2.1513 On this view the picturing relation that makes it a picture also belongs to the picture.

Black (p. 85) prefers ‘picturing relation’ here to P&McG’s ‘pictorial relationship. (Ogden has ‘representing relation,’ which is also OK with Black).

Cf. my comment on 2.13. Anscombe rejects Ogden’s translation in note 1, p. 68. I think her point is that Ogden allows for the misinterpretation that, as Ramsey puts it on p. 271 of Foundations of Mathematics, the elements of the picture “are co-ordinated with the objects by the representing relation which belongs to the picture.” The picturing relation is not a relation between picture and pictured, but between elements of the picture. As Anscombe says (p. 68): “only if significant relations hold among the elements of the picture can they be correlated with objects outside so as to stand for them.” A picture needs a certain kind of coherence to be a picture, just as a sentence needs a certain kind of structure to be a sentence, nonsensical or otherwise. The correlation with something else, in order to give the picture meaning, is something we do.

And yet, in light of what Wittgenstein says next, mustn't the picturing relation belong to both the picture and its relation to what it pictures? Not everything is a picture, so some kind of property or properties must be had in order for something to be a picture, we might think. But then not every picture is a picture of this, so something else would seem to be needed to make a given picture a picture of some given thing or things. If, on the other hand, correlating a picture with something else is something that we do, then couldn't the same be true of the seemingly essential properties of a picture? Couldn't anything be a picture? So perhaps what I've quoted Anscombe as saying here is wrong.