In cryptography there's a concept called "unicity distance". There are
various ways to define it, but basically it is the number of characters of
cypher text you need before there is one and only one unique decipherment.
Take a three character word out of the middle of a ciphertext, say "WKM"
and it might be "you", or "the", or any other three-letter word. For
English Claude Shannon (possibly related?) showed that 28 consecutive
letters of English ciphertext (simple substitution cipher) can, with a high
degree of certainty, only be deciphered in one unique way. Thus, the
unicity distance of English is 28.

I imagine something very similar applies to languages. A very short
encrypted sample might make sense in many different languages, but the
longer that ciphertext sample becomes, the less likely it is to make sense
in more than one _existing_ language.

_However_, that applies only to existing languages. It does not necessarily
apply to a language created solely for the purpose of re-interpreting a
given ciphertext. Thus I could take:

"Ego vos hortor ut amicitiam omnibus rebus humanis anteponatis." Then,
engineer a new language such that it translates into English as:
"The man wanting to befriend everyone must sacrifice himself."

Obviously, the longer the text the more difficult it becomes to construct a
coherent grammar and lexicon that keeps on making sense. While finding a
second _existing_ language that fits would certainly be impossible,
creating one for that expressed purpose might well be possible.


I think it would be an extremely unlikely coincidence if the text made
> sense in two completely different languages, so unlikely that it can be
> rejected out of hand! It is this very unlikeliness which makes
> deciphering extinct written languages possible - otherwise we couldn't
> know whether the language found in which the text makes sense is the
> right one or not!