The present paper deals with the analysis of the U. S. consumptionstructure in the framework of the neoclassical consumer preferencetheory.. We have been analyzing, with a complete system of demand equations,the Japanese consumer demand based on the Family Income andExpenditure Survey and the National Income Statistics (Tsujimura (1964),Tsujimura-Sato (1964), Tsujimura (1968), Maki (1978a) , Maki (1978b) ,and Maki (1979) ).Under a series of research, the following observations have beenmade: (a) The household is the empirical counterpart for the behavioralunit of the theory. (b) Changes in family size in the householdinfluence the pattern of consumer expenditure. (c) The habit formationhypothesis is appropriate to describe both cross section and timeseries data consistently. (d) The linear expenditure system (LES),derived from the Klein-Rubin preference function wl.th family size andhabit potentials, fairly well explains observational data of householdconsumption expenditures, classified into five, twenty five and sixtyitems. (e) For the data classified into sixty items, demand functionsof the LES type, including family size and habit potentials, does notalways uniformly account for observational data of household consumptionexpenditures, and the tendency for the goodness of fit to be unsatisfactoryappears systematically for durable goods such as automobiles andmany kinds of electric appliances. (f) On the other hand, classifyinghousehold consumption into groups of agricultural products, manufacturedproducts and twenty-three service items, we get better estimatesof LES demand equations. (e) and (f) above imply that for perishablegoods and a broader classification of items our model is more suitableto the data and that our model has to be modified to explicitly includethe stock adjustment effect as the next step.These are summaries of what we learned in our past studies. Thepresent analysis, using U. S. data, will test the validity of the LEStype preference function with family size and habit potentials, whichhas proven its applicability to Japanese data. It may be best for thegenerality and validity of a model to analyze the two sets of data bydescribing the different stages of economic growth through the sameexperimental design.