"we as spectators are not in the world of the play itself; we — in a sense — see what is happening in that world, but not in the same sense as that in which we see the actors" (36).1) So Williams distinguishes two senses of seeing what's happening, two senses, that is, of observing the events on stage. We (the audience) who are outside the fictional world witness those events in a different way than the characters, who are inside.
For instance, Othello does strangle Desdemona in Shakespeare's play, and there is a sense in which we (the audience) witness that murderous act: we watch Othello strangle Desdemona when act V has arrived. And yet we don't watch the actor who plays Othello strangle the actress playing Desdemona, for he doesn't: he's just acting. We can say that we watch that actor pretending to perform a jealous murder, and we can say that we watch Othello murdering; and for many purposes, there's not really a difference. But still we're talking about different things.
Williams makes this point in terms of formal identity:
"[the audience] would not be seeing Othello unless they were seeing Sir Laurence [Olivier] or another real man moving around [...]. But we must not say that the reason why, in seeing Sir Laurence, they see Othello, is that Sir Laurence is Othello, at least if that 'is' is the 'is' of identity. For if Sir Laurence is Othello, then Miss Maggie Smith, or whoever, is Desdemona, and since Othello strangles Desdemona, it would follow that Sir Laurence strangles Miss Smith, which is false. [...] I see Othello strangle Desdemona; but that will not entail that I, as part of my biography, have ever seen anyone strangle anyone." (34, 35)In other words, although what these actors do constitute the events in the world of the play, there are still two different things to be observed: the playing, and what's played. And what goes for the events also goes for the physical items, such as props and scenery: "when in a play someone sets fire to the palace, they do not, hopefully, set fire to the scenery." (35)
2) One consequence is that there are, strictly speaking, no spatial relations between the audience and the characters, although there are spatial relations between the audience and the actors (or the scenery). Again, "the audience at such a play are spectators of a world they are not in. They see what they may well describe as, say, Othello in front of a certain palace in Venice; [...] But they are not themselves at any specifiable distance from that palace; unlike Othello, who may be (thus he may be just about to enter it)." (35)
However (and this is where it gets interesting), there is still something like a 'point of view', a perspective from which the whole thing is observed. ('The whole thing' meaning here the world of the play, including the scenery, props, and actors moving around and doing whatever they need to do to constitute the actions of their characters.) When the audience sees the palace, "they are presented with [...] a certain view of that palace, e.g. a view of its front." (35)
Note first, then, that this perspective is not simply constituted by spatial relations. A point of view is not simply equivalent to 'looking from a given spatial direction'. For who's looking here? It's not the characters. There could easily be a scene in which none of the characters looks at the palace, and it would still be there, as seen from a certain point of view. So it must be the point of view of the audience. But once more, "we are, as spectators, at a certain distance from the scenery and the actors, but not from the palace or from Othello" (36). It's not the spatial relationship from which the perspective results.
This becomes even clearer when we switch from stage examples to film, where the perspective can be from any point in space, and typically will even move around: the point of view is now that of the camera, and thus no longer fixed by the spatial location of the audience in their theatre seats. While in a stage play there is only so much possibility of having the fictional world presented from various locations in space, there is an infinity of such possibilities in a movie. (Even though there is still the same basic setup of an audience sitting on seats in rows facing a fixed screen of certain dimensions. That very setup has now lost even the small power it had over point of view with theatre audiences. Of course, with this additional freedom comes also loss: namely, there is no longer the direct physical presence of the actors, which marks one of the primary differences between film and theatre.)
Thus for the worlds of movies holds what holds for the worlds of plays: we're not in those worlds. We're not looking at them in the way an inhabitant of that world would look at them. The point of view from which we watch isn't one from within that world.
"We cannot say [...] that it is our point of view: for we are not, in the usual case, invited to have the feeling that we are near to this castle, floating towards its top, or stealing around those lovers, peering minutely at them. [...] One thing, in the general run, is certain: we are not there. Nor, again, can we say in any simple way that this point of view is the director's [...], since we are no more invited to think of Griffith or Antonioni floating up towers or creeping around lovers. [...] In the standard case, it is not anybody's point of view. Yet we see the characters and action from that point of view". (36–37)3) I have extracted this line of thought from Williams' article partly because it is such a nice illustration for what I mean by the metaphysical apartness of fictional worlds. But there's at least one more interesting aspect to it. Williams' guiding question is whether we can visualize unseen things; the line of example is intended to show that in a stage play or movie, things can happen 'unseen', in the non-trivial sense "in which the playwright can provide the direction 'Enter First Murderer unobserved', and yet still consistently hope that his piece will have an audience, an audience who will indeed see this unobserved murderer" (36). The fact that we visualize things from some point of view, as if we did perceive them from that point of view, doesn't mean that there must be someone within the visualized world perceiving things from that point of view (37).
Perspective doesn't imply an act of perceiving within the world unto which it is a perspective. (That's the idea that Williams uses for his attack on Berkeley.) Another way to put this is that there can be perspectives on instances of unreality and metaphysical apartness can still hold.
 Bernard Williams, "Imagination and the self", in: Problems of the self. Philosophical Papers 1956–1972, Cambridge: Cambridge University Press 1973, 26–45, quoted by page in the text.