Research on Teacher Effects: Critique of a Traditional Paradigm

S ummary . Process‐product studies of the effects of teacher behaviours upon pupil outcomes often employ class mean scores as measures of the outcome variable. Assuming that all the measures used are reliable, a zero correlation between the teacher measure and the pupil outcome implies that the teacher behaviour being studied does not affect pupils in any lawful way. However, such a conclusion may be faulty. This article argues that research on teacher effects should consider pupil variables; teacher behaviours and pupil outcomes may enter into lawful relationships with one another, the nature of the relationship being different for different kinds of pupils. Data from a study of pupils' attitudes to physics are used to support this position.