1: \begin{abstract}

2: For various two dimensional non linear $\sigma$ models,

3: we present a direct comparison between the $\beta$ functions computed with the

4: $2+\epsilon$ renormalization group

5: and the $\beta$ functions measured by Monte Carlo simulations.

6: The theoretical and measured $\beta$ functions match each other nicely for

7: models with a trivial topology,

8: yet they disagree clearly for models containing topological defects.

9: In these later cases, they are compatible with a phase transition at a finite

10: temperature.

11: This indicates that the global properties of the manifold do matter,

12: in contradiction with the assumption used in the $2+\epsilon$ RG computation.

13: \end{abstract}

14: