1: \begin{abstract}

2: We provide explicit nonasymptotic estimates for the rate of convergence

3: of empirical means of Markov chains, together with a Gaussian or

4: exponential control on the deviations of empirical means. These

5: estimates hold under a ``positive curvature'' assumption expressing a

6: kind of metric ergodicity, which generalizes the Ricci curvature from

7: differential geometry and, on finite graphs, amounts to contraction

8: under path coupling.

9: \end{abstract}