Although one-to-one tutoring has been regarded as the most effective method of teaching, surprisingly little is understood about tutoring expertise. Much educational research focuses on classroom teaching, while the few studies that focus on one-to-one tutoring do not offer a precise information-processing account of this skill. This Note describes the authors' initial attempts to study one-to-one tutoring. The goal of the research is to construct a detailed cognitive model of the reasoning and knowledge of an expert human tutor. The method the authors have employed is a variant of knowledge engineering. They videotaped tutoring sessions with expert teachers, subjecting them to a detailed analysis whose aim was to abstract the tutors' knowledge structures. The Note describes some important tutoring techniques they have isolated using these methods. The authors discuss several dimensions along which tutors appear to be intelligent planners and problem solvers. Finally, they note several implications of the research, including its potential impact on the construction of intelligent computer-based tutoring systems.