Val,
Your example raises the issue that language is an acquired social construct which "works" by statistical consenus between interlocuters. Your "philosophy seminar"example which is never actually tested in real life only becomes interesting when fakirs walk across hot coals etc and here we move towards the murky waters where religion, truth and epistemology intermingle.
As Wittgenstein said "meaning is usage"..and by this I think he implied that much vacuous discussion goes on in philosophy with "fragments" out of context. ( E.g The issue of how do I know if "Peter is in pain" is pragmatically irrelevant. All that matters in real life is the function of such a statement in a particular social interaction).
The "phenomenon" is the the interaction of observer and observed, and in as much that believe observers to be "similar" with common sociolinguistic roots there will be concurrence about "external events".
This discussion has been significantly broadened elsewhere to include the so-called metalanguage of mathematics and its role in directing observation in physics. Some workers even believe that ALL that CAN be postulated WILL be observed!
LATER EDIT
I am implying here that "verification" is always functional, It is merely another form of social interaction in which "prediction and control" are covert principles. The very status of "words" themselves such as "flesh" and "burn" implies prediction of persistent properties, and re-classification takes place if counter examples arise. (Consider "alchemy" and its reliance on the "four elements - earth, air fire and water"....to be replaced the more useful (predictive) "chemistry"....to be replaced by....

)