Lecture 1

Types of decisions
 Perceptual choices (objective judgement about physical aspects)
 Value-based choices (which one do I like more?)
 But even for value-based decisions, there is a perceptual-based decision
that labels the two choices – categorize sensory input
 Perceptual choice is useful because it allows us to study the foundations
of decision making
Perceptual decisions
 2x2 tables common
 Different outcomes are associated with different costs and benefits
 False positives and false negatives
 Simple perceptual responses can lean to serious consequences – e.g soviet
officer
Psychophysics
 Investigating the relationship between what’s out their and perceptual
responses
 Particularly interested in errors
 You can have Hits, Misses, False alarms, and correct rejections
 Noise in the nervous system – we don’t care where it’s from, we’re just
going to try and model it. The brain is a noisy place.
 Neuronal responses can be the same for no and low contrast, leading to
incorrect decision making
 System has to place some bound, separating into two parts – “criterion” –
observer places on internal evidence axis to separate absent from present
stimuli
 Criterion – “threshold””
Signal detection theory (SDT)
 Bayesian graphical model
 You have two inputs to a single internal variable (noise and world state)
 Analogous to dice game
 At what point would you go from 0 to 3 (noise to signal?)
 The number is your criterion
 Likelihood ratio is one in the middle (where they overlap)
 Log of the likelihood ratio specifies where we should set the criterion
 Do “intuitions” get better for mathematical ability?
 Dprime is the extent to which the curves are separated – the sensitivity of
the system. (e.g. if the trick dice has a value of 10, the curves would be
very separated)
 Dprime is independent of where you put your criterion
 Dprime = difference between the hit and false alarm rate
 C = the sum of hit and false alarm rates – proportional to the area to the
right of the criterion.
 Negative values of the criterion means that the response is more “liberal”
– more likely to say yes
ROC analysis
  If we have multiple criterion points, can get multiple pairs of hit and false
alarm rates. If you plot these rates you get an ROC curve.
 Receiver operated characteristic – smooth curve. Area under the curve is
also proportional to dprime.
 Need to look at the trade-off between false alarms and hits.
 Non-perceptual factors may affect a subject’s tendency to say seen or
unseen – what individuals report is determined by the criterion they have
placed
 Our top-down expectations also affect where we put our criterion.
 The higher the beta, the more conservative. Rule: say “yes” when
probability likelihood ratio is greater than beta.
We don’t just get snapshots – we have evidence over time
 Random dot motion task – are the dots going to the left or right. Can vary
the coherence (proportion of dots moving together).
 You can’t do this with a single snapshot – you need to integrate.
Performance increases as a function of duration. People have lower
thresholds as duration increases.
Model:
 Compute overall likelihood of all those samples coming from signal or
noise
 Log is even more useful over time.
 Peirce – we should sum up our beliefs if we assume belief is equivalent to
the log likelihood ratio
Sequential probability ratio test
Drift diffusion model
 Problem of SPRT is that it assumes we have knowledge of distributions
 Strength of evidence can be a parameter like Dprime
Lecture 2 – a neural basis of perceptual decision-making
 MT/V5 has cells that are sensitive to motion
 In monkey literature use MT, in human lit use V5
 Parietal lobe also associated with motion
 Cells in visual area MT respond to moving bars of light in directionally
selective fashion
 Different neurons are tuned to different directions of motion
 Measured one neuron, found its maximum preference, and increased
coherence of the dots moving. As coherence increased, response
increased
 Can take firing of neuron and plot “neurometric” function – very similar to
actual performance of animal
 Threshold = arbitrary level – 4% coherence needed.
 Similar thresholds for neurons and animals (whole neural population)
 Why can we not take advantage of the neurons that are better?
 All neurons are correlated with each other because they are all
interconnected locally – they don’t provide animals with independent
samples
o This is why a single neuron can do better than many other
neurons, but performance is not as good as that neuron
 Area LIP cells respond when animal directs attention towards a stimulus
in its receptive field. Will also fire when directing your eye to that space.
 Receptive field of LIP neuron will accelerate or decelerate depending on
motion strength, and will predict where the animal looks.
 MT tells LIP how much evidence there is. But LIP represents choice by
integrating MT.
 Can align stimulus to decision rather than stimulus onset. All the traces
get to the same point – firing rates reach same the same point.
 If you stimulate MT neurons during the presentation of a signal, you get
more behavioural evidence of moving towards the right
How do we study this in humans?
 EEG – keep the dots on the screen the whole time, then every now and
then change the coherence.
 Align response
 CPP – centro-parietal positivity
 fMRI – cannot separate individual neurons attuned to different regions
 But can look at regions that are attuned to faces because these are
topographically mapped
 Integrating mixed face/house info
 What is the equivalent of LIP? Looked for regions that correlated with
difference in evidence.
 Left prefrontal cortex. But also had to mask. Easy should hit the bound
faster – but the problem with the reasoning is that you can’t attribute
signals in fMRI to increases in cell firing for insividual cells. You might
actually predict bigger BOLD signal when choice is harder.
Are decisions made independently of motor plans?
 We don’t need to make a particular motor response – we can abstractly
think “the dots are moving to the right” . In one condition you do a button
 press, in another you make an eye movement There are regions that care
about motor response (red regions) and there are regions that care more
about the button press. But can we identify regions that care about how
easy the decision is independently of whether we are making a button
press or an eye movement?
 Compared high vs low coherence motion. Did analysis separately for the
two tasks. Is there an area that shows increased activity for easy vs hard
decisions independently of task? Identified dorsal frontal region.
How is the criterion/bound set?
 Tradeoff between being speed and accuracy?
 People with greater change in activation in medial prefrontal cortex also
showed greatest changes in strategy from going from accuracy to speed.
 Another type of criterion shift – cost and benefit. (e.g. should we avoid
false alarms or avoid misses?)
 Can we instruct people to change their strategy when categorizing houses
and faces. (e.g. greater cost of saying face incorrectly than house
incorrectly. Changing decisions but not dprime (sensitivity to stimulus)
Frontalstriatal networks
Lecture 3: Confidence

We often make decisions in the face of uncertainty. How do we measure

confidence?

Type 1 decision:
 Is this a face or a house? First order decision about the outside world
Type 2 decision:
 Decision about the type 1 decision – how confident am I? Did I make an
error?
 Can start building up a science of confidence
Peirce & Jastrow:
 Tried to come up with formula to predict peoples confidence
 P = proportion correct
 Proposed confidence scale with performance value
 BUT this just redescribes the problem. If it knows performance/how well
it’s doing, why compute confidence?
 Direct translation model: confidence = f(performance)
 Confidence may dissociate from performance; this is difficult to compute
Extending SDT to accommodate confidence
 Add additional thresholds to accommodate varying amounts of
confidence. So high confidence is above a certain threshold above/below
a certain value on the x axis
 Distance from criterion is measure/proxy for confidence
 Would be the absolute value of that difference (x-c)
Look for tell-tale signs such as the x pattern
 When evidence strength goes up, confidence goes up
 But confidence should go down when you are wrong
 As the trial gets easier, when you’re correct you get more confidence, but
when you’re wrong you get less confident
Neural representations of confidence
 The X goes both ways because as odour mixture goes below 50 it also gets
easier
 Ask rat to make a gamble in the confidence of their decision. The amount
of time they’re willing to wait for a reward. If it is sure it was correct it
should wait for the reward. Whereas if the rat is not confident, it should
just give up because the reward might not come.
 Inactivated the OFC using muscimol
 This did not affect accuracy of decisions, but when OFC is inactivated,
confidence. Altered coupling between type 1 performance and confidence.
Damage in prefrontal cortex in humans damages confidence estimation.
There is still a similar pattern, so hasn’t destroyed inability to do
anything. But when you contrast they are statistically significant
 Log likelihood ratio should provide everything we need. But we need to
put some bounds up. And also we don’t really know what these
distributions are.
 It’s hard to read out confidence from LIP evidence because they always
reach this constant bound
  Maybe we have two accumulating populations and the difference between
the accumulators is a measure of confidence.
Confidence signals in LIP
 Sure target always gives small reward
 Balance of evidence between the left and right tracks confidence levels
Post-decisional processing
 Reflecting on decision after it has been made
 Error monitoring – when you realize after the decision has been made
that you made this error
Difference between sensory certainty and confidence
 Can change coherence of dots
 Or distance between line and the direction of motion
 Regions in visual cortex and parietal lobe were tracking whether the dots
were easy or hard to see. IPS is a human homologue of LIP in monkeys
 Suggests that these regions might be tracking certainty of sensory input
rather than the confidence of the decision that is made
 Regions that track confidence level: medial PFC. Tracks how easy the
decision is (how far the dots are to the reference point).
 This study reconciles previous finding. Dissociation between those that
track sensory certainty and those that track confidence
How coupled are they to a particular sensory modality? Or is it encoded in an
abstract format?
 De Gardelle et al (2014) – subjects asked to make confidence judgements
about two different tasks. Deciding whether stimulus was tilted one way
or another or spatial frequencies (high or low).
 Psychometric function plotted. Signal strength vs proportion correct.
Make fewer errors when they’re confident. Measure slope as function of
how mnay errors made. Look at how slope changed as confidence
changed. People are just as good at knowing when – confidence is
abstract. In a common currency. Visual and auditory tasks the same.
 Another evidence: when you have to judge colour of stimulus (more red
or blue) or judge letter (more Xs or Os). How does confidence of task A
affect confidence on task B? If people are more confident about colour
judgement, more likely to be confident in the letter task even though
these should in theory be completely independent. So confidence may be
coded in a more abstract way
Participants asked to carry out perceptual decisions and then rate confidence.
 In another task, have to remember items they ere shown earlier.
 Confidence about perception and confidence about memory
 Participants brain scanned.
 Crossclassify? If abstract, should be able to cross-classify between the
tasks. If not, then shouldn’t
 Had participants as controls that just pressed button
 See medial PFC activated.
Lecture 4

Interface between CNS and PNS – getting signal from the brain to the spinal cord
to the muscle

Different methods to study neurons:

 Single-unit recording – place electrode to particular location in the brain
and measure electrical activity in the neuron. Neural coding
 Stimulation (e.g. TMS) – drive neurons. By applying electrical currents
(action potential). Magnetic stimulation can be done from the outside of
the head,
 Lesion studies - causal
 EEG – electrodes on scalp. Firing of areas of cortical areas on scalp which
respond to brain activity underneath
 Neuroimaging
What are actions?
 Muscle contractions
 Action – defines all behaviour
 We have a brain so we can make better action
 E.g. creature that eats its own brain when it goes from juvenile to adult
phase
 The only we way we know what we’re thinking is by what we do?
 2 ways to study action: that actions are based on cognitive processes or
neuroanatomical pathways
 5th addition – The cognitive neurosciences
 Principles of neural science
We normally act without any insight into what we are doing
 Programme robot to make an action – you have to explicitly tell the robot
everything to do. But we have very little awareness about these actions
 Things we need to do action: motivation (reason/interest), volition
(want to do), planning (how will I do this? – we think about what kind of
grasp to use – a robot will have to be told exactly which joints), for all the
possible actions we can make, there are an infinite number of movements
which will get us to the action goal.
 We don’t exhaustively explore all possible options. Time consuming –
infinite number of possibilities! Sometimes the cost function is really
shallow – ends are more important than the means – maybe there are two
movements that are both equally optimal.
 The problem about working out the infinite number of actions is not as
straightforward – the selection/inverse problem.
 Must solve the planning (selection) problem before we move.
o How? Familiarity? Doing what has worked before?
 Initiation – from the CNS into the PNS into muscle. Send signal from brain
to body. Primary motor cortex sends primary motor command which
initiates contraction of the muscle and then movement begins.
 Execution – limb movement (reaching to, grasping) – got objective
physical event in the world – body is moving.
  Monitoring – how am I doing now? Need to monitor whether it is moving
as planned. We monitor progress of the action during execution. If action
is failing to meet the goal then it will get corrected – will replan.
 Stopping – hardly discussed in literature. How do we deactivate
systems/actions? Syndromes following lesions in frontal lobe which
causes perseveration – person will do it again and again.
 Before you move – going through series of sequential stages that work in
a feedforward manner.
 Red bits (volition, planning, initiation): are serial, hierarchical,
feedforward, and discrete
 Blue bits (execution, monitoring, stopping): continuous, iterative,
feedback-based
 The red bits are trickier and more resource consuming
Computational model of action control

 Goal-directed action
 Begin with a goal/intention (e.g. to get hand to coffee cup)
 Planner/inverse model – plans how you are going to get to the action.
Takes you from an end to a means. Produces a set of motor commands
which you can send to muscles.
 Take goal as input and produce motor command as output.
 Once you have motor command and send to muscles, limb will begin to
move. Receive sensory feedback that limb is moving. We know something
about the progress of the movement. Can then bring this
 At the junction you compare goal with what is happening by sensory
feedback. If there is a 0 output at comparator – then goal and action
matches and you get 0 output.
 What’s the problem with getting sensory feedback after the action has
occurred?
 But if you get feedback after the action is performed, then it takes time for
the sensory info to get back to the brain and you update. Feedback is
always delayed and you’re reacting to what has already happened.
 The brain solves problem of delayed sensory feedback by the forward
model. 20 ms for signal to reach the brain. But the older you get the
longer it takes because of myelin decay. But this is already a problem!
Instead, the brain uses internal simulation called forward model –
predictive model. Takes a copy (efference copy) of the motor command
before it leaves the brain and makes a prediction as to how the arm is
going to move and makes a guess as to whether it needs to adjust. If the
model predicts a failure, adjusts movement. We think this loop is in the
cerebellum, and the loop takes 12 ms.
Lecture 5 – Motor Pathways

How muscles work

 Nerve comes into muscle and makes special synapse on muscle fibre.
 When action potential reaches neuromuscular junction – muscle will
contract
 Try to keep brain free for the “interesting stuff” – delegate route in stuff to
the lowest level of pathway. A lot of the details of how we move are
fleshed out by the spinal cord mechanisms.
Reflexes – very simple units that are primarily specified by the spinal cord
 Provide basic things that you have to have in order to make useful
movements
 In muscles, there are fibres that contract and fibres that sense. So when
you are wacked in the kneecap it pulls on the tendon which stretches the
muscle. The spindle fires an action potential which sends an afferent
(sensory) peripheral nerve signal which will enter the spinal cord through
the dorsal route and makes a monosynaptic connection back to the motor
neuron which makes the same muscle contract.
 E.g. if you pour water into your coffee cup, the muscle gets stretched more
and more, and sends signal to spinal cord, which causes your muscle to
contract more.
 Spinal circuit compensatorily contracts to keep the same posture
maintained.
 Lowest level of motor pathway is not in the brain but in the spinal cord –
sensing what’s going on and compensating for it to keep things the same.
This is how we stand up! Losing reflex tone results in problems with
posture.
 If you just need to keep still, your brain doesn’t need to be involved – your
spinal cord is what deals with all the low level management of keeping
things the same.
 Reflexes are conservative
 Reflexes tend to be stereotypes – always doing the same thing – but that’s
not the way we move – we do things a little bit differently each time.
 Basic oscillatory movements that are involved walking/swimming,
rhythmic repetitive movements also seem to be spinally led
Central pattern generator – circuit of neurons within spinal cord that can
generate rhythmic behaviour. So you can break a chicken’s neck and it can still
run. Routine activities like walking don’t need the brain
 Lamprey – swims by moving body from side to side. Each time it wiggles
gets a different movement of stretch. Wiggling a lamprey one way is
enough to start it wiggling the other way because of the muscle stretch
which stimulates the muscle on the opposite side. Can begin to swim even
if you cut its head off!
 Same with a cat – if you put a cat on a treadmill, it will begin to walk even
if the brain is separated from spinal cord. Signal that muscles are moving
is enough to set off central pattern generator
As you go up in the pathway, involved with less repetitive movements – more
like initiating a new action, doing something different
Lecture 5: Primary motor cortex
Lower motor neuron – from spinal cord to muscle

Pathway where neurons whose cell bodies are in the motor cortex send axons
down through internal capsule, crossover in the medulla/brainstem (at the top
of the spinal cord), and innovate the motor neurons on opposite side of the body.
Could fire action potential in left motor cortex which will synapse muscle neuron
on right motor neuron. This pathway is called the lateral corticospinal tract
(CST)

CST
Tract is the name for the bundle of fibres which have their cell bodies in the
primary motor cortex and leave it going down to the brain stem and muscle

Function – lesion study

Lawrence and kuypers (1968) – lesioned pyramidal tract bilaterally in medulla

because that is where the fibres leave the region to go to the spinal cord, and
easy to get to as they’re in the back of the head. Essentially lesioning connection
from motor cortex to spinal cord. But the monkey could not recover the ability to
make fine finger movements – it never came back.
 Normal monkey can make precision grip to pick up peanut hidden in table
 Monkey after lesion had clumsy control, using whole hand and unable to
oppose a single finger and thumb.
 Lacks dexterity of normal monkey
 One function of motor connection to spinal cord is dexterity of fine finger
movement
 Finger and facial movements are much more sophisticated than our other
movements eg leg movements.
Strick
 Injected rabies virus in muscle, which is retrogradedly transported back
along the axon into the cell body
 Realized that the time taken for the rabies cirus to travel back along the
axon to the cell body is quite long. The virus will cross the synapse then
retrogradedly transport up to the next cell.
 Takes about 24 hours to cross from one neuron to the neuron upstream of
it.
 Viral equivalent to reverse engineering – inject virus into muscle, 24 hrs
later 1 step up into pathway, 24 hrs later another step up.
 Tricky part is need to sacrifice the animal
 Sacrifice monkey at a particular number of hours after the injection
 After 24 hours, find virus in motor neuron that runs from spinal cord to
muscle.
 It takes one more time-step to get to the CST, which suggests there is one
more synapse
 If sacrificed after 2 time steps, found that some of the cells that were only
one synapse away from the primary motor cortex.
  LEARN THIS STUDY IN DETAIL
 Thre are two pathways – a direct pathway and an indirect pathway that
makes an additional synapse at the interneuron

Primary motor cortex
 Just in front of the central sulcus
 There is a bit that is in the sulcus (in the fold)
 Neurons that were labelled after two time steps were primarily deep in
the sulcus
 Neurons that were labelled after a further time step were on the gyrus
 You have different ways of accessing the motor neuron
 New M1 – humans and monkeys have it, cats probably don’t have it, rats
definitely don’t have it. It is a direct path to the Motor neuron from the
brain which bypasses the interneuron
 Through the evolution of the cerebral cortex, you have more and more
direct pathway for fine motor control
 Also the red areas require far less stimulation to get it to move –
hypersensitive
 Certain part of sulcal M1 is directly associated with fine finger
movements that you don’t get when injected into elbow muscle.
 Not only is this pathway direct, but it is also hypersensitive
 White spots – need to stimulate more anteriorally and need to put more
electricity in order to get muscle to move (you can tell based on the
smaller dots)
 First movement foetuses make – thumb in mouth – hand and face are next
to each other in the primary motor cortex
Penfield homunculus
Lecture 6 – Motor cortex (cont.)

Penfield homunculus
The cerebral cortex is a layered structure – lots of cells in layer 4
 Layer 5 – output layer – cell bodies send axons out of cerebral cortex to
control muscles on opposite side of the body
 Corticospinal tract depends on cells which have cell bodies in layer 5
 Both the pre and post central gyrus contain somatotopic mapping
 Layer 4 is almost absent in the motor cortex (precentral)
 Layer 5 is characterised by cell bodies of Betz cells – one of cells which
contributes to corticospinal tract with big cell bodies. These are only in
the precentral cortex and completely absent in somatosensory cortex
(postcentral gyrus)
 Axons of betz cells god down into white matter and down corticospinal
tract
 Image A – each roman numeral corresponds to a finger digit movement.
You can get thumb movement n lots of different regions of the primary
motor cortex. It is randomly intermixed – and even if you stimulate
behind the central sulcus, you can still get movement in the fingers. This
may be because if you stimulate a cell, you stimulate the cell that is
immediately there but you also stimulate the cells that it is connected to.
Maybe there are also some direct output neurons in the somatosensory
cortex
 B – attempt to draw boundary
 What is actually the wiring diagram? Is it like a marionette? It can’t be
simply like that because the organization is much more fractured
Monkey study
 Electrode going down through layer 5 parallel to the central sulcus – you
can get movement of different digits by stimulating from many places –
not one-to-one mapping
 One-to-one mapping – no redundancy, which would leave animal unable
to move muscle if one are of the brain became damaged. Other wiring
schemes are more robust. Many-to-one is robust because you can damage
without the muscle being compromised. But not very efficient because it
requires space
 Most likely many to many mapping which project to various muscle
groups
 You use the same muscle to do many things
 So maybe you have multiple representations of the same target because
you want to do multiple tasks. Which is why you get neurons projecting to
the index finger because it allows your fingers to do different things.
Maybe particular neurons in motor cortex are all the ones you need to activate in
a particular task
 E.g. drinking – will need muscles to swallow and pick up objects
 Throwing – muscle of arm needs to be coordinated with extension of the
hand
 So maybe there is a “throwing” area and a “grasping” area
  So if we’ve evolved to do a bunch of basic tasks (eg throwing, eating,
grasping
o Maybe the motor cortex has patches of neurons we need for those
skills
o This is the “synergist” perspective
o Everything we do requires lots of muscles
Two approaches to neural coding
 Identified neuron coding studies
 Unidentified neuron/population coding studies
Identified neuron
 Get electrode into neuron and be certain that neuron really is output
neuron which is driving muscles in opposite side of the body (and not an
interneuron or something else)
Population study
 Tricky to specify individual neuron. Record the population of neurons in
the area – as many as can be sampled. Gives general picture of the tuning
Evarts et al
 Had “marionette” view
 Force of contract directly proportion to the firing rate of the neuron
 Placed electrode into layer 5 wrist area in right motor cortex which fires
when monkey moves left wrist. Is this just some tedious interneuron or is
it really a motor neuron?
 To test this, placed second electrode into pyramid in medulla. Output
from motor cortex crosses over in medulla – so he could catch the output
 Stimulated electrode. If this electrode was really picking up activity of
motor neuron, then stimulating the pyramid should backfire – antidromic
effect which causes signal to go back up into primary motor cortex.
Should be able to backfire the neuron.
 Monkey is trained to grasp handle to rotate the handle to a position
 Can record that the flexor muscles are active and extensor muscles are
silent. Corticospnal tract neuron (CTN) is firing just before the onset of
movement – corresponding to driving the flexor muscles
 Added weight either on the left or right which would make it either
difficult or easier for monkey to make that movement. Weight either helps
or makes it harder.
 What does the neuron do in the case that the task is more difficult or
easier?
 When the weight was difficult – see increased activity in flexor muscles
and in CTN. The neuron codes for muscle force
 Primary motor cortex is calculating force needed for muscle contraction
 When the weight was moved to the other side to make the task easier, the
flexor muscles were silent and didn’t need to make a contraction at all,
simply needs to switch off extensor muscles
 It’s all about the force, but not about the position – primary motor cortex
neurons don’t change firing rate occording to position of the hand, only
the force required to move hand to the correct position.
 They don’t code location, they code force
  Athough evarts was right that it’s about forces, it’s not one-to-one
mapping and actually is more sophisticated coding of force
 Spike-triggered averaging – electrode in the primary motor cortex
 Monkey trained to make particular movement (e.g. pick up a peanut)
 Identify neuron in PMC
 Recording from the neuron and from the hand
 Trigger the display of the muscle activity from every spike
 Average across thousands of spikes
 Record from different muscles
 Peak in muscle activity shortly after spike, and also true for some other
muscles
 Peak is direct readout of influence of single muscle.
 These patterns are task specific – when the grips are different, the spike
patterns are different. E.g. precision grip vs power grip. Same neuron has
no impact/activity at all when doing a power grip.
 So it is not like a marionette – primary motor cortex is about the thing you
are trying to do rather than the muscle that sis contracting
 So not just about the forced that we need to contract
 Is primary motor cortex a low-level area? Or is it a higher-level,
sophisticated area that has the vocab of different tasks and actions?
Unidentified neuron/population coding studies
 Saves time
 Removes noise
 Might be able to get enough neurons in the patch which can show how
neurons that run the synergy of muscles might work together
 Monkey trained to grab handle and move to one of the targets
 Trained to use three different grips
 Muscles that monkey needs to use to move handle are completely
different depending on the grasp
 Able to dissociate movement from muscle
 Graphs on right represent single neuron being recorded – interested in
moving handle up and not to the left
 When monkey uses supernated grip, start to see response when monkey
moves handle down that you don’t see for the pronated grip. Intrinsic
neurons – care about which muscle the monkey is using
 Of the 88 neurons that they sampled, 28 statistically showed similar
pattern – no firing in particular direction, but some firing in that same
direction when using different grips
 Extrinsic neuron – no interaction of grip – care which way the handle is
moved irrespective of which muscles are used
 Twice as many that care about where monkey moved handle than there
that cared about which muscle the monkey needs to use to do so
Centre out task –
 Monkey starts in the middle and moves to different positions on clock
 Coarse tuning curve for neuron as a function of direction – has preferred
movement direction
 Plotted 288 neurons
  If you take the vector average of the 288 neurons, corresponds almost
directly to the direction that the monkey moved the handle
 Population vector – representation of the net average firing of all the 288
neurons he recorded from as function of their preferred direction.
Population vector codes the movement
 Can trick the monkey last minute – population gradually shifts as you
move the target
 Problem: each pavement is going to primarily use one muscle but the
muscles work in antagonistic pairs – so of course they’re going to prefer a
certain direction of movement. Maybe these muscles just have a preferred
direction of action and these neurons are projecting to the muscles. So
maybe the patterns we found were just a trivial consequence?
Longer stimulations using TMS at natural timescales produce complete synergic
movements
 Look like motor primitives
 Movement theory rather than muscle theory
 Could just be somatotopy but also “rich”
 If you place monkey’s hand onto different initial postures on different
trials and you stimulated same location, monkey’s hand would move to
same posture
 Irrespective of where you put the hand, will always make same defensive
posture in the end.
 Primary motor cortex seems to be coding for the final position of the
hand, not the muscles you need to use to get it there. Because completely
different muscle groups are needed
 Wants to keep continuous map of external space
 If you code the different behaviours that monkey will need to do to
survive and try to make sure the brain represents external space in close
local regions of the brain and that muscles that are adjacent need to be
close in the brain, you get a map similar to that of penfield’s.
Lecture 7 and 8: SMA and preSMA

Supplementary/pre-supplementary motor areas: the cognitive-motor interface,

voluntary action and the neuroscience of free will

SMA seems to provide drive for action in the absence of an external stimulus –
actions we make for our own internal reasons

Where do our actions come from? How do I decide what to do? Are my actions
free? Does that make me responsible?

Slide shows that there are two strong cortico-cortical inputs in the primary
motor cortex from the area just anterior to it. Can distinguish between lateral
route (1) and medial route (labelled 2)
 2 – input from cortical areas on medial wall of hemisphere (the one that
you don’t see, buried ins interhemispheric fissure
 These two routes provide input and drive into M1 for actions made for
two different classes of reason
 Simple action – finger movement – two reasons why you might make this
movement. Might be reacting to external stimulus (e.g. insect coming) – or
might just feel like it.
 Latter class of action – don’t look for reason for action in external world,
you look for the reason internally
 Why did she do that?
 Broad distinction between the lateral route (concerned with driving
actions that are made in response to environment) and medial route
(driving the same actions but internal)
 What about cases that lie in the middle? Reacting but with free will?
Combination of both external world and internal decisions
 So it’s hard to say for some cases
Lateral route:
 From the parietal cortex (bring multisensory info about external world)
to premotor regions
Medial route:
 From presupplementary motor area/supplementary motor area back into
motor cortex. Where does this area get its input from? Subcortical input
(basal ganglia)
 Will be talking primarily about the medial route
 Whole network of different zones in this route. Rostral cingulate zone,
SMA/preSMA zones. Pretty homogenous broad cortical area.
SMA/preSMA are on top. Cingulate zones are more deeply embedded into
medial wall
 Supplementary motor area classically defined as complex of two areas.
 SMA proper – pretty like M1 (just on border)
 preSMA – more cognitive area, does not project much to M1 and not at all
to spinal cord. Largely connected with prefrontal areas.
 Within frontal cortex – gradient between most anterior parts (completely
cognitive – what should I do next? What’s the long-term strategy of my
 actions?) As you go more posterially towards M1, transition gradually
from something that looks like a thought to something that looks like a
movement.
 preSMA and SMA seem to be the last stages between though to
movement.
Medial frontal areas – concerned with internally generated actions. What do we
mean by voluntary action?
 We think we have free will – but what does that actually mean? There’s no
single thing we can point to, but cluster of features, none of which is
compulsory, but taken together may be sufficient.
 Not based on obvious stimulus or reflex
 Leads to movement
 If you find a brain centre for volition, might find just weak connection to
sensory areas.
 Any brain basis of volition has to lead to movement. Voluntary things are
not just thoughts, they are actions.
 Need strong connections to motor systems
 Deep idea- actions are goal-directed. We make actions for a reason.
 You can ask “why”
 Connected to reward areas – we perform actions because we want
reward. So we can look for these areas. Also planning, monitoring, and
finishing actions.
 Artistic creativity – example  create something no one has ever created
before
 Routine action is not necessarily voluntary – act outside the box
 Something to do with consciousness?
 We have experience that we are controlling what we do - conscious
awareness of what we are doing and we decide to follow through
o This is not necessarily true for reflexes
 Quite philosophical, but implications about how brain circuits should be
wired
 Tensions between things – e.g. reasons-responsive but also
innovative/spontaneous?
 The table shows intuitions that might be related to the way particular
actions are represented in our cortex
 A lot of psychiatric disorders involve change to voluntary action
o E.g. OCD – balance between voluntary action and habitual action
has been lost
o Actions of patients – how do you know if e.g. washing one’s hands
is a voluntary action?
o Are voluntary actions linked to moderation?
o Not necessarily about what you think is a good reason
Anatomy
 Macro and microanatomy of medial frontal cortex in humans
 Immediately in front M1 you have supplementary motor area and then
before that you have preSMA
 Cell counts tell you something interesting – area 4 = M1. Layer 5 – betz
cells. As you move forward you still find some pyramidal tract neurons.
  As you move forward, you see increased interneural connectivity
 As you go more and more anterior, less concerned with movement and
more about bringing in input from other areas
What does this medial wall do?

Internal generation
 Sometimes we make actions for ourselves, not because of anything
external
 Dimension ranging from reflex response --- intentional decision
 E.g. football penalty kick – must choose to kick to the top right or top left
 Competitive games seem to particularly lie on internal generations – don’t
want to disclose information about basis for action
 Most of the actions we perform are responses to external stimuli linked to
what we want to do
Two cortical systems for directing movement
 Passingham (1987) Lesion studies in which he surgically removed regions
of the monkey frontal cortex
 Either removed the lateral route by cutting lateral premotor cortex or
removed the medial route by cutting SMA
 Monkey trained to move handle one way or the other depending on
colour of the light
 Shows number of times the monkey does the wrong thing. Monkey with
lesion to premotor cortex makes lots of errors
 Monkey with SMA lesion also made no errors. Can move fine when there’s
an external environmental cue telling them what to do. Not paralysed, and
can successfully select the right movement if there’s a cue/instruction
telling them what to do.
 The lateral route is the one you need to respond to external signals
 SMA is essential for internally generated actions – but how do you test it?
Normally we tell people what to do, but how do we elicit voluntary
actions in a laboratory setting?
 Passingham developed setting – monkey in cage and infrared light beams
at the top of the cage which are invisible to monkey. If monkey raises arm
will break lightbeam and with a certain probability get a reward – need
some kind of award.
 Only learns because of action-outcome relation and reinforcement of
action
 Graph shows number of arm raising responses monkey does per minute
 Control will learn to raise its arm  will do it whether there is a visible
target or no target
 If you lesion SMA, monkey stops raising its arm when there is no target,
but will continue to raise arm if the target is there.
In humans
 Jenkins (2000) Ask humans to move joystick  when they want. Can let
them choose direction. In other condition – move joystick when they have
heard tone.
 When people decide for themselves  activation in preSMA
  Was a PET study – time resolution is worse than fMRI  non-magnetic
environment in PET. fMRI depends on BOLD signal, but we don’t know
how that relates to firing of neurons. PET injects radioactive ligand, shows
glucose in brain, transported to sites of maximum metabolic activity,
related to firing of neurons. More of a direct measure than fMRI
Individual neurons
 People once again found difference between lateral and medial route
 Halsband et al (1994) – monkey trained to make externally or internally
generated actions. But monkey makes 3 sequence of actions from memory
(internal). External – 3 lights come on to tell monkey which sequence.
Monkey waiting for series of traffic lights. But in internal – monkey has
learned the actions from memory.
 Neurons recorded from supplementary MA or from M1
 Number of neurons active n each area recorded as a function of which
movement of handle and whether Internal or External
 Can do this before the monkey moves and whilst moving
 PMC – externally triggered
 SMA – internally generated
 PMC - The majority of neurons more active when monkey was making
each movement in response to a coloured light
Internal generation
 Review paper by Avamec – tracing studies which have investigated
connectivity between back of monkey brain and front (front = action, back
= sensory) how do we make actions based on sensory input?
 Cluster analysis – worked out main links between parietal lobes and
frontal lobes which use that sensory input for action
 Links found between PMd and PARd, M1 and SS
 Cingulate cortex in medial wall – doesn’t get any substantial parietal input
– don’t get info from back of brain about what’s going on. Main input is
not from multisensory description of environment, but actually the limbic
system. – Doesn’t know much about the world, but only about what you
want; your needs, desires, and motivations.
Philosophical interlude
 What do we mean by “internal generation”? Why is it so fundamental to
our concept of human nature?
 Doesn’t give us a definition about what causes the action
 Can sort of see what basis it has in the brain, but is it really satisfactory?
Strong enough idea to carry all the moral weight
 What is the degree of determination that the stimulus has on the action?
Action memory: sequence and complexity
 Tangji – these areas that we think of as being involved in internally
generated action play role in producing complicated sequences of action
which are replayed from internal memory
 Monkey is trained to produce push, turn, pull
 Each line is individual trial, each tick is action potential
 This neuron shows ramp like increase in firing prior to first movement
  If monkey is also trained to make different sequence of movement with
handle, then neurons that are prepared for “push turn pull” are silent for
“push pull turn”
 So neuron codes for sequence rather than the movement itself
 Monkey chuks to form action chain – neuron is therefore involved in the
action chain rather than any sort of motor movement
 Medial frontall wall – cognitive complexity in representation of
movements
 SMAL proper neurons – other class – concerned with playing out bits of
chunk (e.g. between pull and push)
 Can flow from one sequence to the next
 Tanji – summary of different classes of cells
Cognitive motor frontal hierarchy
 Transition from abstract thoughts through to specific muscle contractions
Inhibition
 Medial frontal wall also involved in suppressing movements
 Not just about choosing what actions to initiate, but also which ones to
inhibit
 Meta-analysis: Rae et al (2014) – free selection in top row (choose for
yourself what you want to do, generate), Stopping (asked to stop an
action which is already happening – e.g. pressing a button then given a
stop signal – can vary delay between go signal and stop signal. If the delay
is short, and successfully delay, but if the stop signal comes late, you can’t
stop yourself. By varying the delay, can calculate the time at which stop
signal needs to happen to cause 50-50 balance.
o What is the problem with the stop signal task? You’ll learn!
Unwanted learning effects
o When you give people stop signals – activation in medial wall,
preSMA
o Deciding what to do and holding back – same mechanisms
 After medial frontal damage, can develop interesting syndromes where
unable to inhibit actions. (GC) Woman had paralysis of right arm, as it
recovered, the arm started to carry out actions without intending to. Had
to use left hand
o After lesion, syndrome in which one hand compulsively responds
to external stimuli – e.g. need cup of tea to get reaching action. So
has lesion in SMA. The rest of us have inhibitory function which
tells us to stop doing what the world tells us to do. (Della Sala et al.,
1991 – anarchic hand)
 Can get a similar thing with corpus callosum
 Suggests that same areas in free will also stop us from doing things that
we don’t
Initiation of voluntary action and “conscious free will”
Kornhuber + Deecke (1966) – unhappy with the notion that all actions are
driven by external stimuli.
 Measured readiness potential, allowed to press the button whenever they
felt like it
 If you ask people to press button when they feel like it and record
electrodes – Cz is SMA proper – readiness potential – ramp-like activity
that occurs over medial frontal cortex 1 or 2 seconds before the person
makes the movement. Person has generated for themselves the
information. Not good spatial resolution from this technique
 Libet (1983) – adapted experiment to investigate brain activity and
conscious experience. Clock serves as chronometer to report what you
experience. After you press button, clock hand stops. Asks – where was
the clock hand where you first felt the urge to move?
o When did people become aware that they wanted to press the
button?
o Readiness potential began a second or more before the person
moved. But the average time using the clock (W judgement) was
206 ms before the button went down.
o Concluded that brain begins to prepare the action when the
readiness potential starts
o You become aware that you’re about to move 200 ms before you
move.
o We develop feeling of intention as a result of unconscious brain
activity
 Since libet – tendency has been to dismiss experience of our own action as
merely illusory
o The brain activity causes our body movement
o We retrospectively postdict reasons for our action
o Free will confabulationism – our free will is just a confabulation to
explain our actions
o We are always constructing retrospectively
o There are some cases where we have a conscious experience of
action before we act and this is not retrospective
o Fried (1991) – implants grid of electrodes into medial frontal wall
of left hemisphere
 For a week before surgery, recording from and stimulating
to work out where they are (e.g. if they start moving you
know they’re in the M1, if they start crying you know
they’re in the limbic cortex, etc)
 Neurosurgically mapped brain
 Supplementary motor stimulated (a4b4, a6b6) – patient
who is completely awake will spontaneously report an
“urge” to move.
 Report experience which sounds like “will” – is language of
volition but have not moved at all. So how could they
possibly operate this system of retrospectively postdicting
the urge when they have no external sensory evidence that
body is moving?
 This is not conscious free will, but it is conscious and does
sound like will
 No externally sensory evidence – except when the current
was turned up. Then they actually moved the same body
part that they wanted to move when stimulated at lower
current
Lecture 10: Action awareness and the automatic dorsal stream

Are we aware of our actions?

 We all make very fast, sophisticated actions all the time
 We probably aren’t aware of all the movements we are making
 How can we study this experimentally?
Motor awareness
 Virtual reality setup – gloves that record hand movements. Recording eye
movements.
 If we know exactly what the movements are, we can ask participants to
track what they know about their movements
Blindsight patients
 Selective lesions to V1
 No visual experiences – report being blind
 But have some ability to access visual information
 Patient TN – able to navigate obstacles
 Perception and action might be mediated by separate visual pathways
Dorsal stream
 Compute absolute metrics of target objects, real-time
 Allows us to interact with objects dependent on the specific actions we
want to make
Ventral stream
 “What” pathway
 Conscious percept
 Independent of orientation or location
There is some information that doesn’t go through LGN, but goes directly
through superior colliculus to posterior parietal cortex

Dorsal stream deficits

 Hemispatial neglect – unaware of contralesional half of space
 Akinetopsia
 Apraxia
 Optic ataxia – impairment in visually guided actions
Accurate grasping despite visual illusion
 Ebbinghaus illusion
 Aglioti (1995) – if you look at size between index finger, grip is accurate
but perception is difference. But failure to replicate
Dot in frame illusion
 Pointing to the dot
Optimal saccadic adjustment
 Making saccades is accurate – participants able to make correct saccade
 But perception – not able to detect that target was displaced
Antisaccade paradigm
 Participants had to look at opposite side to which the cue was shown
 Also incorporated a precue to make things even more confusing
 Look at trials of perceived errors – large contraction in the wrong
direction – participants made big deviation of their eyes towards the
wrong location. But also some cases in which participants made
 contraction in the opposite direction but made fast correction towards
correct location and they don’t even perceive that their eyes made these
incorrect movements/errors.
Visual search
 Eye movements guided by completely non-conscious information
 One target that is more salient than the other
 Difference between moment people make saccade
 The more salient the target it the more likely participants will be at
making fast saccades. Only for slower movements that people were
making saccades
 Can ask participants to report whether or not they saw image.
 Actions are initiated unconsciously
 Some visual information is informing action that we aren’t aware of
Sophistication of grasping movements
 Velocity of hand varies with distance of the object.
 Grip size is invariant regardless of velocity
Automatic pilot?
 Need to correct movement
 Participants are good at correcting their reaches
 For movements that are 250ms, end up at correct location. For
movements that are faster or a bit slower, the accuracy of final reach is
not that good
 Location stop – if the target jumps, this time you have to stop your
movement and not reach for that target
 Participants are able to inhibit their movements quite well fo slow
movements. But for movements that are really fast, a good proportion of
the time they end up reaching for that target
 Participants unable to inhibit the correction of their movements when the
movement is in front of them
 Patients with parietal lesions cannot inhibit corrections at all
 But much easier to inhibit movement when the change is code by colour
change
 Suggests that dorsal stream operates outside awareness, but that
information is coded by different things.
Limited motor awareness
 Virtual reality setup
 Trying to see if participants were aware of distortion in their movements.
Participants sat in front of screen holding a pen and had to draw lines in
front of them. Wouldn’t see what hand is doing but would se position of
pen reflected by the mirror
 Distortion between what you do and what you see. Showed a line that was
distorted to the right of hwat you actually drew. To display a line going
straight you have to make a movement that is slightly tilted to the
opposite direction
 When they asked participants to report what they were doing found 2
groups of Ss
 Graph: dots show actual distortion
  One group under estimated the deviation of their movement. Reported
something less tilted. Individual differences in the abilities
 Another group actually estimated their movement is deviated in the
opposite direction
Neural representation of perceived movements
 Where the dissociation could be in the brain: possible to have info
representing what you think you are doing and what you are actually
doing
 Monkey study: asked monkey to make rotation movement. With curser,
have to make small circles. Gradually start changing the degree of rotation
of what is displaced on the screen. Adapt the visual feedback gradually -
participants are unaware that they are required to change their hand
movements more
 Neurons in the M1 represent the actual actions of the monkey rather than
what the cursor is displaying
 But actually PMv code the intended movement.
 So dissociation between conscious representation of what we think we
are doing and representation of what we are actually doing.
 Fine motor adjustments escape awareness especially for subtle fast
corrections of movements.
Open questions
 Something special about eye movements – are there some kind of actions
that would be more accessible to awareness than others? Are we unaware
or are we just not paying attention? Overload of information?
Action errors and cognitive control
 Sometimes we don’t have the ability
 Sometimes we don’t take into account the information correctly
Error detection and cognitive control
 How do we become aware of our erros?
 Need a circuit that monitors the results of actions
 Error detection – is the action we detected the same as the action we
executed
 Second comparison – info we have in environment vs action we selected
How to study errors
 Frequency
 Circumstances
 Types of errors
 To make people make errors: increase speed, multitasking, present
conflicting information
Error correction in the absence of feedback
 In 60s discovered that we spontaneously correct our errors
 Participants see number and had to say whether higher or lower than 5
 If participants made wrong answer in first place, almost instantaneously
pressed second button to try to correct the errors.
 Later, participants told to not correct, but to press additional button when
made an error
 Ps were slower to report that they made error than to simply detect it
 More errors made in second scenario
 20ms after error could correct
o occurs without instruction
 Error signalling is different – happened only when participants were
instructed and worse when distracted
 Suggested that error correction (because they are so fast) might not be
corrections but might be delayed correct responses.
 Maybe we are preparing several actions simultaneously.
 Seem to be specific behavioural responses to error:
o Graph shows response time in ms before and after error correction
o Mean reaction time when they make error is faster – more likely to
make mistakes
o Error correction seems to be almost at the same time (small
difference) in the time it takes to correct it
o After participants have made an error, even if responding
correctly, slow down response very strongly and become actually
slower than when you were responding previously
o Increase of motor control after we make an error
o After we make a mistake brain slows down
Automatic post-error slowing
 Escapes awareness
 Logan – scientists asked professional typists to type words but sometimes
they inserted mistakes, to look like they made an error
 Other times they corrected the mistakes when participants made
mistakes
  Looked at how participants slowed down
 Looked at time between the inter-key strokes
 Normal errors – participants made an error and was displayed on the
screen – participants slowed dwn.
 What happens when you correct error on the screen – people think they
had correct word but pressed the wrong key – so even when you make a
mistake and it is automatically corrected, your brain registers it and slows
down
 Inserted errors – no post error slowing – although you had impression
your brain doesn’t register it as an error.
 We have automatic post-error slowing
Error-related negativity
 EEG recordings after an error
 ERN activity peaking rapidly that signals a mistake was made
 Negativity is similar no matter the sensory modality
 Generic signal in your brain that you made mistake
 Origin of signal – Anterior cingulate cortex
o Also activated when making judgements of confidence
An ERN during error observation?
 Is it active when looking at someone else make an error?
 Task where you have to press a button an also sitting in front of someone
who is doing the same task
 Compare ERN activity when you make it yourself vs when someone else is
making an error
 Typography is similar – reaction time is slower because we cannot
anticipate it as well but the same brain region is activated
ERN for unaware errors
 Occurs even when you make a mistake that you’re not aware of
 Antisaccade paradigm – with precue.
 In black – when Ps respond correctly – can make aware errors and
unaware errors or partial errors.
 ERN is present in both cases
 Only a later component (error positivity) is presented only for aware
errors. Dissociation between errors that you’re aware of and those that
you are not aware of
Does the ERN reflect confidence?
 Scheffers & Coles and Boldt – how confident you are that you are wrong is
reflected in ERN and later positivity following the error
 ERN – mismatch between evidence you have in favour and against the
decision that you made
Conflict monitoring
 Monitoring level of mutter conflict
Conflict as a cause of errors?
 When there is interference between conflicting evidence of what you
should do that captures attention
 Stroop effect
o Conflict
 Eriksen flanker task
 o Arrow in centre of screen. Respond to direction of error but
surrounded by distractors
o Easier when distractors are congruent
o During congruent trial (>>>), all neurons representing > a
direction will fire
o Incongruent trial: neurons representing < will also fire.
o Response unit – codes whether you will make left or right snwer to
stimulus
o Competition between the response and attention units
o If you have one response, will inhibit activity of the other units.
Allows to make a selection
o Congruent trial: three units activated
o Incongruent – information f central arror goes towards left, but
peripheral arrow activates right
o Create mathematical model – the more bout h of the units are
activated, the higher the level of conflict
o If you just activate one response, then intermediate level of conflict
o But if you activate both responses, get higher level of conflict. Then
triggers attention unit and tells brain to stop paying attention to
the left and right
o In the case you make an error – there may have been some conflict
from before that
o The maximum moment where there is conflict is just after you
made the incorrect response
o And also then have strong negativity ERN occurring in rain
o So suggested that ERN could represent conflict monitoring signal
o Prediction of this model and size and time of ern and this model
explains what is happening very well
Converging evidence of correspondence between conflict monitoring and ERN
 Conflict activates same region that error related negativity ERN reponds
to
 ACC is active in response override (stroop) and also when you have
multiple correct choices (e.g. verb generation) and also when you make
errors.
 Attention unit: what conflict monitoring is for is to monitor level of
attention. When we are not paying attention, it recruits more attention so
we stop making mistakes
 There is some evidence that ACC is modulated by degree of control
 Gratton – reaction times – stroop tasks and look at whether there is a lot
of conflict on the trial you are in (difficult or easy trial?)
 When previous and current trial has low conflict vs when previous trial
has high conflict
 If previous trial had low conglict and low conflict in current trial, respond
quite fast. When low conflict before and high now, RT much slower. But if
previous trial had high conflict, RT is not as slow for current high trial
conflict. – Less surprising so more resources recruited. Activity in ACC is
modulated in the same way as reaction time. The moment ACC is
 activated, high conflict has just appeared. If you had high conflict already,
the ACC is not as activated than Low-high
Contrasting views of action decisions

Stochastic fluctuation model (Schurger et al., 2012)

 Actions are random
 Schurger proposed – if you ask people to press button wheneverthey feel
like it, people rely on internal neural noise which is effectively doing a
random walk until it hits a threshold
 Stochasticly going up ad down – if you look back is the neural activity and
average across what signal does
 Stochastically varying neural signal – if you average all the approaches it
takes to produce a threshold, produce a curve
 If you average a fluctuating neural signal, will find something that looks
like an rp but it isn’t, it’s just a random walk
 Maybe there are decisions about voluntary action, bu they happen not
when readiness potential starts, but the moment you reach the threshold
is when you actually do it. Noise decides when you want to act. Your
action might be a veto process.
 What causes the signals is not a process, but that you just get that curve
because you average the spikes
Role of basal ganglia in action control
What drives the SMA neurons?

Reverse engineering the action chain

 Once you get to the basal ganglia, you find a loop rather than a chain
 Doesn’t just produce a “secret” by reverse engineering, bu that there’s an
element of recurrency and iteration than linear continuity
Basal ganglia
 Collective term for a group of subcortical nuclei linked by a circuit of
excitatory and inhibitory connection
 Through their recurrent connecions to the cortex, regulate control of
action
 Subcortical – don’t see them from the outside
 Tadpole like structure (caudate nucleus
 Encloses steak-like structure – putamen
 Caudate and putamen are jointly called striatum
 Underneath putamen – globus pallidus – pale blob – contains internal and
external segments. Medial – internal. Lateral – external (both terms used
interchangeably)
 Subthalamic nucleus
 Substantia nigra – also divided into two parts – pars compacta (squashed
bit), pars reticulate (stripy bit)
 All these collectively form basal ganglia
 Key thing is where they sit with respect to the cortex – very intimately
connected to the cortex
 Middle bits in particular – sensory parts, motor cortex, SMA,
sensorimotorcortical areas connected to basal ganglia
 All basal ganglia input is inhibitory
 Basal ganglia knows everything going on, sends output to thalamus, which
goes back to narrower set of cortical regions than the regions providing
the input
 Crucially – output is specific – primary motor cortex and SMA, whereas
input draws from a much larger area
 Physiology = anatomy plus thought
 Suggests basal ganglia acts as a filter – need to get this information from
diverse set and pick what to do and procide focal activation of just those
areas that are going to produce actions. What about emotions?
 Regulate amount and appropriateness of action
 Compute movement quantity (how much and how fast?)
 Do NOT regulate movement quality – don’t specify fine motor detail of
what you do
 Cerebellum is associated with this^
 Example – once woman got parkinsons, signatures got smaller – less
movement – quantity
 Connectivity through the loop – excitatory and inhibitory connections
from one nucleus to the next
Inputs to BG
  From the cortex (almost all areas) to the putamen (receiving area, part of
the striatum). Input is excitatory to putamen, making cells fire.
Direct pathway
 “Direct pathway” – dashed outline – cell bodies in putamen which make
inhibitory to GPi – globus pallidus internal. Inside GP – neurons make
inhibitory connection to thalamus (VL). In VL, cells that make excitatory
connection back to cortex.
 Some event happens that makes cortex neurons fire. So becase putamen
inhibit GPi, will no longer exert inhibition f VL, which means thalamus will
be disinhibited, which means they will excite cortex, which means you’ll
do something. Basal ganglia circuit promotes some kind of response.
 GPi neurons are tonically active – always active. Putamen causes the
neurons to stop firing, meaning they no longer
Indirect pathway
 Opposite effect to direct pathway
 Direct pathway facilitates actions, indirect inhibits
 Putamen  inhibitory signa to GPe, inhibiting the GPe neurons. These
normally have an inhibitory connectin to the STN. But since these are now
inhibited, there is a disinhibition, which means STN can now fire, which
project to GPi. The GPi makes inhibitory connection to VL, which means
the VL is inhibited towards the cortex. GPe normally has inhibitory effect
on STN, but become disinhibited
 GPi tonically inhibit thalamus. Net effect of indirect pathway is to stop.
Inhibit yourself.
Nigrostriatal Pathway
 Direct and indirect cancel each other out
 Need to have some mechanism for balancing out these
 Volume knobs for behaviour – are we hyper excitable or inhibited?
 The way it does this is with nigrostriatal pathway
 Substantia niagra pars compacta to the stratum
 Uses dopamine as a neurotransmitter
 Uses dopamine in interesting way – both excitatory and inhibitory –
neuromodulatory
 Axons down direct pathway to GPi
 Neurons in SNc also
 Neurons in putamen that receive input from the
 Depedning on the receptors in the membrane of the post-synaptic cell,
will open channel that goes to
 When neurons in SNc fire, reduces firing rate of striatal rate
 Because of the different dopamine receptors that it targets of different
neurons, nigrostriatal pathway promotes activity of direct pathway, but
inhibit activity of indirect pathway. Same chemical can lead to excitatory
and inhibition
 Upregulates the direct pathway, making you do stuff
 If you don’t have nigrostratal pathway, will turn down volume on action
system.
 Need to have enough projection from SNc
Hyperdirect pathway
  Goes straight from cortex, bypassing striatum entirely
 Drives indirect pathway by exciting STN
 When an event happens in external world
 Has an inhibitory effect of thalamus to cortex
 Provides quick signal to stop – brain’s “brake”
 Helps us not just always respond to external stimuli
Ready set go stop paradigm – cue and button press. In some trials, put stop signal
after the go signal.
 Can they stop themselves from pressing the button
 People can only stop themselves if it comes early enough
 Otherwise get failed inhibition
 Can adaptively vary time at which stop signal is presented so people are
jst able to stop themselves half the time (50%)
 A person who is able to stop themselves even when stop signal is very late
has high subthalamic nucleus activation.
 The more you can get STN active, the better you are at breaking.
Parallel loops
 Four distinct parallel BG loops
 They differ in terms of cortical regions from which originate and the
cortical target to which they project back
 Sensory motor loop receives input from somatic sensory and primary
motor/premotor areas, and projects to SMA
o Action follows from what they did before
 Association loop – underlies thought
o Receives prom posterior parietal cortex, goes through head of the
caudate nucleus. SNr and GPi – goes through different thalamic
nuclei – project to prefrontal areas like dorsolateral prefrontal
cortex.
o Same architecture but not necessarily an action – in association
loop looks like input and output are more to do with “what’s the
general context of everything? What am I going to do about it?”
might not do anything
o Less tied to sensory motor and more to general cognition
 Oculomotor loop
o Like sensorymotor loop, but specific to eyes
o What do I look at next?
o What should I be paying attention to?
o Like the sensorymotor loop but concerned with eye muscles, role
in attention
o Receives from posterior parietal and prefrontal, projects to eyes
 Limbic loop
o Receives from hippocampal, memory areas – medial and lateral
temporal lobe. Projects to anterior cingulate orbitofrontal.
o Given your environment and memory, how should you respond
emotionally?
o What should you feel now?
Motivation for action
  How does the basal ganglia implement what I want to do? I shouldn’t
make an action unless I have a reason to do so
 Medium spiny neuron
 You can classify striatum into two kinds of circuitry
 Two types of cell organizations – matric and striosomes
 Sensorimotor cortices project to matrix
 Limbic cortices project to the striosomes
 What am I doing and how do I feel come together in same place in
striatum
 Not only do they come together, but they then interact at a particular
organization of synapse onto medium spiny neuron
 Input from limbic cortex into striatum – MSN fire. Dopamine neurons in
midbrain are excited by limbic systems, project back up to sensorimotor
striatum. Where on the dendrite does this chain end? Motivational input.
One synapse is further away than the other
 Synapse of the neocortex will only get through if the synapse from the
dopamine neuron allows it. If it allows depolarisation is passed through
 So this can allow or not allow info from the neocortex
 Dopaminergic input into striatum is able to regulate input from
neocortex. It’s all about where the neuron dendrite is.
Hyperkinetic/hypokinetic movement disorders
 Damage in basal ganglia makes people hyper or hypokinetic
 Main classification is to see if it’s hyper or hypo
 Hypo – e.g. strokes. Hyper – e.g. tourettes
 Can classify most movement disorders as hypo or hyper
 We know the neural circuitry for huntingdon’s disease
Parkinson
 Cells in SNc die. Can lose up to 70% of them
 Is parkinson’s just accelerated aging?
 Degenerative condition – cell death in nigra
 Once you lose a particular threshold level of neurons, don’t have strong
nigrostriatal projection.
 Once there are less than 30% left, don’t get enough dopaminergic
regulation
 Neuromodulation is not sufficient to drive direct pathway and inhibit
indirect pathway. So indirect starts to dominate GPi firing, so strong
inhibition of thalamus, so reduced thalamic drive. Don’t move enough.
Symptoms of parkinson’s
 Reduced movement amplitude
 Rigid limbs
 Resting tremor – one of the things you see first. We think it comes from
cerebellum rather than basal ganglia
 Difference in balance of drive from cerebellum loop (which dominates too
much)
 Like rolling a ball of wool
 Arms don’t move when walking
Treatment
  Give L-dopa – which helps brain make more dopamine – helps movement
come back
 In the short term, Parkinsons can be treated by L-Dopa
 However, can have side effects of dyskinesia around face
Deep-brain stimulation
 Want to delay L-dopa treatment for as long as possible
 Stimulate to disable STN normal activitiy
o Will tend to reduce strong drive of STN on GPi, reducing
overactivity in indirect pathway, bringing back into register with
direct pathway, restoring the thalamic input to normal levels.
 Originally done in primates by giving monkeys dopamine antagonists
Non-motor/cognitive functions
 Parkinsons patients complain about not feeling right rather than moving
slowly/less
 Reward-related signalling in BG
 Monkey is expecting to receive food reward
 When monkey gets reward (R) dopaminergic neuron fires. If you train
monkey to expect reward because of a Conditioned stimulus, will fire ass
soon as light comes on. If you put light on an then omit reward, will
actually cease to fire.
 When PD patients given them L-Dopa, try to get sufficient L-dopa into
motor striatum so they can move again But his floods the limbic striatum.
So ventral is overdosed even though dorsal striatum is just right. Can
produce impulse behaviours such as hypersexuality, pathological
gambling, obsessive collecting, pathological shopping
 If we have too much dopamine, hyperkinetic. Tend to do more of things
that they already do a little bit.
Schurger’s paper
 Competenece model
 White noise, pinky noise, very pink noise
 Pink noise – most of the noise is at lower frequencies
 Pink noise – wobbly pattern that you don’t see for white because the slow
up and down are the strong low frequencies
 Pink noise is common for biology – they change slowly
 Put noise through accumulator – adding up white space under curve
(integration of area under curve) as you integrate curve it goes up.
Depends on what the stochastic input looks like
 Nature of noise
 Upper threshold – commitment – point at which action occurs. Prediction
– point at which you know you’re about the reach threshold. Corresponds
to “w” judgement
 When you integrate pink noise, is slower than when it is less pink
 Three columns – different kinds of noise
 Energy as function of frequency
 Fig 3+4 – mean WT – wait time. Arbitrary units
 If people wait a long time, the reason for that is because brain is
producing slow ramp as opposed to faster ramp. Might be reflected by
difference in accumulator or difference in noise
  Black line – result of simulating noise and putting it rthrough leaky
integrator
 Fig 5 – if you have two different ramps – output is rising slowly or rapidly.
Can reproduce this by looking at existing data and looking at whether
 Does the accumulator reset every task?
 Fig 6+7 – button press quick vs slowly. So you become aware that you’re
going to press the button a shorter time before you actually press the
button if the wait is SHORT. Fig 8 – slight but could be individual
differences
 He’s not saying that it’s pink noise that is being accumulated – he’s just
saying that it’s predictive – could account for time at which people act,
signals in the brain that precede the action, and the time at which you’re
aware. Physiological, behavioural, subjective
 One unclear thing – it always starts at the trial? Leak parameter –
accumulate recent noise
 Different shape
 What, when, whether – three aspects of decision making Zapparoli,
seghezzi, paulesu (2017) 28567010 pubmed
o Look for paper by same people PNAS
Brake hypothesis - Nambu
 We might think that the key problem is how do I start something
 But maybe it’s about keeping everything inhibited
 Proactive inhibition – considers that the problem of how you stop acting
till you actually want to. Based on things like getting ready for a race.
Keeps the brakes one
Dopamine function + motivation
 Parkinson’s patients
 Three different types of medication
 Implication of fact that these rewards were fictitious?
 Accumulate as many apples based on combo of apples and awards
 Demand characteristics?
 Different level of effort you have to expend – squeezing hand
dynamometer. Set force to each participant
 Participant decides whether they want to squeeze – if they bother to
squeeze to get the reward offered.
o Only half the time did they actually have to squeeze
 Adapted the test to find the threshold – up/down staircase
 2 sessions
 Trying to piece apart motivation from ability to do it/finish whilst it’s
online.
 Indifference point (IP) point at which Ps will accept a reward 50% of the
time.
 Controls – no difference between sessions
 Main effect of medication PD ON – willing to expend more effort
 Wanted to make sure that these were motivation based effort
 Assessed actual motor skill – no difference in ability to squeeze whether
on or off drug – surprising because you would expect people to squeeze
harder
 Quantity was not affected – which contrasts from what Patrick said about
basal ganglia - CLARIFY
 Does this mean that parkinson’s in general is motivation based?
 Is it to do with risk? No – no difference for the 110% effort levels.
 No difference in accepted but failed trials
 For minimal reward conditions – see deficit in motivation of parkinsons
group
 Saturation effect for upper end of stakes – why is the difference for PD off
more at the lower stakes? Once you get a to a higher reward anyone can
appreciate it. Interaction of group and stake
 No difference between controls and PD patients on squeeze ability –
interesting to see if it would be the same for finer motor tasks
 Does this mean that parkinson’s is a combination of both motivation and
ability?
Metacognition

Focusing on gap between decision making and metacognition

Metacognition – one’s knowledge concerning one’s own cognitive processes

(Flavell, 1979) – originated from studying what children are aware of
 “Thinking about thinking”
 “Cognition about cognition”
 “Object-level” cognitive processes (Nelson & Narens, 1990)
 Formation of explicit judgements about self-performance
 Do I know this topic? Meta-memory decision
 Meta-perception – Am I seeing things clearly?
How do we study metacognition in the lab?
 Second-order reports – judgements/reports about own behaviour or
cognition
 Can be prospective (how well am I going to do?) or retrospective (e.g.
“how well did I do?”)
 Sometimes referred to as “type 2” decisions
 Once you’ve collected second-order reports – can get metacognitive
sensitivity
 Distinguished from first-order performance
 Need lots of trials to assess metacognition
Metacognitive bias and sensitivity
 Confidence in behaviour can be distinguished in these two ways
 For any decision, can have d prime (sensitivity) and c (bias)  type 1
decision
 But may also now rate confidence in decision – can also define sensitivity
and bias in confidence – Type 2
 Metacogntiive sensitivity – how separate are the distributions?
 Bias (overall confidence) and sensitivity (how overlapping the confidence
distributions are) are independent
 How do we quantify the overlap between distributions?
 Vector: 0 when confidence is low and 1 when confidence is high
 Phi: correlation between decision performance and confidence
Study: interested in quantifying metamemory and memory
 4 groups: controls, alcoholic controls, korsakoff syndrome (frontal lobe
atrophy), amnesic patients (but intact frontal lobes)
 Given sentences t learn, then given parts of sentences to complete. Had to
make prospective metamemory judgements – am I going to remember it
later?
 Could compute performance and also confidence (2x2 table)
 Objective performance was similar between the korsakoff and 4 cases
 But the korsakoff group much worse in the korsakoff patients –
suggesting prefrontal cortex involved in metacognition
 Important to be aware of need to control for differences in first-order
performance
 For example, if a group B was worse in the first order tasks and
confidence, differences in metacognition may be spurious
Explicit vs implicit metacognition
 Not all monitoring and control is conscious
 Implicit metacognition – continually monitoring – e.g. action monitoring
of errors – can make corrections regardless of conscious awareness
Dissociating implicit and explicit metacognition
 Logan and Crump (2010) – inserting letters on the screen that were
incorrect, or correcting the mistakes made
 Implicit – impact of errors on subsequent response time. If the system
detected the error has occurred, would slow down on the subsequent
trial. If the typist really did hit the wrong key, then reaction time would
slow down. But if the error was inserted, there was no subsequent
slowing. Correcting had no effect
 Explicit – when subjects were asked, had to make judgements of if they
made an error. If an error was inserted, then the typist did indeed think
they made an error, even though they did not cause any implicit slowing
 So implicit and explicit metacognition are different
 One hypothesis about why these different levels might exist: explicit
might be important for communication and collaboration with others
 Study looked at how people might combine metacognitive judgements
 Bahrami et al (2010) – Ps made decision about whether stimulus was
brighter in interval 1 or interval 2. Individuals were allowed to
collaborate and decide together. Two people do better than the best
individual. Depends on being able to share your confidence together
Origins of metacognition?
 Not all monitoring and control is conscious
 Do we share metacognition with other species?
 Need paradigms to quantify metacognitive sensitivity in other animals
 Waiting time measures, post-decision wagers, opt-out response
Opt-out response
 Does box contain large or smaller number of dots? Animals have to ht box
if large amount of dots, S if small amount of dots, and star if unsure
 If it’s ambiguous if it’s sparse or dense, or likely to use opt out response
 Shows similar patterns between macaques and humans – rudimentary
metacognitive response
 Shown in dolphins as well – pushing buttons for low or high frequencies
 Cautionary note: there are some lower-level explanations for
performance
 It mights just be that you have three options – middle tone is middle
option, high tone is high option.
 To counter this criticism, drive towards testing where animals must first
be trained by 2 levels, then later given an opt out task.
 However, some animals (capuchins) can be trained by 3 option task but
cannot do the opt-out task. So can do low-level task but not the
metacognitive task
 Again, criticism of interpretation: Carruthers accepts that this is implicit
form of uncertainty monitoring, but there’s nothing reflective about it
Hampton criteria
  Said we need to rule out three potential confounds when claiming that
animal has metacognition
 1. Rule out that they’re opting out by default because too much
competition
 2. Rule out that learning that the middle of the stimulus space maps to the
opt-out response
 3. Monitoring one’s own response latencies as a cue
Post-decision wagering
 Start with decision that can either be correct or incorrect, then set a bet to
how correct you were. If you are correct and bet high you get big reward,
if bet low get a low reward. If incorrect and bet high you get no reward,
bet low and you get low reward
 Rules out confounds between stimulus and response
 Monkeys – can correctly bet and wager
 Kornell et al (2007) – can hit high and low bet options. Task is to separate
stimulus that is slightly different to the rest.
 Monkeys tend to make high bet with correct judgements, above chance
correlation for different tasks: number, area, and memory. How can we
tell whether monkeys are consciously aware, or are they just using an
implicit sense?
Frontal lobes and self-knowledge
 Double dissociation of memory and metamemory
 MTL amnesia and korsakoff’s both had worse memory, although those
with prefrontal lesions had good memory
 But Korsakoff’s and prefrontal lesions moth had worsened metamemory,
although MTL amnesics had good memory
 Can get four trial types: right/wrong, confident/unconfident
 Looked at brain regions that correlated with performance – heightened
activation of medial temporal lobe
 Contrast trials of predictions of remembering vs forgetting: prefrontal
cortex
More evidence about feeling of knowing
 Ps asked to give graded feeling of knowing, 1, 2, or 3
 Lasteral medial frontal gyrus – more and more activation in prefrontal
cortex
What makes one person’s metacognition better than another?
 Ps asked to say whether first or second interval contained brighter
stimulus
 Rated metacognition from 1-6
 Isolated variation in metacognitive ability
 Blue triangles: perceptual performance – kept it constant at 70% using a
step procedure
 But substantial variation in metacognition sensitivity – some people have
terrible ability to discriminate
 How does variability relate to ind. Differences in structure?
 fMRI scan for white matter integrity. Subjects with better metacognition –
greater white matter integrity and grey matter volume (Brodmann 10)
  Japan – used functional brain imaging rather than structural (Yokoyama
et al, 2010)p
 Anterior prefrontal cortex – rostrolateral prefrontal cortex (RLPFC), also
called the frontopolar cortex.
Comparative anatomy
 APFC – involved in explicit cognition
 Lateral aPFC – support explicit
 vmPFC – implicit – no communication aspect
Role of lateral aPFC in explicit report
 activation in these areas during explicit judgements
 Temporal profile of activity
 In vmPFC – see early signal, whereas later in the trial you see the lateral
aPFC becoming activated- when asked to explicitly reflect on their
performance
 So vmPFC is more implicit and immediate
 Coupling between these regions (psychophysiological interaction
analysis) – strength of these connections/couplings are stronger in those
with good
 Is the activity between these areas correlated? E.g. is there a dissociation
in implicit and explicit?
Brain development over the lifespan
 These lateral prefrontal regions are some o the last regions to fully
develop and mature. Is this associated with changes in metacognitive
abilities?
 Studied this perceptual discrimination task in differentage groups –
adolescents vs adults. Plotted sensitivity vs age (no difference in first
order performance actually). Metacognitive sensitivity increases with age
in adolescence.
 Decrease in metacognition as you get older – decreased from age 20-45
Theory of mind task
 Child shown cartoon and asked where sally thinks the doll is
 Children pass this test at 4/5
 Quantify level of self-directed metacognition in children
 5 year olds shown bigger difference in confidence between right and
wrong trials
 Implicit metacognition – looked at how long baby searched for the toy.
 18 month babies will search much longer when they are correct, showing
they have some awareness

Journal club

Are psychiatric symptoms linked to biases in self-evaluation or decision-making?

Perceptual decision-making task
 Say which one box contains more dots
 As task becomes harder, confidence rating is lower
 Drift diffusion models to get at parameters
 Looking at responses to confidence rating questions to see if that is
related to psychiatric symptoms
 And metacognitive efficiency
 Is the sensitivity good? Are you good at telling between you being correct
or incorrect? Sensitivity calculated by d-prime
 Metacognitive efficiency is how sensitive you are relative to your
performance -> how good is your metacognition given your performance?
Meta-d’/d’
 D’ is performance, meta-d’ is the performance of your cognition. Efficiency
is the ratio of these two
 Ps then answered psychiatric questionnaires
 Used factor analysis to combine across all the question to figure out what
would be underlying the symptoms
 Eg. question around how depressed you are may also touch on an
anxious-depression – lots of overlap in symptoms people have across
these disorders – getting at something more fundamental cognitively
Results
 Drift diffusion model – model that captures accuracy and response times
in perceptual choice
 How much evidence do people need to accumulate in order to make a
decision?  could be related to different psychiatric symptoms
 People can do the task quite well
 Across all Ps, there is individual variability and this relates to confidence
 Symptoms relate to metacognition, not performance
 Looking at different behavioural measures: accuracy, confidence,
metacognitive efficiency  no relationship between these symptoms and
people’s performance
 People with anxious-depression symptoms give lower confidence levels.
People that have compulsive behaviour and intrusive thought have higher
confidence ratings. No relationship with social withdrawal.
 Anxious-depression are better at distinguishing between whether they
are correct or incorrect
 Difference between metacognitive efficiency and sensitivity? – efficiency
controls for performance in accuracy
 Psychiatric symptoms linked to specific alterations in metacognition
o Important target for therapies and prevention. Could highlight
strategies to prevent mental illness – should also target
metacognition.
o Dissociable metacognitive changes across transdiagnostic
symptom dimensions
 o Comorbidities between different symptoms? – might need
different types of therapy
o Did they look at controls?
o Can metacognitive training shift the bias?
 Exam questions:
 Are they specific to topics? Or do they involve drawing from topics?
 Is there a relationship between metacognitive abilities and accuracy in
healthy controls? General intelligence? Introception
Revision session

Try to get the gist of neuroschematic diagrams – ALWAYS correctly label the axes
with units of measurements. Sketching is VERY useful to demonstrate
understanding. DO include a figure in your essay – at least one or two per essay

Depth is important – make sure examiner knows that you have read MORE than
the abstract of the paper. Have they really shown that they conclude this beyond
reasonable doubt? Communicate depth of your knowledge to examiner for at
least some papers.

Within each of those points you should describe one or two papers in detail; you
can include other papers in less detail.

Essay plan example

How does the primary motor cortex contribute to the control of skilled
movement?
 What do we mean by skilled?
 One of the key ways is CS tract; however, there are also other areas. The
unique contribution of the CS which cannot be substituted by other brain
circuitry is… use the word UNIQUE
 Skill is not distributed equally – certain muscles exhibit more skilled
movements than others (e.g. control of speech and finger movements) M1
has particular representation of both of these areas

 1. Somatotopy – amount of representation/territory dedicated to control
is correlated with precision of movement. Synergy
 2. Corticospinal tract - Lawrence and Kuyper 69. Strick and Dunn – old
M1 and New M1 with the rabies viruses
 3. Synergistic action – can be expressed both in terms of somatotopy and
in CS way. M1 contributes to control of skilled movements by several
principles which allow integrated muscle control to perform functional
tasks. Or could work synergy into the other two points
 4. Movement vs muscles – whilst early studies purely focus of force (e.g.
Evarts), recent studies have looked at force (e.g. graziano) – animal has
repertoire – skill in the sense that it will make the movement regardless
of starting position
 Dedication of area?

Lecture 4
 Neurocomputational model of motor control
o The selection model/inverse problem
o Forward model
o Key papers: Blakemore, Wolpert, & Frith (2002)
Lecture 5
 Anatomical pathways of movement
o Peripheral motor system
 o Spinal control of movement
o Reflexes, central pattern generator
o Corticospinal tract
o Key papers: Lawrence & Kuypers (1969), Rathelot & Strick (2009)
What features (parameter, aspect) of movements are represented in the primary
motor cortex?
 Muscles (somatotopy)
 Force
 Goal of movement (Movement vs muscle controversy) (graziano)
 Not a good idea to waffle on about Lawrence and Kuyper’s CS tracing –
doesn’t present evidence as to what kind of feature. You could probably
work it in at some point
Is there free will in the medial frontal cortex?
What does the medial frontal cortex do?

BG

Affective info could be considered a fourth loop, or could be playing a role in

modulating the other loops (limbic system). Is emotion something we do or is it
something that changes how we do other things?

How is affect expressed in BG circuitry?

Basal ganglia is discussed in terms of objective or affective~

How does the circuitry of the basal ganglia explain the organization of action in
health and disease?
 Immediately: put the circuit diagram of basal ganglia in essay!! Figure
 What does “organization” mean? Give an interpretation of what you mean
by this.
 One important element of organization is the question of which action is
selected to be initiated at which time?
 Model based and model free – advanced material
 Vigour and structure – BG seems more concerned with the vigour
 “By organization I will consider the following”
 Will only get good marks if you mentioned both health and pathology
 1. Explain circuit – draw diagram, explain how the circuitry explains
organization f action. One of the most fundamental Qs – does it happen or
does it not? Regulation/initiation – brain appears to control this by
balance between excitatory or inhibitory drive. Descript in sophisticated
way how this circuit controls the organization of action.
 2. Regulation of the vigour/amount of action. Same principles can also be
used to regulate the degree/vigor of action. General feature of BG
circuitry is idea of specifying the level of action. We know a lot about this
in parkinson’s disease. Hyperkinetic or hypokinetic (talk about both)
 Turn treatment knowledge into answering question. Can give two
treatment examples that are massively different – despite the fact that
both work… Levidopa therapy. Lesion which blocks abnormal activity in
indirect pathway to reduce activity – brain stimulation. Both work and
 seem to restore motor function. So primary motor circuit is to do with
balance of activity. Can state that clinical treatment have shed light on the
mechanisms behind function because those treatments which restore
dopamine levels are the ones that have the
 3. Evidence from different treatments that restore function by acting on
the circuit at different loci
 Spiny neurons – how limbic loop and sensory loop interact in striatum –
importance of context – interaction between motivational circuitry and
motor initiatory circuitry and BG does this… and therefore this is how it
should work in health, therefore perhaps unsurprising that motor
diseases seem to involve an element of inappropriate behaviour
 An action is not merely a movement – it is a complex sequence of blah
blah blah, for actions to be functional, have to coordinate a large range of
cognitive resources all of which are done by the BG but coordinated by
different subloops
 4. Parallel loops – multiple subloops
 5. Fine coordination of movement is not determined by basal ganglia – e.g.
person signing name could still coordinate movement well – amount of
action rather than synergy. Could state what ISN’T in the BG

Do we get better marks for answering difficult questions?