Abstract Whilst teaching about the nature of science (NOS) is a significant goal of science education, there remains debate about specifying the NOS and appropriate pedagogies and approaches to researching the NOS. A neglected, but conceptually and practically significant, problem is the proliferation of NOS-related learning goals such as NOS beliefs, views, understandings and knowledge. In this theoretical paper, we argue that such goals are often poorly defined, and different goals cohere with different pedagogical and research strategies. We propose a novel three-part taxonomy of NOS learning goals (as NOS beliefs, knowledge and understandings) and contend that different approaches are appropriate for teaching and assessing NOS beliefs, views and knowledge. An NOS belief refers to a positive attitudinal stance towards some proposition that lacks justificatory support or cannot easily be judged true or false. NOS knowledge indicates justified true beliefs related to the NOS. NOS understanding denotes a grasping of how a collection of NOS knowledge is related. The goals vary by the extent to which they can be judged true or false and the degree of justification they require. For NOS beliefs, a range of stances is acceptable; NOS knowledge must be a true and justified belief; in the case of NOS understanding, teaching and assessment should focus on the appreciation of relationships between justified true beliefs. The novel taxonomy brings needed clarity to a confused aspect of NOS research and may lead to the development of NOS pedagogies and assessment tools more precisely targeted to well-defined learning goals.