This paper illustrates the results obtained by adapting the model proposed by the authors in an earlier paper [6] to incorporate the method of estimating durability proposed by Nerlove [3]. It is found that in the cases examined the estimates of durability obtained in this way are very low. Possible reasons for these results are briefly indicated but are not examined empirically. IN ANALYSING THE demand for durable goods it is desirable to allow for the possibility of adaptive behaviour which takes time, a difficulty which is not so obviously present in dealing with the demand for food and similar perishables. In two earlier papers [6, 7], the present authors set out a simple dynamic model of consumers' demand which takes account explicitly of (i) the differences in durability between goods, and (ii) the differences in the speed with which consumers try to reestablish equilibrium in their consumption when circumstances change. This model can be applied to data in the form of time-series and, in its initial form, it appeared that it would usually be necessary to assume a value for the parameter associated with durability. In some cases an extraneous estimate of this parameter can be made without too great difficulty, as, for instance, where there is a well organised second-hand market [2] or where detailed records of consumers' inventories are available [4]. But these cases are exceptional. In general, a reasonable guess has to be made, based perhaps on the current practice of valuers or tax assessors. A way round this difficulty has recently been indicated by Nerlove [3]. He has shown that, subj ect to certain reasonable restrictions, a direct estimate of the parameter associated with durability can be made from the time-series themselves, in conjunction with the other parameters of the demand function. This is likely to be an advantage, partly because it provides an additional basis of estimation and partly because the values obtained in this way should be more consistent with the model than any estimates determined extraneously. Some results from the application of this method to British data are presented in the later sections of this paper. It will be convenient to give first a brief statement of the model and the derivation of the basic regression equation.