There is a tendency to overgeneralize in sociology and social psychology, and the logical conditions for extension of conclusions outside the data are not fulfilled. As a result, replications often do not verify the original studies. Some suggestions for improving research design to achieve consistent replications are discussed: the use of theory; statement of categories or variables in generalizable terms; selection of cases throughout the range of the distribution; the use of "realistic" definitions, assumption, and variables.