A fresh attack on the vital problem of the origin of the lunar craters has been made by analysing the surface distribution of craters of given diameter. The distribution shows a general dumpiness in both the lunarite (bright regions) and the lunabase (dark regions). Craters between 30 and 40 km in diameter, situated in the lunarite, are non-randomly distributed at the 2 per cent level of significance. This argues against the impact theory. It is found that the number-density of differently-sized craters is slightly greater in the following half of the Moon then in the preceding half. This result is shown to apply equally to the lunarite and, taken separately, to the lunabase, and again argues against the theory that the craters were produced exclusively by impact. In assessing the origin of the craters on the basis of the observed frequencies and distribution of craters alone, it is concluded that the ratio of the number of impact craters to the number of endogenic craters is not very large. If only one theory is allowed, it must be that the craters are of internal origin.