In this talk, I return to some suggestions I made in the literature back in the early 2000s that the elements that are treated as garden-variety (in)definite articles in Gbe (and presumably in Kwa and Niger-Congo in general) are nominal discourse markers different from typical Germanic- and Romance-type articles (Aboh 2004a, b). In addition, a quick typological survey shows that the Germanic- and Romance-type articles are not commonly observed cross-linguistically. Furthermore, articles emerged in these languages, presumably as an areal feature, between the 8th and 11th century (cf. Perridon and Sleeman 2011: 3), hence their absence or scarcity in older Germanic and Romance varieties. The empirical arguments for the equation “article = D” are therefore not cogent. All we are left with is the semantic argument (cf. Szabolcsi 1987, 1994).

I suggest that the semantic argument does not require duplication of the phase into C and D. More precisely, I argue that (i) there is only one phase, label it “C”, which projects as the left periphery of predicate phrases (cf. Bowers 1993), and (ii) a phase, being a spell-out point (e.g., Bošković 2016), cannot be parametrized.