The purpose of this paper is to provide an explanation of the stability and change of issue-specific opinions during an electoral campaign. To structure our analysis we use a revised version of Zaller's model of opinion formation. Arguing that Zaller's model suffers from a cognitive bias, we amend it so as to better take into account the normative resistance to change, i.e. the role played by "opinion crystallization". Additionally, we assume that the stability of opinions also depends on the characteristics (intensity and direction) of the information flow of a particular issue. From this we derive a number of hypotheses and test them on data from a three-wave panel-study carried out in the context of the 1999 Swiss federal elections. The empirical tests confirm that opinion crystallization has an overriding impact on resistance to change. They also provide encouraging support for our hypotheses that issue-specific campaigns influence the likelihood of change, both directly, and through interaction with political awareness.