marriages from May to December and January for 12,000 marriages in the year

Seasonal pattern of monthly change in numbers of marriages from May to December and January for 12,000 marriages in the year

9We will have little difficulty in explaining them if we bear in mind that Lent occurs during these three months, and if we recall that, in principle, the Catholic Church prohibits its followers from marrying during Lent. We will thus demarcate a period from February to April when marriages are rare, bounded by two periods during which they are more frequent. For a given month, among the three months considered, the proportion of marriages with respect to the annual total of 12,000 depends on the number of days in which marriages are rare, a number which itself depends upon the timing of Lent in the year concerned. As this timing varies substantially from one year to the next, the proportions used to plot the curves for February to April on Figure 1 are also very variable, and this explains their irregularity.

10Let us look more closely at the consequences of Catholic legislation on monthly marriage statistics. As well as celebrating marriage, the Catholic wedding ceremony includes a special Mass, known as the nuptial Mass, during which the priest pronounces the solemn nuptial blessing before the newlyweds. While marriages can be celebrated at any time, the nuptial Mass and accompanying blessing cannot take place during this “closed period” without a special dispensation.

11At present, the closed periods include Lent and Easter, and the period of Advent before Christmas. It is Lent which interests us here. This period today begins on Ash Wednesday, 46 days before Easter Sunday, and ends on Holy Saturday, the day before Easter. The period when marriage is prohibited without special dispensation thus lasts for 47 days, beginning anywhere between 4 February and 10 March and ending between 22 March and 25 April. We see that only the month of March can be wholly contained within this period. From 1927 to 1938, this occurred six times over the period, in 1928,1929,1931,1933,1934 and 1936. On Graph 1, the points corresponding to these six years are marked by a small circle. We see that by limiting ourselves to these six values, we obtain for the proportion of marriages observed in March (for 12,000 marriages in the year) a degree of constancy comparable to that of the proportions observed in the other months of the year from May to December and January. If we count working days only, [3] the overall daily average is 19.9 marriages, and we will assume that this average applies to any working day during Lent.

12*

13Now let us consider a random year, 1936 for example, in which 1,535 marriages were celebrated in April, for an annual total of 12,000. In that year, Easter Sunday fell on 12 April, so the month of April 1936 included 10 working days during Lent and 16 working days after Lent. In the 10 Lenten days, approximately 10 x 19.9 = 199 marriages took place, and in the 16 days after Lent, the number was 1,535 – 199 = 1,336. The same reasoning can naturally be applied to years other than 1936 to calculate the number of marriages during periods containing a variable number of working days after Lent. The result of these calculations is shown in Graph 3 (Curve I), and enables us to state the following:

1. Very few marriages take place on the first working day following the end of Lent, i.e. Easter Tuesday;
2. During the nine following days, the aggregate number of marriages increases more or less proportionally to the number of days, by around 120 per day;
3. The aggregate number of marriages then continues to increase in a linear manner, but more slowly, by around 30 marriages per day.

Figure 3

I. Number of marriages observed over a period of n working days after the end of Lent for 12,000 marriages observed in the year. II. Number of marriages observed over a period of n working days before the beginning of Lent for 12,000 marriages observed in the year

I. Number of marriages observed over a period of n working days after the end of Lent for 12,000 marriages observed in the year. II. Number of marriages observed over a period of n working days before the beginning of Lent for 12,000 marriages observed in the year

14What we did for the period following Lent can also be done for the period preceding it to calculate the number of marriages observed over periods comprising a variable number of working days before Lent This is shown in Curve II of Graph 3. We can make the same remarks as for Curve I, although no day comparable to Easter Tuesday is found, and the period of rapid increase is limited to just three working days.

15The information provided above enables us to plot the curve of the daily number of marriages during Lent and on either side of it (Graph 4). This curve reveals the existence of two periods on either side of Lent when marriages were frequent, although the peak is less pronounced in the pre-Lent period than the post-Lent one. In other words, couples who respect the rules of the Catholic Church prefer to delay their wedding rather than bring it forward.

Figure 4

Daily number of marriages in the days around Lent, for 12,000 marriages in the year

Daily number of marriages in the days around Lent, for 12,000 marriages in the year

16If Easter Sunday, Monday and Tuesday and the two periods of frequent marriage are added to the 46 days of Lent, and taking account of Sundays, we obtain a period of 63 days that we will call the “wedding disturbance” period.

17Let us examine the effect of moving this disturbance period on the proportions f, m, and a that we calculated respectively for the months of February, March and April [4] in order to plot Graph I. Diagram 5 will facilitate this analysis. A chronological scale from February to April is given on the x-axis and an analogue scale for finding the date of Easter on the y-axis. For a given date of Easter, 12 April for example, the disturbance period is located at the points A, B, C and D of our diagram. The segments AB and CD correspond to the two periods of frequent marriage surrounding the period BC of infrequent marriage during Lent.

Figure 5

Position of the disturbance period with respect to the months of February, March and April, according to the date of Easter (does not apply to leap years)

Position of the disturbance period with respect to the months of February, March and April, according to the date of Easter (does not apply to leap years)

18For a given month, say February, we will distinguish three periods of durations t₁, t₂ and t₃ (t₁ + t₂ + t₃ = 28 days) such that the first, represented on the diagram by the segment A′A, has a normal marriage frequency, the second, represented by AB, has a high marriage frequency and the third, corresponding to BB′ has a low frequency.

19When Easter shifts from P₁ (22 March) to P₂ (23 March) t₁ is nil, t₂ increases by one day and t₃ decreases by the same amount, so f increases.

20When Easter shifts from P₂ (23 March) to P₃ (16 April) t₁ increases, t₂ remains constant and t₃ decreases, but more slowly than in the first case.

21When Easter shifts from P₃ (16 April) to P₄ (20 April) t₁ increases, t₂ decreases and t₃ is nil, so f decreases, but the decrease is slower than the increase in the first case.

22Last, when Easter moves from 20 April to 25 April, t₂ and t₃ are nil and t₁ remains constant, so f does not vary. The month of February thus has a normal marriage frequency. [5]

23This is how Graph 6 was constructed. We see that the proportions actually observed from 1927 to 1938 are positioned on this graph very close to the places that we had assigned to them. Hence, knowledge of a sociological fact - observance of Lent by dutiful Catholics - enables us to find order in what, on Graph 1, we first saw as chaos.

Figure 6

Variation, according to the date of Easter, in the proportion of marriages in February, March and April, for 12,000 annual marriages

Variation, according to the date of Easter, in the proportion of marriages in February, March and April, for 12,000 annual marriages

24Some may find that the qualifier “sociological” is somewhat exaggerated; no great sociological expertise was needed to work out this problem. But if the task appears simple, it is doubtless because this custom is an integral part of our current civilization, and we belong to a population that still honours it. But were the Church to change its rules, and lift the ban on marriage during Lent for example, then the demographer of the year 3000 would be very puzzled by the monthly marriage statistics of our current times. Perhaps, with luck, he might find a correlation between the phases of the moon and the fluctuations in marriage frequency from February to April, thereby concluding that in 1936, French people chose their marriage date according to the lunar calendar. But if there is no sociologist on hand to tell him the true reason underlying this choice, he is unlikely to discover the religious origin of the variations he has plotted.

25On Graph 6, the EF part of the February curve corresponds, as we have indicated, to the normal marriage frequency observed a few days before the start of Lent. We can doubtless assume that in the absence of Lent, we would continue to observe this marriage frequency. Likewise, the horizontal part of the March curve corresponds to the average marriage frequency before Lent. Consequently, the ratio of the ordinates δ and Δ represents the proportion of the population that does not observe the marriage rules of the Catholic Church. The ordinates of points G and H, which correspond to the years 1930 and 1935, when Easter was on 20 and 21 April, respectively, give us values that approach Δ. These remarks enable us to construct a particularly useful indicator.

26Indeed, if we draw regional graphs similar to the one we have just produced for the whole of France (Graph 6), we obtain figures which vary considerably from one region to another. The flatter the curves of each month, the less strict the observance of Catholic marriage restrictions during Lent. Graph 7 which corresponds to the 10 départements of Brittany and Graph 8 to the group Yonne, Aube, Seine-et-Marne and Marne, provide a striking illustration. We can thus calculate for each département the quantity p = 1 – δ/Δ which represents the proportion of inhabitants who observe Catholic marriage rules during Lent.

Figure 7

Variation, according to the date of Easter, in the proportion of marriages in February, March and April, for 12,000 annual marriages

Variation, according to the date of Easter, in the proportion of marriages in February, March and April, for 12,000 annual marriages

Figure 8

Variation, according to the date of Easter, in the proportion of marriages in February, March and April, for 12,000 annual marriages

Variation, according to the date of Easter, in the proportion of marriages in February, March and April, for 12,000 annual marriages

27Map 9 shows the results of the calculation for the period 1927-1938. We see that the Catholic legislation is strictly observed to the south-west of a line Caen - Saint Etienne, while to the north-east of this same line, it carries much less weight. We see a rough outline of the French regions where religious practice is traditionally strong. But the place occupied by the north of France, generally considered to be a region of fervent churchgoers, is surprising. This is perhaps because observance of Lent does not have, or perhaps no longer has, the religious importance that we tend to attribute to it. Hence, the lack of similarity between Map 9 and the map of the gross reproduction rate [6] (Map 10) can neither confirm nor refute the frequent assertions concerning the influence of religious factors on fertility. There is a clear correlation, on the other hand, between divorces (Map 11) and non-observance of Lent. Last, Map 12, similar to Map 9 but covering the period 1874-1878, illustrates the weakening of religious traditions governing the timing of marriage.

Figure 9

Percentage of the population observing Catholic rules on marriage during Lent (1927-1938)

Percentage of the population observing Catholic rules on marriage during Lent (1927-1938)

Figure 10

Gross reproduction rate per 100 women (1930-1932)

Gross reproduction rate per 100 women (1930-1932)

Figure 11

Marriages per 10 divorces pronounced (1936-1938)

Marriages per 10 divorces pronounced (1936-1938)

Figure 12

Percentage of the population observing Catholic rules on marriage during Lent (1874-1878)

Percentage of the population observing Catholic rules on marriage during Lent (1874-1878)

28The seasonal variations described above on Graph 2 for the months of May to December and January can now be determined for the entire year. Based on the above, the scale of variation will vary according to the date of Easter. For example, Graph 13 shows the seasonal variation for the whole of France corresponding to a year when Easter is on 1 April and the whole of March is in Lent.

Figure 13

Monthly seasonal variation in marriages for 12,000 marriages in the year and for a year when Easter is on 1 April

Monthly seasonal variation in marriages for 12,000 marriages in the year and for a year when Easter is on 1 April

29Let us now see what meaning we can attach to the irregularities of this variation. We note also that the peaks and troughs on Graph 13 are not independent. Marriages not celebrated at one time must of course be celebrated at another, so the troughs and peaks correspond and the explanation for one applies in turn to the other.

30*

31The troughs in March and the peaks in February and April are linked, as we have just seen, to the marriage legislation of the Catholic Church. The peak in December also has a religious explanation. We saw above that there is another period, in addition to Lent, during which the Catholic Church does not authorize marriage without dispensation, namely the period of Advent which begins on the first Sunday following 26 November and ends at Epiphany on 6 January. If this rule were strictly applied, we would see a dip in marriages as large as that of March. In reality, while the seasonal pattern calculated above (Graph 13) for the whole of France ranks December among the months when marriages are least frequent, it shows that the rule applicable to marriage during closed periods is less well observed during Advent than during Lent. In addition, as is the case for Lent, December should be flanked by two marriages peaks, representing periods when the marriages not celebrated during Advent were able to take place. Graph 13 shows that these peak marriage periods are found neither in January nor in December.

32The seasonal patterns by département give greater substance to this impression. They immediately reveal that the proportions of January and December are independent: while the January figure varies little among départements, that of December changes notably from one region to another. The dearth of marriages in January thus deserves a specific explanation, doubtless linked to the winter season. Likewise, the respective variations in the proportions of December and November are only weakly correlated, while those of October and December vary in more or less the same manner. Contrary to the situation in Lent, the marriages not celebrated during Advent do not appear to be shifted to neighbouring months, but rather brought forward to October. We may thus assume that in the absence of Catholic legislation, the marriages in the last quarter of the year would be equally distributed over the three months.

33The relative difference between the number of December marriages and the three-month average thus represents the proportion of the population that observes Catholic legislation on marriage during Advent.

34For the whole of France, we find a percentage of 17, a rather low figure which shows that the Catholic rule is not widely observed. However, if we calculate the same proportion by département, we find marked differences between the regions, visible on Map 14. This map shows the present day situation, and reveals that the départements of Brittany and above all Vendée are practically the only ones to observe the custom today. Indeed, Map 15, which concerns the period 1874-1878, reveals the progressive disappearance of this custom over the last 60 years.

Figure 14

Percentage of the population observing Catholic rules on marriage during Advent (1927-1938)

Percentage of the population observing Catholic rules on marriage during Advent (1927-1938)

Figure 15

Percentage of the population observing Catholic rides on marriage during Advent (1874-1878)

Percentage of the population observing Catholic rides on marriage during Advent (1874-1878)

35*

36We showed earlier that the proportions of December and October vary in a similar manner. This is only a general observation and there are exceptions to the rule. Certain départements, Creuse for example, which have a normal marriage frequency in December, also have a very high marriage frequency in October. This is because the Catholic rule on marriage in December exists alongside a custom to avoid marriage in November. Map 16 shows that this custom is highly localized, which explains why it is not visible in the pattern for the whole of France. The insets around the map enable us to pinpoint the origin of the custom in the Creuse département around 1874-1878. It developed gradually from then on, and in certain départements, such as Corrèze, it is quite recent, as shown by Graph 17. This same graph shows that the custom is now tending to spread to Gironde and Charente. The populations who accept it explain their attitude by their respect for the dead, to whom they devote the entire month of November. The commemoration, on 11 November, of the Armistice of the 1914-1918 war, may also have played a role in the spread of the custom between the two wars.

Figure 16

Percentage of the population who think that November is not a good month for celebrating marriage

Percentage of the population who think that November is not a good month for celebrating marriage

Figure 17

Percentage of the population who avoid marrying in November

Percentage of the population who avoid marrying in November

37For the months of July, August and September, four types of seasonal patterns can be observed across the different regions:

1. most départements have a peak in September and troughs in July and August;
2. in a small number of départements on the Mediterranean coast, the peak is in August, and the troughs in July and September;
3. in others, the situation is reversed, with a trough in August and peaks in July and September;
4. last, in the Seine département, the peak is in July and the troughs in August and September.

38Map 18 shows the regions where these various patterns are observed. Cases a and b can be explained by economic factors. In many regions, the whole of July and August are devoted to the harvest and marriages are delayed until September, a month when the farming workload is somewhat lighter for rural populations, even in wine growing regions where the grapes are harvested late in the year. In the Mediterranean coast, on the other hand, the grapes are harvested in September and marriages are brought forward to August.

Figure 18

Seasonal variation in marriages over the months of July, August and September (1927-1938)

Seasonal variation in marriages over the months of July, August and September (1927-1938)

39In the Seine département (Paris), these considerations do not apply, and the decrease in marriages in August and September is doubtless linked to the fact that many Parisians go away on holiday in summer.

40In the central-eastern and north-western regions, the explanation is less simple. We believe that the August dip in marriages has a religious origin, although the Catholic Church has not officialized the custom. Indeed, the month of August is devoted, in part at least, to the Virgin Mary, and, in certain regions, it is widely considered inappropriate to marry at this time. We can find confirmation of our interpretation in the fact that marriages are also very rare in May, a month which - even more so than August - is wholly given over to the Virgin Mary; it is the “month of Mary” for Catholic churchgoers. But the question of the month of May is less clear-cut, since numerous influences appear to play a role, though the weight of each is difficult to discern.

41Map 19 shows how the custom is observed today in the French population. It reveals two main areas, the Mediterranean coast on the one hand and the regions between the Loire and the Massif Central on the other.

Figure 19

Percentage of the population who think that May is not a good month for celebrating marriage (1927-1938)

Percentage of the population who think that May is not a good month for celebrating marriage (1927-1938)

42One might be tempted to see this as a custom inherited from Ancient Rome. Indeed, in Roman times, the month of May had a poor reputation. It was the time of the Lemuria festival, a sort of commemoration of the dead, and marriage was discouraged throughout the month. Speaking of the month of May, Ovid declared “Those foolish enough to light wedding torches, may well see them transformed into funeral torches” Many popular proverbs express the same idea today: “Mai commence par une croix, qui s’y marie en porte deux” (May starts with a cross, if you marry in that month you will carry two); [7] “Noce de mai, noce de mort” (May wedding, death wedding); “Mois des fleurs, mois des pleurs” (Month of flowers, month of weeping); “Au mois de mai ne se marient que les ânes” (Only donkeys marry in May); “Les enfants conçus en mai auront les yeux rouges et seront fous” (Children conceived in May will be red-eyed fools). But if this is a relic of Roman times, we should find ever higher levels of observance as we move back in time. The opposite is in fact the case, as shown by Maps 19 and 20. In 1810, a few départements in western France showed a slight tendency to avoid marriages in May. By 1837 the custom had gained strength and was spreading towards the centre, while a new area of observance emerged along the Mediterranean. By 1876, the two zones were clearly visible, and since then have never ceased to expand.

Figure 20

Percentage of the population who think that May is not a good month for marriage

Percentage of the population who think that May is not a good month for marriage

43This suggests that the custom did not exist in the early nineteenth century; to defend the theory of a Roman origin we would need to find out why a forgotten practice later re-emerged. Moreover, it is by no means certain that the custom disappeared a long time ago; only an analysis of monthly marriage statistics before the Revolution would allow us to clarify this point. Unfortunately, these statistics do not exist, and could only be obtained through parish registers.

44As things stand, it seems unlikely that this custom harks back to Roman times. A more plausible explanation is that, like for August, May weddings are avoided in honour of the Virgin Mary, whose cult has developed considerably over the last century. In any case, the current influence of this cult on May marriages is a certainty, as proven by the existence, in all regions, of religious groups of young girls known as “Children of Mary” whose members will never marry in the month of May.

45*

46Hence, a demographic study of marriages in May and November cannot be brought to its conclusion without the help of sociologists and historians. Demographers, for their part, can reveal the emergence, in the nineteenth century, of customs previously thought to take back to ancient times. They can show folklorists that science does not exist outside time and does not consist solely in looking for ancient practices on the verge of extinction. Ours is a living, constantly changing history, enriched each day with new findings, in which the popular imagination finds its expression.

47To conclude, may we express our gratitude to the departmental archivists of Bouches-du-Rhône, Cher, Corrèze, Creuse, Gard, Hérault, Indre, Indre-et-Loire, Var and Haute-Vienne who kindly explained to us why the population of their département avoided marriage in the month of May.