A major challenge in the study of animal communication is distinguishing whether signals convey honest or dishonest information. Biologists infer the honesty of a signal from its correlation with the information being signaled (e.g., fecundity or fighting ability) — the better the correlation, the more honest the signal. However, this view of signaling potentially conflates unreliable indicators with dishonest signals. Just because a trait conveys unreliable information does not mean that the structure serves as a dishonest signal; developmental noise, genetic drift, and environmental constraints can reduce the covariation between a putative signal and an organism's quality. Moreover, a trait must influence the behavior of other organisms to qualify as a signal. We studied how a putative signal, claw size, affects physical performance and social dominance in three species of stream-dwelling crayfish, which fight routinely over resources. For comparison, we measured the relationship between claw size and claw strength in three species of burrowing crayfish, which do not fight. In all species, crayfish with larger claws were not necessarily stronger, indicating that claw size poorly indicates claw strength. However, claw size is unlikely to function as a dishonest signal of fighting ability in these species for two reasons. First, claw size unreliably indicates claw strength in both the stream-dwelling species and the burrowing species. Second, relative claw size poorly predicted whether stream-dwelling crayfish would escalate aggression, questioning whether claw size serves as a signal. Instead, stream-dwelling crayfish escalated aggression based on their relative body size. Our results highlight the importance of observing behavior when determining the honesty of a signal. As such, future studies must distinguish between phenotypes that are unreliable indicators (or reliable indicators) from phenotypes that have been selected for dishonest (or honest) signaling.