Many of the languages have a fortis–lenis contrast in plosive consonants. Lenis/short plosives have weak contact and intermittent voicing, while fortis/long plosives have full closure, a more powerful release burst, and no voicing.
This is close to what Evans (1997) proposed under the name Gunwinyguan (cf. his very different proposal of Arnhem Land languages.)
However, Green (2003) argues that only Maningrida has been established as a valid subgroup, and that the interrelationships of the other languages are as yet unclear. The evidence for Gunwinyguan and perhaps other nodes listed above may simply be reflections of a relationship of all Arnhem languages when only a subset of them was investigated. That is, these groups may be based on shared retentions of Proto-Arnhem rather than distinct historical developments. (However, in reviewing Green, Evans pointed out that much of the Maningrida morphology was also shared by Mangarrayi.) An agnostic view of the family would list each language separately, except for the established Maningrida branch:
*Green does not address Anindilyakwa, Alawa, or Yugul. Yugul is too poorly attested for comparison based on her methods; the other two await validation.