This paper uses a novel survey instrument to identify distinct components of nutrition knowledge, and test for links between knowledge and dietary choices in Southern Malawi. Our first aim is to distinguish respondents’ familiarity with recommended behaviors, such as when to start breastfeeding or introduce solid foods, from respondents’ factual knowledge about mechanisms, such as whether biscuits or papaya and orange fruit or orange Fanta contribute more to future health. We find knowledge of nutrition behaviors to be strongly associated with more schooling, older age and being female, while knowledge of mechanisms is associated only with training and employment as a health professional. We then test whether this expanded definition of nutrition knowledge is associated with dietary intake when controlling for other factors, and find only suggestive evidence for significant links interacting with age of respondents. These findings point to the need for knowledge surveys and public health behavior-change campaigns to address the kinds of information that might have the most influence on actual behavior, potentially including the mechanisms involved in food composition, food safety and disease transmission.