Conventional evaluations of voting systems focus on ballots for which no vote can be recorded (that is, “residual” votes). However, recorded votes that misrepresent voter intent are another potentially important, but less easily measured, source of error. I present evidence that a nontrivial number of voters in the recent California recall election mistakenly voted for one of the four candidates positioned next to the two major candidates on the ballot. I also find that punch-card systems significantly increased the frequency of these errors. These results indicate that future assessments of voting technologies should consider their effects on both recorded and residual votes. Copyright by the President and Fellows of Harvard College and the Massachusetts Institute of Technology.