Learning and Motivation Lecture 3 - Stimulus Control

Thorndike’s Law of Effect

• satisfying outcomes “stamp in” the connection between the stimulus and the response — increases the likelihood of particular behaviour

• association grows as result of the reinforcement

• so next time in presence of the stimuli/context — trigger this instrumental response.

  

Stimulus Response (S-R) Learning

• instrumental behaviours (responses) are controlled by stimuli they are associated with

• discriminative stimulus

Discriminative Stimulus

signal that an action will produce a consequence or outcome (e.g. peck or turn to the pigeon) 

Skinner’s Tripartite Contingency - ABC

• Antecedent

  • the stimulus controlling behaviour

• Behaviour

  • what is the response being reinforced/punished?

• Consequence

  • what is the immediate outcome of the behaviour

Difference between Thorndike and Skinner 

• Thorndike saw role of reinforcing stimulus as stamping in the S-R association

• Skinner considered all 3 things as important he also assumed association can be learnt between the response and the reinforcing stimulus (outcome) 

Stimulus Control Theory

• our decisions and actions are controlled by S-R associations we have learned in the past (cf. experience of free will)

• if we could know an orgs entire learning history — should be able to predict behaviour based on the stimuli present

• very nurture focused

Generalisation & Discrimination

• every instance we encounter (a stimulus or event) is different

• how is a previously learned response (in particular situation) transferred to a new but similar situation — e.g pig closing door

• Generalisation

  • extent to which behaviour transfers to a new stimulus

• Discrimination

  • extent to which behaviour does not transfer to a new stimulus

Factors important for stimulus control

1) How effective the S is learned about (learning-related) 

• basic conditions necessary for associative learning 

• true for classical and instrumental conditioning 

2) How similar or different the S is to previously learned stimuli (performance-related) 

• behaviour in novel (but similar) situations 

• factors dictating generalisation and discrimination 

Generalisation and “Little Albert” Study 

• fear conditioning study 

• set out to test generalisation of leaned fear in an infant 

• US: loud clanging noise 

• UR: fear/shock

• CS: white rat

• CR: fear elicited by the rat  

Generalisation Gradient

Discrimination Learning

• can be learned through training with different schedules of reinforcement

• classical example:

  • high-pitched tone — Food US

  • low-pitched tone — no US

  • (US only contingent on the high tone and CR will differ)

• instrumental example

  • in presence of high-pitched tone: Response —> Rft

  • in prsence of low-pitched tone: Response —> no Rft

  • reinforcement is contingent on the stimulus and response

Generalisation in humans

• more complicated

• physical attributes

• semantic similarity

• rules/analogies that link otherwise dissimilar events

Razran (1939) salivation study using students

• words (CS) paired w food (US) — Salivation CR

• Trained CS: Style, Urn, Freeze, Surf

Trained responses to

• phonologically similar

• semantically similar

More generalisation to semantically similar than phonologically

— humans generalise more on the basis of meaning