The literature on commodity market integration has boomed in the last 15 years, and a sort of consensus is slowly emerging, at least with regard to trends in the last two centuries. This article argues that this consensus is fragile because the research is haunted by serious methodological shortcomings. The results are not really comparable because authors use a bewildering array of statistical techniques, without bothering too much about their assumptions and, more generally, about the theoretical foundations of their work. Market integration is a multi-faceted process and available techniques can be classified according to the issues they are suitable to tackle. In other words, the methodological choices, together with the available data, have steered the research towards a quite narrow set of issues. Thus we know much less than we suppose. The final section sketches out a research agenda beyond pure measurement.