Many attempts have been made to measure curiosity or intrinsic motivation. However, there is no consensus on what curiosity is. In the present study, five previously constructed curiosity inventories were translated to Finnish and modified to be convenient for fifth-graders. The main purpose of the study was to clarify the concept of curiosity. Using conceptual analysis and the results of previous studies, it was possible to construct nine alternative conceptual models of curiosity. The corresponding statistical models were expected to account for the covariances among 15 subscales measuring curiosity-related exploratory behavior. However, before proceeding to the subscale level, the structures of the inventories were tested at the item level. Confirmatory factor analysis was used to investigate the fit of the alternative models. The four-step logic suggested by Stanley A. Mulaik was used in testing the models. A sample of 529 Finnish fifth-graders from southern Finland was used. This sample was divided according to sex (263 girls and 277 boys). The best-fitting model was a three-factor model with two trait factors and one method factor. The trait factors were termed Curiosity and Sensation Seeking. Several mild moderation effects were found. The theme "What to do when an unrestricted model fails?" was also discussed.