MONEY DEMAND FUNCTIONS, as used in macroeconomic theory and in applications to monetary policy, have their theoretical underpinnings in models of portfolio optimization for the individual wealth holder. The use of these functions in an aggregative framework, in the absence of specific premises regarding aggregation, suggests the assumption that all cash holders respond in a uniform way to the same set of variables. Such an assumption is open to question. Separable categories of decision makers can be defined and cash holding behavior may vary across categories. For instance, households, non-financial firms and financial institutions may all determine their demands for cash according to the fundamental principles embodied in the standard portfolio models, but attempts to capture this behavior in money demand functions could properly involve the use of different explanatory variables and structural relations for each sector. Put another way, predictions of the aggregate amount of money demanded based on separate functions for each sector could be more accurate than those based on a single aggregative function. Little attention has been paid to this problem in the extensive literature on the demand for money. The small number of hypotheses that are widely used to generate aggregate money demand functions suggests that it is not difficult to evaluate the empirical importance of the problem. That is, functions considered useful at the aggregate level can be tested at the sectoral level. As well as subjecting these hypotheses to the standard statistical tests on data for a particular sector, their performance can be compared to that of money demand functions that reflect specific sectoral characteristics. The thesis abstracted here reports on such an analysis for the non-financial business sector. Quarterly data on manufacturing corporations in the U.S., from 1947 to 1963, were for the study. A priori suspicions turned out to be justified. Two forms of money demand functions widely accepted at the aggregate level, those based on the permanent income hypothesis and the wealth hypothesis, performed badly on this set of data. The third and simplest of the commonly used hypotheses, that money demand depends on income and the interest rate, fared somewhat better, but not as well as those functions that incorporated additional hypotheses specifically relevant to the firm. The latter hypotheses are based on the argument that the portfolio models normally used to determine optimal cash balances involve an array of assets that is too narrowly defined to represent the firm's behavior. The firm's 'portfolio' involves a complex balance sheet. Thus, the rates of return on assets and liabilities other than the standard marketable ones (e.g., accounts receivable and payable, inventories, bank loans and bonded debt), along with other variables that relate the firm's cash flow to its income flow, are considered in the analysis as well.