In this paper, the evidence collected in the large literature on testing for Granger-causality from money to output is revisited. Using a broad data base of 14 EU-countries plus Canada, the US and Japan, and quarterly data from the mid 60s to the mid 90s, a number of hypotheses from this literature is evaluated. It is found that very few general conclusions can be sustained. For instance, in most countries it is not the case that the use of data in levels creates a bias in favour of finding Granger-causality effects of money on output compared to using differences. Neither does the significance of money lags decline when increasing the number of variables included in the model. What appears to be robust, though, is that allowing for asymmetries clearly increases the likelihood of finding significant causality effects. Based on the Granger-causality test results, a number of country groups are obtained using cluster analysis, which are characterised by a similarly behaviour with respect to the money-output relation.