Defects like micro-cracks have a major influence on the behavior of many materials, even in the elastic range. To model these influences, the defects are usually smeared and equivalent effective values of the elastic properties evaluated: this type of procedure is summarized and expounded in the first part of the lecture. Following these procedures, however, account can be taken of the randomness but only under much limiting hypotheses, including ergodicity. It is rather difficult to account for the actual randomness in the spatial distribution of micro-cracks. This consideration has motivated the numerical research presented in the second part of the lecture: the results obtained seem to indicate that, at least under the simplifying assumptions adopted and within the range of stochastic variability considered, the location of strain localization is not much affected by the randomness, but the effects of the interactions between micro-cracks are the mechanisms that generated such strain localizations.