1: \begin{abstract}

2: We investigate how very large populations are able to reach a global

3: consensus, out of local "microscopic" interaction rules, in the

4: framework of a recently introduced class of models of semiotic

5: dynamics, the so-called Naming Game.  We compare in particular the

6: convergence mechanism for interacting agents embedded in a

7: low-dimensional lattice with respect to the mean-field case.  We

8: highlight that in low-dimensions consensus is reached through a

9: coarsening process which requires less cognitive effort of the agents,

10: with respect to the mean-field case, but takes longer to complete. In

11: $1$-d the dynamics of the boundaries is mapped onto a truncated Markov

12: process from which we analytically computed the diffusion

13: coefficient. More generally we show that the convergence process

14: requires a memory per agent scaling as $N$ and lasts a time

15: $N^{1+2/d}$ in dimension $d \le 4$ (the upper critical dimension),

16: while in mean-field both memory and time scale as $N^{3/2}$, for a

17: population of $N$ agents. We present analytical and numerical

18: evidences supporting this picture.

19: 

20: \end{abstract}

21: