Michael Clarke, "Semantics and Vocabulary," in Egbert J. Bakker, ed., A Companion to the Ancient Greek Language
(Chichester: Wiley-Blackwell, 2010), pp. 120-133 (at 120):
I do not really know Ancient Greek, nor do any of the contributors to this Companion.
To claim knowledge of a language, you must be a member of its speech-community,
open to the possibility that the categories of its grammar and vocabulary may mold
and be molded by the structures of your thoughts and worldview. This cannot happen
if we engage with the language only in a library. Knowledge of language depends on
acquaintance; knowledge by description is not enough.
This leads to an uncomfortable paradox. If I learnt enough Arabic or Chinese to
order a meal in a restaurant, and if I went to Riyadh or Beijing and did so, I would
have a better claim on that language than I have on Homer's mother tongue after
many years of daily engagement with his words.
I know only one person capable of ordering a meal in Ancient Greek. On the one occasion when he took me and his charming wife and daughters to a restaurant, however, he did not speak Ancient Greek. You know who you are, my friend.