Turvey Fitch & Tuller (1982) …and some additional notes.

The focus questions On page 239 - the first page - the notion of the inclusion of reactive forces and inertia in the theory of movement is first mentioned. Give a couple of examples of how these are referred to elsewhere in the paper. Why does Turvey suggest the homunculus concept must be trimmed down (or eliminated)? How do equations of constraint help with the degrees of freedom problem? Define context conditioned variability (CCV) Briefly explain each of the three types of CCV What is the alternative approach to the non-muscular forces suggested at the foot of p.249? How does Turvey characterize the use of the degrees of freedom in learning at the end of the paper?

The “straw person” argument Turvey et al present a simplified version of hierarchical control We’ll go over the most successful (?) theory, from the 70s Richard Schmidt’s Schema theory of discrete motor skill learning (1975)

Schmidt’s schema theory The schema The concept has a rich history Kant (1787) – critique of pure reason Transcendental schema Triangles – how do we know something is a triangle? Categories of things as a means to understand them The rule for the category stays the same, while the individual properties of the items within the category can vary. Triangle: Fixed Rule = 3 sides, internal angles add to 180 Variable properties = size of sides, individual angles.

Schmidt’s schema theory The schema The concept has a rich history Bartlett (1932) – “Remembering” ‘When I [play a tennis shot], I am not producing something entirely new, and I never repeat something old...’ Movements are prototypical. Each movement is novel, because it is generated from an abstraction, rather than a specific command Formation of general rules used to interpret stimuli and drive movement organization

Schmidt’s schema theory Using schema to explain movement (see Schmidt, 1975, 1988, 1991): Three concepts central to Schmidt’s schema theory Generalized motor program (GMP) Basic prototype of the movement Has variant and invariant features (just like Kant’s original idea) Recall schema Provides parameter values (variant parameters) to the GMP Built through experience Recognition schema Expected feeling of the movement

Schmidt’s schema theory Using schema to explain movement (see Schmidt, 1975, 1988, 1991): What is stored after moving? After each movement (for a short while): Initial conditions (body position, context, environment) Response specifications (direction, speed, force) Sensory consequences (during and after moving) Response outcome (desired, actual) After many repetitions, the program parameters are generated from the recall schema, which stores the relationship between: past response specifications past response outcomes past and present initial conditions

Schmidt’s schema theory Using schema to explain movement (see Schmidt, 1975, 1988, 1991): Defining features of a Generalised Motor Program Some features of the GMP were held to be variant: overall force overall timing muscle selection Some features of the GMP were held to be invariant: relative force relative timing

Schmidt’s schema theory Using schema to explain movement (see Schmidt, 1975, 1988, 1991): Example: Walking and running

Schmidt’s schema theory Using schema to explain movement (see Schmidt, 1975, 1988, 1991): Example: Handwriting Change effectors, outcome still the same

Schmidt’s schema theory Reconsider? Do either of these actually show invariant features? What exactly stays the same here from one attempt to the next? What about the bit in the middle?

Back to Turvey… Bernstein’s synergies Degrees of freedom Blacksmith example Car example Links reduce complexity But also reduce d of f to control

Back to Turvey… So synergies reduce complexity, but it’s still very complex Context conditioned variability Changes in movements arising from muscle forces forces due to context into which these forces are “injected” Homunculus must know of context to know the required force

Back to Turvey… So synergies reduce complexity, but it’s still very complex Context conditioned variability Anatomical Muscle contraction has different effect due to initial position of limb segment Mechanical Muscle force has different movement effect depending on context Kinetic energy created by movement in one joint affects others Physiological Neural signals do not descend uninterrupted – they are acted on and interpreted by the assemblies in the spinal cord. It is not a simple hierarchical process

Overall message Muscular and non-muscular forces must complement each other. Learning is about integrating non-muscular forces with muscular forces. Freezing and freeing degrees of freedom as reactive forces are learned, anticipated, and absorbed What’s important about the different perspectives?

Overall message What’s important about the different perspectives? From the schema theory perspective: Mistakes arise from incorrect calibration of recall schema Mistakes arise because the GMP is not well adapted to current consitions All central problems Central organization From the Bernstein perspective Reactive forces are not sufficiently integrated with active forces in such a way as to exploit the “free” energy of the movement E.g. walking, golf swing. Definitely a peripheral problem Listen to the periphery What is it supposed to feel like? [Hula hoop example] Self-organization Turing machine vs. the Watt governor – for next time!