BACKGROUND: The area of self-regulated learning integrates the fields of metacognition and self-regulation and assumes that the student is an active processor of information capable of self-regulating his learning by putting together the cognitive, metacognitive, and motivational components. The Motivated Strategies for Learning Questionnaire (MSLQ) is a benchmark for the measurement of self-regulated learning. However, the field of study does not show adequate evidence of its structural validity. The vast majority of studies involving this question present serious methodological mistakes, compromising the evidence of validity. OBJECTIVE: Our study investigates the structural validity of MSLQ including all 15 scales and corrects relevant mistakes in the previous studies. METHOD: We tested different models through item confirmatory factor analysis in a convenience sample of 670 college students (M = 22.8 years, SD = 5.2) from a public Brazilian university in the technological area. The models with the ML, MLR, MLM and WLMSV estimators. RESULTS: Only WLSMV produced models with acceptable fit. The final model has a bi-factor structure with a general factor (self-regulated learning), 15 components as first-order factors, and four broad components as second-order factors. Twelve first-order components, all second-order components and the general factor had acceptable reliability. The components' elaboration, intrinsic goal orientation and metacognitive self-regulation, did not show acceptable reliability, in terms of McDonald's omega. CONCLUSION: Considering the worldwide importance of the MSLQ, we do not recommend the use of the measurement of these components for clinical practice and psychoeducational diagnosis until new studies show that this low reliability only occurs in our sample. Our study shows new evidence, correcting many previous methodological mistakes and producing initial evidence favorable to the factor structure of the MSLQ.