Well there isn't a nice and easy test, but in linguistics it tends to come down to some measure of mutual intelligibility. How much meaning can a speaker of Variety A garner from a speaker of Variety B? And vice versa. For tricky distinctions, throw in some historical and social analysis to taste.

I don't know much about Mesoamerica but I believe the "30 languages" claim refers to the count of current languages descended from a single ancestor, not the languages of Classic Maya. (EDITED to avoid implying "Proto" and "Classic" Maya were one in the same--they were separated by millenia!)

English is a particularly tricky issue because there are so many "standard" Englishes understood by many but natively spoken by none, and because lingua francas tend to become such a smooth continuum between their different lects. However for the purposes of the mutual intelligibility tests, the vast majority of world Englishes would still fall under one "language". Many that don't are already labelled distinctly, either as creoles/pidgins, or from historical divergence (e.g Scots, Frisian). English' divergence is little compared to that of, say, Arabic, and I suspect any two random English speakers in the world would understand each other much much more easily than any two Mayan speakers.