Conceptions of learning were investigated in three studies. Study 1 piloted a modified version of Purdie and Hattie's (2002) Conceptions of Learning Inventory (COLI-I) with 236 secondary students. Multilog analysis was used to remove items with poor measurement characteristics. Study 2 used a nationally representative sample of 701 New Zealand secondary school students who completed the COLI-II. The measurement model had marginally acceptable fit. Study 3 consisted of 608 secondary students who completed a slightly revised COLI-III and provided standardised academic reading and mathematics achievement data. The Study 2 and 3 samples were found to have configural invariance and the structural model mapping students' conceptions of learning onto their academic achievement had acceptable fit. The conception that learning is a duty predicted lower achievement and the conception of learning as continuous predicted higher achievement. This is the first study to empirically demonstrate a direct relationship between learning conceptions and academic outcomes in the secondary sector.