The A.C. Nielson household scanner panel is used to analyze factors affecting brand-size choices for spaghetti. These data link product purchases, which are scanned, with household demographics and market information on the store where purchased. A multinominal logit model and a nested logit model are specified and estimated. In addition to the effects of inventory and brand loyalty, the most interesting results relate to the price elasticities of choice probabilities. These elasticities are much more elastic than those found in traditional food demand analysis. Different brands and sizes of spaghetti products on the same supermarket shelf are very close substitutes for the consumer.