Background: In the alcohol literature, most cognition research has focused on alcohol expectancies. Overall scores on alcohol expectancy questionnaires predict longitudinal changes in drinking behavior. However, the association between expectancy subscale scores and drinking behavior is inconsistent and modest. Attitudinal information may vary in accessibility (i.e., the ease or speed with which attitudinal information is evaluated), and more accessible (i.e., faster) attitudinal evaluations may predict heavier alcohol consumption. An alternative hypothesis arising from response competition theory is that heavy drinkers will show slow attitudinal evaluations, given that heavy drinkers frequently show mixed, inconsistent, or opposing alcohol-related beliefs. In this study we examined the degree to which speed of processing of alcohol-related evaluative judgments longitudinally predicted alcohol consumption in university students. Methods: Response latencies for alcohol- and non-alcohol-related attitudinal inquiries and measures of alcohol use were administered at two intervals 4 months apart. Results: The key findings were that response times for tension reduction and cognitive impairment items at time 1 positively predicted alcohol consumption at time 2. These results held after we accounted for autocorrelation in drinking and individual variability in reactivity. Conclusion: For items relating to tension reduction, response times fit a response competition hypothesis better than the attitude accessibility hypothesis.