The Human Research Ethics Committee at the University of Western Sydney provided ethical approval for this cluster-randomized controlled trial. The study was conducted in Independent and Catholic schools in Sydney, New South Wales, Australia during 2011. Ten schools were identified using publicly available listings, and invited via telephone and email through publicly available contact details. The first five schools whose principal provided consent were included in the study. Once consent was also received from department heads and teachers of Year 8 PE classes, the research team met briefly with the students to provide basic information about the study. At this time, information sheets and consent forms were distributed to students. Students who were unable to participate in PE classes were not eligible to participate in the study. All other Year 8 PE students who returned signed parental consent forms and provided personal assent were allowed to participate in the study. Students who did not return consent forms still participated in the lessons alongside their peers, but did not wear accelerometers, or complete questionnaires.
Figure provides a summary of the study design, a cluster-randomized controlled trial with four study arms. To begin, a familiarization session was held during a PE lesson. During this lesson, students wore an accelerometer and teachers wore a microphone attached to their shirt that connected wirelessly to an audio recording device. The purpose of this session was to minimize potential reactivity to these monitoring devices during baseline and post-intervention sessions. Data collected during this familiarization lesson will not be analyzed. Baseline and post-intervention data were collected from all participating students, regardless of study arm, during and immediately after two 20-min segments of PE lessons. In between these two lessons, teachers received a brief intervention (three experimental arms) or were asked to continue with usual practice (control arm).
Sample size and power calculation
Lonsdale et al. [11
] found a large difference in PA levels between free choice and structured PE lesson conditions (Cohen’s d
= 1.07). We adopted a conservative approach and estimated a moderate effect size (medium f
= 0.25) for mixed (between and within-subjects) analysis. Based on this estimation, as well as an estimated correlation of r
= 0.5 among the repeated measures of PA (r
= .53 in a previous study [11
]), we required a sample of 76 students in order to achieve 95% power, with alpha set at .05. Taking into account the clustered nature of the data, this sample size was multiplied by a correction factor of 1+(m
- 1)ρ, called the design effect, where m
is the average cluster size and ρ is the intra-class correlation coefficient [35
]. Assuming an m
of 25 students per class and ρ = 0.1 (estimated based on a previous study [36
]), the correction factor is 3.4. As a result, a sample of 258 students (i.e., 76 × 3.4 = 258) was required. To allow for lack of informed parental consent or assent from some students, and to protect against participant attrition across the four teaching conditions, we attempted to increase the sample by 20% and recruit 308 students from 16 PE classes.
The 16 PE teachers, with their students, were randomly assigned to one of four conditions (three interventions and one control). An independent researcher, blinded to the study hypotheses, performed the allocation after the baseline assessment was administered. Randomization was conducted at the class-level and within schools to control for school characteristics, using a computerized random number generator with blocked randomization scheme (block size = 4). School 1 (Catholic – coeducational), School 2 (Independent – girls only), and School 3 (Independent – coeducational) each provided four classes of students and one class from each school was randomly assigned to one of the four conditions. School 4 and School 5 (two classes each) were both Catholic boys schools. Together, these two schools constituted the final block from which four classes were randomly assigned.
Intervention and fidelity check
Prior to the post-intervention PE lesson, each teacher met with a researcher (CL) for approximately 20 minutes to be trained to deliver the assigned motivational strategy. This meeting took place 20–48 hours before the post-intervention lesson. During this meeting, the investigator asked the teacher to share the lesson plan from the baseline lesson or to describe the baseline lesson activities if the lesson plan was not available. The investigator asked the teacher to devise a lesson for the post-intervention PE lesson that was similar in structure to the baseline lesson. He then guided the teacher through a one-page outline of the proposed teaching strategy. This outline included the strategy name, definitions, the rationale behind the strategy, and guidelines for implementation. The teacher had the opportunity to ask questions, and the researcher and teacher discussed plans for strategy implementation in the upcoming lesson.
The researcher trained the teachers randomized to the “relevance” arm (n = 4) to make statements during the PE lesson that explained the rationale behind the activities in the lesson, and made it clear how the activities were relevant to students’ lives. The researcher trained the teachers randomized to the “providing choice” arm (n =4) to provide students with opportunities to make choices from options that were selected by the teacher. In line with meta-analytic evidence regarding the provision of choice [34
], teachers were asked to provide between 2–4 opportunities for choice during the lesson. When providing students with an opportunity a choice, teachers were also asked to avoid overburdening students with too many options from which to choose (i.e., ≥ 5). The researcher asked the teachers randomized to “free choice” (n = 4) to provide the students with complete free choice; meaning that equipment would be provided, but the teacher would not provide any instruction. It could be argued that this strategy does not include adequate lesson structure and may undermine perceived competence [37
]. Previous research, however, has indicated that complete free choice can increase students’ total PA accumulated during a lesson, relative to standard practice [11
]. Therefore, an investigation of potential effects on needs satisfaction, motivation, moderate-to-vigorous PA and sedentary behavior is warranted. The researcher asked teachers randomized to the control condition (n = 4) to continue with their usual practice.
Teachers in all four study arms had their verbal communication audio recorded in baseline and post-intervention lessons. To test the fidelity of the interventions, two independent researchers assessed teachers’ provision of autonomy support. Specifically, the researchers used four items from the Teacher as Social Context Questionnaire [21
] to rate the extent to which: (1) “The teacher gave the students choices about how they do the tasks in PE”; (2) “The teacher talks about how the students can use the things they learn in PE”; (3) “The teacher listens to the students’ ideas”. and (4) “It seems like the teacher is always telling the students what to do”.
As indicated in Table , compared with the control and “relevance” arms, teachers in the “providing choice” and “free choice” arms were expected to exhibit greater increases in provision of autonomy support from baseline to post-intervention on the first item. Teachers in the “relevance” arm were expected to exhibit larger increases than the other three arms on the second item. The raters hold PhD qualifications in a related discipline (e.g., psychology) and have knowledge of motivational theory applied to physical activity contexts, as evidenced by at least five peer-reviewed publications on the topic. Also, raters were blinded to study hypotheses and teachers’ allocation to the control or experimental arms.
Motivation strategy conditions
Different schools and classes were expected to have PE lessons of varying duration and, due to fatigue, it is possible that MVPA levels at the end of longer lessons could be lower than earlier in the lesson. To standardize across lessons of varying durations, all PA data collection occurred during the first 20 min of teaching time during each lesson for the baseline and post-intervention assessments (i.e., after students had changed their clothes and been fitted with an accelerometer). ActiGraph GT3X accelerometers (ActiGraph; Pensacola, FL) were used to assess PA levels (percentage of time spent in MVPA and percentage of time spent sedentary) of each student in this study. Monitors were synchronized with an external clock and initialized to start recording data in three axes of motion in 1-sec epochs, a minimum of 30 min before and after the scheduled PE lessons. Research assistants helped students place an Actigraph monitor, via adjustable elastic belt, over the right iliac crest, prior to the start of each observed PE lesson, to wear for the duration of lesson. Raw accelerometer counts were uploaded to a lab computer, and saved to a customized Microsoft Excel file. Data outside the recorded start and finish time for given sessions was disregarded. Data were checked for spurious values that did not coincide with the direct observation records; all data between start and finish times for all lessons were included in the analyses. Freedson’s MET prediction equation [38
] was used to determine PA intensity, and 100 counts per minute was used as a criterion to determine sedentary time [39
At the 20-min mark of the baseline and post-intervention lessons, each student completed the four items from the Teacher as Social Context Questionnaire described previously. This version of the questionnaire is designed to measure students’ perceptions of their teachers’ autonomy supportive behavior [21
] (e.g., “The teacher gives us choices about how we do the things in today’s class”). Students also completed measures of autonomy need satisfaction [24
], competence need satisfaction [41
], relatedness need satisfaction [42
], and controlled and autonomous forms of motivation (Situational Motivation Scale) [43
]. Each measure has received empirical support regarding reliability and validity in English-speaking samples of this age group [11
Research assistants, who were senior undergraduate students, conducted all data collection. Research assistants received two hours of data collection training from researchers (CL and RR) who had experience with these data collection techniques. All research assistants were blinded to study hypotheses and cluster allocation (i.e., to intervention or control). Students in each PE class were also blinded to allocation; however, teachers were necessarily un-blinded when they received the intervention (i.e., post-baseline assessment) or learned that they were in the control arm of the trial.
Students’ age, gender, place of birth (Australia or other) and main language spoken in the home (English or other) were collected. A proxy measure of socio-economic status was collected, as students reported their home postcode. This postcode was cross-referenced with census data [45
] to determine the economic decile of each student’s home neighborhood.
Wearing accelerometers and completing questionnaires pose little risk of harm to participants. Teaching strategies were based on theory and research evidence, indicating that they can effectively motivate behavior and no adverse events were reported.
Planned statistical analysis
To assess the reliability of the teacher behavior ratings provided by the independent observers, intra-class correlation coefficients (ICCs) will be calculated. The fidelity of the interventions will be assessed using a MANOVA (with follow-up comparisons). If the manipulation of teacher strategies was successful, independent observer ratings of the intended strategy for each condition should be higher than ratings in other conditions. For example, ratings of choice should be higher in the two choice conditions than in any of the relevance and control conditions. To assess differences in primary and secondary outcome variables, linear mixed model analyses will be run on accelerometer counts, motivation and basic needs satisfaction, with students nested within teachers as a random effect. All analyses will be conducted based on the intention-to-treat principle.