We show that the performance of the new factor models of Hou, Xue, and Zhang (2015) and Fama and French (2015) depends crucially on how their investment factor is constructed. Both models use growth in total assets to measure investment. Their ability to price the cross-section of returns decreases significantly when the investment factor is constructed using traditional investment measures, or more complete measures that account for investment in intangibles. The superior performance of the asset-growth-based models manifests only when the economy is in a highly overextrapolative state. In below-median overextrapolation, the new models fail to outperform prior models.