This article examines a commonly used micromechanics approach to evolution of damage in composites. In this approach an initial crack pattern is assumed and on the basis of the local stress fields and a (brittle) fracture criterion, further evolution of the same crack pattern is determined. We question this assumed self-similarity in the damage mode by considering two alternative crack modes, one of which would result in deviation from the assumed crack pattern. Using a simple maximum stress criterion for formation of brittle cracks, it is found that the self-similarity breaks down before saturation of cracking in the assumed crack pattern. This seemingly inherent inadequacy in the damage evolution analysis is discussed.