The purpose of this study is to investigate two issues related to the teaching of thinking skills: (1) transfer across domains; and (2) comparison between individual learning of a thinking strategy and ‘class-like’ setting, which consists of a didactic intervention that takes place in small groups. A microgenetic design was used, in which community college students engaged in investigation of problems (each student participated in 20 sessions). It was found that: (1) the control of variables strategy that was taught in the context of a problem in one domain transferred fully to a new problem in the same domain, hut transferred less well to an isomorphous problem in a different domain; (2) the ‘class-like’ setting contributed to improved performance (as measured by the frequency of valid inferences), with the didactic intervention seeming to contribute to this improvement; and (3) the learning environment described in this study had a larger effect on ‘slower learners’ than on ‘faster learners’. It is proposed to teach for enhanced transfer by focusing explicitly on recognition of underlying logical structures of content-rich contexts.