Individuals care about how they are perceived by others, and take visible actions to signal their type. This paper investigates social signaling in the context of childhood immunization in Sierra Leone. Despite high initial vaccine take-up, many parents do not complete the five immunizations that are required in a child’s first year of life. I introduce a durable signal in the form of differently colored bracelets which children receive upon vaccination, and implement a 22-month-long experiment in 120 public clinics. Informed by theory, the experimental design separately identifies social signaling from leading alternative mechanisms. In a first main finding, I show that individuals use signals to learn about others’ actions. Second, I find that the impact of signals varies significantly with the social desirability of the action. In particular, the signal has a weak effect when linked to a vaccine with low perceived benefits and a large, positive effect when linked to a vaccine with high perceived benefits. Of substantive policy importance, signals increase timely and complete vaccination at a cost of approximately 1 USD per child, with effects persisting 12 months after the roll out. Finally, I structurally estimate a dynamic discrete-choice model to quantify the value of social signaling.