ABSTRACT The retrospective model of voting holds that parties are elected based on their perceived competence in government and punished for incompetence. This theory is complicated in contexts of multilevel governance as it requires attributing responsibility to a particular party in a particular level of government. This paper tests the extent to which this model explains public opinion in a multilevel context (Scotland) on a high-stakes valence issue (the coronavirus pandemic). Using British Election Study Internet Panel data from 2019 to 2021, it assesses how the public responded to the perceived competence of the Scottish Government and British Government. The results find that for each Government those voters who rated their handling of the pandemic more highly increased their liking of the governing party and its leader. Further, those voters who thought a Government had performed well were more likely to switch towards supporting the party in power, but this depended on how much they previously liked that party. Similarly, increases in support for Scottish independence from the UK came from those voters who not only judged the Scottish Government to have handled the pandemic well, but also previously liked the SNP prior to the pandemic.