In this paper we investigate the link between the language in which pupils take the Programme for International Student Assessment (PISA) test and the scores they achieve in this assessment, focusing upon the case of Wales. Using five rounds of PISA data and an instrumental variable approach, we show how pupils who took the test in Welsh score around 0.3 standard deviations (30 PISA test points) lower in reading, mathematics and science than their peers who took the test in English. This finding is robust to different model specifications and statistical approaches. We argue that this may indicate that the academic performance of teenagers in Wales may be underestimated in PISA—particularly amongst those who take this test in Welsh.