The potential impact of sport to promote healthy lifestyle is seriously undermined by the presence of doping practices, which has spread beyond the elite sport [1, 2]. The elevation of the use of performance enhancing drugs among adolescents [3–5] and even pre-adolescents  is a particularly worrying trend. To date, intervention strategies have appealed to moral values and health consciousness of the athletes. Effectiveness has typically been evidenced by changes in self-reported behaviour and/or explicit attitudes towards using prohibited substances. Despite the weak evidence, this approach is based on the assumption that athletes' actions are exclusively motivated by conscious cognitive processes. This may not be true. The fundamental questions to be addressed for effective intervention programs are: what is it that really drives highly skilled and motivated athletes to risk their health, reputation and future participation by engaging in doping practices? Why risk losing a future in one's chosen sport by committing an act that goes against the fabric of fair play and ethical behaviour? Are reasons always objectively evaluated and clearly articulated factors that can be measured by some explicit tools (e.g. questionnaires, interviews) or rather, may decisions also be influenced by values below the explicit awareness?
A whole range of factors impact on athletes' decision to take substances that are intended to provide a performance advantage. Some of these factors are well articulated by athletes themselves. The plethora of reasons athletes put forward to justify their doping practices is well known to the practitioners of sport psychology, whose case studies (e.g. Terry ) report reasons such as i) 'Everybody's doing it', ii) 'I don't want to but it's the only way to compete', iii) 'The doctors can't fix my injury, what other option do I have?', iv) 'I know what I'm doing, I won't get caught'; v) 'I still have to do the work' and vi) 'I'll do whatever it takes to win'.
Psychological analysis of doping behaviour has so far concentrated on individual differences in attitudes towards drug use [8, 9] and towards drug testing programs . What is not well understood are the underlying psychological mechanisms of the use of performance enhancing substance and methods in sport. Few studies have examined social and moral concerns  and achievement orientation  as potential candidates for such mechanisms. More recent reviews have catalogued co-morbidity factors such as ego-oriented achievement striving and motivational climate ; as well as narcissism, depression, lack of self-confidence, eating disorders, body image imbalance, dispositional propensity to risk-taking and suicide . The role of attitudes in shaping behaviours has been widely recognised [15–17] hence understanding athletes' attitudes and behavioural intentions towards performance enhancement is critical to informing anti-doping intervention strategies.
Whilst athletes' attitudes and beliefs are central to most recent social science research into doping [4, 18, 19], they are typically measured by self-reports . Capturing the complexity of these attitudes beyond verbal declarations requires complementary assessments using alternative methods. This paper examines the utility of one such method – the Implicit Associations Test (IAT)  – in assessing implicit doping attitudes in comparison to the self-report derived assessments. Its ability to capture deeply-rooted, more stable, unconscious or introspectively inaccessible representations could complement the traditionally used explicit assessments and make vital contribution to the understanding of drives behind doping behaviour.
The concept of implicit attitudes has been widely used in social psychology with a variety of measurement techniques  and its fundamental propositions have been supported with varied success . Applications of implicit attitude measurement have increased in several areas, including health and exercise, where typical attitude targets include obesity [24–27], body weight in general , exercise  or health-compromising behaviours such as smoking , drinking [31, 32] or drug use . Implicit associations were contributed to the prediction of their respective behaviour in health compromising behaviour studies. Interestingly, while people with obesity showed a more positive implicit attitude towards food , alcohol studies evidenced the opposite: in general, heavy drinkers exhibit negative, neutral or ambivalent attitude towards alcohol in comparison to soft drinks but score higher on the alcohol-arousal [34–36].
However, when personalised implicit assessment was used, the implicit association with alcohol was positive suggesting that alcohol implicit assessment reflect negative extrapersonal knowledge . It was evidenced in all cases that implicit association play an influential role in alcohol use and misuse.
Weak correlation/dissociation is expected for controversial attitude constructs, especially those associated with social stigma . In the IAT literature, there are two fundamental explanations for the hypothesized weak relationship between implicit and explicit attitudes: i) response distortion (both intended and unintended, affecting explicit attitudes) and ii) the existence of dual attitudes. It has been assumed that implicit attitudes may be influenced by: i) early affective experiences, ii) systemic cultural view of the target, iii) cognitive consistency principles , iv) translation between the implicit-explicit representation, v) social desirability, vi) situational pliability and vii) research design issues . This is consistent with the model of dual attitudes , which argues that a new attitude (acquired from experience later in life such as being in sporting situation where doping is present and manifest in explicit attitude) does not necessarily replace the older one but results in dual attitudes by holding different evaluations (implicit and explicit) of the same attitude object. It is assumed that the degree of co-influence of the implicit and explicit attitudes depends on the situation, the cognitive capacity to retrieve explicit evaluation  and many other factors.
On the contrary, the Iterative Reprocessing Model (IRM, ) provides evidence from fMRI investigations showing that while implicit and explicit attitudes serve different purposes, both automatic evaluations (implicit attitudes) and conscious self-reflection (explicit attitude) are not separated and both are equally important. The IRM suggests a continuous evaluative cycle between automatic evaluation (which occurs early in the evaluative process) and the relatively stable reflective evaluation (attitude). Assuming that both implicit and explicit association play an influential role in evaluating doping related situations, intervention strategies should target implicit an explicit doping related expectations.
The distinction between explicit and implicit attitudes raises the question of predictive power of implicit attitudes. This distinction is known to be highly task and context-dependent (for review, see Fazio & Olson ). Both explicit and implicit attitudes are involved in the decisions about behavioural intention and behaviour execution to varying degrees. According to Fazio's Motivation and Opportunity as Determinants (MODE) model [45, 46], explicit attitudes are the amalgamation of implicit attitudes (automatic responses) and verbal responses to the attitude object affected by the motivation and opportunity to deliberate and respond in a strategic way.
The importance of implicit values in behaviour or behavioural intention is a key research question and attempts have been made to incorporate implicit attitudes into the traditional behaviour models [47, 48]. Despite the fact that implicit attitudes are known to have stronger effects on behaviours that are less controlled by deliberate conscious processing , such as spontaneous responses, intuitive action or increased action readiness and sensitivity to the attitude-relevant situational cues, results obtained via implicit assessments, in combination with explicit measures, have improved the prediction of behavioural models in various domains requires deliberation [50–55].
At the time of writing, implicit attitude measurement has not been applied to doping attitudes. The study reported in this paper examined the validity of this doping-related IAT protocol (PE-IAT) by, first, ascertaining the magnitude of the anti-doping evaluative bias among sport and exercise science graduates and, second, by triangulating it against alternative estimates of doping attitudes. One advantage of the method is that it is not susceptible to deliberate response distortions (e.g. faking), which is especially important in doping related research [56–58]. Subjects trying to deliberately manipulate IAT do so by slowing their responses to desirable combinations but the improved IAT scoring procedures tend to cope with this strategy by eliminating very slow responses from the analysis. As the alcohol-related cognition studies [34–37] have shown, the IAT-type tests, however, can be contaminated by associations that are stored in memory but irrelevant to the individual's personal virtues [59, 60]. Implicit assessments are thought to have the potential to tap into unconscious or introspectively inaccessible attitudes that are deeply-rooted in long-term socialisation .
Assessment of doping attitudes
Despite their widespread use in sport psychology, self-report techniques have a range of substantial limitations. Most of these limitations stem from the following two assumptions: the test taker is assumed to (a) be able to self-report and (b) be willing to self-disclose. In other words, the test taker is assumed to have sufficient insight into what's being measured yet no intention to distort his or her responses. Accepting the assumption of the ability to self-report is a relatively safe bet, at least in dispositional trait assessment, as most trait markers tend to be universally understood . The second assumption, however, remains wide open and as such, it should be scrutinised for each individual measurements. In particular, social desirability is known to contaminate questionnaire-based attitude measures, and this contamination tends to escalate with the increasing sensitivity of question content. Doping is a highly sensitive issue for those who are involved in sport, especially for those who derive their livelihood from it. Multiple pressures are likely to prompt athletes to conceal their attitudes towards doping if they are lenient. Not surprisingly, self-reported doping attitudes have shown a significant association with socially desirable responding, even when anonymous questionnaires are used . Violations of either of these two assumptions can compromise the validity of self-report assessment and call for caution about relying solely on data derived from self-declarations .
Towards an alternative attitude measure: Implicit attitudes and associations
Ideally, self-reporting methods should be complemented with alternative measurement techniques. One of the promising alternatives is the Implicit Associations Test (IAT) . Based on the widespread definition of attitude as a relatively stable tendency to evaluate objects with degrees of liking or disliking, implicit attitudes are defined as associative processes reflecting this tendency . Associative evaluations are the result of stimulus-driven, uncontrolled, unintentional, goal-independent or unconscious processes  and as such, they do not require respondents to be aware of these attitudes and thus can provide a solution for the problem of self-presentation distortion.
The IAT is capable of measuring automatic effects of implicit attitudes without relying on self-reports by extending the cognitive task-switching paradigm to the realm of timed semantic classification, typically in a semantical decision task. The use of semantic targets (words) opens the prospect of measuring the evaluative strength of a wide range of implicit associations between social objects . The IAT assesses the generic differences in implicit attitudes by measuring the underlying automatic activation of cognitive processes. Initial validation of the IAT has shown its sensitivity to individual differences in implicit effects of self-esteem and self-identity , attitudes , and stereotyping [65–68], with no evidence of procedural limitations  or familiarity of stimulus  acting as confounding variables. The core of the method is a semantic discrimination task performed under time pressure: participants have to classify sematic targets (usually words or pictures) on computer screen into one of two opposing categories (e.g. pleasant vs. unpleasant) by pressing the respective response keys on the keyboard (e.g. right hand side – left hand side). The attribute-concept associations are assessed by combining a target
First, the categories represent concept discrimination: me – not me, or doping-supplements. Then another category pair is used for attribute discrimination – e.g. pleasant -unpleasant.
Finally, the two tasks are combined: semantic targets have to be categorised into one of four categories (two pairs). This combined task is presented twice, the change in the repeat presentation is the reversal of response keys for one category pair (e.g. first RH response is required to 'pleasant' category and LH response to 'unpleasant' category, then on the repeat presentation LH response is required to 'pleasant' and RH response to 'unpleasant'). The IAT assumes that the simultaneous presentation of the two tasks makes strongly associated (compatible) attribute-concept pairs easier (and hence, faster) to classify when their responses are mapped on the same response keys . Given the instruction to respond with maximum speed and accuracy, response times in combined tasks are assumed to depend on how compatible the categories on each side of the computer screen are in peoples' minds . For example, the me-unpleasant combination has been shown to produce slower response times than the me-pleasant . Further, the difference in response times between the two combined tasks (the initial and the reversed one) were shown to vary substantially across the sample, and were conceptualised as representing the underlying individual differences in the corresponding implicit self-esteem . The response time difference, or the IAT effect , essentially constitutes an estimate of the strength of the subject's implicit attitude.
The IAT method has been particularly effective in capturing associations with evaluative attributes (e.g. pleasant versus unpleasant) and thus in measuring implicit affect/attitudes. The method also has the potential to deal with more descriptive attribute dimensions (such as trait descriptors) which may result in a meaningful assessment of implicit individual differences . Categories in the semantical decision tasks of the IAT can be constructed to represent forms of performance enhancement that are acceptable (e.g. nutritional supplements) or unacceptable, such as anabolic steroids, growth hormone and other substances or methods prohibited by the World Anti-Doping Agency. Combining these categories with IAT 's standard connotative evaluation dimension enables one to estimate the magnitude of automatic evaluative preference (i.e. bias)  in favour or against the performance enhancement categories, which can be interpreted, following the IAT tradition , as athletes' implicit attitudes towards banned substances and methods.
Supplement use to enhance performance, in general, is widely accepted in the sporting community. Therefore the key question was the attitude towards banned substances in comparison to permissible supplements. Hence, the aims of this study were to i) adapt the IAT protocol to categories of performance enhancement and ii) to examine the construct and concurrent validity of the PE-IAT by complementing the test with two different self-reported attitude measures and demographic data.