Pain, Predictive Coding, and the Placebo Effect

Question 1

Name two different models that have been proposed regarding perception by the brain. (2)

Answer 1

Passive reactionary process

Active anticipatory process

Question 2

Describe the ‘passive reactionary process’ model of brain perception. (3)

Answer 2

The brain passively absorbs sensory input,

processes this information in some way,

and then reacts by controlling motor and autonomic responses to these ‘passively experienced’ sensory stimuli.

Question 3

Describe the ‘active anticipatory process’ model of brain perception. (3)

Answer 3

Based on the intended goal of a person’s actions (movement),

the brain actively searches for and anticipates the information that it expects to be present in that particular environmental setting.

The end goal, and a person’s motivation to action that outcome, will influence how the brain perceives and reacts to constantly changing environmental and sensory stimuli.

Question 4

Fill the gaps relating to perception by the brain. (4)

The active anticipatory process model of perception incorporates the activity of neural systems that regulate …………………….., ………………………, and …………………. pathways.
These pathways are partly driven by the neurotransmitter ……………………….
The pathways are necessary to achieve a desired outcome.

Answer 4

attention

motivation

reward

dopamine

Question 5

Fill the gaps relating to perception by the brain. (3)

The goal of the brain is to ……………………. the information that will be given, …………………. unwanted stimuli, and ……………….. the action that it will take.

Answer 5

anticipate

filter out

optimise

Question 6

Fill the gaps relating to perception by the brain. (4)

The methods that the brain uses to process sensory information may have evolved to ………………. identify and focus on events and stimuli in the outside world that are …………………….. or …………………….. and to ……………… to them accordingly.

Answer 6

quickly

surprising

unpredictable

adapt

Question 7

Describe what is meant by a generative model. (3)

Answer 7

The brain generates models of the world

and continually uses data input from sensory organs

to update, refine, and optimise these generative models.

Question 8

Give another name for a surprising or unpredictable stimulus, when thinking of perception as a process of probabilistic inference and Bayesian information processing. (1)

Answer 8

Statistical irregularity

Question 9

Fill the gaps relating to perception by the brain. (8)

The brain has an idea of what the world should look like. It makes ………………………. based on the probabilities and patterns of events occurring in the world, and these models are constantly being ……………………. as the brain is exposed to more …………………..
This means that in unfamiliar situations, the brain can immediately focus in on the ………………………..
The constant updating of models means that the brain can more accurately ……………………… and not be ……………………………

When the brain encounters something that it did not expect and did not predict, a ………………………….. is sent to the brain. This signal then helps to update the ……………………….

Answer 9

generative models

updated

sensory input

unexpected aspect

predict the future

surprised (by the unexpected)

prediction error

generative model

Question 10

The brain is constantly trying to predict future sensory input. What is the signal called which is sent to the brain when we encounter something that was not predicted? (1)

Answer 10

Prediction error

Question 11

What is meant by ‘probabilistic inference’? (3)

Answer 11

The process of calculating the probability of a certain event happening

depending on prior and current evidence.

The evidence is constantly being updated based on new sensory stimuli and experiences.

Question 12

Why does perception by the brain use probabilistic inference to predict the future? (2)

Answer 12

To optimise future motor commands

necessary to achieve a desired outcome.

Question 13

Fill the gaps relating to probabilistic inference and perception by the brain. (6)

Perception is a process of probabilistic inference. ………………….. in the brain need to compute and ……………………. the expected statistical …………………….. (otherwise known as ……………………. patterns) of the outside world.
Thus, the brain applies a probabilistic model to generate …………………….. about future events.
The brain may then generate motor commands based on ………………………………… to occur.

Answer 13

Neural networks

predict

regularities

recurring

predictions

the most (statistically) likely events

Question 14

Is perception a true reflection of reality, or a mixture of reality and what we determine to be the most ‘statistically plausible’ event given our prior experiences in similar situations? (1)

Answer 14

Don’t know - this is the question that scientists are trying to answer.

Question 15

When looking at a graph representing Bayesian probability and information processing, what are the curves called? (1)

Answer 15

Probability density functions

Question 16

When looking at a graph representing Bayesian probability and information processing, what does the area under the curve always equal? (1)

Answer 16

1

Question 17

When looking at a graph representing Bayesian probability and information processing, what is on the Y axis? (1)

Answer 17

Probability

Question 18

When looking at a graph representing Bayesian probability and information processing, what are the three elements represented by the curves? (3)

Answer 18

Prior belief

Posterior belief

Likelihood

Question 19

What is meant by ‘prior belief’? (1)

Answer 19

How we expect the world to behave in a given situation (the brain’s generative model).

Question 20

What is meant by ‘likelihood’ in Bayesian information processing? (1)

Answer 20

The sensory stimuli being received, which will determine the likelihood of an event happening the way we think it will.

Question 21

What is meant by ‘posterior belief’? (1)

Answer 21

The updated belief, based on the prior belief and the sensory stimuli (evidence; likelihood). The event that is perceived.

Question 22

In Bayesian information processing, what is the difference between the likelihood and the prior belief called? (1)

Answer 22

Prediction error

Question 23

Describe what is meant by ‘precision’ in Bayesian information processing, and how this is shown on the graph. (2)

How does the precision affect the posterior belief? (1)

Answer 23

How accurate the prior or sensory information is.

More precise = taller curve and less precise = flatter and more spread out curve

The posterior belief will be weighted towards either the prior or sensory stimuli depending on which is more precise.

Question 24

Describe the precision of the prior expectation and the weighting of the posterior belief in a completely novel situation (in Bayesian information processing). (2)

Answer 24

Imprecise prior

Posterior weighted towards sensory input

Question 25

Describe what is meant by the following sentence:

‘Our brains use probability (Bayesian inference) to predict the future.’

(2)

Answer 25

The brain integrates prior beliefs with incoming sensory information

to predict the future and decide how best to optimise our movements and act.

Question 26

Fill the gaps relating to information processing by the brain. (7)

The probabilistic model that the brain uses is thought to adhere to ……………………. theorem.

The Bayesian brain can be conceptualised as a ………………….. machine that constantly makes ……………………… about the world and then updates them based on information that it receives from the …………………….

The theorem permits the brain to compute an ………………… probability that something is true, based on the ………………….. probability of something being true with the addition of ………………………………

Answer 26

Bayes’

probability

predictions

senses

updated

old

new information

Question 27

In an example of Bayesian information processing in the brain, a person touches a normal-looking fence, but it is actually electric.

What was the prior belief? (1)

What was the prediction error? (1)

What is the likely posterior belief? (1)

Answer 27

This fence will not cause pain

The action potentials produced in nociceptive neurones that were not predicted

Some fences are electric

Question 28

Describe the advantage for the brain of it sending down predictions and not having to process every bit of sensory information. (2)

Answer 28

Processing all information would require too much energy

so by processing only the information that it did not predict, the brain is more energy efficient and can save energy for the future.

Question 29

Fill the gaps relating to Bayesian information processing. (2)

Bayesian learning allows us to minimise the ……………………….. that reach the brain.
Any of these signals that do reach the brain will help to update …………………………

Answer 29

prediction errors

the prior belief system

Question 30

Describe the posterior belief in terms of probability of the hypothesis and the evidence. (1)

Answer 30

P (H/E)

(Probability that the hypothesis (prior) is true given the evidence).

Question 31

When observing a string or sequence of evidence regarding a hypothesis, describe how the posterior and prior change while observing the evidence. (2)

Answer 31

For every new piece of evidence, the old posterior becomes the new prior.

Therefore, the prior (our model of the world) changes with each step.

Question 32

Describe how Bayesian computation in the brain is linked to learning. (3)

Answer 32

Bayesian computation facilitates learning

as the brain ensures our priors are as accurate as possible

by using new evidence to update our predictions.

Question 33

Fill the gaps relating to Bayesian computation in the brain. (6)

Our memory may actually be a ………………… of the brain trying to ……………………..
Our accounts of past experiences may exist in our memory as a byproduct of using them to …………………………….

The brain is always trying to ……………………………….. and be one step ahead. If we encounter something we have never seen before, we have to update our ……………………………. for next time. This is a form of …………………………..

Answer 33

byproduct

predict the future

update our beliefs about the world

predict the future

predictions

learning

Question 34

Fill the gaps relating to Bayesian information processing in the brain. (5)

The brain is thought to be capable of computing …………………………. and making precise predictions with the help of ………………….. that it receives.

Through the coordinated firing of ……………….., the brain represents sensory information in the form of ……………………………

This relates to ………………………., where a group of neurones is activated and produces a neural code. The brain then has to predict the exact sensory input that has caused the particular firing pattern that it receives.

Answer 34

Bayesian probabilities

neural input

neural networks

probability distributions

population coding

Question 35

In the context of Bayesian information processing and population coding, what do neuronal tuning curves represent? (1)

Answer 35

The response of that neurone to different stimuli.

Question 36

Describe what is meant by the ‘free energy principle’ in the context of information processing. (5)

Answer 36

The free energy principle is a theoretical framework

that considers the brain to be a prediction machine

that uses information from previous experiences (memory)

to predict future events (intelligence)

in order to reduce surprise or uncertainty about the outside world.

Question 37

Fill the gaps relating to information processing in the brain. (4)

The brain continuously updates its prior beliefs (also known as ………………………………..) about the outside world based on the ………………… it receives from the ………………………
This is a form of …………………………..

Answer 37

generative models

neural signals

sensory organs

learning

Question 38

Fill the gaps relating to information processing by the brain. (2)

Learning and memory have evolved to more precisely ………………………., which from an evolutionary perspective, is beneficial and necessary for the organism’s ……………………………..

Answer 38

predict the future

continued survival

Question 39

Fill the gaps relating to the free energy principle. (3)

The free energy principle focusses on the brain’s ability to minimise …………………… and optimise its ……………….. by constantly updating its ……………………… about the future.

Answer 39

prediction error

actions

predictions

Question 40

The so-called ‘free energy principle’ posits that perception and action selection are governed by one overarching objective, which is…

(1)

Answer 40

To avoid surprises and minimise prediction errors

Question 41

Fill the gaps relating to the Bayesian brain. (2)

The Bayesian brain attempts to maintain homeostasis at a ‘set point’ of ………………………….. that remains free of ………………………….

Answer 41

neural network activity

prediction errors

Question 42

Briefly describe the two ways that the brain has of avoiding prediction errors under the free energy principle. (2)

Answer 42

Prediction fulfilled by choosing appropriate action.

Brain uses surprise as a teaching signal to adjust prior beliefs.

Question 43

Describe how the brain may minimise prediction error by choosing appropriate actions. (2)

Answer 43

• Moving the sensory organs (eyes, limbs, body) to parts of the environment where the sensory inputs better match the predictions
• Eg. keeping eyes on road while driving to avoid surprises

Question 44

Describe how the brain may minimise prediction error by using surprise as a teaching tool to adjust prior beliefs. (3)

Answer 44

Learning or updating generative model

so that the current prediction error is explained away

and more accurate future predictions become possible.

Question 45

Describe what is meant by ‘predictive coding’. (3)

Answer 45

Instead of the brain computing all of the neural information that is present in the sensory data it receives from sensory organs

it may be preferable to only represent the prediction error

which is the difference between what the brain predicts will happen and what actually happens.

Question 46

Define ‘prediction error’ and give a formula for it. (2)

Answer 46

The difference (or maybe ratio) between the sensory input and the predicted signal (i.e. the prior belief).

Prediction error = input - prediction

Question 47

Fill the gaps relating to predictive coding. (3)

One reason the brain may use predictive coding schemes is that, if the prediction is correct, very little ……………….. is wasted and costly spikes (i.e. …………………………….) are kept to a minimum, thus improving the ……………………… of the brain.

Answer 47

energy

energy intensive action potentials

computational efficiency

Question 48

By what cellular mechanism does the brain send out predictions? (1)

Answer 48

Action potentials

Question 49

Describe how predictions from the brain interact with pain pathways. (4)

Answer 49

• Predictions from higher centres activate endogenous pain modulatory systems
• Which release 5HT and NA onto the spinal cord interneurones
• The interneurones can then release GABA and enkephalins
• To inhibit activity of second order neurones and prevent a large number of nociceptive action potentials going to the brain

Question 50

Give a name for the nociceptive signals in the pain pathway that travel up to the brain because they have not been predicted and inhibited. (1)

Answer 50

Prediction errors

Question 51

What happens in the pain processing pathway if the brain did not make a prediction about pain? (2)

Answer 51

• No descending control
• Lots of action potentials (prediction errors) reach the brain regarding pain

Question 52

Describe a hypothesis relating to spontaneous brain activity and predictive coding. (1)

Answer 52

Spontaneous activity measured in the brain may actually be the brain making predictions and sending the predictions down to the spinal cord.

Question 53

What is the role of ascending feed-forward neural pathways in the context of predictive coding? (1)

Answer 53

Carry stimulus-related information from the outside world.

Question 54

Fill the gaps relating to predictive coding. (9)

Long-range feedback connections from ………………….. and local laterally-projecting interneurones populations provide natural ……………………………. for processing external sensory input.

The brain is thought to control ……………………… (predictions) that cancel out the expected ……………………. sensory input, and the difference between these neural signals, represented as changes in the …………………………. of primary projection neurones, is called the ……………………..

The errors (EPSPs) sum together at the …………….. to generate ……………….. that encode (or amplify) only the ……………………… signals for the brain to process using Bayesian methods.

Answer 54

higher cognitive centres

neuroanatomical check points

inhibitory signals

excitatory

membrane potential

prediction error

soma

action potentials

surprising

Question 55

Fill the gaps relating to predictive coding in the spinal cord. (7)

Primary afferent fibres bring in ……………………… to depolarise second order neurones.
Descending pathways release ……………….. and ……………… onto interneurones.
Interneurones release …………………. and …………………… onto second order neurones.
These signals will partly cancel each other out, so only the ……………………. is transmitted to the brain.

The neurones that modify the dorsal horn response (e.g. second order neurones, interneurones) are together called the ………………………….

Answer 55

glutamate

serotonin

noradrenaline

GABA

enkephalins

prediction error

neuronal ensemble

Question 56

Fill the gaps relating to generative models. (6)

Predictions are made on the basis of an …………………….. of the world.
The brain essentially uses ……………….. to predict the underlying causes of that input.
The better this internal model fits the world, the better the ………………… that can be made, and the lower the ……………………..
Minimising prediction error, therefore, drives a neural circuit to improve its …………………….. of the world.
In other words, minimisation of prediction errors drives …………………….

Answer 56

internal model

sensory input

prediction

prediction error

representation

learning

Question 57

What is meant by an ‘overfitted’ generative model? (1)

Answer 57

A model that is too specific based on individual experiences, and may not apply to other situations.

Question 58

What is the benefit of having a generalisable generative model rather than an overfitted generative model? (2)

Answer 58

More applicable to lots of different situations.

Can extrapolate to unknown situations.

Question 59

What is the accuracy in Bayesian models? (1)

Answer 59

Difference between prior and posterior beliefs.

Question 60

Give two drawbacks of the predictive coding hypothesis of brain information processing. (2)

Answer 60

• Does not explain how predictions (prior expectations) are computed in the brain
• Does not explain how prediction errors should ultimately be used by the brain to update its prior knowledge

Question 61

Give two drawbacks of the Bayesian inference hypothesis of brain information processing. (2)

Answer 61

• Does not specify the underlying neurophysiological mechanisms that regulate probabilistic programming in the brain
• Does not specify how similar sensory situations might be represented and differentiated in higher cognitive centres at the level of neural network activity

Question 62

Fill the gaps relating to predictive coding and Bayesian inference. (5)

Predictive coding describes the differences in neuronal activity between ………………… and …………………. populations of neurons in the …………………, whereas Bayesian inference describes the end-result of computing the ………………… of prediction errors generated by ……………………

Answer 62

higher

lower

sensory pathway

salience

sensory input

Question 63

What is the functional unit of the spinal cord ad brain that captures, processes, transforms, transmits, and stores information? (1)

Answer 63

Nobody really knows, however it could be neuronal ensembles

Question 64

How are neuronal ensembles formed? (1)

Answer 64

Through Hebbian plasticity mechanisms (LTP and LTD)

Question 65

Describe how Hebbian plasticity leads to formation of neuronal ensembles. (2)

Answer 65

Synapses between co-active neurones are strengthened

by repeated exposure to the same stimulus.