Although the trials were conducted in three different countries, with their inherent surface-level cultural disparities, varied in length and intensity, and entailed common and unique intervention components, they all provided substantial support for the hypothesized SDT motivational process model. Indeed, they were the first RCTs to test some or all of the sequential steps put forth by SDT in the PA promotion context (see Figure 1). These findings are aligned with the arguments of Deci and Ryan and existing evidence [73, 74] suggesting that the psychological needs are universal and that the assumed social environmental predictors of more autonomous motivation and optimal functioning apply across nations.
Further, all three trials showed intervention effects (albeit some more comprehensively and robustly than others). In the PAC trial, significant between arm differences in quantity of motivation, quality (autonomous) motivation, and reported PA were exhibited at 6 weeks (mid intervention) as well as 13 weeks (end intervention) and quantity motivation effects were sustained in the post intervention phase. Moreover, intensive intervention arm participants showed greater decreases in body composition from 13 weeks to 25 weeks. In Empower, participants in the intervention arm exhibited lower anxiety scores and also rated their overall health to be significantly more positive at follow-up (6 months). Finally, with respect to the 1-year long PESO intervention, significant differences emerged between arms in perceived need support, perceived competence, autonomous motivation, locus of causality and enjoyment as well as PA at one year and follow-up (two years).
In terms of intervention content, this trio of interventions revealed many similarities with respect to their adoption of SDT principles to create autonomy-supportive contexts. For instance, as it is evident from Table 3, there are notable parallels between the ASSIST phase (5th A) of the PAC trial, the individualized discussions within the Empower trial, and the group session of the PESO trial, particularly with respect to setting PA-related goals, aligning attempts to become more active with life goals, problem solving, and encouraging enjoyment of physical activity.
It is possible that some differences in the three programs might have led to differences in outcomes across the three trials. For instance, it is possible that contact with the HFAs in the Empower trial might have been too infrequent or short, or that group sessions in the PESO trial might have fostered more group support and thus met participants' relatedness needs to a higher extent than in the other trials.
Collectively then, the three particular trials indicate that to the extent that interventions influence psychological needs (through need-supporting environments), more autonomous motivation ensues, which in turn predicts positive PA and/or psychological outcomes. Moreover, these trials suggest that even when interventions are not entirely successful in affecting theory-based constructs (e.g., perceived autonomy support, perceived competence, autonomous self-regulation) versus a control or comparison condition, the same constructs can be found to predict positive behavioral and psychological outcomes in both conditions. Although the level of supportive evidence for SDT is lower in such cases, the theoretically-assumed model can be found to be operative even in the absence of a statistically significant "SDT-based intervention" effect. This is an important aspect bearing in mind the unpredictable and at times inherently need-supportive environments that can be found in many real world scenarios (e.g., fitness counseling, doctors' offices).
In addition, when considering the findings emanating from the three trials, it should be acknowledged that need satisfaction does not automatically ensue from supporting contexts. Rather, need satisfaction is assumed to result through a dialectic relationship between social contexts and individual characteristics . Such findings also speak to the fact that comparative effects between contexts may be weakened by difficulties in creating non-need-supportive control conditions, as advisors/counselors most likely inherently have their participants'/patients' best interest in mind (i.e. support patients' needs). The findings from the process model tested in the Empower trial are consistent with this premise . It would be ill-advised to purposefully create non-need supportive environments solely for the purposes of enhancing internal validity within the RCT design. From this, it is suggested that future research in this area would do well to use standard care as the control group condition, as in many medical-health RCTs, although this might lead to smaller intervention effects due to naturally occurring need support in these contexts. Moreover, future research could determine which techniques need to be delivered in these settings (above naturally occurring need support) to boost need satisfaction.
It is fair to say that the three trials reviewed have their share of limitations, to which we can add certain inevitable challenges that PA researchers face when conducting RCTs in the field. Notably, measurement tools and their implications for related analyses is one of the areas presenting challenging issues. One such concern pertains to the use of instruments assessing exercise motivation in sedentary individuals (e.g., at baseline), with stems like "I exercise because...", such as the Exercise Self-Regulation Questionnaire. These assessments may be troublesome as many individuals have been sedentary most of their lives and may lack the knowledge to formulate adequate answers. To address this limitation, researchers in the PESO trial, for instance, opted to use unadjusted measures (at 4 and 12 months) for some psychosocial exercise variables, and not pre-post change scores, to examine intervention effects. The advantage of having an adequate (standard care) control group was evident in this regard. On the same note, Teixeira and colleagues  observed an interesting phenomenon with exercise self-efficacy, which significantly decreased in the control group during the 1-year program but did not change in the intervention group (with significant time-group effects) . That is, it appears that self-efficacy, and perhaps other related SDT-based constructs, could be artificially inflated at baseline in all subjects, again possibly from a lack of experiential knowledge regarding PA or from an initial social desirability effect. This was also found in the PAC trial with autonomous motivation . This possibility warrants caution in interpreting pre-post change scores of these mediating variables.
Related to the above point, the PAC and Empower trials, among others, may have been affected by ceiling effects on motivational and social environmental variables. That is, high baseline/pre-intervention scores on SDT constructs may have prevented the finding of significant intervention effects on these variables. As mentioned above, in terms of assessing the degree of autonomy support provided by the HFA in the Empower trial , participants' scores on the HCCQ  were generally very high and not significantly different across the SDT intervention and standard provision arms. This finding also speaks to the point mentioned above with respect to limitations in achieving non-need supportive control groups.
Although the HCCQ was developed to assess the degree to which an advisor/behavioral consultant is autonomy supportive , this measure actually taps overall need satisfaction or environmental support. In this regard, it is not unreasonable to suspect that counselors or consultants who are involved in promoting behavioral change would naturally, without purposeful training, vary in the degree to which they are supportive of each of the needs that the HCCQ would appear to tap. Particularly after being exposed to a physical activity advisor (standard care or intervention) who is likely to be quite engaging, interested in the client/patient, enthusiastic about PA, and providing information on PA engagement, it is reasonable to expect that clients/patients might be prone to providing very positive overall ratings on the HCCQ. Such measurement issues point to the need to potentially develop more sensitive objective measures of the content and social environment manifested in PA promotion consultations . Moreover, such issues might warrant that the fidelity of delivery of SDT-based interventions (and control conditions) be independently assessed in future studies.
In the psychology realm, there has been concern over "therapists drift", whereby clinicians, over time, drift away from firm adherence to their intervention training, implementing a previous or other approach . This type of phenomenon may be equally relevant for those intervening in PA promotion and thus it would be important that interventionists' fidelity to SDT training protocol be regularly monitored and reported in the manner of some MI-based intervention studies (for example ). In the PAC trial, implementation was assessed in multiple manners including recording of sessions with the physical activity counselor and assessing for compliance to the IPAC protocol and to SDT . In the PESO trial, although more could have been done to more strongly ensure protocol implementation, this was maximized by holding regular meetings to discuss fidelity issues and by the use of manipulation checks conducted by a senior interventionist during randomly assigned sessions [7, 8]. In the Empower trial, the fidelity of intervention delivery was examined to some degree via observations of a sampling of one-on-one consultations within the SDT-based arm and meetings between members of the research team and the intervention arm HFAs [Rouse, Ntoumanis, Duda: The development of an observation assessment tool examining environmental support within physical activity consultation, submitted]. We suggest that future research in this area systematically include an implementation evaluation in addition to an outcome assessment. Indeed, Marcus et al.  argued that assessment of the fidelity of intervention delivery is one of the methodological issues that needs to be better addressed in future PA promotion RCTs.
In addition to matters of measurement, methodological concerns of a logistical nature also arise in conducting SDT-based PA promotion RCTs. One of these challenges pertains to the characteristics of intervention participants themselves. Specifically, the informed and voluntary participatory nature of the PAC, Empower, and PESO trials may have led to an overwhelming majority of participants with elevated initial levels of motivation, which could be largely autonomous, to make changes to their PA. It might be the case that SDT-based PA interventions differentially affect participants who voluntarily consent to being involved in a trial, namely in the sense that they are already more autonomous than non-participants or are becoming so. Moreover, they already have exercised their autonomy in agreeing to participate in the study in the first place. This complex issue may have compounded previously mentioned ceiling effects and potentially led to a weakening of the influence of these interventions. Although difficult to address, it is a limitation that future researchers should consider.
Another drawback relates to the specificity of participant samples in the aforementioned trials, which make it difficult to generalize the findings across populations. Indeed the typical participants in the three SDT-grounded trials just described were middle-aged overweight or obese women. Future studies should be conducted to evaluate intervention effects on groups that are more diverse in terms of their demographic (e.g., age, gender), motivational, and physical characteristics (e.g., body weight). Perhaps an international RCT promoting PA according to SDT propositions is in order. The three trials conducted to date would suggest that the health care context might be ideal for optimal reach and diversity of samples.
With respect to intervention procedures and to echo our point regarding delivery fidelity, SDT-based PA intervention studies thus far reflect a broader need to better report specific intervention techniques and their theoretical underpinnings . Doing so would not only strengthen researchers' ability to make statements regarding an intervention's theoretical grounding but would also reinforce the external validity of these interventions and allow other researchers to draw on the strategies employed and replicate findings. Importantly, experts are now reflecting on the extent to which strategies used in MI  can be adopted in the future as the standard for what a typical SDT-based consultation centered intervention should look like, highlighting areas of agreement and areas where some discrepancies can be found (see Vanteenskiste, Williams, & Reniscow  this issue).
A related issue is the dearth of SDT-based studies that test intervention techniques and strategies independently to see how well each fairs in fostering the psychological needs and subsequently altering the quantity and/or quality of motivation and levels of PA (e.g, is one technique superior to another?). Undoubtedly, testing the combination of SDT-inspired techniques in the manner of the presented trials has been informative thus far. Yet, it is also by examining the unique effects of distinct techniques (such as the values interview) on mediators of PA changes that we can eliminate redundant components and optimize the most functional strategies. This should lead to more influential and more parsimonious interventions, which may be more cost-effective. Thus we suggest that researchers consider drawing from the "dismantling" approach and assign different groups of participants to receive either different, exclusive aspects of an intervention or combined components . Such dismantling would help to pinpoint the most active ingredients in successful and less successful SDT interventions. Another approach would be to use sophisticated single-subject designs to sequentially test the different techniques in different orders and potentially in different PA promotion settings.
Current research in this area is also limited by failing to conduct economic evaluations of RCTs centered on PA promotion. In order to ascertain if such SDT-based interventions can be optimally implemented on a broader societal level, studies will need to determine whether these interventions (e.g., creating autonomy supportive contexts for behavioral change) are, or can be made, cost-effective (see Angus et al.,  for an example of a cost-effectiveness evaluation from the PAC trial). Additionally, it should be noted that two of the trials presented involved relatively short-term intervention periods. For example, although the intervention in the Empower trial lasted 3 months and the intention was to have more theoretically informed contacts between HFA and the patients , logistical constraints meant that, in reality, there was an initial one hour consultation with many, but not all, of the patients also experiencing an end of the scheme 30 minute exit consultation. In PAC, intensive counseling patients received 6 sessions for an average of 206 minutes of counseling while in PESO participants received 30 sessions lasting 120 minutes each (approx. 3,600 minutes in total), albeit in relatively large groups of 25-30 women . Given the intricacies of altering a complex behavior like PA  as well as the well-documented difficulties in maintaining initial success, future studies will need to consider the effectiveness, cost effectiveness, and feasibility of longer as well as more intense interventions. With more extended intervention programs, participants could be monitored regularly for lengthier time spans (e.g., 2 years or more) and, in some countries, over the course of different seasons. In such future work, it would be pertinent to consider the findings of Hillsdon et al.'s review  regarding the recommended frequency of intervention occasions. Interventionists should also take into account recent research on PA promotion consultations that point to the importance of including a post-program "top up" or follow up to facilitate participants' maintaining levels of PA engagement . Studies that combine strategies focused on behavioral adoption as well as maintenance will provide further insight on the utility of SDT-based interventions in facilitating long-term changes in PA.
Any organized intervention effort toward behavior change in real life settings will hold its share of planning difficulties; attempts towards PA promotion are no exception. Such is the case when we consider the limited time that is available to train the interventionists who are charged with facilitating PA changes in a need-supportive manner. In the real-world, researchers, particularly those in primary care, have to deal with time constraints regarding the training of health care/PA promotion advisers. These advisors typically have limited backgrounds in psychology and models of behavioral change and also face have serious time constraints. In these cases, intervention effectiveness, which generally requires that behavior change programs be implemented in close to ideal conditions, is often sacrificed in hopes of greater efficiency and/or real-world applicability. Ameliorating these challenges may require experts to produce a 'roadmap' of more essential SDT strategies and concepts that should be conveyed in the training stages to those who will deliver these interventions (see , this issue). An alternative would be to teach autonomy supportive counseling in standard University courses, such as those for future Physical Activity Counselors, Registered Dieticians, or Medical Doctors.
Moreover, intervention researchers may wish to examine the threshold level of autonomous motivation that will optimize PA adoption and maintenance. Additionally, similar to what was done in the PAC trial , incorporating qualitative methodologies that draw on participants' experiences throughout the interventions will provide researchers with insight into how and when to optimize autonomous motivation. Finally, as alluded to in the opening section of this paper, progress in terms of the proper testing of mediation effects will need to be made, in order to obtain further support for SDT-assumed motivation sequences in developing and validating PA interventions. The PESO and PAC trials described herein represent important efforts in this direction. In closing, the authors wish to acknowledge the absence of any secondary analyses of trial findings presented in this review. As the number of SDT-based RCTs in the PA domain proliferates in coming years, experts are encouraged to generate a quantitative synthesis of the findings to guide future intervention efforts.