Behavioural: L4 - Attention in PC (Stimuli that become predictive capture…
Behavioural: L4 - Attention in PC
Learning theories attempt to explain and predict which stimuli will be learned about, and how learning happens
Most learning theories are trial based - work out how much is learned on any one trial
Most learning theories are concerned with calculating change in associative strength on a trial– gets a number called ‘V’.
Most learning theories assume that the strength of a conditioned response is proportional to the strength of the association
The Rescorla Wagner model suggests the amount learned on any one trial is proportional to how surprising a US is.
ΔV = α(λ-V)
Change in associative strength (V) - Bigger US's (will support more learning) = big λ
α = Learning rate parameter determined by the properties of the CS
V= Current strength of the CS→US association
α is fixed according to the Rescorla-Wagner model
It is determined by things like how loud/bright/salient a CS is
Is sometimes thought of as attention paid to the CS
Important stimuli in our world can change - so the attention we pay to stimuli has to change too.
How much attention we pay to stimuli determines how much we learn about them.
Animals worse at learning about novel CS than about the familiar CS
= Also much better learning in novel group - effect of pre-exposure reduced how sick they felt.
Amount of learning on a trial is proportional to the attention paid to the stimulus
Following non-reinforced pre-exposure attention to stimulus (alpha) goes down
Hence less learning about pre-exposed stimulus
Stimuli that become predictive capture attention
Mackintosh and Little (1969) - Pigeons were trained so that one dimension was relevant, and one dimension was irrelevant
Easier to learn colour if the pigeon was trained in colour in phase one.
Easier to learn orientation if the Pigeon was trained on orientation in phase 1
Eyetracking is understood in the cognitive literature to reflect attention
Le Pelley, Beesley & Griffiths (2011)
trained people on an associative task, where the predictions of cues was manipulated, and their eye movements were tracked.
Over time, stimuli that are predictive and mean something capture our attention more.
Macintosh Learning Theory (1975)
if CS is better at predicting the outcome than anything else, then the attention paid to it (α) goes UP
If CS is no better at predicting the outcome than anything else, then the attention paid to it (α) goes DOWN
Effect of learning on Attention
Learning has evolved in such a way as to enable us to filter out irrelevant information and focus on stimuli that provide us with useful information
However... It would make sense to pay attention to stimuli that we are not sure what they mean – those stimuli that have an uncertain outcome, or which are inaccurate predictors of their outcomes
Pearce and Hall (1982)
investigated in three groups of rats to see the effects of introducing uncertainty on learning
Trained animals first that a tone predicted a mild shock
Then gave animals trials with no shock delivered – so tone no longer predicted shock.
Then tested how well animals could learn about tone, when paired with a stronger shock
The group that had the tone alone - made the animals learn even more about the tone than group that hadn't had pretrial.
Surprising outcomes get our attention.
If CS always followed by a US – attention to the CS goes down
If CS followed by a US sometimes, attention to the CS = maintained
Hogarth et al., (2008) carried out a study using eye tracking to investigate the effect of uncertainty on attention to stimuli
Participants made ratings of how likely a noise was to occur when two stimuli were presented
AX was always followed by noise
BX was followed by noise 50% of the time
CX was never followed by noise
Stimulus B (uncertain) has more dwell time than stimulus A or C – indicating the role of uncertainty in commanding attention
Pearce & Hall Theory (1980)
Associability of CS (α) on conditioning episode n (αn) is determined by the attention paid to CS
CS commands attention if in the past it has been presented with unpredicted or surprising outcomes
In many ways the two theories, Mackintosh (1975) and Pearce-Hall (1980) are in contradiction
There have been suggestions that a hybrid theory might be required, where both mechanisms are functioning and different mechanisms dominate in different situations
However we can be certain that experience and learning can change the attention that is paid to stimuli, and that the attention paid to stimuli changes what is learnt about those stimuli