Full Idea
Principles of implication imply there is not a purely probabilistic rule of acceptance for belief. Otherwise one might accept P and Q, without accepting their conjunction, if the conjuncts have a high probability, but the conjunction doesn't.
Gist of Idea
I might accept P and Q as likely, but reject P-and-Q as unlikely
Source
Gilbert Harman ((Nonsolipsistic) Conceptual Role Semantics [1987], 12.2.2)
Book Reference
Harman,Gilbert: 'Reasoning Meaning and Mind' [OUP 1999], p.213
A Reaction
[Idea from Scott Soames] I am told that my friend A has just won a very big lottery prize, and am then told that my friend B has also won a very big lottery prize. The conjunction seems less believable; I begin to suspect a conspiracy.