Banks and Sobel (1987) analyze equilibrium selection in signaling games, introducing a new solution concept called "divine equilibrium" that refines sequential equilibria by imposing restrictions on off-the-equilibrium-path beliefs. Divine equilibria require that beliefs about off-the-equilibrium-path events are consistent with the idea that no type of sender would deviate from equilibrium if they expected to lose. This concept rules out implausible equilibria in many examples and is shown to exist by demonstrating that non-divine sequential equilibria cannot be in stable components. However, stable components of signaling games are typically smaller than the set of divine equilibria. The authors also provide a characterization of stable equilibria in generic signaling games, showing that an equilibrium is stable if and only if, for all unused signals, the set of beliefs that can support it is empty. This result is consistent with Cho and Kreps (1987), who also analyze the power of stability to select equilibria in signaling games. The paper concludes that divine equilibria provide a useful framework for selecting equilibria in signaling games, as they capture a minimal restriction on off-the-equilibrium-path beliefs. The authors also discuss the implications of their results for more general extensive-form games and note that divinity may be easier to verify than stability and may be simpler to generalize to games with infinite strategy spaces.Banks and Sobel (1987) analyze equilibrium selection in signaling games, introducing a new solution concept called "divine equilibrium" that refines sequential equilibria by imposing restrictions on off-the-equilibrium-path beliefs. Divine equilibria require that beliefs about off-the-equilibrium-path events are consistent with the idea that no type of sender would deviate from equilibrium if they expected to lose. This concept rules out implausible equilibria in many examples and is shown to exist by demonstrating that non-divine sequential equilibria cannot be in stable components. However, stable components of signaling games are typically smaller than the set of divine equilibria. The authors also provide a characterization of stable equilibria in generic signaling games, showing that an equilibrium is stable if and only if, for all unused signals, the set of beliefs that can support it is empty. This result is consistent with Cho and Kreps (1987), who also analyze the power of stability to select equilibria in signaling games. The paper concludes that divine equilibria provide a useful framework for selecting equilibria in signaling games, as they capture a minimal restriction on off-the-equilibrium-path beliefs. The authors also discuss the implications of their results for more general extensive-form games and note that divinity may be easier to verify than stability and may be simpler to generalize to games with infinite strategy spaces.