Does the replica method give correct thresholds for low-density parity-check codes? This note shows a graph comparing the thresholds derived by the typical set method (McEliece et al), the density evolution method (Richardson and Urbanke), and the replica method (Nakamura and Kabashima).