This paper develops an asymptotic theory of Bayesian inference for tim
e series. A limiting representation of the Bayesian data density is ob
tained and shown to be of the same general exponential form for a wide
class of likelihoods and prior distributions. Continuous time and dis
crete time cases are studied. In discrete time, an embedding theorem i
s given which shows how to embed the exponential density in a continuo
us time process. From the embedding we obtain a large sample approxima
tion to the model of the data that corresponds to the exponential dens
ity. This has the form of discrete observations drawn from a nonlinear
stochastic differential equation driven by Brownian motion. No assump
tions concerning stationarity or rates of convergence are required in
the asymptotics. Some implications for statistical testing are explore
d and we suggest tests that are based on likelihood ratios (or Bayes f
actors) of the exponential densities for discriminating between models
.