Transcript Stimulus Discrimination Training

Learning Factors in Stimulus Control
Learning Factors
• Why does stimulus generalization occur?
– CS transfers to other stimuli with similar physical properties (Pavlov)
– Inability to discriminate between stimuli (Lashley & Wade)
• Stimulus Discrimination Training
– Role of experience in producing stimulus control
– A stimulus discrimination procedure can establish stimulus control
• Experience with CS-US establishes the CS as the discriminative stimuli for the US
• Experience with (S-O) establishes when the reinforcer is available
– Example using eyeblink conditioning
•
•
•
•
•
•
Within subject design
Half of the trials were A+ (2000 Hz tone) signals followed by US
The other half were B- (8000 Hz tone) signals without US
Figure 8.5
Some generalization between A and B tones at first
Then gradual discrimination between the tones
Auditory Stimuli
2000 HZ
8000Hz
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
Stimulus Discrimination Training
• Everyday experiences illustrate the role of learning
• Red light, Green Light, Yellow light combination on traffic lights
• Vanilla, Chocolate, Strawberry etc. flavors of ice cream
• Cabernet Sauvignon, Pinot Noir, Chardonnay, Merlot, Zinfandel wines
• Discrimination training with instrumental conditioning
• Terminology
• Stimulus follow by the reinforcer is S+
• Stimulus not followed by the reinforcer is S-
• Procedures
• Sequential discrimination training
• Alternating trials of S+ or S• Need to keep them in short-term memory to compare them
• Simultaneous discrimination training
• S+ and S- are presented together
• Easier to make the discrimination so it is easier to learn
Effects of Discrimination Training and Stimulus Control
• Experience can produce very fine discrimination between stimuli
– Use generalization gradients to test effects of discrimination training
• Jenkins & Harrison (1960, 1962)
– Trained pigeons to respond to tones
• Grp 1: S+ (1000-cps) S- (no tone)
• Grp 2: S+ (1000-cps) S- (950-cps)
• Grp 3: no discrimination training 1000-cps only
– Generalization Testing with tones from 300 to 3500 Hz
• Figure 8.6
• Grp 3 (Control grp): flat generalization gradient
• Grp 1 & 2:
– stimulus control by tone frequency over responding
– steepest gradient (strongest stimulus control) was shown by Grp 2.
FIGURE 8.6
Generalization gradients of
responding to tones of different
frequencies after various types of
training. Each data point represents
the percentage of all test responses
that occurred to a particular test
stimulus. Prior to testing, one group
of pigeons received discrimination
training in which a 1,000-cps tone
served as the S+ and the absence of
tones served as the S−. Another
group received training in which a
1,000-cps tone served as the S+ and
950-cps tone served as the S−. The
control group did not receive
discrimination training before the
generalization test. (From “Effects of
Discrimination Training on Auditory
Generaliztion,” by H. M. Jenkins and
R. H. Harrison, 1960, Journal of
Experimental Psychology, 59, pp.
246–253; also from “Generalization
Gradients of Inhibition Following
Auditory Discrimination Learning,” by
H. M. Jenkins and R. H. Harrison,
1962, Journal of Experimental
Analysis of Behavior, 5, pp. 435–
441.)
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
Discrimination Training Focused on Interoceptive Cues
• External stimuli –
– Tones, lights, shapes from simple to complex in a variety of modalities
• Internal stimuli “interoceptive cues” –
– related to thirst, hunger, arousal, pleasure
– Of interest in drug discrimination studies
• Schall et al., 1996 Stimulus Control by Drug Stimuli
– Phase 1: 3.0mg/kg cocaine before pecking a key light for food
• Tested with a range of drug doses ( 0, 0.3, 1.0, 1.7, 3.0, 5.6)
• produces a flat generalization gradient
– Phase 2: Discrimination training
• S+ (3.0 mg/kg) and S- (saline) in alternating sessions
• Tested with a range of drug doses ( 0, 0.3, 1.0, 1.7, 3.0, 5.6)
• Produced a Generalization gradient around 3.0 (Figure 8.7)
– Discrimination training increases stimulus control by the internal sensations
created by cocaine
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
Discrimination Training Focused on Compound or Configural Cues
• Configural cue as a combination of two or more simple stimuli
– Pizza sign with colors, shapes and words
– Can configural cues act as stimulus control?
• Procedures used to test configural cues
– Positive patterning
• Reinforce stimuli presented simultaneously (AB+) (noise, flashing light)
• Do not reinforcer either of them separately (A-) or (B-)
– Negative patterning
• Do not reinforcer stimuli presented simultaneously (CD-) (tone, steady light)
• reinforcer each of them separately (C+) or (D+)
– See Table 8.2
– Harris 2008
• Four different stimuli, noise, tone, flashing light, steady light as CS
• Six types of trials (see above) intermixed
– Responded more to (AB)with positive patterning
– Responded more to (C) or (D) with negative patterning
– See Figure 8.8
tone
noise
Flashing light
Steady light
Noise + Flashing light
Tone + Steady Light
Rats in a sign tracking procedure
CS ----- US “Food delivered into food cup”
CR was nosing the food cup
Intermixed six different trial types
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
What is learned in discrimination learning?
• Spence’s theory of discrimination learning
• Do animals learn about S+, about S-, or both S+ and S-?
– S+ rule: "respond only when S+ is present"
– S- rule: "stop responding only when S- is present"
– S+/S- rule: "respond to S+, and not to S-"
• Excitatory conditioning of S+
– Excitatory stimulus generalization gradient centered around the value of S+
• Inhibitory conditioning of S– Inhibitory stimulus generalization gradient centered around the value of S-
What is learned in discrimination learning?
• Drug discrimination study Kearns 2005 see figure 8.9
– Experimental group
• Tone (S+) or Clicker (S+) with Lever pressing on VI schedule for cocaine
• Light (S-) no reinforcement
– Control group
• Tone/light combination with Lever pressing on VI schedule for cocaine
– on half of the trials
– Light no reinforce on half of the trials
– Because is Light only on half of the trials it is not S-
– Summation Testing
• Tone only
• Tone / light combination
– combination of tone and light reduces responding in experimental group
– Evidence that light (S-) has inhibitory properties
• Similar to inhibitory conditioning (CS-) in classical conditioning
– Suggests that drug addiction could be treated with cues that are negative
discriminative stimuli.
• Any reliable cues that predict drug is not available
• Could be context “place” cues
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
Interaction Between S+ and S-: Peak-Shift Effect
• Interdimensional discrimination training uses different stimulus
dimensions (e.g., color vs pitch, time vs number)
• Intradimensional discrimination training uses same stimulus
dimension (e.g., red vs green, 1000-cps tone vs 950-cps tone)
– which results in generalization gradients with a special feature – “peak shift”
• Hanson (1959) Generalization Testing see Figure 8.10
– Discrimination Training: with different colors of light
• Grp1: S+ (550-nm) S- (555-nm)
• Grp2: S+ (550-nm) S- (590-nm)
• Grp3: S+ (550-nm) only i.e. no discrimination training
– Results
• Grp3: excitatory gradient centered at 550-nm
• Grp2: high rate of responding to 550-nm but peak of responding displaced to
540-nm
• Grp1: lower rate of responding to 550-nm than Grp2 or Grp3, and a shift in peak
responding to colors between 530-540-nm
FIGURE 8.10
Effects of intradimensional discrimination training on stimulus control. All three groups of pigeons were
reinforced for pecking in the presence of 550-nm light (S+). One group received discrimination training
in which the S− was a 590-nm light. For another group, the S− was a 555-nm light. The third group
served as a control and did not receive discrimination training before the test for stimulus
generalization. (From “Effects of Discrimination Training on Stimulus Generalization,” by H. M. Hanson,
1959, Journal of Experimental Psychology, 58, pp. 321–333.)
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
Spence’s Explanation of Peak-Shift
• displacement of responding in a stimulus generalization gradient
away from the S+ in a direction opposite the S• It is a function of the similarity of S- to S+ and results from any
intradimensional discrimination
– Excitatory stimulus generalization gradient centered around the value of S+
– Inhibitory stimulus generalization gradient centered around the value of S– Actual responding to a stimulus value on the dimension reflects generalized
excitation to that stimulus value minus generalized inhibition to that stimulus
value
– The exact shape of the excitatory and inhibitory gradients has a big impact on
predictions
– Peak shift requires that the excitatory gradient overlap with the inhibitory
gradient (S+/S- similarity, and spread of the distributions)
Alternative Accounts of Peak-Shift
• Viewing stimuli as single elements or as configural cues
• Spence's explanation is based on absolute properties of individual
stimuli
• Alternative view – Relational
– processing the relationship between or relative difference of several stimuli
– for example discrimination based on the difference between S+ (1000 Hz) vs
S- (950 Hz)
– When S+ is a large circle and S- is a small circle learn to discriminate based on
relative size
– For complex discrimination when there is no easily defined scale such as
wine tasting
Stimulus Equivalence Training
• Difference versus Equivalence
– Discrimination training can make animals respond differently to stimuli that
they would normally respond to in the same way
– Can a training procedure be developed that would make animals respond
similarly to stimuli that they would normally respond to differently?
• Learning procedures to increase stimulus generalization is stimulus
equivalence training.
• Perceptual concept learning which will be covered in chapter 12
– Concepts like "water" "tree" "fish" .etc
– Reinforce the same response to different photos containing the perceptual
category but not to photos in which the category is absent
• Reinforce any photo of tree but not reinforce photos of water or fish
Stimulus Equivalence Training
• Functional equivalence training: Table 8.3
– Initial Training
• For different types of stimuli in set A
• another set of four designs in set B
• reinforced for making the same response (lever pressing) to any of the stimulating
in set A or set B
• this should establish equivalency between the two sets
– Reassignment
• Designs in set A are also trained with a different response (pull a string) for
reinforcement
– Test
• Test the patterns from set B with the alternative response pulling on a string
• if they associate pattern B with pulling the string
• evidence of equivalents between set A and set B
The Principles of Learning and Behavior, 7e by Michael Domjan
Copyright © 2015 Wadsworth Publishing, a division of Cengage Learning. All rights reserved.
Equivalence Class
• An equivalence class has the following three properties
– Reflexivity: A = A, B = B, C = C.
– Symmetry: A --> B, B --> A.
– Transitivity: A --> B, B --> C, A --> C
• For example; if A is the word “Apple” ; if B is an apple; if C is a
picture of apple
• Equivalence training can be used to improve verbal skills of
individuals with intellectual disabilities
• verbal labels in human language facilitates equivalence class
formation
– Transferring a type of response from set a to set B is an example of emergent
relations
– can be used to improve verbal skills
• if a rose is a flower and a flower is a plant than a rose is a plant
• Which is an example of an emergent transitive relationship