Inquiry-based science instruction is widely advocated, but studies based on international large-scale assessments often show inquiry to be negatively associated with achievement. We re-examine this issue by examining whether the association between inquiry and learning depends upon the provision of teacher guidance. Participants were 151,721 students from 5089 schools from 10 highest and 10 lowest science performers in PISA 2015. Multi-group confirmatory factor analyses found that measurement invariance cannot be established, suggesting substantial regional variation in the pattern of inquiry-based instruction. Nonetheless, exploratory factor analyses indicated that at the conceptual level, many regions exhibit a pattern which contrasted between 'Guided inquiry' and 'Independent inquiry'. Results of structural equation modelling showed that inquiry is positively associated with outcomes when it incorporates teacher guidance, and negatively when it doesn't. However, the strength of the positive associations is stronger in regions where guided inquiry is measured with fewer items referring to student-centred activities. These findings are in line with current theories regarding the importance of scaffolding in learning from inquiry. This study suggests that it would be misguided to use PISA findings to support arguments to scale back inquiry and other constructivist approaches to teaching science.