Description:
Strack and Martin (1987) have proposed that although probability samples of general populations are needed for studying attitude content, psychological processes such as context and other response effects can be investigated quite well or even better with homogeneous convenience samples. The present paper shows that, in contrast, when social change is included as an essential variable in a study of a well-known context effect, the process itself is weakened and may even disappear. The distinction between content and process is not always tenable in surveys once we include long-term social change as a variable, as we must as the horizon for survey research extends further and further into the future.