Locomotion and Manipulation (Gibson, 1979, Chapter 13)

Gibson turns his attention now to two major forms of behaviour, locomotion and manipulation, and discusses the optical basis for the control of these action. He considers the visual information available in these tasks, and how these can be used to control the behaviours, without appeal to a central controller (such as the brain). In this chapter Gibson proposes that these and other behaviour are all 'regular without being regulated' (pg. 215) and the question is, 'how can this be'? While he doesn't fully answer this question, he introduces the key optic flow features we will need for these two behaviours. 

Gibson first points out that the theory of affordances implies a deep connection between perception and behaviour; perception is for behaviour. Behaviour also enables and supports perception; we have discussed in previous chapters how the moving point of observation is the basic, necessary case for successful perception. Locomotion and manipulation are two behaviours that exemplify this 'perception-action loop' (not Gibson's term, but how we would refer to these ideas now). 

Locomotion is about moving around a cluttered environment, and depends on a suitable surface that affords walking, etc. There is visual information about such surfaces and their affordances, and there is visual information about the relationship between your feet and the surface (patterns of occlusion and de-occlusion, for example). 

Locomotion is controlled with respect to this information, as is manipulation. They are not reflexes, nor the end result of a motor programme. Instead, these behaviours emerge and unfold according to the perceived affordances of an environment. Gibson makes a big claim about the nature of this control (pg 215):

Locomotion and manipulation are neither triggered nor commanded but controlled. They are constrained, guided, or steered, and only in this sense are they ruled or governed. And they are controlled not by the brain but by information, that is, by seeing oneself in the world. Control lies in the animal-environment system. Control is by the animal in its world, the animal itself having subsystems for perceiving the environment and concurrently for getting about in it and manipulating it. The rules that govern behavior are not like laws enforced by an authority or decisions made by a commander; behavior is regular without being regulated. The question is how this can be.

Information is how this can be; it exists at the right scale (in the present, in the animal-environment system). The medium 'contains' the information for control, in a form of visual kinesthesis. In the case of locomotion, there is information for the (at least) four critical elements of successful locomotion

1. What specifies locomotion vs stasis? Flow of the ambient array over the perspective structure of the information for the environment specifies motion; nonflow specifies stasis. Specifically, flow coming out from a focus specifies approach, and inflow to a focus specifies retreat. The location of the outflow focus in the perspective structure of the information for the environment specifies heading, and the inflow focus specifies where you are coming from. Shifting the location of the outflow focus specifies a change in heading, no shift specifies no change in heading. How the flow is occluded by feet specifies locomotion on foot. (This describes the basic structure of global optic flow).
2. What specifies an obstacle or an opening? Our friend reversible occlusion comes back here. An obstacle is specified when there is a closed contour, and progressive loss of perspective structure during approach (it blocks the view). An opening is specified by progressive gain of perspective structure on approach (things come into view). Gain of structure above a horizontal contour specifies an approaching brink (that could be a step or a cliff). Gain of structure on one side of a vertical contour specifies a corner you are approaching. 
3. What specifies imminent contact with a surface?  Gibson proposes that this is specified by increasing access to finer details on a surface as it magnifies in the global optic array. At this point, he had no notion of anything like tau. 
4. What specifies the benefit or injury that lies ahead? This is where affordances come in, and these are specified by invariants of various order. The perception of affordance replaces the need for foresight, or imagination, or prediction. 

While the research as evolved in the details since 1979, the point remains: there is information in the optic array for all these things, and they can be used in the control of locomotion. Gibson now turns to some potential rules for control. He does not intend these rules to be thought of as implemented in the brain, just a verbal description of what behaviour emerges from. 

Stopping and starting is about making the array flow in particular ways. Steering is about placing the focus of expansion in different places in the perspective structure. Approaching is about magnifying the optical structure located where you are heading; but to different degrees, depending on the goal (reading vs kissing vs grasping). Keeping a safe distance is about maintaining a degree of magnification of the flow from that target; and so on. While none of these are descriptions of motor programmes, they are descriptions of information-based control strategies you can use to control locomotion to serve a goal. 

Manipulation has a similar set of rules and sources of information; more complex, perhaps, but the same kind of things. There is visual information specific to all sorts of manipulation (striking, throwing, opening, uncovering, and so on) and rules can be formulated to describe how the actions can be controlled. But this control emerges, and is not the result of reflex or central command.