The status of Old Icelandic with respect to (argument) configurationality was subject to debate in the early 1990s (e.g. Faarlund 1990; Rögnvaldsson 1995) and remains unresolved. Since this work, further research on a wide range of languages has enhanced our understanding of configurationality, in particular within Lexical Functional Grammar (e.g. Austin & Bresnan 1996; Nordlinger 1998) and syntactically annotated Old Icelandic data are now available (Wallenberg et al. 2011). It is thus fitting to revisit the matter. In this paper, I show that allowing for argument configurationality as a gradient property, and also taking into account discourse configurationality (Kiss 1995) as a further gradient property, can neatly account for word order patterns in this early stage of Icelandic. Specifically, I show that corpus data supports part of the original claim in Faarlund (1990), that Old Icelandic lacks a VP-constituent, thus being somewhat less argument-configurational than the modern language. Furthermore, the observed word order patterns indicate a designated topic position in the postfinite domain, thus reflecting some degree of discourse configurationality at this early stage of the language.