Previous analyses of depreciation have resulted in conflicting claims and evidence concerning the depreciation patterns of durable assets. We demonstrate that heterogeneity in asset quality and usage rates can reconcile disparate results. Specifically, we add both usage and care to the flexible estimation model of Hulten and Wykoff (1981a). By accounting for systematic changes in these additional dimensions and controlling for sample bias and scrappage value, a near-geometric pattern emerges which can otherwise be obscured. The impact of usage (as distinct from age) has implications for the measurement of cost (and hence productivity) and for modeling replacement decisions. Studies measuring the total capital stock and its implied service flows are numerous in the economic literature (Jorgenson, 1971). An accurate measurement of the aggregate capital stock is often essential in analyses of investment. consumption. and productivity at the firm, regional and national levels. But measurement of the stock is difficult, owing to heterogeneity of both the capital stock and how it is used. Also critical is measuring how the capital depreciates as it ages. Despite its importance and the amount of research conducted on depreciation, no clear consensus exists about the depreciation patterns followed by different types of capital goods. Jorgenson and his associates have advocated the use of a geometric depreciation pattern in studies of investment behavior. In addition, several analyses of market data for used capital assets appear to support the geometric pattern, which implies a constant depreciation rate (see Hulten and Wykoff (1981a, b); Wykoff (1989); Hall (1971)). A number of articles have been written challenging the constant rate assumption as both implausible and misleading. Feldstein and Rothschild (1974), for example, point to the unrealistic requirements for asset decay and replacement that are necessary for a geometric pattern. Penson, Hughes, and Nelson (1977) suggest that the rapid decline in productive capacity immediately after purchase is inconsistent with the physical capacity of individual assets. These authors and others (e.g., Taubman and Rasche, 1969) have Received for publication August 2, 1988. Revision accepted for publication August 23, 1989. * Oregon State University and Laurits R. Christensen Associates, respectively. Special thanks go to Moshe Kim, Frank Wykoff, Wes Musser, Michael Kuehlwein and two anonymous reviewers for helpful comments on earlier versions of this paper and to Bette Bamford for typing assistance. Oregon Agricultural Experiment Station Technical Paper No. 9009.