Keeble’s assessment in the mid-1970s of the state, and particularly of the empirical usefulness of industrial location theory was entirely correct. Empirical enquiry into the geography of industry continued but in large part unencumbered by serious theoretical reflection or attention to problems of conceptualisation. One of the commonest empirical methods was (and unfortunately still is) the simple confrontation of two maps, one representing cause, the other effect. The effect, the phenomenon to be explained, might be a geographically-differentiated pattern of employment change. The method of establishing cause was to draw up the same map-outline but this time showing the geographical pattern of area characteristics (location factors) which were hypothesised to be influential in producing the distribution of employment growth or decline. The data on the two maps were then correlated and the production of a set of significant coefficients was assumed effectively (after all the usual caveats) to imply cause. The location factors in one map (the characteristics of areas) were deemed to explain the pattern of employment change in the other. More generally, and leaving aside the obvious ‘correlation-does-not-imply-cause’ objections, geographical characteristics were taken to be a significant explanation of geographical distributions.