Problem-based learning (PBL) spread from the medical school to other university and K-12 contexts due, in part, to the stated promise that PBL produces the target outcomes of deep content learning, increased problem-solving ability, and increased self-directed learning (Hmelo-Silver, 2004). However, research results have been unclear. This paper examines how the three target outcomes were measured in 33 empirical studies. Results indicate that few studies included (a) theoretical frameworks for the assessed variables/constructs, (b) rationales for how chosen assessments matched the constructs measured, or (c) other information required for readers to assess the validity of authors' interpretations. Implications for future research are discussed.