Regular physical activity and high physical fitness are associated with numerous health benefits such as reduced risk of premature death from all causes and cardiovascular diseases, type 2 diabetes, specific types of cancer (i.e., colon and breast cancer) and osteoporosis (see Warburton, Shannon and Bredin  for review). However, depending on survey methods, it is estimated that only 25% to 40% of the adult population reaches the physical activity level recommended for health benefits [2, 3]. This observation highlights the need to devote more attention to factors explaining why some people are active while others remain sedentary.
In the Theory of Planned Behaviour (TPB; Ajzen ), considered as one of the most useful theories to study the cognitive determinants of behaviour, intention is a key predictor of behaviour in a wide range of health domains. Reviews and meta-analyses provide empirical support for the predictive power of intention, indicating that it accounts for 20% to 40% of the explained variance of physical activity behaviour [5–7]. According to Cohen , this constitutes a medium to large effect size. Nonetheless, there remains a gap between intention and action , caused mainly by those who express a positive intention to exercise but do not act; this group represents about one third of the population . This observation is congruent with the analysis of Sheeran , who also identified this group as the main source of the lack of consistency between intention and behaviour.
In fact, this gap between intention and behaviour could be attributed to differences in cognitions or other unknown factors. Interestingly, evidence suggests that differences in the intention-behaviour relationship between active and inactive intenders were not attributable to differences in cognition (i.e.: attitude, subjective norm, perceived behavioural control and intention) [11, 12]. Thus, other specific factors must be investigated. According to Baron and Kenny , the absence or the inconsistence in the presence of a theorized association between an independent variable (e.g.: intention) and a dependent variable (e.g.: behaviour) may indicate that a third variable affects the direction and/or the strength of this relationship. That is, the intention-behaviour might vary according to different levels of a third variable, known as an effect modifier or a moderator. Therefore, one avenue of research to help understand the gap between intention and behaviour is to investigate moderators of this relationship. Indeed, for a given level of moderator, the intention-behaviour relationship should be higher.
The adoption of physical activity depends not only on intention, but also on a variety of control factors such as abilities, resources and conditions. That is, the performance of physical activity is not fully under complete volitional control. According to the TPB , the construct of perceived behavioural control (PBC) deals with human behaviour not under complete volitional control. As indicated by Ajzen , PBC can be viewed as the combined influence of two components: self-efficacy (ease or difficulty of adopting a behaviour) and controllability (the extent to which the behavioural performance is up to the actor). Within the TPB, perceived behavioural control (PBC) plays different roles in the prediction of behaviour . First, PBC can contribute to the prediction of behaviour along intention (PBC → behaviour). Second, among individuals who express the same level of intention of performing a given behaviour, those with a higher PBC are likely to try harder to perform the target behaviour compared to individuals with lower levels of PBC; this is the intention × PBC interaction hypothesis (intention × PBC → behaviour). However, in order to play these roles, two conditions must be met: 1- the target behaviour must not be completely under volitional control; 2- PBC must reflect accurately "actual" or "real" control over the target behaviour . This latter explanation is congruent with Sheeran, Trafimow and Armitage , who for exercise behaviour observed a significant difference between PBC and a proxy measure of actual control among 73% of their study sample (N = 226). They reported that when PBC was realistic, PBC explained twice as much variance in behaviour compared to unrealistic PBC. In addition, when PBC was realistic, PBC moderated the intention-behaviour relationship, whereas it did not when PBC was unrealistic. Therefore, investigating PBC as a moderator of the intention-behaviour relationship and scrutinizing moderators of the PBC-behaviour relationship is justifiable.
In scientific literature, PBC has received mixed evidence for its moderating effect of the intention-behaviour relationship [16, 17]. About half of the studies reviewed showed a moderating effect of PBC on the intention-behaviour relationship. In each case, higher PBC was associated with better intention-behaviour consistency. In the specific field of exercise/physical activity, four studies have reported a moderating effect of PBC on the intention-behaviour relationship. In the studies by Kimiecik  and Payne, Jones and Harris , this moderating effect was observed among a sample of 332 and 289 employees, respectively, over a period of two weeks, whereas in the study by Terry and O'Leary , this effect was reported among 146 undergraduate students over a period of four weeks. In the fourth study , the moderating effect of PBC among two samples of 105 undergraduate students over a period of two weeks was borderline (p = 0.075). Another study reported a moderating effect for self-efficacy, a variable close to the concept of PBC . In their study, Courneya and McAuley  investigated the moderating role of self-efficacy on the intention-behaviour relationship among 170 undergraduate students. They observed that higher perception of self-efficacy was associated with higher intention-behaviour consistencies at four-week follow-up. Given that Ajzen  considers self-efficacy as one of the sub-dimensions (i.e., perceived difficulty) of PBC, the moderating role of self-efficacy on the intention-behaviour relationship was also assessed.
It was also reported that individuals who both anticipated regret and had the intention to practice physical activity showed better intention-behaviour consistency, compared to intenders who did not anticipate regret among four samples (97 ≤ N ≤ 254) of undergraduate students [23, 24]. Anticipated regret is the perceived feeling of regret if the target behaviour is not performed. Among the theoretical reasons why anticipated regret could be a moderator of the intention-behaviour relationship was its association with greater intention stability [23, 24]. Intention stability has been clearly demonstrated as one of the most important moderators of the intention-behaviour relationship, including exercise/physical activity [24, 25]. Thus, the intention of individuals who anticipate regret for not exercising should better predict behaviour, compared to those not anticipating regret for not exercising.
Moderation effect was reported for past behaviour for both the intention-behaviour  and PBC-behaviour  relationships. However, two opposite findings were reported concerning the intention-behaviour relationship. On the one hand, some researchers observed that the intention-behaviour relationship was stronger when past behaviour was high. For instance, among a sample of undergraduate students (N = 185), Sheeran and Abraham  observed that when the frequency of past physical activity was moderate or high, intention was a significant determinant of behaviour whereas it was not when the level of past behaviour was low. The main reason offered for this moderating effect was that past behaviour was related to intention stability (see also Conner and Godin ). However, it has also been documented that the intention-behaviour relationship was low when the behaviour had been frequently performed in the past (e.g.: Verplanken, Aarts, van Knippenberg and Moonen, ; for travel mode choice). This latter observation is congruent with Triandis' Theory of Interpersonal Behaviour (TIP; Triandis ), suggesting that behaviour falls less under the control of cognition when it is performed frequently. Thus, this controversy in findings justifies performing additional tests on the moderating effect of past behaviour on the intention-behaviour relationship.
Concerning the PBC-behaviour relationship, Ajzen  stated that past behaviour should moderate the PBC-behaviour relationship. The potential mechanism by which past behaviour may act as a moderator relates to actual control. If individuals are familiar with the behaviour to be adopted, which is likely to be the case for individuals with high levels of past behaviour, PBC should be more accurate and, consequently, PBC should adequately reflect actual control and the PBC-behaviour relationship should be stronger. Notani's meta-analysis  provides some empirical support for the PBC × past behaviour interaction in predicting behaviour. Indeed, PBC was a significant determinant of behaviour for samples familiar with the behaviour, whereas it was not a significant determinant of behaviour among samples unfamiliar with the behaviour. In the specific field of exercise/physical activity, Norman, Conner and Bell  reported a better PBC-behaviour relationship among patients attending health clinics (N = 87) who frequently exercised in the past. When the frequency of past physical activity was moderate or high, PBC was a significant determinant of physical activity whereas it was not a significant determinant when the frequency was low.
Age was found to moderate the intention-behaviour relationship. According to the meta-analyses of Hagger, Chatzisarantis and Biddle  and of Downs and Haussenblas , a lower intention-behaviour relationship was observed among younger individuals compared to older age groups. In their meta-analyses, Hagger, Chatzisarantis and Biddle  found that younger samples from 72 studies (adolescent and college, aged under 25, based on mean age of study samples) showed a significantly weaker intention-behaviour relationship when compared to older samples (aged 25 or older, based on mean age study samples). Similarly, among samples of 111 studies, Downs and Hausenblas  reported a weaker intention-behaviour relationship for samples of children/adolescents (aged 8 to 17) compared to samples of young (aged 18 to 25), middle-age (aged 26 to 64) and older (aged 65 or older) adults. They also reported a stronger intention-behaviour relationship for samples of young and older adults compared to middle-age adults. Two explanations have been suggested for the moderating effect of age on the intention-behaviour relationship. First, younger individuals may have unstable intentions and, second, they may have a lack of direct experience with the behaviour compared to older individuals [6, 31]. This argument was also suggested in Natoni's meta-analysis  to explain why the PBC-behaviour relationship for non-student samples (older individuals) was significant, whereas it was non significant for student samples (younger individuals).
The presence of physical activity facilities (built environment) should also receive more attention due to its possible association with a higher level of physical activity . For instance, perceived access to a low-cost recreation facility in a neighbourhood was reported to moderate the intention-behaviour relationship for walking behaviour . A strong perception of closer access to the facility was associated with a more consistent intention-behaviour association. We consider that the perceived built environment should be viewed as a potential barrier to a behaviour and, therefore, the mechanism by which it can moderate the intention-behaviour relationship could be similar to that of the intention × PBC interaction, as suggested by Ajzen . However, given that this latter study (N = 351 adults) was based on a cross-sectional design and used a measure of past behaviour instead of future behaviour, it is advisable to conduct additional tests of its moderating effect.
Other potential moderators of intention-behaviour and PBC-behaviour relationships
Baranowski, Anderson and Carmack  indicated that the effectiveness of physical activity intervention programs based on theoretical models would be increased if study samples were stratified on factors that may moderate model predictiveness. Consequently, new tests for moderation effects of the intention- and PBC-behaviour relationships are needed. In this regard, factors such as annual income and educational level have all been reported to be consistently associated with physical activity (see Trost, Owen, Bauman, Sallis and Brown  for review), although there is a lack of information on their potential moderating effect. It is reasonable to surmise that the intention- and PBC-behaviour relationship will be stronger for individuals with higher annual incomes (higher financial resources and access to equipment) and higher levels of education (greater knowledge). In addition, in the field of physical activity, very few studies have attempted to include biological and genetic variables in a conceptual framework. Because physical activity is a biologically-based behaviour, biological and genetic factors could moderate cognition-behaviour relationships [34, 36]. As such, the relationship between genetic susceptibility to obesity (familial history) and physical activity should be tested. Indeed, parental overweight/obesity increases an individual's risk of obesity in both childhood and adulthood [37–39]. It is suggested that members of a family share genes and familial environment, including lifestyle and behaviour, leading to obesity, poor nutrition and sedentary activities. In turn, obesity, characterized by high body mass index (BMI), is inversely associated with physical activity level [40, 41]. It is likely that genetic susceptibility to obesity assessed by a positive family history of obesity (FHO) may also influence participation in physical activity. Thus, in the context of the present study, the moderating effect of BMI and FHO status was investigated.
In summary, this study aimed at testing psychosocial moderators as well as socio-demographic and biological moderators of the intention-behaviour and PBC-behaviour relationships for leisure-time physical activity.