The field of public opinion research has gone through three major periods with respect to the dominant method of data collection, and may now be on the verge of entering a fourth. In the first of these periods, which lasted until the mid 1930's, the primary instrument was the straw poll in which a magazine or newspaper would ask a self-selected sample of their readers to express their opinion by returning selfadministered ballots or questionnaires. In the middle 1930’s, notorious failures of this method had led to a second, more scientific period, characterized by doorto- door interviewing and area probability samples of the general public. By the early 1960's, sharply increasing costs and declining response rates of door-to-door surveys led to a third period in which the instrument of choice became the telephone survey, conducted from a central location and most often using some form of random digit sampling.

In recent years, telephone surveys have seen a marked increase in costs, along with a sharp decline in response rates, leading many to look for new and more efficient methods for data collection. This, in turn, has led to much speculation on ways of using the emerging “Electronic Highway” as a vehicle for conducting surveys. There remain, however, many issues to be resolved before this approach can be used effectively. In this paper we report the results of a series of experiments in using E-Mail and on-line surveys to

Services Company among its subscribers. The primar focus will be on a tracking study of presidential and congressional approval that was conducted continuously from January 1993 through February 1995. We also report on a number of on-line surveys conducted during the presidential election campaign of 1992.

The most significant, and somewhat surprising, conclusion from these experiments is that they produced results which are in many respects comparable to those obtained in national surveys by the major polling organizations.