This paper presents an experimental study on cheating, specifically focusing on lying behavior. The study uses a novel design where participants roll a die privately and report the number to receive a payoff based on that number. The design has three key advantages: it cannot detect individual cheating, it is easy to implement, and it allows for testing theoretical predictions. The results show that about 20% of inexperienced subjects lie fully, while 39% remain honest. A significant portion (20%) of subjects lie partially, not reporting the payoff-maximizing number. The study explores various motives for these behaviors, including moral concerns, maintaining a positive self-image, and avoiding being perceived as greedy. The findings suggest that while some people lie to maximize their payoffs, others try to disguise their lies to maintain a favorable self-concept. The study also examines how lying behavior changes under different conditions, such as varying stakes, externalities, and anonymity, and finds that the observed patterns remain robust.This paper presents an experimental study on cheating, specifically focusing on lying behavior. The study uses a novel design where participants roll a die privately and report the number to receive a payoff based on that number. The design has three key advantages: it cannot detect individual cheating, it is easy to implement, and it allows for testing theoretical predictions. The results show that about 20% of inexperienced subjects lie fully, while 39% remain honest. A significant portion (20%) of subjects lie partially, not reporting the payoff-maximizing number. The study explores various motives for these behaviors, including moral concerns, maintaining a positive self-image, and avoiding being perceived as greedy. The findings suggest that while some people lie to maximize their payoffs, others try to disguise their lies to maintain a favorable self-concept. The study also examines how lying behavior changes under different conditions, such as varying stakes, externalities, and anonymity, and finds that the observed patterns remain robust.