This paper aims to study, by means of a laboratory experiment and a simulation model, some of the mechanisms which dominate the phenomenon of knowledge diffusion in the process that is called ‘interactive learning’. We examine how knowledge spreads in different networks in which agents interact by word of mouth. We define a regular network, a randomly generated network and a small world network structured as graphs consisting of agents (vertices) and connections (edges), situated on a wrapped grid forming a lattice. The target of the paper is to identify the key factors which affect the speed and the distribution of knowledge diffusion. We will show how these factors can be classified as follow: (1) learning strategies adopted by heterogeneous agents; (2) network architecture within which the interaction takes place; (3) geographical distribution of agents and their relative initial levels of knowledge. We shall also attempt to single out the relative effect of each of the above factors.