Generalization, & Discrimination Learning

Generalization

Once learned, behavior often appears in situations other than the learning situation
 The closer the new situation is to the one in which the behavior is learned, the more probable the behavior will appear
- When graphed, this is called the generalization gradient

 People often have difficulty generalizing, particularly
- Problem solving skills
- Drawing analogies
 can be improved by training in different situations
 Punishment also has generalization effects

Discrimination Learning

 Learning when not to do something
- Learning when a response is not necessary (Classical Conditioning)
- Learning when a behavior is not appropriate (Operant Conditioning)
 The idea here is to get the organism NOT to generalize (at least not to that stimulus/situation)

Pavlovian Discrimination Learning

• Involves 2 “conditional stimuli”
   CS+ is regularly paired with the US
   CS- is not paired with the US

• CS+ comes to elicit the CR, but CS- would not

• As long as the organism can perceive the difference between the CS+ and CS- they can be very similar

• The similar the CS+ and CS-, the steeper the generalization gradient

Operant Discrimination Learning

• An antecedent stimulus (situation) acts as a cue that reinforcement will be given for perform the behavior
   This stimulus (or situation) is denoted S+
  

• Another stimulus (or situation) (usually the opposite) indicates that reinforcement will not occur, regardless of the performance of the behavior
   This stimulus is denoted S-

Operant Discrimination Learning Procedures

• Successive discrimination training
   Some trials are S+ trials, some are S- trials

• Simultaneous discrimination training
   The organism is given a choice between two options
  > One option has the S+ indicator
  > The other has the S- indicator

• Matching to sample
   The organism is shown a stimulus (S+)
   Task is to select the match from 2 (or more) comparison stimuli

Differential Outcomes Effect (DOE)

• Procedure
   Different behaviors result in different reinforcements
   The S+/S- indicate which behavior will be reinforced

• Finding
   As long as the behavior/reinforcement pairing is consistent, organisms show improved discrimination learning

Stimulus Control

• After discrimination learning has occurred, the behavior is said to be under stimulus control
   That is, the behavior is performed in the appropriate situation (S+) and not the
  inappropriate one (S-)

• If we know what stimuli “control” (or influence) our behavior, then we can change
  our behavior by eliminating or increasing those stimuli

Spence’s Theory

• The CS+ produces an excitatory generalization gradient
   increased probability or strength of CR

• The CS- produces an inhibitory generalization gradient
   decreased probability or strength of CR

• The inhibitory probability or strength (represented by a negative number) is added to the excitatory probability or strength to get the actual probability or strength of a CR (given a CS that is between the CS+ and CS-)
   Predicts the peak shift phenomenon

Peak Shift Phenomenon

 If an individual is trained on a discrimination task, they are likely to respond more strongly (shift the response peak) to a stimulus that is further (away) from the S- than the original S+
 Example
- Pigeon trained to peck a key when a light of 550nm (S+) (green) is shown
- Pigeon trained to discriminate between 550nm and 590 nm (S-) (orangish yellow) show increased responding to light of 540nm
- If S- is 555nm there is almost no response to the original S+ but increased responding to lights of 540nm and 530nm
 When asked to make a discrimination, organisms “err on the
side of caution”

Lashley-Wade Theory

 Generalization occurs because the organism has NOT learned to discriminate between the original stimulus/situation and the new stimulus/situation
 Steepness of a generalization gradient influenced by the organism’s experience with that type of stimuli before/during training