The nature of contextual facilitation effects for items derived from Raven's Progressive Matrices was investigated in two experiments. For these, the original matrices were modified, creating either abstract versions with high element salience, or versions which comprised realistic entities set in familiar contexts. In order to replicate and extend previous research, two versions of the contextual matrices were created, either comprising black and white or colored elements. The best performance was for colored contextual items, then equal performance for black and white contextual and colored abstract versions. By-item and by-participant correlations were invariably positive and significant, showing that irrespective of appearance and content, the best performers at one set of items were the best at another, and that the hardest items for one set were the hardest for another. Overall, these findings suggest that contextual facilitation effects are unlikely to be due to the recruitment of special domain-specific cognitive processes, and that differences in difficulty between versions can be fully accounted for by element salience, verbal tagging, and item engagement. (Contains 4 figures.)