Full Idea
Most accounts of the concept of mathematical truth can be identified with serving one or another of either semantic theory (matching it to ordinary language), or with epistemology (meshing with a reasonable view) - always at the expense of the other.
Gist of Idea
Mathematical truth is always compromising between ordinary language and sensible epistemology
Source
Paul Benacerraf (Mathematical Truth [1973], Intro)
Book Reference
'Philosophy of Mathematics: readings (2nd)', ed/tr. Benacerraf/Putnam [CUP 1983], p.403
A Reaction
The gist is that language pulls you towards platonism, and epistemology pulls you towards empiricism. He argues that the semantics must give ground. He's right.