As more survey and opinion polling agencies collect their data using the Internet, questions arise about how representative Internet users are of the American population as a whole. We use a targeted module of IT-relevant questions added to the 2000, 2002, 2004 and 2006 samples in the General Social Survey (GSS) to analyze differences between Internet users and non-users. We devote particular attention to the issue of whether Internet use is associated with more or less diverse or 'liberal' political opinions and to how these associations have changed since 2000. In general, we found that where differences existed, they were in the direction of Internet users being more supportive of diverse and tolerant points of view than were non-users, consistent with the premise that going online is a way of expressing openness to opposing points of view and new experiences. However, the differences were often non-monotonic - that is, openness did not always increase progressively with the amount of Internet use. Moreover, there were differences for some racial/family/sexual/political attitudes but not others, in patterns that do not fit easily under standard labels such as liberal, conservative or even libertarian. Internet users also expressed slightly more optimistic and sociable attitudes on certain other GSS questions. On most GSS items, however, either there were no differences between Internet users and non-users, or the differences could be explained by age, education, race, gender or income factors. Thus, respondents in Internet surveys can be expected to differ from non-Internet respondents in being selectively more tolerant on certain political issues, but not on most other political issues. The results held about as strongly in 2006 as in 2000.