Political scientists often debate how much information people have and deploy when making electoral decisions. Some scholars suggest that voters are aware of which party is likely to win in their local constituency at British general elections; however, this might not be the case in situations when there is substantial and spatially varying change in the relative fortunes of two or more parties between elections. That argument is evaluated here using as a case study the 2015 and 2017 general elections in Scotland: at the first, the SNP's vote share more than doubled, and it won 56 of the country's 59 seats, having won just six at the previous contest; at the second, its vote share fell by about a third, and it lost 21 of those 56 seats. Analysis of British Election Survey data collected before and during the campaigns preceding those elections shows that most respondents were aware of the SNP's surge in 2015 and expected their victory in every constituency. In 2017, most voters were aware which of the SNP's three competitors was the biggest threat in each constituency, and that awareness became clearer during the campaign; yet, voters – especially those who identified with the SNP and were contacted by it during the campaign – still (incorrectly) anticipated a local SNP victory.