Two experiments were conducted, using Moroccan Arabic data, to evaluate conflicting predictions of autonomous and interactive models of spoken word recognition. In Experiment 1, lexical decision response times indicated the presence of strong lexical effects both with monosyllabic and bisyllabic words. In experiment 2, a General Phoneme Monitoring task was used in which subjects were asked to monitor for a target phoneme located at four different positions before and after the Uniqueness Point (UP). Strong lexical effetcs were obtained before UP. The bearings of these results on current autonomous and interactive models are discussed.