So far we have been discussing constructive mechanisms almost entirely in general terms. It is now time to consider more specific examples in detail. The main principle in analysing these is firstly the idea that there are very specific skills that play a pivotal role in development and act as stepping stones to higher achievements, and secondly the concept that there should be interrelationships between skills that facilitate the construction of one skill on the basis of the possession of others.
Our first example is the process of learning to walk, where we have already discussed the potentiating skill of rising to a vertical position (itself supported by the auxiliary skill of getting support from objects). Subsequent pivotal skills include being able to stand without support, taking single steps, progressing without support and without falling over, walking to a visible object where there are no intervening obstacles, etc. These skills have the desired feature that it is in some sense a straightforward task to advance from possession of one skill to the next. For example, when the above progression has reached the stage where the child is able to assume a standing position for a moment then further processes can be invoked centred around the process of balance involved in sustaining such a situation over a period of time. Again, after the skill of balancing in a standing position has been achieved then, given the appropriate hardware, the compensations that apply to ensure balance while standing can be usefully used as components of a system to avoid falling over while walking. On this basis, initially small but gradually greater steps in the direction of expert walking can then be taken, under the control of the generative system that oversees this part of the process.
Let us now skip a few stages, and go on to consider the transition from the ability to walk without falling over to the skill of walking to a location which is visible and where there are no intervening obstacles, again a pivotal skill in that it gives the person some ability to be located in a place other than where he currently is should another location be more suitable for what he wants to do. This transition can be modelled in terms of Brooks' concept [3] of adding additional layers to an existing system to give it more advanced capabilities. A basic walking system contains no provision for making changes in direction, but this can readily be included by adding new systems to the design. Firstly there can be incorporated into the system architecture signals intended for directional control, which affect the motor centres in such a way as to cause a person standing or walking to change direction. Such movements will affect balance and so the systems responsible for balance must relearn balance again in the new situation. When this has happened, a new specialised diagnostic system can become active that is sensitive to relationships between the direction of walking and objects in the visual field. Training algorithms can then teach a system driven by appropriate visual information to generate the appropriate directional signals, the ultimate aim of such a system being that of being able to achieve a `balanced' situation where walking without applying directional control leads to a situation where the direction of the selected object does not change significantly when one approaches it (or alternatively one of controlled unbalance leading to passing on the side of an object).
How does the cognitive system develop beyond the stages discussed so far (leading to being able to walk directly to a visible location)? We explain these processes in terms of the nervous system having built-in capacities to handle particular themes, such as obstacle, gap, search, route, preference and alternative. These themes correspond to particular abstractions that can be usefully made in particular situations, and which are able to mediate further constructions of various kinds.
As an example, the route or sequence theme is one that can be invoked to record, using some symbolic representation, the sequence by which a certain situation is arrived at. The hardware that does the recording must have the feature that the sequence information can be unpacked on a later occasion and used to repeat the sequence either exactly or in some modified form, e.g. with skips or insertions.
Informally, the sequence mechanism is able to mark the route the first few times until the person has had a chance to learn to retrieve the route without having to make use of this dedicated mechanism. The mechanism thus acts as a bridge between a simpler situation and a more advanced one. One may for example first of all be in a situation A and get to a situation C via an intermediate situation B, and then later when one is at A evoke, using the sequence mechanism, the sequence ABC again, and then follow it up in actuality if the outcome C is a desired one. Bottom-up constructive mechanisms of this type combine with mechanisms that try to adapt the totality of what is learnt to the demands of higher level skills, as a result of which there is strong selection in favour of what is most useful in an individual's life.
Equally crucial is the theme of there being an obstacle or problem, an occasion for the theme being invoked being the failure of a process that normally succeeds. There may be various diagnostics specific to locating or defining a problem, and special mechanisms are required for integrating the resolution of the problem that may be found into the original activity. Again we have the situation where specialist mechanisms can form a bridge between a simpler situation and a more advanced one.
We move on now to consider a different avenue of development, that associated with symbol use. We conceive of symbols in the following terms. A symbol has power through its having a referent or referents. The kinds of constructive processes we have been referring to lead accumulatively, through an exploration of what is possible, to the symbols being related to their referents in more and more subtle ways, always consistent with the demands of utility. For example, a symbol may initially be created in response to a perceived object, and a link formed between it and the activities of the nervous system related to the actual perceiving of that object. Other links may be created with more complex situations involving the object, for example with the skill of looking for the object; in such a way the invocation of the symbol can become a means of finding the object, and then built into higher skills. Or, again, it may be used in connection with the process of anticipating an event. Further possibilities have been described in the theory of language due to Josephson and Blair [5].