Continuous and overlapping sounds in connected speech yield an output that has very few reliable cues to detect word boundaries. This may obscure listeners’ recognition of spoken words. Therefore, investigations to find the processes native listeners use to start lexical access have been the focus in psycholinguistic studies. Segmentation was identified as one of the processes. Research conducted on different languages identified different prosodic units employed in segmentation and the recognition of spoken words. The following paper reports on one of the first studies conducted on Arabic connected speech investigating the role stress plays in the segmentation and recognition of words in spoken Cairene Arabic. Phonologically Cairene Arabic is identified as a stress-timed language. However, few empirical studies have been conducted to validate or refute this classification. Using the ‘word spotting’ technique, the present study found an effect for stress as a segmentation unit, hence providing empirical evidence for the theoretical classification.