Case study evidence suggests that exporting firms learn from their clients. But econometric evidence, mostly using exporting and TFP growth, is mixed. We use a UK panel data set with firm-level information on exporting and productivity. Our innovation is that we also have direct data on the sources of learning (in this case about new technologies). Controlling for fixed effects we have two main findings. First, we find firms who exported in the past are more likely to then report that they learnt from buyers (relative to learning from other sources). Second, firms who had learned from buyers (more than they learnt from other sources) in the past are more likely to then have productivity growth. This suggests some support for the learning-by-exporting hypothesis.