How we are misinterpreting physical activity intention - behavior relations and what to do about it CQU.pdf (1.09 MB)

How we are misinterpreting physical activity intention - behavior relations and what to do about it

Download (1.09 MB)
journal contribution
posted on 2023-05-16, 02:03 authored by Amanda RebarAmanda Rebar, RE Rhodes, B Gardner
Background: Studies of the physical activity intention-behavior gap, and factors that may moderate the gap (e.g., habit, perceived behavioral control), can inform physical activity promotion efforts. Yet, these studies typically apply linear modeling procedures, and so conclusions rely on linearity and homoscedasticity assumptions, which may not hold. Methods: We modelled and plotted physical activity intention-behavior associations and the moderation effects of habit using simulated data based on (a) normal distributions with no shared variance, (b) correlated parameters with normal distribution, and (c) realistically correlated and non-normally distributed parameters. Results: In the uncorrelated and correlated normal distribution datasets, no violations were unmet, and the moderation effects applied across the entire data range. However, because in the realistic dataset, few people who engaged in physical activity behavior had low intention scores, the intention-behavior association was non-linear, resulting in inflated linear moderation estimations of habit. This finding was replicated when tested with intention-behavior moderation of perceived behavioral control. Conclusions: Comparisons of the three scenarios illustrated how an identical correlation coefficient may mask different types of intention-behavior association and moderation effects. These findings highlight the risk of misinterpreting tests of the intention-behavior gap and its moderators for physical activity due to unfounded statistical assumptions. The previously well-documented moderating effects of habit, whereby the impact of intention on behavior weakens as habit strength increases, may be based on statistical byproducts of unmet model assumptions. © 2019 The Author(s).


Category 1 - Australian Competitive Grants (this includes ARC, NHMRC)




Start Page


End Page


Number of Pages





BioMed Central, UK

Additional Rights

CC BY 4.0

Peer Reviewed

  • Yes

Open Access

  • Yes

External Author Affiliations

University of Victoria, Canada; King’s College London, UK

Author Research Institute

  • Appleton Institute

Era Eligible

  • Yes


International Journal of Behavioral Nutrition and Physical Activity