The evolution of skilled action from an information processing perspectivevariability and chaos: constructive vs destructive combinations Representing skills in DNA through languageinnate skills as programs Evolution and 'inner language' as learning mechanismsevolving skills The research (framework, presentation) is divided into n stages, roughly matching evolutionary progress. Each stage presents a highly simplified but feasible model of the information processing at that stage. The emphasis is on feasibility rather than on detailed correctness.

The first stage models basic physical action with minimal integrated perception and communication. Physical action is seen as based on joint rotations caused by muscled that, in turn, receive muscle tension information at roughly 10msec intervals. An action frame is likely to cover as much as 1 second, with change and timing information for each of the planes of motion for each of the joints. It is inferred that an information specification for skills must inlude action, timing, and geometric information (joint angles).

The first stage also explores information requirements for the perception component for simple skills such as mimicking (e.g. imprinting), chasing a prey, and exploring for food. Limb orientation (or joint angles) must be perceived for mimicking. Direction must be perceived for chasing. Spontaneous variations in direction have to be generated (exploring for food).

The second stage models evolution as an information processing problem that copies information from parent to child, with variation and selection. Accordingly, DNA must encode the skill information and pass it to the brain. To have evolution, there must be variation and selection as well. Any process that exhibits these information forwarding features should suffice to produce evolution.

Mimicking is suggested as an alternative information passing process that could produce evolution. Information representing the action must be passed into, and stored by the brain - to be used by later repeatedly producing the same action. Variability can result from the mimicking copy process.

Experimentation with alternative behaviour patterns is suggested as an alternative information passing process that could produce evolution. Information representing the most successful action is selected and stored by the brain. The model works within individuals and assumes that skills can be forgotten and 'refreshed', like computer memory.

The third section explores the internal (within brain) representation of complex skills with integrated perception. A simple internal programming language is suggested. One of the advantages of such a representation is that complex skills can call on simpler skills as components. Overall a 'higher level' programming language representation of skills is a more compact representation, i.e. would require fewer bit (less space) in the brain or in DNA.

The fourth section explores inter-individual communications to explore complex cooperative tasks with division of labour etc. that require real-time two-way communication.

The fifth section explores the evolution of human spoken language as a set of stages starting with within-individual evolution, starting with babbling. It also involves the mimicry of sounds as corresponding action such as pointing. School involves more experimentation and mimicry. Skills conveyed by DNA play a rather lesser role, but may include the ability to mimic sounds, etc.

The sixth section explores evolution based on language as copying mechanism.