Introduction: To engage effectively and efficiently in self-directed learning and knowledge-seeking practices, it is important that physicians construct well-formulated questions; yet, little is known about the quality of good questions and their relationship to self-directed learning or to change in practice behavior. Methods: Personal learning projects (PLPs) submitted to the Canadian Maintenance of Certification program were examined to include underlying characteristics, quality of therapeutic questions (population, intervention, comparator, outcome [PICO] mnemonic), and relationships between stage of change and level of evidence used to resolve questions. Results: We assessed 1989 submissions (from 559 Fellows of the Royal College of Physicians and Surgeons of Canada [RCPSC]). The majority of submissions were by males (69.2%) aged 40-59 (59.4%) with an average of 24.3 (range 6-58, SD 11.1) years since graduation. The most frequent submissions were treatment (36.6%) and diagnosis (22.3%) questions. Half of all questions described [greater than or equal to] 2 components (PICO), and only 3.7% of questions included all 4 components. Cross tabulations indicated only 1 significant trend for the use of narrative reviews and the outcome "integrating new knowledge" (P less than 0.000). Discussion: Self-directed learning skills comprise an important strategy for specialists maintaining or expanding their expertise in patient care, but an important obstacle to answering patient care questions is the ability to formulate good ones. Engagement in most major learning activities is stimulated by management of a single patient: formal accredited group learning events are of limited value in starting episodes of self-directed learning. Low levels of evidence used to address learning projects. Future research should determine how best to improve the quality of questions submitted and whether or not these changes result in increased efficiencies, more appropriate uses of evidence, and increased changes in practice behaviors. (Contains 5 tables.)