Instruments measuring understanding of the nature of science (NOS) are required if educational institutions intend to use benchmarks or examine the effects of interventions targeting students' NOS development. Compared to other constructs, NOS understanding is complex, having been the subject of debate among scholars in both its substance and its dimensionality. This complexity invites challenges in defining what is to be measured. Drawing from the perspective that policy reform documents provide pragmatic consensus-based definitions of NOS, this study investigated how well the dimensionality described in the NOS component of the Next Generation Science Standards (NGSS) framework matched the empirical structure of data collected from a set of secondary-school students' responses to an NOS instrument comprising multiple-choice and Likert-scale items. Using multidimensional item response modeling to compare structures of NOS dimensionality, we found that treating NOS as comprising multiple dimensions-as defined by the themes in the NGSS NOS framework-resulted in a better fitting model than when treating NOS as a single dimension. The multidimensional model also had fewer poorly functioning items and revealed NOS profiles that otherwise would have been masked in a model treating NOS as a single dimension. These results provide support for the NOS NGSS framework and contribute to the ongoing discussion about the dimensionality of NOS.