## Figure 912

Schematic diagram of Kosslyn's computational model of imagery. Images are constructed and manipulated (using the PICTURE, ROTATE, SCAN, and TRANSFORM processes) in the highest area of resolution in the spatial medium using the information stored in image files and propositional files in long-term memory.

### The spatial medium

The spatial medium in which the duck is to be represented is modelled as a television screen in Kosslyn's computational model (see Kosslyn & Shwartz, 1977). That is, the medium has a surface that can be divided up into dots or pixels each of which can be characterised by co-ordinates indicating where a dot is on the screen. The theory mentions four properties of this spatial medium. First, that it functions as a space, in the sense that it preserves the spatial relations of the objects it represents. So, if an object is represented in the extreme top left of this space and another object in the extreme bottom left, then the relative position of the two objects will be preserved (i.e., the second object will be beneath the first object). The spatial medium is also like a physical space in that it has a limited extent and is bounded. If images move too far in any direction they will overflow the medium, like a slide projected on a screen. Finally, the space has a definite shape; while the central area of highest resolution is roughly circular, the medium becomes more oblong at the periphery.

The second main attribute of this spatial medium is that it does not necessarily represent images at a uniform resolution. Rather, at the centre of the medium, an image is represented at its highest resolution.

According to Kosslyn's theory, images are constructed in parts, so one might first form (a) a skeletal image of a duck, and then (b) add a wing part to this initial skeletal image.