Many have talked about the fact that, in Derrida's deconstructive readings, the signifier he obsesses over doesn't need to be really there. To be sure, in many of his analyses he finds a specific word whose flickering of meanings will reveal the symptoms of différance at play in the given text (think of Rousseau's supplement, or Plato's pharmakon). But that is not always the case, and some have stressed that such a privileged signifier does not need to actually feature in the text. Its importance, however, is situated by the other signifiers, so that it's silhouetted against them. It may not emerge once, but its 'presence' is secured precisely by the strategic placement of all the other words. Like this:
There is no white triangle in the picture; we know that. But its place or its right of a place is there, engendered by the other shapes. Even though there is no white triangle, the effect and function of the image depends on it as much as on the 'present' shapes.
In a way, this 'radical' technique of Derridean reading is not that crazy, since 'the silhouetted signifier' is nothing but the definition -- or the very functioning -- of the signifier (or the sign as a whole). Each signifier is precisely the space left to it by its fellows, so that its 'nature' will precisely coincide with such space, shaped as it is by what surrounds and silhouettes it.
Escher gives us a more accurate representation of the differential relationships among signs and how they shape each other. In the image above, it is not really possible to say which animal shapes which, which silhouette is silhouetted by which. Everything is a silhouette of a silhouette.