Field experiment of a very brief worksite intervention to improve nutrition among health care workers

The analyses were rerun using per protocol and item imputation, and there were no substantive differences to the pattern of findings. Moreover, there were no significant differences in key variables at baseline between those who remained in the study and those who dropped out, F(5, 73) = 0.30, p = .91, η _p ²  = .02, d = .29.

Participants in the control condition were asked to form plans because it controls for the possibility that being asked to plan exerts effects that are independent of being asked to form an implementation intention (it is not possible to ask people to form implementation intentions without asking them to plan, but it is possible to ask people simply to plan). Importantly, research shows that simply asking people to plan is equivalent to the passive/questionnaire-only control groups that have commonly been used in implementation intention research (Armitage, 2008). In addition to controlling for “generic planning,” the present control group may also be preferable to asking participants to form goal intentions (another control condition employed in implementation intention research) because prompting intention formation is a behavior change technique in its own right (Abraham & Michie, 2008) that brings about a deliberative mindset that works counter to implementation intentions (Brandstätter & Frank, 2002; Heckhausen & Gollwitzer, 1987). Generic planning can therefore be considered an active but neutral control group.

Note that one possible alternative explanation for this pattern of findings is that the nature of the implementation intentions differed between the volitional help sheet condition and the self-generated condition. However, consistent with previous research (Armitage, 2008), differences in the nature of implementation intentions formed within the self-generated condition were larger than the differences in the nature of implementation intentions formed between the self-generated and volitional help sheet conditions. Moreover, in the volitional help sheet condition, implementation intentions are formed by drawing lines; consequently, it is not possible to judge the quality of the implementation intention accurately.

Note that the direct paths between independent variable and dependent variable were statistically significantly reduced. However, because none of the direct paths was reduced to zero, technically, this is best described as partial (as opposed to full) mediation.

Note that it could be argued that the present intervention improves self-monitoring per se, by drawing participants’ attention to critical situations. However, this explanation is ruled out by research that has tested volitional help sheets against control conditions in which participants are asked to identify critical situations. Specifically, participants in these control conditions are presented with the volitional help sheet and are given identical instructions to the experimental condition, but are asked to tick critical situations/appropriate responses that might work for them (as opposed to linking critical situations/appropriate responses in the experimental condition and in the present study). Thus the operation of implementation intentions cannot be ascribed solely to increasing self-monitoring of critical situations per se (see Arden & Armitage, 2012; Armitage, 2008; Armitage et al., in press; Armitage & Arden, 2010, 2012; Armitage et al., 2014).

Thanks are due to the anonymous reviewer who noted this potential limitation.