8. Cerebellar Model Flashcards
Can theories of cerebellar function help predict which is the site of plasticity?
David Marr
James Albus
Albus-Marr framework
models are not identical but are similar enough to be referred to as the Albus-Marr theory/framework
- trying to understand how the cerebellum is important for motor control
- start with trying to understand what motor control involves (task analysis)
task analysis
complex informational processing problems need to be modelled at more than one level
- the most abstract = computational model
(what problem is the cerebellum trying to solve?)
- second level = algorithmic level
(how does the cerebellum solve the problem?)
- third level = implementation
(how does the cerebellum implement this?)
Marrs first level
computational level - what is the problem
- use a simple task
- target appears on the retina to a degree either left or right of centre
- problem is to look accurately at the target
- how big of a command do we need to send to the muscles to look at the target wherever it is located
Marrs second level
algorithmic level - how does the cerebellum tackle this problem
- precise motor commands are learnt using supervised learning (infants spend a lot of time investigating this, looking around)
- a signal that tells us whether our movement was too big or small (error signal that corrects motor command)
- learning rule = if movement is too small, increase the command. if movement is too big, reduce the command. if accurate, do nothing.
Marrs third level
implementation - how does the cerebellum implement this?
- we have to simplify > parallel fibre signals location of the target. When parallel fibre fires, so does the Purkinje cell (motor command)
- how much the Purkinje cell fires depends on the weight of the synapse between parallel fibre and Purkinje cell
- weight of synapse adjusted by climbing fibre (error signal) - increase movement if too small, decrease movement if too big
cerebellum and motor learning
Albus-Marr
- simplified account of model trying to explain cerebellar role in learning accurate movements
- example of supervised learning (tells you whether you have done it wrong and to what degree)
- 2 plausibilities
plausibility 1
- explains why there might be so many granule cells (80% of all cells in brains)
- a lot of factors can affect the size needed for an accurate motor command (climbing fibres - music or sport)
- a lot of information is needed to get it right
= climbing fibres and Purkinje cells wrapped around each other acting as 1 large synapse (lots of info transmitted)
= contrast with mossy fibre input gives rise to parallel fibres (150’000 different synapses per Purkinje cell)
plausibility 2
- also explains why climbing fibre inout is so unusual
- you dont want error signal to drive signal directly, you want it to correct the system
- whenever a climbing fibre fires, the Purkinje cell also fires (along with the 150’000 synapses between granule cells and Purkinje cells)
- all climbing fibres affect the dendritic tree
simplifications
- role of granule cells is not well established (expansion recoding)
- learning rule is more complicated than portrayed
- Marr-Albus theory sometimes known as the decorrelation rule - current version is the adaptive filter model
relating Marr-Albus model to NMR conditioning
- initial movement is too small (eye open)
- error signal (US) needs to increase movement size (teach Purkinje cell to fire less as its inhibitory)
- information about CS (tone) arrives at cerebellum via mossy fibres
- information about the US (periorbital shock) arrives at cerebellum through climbing fibres
- before conditioning
- during conditioning
- after conditioning
before conditioning
- net effect of parallel fibres on Purkinje cell is 0 at the at the start of conditioning
- error signal is US (periorbital shock), sent by inferior olive
- CS is the tone, via mossy fibres
- CR is produced by chanhe in Purkinje cell firing
- before conditioning CS produces no response (no change in PC firing)
- possibly becaiuse inhibitory and excitatory inputs to Purkinje cell cancel out
—– CS tone comes on, parallel fibres excite the Purkinje cell dendrites whilst the collateral from stellate and basket cells inhibits it so Purkinje cell output doesnt change.
during conditioning
- error signal (US - periorbital shock) reduces the weight/efficacy of active Purkinje cell synapses (gets smaller/less effective = LTD)
- parallel fibre Purkinje cell synapse smaller (less effective) so inhibits the cerebellar nuclei less = conditioned response
- parallel fibre conditioned stimulus is now pared with climbing fibre unconditioned stimulus
- pairing produced LTD in excitatory synapses between parallel fibres and Purkinje cells
- CS predicts periorbital shock (an error) = so do it less (ERROR)
- predicted site of plasticity = synapse between granule cells and Purkinje cells
general conclusion
Marr-Albus type general models offer specific account of classical conditioning
- predicted site of plasticity = synapse between granule and Purkinje cells
