Classroom discussion, despite its association with good academic outcomes, is exceedingly rare in U.S. schools. The Word Generation intervention involves the provision of texts and activities to be implemented across content area class, organized around engaging and discussable dilemmas. The program was evaluated with 1,554 middle grade students in 28 schools randomly assigned to treatment or control conditions. There were large effects on classroom discussion quality across all content areas, especially in math and science (Cohen's d =?0.38-1.13). The program also produced significant, though small, effects on taught vocabulary (effect size?=?0.25, p <?0.01) but no effects on a standardized assessment of general vocabulary. Quality of classroom discussion mediated 14% of the treatment effect on vocabulary outcomes.