We use natural experiments – plausibly exogenous, anticipated increases in the piece rate – to study how effort responds to incentives. Our first finding, like some previous studies, lends little support to the view that incentives increase effort: raising the piece rate has zero effect on total daily effort. Previous studies have speculated that changes in motivation over the course of the workday, caused by the increase in the piece rate, may lead to this result, but have relied on data aggregated to the day. Our data allow us to look within the workday. We find that workers do respond to incentives within the day: they work significantly harder in early hours of work, but significantly less hard later on, with a net effect of zero on total daily effort. We consider different possible explanations for this behavior. The most parsimonious explanation is a model in the spirit of Loewenstein and O'Donoghue (2005), in which a cognitive system, assumed to behave like the standard economic model predicts, is in conflict with the affective system. We review evidence from psychology and neuroscience to argue that the affective system may be strongly influenced by within-day changes in earnings, relative to an earnings goal. The affective system cares less about income once the goal is surpassed, providing an explanation for a drop in effort later in the day, and for the findings of earlier studies.