It couldn't, by the Euclidean definition of "triangle." Adding up to 180° is not a part of the explicit definition of "triangle" in Euclid's system, but it is logically deducible from that definition, together with other propositions in the system. We are justified, then, in saying "if not 180°, then not triangle" : this is analytic within the Euclidean system. But what about the actual field out there? Suppose you kept getting this peculiar result that you could no longer put down to observational error. Then you would have to say, puzzling as it might seem at first, that Euclidean geometry does not describe our actual space-that actual space is not Euclidean. The deductive system is one thing, actual physical space another. Whether physical space is obliging enough to follow the simple Euclidean system, only careful observation can tell.