Transcript Discrimination training - Webcourses

Basic Learning Processes
Robert C. Kennedy, PhD
University of Central Florida
[email protected]
12/7/15 Plan
• Ch 11 Generalization, Discrimination and
Stimulus Control (part 2)
• Ch 12 Forgetting
• Exam discussion
Chapter 11: Generalization,
Discrimination, and Stimulus Control
(continued)
Chapter 11 Vocabulary
•
•
•
•
•
•
•
Concept: Any class (i.e., group, category) the members of which share one or more defining features.
The term may also be defined in terms of the tendency to respond in a particular way toward a
particular class of stimuli.
CS-: A stimulus that is not followed by a US.
CS+: A stimulus that is reliably followed by a US.
Differential outcomes effect (DOE): The finding that discrimination training proceeds more rapidly
when different behaviors produce different reinforcers.
Discrimination: The tendency for a behavior to occur in certain situations but not others. A variation of
this is the tendency to respond in the presence of certain stimuli, but not in their absence.
Discrimination training: Any procedure for establishing a discrimination. A distinction is made between
Pavlovian and operant discrimination training.
Discriminative stimulus: In operant discrimination training, any stimulus that signals either that a
behavior will be reinforced (an SD) or that it will not be reinforced (an S∆). Some instructors object to
the word signal, so the term may also be defined as a stimulus that reliably precedes reinforcement
(an SD) or non-reinforcement (an S∆). In Pavlovian discrimination training, discriminative stimulus
refers to a CS that is associated either with the appearance of a US (the CS+) or with its absence (the
CS–).
Chapter 11 Vocabulary
•
•
•
•
•
•
•
Errorless discrimination training: A form of discrimination training in which the S∆ is introduced in
very weak form for short periods and gradually presented at greater strength for longer periods.
Excitatory gradient: In Spence’s theory of generalization and discrimination, a generalization
gradient showing an increased tendency to respond to the SD or CS+ and stimuli resembling them.
Generalization: The tendency for the effects of a learning experience to spread. In the case of
stimulus generalization, this means that a behavior tends to occur in situations different from the
one in which it was reinforced.
Generalization gradient: Any graphic representation of generalization data.
Inhibitory gradient: In Spence’s theory of generalization and discrimination, a gradient showing a
decreased tendency to respond to the S∆ or CS– and stimuli resembling them.
Matching to sample: A discrimination training procedure in which the task is to select from two or
more comparison stimuli the one that matches a sample.
Mismatching: A variation of matching to sample in which reinforcement is available for selecting
the comparison stimulus that is different from the sample.
Chapter 11 Vocabulary
•
•
•
•
•
•
•
•
•
Oddity matching: Another term for mismatching.
Peak shift: The tendency following discrimination training for the peak of responding to shift away
from the S∆ or CS–.
S+: Symbol for discriminative stimulus that signals the availability of reinforcement.
S– : Symbol for discriminative stimulus that signals the reinforcement is not available.
SD : Symbol for discriminative stimulus that signals the availability of reinforcement contingent upon
availability of an appropriate response.
S∆ .: Symbol that signals the availability of reinforcement is not available and a response will not be
effective in producing reinforcement (extinction stimulus).
Stimulus discrimination training: A discrimination training procedure in which the SD and S∆ are
presented at the same time.
Stimulus control: The tendency for a behavior to occur in the presence of an SD but not in the
presence of an S∆.
Successive discrimination training: A discrimination training procedure in which the SD and S∆ are
presented one after the other in random sequence.
Discrimination Learning
• Discriminative stimulus - A stimulus that signals the
availability or unavailability of reinforcement
• Discriminative operant - An operant behavior that is
under the control of a discriminative stimulus.
• SD - A stimulus that indicates the availability of
reinforcement contingent upon the occurrence of an
appropriate operant response.
• SΔ – A stimulus that indicates that reinforcement is
unavailable and that the operant response will be ineffective.
Discrimination Learning
• To interact effectively with our environment, we must learn to
discriminate the conditions that indicate reinforcement availability from
the conditions that do not.
• Discrimination learning involves discovering not only when reinforcement
is available or unavailable, but also when aversive events may or may not
occur.
• Discrimination training: any procedure for establishing a discrimination
Example: training a dog to salivate at the sound of a buzzer, but not at the
sound of a bell
• Discriminative stimuli: stimuli that signal different consequences for a
behavior
- S+ or SD indicate behavior will have reinforcing consequences
- S- or S△ indicate behavior will not have reinforcing consequences
Example: When lamp is on (SD) the rat presses the lever, but when the lamp is
off (S△)it does not press
Discrimination Paradigms
• Two-choice discrimination learning
– A task when the SD and SΔ are on the same stimulus
dimension.
– Responding to the SD produces reinforcement or
punishment, and choosing the SΔ leads to neither
reinforcement nor punishment.
– Sometimes the two are presented simultaneously,
and sometime sequentially.
– In this paradigm, they must choose to which to
respond.
Two-choice discrimination learning
• Research shows that initially, subjects will
respond to the SD and SΔ equally.
• With continued training, responding to the SD
increases and responding to the SΔ declines.
Conditioned Discrimination Task
 Conditioned discrimination
 A situation in which the availability of reinforcement
to a particular stimulus depends upon the presence
of a second stimulus.
 In some circumstances, a particular cue indicates that
reinforcement is contingent on the occurrence of an
appropriate response, whereas under other
conditions, the cue does not signal reinforcement
availability.
Behavioral Contrast
• Behavioral contrast
– In a two-choice discrimination task, the increase in
response to SD that occurs at the same time as responding
to SΔ declines
• Local contrast
– A change in behavior that occurs following a change in
reinforcement contingency. The change in behavior fades
with extended training.
• Sustained contrast
– The long-lasting change in responding due to the
anticipated change in the reinforcement contingency.
– Also called anticipatory contrast
Hull-Spence Theory of Discrimination Learning
• The idea that conditioned excitation first develops
to the SD, followed by the conditioning of inhibition
to SΔ.
• Discrimination learning develops in 3 stages.
– First, conditioned excitation develops to the SD as a
result of reinforcement.
– Second, nonreinforcement in the presence of the SΔ
results in the development of conditioned inhibition to
the SΔ.
– Finally, the excitation and inhibition generalize to other
stimuli.
The Aversive Character of SΔ
– Terrace (1964) suggested that behavioral contrast
is responsible for the heightened response seen
with discrimination training.
– Argued that exposure to the SΔ is an aversive
event and that the frustration produced during SΔ
periods increased the intensity of the response to
other stimuli.
– The effect of drugs that eliminate frustration
induced behavior supports Terrace’s view.
Errorless discrimination learning
– A training procedure in which the gradual
introduction of SΔ leads to responding to SD
without any errors to SΔ.
– Some discriminations are more difficult to acquire
than others.
• Pigeons can discriminate colors better than line tilt.
Nonaversive SΔ
• The behavioral characteristics found with standard
discrimination training are not observed with errorless
discrimination training.
• The peak shift does not appear.
• Drugs that inhibit frustration induced behaviors have
no effect on responding to the SD.
• Demonstrates that SΔ is not aversive.
• Argues that with errorless discrimination, the SΔ does
not develop inhibitory control; instead, subjects learn
only to respond to SD.
Discrimination
• Oddity matching/mismatching: bird may be
required to peck a disc that is different from
the standard
• Errorless discrimination training: S△ is
presented in weak form and for shorter and
shorter periods
Discrimination
• Implications:
- Helping dogs learn how to sniff out illegal
drugs
- Training rats to identify landmines through
discriminating between explosives and other
substances
Chapter 12: Forgetting
Forgetting Vocabulary Terms
•
•
•
•
•
•
•
•
•
•
Autobiographical, or event, memory: Alternative names for episodic memory.
Context: Stimuli present at learning that are not directly relevant to what is learned.
Cue-dependent forgetting: Forgetting that results from the absence of cues that were present
during training.
Declarative memory: Memory of events that can be expressed, usually in words. Also called
explicit memory. Declarative memories include semantic and episodic memories.
Delayed matching to sample: A method of measuring forgetting in which the opportunity to
match a sample follows a retention interval.
Distributed practice: Practice sessions spread out over time. Also called spaced practice.
Episodic memory: Memory for personally experienced events. Also called autobiographical or
event memory. It is a kind of declarative memory.
Extinction method: A method of measuring forgetting by comparing the rate of extinction after a
retention interval with the rate of extinction immediately after training.
Fluency: The rate of performance, typically the number correct per minute. (A measure of
learning, it can also be used to measure forgetting.)
Forgetting: Deterioration in learned behavior following a retention interval.
Forgetting Vocabulary Terms
•
•
•
•
•
•
Free recall: A method of measuring forgetting that consists of providing the opportunity to
perform the learned behavior. Some psychologists suggest that there is no such thing as free recall,
since context cues and prompts are always involved in recall.
Gradient degradation: A method of measuring forgetting in which a behavior is tested for
generalization before and after a retention interval. A flattening of the generalization gradient
indicates forgetting.
Massed practice: Learning or practice sessions separated by little or no time. Typically the term
refers to all of the time spent learning occurring in one session.
Mnemonic: Any device for aiding recall. Typically they involve learning cues that will later prompt
recall.
Non-declarative memory: Memory of things that cannot be expressed. Non-declarative memories
include Pavlovian conditioning and procedural memory. Also called implicit memory.
Overlearning: Continuation of training beyond the point required to produce one errorless
performance. Some instructors may object to the language “errorless performance,” since the
standard for mastery may not always be zero errors. The phrase may be replaced with “satisfactory
performance,” though that language is problematic because of its subjectivity.
Forgetting Vocabulary Terms
•
•
•
•
•
•
•
•
•
•
•
•
Paired associate learning: A learning task involving pairs of words or other stimuli in which the subject is
presented with the first item of a pair and is expected to produce the second item. Some authors, especially in
decades past, have considered PA learning analogous to Pavlovian conditioning, but I believe most learning
researchers today consider it an example of operant learning.
Proactive interference: Forgetting caused by learning that occurred prior to the behavior in question.
Procedural memory: Memory for procedures; the ability to perform a skill. It is a kind of non-declarative memory.
Prompted, or cued, recall: A method of measuring forgetting in which hints (prompts) about the behavior to be
performed are provided.
Recognition: A method of measuring forgetting in which the subject is required to identify stimuli experienced
earlier.
Relearning method: A method of measuring forgetting in which a behavior is learned to criterion before and after
a retention interval. Ebbinghaus was probably the first to use this method.
Retention interval: A period during which learning or practice of a behavior does not occur.
Retroactive interference: Forgetting caused by learning that occurred subsequent to the behavior in question.
SAFMEDS: An acronym for a way to use flash cards. The letters stand for "say all fast minute each day shuffle."
Savings method: Another name for the relearning method.
Semantic memory: Knowledge of the world. It is a kind of declarative memory.
State-dependent learning: Learning that depends on the physiological state (such as alcoholic intoxication)
present during learning.
Forgetting
• Forgetting: a deterioration in performance of
learned behavior following a retention interval
• Retention interval: a period during which learning
or practice of the behavior does not occur
• Usually leads to a decrease in the probability of a
behavior, but can also mean an increase in
probability
Measuring forgetting
• Free recall: individual given opportunity to perform a
previously learned behavior
Example: student is required to learn the recipe for baking a cake
and then is asked to recite it again after a period of time
• However, it is a crude way to measure forgetting
Example: student cannot recall a French word he earlier studied,
but he may be able to say things about the word such as number
of syllables, word starts with letter “f”, etc.
Measuring forgetting
• Cued recall: another variation of the free recall technique
- Present hints or prompts to increase the likelihood that the
behavior will be produced
- Relearning method: measures forgetting in terms of the
amount of training required to reach the previous level of
performance
- Recognition task: participant has to only identify the material
previously learned
Measuring forgetting
• Delayed matching to sample: task is to select from two or
more alternatives the stimulus that matches a standard,
but prevented from performing following presentation of
the sample
• Extinction method: put the behavior on extinction after a
retention interval; forgetting occurs when extinction
proceeds more rapidly than it would have immediately
after training
• Gradient degradation: Forgetting can be measured as a
flattening of a generalization gradient
Sources of forgetting
• The longer the interval between training and
relearning, the greater the forgetting
• However, time itself may not cause forgetting
Sources of forgetting
• Degree of learning: the better something is
learned, the more slowly it is forgotten
• Learning continues even after we seem to
have achieved mastery (overlearning)
Example: good indication of how well a person
remembered Spanish 20-50 years after studying
was how thoroughly he or she had learned it
Prior Learning
• Forgetting occurs rapidly when we learn
unrelated words, random digits, and nonsense
syllables
• More meaningful material is easier to remember
• Meaningfulness: related to the importance of
prior learning
• Previous learning can reduce forgetting, but can
also lead to interference with recall (proactive
interference)
Prior learning
• Paired associated learning: typically, person
lists word pairs so that when given the first
word, the participant produces the second
Example: hungry-cake. Think of word “hungry”,
then think of the word “cake”.
Subsequent learning
• Retroactive interference: what we learn
increases forgetting of previous learning
• Often studied by having people learn two or
more lists of paired associates
Changes in context
• Context: stimuli present during learning that are
not directly relevant to what is learned
• Changes in the context in which learning occurs
may affect forgetting
• The context may offer cues that aid with recall;
absence of cues may lead to performance decline
(cue-dependent forgetting)
Changes in context
• Students learn list of words either when
standing up or sitting down. Those who learn
lists while standing up recall better when
standing up; those who learn lists while sitting
down remember better when sitting down
Applications
• Eyewitness testimony
- Students more likely to report seeing a broken
headlight if asked about “the” broken headlight
than if asked about “a” broken headlight
- Students report seeing broken glass more often if
asked about cars “smashing” than if cars “hit”
each other
Applications
•
•
•
•
•
•
•
•
Learning to remember
Overlearning
Practice with feedback
Distribute practice
Test yourself
Mnemonics
Use context cues
Take a problem solving approach
Next Class 12/14: EXAM 3
•
•
Exam #3: 12/14, 7pm, RIGHT HERE!
Your overall exam grade for the class will be based on your two highest exam
scores.
–
–
–
–
–
There are 3 possible exams
Your overall grade for exams will be based on an average of the 2 highest scores.
You need only take 2 of the exams
If you only take 2 exams, your overall exam grade will be based on those 2
If you take all 3 exams, your lowest exam grade will be dropped, and your overall exam grade
will still be based on the average of the highest 2 exam scores.
– If you take all 3 exams, YOUR OVERALL EXAM GRADE WILL NOT CONSIDER THE LOWEST OF
THE 3; only the 2 highest.
•
•
•
•
100 items/2 hours
YOU MUST BRING BROWN SCANTRON
Bring your student ID or other form of identification
100 items/2 hours