A mixed-methods sequential research design has been used to test the effect of introducing teacher science demonstrations to a traditional book-learning sample of 384 Grade 7 boys and girls from five schools in Lahore, Pakistan. In the quasi-experimental quantitative study, the eight classes of comparable ability were designated either Experimental or Control. After pre-testing for attainment and attitudes to science, the Experimental classes were taught the physics topics of Electricity and Light/Colour by means of teacher demonstrations. The Control classes were taught the same topics through textbook exposition. Post-tests on the original measures and residual gain analysis resulted in the girls in the Experimental classes showing significantly greater improvement on both topic tests. Boys in Experimental classes recorded significantly higher scores on the Electricity test only. Attitudes to science show distinct gender-dependent factor structures reflecting national cultural characteristics. Initially, puzzling outcomes as Experimental class students reassessed their attitudes were explained by post-hoc interviews in a qualitative follow-up. Interviews with teachers showed that they were fully aware of the deficiencies of the science presented but attributed this to an overloaded, rigid, test-oriented curriculum. Implications for the education of girls and curriculum improvement are discussed. (Contains 9 tables.)