Objectives.Walking is poorly represented in memory, making it difficult to measure using self-report and even harder to predict. To circumvent this, we used the affective priming paradigm (Fazio, Sanbonmatsu, Powell, & Kardes, 1986) to assess implicit attitudes towards walking. METHODS: Royal Air Force trainee aircraftsmen (N=188) wore pedometers for 1 week prior to completing the affective priming paradigm, questionnaire and interview. The affective priming paradigm involved a computer-based response latency task containing physical activity words as primes followed by adjectives as targets to be evaluated. Targets were drawn from two bipolar dichotomies, good-bad (the original Fazio et al. items) and happy-sad (mood). RESULTS: Priming for mood items was related to levels of physical activity with high frequency participants priming for the positive (happy) pole and low frequency participants priming for the negative (sad). Both groups primed for the negative element of the Fazio (good-bad) dichotomy. Regarding walking and running, there was no differentiation on the basis of participation level. Instead, facilitated responses to happy targets contrasted with inhibited responses to sad targets for both types of locomotion. There was weak evidence that intentions to run were associated with priming of positive target items, irrespective of category. CONCLUSIONS: The relationship between implicit attitudes and behaviour is complex. Whereas implicit attitudes were related to overall exercise participation, they were not related to the specific activity of walking, despite the behaviour being mainly under automatic control.