Daniel Rothschild asked me to comment on a point he makes in his paper on NPI-licensing and definites:

It seems to me that definite descriptions, if understood in the standard way as presupposing uniqueness, are strawson-downward-entailing in their restrictor.

For in any situation in which “the dog is happy” is true then in that same situation “the dog with red ears is happy” is true as long as there is a unique dog with red ears. In general, “the F” presupposes that there is just one F, whereas “the F and G” presupposes that there is just one F and G. So if both presupposition are satisfied then “the F is the F and G” will always be true.

However, NPIs are infelicitous in the restrictor of most singular definite descriptions:

?”The dog who ever went to the park was in my living room.”

It turns out that the issue with definites has been pointed out to me a number of times – in fact, the first time was just days after the final proofs of my Strawson-DE paper had gone out to the printers. In my recollection, the point was made to me independently by Philippe Schlenker and by Bernhard Schwarz. It has since appeared (in some fashion) in some critical passages by Anastasia Giannakidou as well.

What I observed in the conversations with Philippe and Bernhard was the following. Definites are not just Strawson-DE, they are also Strawson-UE: if “the dog is happy” is true and the presuppositions of “the animal is happy” are satisfied, then “the animal is happy” must be true. So, my suggestion, which I have never publicized (but probably should, together with some other remarks occasioned in particular by Giannakidou’s work), was and is that the NPI-licensing condition requires that the licenser be Strawson-DE, but properly so, i.e. not Strawson-UE at the same time. This seems only reasonable.