Version classiqueVersion mobile
OpenEdition Books

Sexual Behaviour and Risks of HIV Infection

 | 
Michel Hubert

II. Qualitative and Quantitative Data Collection Methods in the Study of Sexual Behaviour

Notes on the Design and Construction of a National Survey of Sexual Attitudes and Lifestyles

Kaye Wellings, Julia Field, Anne Johnson, Jane Wadsworth et Sally Bradshaw

Texte intégral

1The British National Survey of Attitudes and Lifestyles (NATSAL) was conceived in response to a need for data which would

2a) facilitate the development of a well targeted and effective AIDS public education campaign

3and

4b) enable more accurate estimates to be made of the possible future pattern of HIV transmission.

5Originally, two surveys were set up in an attempt to meet this need. With a predominant emphasis on the first of the two aims, the Health Education Authority (HEA) in conjunction with Social and Community Planning Research, in 1987 set in progress plans for a survey of sexual activities and behaviour together with their social determinants.

6Concurrently with this survey, an academic team centring on St Mary's Hospital Medical School, Middlesex Hospital Academic Department of Medicine and Imperila College, London were funded by a private benefactor to develop a survey whose chosen emphasis was the second of the two aims, that of helping predict the likely spread of HIV.

7In 1988, the two studies were combined and the two research teams began preparatory work on a national feasibility survey which was completed in October of that year. There were a number of reasons for the merger, not least among them a concern for economies of human and financial resources, and an awareness of the advantages in terms of pooled expertise and shared experience and perspectives.

8The focus of this paper is on technical aspects of the design and construction of the questionnaire for the survey. A number of factors were influential in the development of the research instrument. We were guided by the purposes for which the data were to be used; by the need for comparability with past and present studies, by the possibilities and problems revealed by a surveillance of work carried out by others, and not least by our own ideological and theoretical stance on particular issues. In relation to this last, we were fortunate in sharing a common perspective, and where there were differences of opinion, they were of a nature likely to lead to fruitful discussion rather than friction.

I. Establishing the variables

9The National Survey of Sexual Attitudes and Lifestyles was conceived and planned within the public health context of the AIDS epidemic, its aims, as already stated, being to assist in the prevention of further spread of the virus and in the planning and provision of health care services for those already affected. Necessarily then, the investigative focus of the study reflected these aims and those involved in its design were ever mindful of the uses for which the data were to be employed.

10There was, however, a further concern which guided the measurement objectives. It has been commented on elsewhere (Wellings, 1989) that surveys set up in response to a particular contemporary problem are often of limited value in relation to other social or health issues. The urgent need to predict and prevent the spread of AIDS and HIV has served to demonstrate that knowledge about sexual behaviour is now clearly essential. Yet the lack of good information on the subject has for some time had serious implications for understanding fertility patterns, trends in contraceptive usage and the epidemiology of other sex-related diseases (cervical cancer, for example), and for placing debate about the changing forms and stability of the family and the provision and timing of sex education in a firmer context.

11Since this survey draws extensively on public funds, and since there is no certainty that such funds will be available on a regular basis, we felt it incumbent upon us to extend the research instrument wherever possible to provide data which would assist other health care professionals working in the field of sexual health; in psycho-sexual counselling, in the prevention of other sexually transmitted diseases, in family planning, etc.

12The measurement objectives of the study were therefore:

  1. To measure rates of sexual partner change and sexual history in a representative sample of the British population (including age at first intercourse).

  2. To measure the prevalence and distribution of different patterns of sexual orientation, eg. homosexuality/bisexuality/heterosexuality.

  3. To measure the extent of experience of injecting drug use and sex for payment.

  4. To determine the demographic characteristics of those whose current sexual lifestyle puts them at greatest risk of HIV and other STDs.

  5. To measure attitudes to sexual behaviour, sexual relationships and knowledge of possible associated risks, and to examine the relationship between attitudes, patterns of socialization, and behaviour.

  6. To assess changes in sexual lifestyles through generational comparisons of sexual histories.

Research Variables

13These measurement objectives, in turn, generated the following variables.

14Past sexual history:

Age at first sexual encounter; nature of encounter
Age at first intercourse; identity of partner, nature of relationship (ie. casual encounter, permanent one-to-one relationship, etc.)
Timing of subsequent experience

15Recent experience:

Current relationship
Nature of relationship
Experience of marriage/divorce/separation

16Sexual orientation:

Predominant mode; celibate/heterosexual/bisexual/homosexual
Extent to which categories are discrete/continuous in thought and in practice
Stability of sexual orientation over time
Extent and nature of experience of alternative orientation(s)
Attitude towards alternative orientation(s)

17Sexual practices:

Past and current experience of different sexual practices (mutual masturbation, penetrative intercourse, vaginal and anal intercourse, oro-genital sex)
Combination of sexual practices

18Sexual partners:

Number and gender of sexual partners including frequency of change of partners, nature of sexual practice
Duration of relationship
Experience of monogamy/serial monogamy/concurrent multiple partners, etc.
Experience of paying for sex

19Sexual health:

Knowledge and awareness of risks of ef. STDs, AIDs and HIV, cervical cancer, unplanned pregnancy/abortion
Sources of advice on issues relating to sexual health
Experience of recreational drug use

20Risk reduction behaviour:

Perceived degree of control over sexual health and life course
Knowledge of, motivation towards, and experience of safer sex practices
Attitudes towards, and use of, prophylactics/contraceptives

21Psycho-social factors influencing sexual behaviour:

Extent and strength of social influences, the influence of parents, friends, school, etc.
Personal beliefs; philosophical, religious, ethical: nature, strength and influence on behaviour and persistence over time (family, friends, school, etc.)
Attitudes towards eg. celibacy, monogamy, permissiveness, etc.
Attitudes towards abortion, pre-marital and extra-marital sex, homosexuality, etc.

II. Wording the questions

22The second stage in constructing our questionnaire was to move from a simple description of the instrument to a detailed specification in operational terms.

23This work was informed from several sources. We searched the literature carefully to determine how various authors use different concepts. We examined other researcher's approaches to the task, in particular, their wording of questions. Most important perhaps, we relied on qualitative research, carrying out in-depth semi-structured interviews with members of the population we were interested in, in order to familiarize ourselves fully with respondents' use of language, willingness to respond, and the optimal conditions under which they were likely to do so. This work is more fully reported elsewhere (Spencer et al, 1988).

24Our aim throughout these efforts was to describe, in precise and measurable terms, the range and nature of sexual attitudes and behaviours in the general population. Given this operational specification, we were able to proceed to prepare the questions (or attitude statements) which would constitute the measuring instrument.

A. Standardising the sign

1. Choice of Language: formal/informal

25Much of our preparatory work, reading, studying, listening, etc., seemed to point naturally to the use of the respondent's own language in the wording of the questionnaire. Many researchers counsel matching the language of the interview with that of the respondent on the dual grounds that only thus will a rapport to be forged with the interviewer facilitating maximum possible disclosure, and only thus can full understanding by the respondent be vouchsafed. Kinsey (1948:52), for example, cautions against the use of scientific terms, advising instead the use of the vernacular:

"Adapting the form of the question: While the point of each question has been standardized, the form of each question has varied for the various social levels and for the various types of persons with whom the study has dealt. Standardized questions do not bring standardized answers, for the same questions mean different things to different people. In order to have questions mean the same thing to different people, they must be modified to fit the vocabulary, the educational background, and the comprehension of each subject. It is especially important to use a vocabulary with which the subject will feel at home, and which he will understand".

26Several recent surveys have successfully adopted this formula. Project Sigma, for example, ask their sample of gay men for their meaning of a particular term before imposing their own definition, in order to determine precisely what is being measured:

"Suppose someone asked you 'How many sexual partners have you had this month? 'What must have happened sexually for a person to 'count' as your sexual partner?"

27and add as a note to the interviewer:

"(This is often a difficult question for the subject to answer, and he might well never have thought about it before. The important thing is to elicit his definition of 'sexual partner' and the criterion he uses to mark a sexual encounter.)

28Similarly, with a mixed race sample in the US the AIMN (AIDS in Multicultural Neighbourhoods, AIMN 1988), during the face to face interview, the interviewer hands the respondent List A, which reads as follows:

29Preferred terms for:

  • Penis…………………………………

  • Semen…………………………………

  • Orgasm/Ejaculation…………………

  • Condom…………………………………

  • Masturbation……………………………

  • Fellatio…………………………………

  • Cunnilingus…………………………………

  • Anal intercourse…………………………………

  • Vaginal intercourse…………………………………

  • Oral/anal stimulation…………………………………

30asking "Before we start, can you please tell me what word you use in ordinary conversations to refer to each of the following sexual terms". Having translated the concept into the respondent's language, the interviewer then substitutes the respondent's preferred term for each formal term contained within the instrument

2. Application to NATSAL

31The arguments in favour of using the respondent's own terms are self evident and do not need rehearsing. Ideally, respondents should be able to select their own preferred terms. In the context of our own survey, however, we were less convinced of the wisdom of using the respondents' own language. There were, firstly, practical problems generated by the scale of the survey, the sample population, and the chosen format A sample of 20,000 is costly enough as it is, and extra time spent in negotiating meaning would add considerably to the overall expense. Further, tailoring language to participant is somewhat less feasible in the case of a large scale heterogeneous general population sample where there is wider diversity in terms of sexual orientation, practices, etc., than it is in that of a purposive sample relatively more homogeneous, in terms of at least one variable. Added to which was the problem of format; part face-to-face standardised interview, part self completed questionnaire, which presented fewer possibilities for negotiating meaning. The self-completion (Booklet) component of the schedule did not present the same opportunities for selecting preferred terms, and so the exercise became less feasible.

3. Respondents’ Preferences

32Our own development work revealed difficulties of a more substantive nature. In the course of this work, we sought to explore preferred patterns of usage and their meaning by investigating both revealed and expressed preferences. Interviewers at the qualitative stage were instructed to encourage and prompt respondents to provide an account of a particular experience, behaviour, practice or relationship and then to ask what term they would use to describe such a phenomenon (eg. "What would you generally call the person you've just described?", "What term would best fit the person you've just described?").

33On other occasions, the interviewer picked up a term used by the respondent and asked what experience or phenomena he or she was describing (eg. "What would count as a sexual partner for you?"). In this way, the interviewer was able to move backwards and forwards between signifier to sign and from sign to signifier in an attempt to probe language preferences.

34This work revealed an extraordinary diversity in relation to language styles, ranging from the biblical ("couple", "copulate", "fornicate") to the vernacular ("screw", "fuck"); from the euphemistic ("doing it", "having it") to the romantic ("making love"); and from the lay terms ("having sex”, "sexual intercourse") to the scientific "(coitus”). The general population seemed to show a greater tendency to sexual polyglotism.

35The terms used varied not only between respondents but whithin the pattern of usage of each individual respondent. Respondents were seldom consistent in their style of language throughout the interview. For example:

M.

In my own mind, sexual experience involves intercourse

I.

As you say...

M.

Screw, I screw. I think if we're talking about intercourse we are talking about a man and a woman Copulatin

M; 20-24 years; single

36There was some evidence, too, that the use of terms changes with experience and through time. Use of different terms for the same act varies it seems with the nature of the relationship and with the life stages of the respondent:

M.

I think you grow up with different words as you go through life. Originally, you'll probably go on to "I've had sex with her”, "I've screwed her" and later on when you've got a more permanent kind of relationship you might talk about "making love".

37Another difficulty which came to light as a result of the qualitative work was the possible discordance between words the interviewer felt comfortable using, and those the respondent felt comfortable with. Some respondents did express a clear preference for the vernacular

M.

"Let's talk one another's language, the first screw I had was when I was about 27"

M; 26-35 years; single

38Others used a term themselves, but were less happy hearing it used by the interviewer. The printed word poorly conveys the affrontery expressed in the tone of the spoken word in the following exchange:

M.

No, you've got to be very careful because if you try to come down to somebody's level too much... if you try to use some words too much, or use them at all, for instance, snogging or er... (hesitation)

I.

Fuck?

M.

Yes. Well, I mean you could use that, I think that'd be... well, you've got to be very, very careful...

M; 21-25 years; single

39Conversely, although it was only anecdotal, there was some evidence on the part of interviewers of discomfort when over-frank language was used by respondents.

40At the same time as espousing the use of the vernacular, Kinsey acknowledges this and issues the following word of caution (1948:52):

"One must know and use the vernacular terms with a fine sense of their properties and their exact meanings in each group. Their awkward use may damage instead of building rapports".

41Our own experience, therefore, failed to give us confidence that the use of respondents' own language would provide the required standardization in terms of what was signified. Instead, it had rather the reverse effect showing us how context dependent and mercurial were the connotations attributed to different terms, such that we could not rely on there being an underlying constant to which different terms applied.

4. Expressed Preferences

42Towards the end of the in-depth interviews conducted at the qualitative stage of fieldwork, respondents were asked to explicitly state their preferences in terms of the style of language used. The previously identified consensus on more formal language was broadly confirmed in these expressed preferences. The majority of respondents favoured the use of more formal terminology with the proviso that explanations be provided where necessary. Typical responses included:

M.

No, I think the formal language is the thing to use because everyone knows what formal language means, it has a meaning. If you're talking in slang, a lot of people can get the wrong idea.

M; 26-35 years; single

M.

It depends, I mean if you're talking to someone who understands the terms it’s okay. If you're talking to someone who doesn't know the medical terms you'd have to explain.

M; 3645 years; divorced

M.

Instead of using the penis, you can use the word baked beans, providing we both understand baked beans is actually a penis, it doesn't matter what words you use.

M; 26-35 years; single

43The decision was therefore taken to use formal terms throughout the questionnaire, both in the face to face interview and the self completion Booklet.

5. Respondents' Understanding of Terms

44In deciding on formal language, the concern was to use relatively neutral terms and expressions which could not cause offence, but avoided coyness and lack of clarity. At the same time, they needed to be easily understood, and it was not always easy to avoid resorting to technical terms. The qualitative research taught us that there were few such terms whose meaning could be taken as universally understood. Even those terms which are widely used in health educational literature provided problems of understanding which came to light at the development stage, 'vaginal sex', for example:

I.

What about vaginal sex, do you know what that means?

F.

No.

I.

Right, urn, oral sex?

F.

That's just ordinary sex, isn't it? Just having sex in the ordinary way.

I.

And, penetrative sex?

F.

No, haven’t heard of it, urn, is it sort of going half way?

F; 19 years; single

F.

No, I don't know what that means

F; 45-65 years; married

F.

No, never heard of that actually

F; 25 years, single

45In addition to a simple lack of understanding of many terms, in which case explanation would have been enough to clarify meaning, the developmental work unearthed another difficulty, namely a tendency to attribute to unfamiliar terms meanings associated with unorthodox or bizarre sexual practices.

I.

What about vaginal sex-are your clear? What do you understand by that?

M.

Yeah, there again, it's something that doesn't appeal to me-I don't know why, it revolts me

M; 35-44 years; married (and reported regular intercourse with his wife)

I.

Heterosexual?

M.

Well, it’s all the same to me. Heterosexual, bisexual, they're all bloody queers as far as I'm concerned

M; 26-25 years; single

I.

Penetrative sex?

M.

Well, I suppose it's something queers do is penetrating-penetrative sex

M; 31 years; single

46By contrast, what was classified as 'normal' sex did not seem to need a distinguishing term to define it.

I.

Vaginal sex?

M.

That's just what one does, isn't it? Would have done just to say sex there.

47(It is not the intention here to examine the apparent difference in responses between the sexes, though there docs seem a tendency for men to dismiss what is unfamiliar as unacceptable, and for women to simply acknowledge ignorance - perhaps because men are culturally held to be more knowledgeable about sex).

B. Standardizing the signifier

48Having taken the decision to standardize the terms, we needed to ensure that the meanings attached to those we used were also standardized, precise, that the underlying behaviour represented by them was also constant throughout our sample population. This was not simply a question of defining medical/scientific terms. Everyday English as it is used in practive is ambiguous and for the purposes of the survey we needed definitions which were tight enough to measure accurately the range and nature of sexual behaviours.

49Clearly, any variation in meaning would threaten content validity ("the degree to which the measurement used represents the concept about which the generalisation is being made", Smith 1976:76). Yet meanings vary both between individuals, from one person to another, and within one individual's experience, from one time or place to another.

50We were aware, from the qualitative work, that the meanings attached to standardized signs were not always constant but were continually negotiated, often during the course of the interview. Sexual behaviour is rarely spoken about publicly and as a result the language used to describe it is impoverished, inadequate and inappropriate. Given the chance to explore the language and concepts associated with this area of behaviour, respondents often took this rare opportunity to reflect on and experiment with their accepted meanings. As a result, even when the sign itself is standardized, it is still possible for the experience signified by it to vary throughout the interview.

51Thus:

I.

So, in terms of having a sexual experience now, what would it mean to you?

M.

Well, I may have just altered my opinion slightly, up until five minutes ago if you would have said to me a sexual experience, I would have said that meant intercourse.

I.

Yes?

M.

ie, a screw. I suppose looking at it harder which is something I've not done until now, in other words, heavy petting, to the point of getting an erection is vaguely definable as sexual experience but I think you're treading on a very wet marsh (laughs).

M; 26-35 years; single

52Terms also meant very different things to different people. This was clearly illustrated for us in the use of the term 'sexual partner' a term which held a wide variety of meanings for respondents. Qualitative work showed the concept to be variously defined according to the strength and durability of the relationship, and the extent to which it was socially institutionalised. For most people, the term signified some kind of sexual activity, but for some, this alone was sufficient criterion for them to 'count' as a sexual partner.

M.

A sexual partner is somebody who-just for the purposes of having sex-somebody to have sex with

M; 36-45 years; single

M.

Sexual partner to me would be when one has a full sexual relationship be it within marriage or beyond marriage, someone one couples with.

M; 56-60 years; married

53For most people though, the physical element alone would not satisfy the conditions of the definition, an additional dimension was needed:

F.

Well, it's not just someone who you sleep with, it's a friend, a...

F; 19 years; single

54For a minority of others, the sexual component was not even essential to the definition:

F.

I can't remember the last time we had what you’d call full sexual intercourse, but you see that doesn't really matter to us. He's still my sexual partner. I couldn't think of anyone else in that way.

F; 3544 years; married

55For such respondents, a sexual partner may not necessarily be someone with whom sexual acts have or are taking place currently, but rather may be someone for whom sexual thoughts, fantasies, reminiscences etc, focus on and are reserved for.

56In other words, the essential component of the definition as far as we were concerned in the context of this survey - the actual physical experience of sexual activity - was neither necessary nor sufficient in terms of the meanings respondents imputed to the concepts. Yet it was crucial to the validity of the study that the indicators used measured what we wanted them to measure, only that and nothing else.

57As a result, it was necessary to define our terms carefully, not simply in terms of dictionary definition, but in terms of a specified practical reality. For example, "sexual partner" we defined as "people who have sex together just once, or a few times, as regular partners or as married partners". In the faceto-face section of the interview schedule, interviewers were provided with explanations to offer respondents. Whenever possible, we used the language provided for us by respondents in the qualitative work to amplify questions. Parenthesizing lay terms as alternatives for interviews:

"... at what age did you stard menstruating, (having periods)?" (pX)

58and in the response options for contraceptive use we used the coil in addition to IUD, "safer period” in addition to "rhythm method", etc. In the self-completion section, we prefaced the schedule with a Glossary of key terms eg. vaginal sexual intercourse: A man's penis entering a woman's vagina (the text of the Glossary can be found in the Appendix). When we used terms which were not specific, additional explanation was provided. For example, "When if ever was the last occasion you had sex with a man?" (This means vaginal intercourse, oral sex or anal sex). (Emphasis added).

The case for the operational definition

59By using standard terms, and providing standard definitions for them, we lay ourselves open as researchers to the charge of narrow operationalism; and to the possible criticism that we neglect of disregard the meaning of respondents. This was not seen as an unimportant issue. The meaning of different sexual behaviours, orientations etc., is a vital area for inspection and not nearly enough work has been carried out in this field. Far too much risk reduction advice urges people to substitute one practice for another in their sexual repertoire with far too little consideration for what they mean to the individual.

60In our defence, the justification for our decision lies partly in the genesis and purpose of the study, one of the aims of which, as previously mentioned, was to help predict the spread of HIV infection. In this context, it is essential to know, as precisely as possible, at which intervals, with what frequency and with whom, sexual acts involving the exchange of bodily fluids have taken place. Thus, for example, in relation to sexual partners, the variable of interest was "number of partners" so that it was essential that the only variable was the number and that the concept "sexual partner" remained invariable.

61Not only was it not, unfortunately, within our remit in this study to explore the meanings which different forms of sexual expression and relationships have for actors, but also the study design was not appropriate to such a purpose. It has to be said that large scale, quantitative surveys of this kind are not best suited to exploring the richness and complexity of sexual expression. Far better to employ in-depth qualitative techniques which should exist alongside the quantitative work to put the flesh on what might otherwise be rather dry and sparse bones.

62This said, we were always at pains to convey to respondents that we were not attempting to impose a 'correct' meaning to terms, but that we were giving standard definitions to the terms "so that everyone attaches the same meanings to the terms" (p. X), and whenever possible those standard definitions derived from the meanings respondents provided us with.

III. Sexual orientation and practices

63Our task in formulating questions relating to sexual orientation was made easier by the abundance of research ongoing in the field. In addition to our own qualitative work, the experience of those whose survey instruments were currently being used and a surveillance of those used in the past informed our choice of questions and greatly facilitated the drafting of this section of the questionnaire.

A. Question Wording

64A primary concern was with the words used to describe sexual activity in a homosexual relationship. It was clear from the data from the developmental stade that although the tendency to equate 'sex' with 'intercourse' was widespread, if not nearly universal, amongst those who had experience of predominantly heterosexual relationships, this was not true of those whose relationships were mainly homosexual. For gay men and women, sex has a broader meaning, reflecting a more varied sexual repertoire, less exclusively focussed on penetrative intercourse.

65This is confirmed in the Project Sigma's data. Only 40% of their sample of gay men reporterd having had active or passive anal intercourse in the month prior to interview (Project Sigma, March 1989).

66Kinsey (1948:616), too, cautions against heterosexual assumptions of what constitutes homosexual experience:

"... the homosexuality of certain relationships between individuals of the same sex may be denied by some persons, because the situation does not fulfil other criteria that should be attached to the definition. Mutual masturbation between two males may be dismissed, even by certain clinicians, as not homosexual, because oral or anal relations or particular levels of physic response are required, according to their concept of homosexuality".

67We were thus alerted to the inadequacy of measuring the prevalence of homosexual behaviour by using intercourse as an indicator. Attempts to do so would be likely to result in a significant under-reporting.

68This may account for the low rate of homosexual behaviour recorded in other studies. Forman and Chilvers (1989), for example, in the course of a case-controlled study investigating risk factors for testicular cancer, asked of their sample of young and middle-aged men "Have you ever had homosexual intercourse" to which only 8 out of 480 (1.7%) gave a positive reply. Whilst this may well have been an accurate record of the number of men who had had sexual intercourse with another man, it is unlikely to record the true prevalence of sex between men, which is more extensive in scope. We were thus convinced of the need to record the incidence of any practices which take place between people of the same gender, as a measure of homosexual behaviour, and not those which are stereotypically taken as characteristically homosexual.

B. Semantic Limitations

69Another key problem in formulating sexual orientation questions stems from the fact that the English language provides one word -'sex'- to describe both 'gender' and, for want of a better term, 'libidinous expression'. This gives rise to considerable opportunities for misinterpretation. Again, experience of other survey researchers prewarned us of certain pitfalls in respect of this semantic difficulty. A question on sexual orientation is included in the AIDS Strategic Monitor, a survey conducted by the British Market Research Bureau and commissioned by the Health Education Authority, in order to monitor response to AIDS public education. The survey is conducted on a rolling basis with some 300 respondents questioned every fortnight, accumulating roughly 8,000 each year. The question wording on this particular item was: "Have you ever had sex with someone of your own sex?"

70In the region of 4% the sample over a period of a year reported having such sex at some time during their life (whether once only, sometimes, often or always). Closer examination of the questionnaires, however, revealed that a small but significant number of these respondents (roughly a quarter of this subgroup, accounting for about 1% of the total sample) were at the same time reporting a lifetime's exclusively heterosexual experience and were also ranking themselves at TO'("exclusively heterosexual') on an identity scale on which 0 represented' exclusively homosexual'. What seemed to be happening was that these respondents were interpreting the phrase "own sex" not in terms of gender but in terms of sexual preference, ie. "People who have my own kind of sex", and so was a measure of predilection rather than orientation.

71This salutory experience was taken into account in the wording of the NATSAL questionnaire, and the wording 'same sex' used in preference to 'own sex', thus the decision was taken to use the terms 'opposite sex' and 'same sex' (but see below).

C. Devising a Rating Scale

72In addition to concerns over actual question wording, we were concerned to represent the gamut of sexual experience and activity as a scale rather than a dichotomous choice. It was obvious from both theoretical and empirical literature that most people are exclusively neither homosexual nor heterosexual, in inclination if not in practice, and that sexual experience is better expressed on a continuum than by a pair of fixed points.

73Again, Kinsey (1948: 617) cautions:

"One must learn to recognise any combination of heterosexuality and homosexuality in the histories of various individuals.
It would encourage clearer thinking on these matters if persons were not characterized as heterosexual or homosexual, but as individuals who had certain amounts of heterosexual experience and certain amounts of homosexual experience. Instead of using these terms as substitutes which stand for persons, they may better be used to describe the nature of overt sexual relations, or the stimuli to which a person eroticaly responds."

74The other influential factor in this decision was the poor result achieved in Gallup's earlier pilot study for the academic team centering on St Marys/Middlesex/Imperial, before the teams combined. Only two of the achieved sample of 500+ respondents reported same gender sex. Despite the clear and unambiguous wording of the question ('Have you ever had sex with a man?"-on the male version), the use of a simples "yes/no" dichotomy in preference to a scale might conceivably have allowed people to discount one-off experiences and report only their mainstream activities.

75The decision was therefore taken to use a Kinsey-style rating scale, but that it should be respondent-rated and not, as Kinsey's was, rated by the researcher. Additionally, unlike the scale used by Kinsey and colleagues, which combined both psychological reaction and overt experience, we opted for scales which were separate (but identical in terms of rating) for both sexual attraction and sexual experience, as used by Project Sigma.

76Project Sigma's questionnaire reads as follows:

We turn now to sexual orientation... basically, the question of how you consider yourself in terms of both your sexual activities and feelings.
Activity
So first of all, we are interested in what you do. How would you rate yourself in terms of your sexual activity?
... Are you?

0

Exclusively heterosexual

1

Mainly heterosexual but with a small degree of homosexuality

2

Mainly heterosexual but with a substantial degree of homosexuality

3

Equally heterosexual and homosexual

4

Mainly homosexual but with a substantial degree of heterosexuality

5

Mainly homosexual but with a small degree of heterosexuality

6

Exclusively homosexual

Feelings
This scale was repeated following the statement "Now I'd like you to rate yourself in terms of your sexual feelings".

77Note that sexual identity here is assumed to derive from sexual practice or activity. Because the sample used is a purposive one, one of the selection criteria being disclosure of some degree of homosexual activity, there is likely to be less of a problem of disparity between identity and activity (though there may be some). In the case of a general population sample, the same assumption cannot be made. Because of the stigma attached to a homosexual activity, people are likely to be more willing to describe same sex practices than to label themselves homosexual, and this was reflected in our questionnaire.

78Continuous piloting following the qualitative stage of fieldwork enabled us to test out the wording arrived at as a result of these considerations and by the time we were ready to embark on the feasibility stage of the survey we were satisfied that the question on orientation reflected the following concerns:

  1. that we should be measuring sexual activity and experience rather than self-perceived identity in relation to sexual orientation

  2. that sexual experience and activity should incorporate/include a variety of sexual acts and not simply penetrative sexual intercourse

  3. and that we should be measuring approximate position on a continuum, rather than polar oppositions.

79These considerations pointed to the following formulation:

80I have felt sexually attracted

  • only to a person, or people, of the opposite sex

  • mostly to people of the opposite sex

  • about equally to people of the opposite sex and to people of the same sex as me

  • mostly to people of the same sex as me

  • only to a person, or people, of the same sex as me

81I have never felt sexually attracted to anyone at all.

82I have had some sexual contact or experience:

  • only with a person, or people, of the opposite sex

  • mostly with people of the opposite sex

  • about equally with people of the opposite sex and with people of the same sex as me

  • mostly with people of the same sex as me

  • only with a person, or people, of the same sex as me

83I have not had any sexual contact or experience with anyone at all.

84It should be noted that the scale was condensed to a five point one, though additional opportunities for amplifying responses were provided by questions on same-gender experience in the self completion Booklet

D. Ordering the Rating Scale

85A final question to be addressed was the order in which the points on the rating scale should be listed. What remained to be determined was whether it would work better with the statistically more widespread behaviour pattern (sex with people of the opposite sex) at the top of the scale, or in reverse order.

86On the existing evidence, there seemed to be sound arguments for both versions. One form of question bias which often affects the answers of survey respondents is the order in which the choice of answer elements are offered, especially the order of presentation of the items in a rating scale. Although the literature is not entirely conclusive on this point, (Rugg and Cantril 1942, Payne 1951, Berg and Rappaport 1954), generally the evidence points to greater endorsement of the first item on a scale than the last, and to greater endorsement of the more 'favourable' or 'positive' response, and since questionnaires are conventionally set out with the favourable items first, these two effects are often additive.

87One way of attempting to eliminate or reduce this effect is therefore to reverse the scale so as to offset one effect against the other, that is, to place the least acceptable response first Given the 'first item' bias, and given the stigma attached to same sex (making 'opposite sex' the favourable option), it would seem logical to assume that any 'order effect' bias would be increased by putting the 'favourable' option of 'opposite sex' last on the scale.

88At the same time, we had to balance these possible question bias effects against considerations for the sense of the question. A scale with 'opposite sex' at the top might help clear confusion discussed above as to whether sex referred to gender or proclivity, since 'opposite sex' more obviously implies the first of these two.

89The opportunity was taken, in the feasibility survey, to pilot this part of the questionnaire in two versions. One version (A) was printed with 'same sex' at the top, while the other was printed with the rating scale in reversed order ('opposite sex' at the top). Approximately half the survey sample received version A and the other receive version B.

90Fortunately, questions contained in the self completion questionnaire (which formed the content of the booklet) afforded an opportunity to check for concurrent (or convergent) validity. Table X summarises the results for both interview and questionnaire.

Summary of homosexual questions

Summary of homosexual questions

91As is evident from these figures, the results from Version B (’opposite sex' first) showed greater consistency with the data from the self completion booklet, with 5.0% of men and 2.6% of women reporting homosexual contact or experience in the interview, and 8.9% and 2.6% respectively in the questionnaire, compared with 14.3% of men and 5.9% of women reporting homosexual contact or experience in the interview schedule using Version A ('same sex' first). It seemed to us unlikely that respondents would under-report in the Booklet which allowed greater privacy, and this was borne out by an analysis of the data. Of the total of 352 men who denied having had any same sex experience in the interview, 26 subsequently reported such experience in the self completion questionnaire in the Booklet.

92These results seemed to justify the conclusion that the source of possible bias was more likely to be misunderstanding of the terms used, and less likely to be the result of the order effect. It was therefore decided to abandon the terms 'same' and 'opposite' sex, and opt instead for the terms 'sex with a woman' and 'sex with a man'. Our earlier resistance to use the terms both woman and man (on the grounds that younger respondents might not identify as such), was waived in the interests of increased validity.

E. Sexual practices

93As previously discussed, it was clear from the developmental work that for heterosexual respondents, having sex was generally equated with vaginal intercourse. It followed that we could assume that questions containing the phrase "having sex" would be interpreted as having that meaning, so that it would be necessary to state explicitly which activities were being referred to. Thus questions intended to investigate the occurrence of any kind of sexual activity which could result in the exchange of bodily fluid was expanded for clarity.

eg. Q 2a) (Booklet)
When was the last occasion you had sex with a woman (that is vaginal intercourse, oral or anal sex)? (Emphasis added)
Q 3 (Booklet)
When, if ever, was the last occasion you had vaginal intercourse with a woman? (Men's version; enphasis added)

94We were alerted to, and therefore vigilant for, problems of misunderstanding of 'terms relating to sexual activity'. Most of the problems of comprehension were ironed out at the early qualitative stages of investigation, or else at the pre-feasibility study piloting stages. However, one or two emerged during the course of the feasibility study, an example of which is provided below.

95Since one of the main thrusts of health educational interventions encouraging risk reduction strategies has been to encourage people to adopt sexual practices which do not involve the exchange of bodily fluids, a central interest was in investigating how widespread such practices already are, that is, forms of sexual expression which avoid vaginal and anal intercourse and oral sex. The intention was to assess the prevalence of mutual masturbation, though it was clear from the development work that the term 'masturbation' was not generally acceptable.

96Question 3 in the Booklet addresses the issue of sexual practices and, following on from Questions 3a, b and c, which ask about vaginal intercourse, oral sex and anal intercourse, Question 3d originally asked:

When, if ever, was the last occasion you had any other form of genital contact with a woman, that dit not also involve vaginal intercourse, oral or anal sex?

97This was piloted in advance of the feasibility study with apparently no misunderstanding, but the result of the feasibility study was an astonishingly high proportion-40% (157) of men, and 44% (225) of women claiming never to have had genital contact without intercourse. The logical conclusion from the data was that some four out of ten respondents embarked on their sexual experience by being initiated into full on their sexual intercourse, a conclusion which we felt intuitively difficult to accept. Accepted wisdom has it that most people have had some sexual experience without intercourse, if at no other time in life, before their first time experience of full intercourse.

98It seemed more likely that this was the result of a large number of respondents misunderstanding the question and again, the qualitative work shed some light on this. In the knowledge that what sounded unfamiliar was often perceived as unorthodox and unacceptable, the expression "genital contact without intercourse" could have been perceived as some kind of sexual aberration or bizarre practice. Repiloting the question produced a more easily understood, if rather convoluted wording:

When was the last occasion you had genital contact with a woman NOT involving intercourse. (For example, stimulating sex organs by hand but not leading to vaginal, oral or anal intercourse).

IV. Veracity and recall

99Issues relating to validity and reliability are addressed more fully in another of this series of Working Papers, which specifically examines evidence from the data that the investigative tools measured what was intended and that the results were generalisable to the population as a whole. The concern here, as previously stated, has been with the opportunities for pre-empting possible sources of invalidity and unreliability and with devising strategies and mechanisms for minimising their effect.

100There are two outstanding issues, however, which particularly exercice those in doubt as to whether data from a study of sexual behaviour can be regarded as valid. These are the twin problems of veracity and recall, which, as is illustrated below, tend to be inter-related when we are dealing with, aspects of experience upon which the unconscious mind, in the form of involuntary responses and influences, exerts itself.

101In practice, of course, researchers into sexual behaviour have no monopoly on difficulties relating to veracity and recall; potential problems relating to people's ability to recollect accurately and report honestly their experience beset investigation into many aspects of human behaviour.

102Difficulties encountered in this respect were no greater than anticipated, and if anything, rather less so. Nevertheless, because of the widely articulated concerns in this area, the issues of veracity and recall have been amplified here.

A. Veracity

103It is widely assumed that people will be reluctant to tell the truth about aspects of their sexual experience. Yet we have no reason to believe that this is more true of sexual practices than it is of other aspects of human behaviour. Belson (1968:18) lists a number of possible topics on which people might be tempted to give less than honest replies, including drinking and smoking behaviour, frequency of taking a bath or shower, reasons for absences from work, negative views about racial minorities, earnings from all sources, etc. Researchers display few reservations about investigating these areas of behaviour on the grounds of doubtful disclosure.

104Clearly though, this is one area of research in which interviewer effect could threaten validity. The greater the sensitivity about an issue, the stronger the temptation to modify the truth. A question may occasion sensitivity for several reasons, because it probes socially disapproved attitudes or behaviour, because it deals with very private matters, or because it requires low prestige answers.

105A question on sexual behaviour may qualify on all these counts. Kudos attaches to some, and approbrium to other, sexual practices; just as it does to consumer practices, or eating and drinking, or earning money. Some of the behaviours we were asking respondents to reveal are not only socially disapproved, but are actually illegal in the UK (sex between a man and a woman, for example).

106There are a number of design features which can be built into the methodology to reduce this effect. One solution lies in posing "permissive" questions, those which by their wording imply acceptance of the behaviour in question. Some researchers on sexual behaviour (Coxon et al., 1988; Kinsey, 1948) advocate using leading qustions of the type "When did you last... ?" or "When was the first time you... ?". The use of such questions is generally criticized and warned against, on the grounds that they introduce bias in favour of over-reporting the behaviour in question. However, the inherent bias produced by leading questions may on some occasions intentionally be used to correct bias in the opposite direction, ie. the tendency to under report behaviour which is socially unacceptable, disapproved of, or felt to be unorthodox. Kinsey (1948: 53-54) gave the following advice:

"The interviewer should not make it easy for the subject to deny his participation in any form of sexual activity. It is too easy to say no if he is simply asked if he has ever engaged in a particular activity.
... Consequently we always begin by asking when they first engaged in such activity... and since it becomes apparent from the form of our question that we would not be surprised if he had had such experience, there seems to be less reason for denying it. It might be thought that this approach would bias the answer, but there is no indication that we get false admissions of participation in forms of sexual behaviour in which the subject has not actually been involved".

107This is the rationale behind the use of such questions by Project Sigma in the survey of the sexual behaviour of gay men. Some of their questions, "At what age did you have your first sexual experience with an animal?", however, come perilously close to being potentially offensive when used in the context of the general population.

108The technique of using leading questions was used in a more moderate form in the National Study of Sexual Attitudes and Lifestyles (NATSAL), questions of the type "When did you first...?" including a 'never' option amongst the response options.

109A guarantee of confidentiality can also do much to overcome this problem. Reassuring respondents of the anonymity of the survey was of greatest importance in relation to the self-completion booklet which contained the more intimate and personal questions. Answers to these were unseen by the interviewers who gave respondents instructions to place the booklet in the envelope and to seal it on completion. The interviewer was asked to add, if necessary (meaning if the respondent expressed any anxiety over confidentiality)

"The questions are quite personal and this way your answers will be completely confidential and I won't see them"
adding, to explain the identification process,
"We need to have a number on it in case it gets separated from the questionnaire. Our office can then check that all the documents for one person are completed, but names are never attached to answers".

110All of this was reinforced in the printed instructions on the booklet itself, which stated:

Confidentiality
The questions in this booklet are mostly very personal. Your answers will be treated in strict confidence; the interviewer does not need to see them.
When you have finished, put the booklet in the envelope and seal it. Your name will not be on the booklet or envelope.

111The importance of answering honestly was at the same time underlined:

It is very important to the whole study that you answer these questions completely honesdy and as accurately as you can.

112Finally, the status and authority of the agencies providing funding and acting as consumers of the data is not insignificant as a motivator to honesty. Respondents were asked, at the end of the interview they had answered honestly and whether they believed others would do so. The vast majority asserted that they had indeed given truthful responses, the stated reason in a large number of cases being the urgent need for the data and the credentials of the originators-a Government agency and well-reputed Academic Departments of Medicine.

B. Recall

113The problem with recall, in common with that of veracity, is not exclusive to research into sexual behaviour. Social survey researchers regularly rely for their data on the memories of respondents. Inevitably this introduces issues of validity. The literature on memory research is not extensive, but what studies there are point inescapably to the fallibility of memory-dependent information.

114Internal consistency checks can be valuable in checking on recall and their use is described in a later Working Paper. The concern here, however, is with methodological possibilities for reducing the effect of memory on validity.

1. Memory of Sexual Behaviour

115The most commonly voiced concern in relation to surveys of sexual behaviour is that people simply will not remember events in their past, the assumption being that memory decays progressively with time so that recent events are more reliably recalled than remote ones. Yet there are many instances where this effect is reversed. Older people often remember events in their distant past in detail whilst forgetting what happened yesterday.

116According to the psychological literature on the subject we cannot assume a direct correlation between the rate of forgetting a particular event and the time lapse since its occurrence. There may be a number of intervening variables, the emotional significance of an experience, its associations with other events, the motivation to retain the memory, etc.

117Certainly, events like first intercourse seem sufficiently emotionally salient to be recalled by most people without major difficulty. At feasibility stage the question was asked:

"How old were you when you first had sexual intercourse with someone of the opposite sex or hasn't this happened?" (Q 17)

118Only 8/557 (1.5%) of women and 7/420 (1.7%) of men failed to remember the age at when they first had sexual intercourse. The proportion of people who could not remember how long after their first sexual partner they had sex with another person was also small, 3.3% for both men and women.

119Particularly in relation to the disclosure of sensitive information, it is never easy to know whether inaccuracies are a product of simple lapse of memory, of uncounscious repression or of deliberate concealment. Forgetting may be a process of active blocking rather than passive decay. Kinsey reminds us that the effect of this can be to reverse any possible advantageous effects of immediate recall on accuracy since recent events may be most subject to censorship.

"The possibly greater accuracy of recent memory is at least partially offset by the greater extent of cover up on recent events... many subjects in a case history study will admit participation in the more taboo sexual activities at some time in the past; while insisting that such activities are no part of their current histories, or that frequencies are now very much reduced" (Kinsey, 1954:149).

120This may also affect accounts of numbers of partners, practices engaged in; etc. Investigating reported earnings against checks on other sources, Withey (1954) for example found that errors in recall were not random but varied with social expectations. Young people underestimated their earnings, older people over-estimated them.

121In this sense recall seems prone to something akin to the social desirability response. Here the boundary between the twin problems of veracity and recall becomes blurred. People may not lie intentionally but may be incapable of giving a truthful account because of the difficulty of retrieving censored information. To some extent then, the problem seems likely to respond to the same kinds of methodological devices designed to deal with the problem of honesty, ie, the formulation of neutral questions, a non-judgemental interviewing style and provision of assurances of confidentiality and anonymity.

122Another possible source of bias may occur in a reliance on respondents recollections of their lifetime sexual partners. Recall will be easiest for those with fewest partners, and for those whose relationships have always been socially institutionalised in some way, eg. through engagement or marriage. The resultant increased relative invalidity in those respondents who have large numbers of casual partners is worrying since the experience of these respondents may be of greater interest to those requiring the data. Nevertheless, it seems worthwhile exploring by what means and what efforts bias caused by imperfections in recall might be reduced.

2. Aiding the Process of Recall

123Large scale quantitative surveys admittedly have their limitations in eliciting data which is difficult to recall. Recall, for most people and for most experiences in life, can be more a process of gradual reminiscence than instant recollection. Probably the most successful techniques facilitating this are those involving a slow, careful process of retrieval more akin to the therapeutic method than scientific research.

124Good qualitative work can do much to uncover hidden experiences in the respondent's past, and our own experience in the early in-depth interviews was that respondents did a good deal of 'remembering' during the course of an interview. By contrast, the fully structured questionnaire allows only a 'one shot' approach to response which does not permit reflection and revision. What comes to mind first may not always bear the closest correspondance with reality.

125We were forcibly reminded of this in the following dialogue in one of our early in-depth interviews:

I.

Let's keep to that first sexual experience, in your mind was it heterosexual or homosexual?

M.

What I would term as a sexual experience is the heterosexual, up until then it had been... (hesitation)

I.

Yes?

M.

I didn't really consider it important.

I.

You wouldn’t judge it as a sexual experience?

M.

Technically, it was a sexual experience but in my own mind it's something I would prefer to forget

M; 26-35 years; single

126As is often the case, this man had reconstructed his experience by filtering it through a lens of what is culturally considered normal and acceptable. Asked baldly and simply whether his first sexual experience was with a man or a woman, he would no doubt have answered 'with a woman'. The kind of reflection and retrospective analysis which is needed to tease out the effects of this self-censoring mechanism (which may not even be conscious) is not practically possible using the research techniques and instruments planned in this survey.

127All the same there are some kinds of considerations which can be brought to bear on the design of the questionnaire to make the process easier. Scheduling the question order such that the early section includes a fair number of the 'milestone' variety of lifestage questions ("When did you first... ?") provides respondents with a framework in which to locate less easily recalled experiences, and may trigger associations between life-events.

128Questions on sexual behaviour were prefaced in the NATSAL interview schedule by a number of life-event questions which prompted respondents to cast their minds back and probe the past in advance of the most important stage of investigation.

129For example:

Q2d)

"In the last five years have you had any (other) illness or accident that affected your life for at least 3 months?"

Q-

"How old were you when you first lived or stayed away from your parents' home for any length of time?”

Q 14b)

"How old were you when your (first) child was born?"

3. Single Occurrence Events Versus Habits

130Events like first intercourse, marriage, births of children, etc., occur only once and by that fact alone have advantages for recall. The literature on memory research confirms that incidence is easier to recall than frequency.

131Thus questions such as "When was the first time you... lived with a partner/lived away from home?", for example, or single occurrence events (birth of babies, marriage, etc.) might therefore be expected to produce more valid resultats than regular occurrences (eg., frequency of intercourse). Knowledge of this effect might be turned to methodological advantage simply by converting as many questions as possible into those requiring incidence as the answer in preference to frequency. For example, if one purpose of the instrument is to establish average frequencies of intercourse in the general population, a more reliable method might be to avoid questions of the type "How often do you have sex?'" substituting instead "When did you last have sex?".

4. Fixed Versus Open-Ended Time Periods

132The evidence is that forgetting mainly takes the form of forgetting when, rather than forgetting completely. Using data from the British Survey of Sickness, Gray (1955) found that people had difficulty remembering the actual dates when medical consultations or bouts of illness occurred even though the reported values for incidence closely approximated to the true values (as entered in medical records). We were mindful of this effect in asking ourselves whether we should ask about a period fixed in time, eg. a particular week, or whether we should ask about a period ending on the day before the interview, ie, in the last seven days. This issue was considered especially in relation to sexual partnerships and sexual practices.

133Gray's experience was that people tend to transfer into the period in question, events which actually took place outside of it, at either end. This seemed to present a case for taking the memory period up to the period immediately before the interview since this closes the door to one set of transfers. Bringing the period in question right up to the present gives less scope for over-reporting since respondents can bring events forward into the time period but they cannot anticipate those which have yet to occur.

5. Question Order

134We were aware that the order in which different time periods are set out in the format of a question might also influence recall.

135The issue of recall wass particularly important in relation to our questions requiring of respondents an estimate of their numbers of sexual partners over periods of different lengths, a lifetime, five year period, two years, etc. Ranking more recent events higher, and more distant ones lower might, we reasoned, act as an aide memoire, facilitating the process of summation. In order to assess the effect of order in our questionnaire two versions of the questionnaire were piloted. Version A set out the shorter time period at the top, and the longer at the bottom, and Version B was in reverse order. In fact, Version B proved the more satisfactoy alternative.

Bibliographie

References

References include both those cited in the text together with other useful publications.

AIMN (1988), AIDS in a Multicultural Neighbourhood. Bayview-Hunter's Point Foundation for Community Improvement, Questionnaire provided by Director, Mindy Fullilove, 6025 Third Street, San Francisco, California.

Belson W.A. (1981), The Design and Understanding of Survey Questions, Gower, London.

Belson W.A. et al. (1968), The development of a procedure for eliciting information from boys about the nature and extent of their stealing, Surrey Research Centre, London.

Belson W.A., Didcott P.J. & Millerson G.L. (1968), Identifying difficulties and facilitating factors in getting information from boys about their stealing and about associated matters: an exploratory study, Survey Research Centre, London.

Belson W.A. (1966), The effects of reversing the presentation order or verbal rating scales, in Journal of Advertising Research, vol. 6 (4).

Belson W.A. (1966), The ability of respondents to recall their purchases of chocolate confectionery, Survey Research Centre, London.

Benson S., Booman W.P. & Clark K.E. (1951), A study of interview refusals, in Journal of Applied Psychology, vol. 35 (2).

Bradburn N.M. & Mason W.M. (1964), The effect of question order on responses, in Journal of Marketing, vol. 1 (4).

Cantril H. & Fried E. (1944), The meaning of questions, in Cantril H. et al., Gauging public opinion, Princetown University Press, Princeton.

Cati, lah & IDS Questionnaire, available from Research Unit in Health and Behavioural Change, University of Edinburgh, 17 Teviot Place, Edinburgh EH1 2Q2.

Collins M. & Courtney J. (1983), The effect of question form on survey data, Market Research Society.

Colombotus J. (1965), The effects of personal vs telephone interviews on socially acceptable responses, in Public Opinion Quarterly, vol. 29.

Ebbinghaus H. (1885, translated 1913), Memory: a contribution to experimental psychology.

Fay R.E., Turner C.F., Klassen A.D., Gagnon J.H. (1989), Prevalence and patterns of same gender sexual contact among men, in Science, 243: 338-48.

Forman D., Chilvers (1989), Sexual behaviour of young and middle aged men in England and Wales, in Br. Med. J., vol. 298: 1137-1142.

Gray P.G., The memory factor in social surveys, in Journal of the American Statistical Association, vol. 50 (270).

Hanson, R.H. and Marks E.S. (1958), Influence of the interviewer on the accuracy of survey results, in Journal of the American Statistical Association, vol. 53.

Hunter I.M.L. (1957), Memory, Pelican.

Johnston A.M. (1988), Social and behavioural aspects of the HUV epidemic- a review, in Journal of the Royal Statistical Society, series A, 151: 99-114.

Keating E., Paterson D.G. & Stone C.H. (1950), Validity of work histories obtained by interview, in Journal of Applied Psychology.

Krugman H.E. (1977), Memory without recall, exposure without perception, in Journal of Advertising Research, vol. 17 (4).

Mant D., Vessey M., Loudon N. (1988), Social class differences in sexual behaviour and cervical cancer, in Community Med., 10: 52-6.

McGeogh J.A. (1932), Forgetting and the law of disuse, in Psychological Review.

Project Sigma (March 1989), Briefing Document: A Digest of Early Findings of Project Sugma, Department of Social Sciences, South Bank Polytechnic.

Quinn S.B. & Belson W.A. (1970), Thought processes and accuracy in the recall of purchases: petrol buying, in Survey Research Centre, London.

Quinn S.B. & Welson W.A. (1969), The effects of reversing the presentation order or verbal rating scales in survey interviews, in Survey Research Centre, LSE, London.

Ruhbc (March 1989), CATI: LAH AIDS Questionnaire by month, Research Unit in Health and Behavioural Change, University of Edinburgh, 17 Teviot Place, Edinburgh, EH1 2QZ.

Schofield M. (1973), The sexual behaviour of young adults, London: Allen Lane.

Schofield M. (1965), The sexual behaviour of young people, London: Longman.

Smith H.L. & Hyman H. (1950), The biasing effect of interview expectations on survey results, in Public Opinion Quarterly, vol. 14 (3).

Spencer L. (1988), Talking About Sex, SCPR.

Stember H. & Hyman H. (1949), How interviewer effects operate through question form, in International Journal of Opinion and Altitude Research, vol. 13.

Sudman S. & Bradburn N.M. (1982), Asking Questions, Jossey Bass, San Francisco.

United Kingdom Family Planning Research Network (1988), Patterns of sexual behaviour among sexually experienced women attending family planning clinics in England, Scotland and Wales, in British Journal of Family Planning, 14: 74-82.

Wellings K. (1989), Researching sexual behaviour, Radical Community Medicine, first quarter.

Withey S.B. (1954), Reliability of recall of income, in Public Opinion Quarterly, vol. 18.

Table des illustrations

Titre Summary of homosexual questions
URL http://books.openedition.org/pusl/docannexe/image/14138/img-1.jpg
Fichier image/jpeg, 103k

Auteurs

Academic Department of Public Health, St Mary's Hospital Medical School, London.

Social and Community Planning Research, London.

Academic Department of Genito-urinary Medicine, Middlesex Hospital, London.

Academic Department of Public Health, St Mary's Hospital Medical School, London.

Academic Department of Public Health, St Mary's Hospital Medical School, London.

© Presses de l’Université Saint-Louis, 1990

Conditions d’utilisation : http://www.openedition.org/6540