Two experiments examine the adequacy of the expectancy hypothesis explanation (Marschark & Paivio, 1979) of the congruity effect in symbolic magnitude comparisons.  According to that hypothesis, an instruction to select the larger or the smaller object primes objects of large or small magnitude, respectively.  Accordingly, large objects will be compared more rapidly under the instruction to choose the larger, but small objects under the instruction to choose the smaller.  A new paradigm in which subjects indicate whether a second object is larger or smaller than the first tests this model.  In this new paradigm, there is no differential priming by an instruction.  Nevertheless, comparable congruity effects occur for the same materials across priming and nonpriming paradigms, demonstrating that congruity effects are not likely to be due to expectancies.  These experiments support the discrete code model account (Banks, 1977) of congruity against models that ascribe the effect to differential processing induced by an instruction prior to or subsequent to a trial.