In two experiments we tested the hypothesis that children have a basic problem in mastering the attributive relation because it involves a two-step logical–semantic integration process of the head-noun and the attributive adjective. Hebrew-speaking children were asked to interpret highly familiar adjective–noun combinations by selecting a photo that depicted the correct referent. In Experiment 1 there were four choices for each adjective–noun pair: correct object/correct property, correct object/wrong property, wrong object/correct property, wrong object/wrong property. 170 children (1;6–4;4) participated. Analyses of errors and spontaneous self-corrections indicated that children initially ignored the adjective and based their responses only on the noun. In Experiment 2, in addition to the 4-choice condition, there were two simpler conditions with only two choices: the correct object/correct property, and either the correct object/wrong property, or the wrong object/correct property. 30 children (1;9–4;11) were tested. The children, and in particular the lowest-scoring third of the sample, did significantly better in the 2-pictures conditions. The results suggest that young children do possess a basic adjective vocabulary and can use it in simple discriminations, but have a considerable difficulty in integrating the information furnished by the adjective with information furnished by the noun.