Foley (2018) notes that Grass languages share very few lexical items with the other Ramu languages, with virtually no lexical cognates Banaro and Ap Ma.
However, the Grass languages are still classified as Ramu due to widely shared morphosyntax and typology.
[1] The original Grass language proposal, also known as Keram, included several languages, such as Banaro and Kambot (Ap Ma), that are no longer thought to be closely related to Adjora and Gorovu.
Laycock (1973) rejected Kambot and noted that Banaro was lexically divergent, and therefore grouped it with the Grass family in a higher-level Grass stock,[2] a position accepted by Pawley (2005).
[3] Grass/Keram (Laycock) Timothy Usher (as reported in Glottolog) broke it up still further, with only Abu (Adora) and Gorovu kept together (in a "Porapora River" or "Agoan" branch),[4] Aion (Ambakich) and Kambot (Ap Ma) grouped with the Mongol–Langam languages, and Banaro left as a primary branch of Ramu proper.