Bayesian and maximum likelihood methods and quarterly United States data are employed to estimate parameters of an aggregate consumption function, $C_{t} = \lambda C_{t-1} + (1 - \lambda) k Y_{t} + u_{t} - \lambda u_{t-1}$, under four different assumptions about the error terms, $u_{t} - \lambda u_{t-1}, t = 1, 2, \ldots, T$. After a discussion of the results of estimation, Bayesian techniques for computing posterior odds on alternative models are described and applied to obtain posterior odds relating to different formulations of the consumption model with alternative prior probability density functions for the parameters.