Relationships between perceived voter use of information sources and perceived source credibility and gubernatorial candidate preference were examined in a cross-lagged correlation design employing data from Illinois primary and general election surveys. The only significant test-retest correlation for source use and believability was for newspaper credibility. The media use and credibility items were also uncorrelated with gubernatorial candidate preference at both primary and general elections and with post-general election vote reports. All cross-lagged correlations between media variables and candidate preference were non-significant. All test-retest correlations for candidate preference and political party preference were statistically significant. The findings suggest that the media use and credibility variables as measured by Roper-type questions are either sensitive to differing environments of primary and general election campaigns or are unreliable measures of media use and believability. (Author)