Do computer scientists deeply understand the traditional Arabic morphology? (What to keep and what to drop from this tradition?) Alexis Amid Neme and Eric Laporte ϥوّ يΑسوΎ الحϥهندسوϤ يفهم ال هل ؟  ΎيقϤ ع  ΎϤ فهم الصرف عل Since 1990, several teams of computer scientists have implemented the traditional model of Arabic morphology in systems of Natural Language Processing (NLP) without questioning its aims, assumptions, definitions, and purposes. Early grammarians and lexicographers had designed Arabic morphology and lexicography for human minds tooled up with paper; whereas we should design Arabic computational morphology for humans equipped with processors and memory devices. These two statements do not seem obvious to computer scientists. The aim of forerunners of grammar in the eighth century was to discover the features of the Arabic language, and they had political and religious incentives. These pioneers accumulated knowledge in semantics, syntax, morphology, phonology and lexicography, produced inventories in order to standardise the language, generating the massive grammatical production of that time. Teaching for native and non-native speakers probably soon became a pressing goal due to geographical expansion. Language teaching has always been focused on vocabulary, word meaning and text understanding. As for other Semitic languages, Arabic morphology was established around theabstract notion of root, three consonants representing a meaning, whether precise or vague. The traditional derivational morphology based on the root-and-pattern model has developed around this abstract consonantal root. In this model, each word is represented by the intersection of a root and a pattern, such as kitaAb = [ktb &1i2aA3] (kitaAb, book, ΏΎكت) i . A pattern is a discontinuous affix (or transfix), made of vowels and non-radical consonants inserted around slots for the root consonants. To each pattern, traditional grammar associates a morphological category and/or inflectional features, and/or semantic features such as agent (kaAtib, writer, ΐتΎك),