Transcript Matching to Sample and Language

Verbal Behavior
An Operant Perspective
Defining language and verbal behavior
• Language typically refers to linguistic behavior
• By definition, must have several functions
–
–
–
–
Uses symbols, syntax and semantics
Used to get access to items/activities
Can be used to stimulate additional symbols
Can be used to prompt social behavior from others
• Verbal behavior:
– Skinner
– Deals with performance of a speaker and the environmental
conditions that establish and maintain verbal performance
– Focus on the FUNCTION of the verbal behavior, not the form of
the language
Basic Distinctions for Verbal Behavior
• Verbal behavior =
– vocal, written, signed behavior of a speaker, writer or
communicator
– Behavior operates on the listener, reader, or observer
– Reinforcement of verbal behavior occurs in particular
settings
• Allows the communicator to affect environment
indirectly, as opposed to nonverbal behavior
(which has a direct effect)
Range of Verbal Behavior
• Verbal Operants include
– Speaking
– Signing
– Writing
• Function of the speaker functionally different
from behavior of listener
– Rule-governed behavior = listener
– Verbal behavior = speaker
Rule Governed Behavior
• Refers to effects of words in form of
– Instructions
– Advice
– Maxims
– Laws of listening
• Rules = complex discriminative stimuli
– Principles that govern sD’s govern stimuli that
control behavior of listener
Verbal Behavior
• Govern behavior of speaker
• Verbal behavior is mediated by actions of others
• Way a person speaks is shaped by consequences from the listener
• Reinforcement shapes
–
–
–
–
–
Style of speaking
Dialect
Tonal quality
Words used
Word structure, etc.
• All speakers = part of a verbal community
– Practices of a cultural/linguistic group
– Verbal behavior established and maintained by reinforcing practices of one’s
verbal community
How does Verbal Community control
social use of words?
• Social reinforcement shapes the way that
speakers use words
– Words don’t have own unique meaning, but only
meaning given by verbal community
– What we say = function of social contingencies
involving effects/consequences arranged by members
of our verbal community
• Long range consequences of speech/word use =
access to social and economic resources
Types of social word use
• To establish facts, persuade others: use words to give
accounts of own or others’ actions
• Attributions = verbal strategies to persuade others of your
current state
– I am sad  access to comfort
– Use as evidence or to prevent challenges to “factual”
statements
– Use words to support actions
– Use words to gain access to social reinforcement:
• Good behavior, then words that emphasize dispositional
• Bad behavior, then words that emphasize situational
Operant Functions of Verbal Behavior
• Two broad operant categories of verbal behavior
– Mands
– Tacts
• Mands, or manding, =
– Class of verbal operants whose form is regulated by
establishing operations
– Deprivation states, fear or aversive states
– Use to gain access to solutions to establishing operations:
• May I have a glass of water resolves thirst when get water
• Stop scaring me avoid fear stimuli
• Buy this toy for me gain access to toy
Operant Functions of Verbal Behavior
• Tacts or tacting
– Class of verbal operants whose form is regulated by nonverbal discriminative
stimuli and maintained by generalized conditioned reinforcement in verbal
community
– Descriptions of the environment, other facts
– Use to gain access to generalized conditioned reinforcement
• The sun is yellow; grass is green
• Reinforced by the verbal community (teacher says, “That is correct”)
• Look at function of the verbalization to determine if it is manding or
tacting
– Often manding hides as tacting
– “You look wonderful tonight”- description or gaining access to something?
How can we train Mands and Tacts?
• Manding relations:
– Use conditioned establishing operation (CEO):
• Blocked response CEO: impede response by blocking access to a
stimulus or event
• As child reaches for cookie- prompt “cookie”; only gets the cookie if
make the correct mand (sign or verbalization of “cookie”)
– Can make these very complex: e.g., our test is a series of mands:
• “define reinforcement”
• Only gain access to points if use correct words to answer the question
– Can also mand to make things go away (negative reinforcement)
or to avoid/reduce punishment
How can we train Mands and Tacts?
• Tacting relations:
– Speaker must emit a verbal operant whose form depends on nonverbal
discriminative stimulus AND operant class should be acquired and maintained
by nonspecific reinforcement
– Nonspecific reinforcement = reinforcer for one response exerts no stimulus
control over the form of the next response
– Thus: can use food reinforcement as long as it does not set the occasion for a
subsequent verbal response or the selection of the next stimulus
• E.g., identification of pictures: shown a picture of a tiger, say tiger, get food reward
• The food reward does elicit another dependent response
• Learning to name objects or describe relationship between objects = form
of tacting
• Manding training facilitates acquisition of tacting, but not vice versa
Complex verbal relations
• Intraverbal relations:
– Class of verbal operants regulated by verbal
discrmininative stimuli
– Verbal stimuli elicited by verbal stimuli
• E.g., counting: one, two _______
– Answers to questions: “I went out to eat today”, “oh,
where did you eat?”
– Relatively finite set of appropriate (reinforced)
responses
Complex verbal relations
• Autoclitic relations
– Form of verbal behavior that modifies the consequences produced by other
verbal responses
– Used to exert control by a nonverbal stimulus over the speaker’s tact
– Five categories:
•
•
•
•
•
Descriptive: What color is that?
Qualifying: Do you think this is enough rice?
Quantifying: Put a place setting for each person, please.
Manipulative: I will give you a cookie if you say you love me!
Relational: Which one of these matches that one?
• Controlled by motivating operations (MO): make it reinforcing for the
speaker to modify the mand, thus increasing control over listener behavior
• Get me water vs. Could you please get me some water
• Addition of “please” increases likelihood of listener bringing the speaker water
Echoic vs. Textual relations
• Intraverbal relations:
–
–
–
–
Class of verbal operants regulated by verbal discrmininative stimuli
Verbal stimuli elicited by verbal stimuli
E.g., counting: one, two _______
Answers to questions: “I went out to eat today”, “oh, where did you
eat?”
• Point-to-point correspondence between stimulus and response
defines echoic vs. textual relations
– Defined by formal similarity:
– requires that the verbal stimulus and the product of the response be
in same modality and have exact physical resemblance
– Counting with 1:1 correspondence
Echoic vs. Textual relations
• Echoic: class of verbal operants regulated by a verbal
stimulus in which there is correspondence and formal
similarity between stimulus and response
– this is a dog, say dog; dog
– imitation
• Textural:
– No formal similarity between the stimulus and response
– Reading out loud: not naming the letters, but the letter
combination creates a unique word.
– Silent reading assumes that there is textural relations, but
the verbalization has become silent
Matching to Sample and
Language
Is metacognition or metalinguistic
awareness necessary?
Language
• Words act as symbols
• Verbal humans can
– Manipulate symbols
– Map words onto internal concepts
– Use words to “refer” to objects, events, relations
• Behavioral perspective:
– Meaning of word established through direct contingencies embedded
in interactions with verbal community
– Contingencies largely social
• Symbols = discriminative stimuli
– Symbols = stimuli that are “thrown together” with other stimuli
– Is bidirectional!
Equivalence Class
• Three defining relations:
– Reflexivity:
• generalized identity matching
• Matching novel stimulus to itself
– Symmetry
• Functional reversability of conditional relation
• If A then B; if B then A
• Occurs without direct reinforcement
– Transitivity
•
•
•
•
Three stimuli: A, B, C
A=B
A=C
Therefore, B = C
Equivalence Class
• Stimulus equivalence:
–
–
–
–
Symbol and referents form functionally substitutable elements
Relation between symbol and referent not unidirectional
Deal with verbal or symbolic activity
Picture of a dog = word dog = picture of a dog
• Many animals show stimulus equivalence
–
–
–
–
–
–
Monkeys
Chimps and bonobos
Parrots
Dogs
Pigeons to lesser degree
Sea animals
• What cognitive abilities are necessary for this?
How test “concept formation”
• Concept =
– problem solving strategy that is based on relations
between stimuli
– NOT strategy based on particular aspects of individual
problems
– Start with # of exemplars then applied to novel problems
• Use matching to sample
– Shown an exemplar
– Pick the matching concept from stimulus array
– Sameness-different-ness
Testing animals for Concepts
• Several important criteria for testing across species:
– Exclusion effect: novel vs. familiar
• Correct answer = novel stimulus
• Are you shaping “choose the new” or “choose the concept”?
– Effects of novelty:
• Can be disruptive
• Is it the stimulus or the novelty that the animal is responding to?
• Using large pool of stimuli helps reduce this effect
Pigeons:
• Maki and Hegvik (1980) directed forgetting
– Assume that updating of memory is critical
– Human data suggest that this depends on mnemonics
• Directed forgetting = cueing what to forget
• Can animals do this?
– Use MTS task again
– Now add a delay: DMTS
Pigeons:
• Method:
– 6 pigeons
– 3 key conditioning chamber; Center key lights up white; peck it
– Peck would then result in one of two equally probable events:
• 2 sec access to grain
• 2 sec with no stimuli presented
– Used different delays: 6-15 seconds
– Then 2 keys light up: Red and Green
• Red reinforced if trial begun with NO food
• Green reinforced if trial had begun with NO stimulus
presented
• Had to remember first event: if correct, got food; if
incorrect, got TO
Training
• Training:
– Group “light”:
• comparison stimuli omitted for trials containing the house
light during delay
• Cue to remember : dark
• Cue to forget: light on
– Group “dark”:
• comparison stimuli omitted for trials containing NO house light during
delay
• Cue to remember: Light
• Cue to forget: Dark
– Ran probe tests on last 20 days: Total of 40 F-cue and 40 R cue
probe trials
Results for early Pigeons
• Obtained Mean percentages correct for F (forget) cue
and R (remember) cues, dark and light, and short or
long delays
• Results:
– Decrease in matching accuracy in F-cue probe trials
relative to R-cue trials for both Part A and Part B training
• Remembered less when cued to forget!
– R-cue trials were more accurate than F-cue, particularly
when F-cue was house light and not darkness
Experiment 2
• Examined effects of cuing and the predicted time course of
cuing
• Also examine feature positive vs. feature negative effect
• Method:
–
–
–
–
–
6 birds again
Trained on basic task with no delay
Trained to flashing vs. steady houselight
Trained to dark vs. light
Added probe trials
Results of Exp 2 with pigeons
• Again, performance during house light as cue for
forgetting was worse compared to house light as cue to
remember
• Matching following forget cues was less accurate than
following remember cues
• Delay decreased performance
• Presence of House light as forgetting cue was
disruptive!
Conclusions:
• Cuing effects can vary with nature of to-beremembered sample (remember the feature
negative effect!).
• Did NOT support the rehearsal hypothesis, but
appears pigeons “did something else” when
prompted to forget
• Suggests must engage in mediating behaviors to
maintain remembering
Memory without Awareness in Pigeons?
•
Metamemory?
– People can report state of memories, don’t know if animals can
– Why is reporting on memory important?
•
•
•
•
Adaptive
Provides functional guide to how to remember
Memory monitoring = form of Memory awareness
How test metamemory:
– Test of sensitivity to memory strength: choose whether to take a memory test
•
Rejecting the test = less reward than taking test
– Used Direct test of memory (DMTS)
– Animals with strong memory awareness should choose to take test more often
•
Who has shown metamemory
–
–
–
–
–
–
Humans
Monkeys
Other primates
Dogs
Rats!?!
Now test pigeons!
Metamemory experiments
• Several experiments with 3 pigeons
• Exp 1: birds chose between fixed small # of pellets and
varying probabilities of larger # of pellets
– Fixed = escape: nNO test
– Variable = larger reward for remembered or none for miss
(error)
• Over time all birds chose constant reinforcer and
switched away from higher variable payoff
– No differences in memory
– Just a difference in preference
Metamemory experiments in pigeons
• Experiment 2:
– Trained on task, then tested memory:
– Given the option of a test or no test
– If chose no test got an easy discrimination task rather than time out
• Examined choice and accuracy during choice testing vs. forced
trial testing
– Predicted that less memory = easy discrimination choice when had
choice
• Results:
– No effect on memory: all pigeons remembered regardless of forced
or choice test
– Found longer delays = poorer performance regardless of forced or
choice
– Pigeons still preferred to not take the test!
Metamemory experiments in pigeons
• Exp 3:
– Now present choice to escape simultaneously
– Again examined forced vs. choice testing and length of
delay
– In general, more accurate when choice test, but not
significant
• Exp 4:
–
–
–
–
Had pigeons rate their confidence level for “answers”
Trained on conditional discrimination (DMTS)
Again, choice vs. forced and the delay interval examined
Results:
• None of pigeons chose to gamble after incorrect response
• Ambiguous choice for gambling after correct responses
• Accuracy dropped with longer delay intervals, but no affect on
confidence
Summary of pigeons and metamemory
• Bottom line: little evidence for metamemory in
pigeons, even with high accuracy for DMTS task!
• Can perform the task, BUT appear to be
“unaware” of how strong or accurate their
memory might be
• Unlike other higher mammals and some birds
– Interesting because of strong performance
– But absence of metamemory
Sea Lions
• 2 female Sea lions: Rocky and Rio
• Procedure:
– All initial stimulus training = problems with 2 stimuli
– Novel stimuli always paired with novel stimuli
• Large number of stimuli used
– Allowed assessment of novelty effects
– Each comparison stimuli had an equal probability of appearing
as S+ and S-: maintains conditionality (A=B or B=A)
• Hypothesis: Identity matching experience that is gained by
completing one test will facilitate performance on
subsequent tests.
Experiment 1
•
At beginning of trial: sample stimulus exposed for 4 sec,
then 2 side doors opened revealing 2 choices (S+ and S-)
–
–
•
Training:
–
–
–
•
Additional experience
Dependence on context eliminated
Minimize control by other unintentional stimuli
Testing: 30 novel stimuli in 15 paired problems
–
–
•
Exclusion phase: correct vs. familiar, familiar = STrial and error: all novel stimuli, trained until 90% correct
Reshuffling: any stimulus could appear with any other within
the concept
Why reshuffling?
–
–
–
•
Simultaneous conditioning
Nose poke was operant response; Sr = fish
4 training sessions
Then test stimuli
Assessment:
–
–
First trial performances compared with chance (50%)
Test trials into 2 groups:
•
•
•
–
Trials 1-4 (trial 1 alone, too)
Trials 5-8
Compared these 2 groups compared to baseline
Examined 4 test groupings as a test to be passed or failed: 3/4
Results of Sea Lion MTS
• Training:
– Rio learned first problem more easily (90% or higher)
– But: Initial probe trials with novel stimuli , she scored
only at chance (50%).
• Reshuffling:
– No decrement in performance for either sea lion
– Showed transfer of training between stimuli
– Appeared to be responding according to identity
relationship
Reshuffling: Test 1:
– Rio:
•
•
•
•
•
first set NOT better than chance
But further training: 90% or better
Treated novel problems differently than familiar
Interestingly, vocalized and touched novel stimuli more
Novel stimuli appeared to disrupt Rio’s behavior
– Rocky:
•
•
•
•
•
No apparent reaction to novelty of stimuli
Test 1 was ambiguous
Test 2 was improved
But: overall, performed well on both tests
In general: performed as well as during baseline
– Pass/fail analysis:
• Rio passed all
• Rocky passed 14/15 then 15/15
Experiment 2
• Assess degree to which generalization of
matching rule would occur with stimluli
previously encountered in nonidentity context.
– Retrained sea lions on arbitrary MTS task that they
previously had extensive training on (changed the
pairings)
– Tested on MTS using familiar stimuli from old task
• Same procedure, but different pairing of stimuli
• Test with familiar, not novel stimuli
Results of Experiment 2 with Sea Lions
• Pass/Fail analysis:
– Rocky: 10/10 for first set; 8/10 for second set
– Rio: 7/10 and 8/10
– Not as many 100% as in novel training
– Did show reflexivity among elements previously
related only to dissimilar or nonmatching stimuli
• Suggests that there was some interference
from previous training
Conclusions regarding Sea Lions
•
Sea lions were able to transfer identity concept to novel problems in visual MTS
test
•
Why successful?
– Large number of exemplars
– Extended training
– Reshuffling phase allowed problems to be broken up, reducing reliance on unintentional
attributes….This way, selecting only for the identity relation
– Habituated to novelty
•
Individual differences apparent:
– Experience
– “intelligence”
•
How transferable is MTS? May depend on “abstractiveness” of concept
– Context important: what can the abstract concept be grounded with?
– MTS not necessarily limited by precise context that is learned
DTMS and human children
• 12 children in 3 groups (MA 14-36 mos)
– Normally developing preschoolers
– Mentally retarded with near typical language
– Mentally retarded with no language development
• Stimuli:
– 4 conditional discriminations:
•
•
•
•
•
If A then B
If D then E
If A then C
If D then F
Matching made up animal like figures using MTS
DTMS and human children
• Training: Presented A or D as sample, B,E or C,F as comparisons
– 3 stimuli presented on paper
– Sample at top, two choices at bottom
• Test: equivalence indicated by matching B and C or E and F
– B or E as sample with C and F as comparisons
– C and F as sample with E and B as comparisons
•
•
•
•
Each child taught and tested individually
Reward = short activity or treat
Did use visual prompting
Obtained interobserver agreement and reliability estimates
Language and kids: Results
• Looked at group and individual data
• Data graphed as percentages of unprompted
correct responses in blocks of 10 consecutive
trials
• Performance varied across 3 groups:
– Typical and retarded/language required fewer trials to
mastery
– About 100 for typical; 225 for retarded/language; 500
for retarded/no language
Language and kids: Results
• Equivalence test:
– Individual data
– Did track # of no responses made by child during each block of 10 trials: did
not differ x group
– % of correct responding = # correct responses/total number of responses in
block
• Normal group: 84.5% correct
• Retarded/language: 78.25%
• Retarded/no language: 44.5%- very close to chance
• Typical and retarded/language children improved across the equivalence
testing phase:
• Normal: 77.5 to 95.5
• Retarded/language: 69.75 to 88%
• Retarded no language: 46.25 and 39.25%
Language and kids: Results
• Data suggest that language/symbol use may be necessary
for development of stimulus equivalence in young children
– Not that couldn’t learn discriminations
– Couldn’t learn conditional discriminations under these
conditions
• Literature shows can learn with overtraining
• Seems to be lack of symmetrical responding rather than
inability to show transitivity
• Slower to learn overall
• May just take longer
Conclusions
• Which comes first: equivalence class learning or symbol use?
– Animal data suggest equivalence class
• Pigeon data: could do task, but not aware
• Sea lion data: better transitivity and symbol use
– Higher mammals, primates, dogs show transitivity and symbol use
• Is language learned, innate?
– Is it a process that requires multiple inputs from genetics,
environment
– Synergistic interactions between nature and nurture?
• But is it required to discriminate complex stimuli?
– Answer seems to be, depends on the type of complex stimuli!