We measure the impact of measurement error in labor-supply elasticities estimated over recalled usual work hours, as is ubiquitous in the literature. Employing hours of work in diaries collected by the American Time Use Survey, 2003-12, along with the same respondents' recalled usual hours, we show that the latter yield elasticities that are positively biased. We argue that this bias arises from the salience on recalled hours of differences in wage rates.