The current article offers a comparative account of the address system of two dialects of Arabic, Najdi and Tunisian Arabic. Capitalizing on the theory of Idealized Cognitive Model, the article defends the view that the two systems display Idealized models, which are central to the system, and non-Idealized models, which are peripheral to it. Najdi Arabic includes Idealized terms such as first names, teknonyms, and kinship terms while non-Idealized models include a battery of terms of address. Tunisian Arabic Idealized models hinge on Si/Lalla + first names, first names, and kinship terms while non-Idealized models make use of endeared first names, kinship terms, and diminished kinship terms. The two systems are shown to differ at the level of types of encounter (including formality, closeness, and deference), availability of address options, social horizontality-verticality, and use of metaphor and metonymy.