I would like to make a few comments in reply to Robert Sack's response to my review of his book (Sack, 1980; Sayer, 1982a). These kinds of interchange can easily become rather private and exclusive affairs, not to say unduly sensitive, and they can quickly lose contact with geographical questions. In order to avoid this I shall therefore follow up the more philosophical points with some suggestions as to their wider significance for geography. First a concession: in retrospect I realize my title was unfair for it was not so much Sack's conception of space that I questioned--on the contrary, I commended it--but his conceptions of social thought. Secondly, some points of agreement: (i) of course no philosophy is complete and consistent; (ii) of course not all the doctrines of 'realism', however defined, are unique to it (p. 504); (iii) of course 'not everything is a matter of philosophy' (p. 509); (iv) of course we should use philosophy to focus on substantive matters such as geographical questions (p. 504); (v) of course 'science' is not merely 'subjective' (p. 506). Thirdly, the philosophical ideas to which I am committed are clear from the arguments and references in my review but if these are not sufficient, I have discussed them elsewhere (e.g. Sayer, 1981, 1982). Fourthly, I think some difficulties may have arisen from the ambiguity of the word 'generalization'. On the one hand there are statements about classes of objects which attempt to convey their common properties or what is believed to be true of them 'on the whole', as an approximation. Such statements (hereafter 'generalizations') include implicit or explicit assessments of quantities: e.g. 'most free-standing towns have grown'; '80 per cent of the population is urban'; 'people usually overestimate distances to unfamiliar places'. On the other hand, there are statements which have a wider and vaguer sense of 'hypotheses', which need not include references to quantities and can be about individual objects rather than classes of them (hereafter generalization2). I use the term only in the first sense, and would argue that distinctively realist theoretical claims concern necessity rather than generality or regularity, and hence are not generalizations'. To take an example from my review: 'To exist as a nation state, certain social relations must be established on the basis of some degree of understanding (not necessarily a clear or very correct one) of the meaning of its concept'. If this were to be recast as a generalization' it would read: 'In all (or most) nation states, certain social relations are established on the basis of some degree of understanding. . .' etc.). What was, in the realist version, a claim about a necessary relation has become diluted, as a generalization', to a statement which could equally refer to a state of affairs which is neither necessary nor impossible, even if it happens to have occurred in all cases. In the former, the claim is much stronger--as Popper would say, it 'sticks its neck out more' and consequently its falsification would have to be taken seriously. In the case of the generalization', falsification would be of little significance as the statement did not say things had to be that way. (This is not to argue, of course, that one should treat contingent matters as necessary merely to derive