Mazur Chapter 9: Stimulus Control & Concept Learning Flashcards
(36 cards)
Probe Trials
Stimuli are presented to measure the responding but no reinforcer is given
Method of measuring generalization gradients
What causes generalization gradients?
Pavlov
Innate
Generalization is an automatic byproduct of the conditioning process
Conditioning somehow spreads across to nearby neurons in the cerebral cortex
What causes generalization gradients?
Lashely and Wade
Learned through experience
Explicit discrimination training along the dimension in question is necessary before the typical peaked generalization gradient is obtained
e.g. person needs to be reinforced when a certain color is delivered but not when it is absent
Without discrimination the individual would respond just as strongly to all colors
nondifferential training
Every trial is the same
Produces a flat generalization gradient
e.g. same response rate for all frequencies
presence-absence training
Produces typical generalization gradients
Intradimensional training
same stimulus (e.g. tone) just different dimension ( e.g. tone frequency)
Produces narrower generalization gradients
e.g.frequency is only reliable cue.
Sensory deprivation and generalization
Sensory deprivation can sometimes produce peaked generalization gradients even when animals have no
prior experience with a particular stimulus dimension
In other words, experience may not always be necessary to obtain
Simultaneous discrimination procedure
Present two stimuli together (a dark gray card & a medium gray card) and the animal/human must choose between them
Differences in opinion if learning is absolute or relational [refer to other flashcards]
absolute theory of stimulus control
animal has learned about the 2 stimuli separately
e.g. the medium gray gives me food, but the dark gray does not
Relational theory of stimulus control
animal learned about the relationship b/w the 2 stimuli
it knows that the lighter gray card is associated with food
Transposition
transfer of the relational rule
e.g. if medium gray (which before predicted reinforcer) and lighter gray are presented→ they chose the lighter gray
they realized that the lighter gray always gets the food
Peak Shift
Shift in the generalization gradient in a direction away from the S-
Spence’s Theory of Excitatory & Inhibitory Gradients
Accounts for both transposition & peak shift
In intradimensional training
excitatory generalization gradient develops around the S+ inhibitory generalization gradient develops around the S-
Spence’s theory does not do a good job predicting the results on a test called intermediate size problem
Some situations respond to the absolute properties of stimuli, however, considerable evidence points to and favors the relational approach to stimulus control
Spence: associative strength
The ability of each stimulus to elicit a response
Spence: net associative strength
Determined by subtracting a stimulus’ inhibitory strength from its excitatory strength
Behavioral Contrast
Phenomenon in which responding to one stimulus changes as a result of a change in reinforcement conditions for another stimulus
Gutman and Behavioral Contrast
Gutman’s study used a type of successive discrimination procedure known as multiple schedule: 2 or more reinforcement schedules that are presented at one time, in an alternating pattern, and each schedule is associated with a different discrimination stimulus (light & noise)
1st trial- light & noise were presented alternating on the same VI schedule, and eventually response rates were the same
2nd trial-extinction trial for the noise and eventually response rates reduced for the noise stimulus
surprisingly although the reinforcement schedule did not change for the light, the response rates increased
This is called behavioral contrast
Gutman found a Positive contrast- b/c it involved an increase in responding during the unchanged light component
If the light stimulus had decreased in responding during the unchanged light component it would be called-negative contrast.
Behavioral reallocation hypothesis
Rapid responding in the unchanged component is possible because of the slower responding that occurs in the component that is changed to extinction
This gives the subject time to recover from fatigue, and they can respond faster to the unchanged component
reinforcer habituation/satiation hypothesis
The more frequently a reinforcer is presented over a short priod of time, the less effective it becomes, because of habituation, satiation, or both
Errorless Discrimination Learning
The learner typically makes few or no responses to the S-
Rather than waiting for steady responding to the S+, the experimenter introduces the S- in early in the training
Fading: An effective way to decrease the number of responses to the S – and improving the learners long-term discrimination performance
Terrace example of Fading
Terrace knew that pigeons do not usually pack at a dark key, so at first, the S- was not illuminated green key but a dark key
Using a feeding procedure, he gradually progress from a dark to a dimly lit green key, and over trials intensive the green light was increased
In other words, the S- was introduced early in training, presented very briefly at first, and it was initially a stimulus that was unlikely to elicit responding
Errorless Learning: education application
Pavlov explains that if we do not want our children to avoid school, then we need to make sure that they do not make incorrect responses b/c it could be aversive
Used fading to teach children with down syndrome shapes
First the cards only had a correct shape and two blank spaces
then very small incorrect shapes were added and over trials the sizes were increased.
Drawbacks of errorless learning
children may have difficulty learning discrimination reversal in which the S+ and S- are reversed.
may also have trouble generalizing
Learning Set
If an animals is given many discrimination tasks, one after another, each of which have a different pair of stimuli as the S+ and S-, they get faster and faster at solving the problem
Different species vary on their ability to develop new learning sets
Animals higher on the phylogenetic scales develop stronger learning sets.
Higher species have the ability to acquire more abstract info and can recognize similarities btw problems at developing a behavior strategy that improves performance on subsequent problems.
A learning set can also develop in discrimination reversal procedures- subject first acquires one discrimination and then the S+ and S- are switched. they learn quickly to switch to other stimulus
difficulty doing this may be indicator of brain damage (schizophrenics have trouble).
