We use a field experiment to investigate the effect of incentives on food purchase decisions at a grocery store. We recruit over 200 participants and track their purchases for a period of 6 months, permitting us a glimpse of more than 3,500 individual shopping trips. We randomize participants to one of several treatments, in which we incentivize fresh fruit and vegetable purchases, provide tips for fruit and vegetable preparation, or both. We report several key insights. First, our informational content treatment has little effect. Second, we find an important price effect: modest pecuniary incentives more than double the proportion of dollars spent on produce in the grocery store. Third, we find an interesting pattern of consumption after the experiment ends: even when incentives are removed, the treatment group has higher fruit and vegetable purchases compared to the control group. These long-term results are in stark contrast to either a standard price model or a behavioral model of 'crowd out.' Rather, our results are consonant with a habit formation model. This opens up the distinct possibility that short term incentives can be used as a key instrument to combat obesity.