1: \begin{abstract}

2: This paper brings a contribution to the Bayesian theory of

3: nonparametric and semiparametric estimation. We are interested in the

4: asymptotic normality of the posterior distribution in Gaussian linear

5: regression models when the number of regressors increases with the

6: sample size. Two kinds of Bernstein--von Mises theorems are obtained in

7: this framework: nonparametric theorems for the parameter itself, and

8: semiparametric theorems for functionals of the parameter. We apply them

9: to the Gaussian sequence model and to the regression of functions in

10: Sobolev and $C^{\alpha}$ classes, in which we get the minimax

11: convergence rates. Adaptivity is reached for the Bayesian estimators of

12: functionals in our applications.

13: \end{abstract}