The paper proposes a payment rule for crowdsourced tasks. This rule is intended to incentivize workers to accurately report their confidence (e.g. by skipping a task when they have low confidence), and to pay little to spammers. Payment is based on the product of the evaluations of a worker's responses to a set of gold-standard tasks; if the worker gets a single gold standard task wrong and asserts high confidence, the overall payment is zero.