In most nutrition intervention studies it is not nutritional status but 1 or more specific aspects of nutrition that are measured and the difficulty of measurement has increased with the complexity of the intervention models. The direct intervention model such as the iodization of salt includes the provision of a nutrient supplement to control a specific nutritional deficiency disease with the end effect easy to measure. The essence of this model is that confounding variables are randomly assigned to the experimental and control groups; they are not eliminated or measured but are balanced. 2 problems arise in applying this model in the community: 1) violation of the random assignment rule and 2) meeting design requirements may force a change in the questions being asked. In a multi-stage model there are some elements that increase the complexity of the study: 1) the expected end effect has less specificity 2) the influence of the intervention may be conditioned to a degree by other variables and 3) the hypothesis linking the intervention to the expected outcome has a less severe foundation. The design of this model has moved from simple direct intervention to one of epidemiology. Pathway analysis of intervention programs may help to explain negative evaluations and lead to program improvement. The multi-effect model considers the many pathways which program effects might take; the author argues that linear thinking may result in negative assumptions about program effects. As interventions have changed methodological demands have done the same. There is now a need to measure both intervening and confounding variables as well as outcome variables a recognition that researchers should be looking toward functional measures as outcome variables and that analytic approaches used in more elaborate designs call for data analysis of the individual rather than at the group level. Other methodological considerations include definition of goals identification of the control group duration of observation field staff data handling and statistical advice. The most crucial element in this area is establishing proper goals at the onset of the study.