There has been a long history of challenges regarding aspects of condom use. Numerous studies have attempted to understand factors that account of variation in use, be these related to user characteristics (knowledge, attitudes, gender, etc), actual and/or perceived partner characteristics (gender, perceived risk category, assertiveness skills, etc), immediate contextual factors (relationship status, location, condom availability, the impact of desire, alcohol, etc), distal factors (pricing, perceptions of quality, the impact of sex education programmes, etc), and other factors. Very many articles have been published in a very wide range of journals and more popular media.

The vast majority of such articles have used just one or two simple questions to assess condom use as the outcome variable. Some studies explore intended use, while others ask about actual use. Questionnaire items normally cover one or more of the following—use at first intercourse, use at most recent intercourse, use at first intercourse with most recent partner, etc—using a binary response option. Others ask about regularity of usage over some time period, generally using a Likert type scale ranging from ‘never’ to ‘always’ with varying numbers of points (normally three or five) in between these extremes.

There are, of course, immense challenges regarding the accuracy of the data collected. Although the selection of ‘always’ and ‘never’ provide unambiguous anchor points, there is likely to be fairly wide variation in the ways in which terms like ‘sometimes’, ‘almost always’, and other common scale point are interpreted. Many of the other more obvious shortcomings of measurement have been fairly well documented, including poor memory, not being able to recall due to intoxication or drug use, falsification of responses, the ambiguity of terms (including whether having ‘used’ a condom can be regarded as indicating use on all occasions of sexual activity during that particular episode), and so on.

In more recent years, greater attention has been given to the reported quality of condom use, as opposed to just the reported quantity. So, rather then just asking about whether one was used or not, and/or how often, participants in such studies are also asked about whether there was any breakage, slippage, late application, early removal and so on; clearly, such questions can only be asked retrospectively.

Research on these so-called condom ‘errors’ or ‘problems’ has been carried out for many years in different parts of the world. Richters et al1 reported on slippage and breakage in Australia way back in the early 1990s, Spruyt et al2 and Warner et al3 reported US data; Mekonen and Mekonen4 reported prevalence in Ethiopia, Bradley et al5 reported on sex workers in India, and de Visser and Smith6 wrote about the methodological implications of inefficient usage. Data from the UK were reported by Hatherall et al.7

From the USA, there has been a veritable outpouring of papers on the topic in recent years, many of these emanating from the Condom Use Research Team coordinated from The Kinsey Institute; the team members are Sanders, Yarber, Graham, Milhausen and Crosby. A range of methods has been used, including traditional questionnaires, diaries, interviews and electronic devices, and a range of populations have been sampled, including sexual enhancement product mail order clients, inner city sexually transmitted infection clinic users, and, of course, the ubiquitous university undergraduates. (see http://www.kinseyinstitute.org/research/condom_errors.html)

A recent review published in a Special Issue of Sexual Health8 listed prevalence levels of incomplete condom use and other forms of error reported in 50 articles from 14 countries with different definitions, populations and reporting time scales being used, prevalence of errors varied widely, but most studies reported genuinely high rates of incorrect usage of one form or another (up to 50% of occasions during the target time period for some errors). One study, based on a dairy study in the UK, reported that over one third of participants had used condoms incorrectly and/or experienced a breakage/slippage on at least one occasion during the previous ten occasions of intercourse,9 and this rate is not out of line with many other studies.

So, reported prevalence of errors is disturbingly high. This has immense implications for various reasons, including those concerned with the development of theoretical models of health behaviour as well as those using self-reported behaviours to assess public health interventions of one kind or another. The use of one or two simple questions regarding the use or not of condoms per se can no longer be taken as a reliable or valid index. One can assume that all those who used one incorrectly would still have answered ‘yes’ to the question ‘Did you use a condom?’, thereby introducing immense levels of noise and uncontrolled variance into the system.

But self-report, with all its concomitant problems, is one matter. What is so valuable about the Crosby et al¹⁰ paper in this current issue of this journal is that it directly relates self-reported errors to actual outcomes, and shows that reported consistency of use does not predict sexually transmitted infection acquisition, whereas reported quantity of use does so; the use of contemporaneous diary keeping—and biological verification of infection—add to the likely reliability of the data. This study provides much clearer and more direct evidence than has hitherto been available that quality of use matters.

However, what the field in general lacks is a good solid theoretical analyses of specific (personal, relationship and contextual) characteristics that are associated with specific types of errors and problems. Many studies report little more than simple descriptive data on generic ‘errors’ and ‘problems’. Sanders et al8 list many different categories of ‘error’, but their Condom Experience Model makes no distinction between types of ‘error’—although it is a start.

It may be worth considering a parallel from work on young drivers and accident probability. Parker et al11 proposed that driving problems could be categorised into violations (intentional breach of rules or advice), errors (misjudgements) and lapses (absent minded errors). Future research in sexual activity could explore whether, for example, condom use violations (intentionally not using one when it is clear that one should be used—possibly because of not wishing to spoil the experience, not wishing to expose oneself to ridicule, not wishing to risk losing an erection, being a reluctant user, etc) are associated with different demographic, psychological and contextual processes than are errors (simply not knowing how to best avoid the risk of breakage, or tearing accidently, or just using for ejaculation, for example) and lapses (merely forgetting or not preparing properly due to the effect of drink or drugs or tiredness or desire). Different aetiologies of ‘problems’ may well turn out to have different implications for possible public health interventions.

Meanwhile, however, the simple question ‘Did you use a condom’ will no longer be sufficient to use for improving models of health behaviour or assessing the impact of interventions, other than at a very general level of analysis.