Swanston (1991).
Visual frames of reference
Figure 3.16 shows one general account, put forward by Wade and Swanston (1991), of various sorts of coding schemes that have been assumed to be present at dif- ferent levels of the visual perceptual system. We have already discussed the notion of retinal coding whereby the position of the proximal stimulus is defined in terms of a retinal co-ordinate system. It has been sug- gested that one way to think about this is in terms of the space mapped out by a simple 2D graph in terms of x and y axes. What is critical in this scheme is where these two axes intersect – the point of origin – and the points in the 2D space defined by the axes are specified in terms of x/y co-ordinates that are fixed relative to this point. In this general scheme the axes are sometimes referred to as frames of reference and
Figure 3.15 The eye has it
How to achieve cyclopean perception without having to integrate information from two eyes. The purple hair is optional though.
Source: 20th Century Fox/Everett/Rex Features.
Figure 3.16 Visual frames of reference
Schematic representation of the different frames of reference used to code information in the human visual system (from Wade and Swanston, 1991).
Source: Wade, N. J. (1996). Frames of
reference in vision. Minimally Invasive
Therapy and Allied Technologies, 5,
435–439 (fig. 1, p. 436). Reproduced with permission from Taylor & Francis Ltd.
here the axes provide the frames of reference for every point in the space.
Monocular retinocentric frames of reference
If we take this graph analogy and apply it directly to the idea of retinal coding, the centre of the eye – the fovea – offers itself as a natural point of origin. Furthermore we can also define an x axis as passing through the fovea (the horizontal midline) and a y axis intersecting the x at the fovea (the vertical midline) so that such midlines provide the retinal frames of reference for each point on the retina as defined by the individual photoreceptors. Wade and Swanston (1991) referred to this form of coding as monocular retinocentric. The implication is that such a monocular retinocentric
frame of reference provides the initial co-ordinate
system for coding the stimulus information. Import- antly, each eye contains its own retinotopic co-ordinate system (see Figure 3.17).
Cyclopean (binocular) retinocentric frames of reference
Given that we do not normally see the world through a single eye, there must be a stage at which informa- tion across the two eyes is combined to form a so- called cyclopean view of the world – we see one world, not two. Wade and Swanston (1991) referred to such
a combined representation as being coded in terms of a cyclopean retinocentric frame of reference. At this level, information from the two eyes is combined to form a binocular retinocentric representation. Here we can think of taking the left and right retinal graphs and super-imposing them over one another such that they share a common origin (see Figure 3.18). Psychologically the origin occupies a point located between the two eyes known as the cyclopean eye or the egocentre. Such a point will be located in the centre of the head if information from both eyes is weighted equally. The egocentre may, however, be offset slightly in the direction of the dominant eye. At this level, information from the two eyes is combined to form a binocular retinocentric representation. Wondering which one is your dominant eye? Very easy. With both eyes open, create a circle with both your hands and capture a small object in its bull’s eye. Now close your left and right eye in turn. Whichever eye is open when the object moves less is your domin- ant eye.
Figure 3.17 A monocular retinocentric frame of reference
The image on the left is the environmental scene. Superimposed over this is a co-ordinate system projected onto the image that specifies x and y axes whose origin is fixed at the nose. The image on the right is supposed to convey the retinal image – the image is inverted because the lens of the eye projects an inverted image. In this case, the frame of reference preserves the spatial nature of the dimensions in the scene but the x and y co-ordinates are defined relative to an origin specified on the viewer’s retina. Therefore co-ordinates in the right image provide a monocular retinocentric frame of reference. Each eye of the viewer possesses its own monocular retinocentric frame of reference. The scene-based co-ordinates are fixed on an origin (the nose in this case) in the scene but the retinal co-ordinates are fixed on an origin in the retina, for instance on the fovea.
Figure 3.18 Achieving a cyclopean view of the world
Information is combined from the two eyes so that information contained in two separate monocular retinocentric frames of reference are then registered in a single binocular co-ordinate frame (see text for further details). The origin of this co-ordinate system is now in terms of some form of cyclopean frame of reference.
Egocentric (head-based) frames of reference
If the eyes remain still, then any movement in the visual field will be coded relative to the egocentre. However, whenever the eyes move this must be taken into account. It would be catastrophic if every time we moved our eyes the world were seen to move in the opposite direction. In fact, if you keep your head still for a minute and look around, it’s pretty amazing exactly how stable things appear despite the eye movements. Wade and Swanston (1991) therefore discussed the egocentric frame of reference. At this level, the binocular retinocentric re- presentation is combined with information about eye movements so that the movement of the eye in the head can be discounted. In this regard, the egocentric frame
of reference is also known as a head-based frame of
reference. At this level it is as if there is a single eye in
the centre of the head that moves with every head move- ment, so although we may not much look like a Cyclops, the visual system is certainly cyclopean in nature. This sort of co-ordinate system codes the relative distance of objects in the environment relative to the observer’s head independently of where the eyes are pointing.
Geocentric (scene-based) frames of reference
The final frame of reference Wade and Swanston (1991) discussed is the geocentric frame of reference.
So far we have discussed (i) the retinocentric frame that codes information in term of retinal co-ordinates but this will change every time the eyes move, and (ii) the egocentric frame that codes information in terms of a head-based co-ordinate system but this will need to be updated every time the head moves. The geo- centric frame codes information about the disposition of objects in the environment independently of move- ment of the observer – this has been alternatively termed a scene-based frame of reference (Hinton & Parsons, 1988). This level of representation is the most stable of the three because it remains fixed when the observer moves, the objects in the visible environment move or both move. ‘ See ‘What have we learnt?’,
below.
‘What have we learnt?
The main reason in discussing the various putative frames of reference in vision at this juncture is that this framework is useful for thinking about the nature of iconic storage – just what sort of co-ordinate system captures the icon? Aside from this, the framework comes in useful when we discuss visual object recog- nition (see Chapter 13). So when we get round to discussing visual object recognition you will already be in a position to appreciate some of the issues. For example, what sorts of co-ordinate systems are used when internal representations of objects are derived? Nonetheless, it should now be clear that it is quite reasonable to discuss the icon as preserving informa- tion in terms of a retinal co-ordinate system – this codes the position of the stimulus elements in the target display – without also committing ourselves to the idea that iconic memory necessarily resides in the retinae (Wade & Swanston, 1991, p. 91).
In this regard therefore the abstract framework for thinking put forward by Wade and Swanston (1991) provides a very useful means to try to under- stand the detailed nature of iconic representation. We have accepted that the icon preserves the visual characteristic of the stimulus (e.g., red items, large items, items in a particular position in the display, etc.), but we have yet to understand fully what sort of co-ordinate system this information is captured in. For instance, one plausible suggestion is that iconic storage resides at a level at which informa- tion is integrated across the two eyes and is there- fore a form of binocular retinocentric representation as described by Wade and Swanston (1991). Fur- ther evidence that bolsters this claim resides in the literature on visual masking, and it is to this that discussion now returns.
Pinpoint question 3.7
According to Wade and Swanston (1991), what are the four visual frames of reference?
retinal co-ordinate system Defining retinal positions in terms of a 2D co-ordinate system. For example, using x (left/right) and y (up/down) values with the origin fixed at the fovea. Such codings are preserved at some of the higher levels of the visual system. It is possible to find retinotopic maps in the cortex.
visual frames of reference After Wade and Swanston (1991), a collection of coding schemes that exist at different levels of the visual system. Each codes the spatial disposition of the parts of a stimulus in its own co-ordinates such as positions relative to the fovea – as in a retinal frame of reference.
monocular retinocentric frame of reference Spatial co-ordinates derived within an eye and defined in terms of some form of retinal co-ordinate system.
cyclopean retinocentric frame of reference Spatial co-ordinates derived from combined information from both eyes.
binocular retinocentric representation A representation of the visual world derived from both eyes but which preserves locational information in terms of retinal co-ordinates.
egocentric frame of reference A head-based frame of reference used by the visual system to specify object positions relative to the viewer’s head.
geocentric frame of reference A scene-based frame of reference in vision which takes into account both body-centred and scene-based attributes.
Research focus 3.2