- The person is interested, has studied a bit of linguistics at uni, and tells you what they found most interesting about it. This response is rare.
- The person asks you what linguistics is. This response is good, because you get to tell someone what linguistics is and the world is a slightly better place.
- The person looks bored/scared/turned off and moves away. This is an undesirable result.
- The person gives you well-meaning but useless information or asks you a well-meaning but irrelevant question about something that isn't linguistics. This is quite nice, really, if not that helpful, but it's hard to get them back on the right track after they've put that effort in.
- The person asks you about split infinitives, semi-colons or ending sentences with prepositions. You then either a) tell them that linguists don't give a monkey's about that kind of thing and they go away all hurt, or b) give in and tell them the answer and perpetuate their misconception.
- The person asks you how many languages you speak. To this the correct reply is:
Would you ask a doctor how many diseases they have?It's what I like to say, though I don't know who originally came up with it.
Anyway, Speculative Grammarian, the satirical linguistics magazine (yes, there is one) has a nice article about linguists at cocktail parties, in which this crops up. It shows, by means of statistics, that the diseases line plays the best. There's an alternative which is nearly as good but not as reliable, but that I like:
Would you ask a geologist how many rocks they have in their head?And it also shows that you shouldn't conduct surveys at cocktail parties. As the article notes,
Taking surveys at cocktail parties is not a good way to improve the perception and reputation of our field or ourselves
— that must be why syntacticians only ask themselves and their indentured- servant grad students for data.