We report on a study of a laboratory teaching apprenticeship program designed to improve graduate teaching assistant (GTA) performance. To catalyze GTAs as laboratory teachers we constructed learning goals, synthesized previous literature into a design model and a developmental path, and built two instruments to measure 12 strategic pedagogical interactions. The resulting model is a cognitive apprenticeship. The instructors encourage GTAs to teach chemical knowledge by guiding their undergraduates' reasoning--in addition to explicit transmission of chemical knowledge. The new graduate course was refined over 5 fall semesters and 83 graduate students. Qualitative audio-video supplemented statistical data. Valid and reliable data, collected from coding 12 strategic interactions while GTAs taught a lab, were used to judge effectiveness of teaching in terms of pedagogical chemical knowledge (PChK). Three types of laboratory PChK are defined and described. The results show that 45% of new GTAs developed higher forms of PChK. Exemplary GTAs found that the extended time in lab is an advantageous occasion to teach UGs how to reason with chemical concepts; their actions utilized a constructivist learning model focused on directing students to reason, thereby increasing students' abilities to reason. Some 8% of new GTAs developed only the lowest form of PChK, procedural teaching.