Discussions of political issues may influence people's opinions. Is there any systematic difference in opinions between those who discuss frequently and those who do not? We measured the association between self-reported discussion frequency and the probability of holding the more liberal opinion on moral issues, using data from the General Social Survey (81 issues, n = 4,395) and the American National Election Studies (27 issues, n = 17,653). This association looked different among liberals and among conservatives. Having more frequent discussions is associated with a higher probability of holding more liberal opinions among liberals, while there is little association between discussion frequency and opinions among conservatives. These findings can be explained by the moral argument theory, which is an account of the long-term liberalization of public opinion on moral issues as an outcome of repeated discussions. The key assumption of this theory is that opinions that are justified by the kinds of arguments that only conservatives accept have a disadvantage compared to opinions that are justified by the kinds of arguments that everyone accepts. Consistent with this theory, we find that the effect of discussion frequency is stronger for moral opinions that have a bigger argument advantage.