The development of an epidemiology of performance enhancing drug (PED) use in sport has been restricted by the absence of a reliable and valid indicator of drug use [1, 2]. Where conventional drug testing indicates prevalence around 2% , estimates from "how many people do you know who use PED" indicate the prevalence to be 6%  and in anecdotal reports up to 95% . The absence of a reliable indicator has significant implications for assessing the value of interventions to ameliorate or eliminate drug use in sport. There is, however, the potential to develop a self-report measure using a known bias in human perceptions of social behaviour, the False Consensus Effect (FCE).
As noted above, prevalence estimates for PED use in sport range from 2% to 95%. Such a wide range of estimates indicates that there is poor evidence about the actual prevalence rate. That is, there is no reliable epidemiology of PED use among athlete populations . Laure  reports an attempt to develop an epidemiology of androgenic-anabolic steroid (AAS) use in France, suggesting that 10–20% of athletes use such substances regardless of age, sex or sport. A comprehensive study across six European countries that relied upon self reports among university students indicated that 2.6% were willing to admit use of PED .
The failure to generate an acceptable epidemiology is predicated based upon the methods used to detect the prevalence of PED use in sport being flawed. The administrative, financial and scientific constraints of biomedical testing have become received wisdom with acknowledgement of the drugs in sport 'arms race' between new drugs and detection technologies. That is, biomedical detection is unlikely to give an accurate indication of prevalence due to the combination of an inability to test universally and the introduction of drugs undetectable by contemporary methods .
The application of typical social science methods to generating estimates of prevalence leads to problems such as the reliability of self report, non-response bias or social desirability . Who is asking the question may also contaminate the response; it would be a brave athlete who admitted to PED use on a survey run or sponsored by a National Anti-Doping Organisation. Likewise, it would be scientifically invalid to infer anything about substance use behaviour from stated attitudes or intentions towards PED use given the tenuous relationship between the two.
With the failure of typical biomedical or social science approaches to provide a basis for developing an epidemiology of PED use in sport, atypical approaches are called for. One such atypical technique called the 'Random Response Technique' (RRT) has proven to be more reliable when sensitive issues such as abortion, illicit drug use, opinion about capital punishment or shoplifting are investigated [9–11]. Using the RRT, Simon and colleagues recently showed a comparatively high (12.5%) prevalence of doping use among gym users .
The aim of this paper is to propose an alternative indirect approach which has been used in sociology but is new to doping research and relies on social projection. The notion of social projection was introduced more than 80 years ago  and the method has been extensively used in social psychology [14–19]. The false consensus effect arose from psychology's efforts to explain discrepancies in social judgement. Specifically, the effect describes the considerable overestimation of behaviour in which a person engages, and a slight underestimation of behaviour absent from a person's repertoire . That is, over-estimating a particular behaviour indicates that the person who makes the estimate (and overestimates the behaviour) is likely to be engaged in the same act. Research regarding attributive projection (the tendency of people to project their own characteristics onto others) , the FCE and uniqueness bias have been particularly pervasive in social psychology . According to the FCE theory , individuals often tend to overestimate the extent to which others behave the same way as they do, especially if the behaviour in question is deemed to be socially questionable or unacceptable. This phenomenon is explained by a part motivational, part cognitive process resulting in people believing that their own action is a relatively common behaviour. The effect appears to be present even when objective statistics and information on the bias effect are provided, indicating the intractable and egocentric nature of this biased social perception .
For example, self reporting marijuana smokers overestimated the proportion of users in the general population by 28% whereas non-smokers of marijuana overestimated the rate of use by 14% . The directions of these estimations were congruent with the self-reported behaviours (i.e. non-users under-estimated and users over-estimated) in a study regarding students' use of amphetamines. In this report students who abstained from amphetamines typically underestimated (estimate 29% versus 35% reported) and users overestimated (estimate 48%) prevalence of amphetamine use but not other behaviour, suggesting that this FCE is behaviour-specific and does not generalise to other similarly ostracised acts .
Recent marketing research investigating consumer behaviour demonstrated that overestimation is greater when an individual holds positive feelings toward the subject . In addition to finding further evidence for the FCE, Monin & Norton  also demonstrated the existence of a strategy people use to justify their undesirable behaviour. This strategy typically involves justification based on the sense of comfort in large numbers (i.e. many are doing so) or citing special mitigating circumstances. It was also shown that bias estimation (whether over or under-estimation) is rooted in the social perception of the behaviour, not in the behaviour itself . The estimation of others' behaviour was influenced by the combination of two conditions: i) the person's own behaviour and, ii) what was desirable in the given situation. As such, estimation bias may change over time as one or both of these conditions change.
Whilst its causality has remained unknown, the relationship between self-involvement and overestimation has been repeatedly evidenced with regard to smoking, drinking and illicit drug use [24–26]. It has been suggested that perceived prevalence may act as a normatively prescribed behaviour  and actually initiates the behaviour. For example, if emerging athletes believed that using PEDs is necessary to be successful in high performance sport and that everyone uses PEDs, this belief may work as a perceived norm for these athletes and motivates them to do as the others and start taking PEDs. While it is a plausible application of the FCE, its validity requires further evidence, preferably from longitudinal studies. For the purpose of the present proposal, it is sufficient to assume that significant overestimation signals involvement, namely doping use or intention to use.
Estimation of prevalence has also appeared in doping research. Pearson & Hansen's study of athletes at the 1992 Winter Olympics provides an insight into how the FCE might work in an anti-doping context . In this study, athletes were asked to estimate the prevalence of doping or certain PEDs among their peers. For example, where the reported positive cases vary around 2% , 67 of 155 athletes (43%) surveyed by Pearson & Hansen thought that more than 10% of athletes in their sports used anabolic steroids, and a further 53 (34%) gave an estimate between 1% and 9% . A survey conducted among Finnish Olympic athletes revealed similar results. Whilst none admitted using PEDs, 42.5% from stress power sports and 37.0% of endurance athletes reported that they personally know another athlete who uses PEDs .
In the context of a review for WADA, Backhouse and colleagues report that unvalidated self-reported PED use among elite athletes typically ranges between 1.2% and 8% . Conversely, projective techniques where athletes are asked to estimate how many team mates or competitors used PED, the estimate increased to between 6% and 34%. This divergence in estimates appears large for random sampling differences and may be better explained by the FCE. Using FCE-based surveys may equip researchers, policy makers and health care professionals with a more realistic estimate of PED use by the athlete population. It is envisaged that in its broader aspects, this study would help to provide guidance for the general population with respect to PED use, particularly for non-prescription anabolic steroids, amphetamines and/or analgesics.