The paper describes and discusses the sample design to measure television viewing among the Hispanic population in the U.S. using a peoplemeter. The sample design is a two phase sample. The first phase sample is a sample of U.S. households in which screening interviews are conducted to find Hispanic households. Methods used to define and identify Hispanic households are described. An operational definition of Hispanic persons is needed to determine the households to be included in the meter panel. The definition used is based on national origin and is consistent with the definition used by the U.S. government which is the source of Hispanic population estimates. Data is provided regarding the distribution and concentration of the Hispanic population. How this information was used to develop a sampling frame for the screening sample is described. The possible benefits and disadvantages of oversampling of areas with high concentrations of Hispanic households to reduce screening costs are discussed. Weighting to compensate for different selection probabilities may increase design effects requiring a larger meter panel to achieve the same precision as a proportionate sample. The paper shows how the extent of oversampling is related to the relationship of screening costs to the cost of collecting tuning and viewing data. There was concern that less acculturated Hispanic households might be less willing to participate in the peoplemeter sample. Language use by Hispanic households was obtained during the screening interview and used as a proxy measure for acculturation. Language use and concentration of Hispanic households in the neighborhood in which Hispanic households resided were used to stratify the sample to control for possible acculturation differences.