Allocation Research (Continued) By Michael Baksh (UC-Los Angeles) In previous installments I reviewed the advantages and disadvantages of the spot observations technique in time allocation research (Volume 1, Issue 2), and described problems that arise when doing spot observations in the field (Volume 2, Issue 2). Here I focus on the issue of sample size. How many individual observations are needed to gain representative information on how people spend their time? How many visits should be made to study households? Questions as these are important not only for reasons of representativeness, but also for directing the amount of ficldwork that should be devoted to collecting data. Because the spot observations technique is highly efficient for collecting data, field workers occupied with other research priori ties can still produce good estimates of a population's time use with minimal effort. In my research I collected 4,182 observations of the Machiguenga (41 households) and our Kenya research team collected 82,050 of the Embu (169 households). I am fairly comfortable with each sample although, for reasons described below, little would be lost for many research purposes if these samples were considerably smaller. The most critical factor to consider when designing the sample size of a spot observations project is the smallest unit of analysis for which the time allocation data will be used. Will the time allocation data be analyzed at the community level, or will data be requi red to address hypotheses at the household or even individual level? Clearly, a larger total number of observations is required if a researcher plans to describe how each woman in a community spends her time, rather than how the average woman in the community spends her time. A second factor, also related to research needs, is the level of detail for which information on time use is required. If a research objective is to study total interhousehold time, fewer observations need be made than in a project concerned with capturing the time that different households devote to very specific or unusual work activities such as planting specific types of crops or taking cattle to a cattle dip. A third factor that should be considered is the level of sociocultural complexity. Peasant societies, for example, are more complex than family-level horticultural societies, as reflected by levels of social and political organization, types of economic exchange, and other sociocultural variables. As such, peasant societies engage in more unique kinds of activities than family-level societies, and their interhousehold variations in time use arc likely to be greater due to (if nothing else), greater economic specialization.