The existence of nominal classification systems has long been thought of as one of the defining features of the Papuan languages of island New Guinea. However, while almost all of these languages do have nominal classification systems, are, in fact, extremely divergent from each other. This paper examines these systems in the East Papuan languages in order to examine the question of the relationship between these Papuan outliers. Nominal classification systems are often archaic, preserving older features lost elsewhere in a language. Also, evidence shows that are not easily borrowed into languages (although can be). For these reasons, it is useful to consider nominal classification systems as a tool for exploring ancient historical relationships between languages. This paper finds little evidence of relationship between the nominal classification systems of the East Papuan languages as a whole. It argues that the mere existence of nominal classification systems cannot be used as evidence that t he East Papuan languages form a genetic family. The simplest hypothesis is that either the systems were inherited so long ago as to obscure the genetic evidence, or else the appearance of nominal classification systems in these languages arose through borrowing of grammatical systems rather than of morphological forms. 1. BACKGROUND (1) The languages to be evaluated, the East Papuan languages, include all but one of the non-Austronesian languages spoken in the islands off the east of mainland New Guinea (one language, Kovai, spoken on Umbai Island, off the coast of mainland Papua New Guinea, is clearly of the Trans New Guinea Phylum [Ross 2001:301]). The East Papuan languages as a whole are currently considered as a in this paper solely on the basis of their geographical proximity. The best known classification of East Papuan languages is that of Wurm (1982: 231-244, basically repeated from Wurm 1975), while a more recent classification comes from Ross (2001). Ross (to appear, b) is a fuller treatment of the same data. Wurm's classification is based mainly on lexical and typological features including verb systems and forms, gender systems and the form of the gender markers, etc. (Wurm 1982:231). However, Wurm does not indicate the data on which his analysis was based, and at least one of the languages included (Kazukuru) became extinct before grammatical data were available on it. It appears that some languages were included in the genetic classification simply because of their geographical location, rather than because of specific shared features. Malcolm Ross has revived interest in this area recently, using the comparative method to undertake a comparison of the pronoun systems of these languages (Ross 2001; to appear, a; to appear, b), and on that basis argues for eight separate genetic groups. Ross leaves out of his analysis the languages of Reefs-Santa Cruz, on the grounds that have had complex contact histories with Austronesian and Papuan mixing (to appear, b). This notwithstanding, he still states that they have no genealogical relationship to any other east Papuan group (2001:310). Figure 1 in Dunn, Reesink, and Terrill (2002) (31) compares Wurm's and Ross's classifications. The two concur in many of the lower-level groupings, but differ significantly in that for Wurm all the East Papuan languages belong in one genetic grouping, while for Ross there are eight separate genetic groups involved. Dunn, Reesink, and Terrill (2002) examine various linguistic features, including clausal and NP constituent order, pronominal systems, and the structure of verbal morphology, and conclude that these grammatical subsystems show clear similarities between Anem and Ata (New Britain), and also between some of the languages of Bougainville, but not for the East Papuan languages as a whole. The present paper focuses on the smaller subsystem of nominal classification, with a view to elucidating the same questions. …