"Someone, let's say a baby, is born; his parents call him by a certain name."

That line-recently circulated on social media by Eric Winsberg (South Florida / Cambridge) as "the funniest sentence in the history of philosophy" -- is from Saul Kripke's Naming and Necessity.

I'm not sure its the funniest sentence in the history of philosophy, but it is pure poetry.

*UPDATE: Some are suggesting Kripke's line was meant to be funny, in which case I suppose we should broaden the request for suggestions to include intentional self-parodies by analytic philosophers.

In the comments, Eric Winsberg responded:

I don't think he's joking. You have to remember this is the transcript of a talk. He's just throwing in very standard though [sic] experiment language. Compare "a number of people, say five, are tied to a railroad track". You're just signaling that the number 5 is only important in that it's more than one. But in Kripke's case, the whole point of the example is that there's nothing distinctive about the person at all except the ensuing baptism. So he's going "someone is born" and he's brain goes "we need a "say" clause, and ends up making the funniest sentence ever.

Kenny Easwaran added:

I've always interpreted this line as a clear self-parody. When he's trying to stay object-level, he has no trouble coming up with specifics, like naming his pet aardvark Napoleon, or any example involving, say, Nixon. In this case, I think he's observing the need to include "say X" afterwards, and then filling it in with the one thing that actually adds nothing, rather than "someone, let's say, the future teacher of Aristotle, is born".

FWIW, Eric Winsberg's comment about "standard thought experiment language" seems persuasive to me.

In Naming and Necessity, "let's say" is used 18 times as a rhetorical device to introduce a specific but arbitrary assumption for the purposes of the current argument, e.g. on p. 80:

Let's see if Thesis (2) is true. It seems, in some a priori way, that it's got to be true, because if you don't think that the properties you have in mind pick out anyone uniquely -- let's say they're all satisfied by two people -- then how can you say which one of them you're talking about? There seem to be no grounds for saying you're talking about the one rather than about the other.

He also uses plain "say" for the same rhetorical purpose, e.g. on p. 17:

Nor, when we regard such qualitatively identical states as (A, 6; B, 5) and (A, 5 ; B, 6) as distinct, need we suppose that A and B are qualitatively distinguishable in some other respect, say, color. On the contrary, for the purposes of the probability problem, the numerical face shown is thought of as if it were the only property of each die.

And on p. 87, he uses "let's say" to introduce an assumption arguendo that's vague nearly to the point of emptiness, probably because he didn't care to come up with anything more specific:

What's going on here? Can we rescue the theory? First, one may try and vary these descriptions -- not think of the famous achievements of a man but, let's say, of something else, and try and use that as our description. Maybe by enough futzing around someone might eventually get something out of this; however, most of the attempts that one tries are open to counterexamples or other objections.

It might still be true that Kripke was intentionally parodying this device when he said "Someone, let's say a baby, is born" -- but Eric Winsberg's hypothesis is consistent with Kripke's overall patterns of usage.