1. Invisible but prevalent?

6 Beginning in the 18th century, France experienced a first demographic transition that meant ‘that no generation born after 1870 had more than 2.7 children (on average) per woman still living at age 50’ (Leridon et al., 1987, p. 7). While many demographers and historians have examined the modalities of this transition, they seem mainly to have focused on the reasons why individuals chose to limit births: as summarized in the famous article by Dupâquier and Bardet, ‘the question of why is much more important than that of how’ (1986, p. 11). The nature of the methods used remains even vaguer since ‘the limitation of births does not elicit any explanatory discourse from those doing so’ (Bardet and Le Bras, 1988, p. 359). Articles addressing the ‘how’ are limited to those referring to the literature for suggestions (van de Walle, 2005). Nevertheless, withdrawal emerges as the most likely answer. Corbin qualified coitus interruptus as ‘the most widespread contraceptive technique in France’ [3] (1986, p. 269) and noted that since the early 19th century, ‘the prevalence of coitus interruptus is now documented in many regions’ (2008, p. 336); Flandrin labelled it ‘the primary method of contraception in the modern era’ (1981, p. 114). Dupâquier and Bardet also asserted that ‘it is generally believed, though not proven, that most couples mainly used coitus interruptus, the most natural and least dangerous contraceptive method’ (1986, p. 8). But even an imprecise assessment of the actual use of this most widespread of ‘dark secrets, unknown to any animal but man’ (Moheau, 1778, Livre II, p. 102) remains challenging:

7

In all likelihood, coitus interruptus was the most widely practised method, but even this hypothesis is fragile; it is based on various ambiguous references gleaned from administrative files, medical reports, or elegantly phrased ecclesiastical discussions. In short, we have no statistics on the practice of this most basic method of contraception. (Bardet and Le Bras, 1988, p. 359)

8 The use of the withdrawal method from the early 20th century until the 1960s (before the legalization and distribution of medical contraception) has been attested through historical research based on sources such as judicial archives (Sohn, 1998) and the analysis of letters and intimate diaries (Rebreyend, 2006), with no accurate statistics yet available. However, since the 1960s, various surveys have started to put figures to the use of this method.

9 The first that might be mentioned is a study cited in La Seconde révolution contraceptive and based on surveys conducted in maternity wards in 1961–1962 in Lyon and Grenoble by Dr Jean Sutter. Though not representative, these surveys provide ‘an idea of the relative popularity of each method’: 74% of women surveyed cited withdrawal as the method used before their most recent pregnancy (61% alone, 13% in combination with the rhythm method), 10% the rhythm method or temperature method, 4% condoms, and 12% another method or combination of methods. As the authors emphasized, these results are ‘hardly surprising’ and confirm a ‘monopoly’ unrivalled by methods of periodic abstinence or condom use (Leridon et al., 1987, p. 62). Secondly, a study was conducted in 1965 by the Christian magazine Clair Foyer [4] on ‘Households and Birth Control’, subsequently published under the title 3 000 foyers parlent. Une enquête de Clair Foyer sur la régulation des naissances (3,000 Households Speak Out: A Clair Foyer Survey on Birth Control; Lambert and Lambert, 1966). Of the 2,700 responses, a random sample of 2,000 were transposed onto punched cards and used to produce various statistics, clarified by the authors as needing to be considered ‘in relation to the Clair Foyer readership’ and as not constituting ‘a true survey’ (Lambert and Lambert, 1966, p. 290). According to this study, 52% of the sample (most respondents were between 25 and 44 years old) reported using ‘pulling out’ (alone or in combination with another method), 42% the temperature and rhythm methods, and 12% condoms (Lambert and Lambert, 1966, p. 291). Both these studies illustrate the popularity of the withdrawal method, but the surveys used quite particular survey methods (in hospital after a pregnancy, anonymous letters sent to a magazine) to reveal its prevalence.

10 The 1970s saw the advent of the first large-scale surveys on contraception and sexuality. However, their representative nature would be undermined by the obstacles to individual reporting of this method.

2. Church, medicine, and sexuality: the construction of stigmatization

11 Since the very first surveys on the topic, which coincided with the circulation of medical contraception, it has proved difficult to capture the practice of non-medical methods of contraception. In 1971, the first large-scale survey on contraception [5] came up against the fact that ‘the public, still poorly informed in 1971, preferred to report that they were not using any form of contraception at the time of the survey rather than to report traditional contraceptive practices such as withdrawal (35% of women at risk [of pregnancy] were in this situation)’, since the practice was deemed ‘not worthwhile’ (Collomb, 1979, p. 1047). Even admitting to the use of contraception may have been problematic for some people, in a context where religious practice was still important. The Catholic Church’s prohibition of any method except for periodic abstinence [6] was confirmed by the 1968 encyclical Humanae Vitae, which specifically forbade the use of any medical methods of contraception (Sevegrand, 1995). But opponents of the Church and campaigners for the legalization of contraception were also opposed to the withdrawal method: the French Family Planning Movement encouraged only the use of so-called modern medical methods, such as the diaphragm and subsequently the pill (Pavard, 2012). Pressure from these two camps—religious and medical—contributed to the stigmatization of withdrawal, in the sense that a stigma always results from an interaction that discredits the person with the stigma that was until that point only discreditable (Goffman, 1963). In the case of withdrawal, this stigma does not operate in all social spheres. Initially, it emerged within the religious sphere, but, over the longer term, it was primarily the medical sphere that would contribute to delegitimizing this practice deemed unreliable and no longer necessary, perhaps particularly so because it lay outside the control of doctors (Roux, 2020). As contraception rapidly became medicalized (and feminized) in France (Toulemon and Leridon, 1991), this medical stigma gradually became more permanent, rendering the withdrawal method something not to be mentioned to the doctor and, ultimately, something not to be mentioned at all.

12 But the withdrawal method is not the only contraceptive method affected by the stigmatization that takes place during discussions with healthcare professionals specifically, even if the reasons for it may be different. First, from a closely related perspective, it also affects other so-called natural methods of contraception, including periodic abstinence, such as cycle-tracking or the Knaus–Ogino (rhythm) method. The latter has been nicknamed ‘Vatican roulette’ (Régnier-Loilier, 2007, p. 31) and, in France, babies conceived despite the use of the rhythm method are referred to as ‘Ogino babies’. The external (or male) condom has also suffered from a poor reputation for many years, initially because it was associated with prostitution (Barrau and Perrin, 1966) and subsequently because, with the emergence of medical methods, its status as a barrier method, and one that involves a pause in the sexual act to put on, makes it seem very old-fashioned (Thomé, 2016). It was the HIV-AIDS epidemic in the 1980s that would make its use more commonplace (Toulemon and Leridon, 1995; Rossier et al., 2004). This stigmatization of certain methods also extends to another practice aimed at regulating births, abortion, the stigmatization of which extends far beyond the medical sphere. While initially criminalized (Cahen, 2016), since its legalization (in France) in 1975, it is gradually becoming a ‘legal deviance’ (Divay, 2004). The stigmatization of pregnancy termination by the medical establishment (Mathieu and Ruault, 2014) has a strong impact on women, to the point that they try to conceal or ‘invisibilize’ the act among their close contacts (Thizy, 2021). This attitude makes it difficult to measure and, again, requires the development of solutions for avoiding its under-reporting (Guillaume and Rossier, 2018; Lindberg et al., 2020). We can see, then, that the challenges faced when measuring use of the withdrawal method should be viewed within the context of a more general stigmatization by the medical establishment of all methods considered ‘traditional’ in France, and of abortion, while ‘modern’ methods are promoted.

13 This stigmatization of withdrawal, the gradual emergence of which can be traced, persists and impacts its reporting in surveys often related to public health and, therefore, to the medical domain. The withdrawal method continues not to be considered an actual method (Jones et al., 2009), as clearly demonstrated in the rare qualitative surveys on the topic, in France and elsewhere. For example, a study conducted in Türkiye in the early 2000s showed that out of 20 men who ultimately reported using the withdrawal method during face-to-face interviews, only three of them had actually reported doing so in the preliminary questionnaire about their contraceptive practices (Ortayli et al., 2005). Likewise, during interviews on sexuality and contraception conducted in France in the late 2010s, it emerged that the respondents only reported using the withdrawal method when the question was explicit, rather than at the start of the interview when they were asked to describe all the various methods and techniques they had ever used (Thomé, 2019). This is primarily because the medical establishment still plays a significant role in delegitimizing the method (Jones et al., 2014); various respondents explicitly apologized for having used it, thereby admitting that they were conscious of contravening the norm of medicalized contraception (Le Guen et al., 2017). But we should also factor in a sexual dimension that can complicate its reporting. Withdrawal directly impacts the sex act, the penetration of the vagina by the penis, by interrupting it: this means that talking about it evokes an intimate sexual technique that is much more difficult to discuss than taking a physician-prescribed drug (Thomé, 2022). While addressing the topic might be expected in the context of a survey on sexuality, it is more problematic when the respondents are only expecting to answer questions considered purely medical.

14 Invisible for a long time, with the promotion of medical methods of contraception, withdrawal is becoming a stigmatized practice. This has a dual impact on reporting: first, it is not considered a genuinely legitimate method, and respondents do not necessarily think to report it when questioned on contraception; secondly, given its stigmatization by modern medicine, its use might even be concealed, at least initially, through fear of potential judgement by the interlocutor. However, various solutions for overcoming this twofold issue emerge from a closer look at French surveys and the international literature.

1. Asking the right questions: wording, vocabulary, and repetition

17 The crucial factor for a correct measurement of the withdrawal method is the way the question is phrased. Since the 1971 survey, the limitations of an open question have been demonstrated: ‘In response to the question “As far as you know, what can be done to avoid a birth?”, the pill was spontaneously cited as a contraceptive method with a frequency of 41% (6% for the withdrawal method)’ (Collomb and Zucker, 1971, p. 67).

18 It appears essential to name the method in order for it to have a chance to be reported (as is the case in the rest of the questionnaire), though this still fails to fully remove a bias in favour of medical methods, leading to the under-reporting of ‘traditional contraceptive practices’ that might ‘seem less respectable in the context of more modern practices or not to constitute a real “form of contraception”’ (Collomb and Zucker, 1971, p. 66). This is particularly important in a context where withdrawal is, by far, the most widely used method. Moreover, the same type of problem can be found in contemporary surveys conducted in different parts of the world, for example (Fruhauf et al., 2021). The Demographic Health Surveys, [7] standardized surveys conducted in many low- and middle-income countries, underestimate the use of ‘natural’ methods (withdrawal or calendar-based contraception) because the general questionnaire only frames the question in an open way (Rossier and Corker, 2017).

19 In France, the 1978 survey included introductory text that encouraged respondents to report all methods used ‘to space out births’, emphasizing that ‘almost all couples do something in this regard (even if that is simply abstaining from sexual intercourse, for example) as otherwise they would all have 8 or 9 children’ (Sardon, 1986, p. 77). Reference might also be made within the question itself, as in the FECOND 2013 survey, which explicitly references non-medical methods: ‘Do you or your partner use any method of avoiding pregnancy, including any natural method?’ This is a way of endorsing the legitimacy of reporting certain methods. The same type of strategy can also be used to measure use of abortion as accurately as possible (Mueller et al., 2022).

20 Another way of improving measurement of the withdrawal method is to ask directly about its use. To do this, however, the method has to be given a name, which may itself be problematic—and may have contributed to its invisibility. For a long time, it was not referred to using any precise term, as demonstrated by research based on both archive and oral history data (Thomé, 2019). This emerges in the variety of vocabulary used to refer to it, which tends to be metaphorical (Bozon, 2001). Certain expressions are particularly allusive or even euphemistic, and their meaning is only evident within a very precise context, such as, in French, faire attention (be careful) or se gêner (hold back). Other expressions, though, refer to the necessary interruption of the sex act required by this method: ne pas faire son affaire complètement (not quite finish), ne pas aller au bout (not go all the way), interrompre (stop), or the more colourful sauter en marche (jump off). Such phrases are used across the world with local variants. Historians Szreter and Fisher, who conducted an interview-based survey on sex lives in the town of Blackburn, England, in the first half of the 20th century, reported that the accepted expression was ‘get off at Mill Hill’, from the name of the last suburban stop before the central station [8] (2010, p. 239). In another register, we find, in Spanish, the term cuidar (to take care of): el me cuida (he takes care of me) refers to both the withdrawal method and periodic abstinence (Hirsch and Nathanson, 2001, p. 420). Certain expressions are derived from slang, such as the French phrase arroser le persil (literally, to water the parsley); this expression originally meant ‘make love’, but its meaning changed in the 1960s to mean either to ejaculate early, or, as potentially happens with the withdrawal method, to ejaculate ‘into the pubic hair’ (Amerlynck, 2006, p. 144). The terms retrait (withdrawal) and coït interrompu (coitus interruptus) are only rarely used.

21 When surveys list various methods to avoid individuals omitting them, the process is described in full: ‘interruption of intercourse before ejaculation’ (1971); ‘practising withdrawal, i.e. withdrawing early (before ejaculation)’ (1978, 1988); ‘withdrawal by the partner before ejaculation’ (2000, 2010). The method is described precisely and explicitly so that each respondent can recognize whether they use it; it would also be useful to determine which phrasing produces the most accurate reporting of withdrawal, something not currently possible given the disparity of the surveys. In any case, highlighting non-medical methods gives them some visibility and, therefore, legitimacy at the point of reporting. This could be further improved by distinguishing between medical methods and non-medical methods, and then listing all the methods of each type. A recent American study opted for this solution and positioned withdrawal at the top of the list of non-medical methods, based on the hypothesis that ‘when a single, long list of methods is provided, respondents might stop after ticking methods considered to be “real” or “effective”’ (Jones et al., 2014, p. 417). This method resulted in much higher reporting of the practice than in previous surveys.

22 Lastly, when methods are not listed, double-check questions need to be included, as observed by Sardon based on the 1978 survey:

23

After asking the survey respondent the question ‘Do you and your husband currently use any method to avoid having a child?’, we then, if the response was negative, asked a second question, worded ‘So neither you nor your husband takes any precautions?’ This simple repetition increased our estimate of the number of women using some form of contraception by 25%. The additional information obtained with this second question is clear evidence of the difficulties associated with communication in this area. In almost all cases, the method overlooked when answering the first question was withdrawal, which many women do not consider a form of contraception, perhaps because it is actually practised by the man and does not involve any mechanical means. (Sardon, 1986, p. 78)

24 This double-check question does not refer directly to the withdrawal method, but it aims to capture its practice through the euphemistic expression ‘take precautions’. The intention was clearly to use phrasing similar to the respondents’ vocabulary and to avoid referring outright to a practice that respondents may have never actually named and that they do not really consider a ‘method’ as such, as referred to by the questionnaire. While the same approach was used in 1988, more recent surveys have opted to refer to the method explicitly (in FECOND 2013: ‘You/your partner pull/s out before ejaculation’). These double-check questions still appear essential when only a single, open question is asked; moreover, we see this in very different contexts, such as in contemporary Burkina Faso, where the use of additional questions allows much more accurate measurement of use of non-medical methods than the standard DHS open question concerning the contraceptive method used (Rossier et al., 2014).

25 To achieve the most accurate reporting of withdrawal use, the method needs to be listed using the clearest possible vocabulary, or, where an open question is used, at least one double-check question should be routinely included. Where methods are listed, distinguishing two ‘sets’ of methods (medical and non-medical) is particularly relevant in that it reinforces, for respondents, the legitimacy of reporting non-medical methods.

2. Situating withdrawal within the ‘mosaic’ of methods

26 Another reason why measuring use of the withdrawal method can be problematic is that, these days, it is rarely practised alone as the principal method of contraception, unlike most medical methods (Gray et al., 1999; Jones et al., 2009). Studies from the United States looking at withdrawal show that the specific nature of this practice needs to be considered in surveys:

27

Rather than asking respondents to choose from a list of contraceptive methods, researchers could probe about use of withdrawal (and, perhaps, other coitally dependent methods) for each time period under investigation (e.g. ‘And did you use withdrawal during that month?’ or ‘And can you tell me which months in that year you used withdrawal?’). It is likely that many couples use withdrawal inconsistently, or in combination with other methods, and asking questions such as ‘When you and your partner have vaginal intercourse, about how often does/do he/you “pull out” or “withdraw” before ejaculating?’ would help further clarify people’s practices with this method. (Jones et al., 2009, p. 409)

28 Moreover, withdrawal can also be given more or less emphasis when the collected data are processed and the results presented. Both the surveys themselves and the publications based on them are focused on the idea of ‘principal method’, dividing the users of the various methods into different groups that total 100% (for FECOND 2013, see, for example, Bajos et al., 2014). There are two ways of determining ‘principal method’, and neither of these favours an accurate assessment of withdrawal use.

29 For the first way, a hierarchy of methods is defined within the context of the survey, and the method in the highest position within this order of precedence is designated as the ‘principal’ method. In the 1978 survey, where a combination of methods was reported, the one used in the middle of the cycle (i.e. when risk of undesired pregnancy is highest, and the most reliable method is therefore used) was considered the principal method; where multiple methods were reported mid-cycle, the one considered the principal method was the first one to appear from the following list: (1) pill; (2) intrauterine device; (3) diaphragm, cap, spermicidal foam, or pessary; (4) lavage, injection; (5) male condom; (6) withdrawal; (7) abstinence, periodic abstinence; (8) abortion (Leridon et al., 1987). Withdrawal only appears 6th on the list, reducing its likelihood of being reported.

30 For the second way of determining principal method, found in more recent surveys, respondents are asked to indicate which one of the methods is used and considered the principal method; however, as we have seen, withdrawal is rarely considered a legitimate method, or even as a method at all. When used in combination with another method, therefore, it is rarely referred to as the principal method, and this is the case in contexts as varied as the United States (Jones et al., 2014) and Ghana (Marston et al., 2017). For France, taking the example of the FECOND 2010 survey, we see that in men who use contraception, while only 2.3% use the withdrawal method alone, the figure is more than 3 times as high (7.7%) if we include use in combination with the pill or intrauterine device (0.4%), condom (1.4%), or periodic abstinence (3.8%) (Le Guen et al., 2015).

31 The withdrawal method should, therefore, be thought of as part of a set of methods that can be used in combination by individuals, potentially during certain limited periods (e.g. the post-partum period or between two medical contraceptive methods) or only at certain points in the cycle or even during certain cycles. The idea of a ‘mosaic’ of methods enables us to consider both the timing and combination of methods:

32

The combination [of methods] used in each cycle might vary, over time forming a ‘mosaic’ of methods. The pattern over time (i.e. the combination of combinations) is what we refer to as the ‘mosaic’, a term we use to distinguish this type of pattern from simple combinations of methods that might be used in a given menstrual cycle. (Marston et al., 2017, p. 114)

33 To assess the actual use of withdrawal (and other non-regular methods, such as emergency contraception), it could be helpful to ask survey respondents not only about the method ‘currently’ used (FECOND 2010, 2013) but about methods used over a longer period (e.g. the last 4 weeks or even the last 12 months).

34 As a method lacking legitimacy, withdrawal risks disappearing behind other methods, even when it is reported. To measure this practice as accurately as possible, we need to ensure that we are not influenced by the hypothesis of its quasi-total disappearance, at the risk of artificially increasing the latter. In practical terms, one possibility would be to list the least legitimate methods at the top, when lists are given; another would be to look at the time-related aspect of questions on contraceptive practices. This second option might also prove a fruitful avenue within the context of large-scale surveys when it is not possible to reconstruct precise contraceptive trajectories, which would be too costly.