This article aims to compare the meanings of "Non Response" (NR)[1] and "Don't Know" (DK)[2] responses to questions asked through questionnaires in quantitative surveys on sexual behaviour.

NRs are usually treated by researchers as "obstacles to data interpretation" (Simon, Gondonneau, Mironer, Dorlen-Rollier : 1972), as "measurement errors" (Hyman, 1954) or as "one of the most complicated subjects in the theory of surveys (...) which are unpleasant for the investigator" (Ardilly, 1994, p. 253). Facing NRs, a statistician "finds himself [3], consciously or not, in a position to make hypotheses on the behaviour of populations (...) and to take up particular responsibilities in relation to that point" (Ardilly, 1994, p. 254). When NRs deal with large numbers of data, the pertinence of the question itself is challenged.

In methodological type approaches, NRs can be considered as being caused by the theme of a question, the way it is asked, the respondent's characteristics and those of the interviewer. On the contrary, sociological approaches that have attempted to overstep the limits of the methodological thought frame consider that NRs constitute an important source of information. Treating them as such allows reinforcing or qualifying interpretations elaborated from more substantial responses. The main argument put forward to study the meaning of NRs, notably in surveys on political opinions (Michelat and Simon, 1982), relates to their non-random character. NRs, for instance, have been considered as "indicators of various forms of social exclusion" (Francis and Busch, 1975). These responses therefore constitute full responses: they have a meaning worth interpreting in the same way as any other response to a given question.

DKs have another status. Whereas researchers are subjected to NRs, they can choose to integrate DKs in the construction of questionnaires in order to measure respondents' self-proclaimed ignorance (see : Bauer in this journal). Hence, they provide the respondents with an implicit honourable way out, allowing them not to answer a question without coming into conflict with the interviewers. By neglecting to offer a DK modality, researchers force respondents to answer. Omitting to offer a DK modality is entering a questioning strategy which can be defined as "imposing a problematic" (Bourdieu, 1973). A respondent can but fall back on a NR to express his/her ignorance.

NRs and DKs IN SURVEYS ON SEXUAL BEHAVIOUR

Surveys on sexual behaviour are expected to bring up higher NR rates than other surveys which are perceived as less threatening (Bradburn and Sudman, 1981). NRs are doubly defined as being a refusal to answer a question and the denial of a practice which was actually performed : "some respondents may find it more socially acceptable and less confrontational to report non-performance rather than to refuse to answer a question" (Catania, Gibson, Chitwood and Coates, 1990, p.342). Assimilating a NR to a respondent's denial of a specific sexual practice relies on the hypothetical attribution of a profile to the respondent from which one implicitly expects he will respond positively to a question. Even though it does not have much of a scientific basis, this assimilation contributes in these authors' argumentation to the favouring of strategies which reduce NRs instead of interpreting them as full responses.

To decide on the choice of the method they would use, ACSF[4] investigators (Spira, Bajos and ACSF Group, 1994) carried out a pilot survey which compared the telephone method and the face to face method (which included a self-administered section, namely for questions about sexual practices). It enabled the researchers to bring to the fore the fact that NR rates did not vary -in most cases- whatever method was used. NR rates, for the same question, were always lower if they had been collected by phone, and when they concerned self-administered questions from the face to face questionnaire (ACSF Investigators, 1992). Moreover, this pilot survey served to eliminate all of the questions in which the NR rate was considered as too high. These results contributed in the choice of the telephone method. The final ACSF survey only includes overall low NR rates (less than 5% on average). Actually, the respondents were offered DK modalities in a great number of questions, especially in questions on "knowledge" which were later subjected to processing and interpretation.

VARIETY OF DATA GATHERED IN SURVEYS ON SEXUAL BEHAVIOUR:

Surveys on sexual behaviour aim to gather a great variety of data produced by the information respondents agree to provide. The data pertain to:

- respondents' practices, representations, feelings, attitudes at different times - more or less distant from the time at which the data are collected;

- other people than respondent : sex partners, parents, confidants, friends;

- the knowledge they have or do not have of some information or techniques (contraception, prevention, etc.,...);

- respondents opinions or attitudes towards a certain number of subjects which are more or less related to the field of sexuality.

The variety of the information gathered in these surveys calls on different cognitive processes according to the content elicited in the questions: recollection of present and past events, expression of knowledge, assessment of one's own life and relationships with sexual partners, friends and relatives, expression of opinions and attitudes.

On account of the variety of themes explored in these surveys, NRs and DKs do not have the same meaning in the context of different questions. In addition, DKs are not always offered to a respondent.

SURVEY SITUATION

A survey situation constitutes a social interlocutory situation. A certain number of measurement errors are attributed to the interviewer and to the characteristics of the relation established between the interviewer and the respondent (Hyman, 1954, Johnson and Delamater, 1976). Notably, the gender of an interviewer often introduces variations in the distribution of responses given to questions considered as "sensitive". For instance, it has been established that women have a greater tendency to report a masturbation practice when they are facing a man rather than a woman interviewer (Béjin, 1995).

Nevertheless, since sexuality is replete with emotion and culture, the survey situation can be experienced, according to G. Devereux (1980) as an erotic situation on an imaginary plane or as an admission or confession process.

If the act of collecting a questionnaire is considered as being a social interaction between an interviewer and a respondent, producing data must be considered - from the respondent's point of view - as a "declaration" process. The respondent is bound to adopt a strategy which allows him to manage both his own connection to the object of the survey and his environmental and inter-personal relationship. Moreover he has to cope with the interaction with the interviewer. The interviewer, for his part, has to manage his relationship to the respondent and meet the validity and consistency requirements of the data he collects. From this angle, NRs and DKs constitute a form of dissonance for the interviewer -which he is compelled to reduce by inventing responses. We were able to observe the fact that, when carrying out a questionnaire by telephone, some interviewers tended to set the cursor allowing them to "key in" the response automatically in advance in front of the expected response modality -before it was even formulated by a respondent - according to the "profile" representation they had elaborated for him. NRs can also express coding mistakes on behalf of interviewers.

LIE, DISTORTION AND MISKNOWLEDGE

The problem with NRs and DKs in surveys on sexual behaviour stands in a "declaration" dynamic, that is, in communicating to the interviewer certain information that might be subject to voluntary concealment. Criticism of surveys on sexual behaviour mainly concerns respondents' "lies" (Turner, Danella, Rogers, 1995) The information collected is said to be distorted because of the voluntary concealment of the respondent. Some of those who rely on this assumption can generally bring no evidence other than their "surprise" facing the quality of the responses they are faced to (Lewontin, 1995).

Other criticisms focus on memory distortion : experiences which took place a long time before the survey would be more often prone to memory distortion than more recent information. The example which is most often drawn on concerns the gap between the number of lifetime partners men and women report they have had and the reduction of this gap as concerns the number of partners they had in the year before the survey (Lagrange, 1991). Furthermore, all experiences do not have the same ideological and emotional importance, (salient or ordinary experiences) and are not recalled at random when filling in a questionnaire.

Forms of "misknowledge" are defined as factors which infer distortions to declaration when data are not ignored by the respondent. Respondent denial or refusal to express a response, allows him to carry on as if he did not know the responses and to preserve the state of cognitive and emotional equilibrium he is in. Feeling embarrassed in making a "declaration" may also constitute a factor of "misknowledge".

Along those lines, producing data by means of a questionnaire implies, for the respondent, an elucidation dynamic in the process of an interaction which, in some cases, can favour the production of data and in others the expression of resistance.

Producing data on sexual behaviour through surveys results from a complex declaration process which is part of a social interaction between interviewer and respondent. It calls on differentiated socio-cognitive and emotional processes, mediated by questionnaires in which a respondent is subjected to misknowledge.

Reflection on the meaning of these two types of responses oversteps the limits of a methodological analysis in the strict sense. The analysis does not aim, in the first place, to improve questionnaire elaboration by reducing NRs and DKs. It relies on residual data constituted of significant NR rates in certain surveys that have not been analysed by researchers and that we have attempted to explain the meaning of on one hand, and on the interpretation of data including the DK modality on the other. It aims to bring to light the plurality of meanings of these two types of responses.

The analysis of NRs and DKs is grounded on the established fact of presence and absence of the DK modality in some surveys on sexual behaviour.

When the DK modality is not provided, NRs are the only possibility for a respondent to express a set of reactions to a question: voluntarily refusal to answer, misunderstanding the question, not feeling concerned by the question, not knowing the response (whatever it may be), denying a situation proposed in a question, feeling uneasy about answering the interviewer. A NR globally appears, at first sight, as an incoherence.

When respondents are offered a DK modality, their are allowed to express their self-proclaimed ignorance -but also other forms of ignorance - or reactions which might be similar to those appearing in response to a NR and considered as "misknowledge".

It appears to be necessary, therefore, to dissociate the phenomenon of ignorance from the way it is measured through DKs. DKs could reflect expression of "misknowledge". NRs could be a refuge where ignorance can be expressed when the DK modality is not offered to the respondent. The following analysis, based on a few examples, aims to explore this working assumption.

STUDYING MATERIAL

The analysis proceeded in a secondary reading of two national surveys on sexual behaviour which had been carried out in France, more than twenty years apart (Simon et al., 1972; Spira et al. 1994) in different social contexts and according to different problematic. The "Simon Report" mainly explores heterosexual sexuality from a contraceptive viewpoint; ACSF explores a larger range of sexual behaviour in a context of the rise of the HIV epidemic (Giami, 1991, 1995). Only some of the items from these two questionnaires can be compared due to questions being asked in a different way. Furthermore, the two studies also present differences in their technical conception. ACSF was realised by using the telephone method. The Simon Report was face to face -with a self-administered section for questions on sexual practices and on attitudes relating to sexuality. Whereas the Simon Report only offers the DK modality in a few questions, the tables appended to the report systematically include a "Do Not Declare" category. Reading them reveals the non random character of these NRs according to socio-demographic and gender characteristics. ACSF was conceived according to another strategy. DK modalities were offered in questions on knowledge and in particular on knowledge of partners' sexual experience.

We have chosen to present, first of all examples of NRs which can be interpreted as full responses. Then we have attempted to point out the limits to the interpretation of these NR by discussing an example allowing to compare NRs and DKs obtained in for a same question. Eventually we treated an example which underlines the fact that using a DK modality can produce interpretations otherwise than a respondent's self-attributed ignorance.

RESULTS

Interpreting NRs

NRs : Imposing a problematic, embarrassment or ignorance

Analysing "Do Not Declare" responses from the Simon report reveals a gender effect concerning the question on the size of a man's penis.

Table 1 - According to what you know or believe, what is the average size of a man's penis during sexual intercourse ?

NR[5]

Overall

20 -29

over 50

married over 50

Men

28 %

12 %

37 %

Women

62 %

44 %

77 %

75 %

Question 5.4 - pp. 808-809 in Simon et al. 1972

First of all, the high percentage of NRs among both women and men rises commentaries about the way the question was formulated.

Table 1 reveals a large gap between men and women's overall NR (28 % versus 62 %) on one hand and, on the other, an important gap between women of different generations (44 % for the youngest versus 77 % for the oldest).

Young women are likely to be less embarrassed about answering this question than older ones. The gap between the youngest men and women is almost as important as that of all age brackets put together. In other tables, differences in standards of education contribute to an even larger gap. Women and men with an elementary school level have a greater NR rate than average. A higher education level reduces the gap between genders but does not neutralise it.

As far as this question is concerned, correlating socio-demographic characteristics and NRs confirms the major gap between genders.

The question seems to call for "knowledge". The importance of women's NR rates -notably as concerns the most elderly- induces us to believe that their ignoring the response to the question is not the only reason for the presence of NRs since they are actually not sexually less experienced than young women. One can consider that a woman could feel embarrassed to report that she implicitly "knows" the size of a man's penis by giving any answer to this question. Women choose not to answer instead of giving an indiscriminate response. The importance of the gap between men and women's NR rates could reflect the fact that the topics of the question is more masculine than feminine. Most women probably feel less concerned by the matter than men.

This is a case of what we can qualify as "imposing a problematic". Since there is no DK modality -which we could have compared with the NRs, conclusions cannot be drawn concerning ignorance and various other dimensions operating in this "Do not Declare" response.

Memory distortion or effect of a social norm :

Questions involving a first sexual partner's characteristics (age, marital status, circumstances of first meeting with...) show higher NR rates for men than for women.

Table 2 - How old was the woman (or man) with whom you had your first complete sexual intercourse ?

Overall

20 - 29

over 50

married over 50

Men

11 %

4 %

17%

21 %

Women

3 %

4 %

Question 2.2. - pp. 648-651 in Simon et al. 1972

Men aged 20-29 had lower NR rates than men overall and equivalent to women's rates overall. More men over fifty, even more when they are married, omitted to answer this question. But women's NR rates do not increase as women grow older. Men's age related increase could have been interpreted as a deficit in memory processes. The youngest men would remember more frequently their first sexual partner's characteristics because this experience is more recent for them. However, this assumption can be invalidated by evoking the absence of increase concerning older women's NR rates. Older women are more likely to remember their first sexual partner than men.

Furthermore, these findings must take into account the number of men who had their first sexual intercourse with a prostitute: 9 % overall and 4 % of the men who are between 20 and 29 years old. It must also take into account the percentage of women who had their first sexual intercourse with their spouse: 56 % versus only 20 % for men. The gender biased different characteristics of this first so-called sexual relationship can explain why men's NR rates are so high whereas women's are so low. The ability to forget and taboo concerning prostitution may be interlinked.

Comparing NRs and DKs

The question about the first sexual partner's virginity was one of the rare questions in the Simon Report in which respondents were offered a DK modality. This is quite an interesting point to note in relation with the implicit assumptions of the investigators.

Recollection selectivity also arises strongly on this subject. Although men's NR and DK rates increase as they get older, men have much lower NR and DK rates overall than women as to knowledge of their first sexual partner's virginity.

Table 3 : Was this woman (or man) with which you had your first sexual intercourse a virgin or had she (or he) already had sexual intercourse?

Overall

20-29

50 and over

married

Men

Women

(Question 2.4. in Simon et al. 1972, pp. 654-655)

Whereas recalling the age of their first sexual partner provoked more NRs from men than women, virginity appears - to men - a salient criterion. Furthermore, in other tables of the Simon Report, twice as many men "know" whether their first partner was a virgin (30 % versus 14 % of the women). 50 % of the men reported they had, at least once, had sexual intercourse with a virgin versus only 16 % of the women. These findings tend to reinforce the assumption according to which men pay more importance to their sexual partner's virginity than women. More elderly women "do not respond" more frequently than younger women. Perhaps the question was more often perceived by these women as being irrelevant.

Comparing DKs and NRs reveals the difference between men and women self-proclaimed ignorance according to the virginity of their first sexual partner. Men's DK and NR percentages are roughly equivalent. Women chose more frequently to use the DK modality rather than do not responding to this question. Women from all generation's choice - in a strong minority of cases - of the DK modality related to their first sexual partner's virginity can be interpreted as their expressing their self-proclaimed ignorance. It is, in fact, more difficult for a woman to "know" whether a man is a virgin; on one hand for anatomic reasons and on the other because virginity is not an important value as far as men are concerned. On the contrary, mentend to hide their virginity rather than to claim it. However, the meaning of NRs - which are less important when related to women - remains obscure.

By offering a DK modality, this question provided means to shade the assumption on women's NRs according to which the question about the size of a man's embarrassed or did not concern women. Part of their NRs in that question could also have expressed their self-proclaimed ignorance about this matter. Therefore the presence of a DK modality in this question of the Simon Report gives way to more subtle interpretations as opposed to other questions in which this modality is not proposed.

SELF-PROCLAIMED IGNORANCE : DOUBT OR DENIAL ?

Studying the characteristics of sexual partners constitutes one of the main approaches developed in surveys on sexual behaviour in the time of AIDS. The information provided is essential for determining the probability a respondent has of meeting a partner who is at risk of infection or already contaminated by the virus and to evaluate the respondent own degree of exposure to risk. Furthermore, analysing this data facilitates the understanding of some modalities which structure relationships between sexual partners.

In ACSF, it was established that 16 % of individuals reported that they had "changed their sexual behaviour since AIDS appeared" . Changes related primarily to partner selection strategies: "I only have sexual intercourse with people I know" (76 %), "I only have sexual intercourse with people I'm in love with" (70 %) (Spira et al., 1994, p. 151).

Since selection strategies which were chosen in the majority of groups were based on "knowing a partner", it seemed appropriate to explore the "knowledge" that respondents said they had of their sexual partners' sexual experience. The "knowledge of a sexual partner's sexual experience" was defined as "knowing whether or not a partner with which a respondent reports he had his last sexual intercourse had other partners than the respondent during the time they were in contact with each other". The "knowledge of a sexual partner's sexual experience" was determined according to whether the respondents reported they had one partner (only one partner in the twelve months before the survey) or multiple partners (two or more partners in the last twelve months). In addition, these criteria were compared according to the type of partners with which respondents reported they had their last sexual intercourse ("cohabitant", "regular", "occasional" or "new"). Respondents were offered a "Don't Know" modality.

Table 4 and 5: Knowledge of a partner's sexual experience (knowing if he/she had other partners at the time he/she was having a relationship with the respondent) (%):

a/Men

Knowledge of partner

Respondent's status

type of partner

one partner

multiple partners

one partner

Overall

cohabitant regular

occasional

new

1
9
31
6
3

3
8
17
11
4

multiple partners

Overall

cohabitant

regular

occasional

new

76
70
56
54
68

9
14
29
21
14

15
15
15
25
18

b/Women

Knowledge of partner

Respondent's status

type of partner

one partner

multiple partners

one partner

Overall

cohabitant regular

occasional

new

81
75
70
93

1
11
7

19
3

3
8

multiple partners

Overall

cohabitant

regular

occasional

new

75
63
63
64
67

5
29

27
18
19

20
8
11
18

Spira et al. 1994, p. 153

The majority of respondents, male and female, who define themselves as having had one partner during the past twelve months considered that the partner with whom they had their last sexual intercourse also had only one partner. They were also less likely to report that they ignored their partner's sexual experience. Only 3 % of those (men and women) with one partner reported ignoring their cohabitant's experience; 8 % ignored regular (non-cohabitants) partners' experience. DK percentages increase when the last sexual partner was "occasional" or "new". One can observe, however, that this rise is related to the rise in percentages concerning respondents who reported they knew their sexual partner had other partners. Therefore, when one type of partner ("occasional" or "new") is more often supposed to have had other partners, he is all the more subjected to DK responses. Hence, DKs can, in this case, express a doubt - fairly inclined to favour a positive response to the question.

The majority of respondents with multiple partners also reported that they "knew" their partner had had no other partner than themselves. But, more often than respondents with only one partner, they reported that they "knew" that their partner had other partners than themselves. In addition, they reported, more often than respondents with only one partner, that they ignored their partner's sexual experience.

Among respondents with multiple partners, 20 % of the women and 15 % of the men reported that they did not know whether their cohabitant had another partner. This "ignorance" also concerns 25 % of the men and 18 % of the women with multiple partners reporting they had their last sexual intercourse with a "new" sexual partner. However, more women than men with multiple partners reported that they "did not know" whether their sexual partner had other partners when the partner was the "cohabitant". It is as if those women preferred not to know whether their cohabitant had other partners than themselves

Thus, one can observe:

- a correlation between a type of partner and the sexual experience attributed to that type of partner;

- a correlation between the sexual experience (with one or multiple partners) a respondent report he's had and the experience he reports his partner's had;

- a correlation between the fact that a respondent has multiple partners and his reporting he does not know his partner's experience, including when that partner is described as a cohabitant;

- cohabiting women, more often than men, report that they "do not know" about their cohabitant's experience.

A large minority of the respondents "do not know" about their partner's experience, and this becomes more pronounced among women when they are married and have multiple partners themselves.

DK responses can then be interpreted whether as a connivance between partners who both have multiple partners or as a projection of the respondent's sexual experience onto his partner, or as a denial of knowledge of such experience.

The selective use of condoms according to a partner's sexual characteristics provides food for thought concerning the meaning of DK in relation to questions about the knowledge of a partner's sexual experience.

Table 6 and 7: Condom use at last sexual intercourse according to sex, sexual activity and knowledge of a partner's sexual experience (%):

a/Men Partner has other partners	with only one partner	with multiple partners
No	10.4	22.2
Yes	21.2	36.7
Don't know	21.8	29.3

b/Women Partner has other partners	with only one partner	with multiple partners
No	7.6	15.4
Yes	14.1	18.1
Don't know	16.7	18.4

Knowledge and non-knowledge of a partner sexual experience are both related to the selective use condom at last sexual intercourse.

Men and women with only one partner reported that they used condoms when they last had sexual intercourse, more frequently when they "knew" or "did not know" if their partner had other partners than when they "knew" that he did not have any. However, men with multiple partners used a condom more frequently when they "knew" that their partner had an other partner than when they "did not know" it. Women with multiple partners reported use of condom more frequently when they "did not know" if their partner had an other partner.

Hence, the question about the use of condoms throws light on the global meaning of DK responses concerning the knowledge of a sexual partner's experience with other partners. When men and women report they "don't know" whether their partner has other partners, they are more likely to use condoms. On the contrary, men with multiple partners do not use condoms quite as often as women do when they "don't know" whether their partner also has multiple partners.

Most likely, men who "don't know" whether their partner had multiple partners, sometimes consider that this is not the case. Whereas women with multiple partners are more strict in the way they act when they "don't know". In addition, when women give a "don't know" response, they could either be "knowing" the fact or denying it.

CONCLUSION:

Starting from the comparative analysis of NRs and DKs to questions on sexual partners in surveys on sexual behaviour, it has been demonstrated that the meaning of NRs and DKs vary according to the type of information required and to the context the question is in.

- When a majority of women fails to answer a question about the size of a man's penis, it did not appear as an inconsistency. However, the high rate of NRs in women's responses to this question makes their statistical processing difficult for the investigators - as the question was grounded on an assumption of "knowledge". The absence of a DK modality for this question makes it difficult to distinguish between self-proclaimed ignorance, strain to answer or feeling not concerned by the topics.

- Finding a high NR rate among men to questions concerning the characteristic of a first sexual partner and a much lower NR rate to a question on the "knowledge" of a first partner's virginity reflects how much men are involved with their first sexual partner. Especially when one considers that a proportion, equivalent to the number of men who did not answer the questions, had a prostitute as a first partner on one hand and that for a high proportion of women the first sexual partner was the spouse. The differential NR rates, as far as men and women are concerned, confirm the fact that this first relationship is perceived in a very different way by both genders.

- The comparison between DKs and NRs in the question about the virginity of the first sexual partner showed that women declare their ignorance more frequently than men. Moreover the high proportion of DK to this question allowed us to shade the assumption about women feeling embarrassed to answer the question about the size of a man's penis. Women might be more likely not to know the answer to this question.

- When a respondent reports that he "doesn't know" if his sexual partner had other partners, this can reveal a projective attribution process. Respondents with multiple partners themselves report more frequently than respondents with one partner that they "know" that their partner had an other partner. DK could also express doubts concerning the partner's sexual experience. If condoms are so more used when a respondent "does not know" whether his partner has had other partners than himself, it could demonstrate the fact that reporting this lack of knowledge is denying the knowledge he has about that experience.

This article calls for the development of qualitative forms of analysis of NR and DK. To give readers a sense of the distribution of these responses in the population, studies drawing on large, representative samples were chosen. However, as S. Kracauer wrote: "An example of the limitations placed on qualitative analysis may be found in Berelson's statement that: 'Whenever one word or one phrase is as "important" as the rest of the content taken together, quantitative analysis would not apply. Qualitative analysis would; and it would make its breakdowns hinge on this one word or one phrase' " (Kracauer, 1952, p. 639). Qualitative analysis of NRs and DKs in quantitative surveys could therefore contribute in filling out the meaning of quantitative results.

References :

ACSF Investigators, (1992) : Analysis of sexual behavior in France : A comparison between two modes of investigation, telephone survey and face-to-face survey. AIDS . 6 : 315-323

Ardilly P. (1994) : Les techniques de sondage . Paris, Editions Technip.

Béjin A. (1995) : Female Masturbation in France. The estimation of the underreporting of a practice in : M. Bozon, H. Léridon eds. Sexuality and the Social Sciences. Dartmouth, Aldershot. (in print)

Bourdieu P. (1973) : L'opinion publique n'existe pas. Les Temps Modernes 318 : 1292-1309

Bradburn N. , Sudman S. , (1981) : Improving Interview Method and Questionnaire Design . San Francisco, Jossey-Bass Publishers.

Catania J.A., Gibson D. , Chitwood D.D. , Coates T. , (1990) : Methodological problems in AIDS behavioral research : Influences on measurement error and participation bias in studies of sexual behavior. Psychological Bulletin. 108 : 339-62

Devereux G. (1980) : De l'angoisse à la méthode dans les sciences du comportement . Paris, Flammarion.

Francis J., Busch L. (1975) : What we know about "I don't knows". Public Opinion Quarterly . 39 (2) : 207-218

Giami A. (1991) : De Kinsey au sida: l'évolution de la construction du comportement sexuel dans les enquêtes quantitatives, Sciences Sociales et Santé. 9 (4) : 23-56

Giami A. (1995) : The ACSF survey questionnaire : The influence of an epidemiological representation of sexuality in : M. Bozon, H. Léridon eds. Sexuality and the Social Sciences. Dartmouth, Aldershot. in print

Hyman, H. (1954) : Interviewing in Social Research

Johnson W., Delamater J. (1976) : Response effects in sex surveys. Public Opinion Quaterly. 40 : 165-181

Kracauer S. (1952) : The challenge of qualitative content analysis. Public Opinion Quarterly. Vol. XVI, : 631-641

Lagrange H. (1991) : Le nombre de partenaires sexuels : les hommes en ont-ils plus que les femmes ? Population . 2 : 249-278

Lewontin R. 1995 : Sex, Lies and Social Science. The New York Review of Books . April 20 : 24-29

Michelat G., Simon M. (1982) : Les "sans réponse" aux questions politiques : rôles imposés et compensation des handicaps. L'Année sociologique 32 : 81-114

Simon P., Gondonneau J., Mironer L., Dorlen-Rollier A.-M., (1972) : Rapport sur le comportement sexuel des français. Paris, Julliard, Charron.

Spira A., Bajos N., and ACSF Group, (1994) : Sexual Behaviour and AIDS. Aldershot, Avebury.

Turner C., Danella R., Rogers S. 1995 : Sexual Behavior in the United States, 1930-1990 : Trends and methodological problems Sexually Transmitted Diseases . 22 (1) : 174-190

Author's address :

INSERM U. 292

Hôpital de Bicêtre

Secteur Bleu "Pierre-Marie" - Porte 26

82, rue du Général Leclerc

94246 LE KREMLIN-BICETRE Cedex