Associative Learning

Associative Learning

Associative learning is where we link two stimuli (events) together and condition a response with them.

Pavlovian (classical) Conditioning

Pavlov's classic experiment where he conditioned his dogs to respond to a bell ringing by salivation changed the way people think about conditioning.

By defining four variables we can explain a relationship between different events, and examine how linking them together can cause a Conditioned Stimulus to predict an Unconditioned Stimulus, which elicits a Conditioned Response.

  • Unconditioned Stimulus (US) - stimulus that causes a response (biologically relevant, usually)
  • Conditioned Stimulus (CS) - neutral stimulus that is being linked to US
  • Unconditioned Response (UR) - (unlearned) response to US
  • Conditioned Response (CR) - response to CS (after learning)

Screen%20Shot%202012-10-19%20at%208.44.55%20AM.png

Necessary Relationships

  • Contiguity: The CS and US need to be temporally related (occur at the same time).
  • Contingency: The CS and US should be dependent on each other.
    • Introducing the CS without the US (pre-exposure) reduces the likelihood of them being linked. This is referred to as latent inhibition.

Screen%20Shot%202012-10-18%20at%2012.16.14%20AM.png

  • Surprise: The US shouldn't be already conditioned with something else.
    • Introducing the US without the CS reduces the learning as well. This is known as the blocking effect.

Examples of Emotional Conditioning

By linking an occurrence with an emotionally charged event - perhaps fear of being bitten, or sexual arousal - we can condition humans/animals into certain emotional responses. An example is associating a certain food, the US, which has previously had a UR of a stomach bug, with a CR of nausea.

Sexual Quails

Akins (1998) and Matthews, Domjan, Ramsey & Crews (2007) examined Quails and sexual activity, and found that mailes placed in context predictive of female access were quicker to engage in sexual activity.

Equipotentiality

The Principle of Equipotentiality is the idea that the rate of learning is independent of the combination of conditioned and unconditioned stimuli that are used in classical conditioning.

Food Avoidance Experiments

Some things appear to violate the theory of equipotentiality; Garcia and Koelling's 1966 study of rats, for instance. They found that pairing thirsty rats with one of four combinations changed the learning rates:

  1. Audiovisual CS, Shock US
  2. Audiovisual CS, Illness US
  3. Taste CS, Shock US
  4. Taste CS, Illness US

Screen%20Shot%202012-10-17%20at%2011.39.50%20PM.png

After measuring the laps per minute of each group, they found that the Taste CS, Illness US group, and the Audiovisual CS, Shock US group learnt far faster than the other two.

Shock Conditioning

After conditioning human participants that a blue square on a screen (CS+) predicted shock, and a red square (CS-) did not, participants measured a higher Electrodermal Activity (EDA) reading for the CS+ than CS-.

Shock conditioning also seems to violate Equipotentiality; phobic images are more likely to have a higher change in EDA than neutral images (e.g. spider vs mushroom).

Covariation Bias

Toumarken et al (1989) and Pauli et al (1996) studied various patients and asked them questions of how often a shock followed positive (flower), neutral (mushroom) and negative (spider) fear images - the negative images rated higher.

Observational (Vicarious Leanning)

Olsson and Phelps (2007) showed that humans can learn to fear a stimulus, even if they simply observe that it shocks others.

Gerull & Rapee studied this in toddlers, and found toddlers whose mothers gave negative responses to objects responded in a more negative manner themselves.

Animals can learn this way too (can be unconditioned to be afraid of snakes, by being shown fake negative reactions to flowers and positive reactions to snakes).

Opponent Process Model

CR doesn't always follow classical conditioning and stimulate the US->UR, sometimes it has an opposite response.

For instance, Siegel's 1972 experiment feeding rats insulin determined that after the rats learned to associate a CS with an insulin injection US, the CR was to up the BG level to maintain homeostasis, rather than reduce BG level.

The BG level (observed behaviour) was the sum of the CR and the UR (i.e. an apparent reduced reaction to insulin).

More evidence for this comes from the 1978 report; Conditioned morphine tolerance (Siegel, Hinson, & Frank).

Instrumental/Operant Conditioning

Also known as operant conditioning, instrumental conditioning is a form of conditioning where an individual's behaviour is modified by its consequences; changing in form, frequency and strength; this type of behaviour is goal directed.

Instrumental conditioning also requires contingency between response and outcome, but the contingency is not just a temporal link, it's an aversive or appetitive.

Thorndike's Law of Effect

Edward L Thorndike was the first to invent a methodology for studying instrumental conditioning.

His Law of Effect states that responses that produce a good effect in a situation are more likely to occur again, and conversely responses that produce an uncomfortable effect become less likely to occur again.

So it's not learning a link between response (behaviour) and outcome, it's learning that a response is good or bad. He hypothesised that the reward "stamped in" the correct response.

e.g. Puzzle Box learning:

Screen%20Shot%202012-10-19%20at%203.12.24%20PM.png

Types of Instrumental Contingencies

Positive: Adding something in
Negative: Taking something away
Reinforcement: Rewarding, hence making the situation more good.
Punishment: Making the situation less good.

  1. Positive Reinforcement: Giving someone candy
  2. Positive Punishment: Playing a loud noise
  3. Negative Reinforcement: Turning off a loud noise
  4. Negative Punishment: Taking away an x-box

Screen%20Shot%202012-10-19%20at%203.13.49%20PM.png

Influencing Factors

Delay

The delay between response and outcome is an influencing factor; it appears to be better late than never, and earlier being the best (higher correlation).

Reinforcement Schedule

This determines how and when a behaviour/response is followed by the delivery of a reinforcer/punishment/extinguisher.

There's two things that can be adjusted here - interval (amount of time before a reinforcer), and ratio (number of responses before a reinforcer). Either type can be fixed or variable.

Screen%20Shot%202012-10-19%20at%203.15.29%20PM.png

Applications of Principles

Infant Memories

There's a belief that events that occur early in development have a more pronounced effect.

In any case, the characteristics of infant memories have three features:

  • Resistant to extinction (hard to unlearn)
  • Context independent (apply learning to a new room, kind of thing)
  • Broadly generalised

Examples

Johanson and Hall in 1979 learnt that rats can learn associative behaviour quite young.

Fillion and Blass in 1986 reared rats with either citral-treated nipples or saline-treated nipples, and found that both rat groups responded better (faster) to a test female of the same treatment.

Screen%20Shot%202012-10-23%20at%201.03.14%20PM.png

Sullivan in 2001 studied pairing an odour CS with a mildly unpleasant US (e.g. shock) and measuring odour preference. She found out that very young rats preference the paired odour - possibly explaining an infant's attachment to an abusive caregiver.

Anxiety Disorders

Jacobs and Nadel have a model of psychopathology that suggests that many anxiety disorders have the same three prominent characteristics as infant memories; suggesting that anxiety disorders are expressions of early memories.

Pavlovian Conditioning and Criminality

Gao et al, 2010 found in a study titled Association of Poor Childhood Fear Conditioning and Adult Crime that children with poor fear conditioning (i.e. less response than expected to fear conditioning tests with electric shocks) predisposes a person to crime at 23, suggesting a link between the brian's ability to respond to fear, and disposition for crime.

Birbaumer et al, in 2005, did a study examining Pavlovian fear conditioning in psychopaths and found that they have less activation in the amygdala and orbitofrontal cortex than a control group.

Extinction

Extinction is the decrease in the amplitude/frequency of a CR as a function on non-reinforced presentations of the CS.