The widely accepted conception of the triliteral root as the basis of the lexical system of Semitic languages is refuted by findings of two studies described by Abelhadi Razouk (1999-2000). In study 1, Arabic speakers enrolled in university programs of Arabic language & literature in Morocco & France (N = 107 & 25 respectively) were tested on the identification of the root in 20 selected Arabic words; results show only 0%-18% correct responses for all but three of the test items. Study 2 presented the same words in a connected text to first-year students of Arabic (N = 10) for sight reading, after which the test of study 1 was repeated. Results indicate a reading error rate of 3% & an error rate of 93% in root identification across all Ss & are argued to show that the root is irrelevant to the reading process. It is proposed instead that the Arabic lexicon is organized at three levels: (1) the phonetic matrix, an unordered combination of matrices of phonetic features linked to a semic nucleus; (2) the etymon, an unordered combination of phonemes that particularize the semic nucleus of the matrix; & (3) the radical, an expansion of the etymon by adding at least one vowel & either a resonant or a copy of the last consonant, with expansion of the semic nucleus.