In this article, we examine current practices in the measurement of syntactic complexity to illustrate the need for more organic and sustainable practices in the measurement of complexity, accuracy, and fluency (CAF) in second language production. Through in-depth review of examples drawn from research on instructed second language acquisition, we identify and discuss challenges to the evidentiary logic that underlies current approaches. We also illuminate critical mismatches between the interpretations that researchers want to make and the complexity measures that they use to make them. Building from the case of complexity, we point to related concerns with impoverished operationalizations of multidimensional CAF constructs and the lack of attention to CAF as a dynamic and interrelated set of constantly changing subsystems. In conclusion, we offer suggestions for addressing these challenges, and we call for much closer articulation between theory and measurement as well as more central roles for multidimensionality and dynamicity in future CAF research.