1 ABA 635 Concept Formation Caldwell College Applied Behavior Analysis Dr. Ken Reeve
2 In the beginning… In the beginning… A person learned a skill (or response or behavior) A person learned a skill (or response or behavior) And it was good And it was good And the almighty behavior analyst said… And the almighty behavior analyst said…
3 Let there be different “contexts”… … …in which the behavior should and should not occur …in which the behavior should and should not occur And the behavior analyst said, “let us now study stimulus control!” And the behavior analyst said, “let us now study stimulus control!” And this begat the notions of And this begat the notions of Generalization Generalization Discrimination Discrimination Discriminative stimuli Discriminative stimuli Prompts of all manner and kinds Prompts of all manner and kinds Simple and complex antecedent stimuli Simple and complex antecedent stimuli Prompt & Stimulus Fading Prompt & Stimulus Fading Errorless learning MORE>> Errorless learning MORE>>
4 Relevant and irrelevant stimulus characteristics Relevant and irrelevant stimulus characteristics Stimulus “features” Stimulus “features” Stimulus “dimensions” Stimulus “dimensions” Intra-, Inter-, and Extra-dimensional generalization and discrimination Intra-, Inter-, and Extra-dimensional generalization and discrimination Simple and conditional discriminations Simple and conditional discriminations Delayed match-to-sample Delayed match-to-sample Tacting (Object labeling) Tacting (Object labeling) Multiple exemplar training Multiple exemplar training Over-selectivity problems Over-selectivity problems Programming for generalization MORE>>> Programming for generalization MORE>>>
5 Concept formation Concept formation Perceptual (Natural, Fuzzy) categories Perceptual (Natural, Fuzzy) categories Relational classes Relational classes Equivalence classes Equivalence classes Relational Frame theory Relational Frame theory Etc. Etc. And the behavior analyst said, “We have a lot to study, folks!”
6 Stimulus Control controlling stimuli = antecedent events that precede operant (or respondent) behavior that affect likelihood of occurrence (we will mostly focus on operant behavior) controlling stimuli = antecedent events that precede operant (or respondent) behavior that affect likelihood of occurrence (we will mostly focus on operant behavior) A controlling stimulus (S) is said to alter the probability of an operant, in the sense that the response is more (or less) likely to occur when the stimulus is present. A controlling stimulus (S) is said to alter the probability of an operant, in the sense that the response is more (or less) likely to occur when the stimulus is present. Discriminative stimulus (S D ) or (S+) - a controlling stimulus that sets the occasion for reinforcement of an operant. Discriminative stimulus (S D ) or (S+) - a controlling stimulus that sets the occasion for reinforcement of an operant. S-delta (S Δ ) or (S-) or extinction stimulus - a stimulus that sets the occasion for extinction of an operant. S-delta (S Δ ) or (S-) or extinction stimulus - a stimulus that sets the occasion for extinction of an operant.
7 Emitted versus evoked Operants can and do occur in the absence of any eliciting stimulus (they are said to be “freely emitted”). However, when an S D comes to control occurrences of an operant (to alter its probability of occurring), then we say that the S D evokes the operant. The term evoke dictates that the operant is under the stimulus control of an antecedent stimulus.
8 Differential Reinforcement for S D or S Δ When a person responds in one situation but not in another, we say that the person demonstrates a discrimination between the situations. When a person responds in one situation but not in another, we say that the person demonstrates a discrimination between the situations. simplest way to teach differential responding (a/k/a discrimination) is to reinforce an operant in one situation and withhold reinforcement in the other simplest way to teach differential responding (a/k/a discrimination) is to reinforce an operant in one situation and withhold reinforcement in the other “Stimulus control” refers to a change in behavior that occurs when either an S D or S Δ is presented. “Stimulus control” refers to a change in behavior that occurs when either an S D or S Δ is presented. When S D presented, probability of response increases; when S Δ is presented, probability of response decreases. When S D presented, probability of response increases; when S Δ is presented, probability of response decreases.
9 Stimulus Generalization vs. Discrimination DISCRIMINATION = target behavior occurs in one situation but not in another DISCRIMINATION = target behavior occurs in one situation but not in another We discriminate “among settings, people, stimuli” We discriminate “among settings, people, stimuli” GENERALIZATION = respond similarly in different situations GENERALIZATION = respond similarly in different situations We “generalize across settings, people, stimuli” We “generalize across settings, people, stimuli”
10 Stimulus Generalization operant is emitted in presence of new stimuli (called “transfer of control”) operant is emitted in presence of new stimuli (called “transfer of control”) new stimuli presumably share common properties with the original discriminative stimulus new stimuli presumably share common properties with the original discriminative stimulus Generalization and discrimination refer to differences in the precision of stimulus control. Generalization and discrimination refer to differences in the precision of stimulus control. Discrimination and generalization are inversely related (as one goes up the other goes down) Discrimination and generalization are inversely related (as one goes up the other goes down)
11 Generalization Gradient A generalization gradient shows the relationship between the probability of response and stimulus value. A generalization gradient shows the relationship between the probability of response and stimulus value.
12 Studying Stimulus Control To study generalization gradients, the researcher may arrange the presentation of S D or S Δ so that one follows the other (called successive discrimination training). To study generalization gradients, the researcher may arrange the presentation of S D or S Δ so that one follows the other (called successive discrimination training). In alternative procedure, simultaneous discrimination, the S D and the S Δ are presented at the same time and the organism responds to one or the other. In alternative procedure, simultaneous discrimination, the S D and the S Δ are presented at the same time and the organism responds to one or the other. In conditional disctrimination, the presence of a “sample stimulus” dictates which of two or more “compariosn stimuli” organism should select. In conditional disctrimination, the presence of a “sample stimulus” dictates which of two or more “compariosn stimuli” organism should select.
13 Errorless Discrimination and Fading When the S D and the S Δ are alternately presented as in successive discrimination, the organism initially makes many errors. When the S D and the S Δ are alternately presented as in successive discrimination, the organism initially makes many errors. That is, person continues to respond in the presence of the S Δ because generalization often occurs unless you teach otherwise. That is, person continues to respond in the presence of the S Δ because generalization often occurs unless you teach otherwise. As discrimination training progresses, a differential response occurs to the S D and S Δ. As discrimination training progresses, a differential response occurs to the S D and S Δ.
14 Problems with “Regular” Discrimination Training Procedures BASIC RESEARCH IN “REGULAR” DISCRIMINATION TRAINING: pigeon taught to peck (R) a green key ( S D ) for food ( S r ). BASIC RESEARCH IN “REGULAR” DISCRIMINATION TRAINING: pigeon taught to peck (R) a green key ( S D ) for food ( S r ). Once behavior is well established, color on key is changed to blue ( S Δ ) and pecking not reinforced Once behavior is well established, color on key is changed to blue ( S Δ ) and pecking not reinforced blue and green colors are alternately presented blue and green colors are alternately presented During early training sessions, onset of extinction will generate emotional behavior that interferes with ongoing operant behavior. (Bird is Mad!) During early training sessions, onset of extinction will generate emotional behavior that interferes with ongoing operant behavior. (Bird is Mad!)
15 Problems with “Regular” Discrimination Training Procedures Extinction is an aversive procedure! Extinction is an aversive procedure! Pigeons flap their wings aggressively and will work for an opportunity to attack another bird during the presentation of S Δ ! Pigeons flap their wings aggressively and will work for an opportunity to attack another bird during the presentation of S Δ ! Birds will even peck a different key if pecking turns off the extinction stimulus (S Δ), implying that the stimulus is aversive. (this is an avoidance behavior!) Birds will even peck a different key if pecking turns off the extinction stimulus (S Δ), implying that the stimulus is aversive. (this is an avoidance behavior!)
16 Problems with “Regular” Discrimination Training Procedures Because “emotional behavior” is elicited by the frustrating trial and error, discriminative responding takes a long time to develop! Because “emotional behavior” is elicited by the frustrating trial and error, discriminative responding takes a long time to develop! So…is there a better option? So…is there a better option?
17 Yes: Errorless Discrimination Here, the teacher does not allow the organism to make mistakes by responding to the extinction stimulus S Δ. Here, the teacher does not allow the organism to make mistakes by responding to the extinction stimulus S Δ. Errorless discrimination involves gradually introducing the S Δ initially at a very weak “intensity” so that responding to it is very low in probability. Thus, person is actually practicing NOT responding to it. Errorless discrimination involves gradually introducing the S Δ initially at a very weak “intensity” so that responding to it is very low in probability. Thus, person is actually practicing NOT responding to it. Over repeated trials, intensity of the S Δ gradually increased. Over repeated trials, intensity of the S Δ gradually increased. Eventually S Δ is presented in full intensity and the person will not respond to it. Eventually S Δ is presented in full intensity and the person will not respond to it. So, a discrimination between the ( S D ) and the S Δ was acquired without the errors of responding to the S Δ. So, a discrimination between the ( S D ) and the S Δ was acquired without the errors of responding to the S Δ.
18 FADING When some stimulus ALREADY occasions a response, but we want to have a DIFFERENT stimulus occasion the response, then we might use a FADING PROCEDURE When some stimulus ALREADY occasions a response, but we want to have a DIFFERENT stimulus occasion the response, then we might use a FADING PROCEDURE PROMPT = an “artificial” S D that we use to make a behavior likely to occur when the “natural” S D is ineffective PROMPT = an “artificial” S D that we use to make a behavior likely to occur when the “natural” S D is ineffective We then gradually reduce (FADE OUT) the prompt so that the control of the behavior transfers to the “natural” S D We then gradually reduce (FADE OUT) the prompt so that the control of the behavior transfers to the “natural” S D Sometimes we FADE IN the “natural” S D Sometimes we FADE IN the “natural” S D
19 More on FADING Sometimes we can “morph” an “artificial” S D into the “natural” S D Sometimes we can “morph” an “artificial” S D into the “natural” S D This can be done with morphing software. This can be done with morphing software. Often done where a picture is morphed into a word over successive trials so that the word will control a verbal label such as CAT. Often done where a picture is morphed into a word over successive trials so that the word will control a verbal label such as CAT.
20 COMPLEX STIMULUS CONTROL Refers to situations in which the behavioral function of an antecedent stimulus changes depending on the presence of another stimulus. Refers to situations in which the behavioral function of an antecedent stimulus changes depending on the presence of another stimulus. That is, in some situations, a stimulus might be an S D and in other situations the S Δ That is, in some situations, a stimulus might be an S D and in other situations the S Δ This is called CONDITIONAL DISCRIMINATION LEARNING because the correct response is “conditional on” (dependent on) another antecedent This is called CONDITIONAL DISCRIMINATION LEARNING because the correct response is “conditional on” (dependent on) another antecedent A CONDITIONAL DISCRIMINATION is an “IF- THEN” rule A CONDITIONAL DISCRIMINATION is an “IF- THEN” rule
21 CONDITIONAL DISCRIMINATION = MATCHING TO SAMPLE Because in conditional discrimination training, the conditional stimulus is referred to as a SAMPLE and the choices we respond to are called COMPARISON STIMULI, this is also called MATCHING TO SAMPLE (MTS) because you are trying to make the correct response in the presence of the sample Because in conditional discrimination training, the conditional stimulus is referred to as a SAMPLE and the choices we respond to are called COMPARISON STIMULI, this is also called MATCHING TO SAMPLE (MTS) because you are trying to make the correct response in the presence of the sample Did you know? A multiple-choice exam question is a type of matching to sample or conditional discrimination trial! Did you know? A multiple-choice exam question is a type of matching to sample or conditional discrimination trial!
22 EXAMPLE OF CONDITIONAL DISCRIMINATION For a child, the stimulus of MOMMY being present is usually an S D for talking to her But WHAT IF mommy and the child are in church or temple? Note that the correct response to mommy (being quiet) is conditional on (dependent on) the setting in this case. What instructions would REVERSE the behavioral functions of the stimuli?
23 Traditional cognitive psychology talks about the internal mental rules defining what things “go together” and why; they assert that the mental rule controls behavior Traditional cognitive psychology talks about the internal mental rules defining what things “go together” and why; they assert that the mental rule controls behavior Behavior analysts refer to concepts as “sets of stimuli that occasion a common response” Behavior analysts refer to concepts as “sets of stimuli that occasion a common response” Those “sets of stimuli” can be incredibly complex such as exemplars of “immoral conduct” Those “sets of stimuli” can be incredibly complex such as exemplars of “immoral conduct” The “common response” can also be complex The “common response” can also be complex Concept Formation
24 Behavior analysts identify the characteristics of the stimuli in the “concept” that controls the behavior Behavior analysts identify the characteristics of the stimuli in the “concept” that controls the behavior “One does not have a concept, just as one does not have extinction – rather, one demonstrates conceptual behavior, by acting in a certain way” (Keller & Schoenfeld, 1950, p. 154). “One does not have a concept, just as one does not have extinction – rather, one demonstrates conceptual behavior, by acting in a certain way” (Keller & Schoenfeld, 1950, p. 154). Behavior analysts examine how training (discrimination training, programming for generalization) affects the likelihood of stimulus classes (see work of Reeve, haha!) Behavior analysts examine how training (discrimination training, programming for generalization) affects the likelihood of stimulus classes (see work of Reeve, haha!) Concept = stimulus class
25 PERCEPTUAL CLASS = stimuli in the set share some physical characteristics PERCEPTUAL CLASS = stimuli in the set share some physical characteristics RELATIONAL CLASS = stimuli in the set share some abstract relationship, such as examples of “bigger than.” RELATIONAL CLASS = stimuli in the set share some abstract relationship, such as examples of “bigger than.” EQUIVALENCE CLASS = stimuli do NOT share any physical characteristics. Stimuli belong together just because society says so. EQUIVALENCE CLASS = stimuli do NOT share any physical characteristics. Stimuli belong together just because society says so. Example: 1 = one = spoken word “One” Example: 1 = one = spoken word “One” Stimulus Class Types
26 Non-Human Studies on Natural Concept Formation Herrnstein - Pigeons Herrnstein - Pigeons Pigeons can learn to form concepts of a person as discriminated from non-persons, i.e., statues, mannequins, etc., a specific person from others, etc. The abstract stimulus class of “person” readily learned. Pigeons can learn to form concepts of a person as discriminated from non-persons, i.e., statues, mannequins, etc., a specific person from others, etc. The abstract stimulus class of “person” readily learned. Pigeons can learn to form natural concepts of persons, trees, fish, etc., with a high degree of accuracy. Artificial concepts such as buildings also learned. Pigeons can learn to form natural concepts of persons, trees, fish, etc., with a high degree of accuracy. Artificial concepts such as buildings also learned. Pigeons appear to respond to two-dimensions objects (photographs) as representations of three dimensional objects Pigeons appear to respond to two-dimensions objects (photographs) as representations of three dimensional objects
27 Benefits of Conceptual Behavior Something for nothing Something for nothing New behavior occurs without training (behavior is “emergent”) New behavior occurs without training (behavior is “emergent”) Reduced teaching load Reduced teaching load Less need to explicitly teach to everything Less need to explicitly teach to everything Economical Economical Saves time and effort Saves time and effort “Psychic” power! “Psychic” power! If you learn a concept, you can make predictions about something even if you’ve never encountered that specific something before If you learn a concept, you can make predictions about something even if you’ve never encountered that specific something before