In Stecker and Gracyk, Aesthetics Today (2010) 

This document is a summary of Wollehim. 
My personal comments are in red. 

(The background to this discussion is Ludwig Wittgenstein's influential discussion of "seeing as." Wittgenstein discussed the case of the duck-rabbit figure, which we can see as a duck, or see as a rabbit, but not both at the same time.)

 In earlier writings, Wollheim used the phrase "seeing-as" to ALSO cover the case of seeing who is pictured in an ordinary pictured. Wollheim now thinks that this approach was too crude. We can see the duck-rabbit as a duck, but we also see the duck in the picture.

KEY IDEA: Our ability to see pictures is due to a general psychological ability of seeing x in y, which is not the same as seeing x as y. Seeing-in is REPRESENTATIONAL SEEING. Seeing-as isn't.

Three reasons to make the distinction when discussing pictures and paintings:

• Seeing-as is restricted to seeing a particular object as being an object of a certain kind. Seeing-in also allows us to see a more general situation or a scene (what Wollheim calls "a state-of-affairs").  (Is the sky an object? If it's not, then you cannot see a blue patch of a painting "as" the sky. Instead, you see the sky "in" the painting.)

• Related to the first point: When I see x as y, I see some particular real feature of x as a particular feature of y, and can point to the specific place. But some of what I see in a representation is not localized, as when I see a gathering storm or that an event is taking place. (It may be obvious to say this, but an unchanging visual object is not event. So if an artwork shows an event, the event is seen in the artwork, but the artwork is not seen as the event.)

• The TWO-FOLD THESIS: Seeing-in requires one to attend simultaneously to object (what is "in" the picture) and medium (to the physical stuff of the picture, arranged in a particular way). Seeing-as does not permit this simultaneous attention.

With seeing-in, the intentions of the creator give us a standard for correctness. (This is basically Iseminger's point that I need the intentions of the creator to remove ambiguity.)

Arguments in favor of the two-fold thesis:

• We can change our viewing point on a pictorial representation without changing our perception of how perspective operates in the picture. The same is not true when looking at a real object of that sort. The difference is that the SURFACE properties of the representation are controlling what we see in it.

• When we evaluate pictorial representation, we generally pay attention to how the artist accomplishes it (how the manipulation of the medium leads to the picture being what it is).

This is a distinction between two kinds of perception, not two kinds of object. In some cases, a single object can facilitate both seeing-as and seeing-in (e.g., clouds). (The duck rabbit!)

Interesting case: Jasper Johns’ flag paintings. We can see them as a flag! So is it a representation at all? Of course, it is intentionally raising this issue.