Increasing evidence suggests that implementation intentions are effective in moving people towards achieving health behaviour goals. However, the type of health behaviours for which they work best is unclear. Furthermore, implementation intentions appear to be less effective when studied in clinical rather than student populations. This prospective study tested implementation intentions with a complex, repeated health behaviour in a patient sample. A total of 120 cardiac patients in the UK were asked to increase their daily fruit and vegetable consumption by two portions and to maintain this over 3 months. Participants were randomly assigned to three groups (control, Theory of Planned Behaviour (TPB) questionnaire, TPB questionnaire+implementation intention) and telephoned at 7, 28 and 90 days follow-up to record daily consumption (24-h recall measure); 94 participants completed the study. Daily fruit and vegetable consumption increased from 2.88 portions (SD=1.67) at recruitment to 4.28 portions (SD=2.25) at 90 days. A 4x3 (time by group) mixed design ANCOVA was computed with daily fruit and vegetable consumption at recruitment entered as a covariate. This revealed a significant time effect (F (3, 270)=29.79, p<0.001) (eta2=0.25) but non-significant group (F (2, 90)=0.32, p=0.73) (eta2=0.07) and time by group effects (F (6, 270)=0.48, p=0.82) (eta2=0.01). There was also a significant main effect of the covariate (F (1, 90)=48.51, p<0.001) (eta2=0.35) and a significant time by covariate effect (F (3, 270)=12.14, p<0.001) (eta2=0.12). Substantial increases in fruit and vegetable consumption were achieved particularly by participants who were eating low levels at recruitment. Consumption was not improved by implementation intentions. These findings are discussed in the context of the targeted health behaviour and sample.