Most studies of the digital divide are concerned with the simple criterion of access, usually in the convenient locale of one's home. That divide could be exacerbated by usage differences after such access has been achieved, however. This article takes advantage of usage data from the General Social Survey and other surveys to examine whether more highly educated respondents also have such advantages in usage processes after access has been achieved. Education has emerged from the NTIA and other national surveys as a more important multivariate predictor than income.

Using a framework developed by DiMaggio and Hargittai (2001), it is found that college-educated respondents possess clear advantages over high-school educated respondents in using the Internet to derive occupational, educational and other benefits. The clearest advantage appears in terms of the types of sites visited, uses made and political discussion. Here, multivariate evidence shows that education - and occasionally income, age and marital status - is associated with consistently more long-term uses related to enhanced life chances via work, education, health or political participation; education is also related to less use for simple, short-term, entertainment or personal purposes. The advantages to the college educated are also evident in their keeping in contact with a wider range of friends and relatives, particularly by email. On the other hand, in several areas (e.g. search strategies employed; receiving assistance from relatives) little gap by education exists.

Most analyses of the digital divide have conceived of Internet access as binary - either someone is an Internet user or is not. Using data from a 2002 national random digit dial survey, this article visualizes online access as a continuum. Internet access may be intermittent for some users, nearby for others (such as nonusers household in which another person uses the Net), and a remote possibility for others (given their preferences, perceptions and concerns about the Internet).

This article then proceeds to analyze the social, demographic and psychological predictors of Internet users and nonusers. Demographic factors (being white, well educated, and having a high income) are associated with more Internet adoption, as are high levels of trust, social contentment and media use. Controlling for other variables, Hispanics and African-Americans are less likely to be online, as are people who report frequently socializing with family and friends and being members of social groups or clubs.