A central assumption of the canonical cheap talk literature is that people misreport their private information if this is to their material benefit. Recent evidence from laboratory experiments with student subjects suggests, however, that while many people do report the payoff-maximizing outcome, some report their private information truthfully or at least do not lie maximally. We measure truth-telling outside the laboratory by calling a representative sample of the German population at home. In our setup, participants have a strong monetary incentive to misreport, misreporting cannot be detected, and reputational concerns are negligible. Yet, we find that aggregate reporting behavior closely follows the expected truthful distribution. Our results underline the importance of lying costs and raise questions regarding the influence of the decision-making environment and the elicitation mode on reporting behavior.