Interaction Between Automatic And Reflective Processes In The Prediction Of Accelerometer-derived PA Behaviors

Abstract

Grounded in dual process theories related to physical activity (PA) behavior, both reflective and automatic processes have been shown to predict PA. However, relatively little is known about the interaction or relative contributions of these two processing systems in relation to PA behavior, and even less is known about how individual factors alter these relationships. PURPOSE: To examine the relationships among automatic and reflective processes (i.e., PA intentions) across four different PA domains (i.e., exercise [EX], light [LPA] and moderate-to-vigorous intensity non-leisure time PA [MVPA], and sedentary behavior [SED]), and explore how individual factors modulate these associations. METHODS: Participants (≥18 years old) completed a computerized Single Category Implicit Association Test (SCIAT) measuring implicit attitudes towards PA and SED behaviors using images depicting the four different PA domains. Positive SCIAT scores indicated positive affective evaluation (AE) toward the target behavior. PA intention and PA self-efficacy (SES) were measured using questionnaires. Habitual PA was measured using an accelerometer. Multiple regression analysis was used to examine whether AE predicted unique variance in habitual PA levels after controlling for PA intentions and SES. RESULTS: Sixty-six participants (63% women; 92% students) were included in our final sample (M±SD: 23±8 y, 26.2±19.1 kg/m2, 53.4±13.0 min/d and 505.3±315.1 min/d of MVPA and SED, respectively; 80% met current PA recommendations). AE of the different PA domains did not differ: EX=0.00±0.67, LPA=0.00±0.89, MVPA=-0.08±0.46, SED=-0.08±0.72. AE of EX was strongly and positively correlated with LPA (r=0.53, p<0.001). AE of EX and SES interacted to predict min/d of MVPA (β=-0.21; p=0.04) and total activity counts per day (a measure of total PA) (β=-0.26; p=0.01). Simple slope analysis revealed that at high SES, AE of EX was negatively correlated with both MVPA (β=-0.29, p=0.05) and total PA (β=-0.31, p=0.04). CONCLUSIONS: Our results, although somewhat unexpected, support the role of AE in relation to PA behavior and suggest that its effect may depend on self-regulatory resources such as SES. Caution is warranted in generalizing our findings to other samples as ours was comprised of young, healthy, and active adults.