Automatic acquisition of qualia structures is one of the directions in information extraction that has received a great attention lately. We consider such information as a possible input for the word-space models and investigate its impact on the categorization task. We show that the results of the categorization are mostly influenced by the formal role while the other roles have not contributed discriminative features for this task. The best results on 3-way clustering are achieved by using the formal role alone (entropy 0.00, purity 1.00), the best performance on 6-way clustering is yielded by a combination of the formal and the agentive roles (entropy 0.09, purity 0.91).