Finally, feature set theor (for theoretical) generalizes and adds to the theoretical properties used in the first experiment (Table 3 in Section 4.2). Upon inspection of the clustering solutions (not reported here for space reasons), some further potentially relevant distributional pieces of information cropped up that were included in the theor features of the present experiment. The new features, summarized in Table 9, cover several aspects of the noun phrases (NPs) in which adjectives occur: The type of determiner of the NP, agreement properties (as these can correlate with semantic properties), the syntactic function of the head noun, and the presence of a potential adjective complement. The latter are usually headed by prepositions (El Joan està gelós d’en Pere, ‘Joan is jealous of Pere’). Finally, feature distance to the head is a reformulation of feature adjacent from Section 4.2. It encodes the mean distance of the adjective to the head, in number of words, as this is a more general definition that alleviates data sparseness.