ian stewart. a1 b2b1 a2 b2b1 c2c1 b2 c2c1 b1 a2a1 b2 a2a1 c1 b2b1 c2 b2b1 a1 c2c1 a2 c2c1 a2a1 c2...
TRANSCRIPT
Ian Stewart
A1
B2B1
A2
B2B1
B1
C2C1
B2
C2C1
B1
A2A1
B2
A2A1
C1
B2B1
C2
B2B1
A1
C2C1
A2
C2C1
C1
A2A1
C2
A2A1
Directly Trained Baseline Relations
Derived Symmetrical Relations
Derived Transitive Relations Derived Equivalence Relations
A B CSymmetry Symmetry
Equivalence
Transitivity
A Schematic Representation of Stimulus Equivalence
Derived Relations
A
B C
Why the interest?1. Untrained Performances
• Emergent
• Derived
• Generative
2. Difficult to explain using traditional behavioural concepts
3. Relevant to human language and cognition
Non-Humans and Derived Relations
• Apparently absent in nonhumans
• Present in all normal humans
0
20
40
60
80
100
Baboon 1 Baboon 2 Rhesus 1 Rhesus 2 Child 1 Child 2
Per
cent
Cor
rect
Performance on Tests for Symmetry Responding
Sidman et al. (1982)
0
20
40
60
80
100
Lana Sherman
Per
cent
Cor
rect
Performance on Tests for Symmetry Responding
Dugdale & Low e (2000)
Non-Humans and Derived Relations
C1
A1 A2
A goes with B and B goes with blah,
blah, blah . . . Who cares, I’m a duck!
Non-Humans and Derived Relations
If A goes with B, then B goes with
banana? Doh! Got it wrong again.
Emerges in infancy Emerges in infancy in humansin humans
0
50
100P
erce
nt
corr
ect
Learn Object-Name, Test Name-ObjectAge: 17 months
87.5%(4 pairs)
Lipkens, Hayes, & Hayes (1993)
Development and Derived RelationsDevelopment and Derived Relations
Emerges in infancyEmerges in infancy Can combine multiple Can combine multiple
relations at least by relations at least by 23 months23 months
Object Name
Sound 0
50
100
Per
cen
t co
rrec
t
Name-SoundSound-Name
Age: 23 months
90%(4 pairs)
Lipkens, Hayes, & Hayes (1993)
Development and Derived RelationsDevelopment and Derived Relations
Emerges in infantsEmerges in infants Can combine multiple Can combine multiple
relations at least by relations at least by 23 months23 months
Multiple relational Multiple relational forms develop in forms develop in infantsinfants
020406080100
Per
cen
t C
orr
ect
16 24 26 27
Age in Months
Expose Novel Name-Picture
Test Novel Picture-Name
Lipkens, Hayes, & Hayes (1993)
Development and Derived RelationsDevelopment and Derived Relations
Development and Derived RelationsDevelopment and Derived Relations
Emerges in infantsEmerges in infants Can combine multiple Can combine multiple
relations at least by 23 relations at least by 23 monthsmonths
Multiple relational forms Multiple relational forms develop in infantsdevelop in infants
Correlates with Correlates with cognitive/verbal abilitycognitive/verbal ability
0
25
50
75
100
1 2 3 4Blocks of Testing (No
Feedback)
Per
cent
age
Cor
rect
LD: No receptive
LD: Receptive
Normal
Devany, Hayes, & Nelson (1986)
Chance
Development and Derived RelationsDevelopment and Derived Relations
Emerges in infantsEmerges in infants Can combine multiple Can combine multiple
relations at least by 23 relations at least by 23 monthsmonths
Multiple relational forms Multiple relational forms develop in infantsdevelop in infants
Correlates with Correlates with cognitive/verbal abilitycognitive/verbal ability
Correlates with the Correlates with the development of specific development of specific verbal skillsverbal skills
Barnes, McCullagh, & Keenan (1990)
0
25
50
75
100
1 2 3 4Blocks of Testing (No
Feedback)
Per
cent
age
Cor
rect
Normal
Hearing Impaired:Language =>2 yrs
Chance
Hearing Impaired: Expressive NamingNo Receptive Naming
Language and Derived RelationsLanguage and Derived Relations
Derived relations Derived relations correlate with verbal correlate with verbal ability on the WAISability on the WAIS
0
20
40
60
80
100
Low Verbal Scores High Verbal Scores
Per
cent
Cor
rect
Pelez-Nouregas, O’Hora, & Barnes-Holmes (in press)
500
700
900
1100
1300
1500
1700
Me
an
Mill
ise
con
ds
Priming Among Equivalent and Non-Equivalent Stimuli
Reaction Time
Directly Trained
Symmetry EquivalenceEq. Member - Nonmember
Eq. Member - Nonsense
Nonsense - Eq. Member
Nonsense - Nonsense
Transitivity
Language and Derived RelationsLanguage and Derived Relations
Derived relations Derived relations correlate with verbal correlate with verbal ability on the WAISability on the WAIS
Derived relations Derived relations produce priming effectsproduce priming effects
Staunton, Barnes-Holmes, Whelan, & Barnes-Holmes (2002)
Language and Derived RelationsLanguage and Derived Relations
Derived relations Derived relations correlate with verbal correlate with verbal ability on the WAISability on the WAIS
Derived relations Derived relations produce priming effectsproduce priming effects
Derived relations Derived relations produce differential produce differential ERPs measuresERPs measures
McIlvane, et al. (1999)
Language and Derived RelationsLanguage and Derived Relations
Derived relations correlate Derived relations correlate with verbal ability on the with verbal ability on the WAISWAIS
Derived relations produce Derived relations produce priming effectspriming effects
Derived relations produce Derived relations produce differential ERPs measuresdifferential ERPs measures
Derived relations produce Derived relations produce neural activation patterns neural activation patterns (recorded using fMRI) that (recorded using fMRI) that resemble those involved in resemble those involved in semantic processingsemantic processing
Dickins, Singh, Roberts, Burns, Downes, Jimmieson, & Bentall (2001)
Equivalent
This model captures one of the ‘core’ properties of analogy -- the relating of relations
If derived relations provide a behavioral model of semantic relations, then the relating of derived relations to derived relations may provide a model of analogical reasoning
APPLE
SHEEPPEACH
DOG
Equivalence-Equivalence Equivalent
So What About Analogy?
Stewart, Barnes-Holmes, Roche & Smeets (2001) proposed a model of analogy as
equivalence-equivalence responding based on the abstraction of common physical properties
However, analogy is not simply relations between arbitrary relations. Analogies are useful because they abstract out non-arbitrary relations between events. In other words,
analogy has its origins in the control of behavior by non-arbitrary environmental relations
A1
D1
C1
B1
A2
B2
C2
D2
W2
X2
Y2
Z2Z1
Y1
X1
W1
Test 1 -- ‘Property Abstraction’
Test 2 -- Equivalence-Equivalence
Z1
W2W1
Z2
W1 W2
W1/Z1
Y1/X2Y2/X2
W1/Z2
Y1/X1 Y2/X1
The two central features of this theoretical and empirical model of analogy are as follows:
(i) Relations between relations
(ii) The relations to be related involve non-arbitrary properties
Stewart, Barnes-Holmes, Roche and Smeets (2002)
2 --- Training and Testing of four 5-member EQUIVALENCE relations
1 --- BASELINE Block Sorting Task
4 --- Test for the discrimination of formal similarity
3 --- EQUIVALENCE-EQUIVALENCE Testing, biasing towards color or shape
5 --- Block Sorting Task testing for TRANSFORMATION OF FUNCTION
?
?
?
Stage 1 - Block Sorting
B1
B2
B3
B4
C1
C2
C3
C4
A1
A2
A3
A4
Stage 2 - Equivalence Training & Testing
B1/C1
B2/C2 B3/C4
Stage 3 - Equivalence-Equivalence Testing
Equivalence-Equivalence
Discrimination of FormalSimilarity (“Insight”)
B1/C1
B3/C3 B2/C4
Equivalence-Equivalence Testing (Shape-Bias)
Equivalence-Equivalence
Discrimination of FormalSimilarity (“Insight”)
Color
Color Color
Color
Shape
ShapeShape
Shape
Stage 4 - Test for Discrimination of Formal Similarity
COLOUR GROUP
SHAPE GROUP
Stage 5 - Transformationof Function
Cognitive Development, Analogy and Derived Relations
• If the current model of analogical reasoning has some validity, equivalence-equivalence tests should produce outcomes similar to those observed with traditional tests of analogical reasoning
• Adults and older children readily demonstrate analogical reasoning
• Young children (4-5 year olds) do so less readily, and when they do some researchers have argued that the performances are primarily associative or thematic rather than analogical
• Recent research has examined developmental differences in equivalence-equivalence responding in order to assess the validity of the model
Stage 1 – Equivalence Training and Testing
Stage 2 – Equivalence-Equivalence Testing with No Associative comp.
B1C1
B3C3 B2C3
Stage 3 – Equivalence-Equivalence Testing With Associative comp.
B1C1
B3C3 B1C3
B1C2
B3C3 B2C3
B1C2
B2C2 B2C3
0
20
40
60
80
100
No.
Par
ticip
ants
(%
)
Equiv Equiv-Equiv
9-yr Olds Equiv Equiv-Equiv
Adults Equiv Equiv-Equiv
5-yr Olds
Participants Who Passed the Equivalence and Equivalence-Equivalence Tests
Carpentier, Smeets, & Barnes-Holmes (2002)
Experiment 1
B1
B2
B3
C1
C2
C3
A1
A2
A3
Stage 1 – Equivalence Training and Testing
Stage 2 – Equivalence-Equivalence Testing with No Associative comp.
B1C1
B3C3 B2C3
Stage 3 – Equivalence-Equivalence Testing With Associative comp.
B1C1
B3C3 B1C3
B1C2
B3C3 B2C3
B1C2
B2C2 B2C3
Experiments 2 & 3 Perhaps the 5-year olds failed to treat the BC compound stimuli as functionally equivalent to the corresponding sample-comparison configurations? Extra trial-types were inserted after (Exp 2) or before (Exp 3) a child failed an Equiv-Equiv test.
B1C1B3C1 B1C3
B1
C1 C2 C3
+
Carpentier, Smeets, & Barnes-Holmes (2002)
0
20
40
60
80
100
No.
Par
ticip
ants
(%
)Participants Who Passed the Equivalence and Equivalence-Equivalence Tests
Equiv Equiv-Equiv
Adults Equiv Equiv-Equiv
9-yr Olds Equiv Equiv-Equiv
5-yr Olds
B1
B2
B3
C1
C2
C3
A1
A2
A3
Stage 1 – Equivalence Training and Testing
Stage 2 – Equivalence-Equivalence Testing with No Associative comp.
B1C1
B3C3 B2C3
Stage 3 – Equivalence-Equivalence Testing With Associative comp.
B1C1
B3C3 B1C3
B1C2
B3C3 B2C3
B1C2
B2C2 B2C3
Experiment 4 Perhaps the 5-year olds failed the Equiv-Equiv tests because they did not encounter this particular task format until they were presented with the test? If so, test exposures to AB-AB and AC-AC matching tasks should remedy this problem.
A3B1
A3B3 A3B2
A1B1
A3B3 A3C2
Carpentier, Smeets, & Barnes-Holmes (2002)
0
20
40
60
80
100
No.
Par
ticip
ants
(%
)
Equiv Equiv-Equiv 5-yr Olds
Participants Who Passed the Equivalence and Equivalence-Equivalence Tests
B1
B2
B3
C1
C2
C3
A1
A2
A3
Stage 1 – Equivalence Training
B1
B2
B3
C1
C2
C3
A1
A2
A3
Stage 2 – Equivalence-Equivalence Testing with No Associative comp.
B1C1
B3C3 B2C3
Stage 3 – Equivalence Testing
B1C2
B3C3 B2C3
Experiments 5 - 8 Given that 5-year olds can pass an Equiv-Equiv test following exposure to AB-AB and AC-AC matching tasks, could they also pass this test BEFORE being exposed to an Equiv test? Barnes et al., (1997) demonstrated this with adults.
A3B1
A3B3 A3B2
A1B1
A3B3 A3C2
B1
C1 C3C2
B3
C1 C3C2
No Equivalence Test
No Equivalence Test
0
20
40
60
80
100
No.
Par
ticip
ants
(%
)
Participants Who Passed the Equivalence-Equivalence Test Before and After Passing the Equivalence Test
Equiv-Equiv Equiv-Equiv Before Equiv After Equiv Adults
Equiv-Equiv Equiv-Equiv Before Equiv After Equiv
5-Year Olds
Carpentier, Smeets, & Barnes-Holmes (in press)
So What?• These findings are broadly consistent with previous
research on analogical reasoning in children and adults
• Adults and older children demonstrated equivalence-equivalence responding with relative ease but the 4-5 year olds did not
• However, the 5-year olds readily demonstrated equivalence-equivalence responding following, but not proceeding, a successful equivalence test
• What does this mean?
So What?• Some developmental/cognitive researchers have argued
that when a young child solves a verbal analogy the solution is “primarily but not exclusively associative” (Sternberg & Nigro, 1980, p. 36; see also Gentner, 1989)
• Others have argued that genuine analogical reasoning “is an important building block from an early age” (Goswami & Brown, 1990, p. 207)
• The extent to which the current behavioural data support one of these positions depends on what is meant by “associative”
So What?• If associative means based on directly taught,
reinforced, or explicitly instructed stimulus pairings, then the current data refute the former position
• The 5-year olds clearly demonstrated equivalence-equivalence responding based on untaught or derived equivalence relations
• If associative means based on stimulus pairings that occur either via direct reinforcement or derivation the data appear to support the former position rather than the latter
• Almost all of the 5-year olds required an equivalence test, during which the derived “associations” could occur, before successfully passing the equivalence-equivalence test
Ian Stewart
Thus far we have provided basic behavioral models that capture some of the core features of both
analogy and metaphor
However, what about differences between these phenomena?
For example, one important difference may be that analogy is bidirectional, whereas
metaphor is unidirectional
M E T A P H O RA N A L O G Y
“An atom is like the solar system”
ATOM SOLAR SYSTEM
ATOMSOLARSYSTEM
(A)
(B)
e-
“Cats are dictators”
CATS
CATSDICTATORS
(A)
(B)
A works . . . but B doesn’t
DICTATORSObvious qualities:
Small, furryNon-obvious
qualities:
Demanding, willful
Obvious qualities:
Demanding, willful
Obvious qualities: Demanding, willful
Obvious qualities: Small, furry
Non-obvious qualities:
Demanding, willful
Modeling analogy and metaphor - An overview of progress so far
--- Barnes, Hegarty and Smeets (1997) provided a model of analogical reasoning as equivalence-equivalence
responding
--- Stewart, Barnes-Holmes, Roche & Smeets (In press) provided a more ecologically valid model of analogy as
equivalence-equivalence responding based on the abstraction
of common formal properties
--- In more recent research in the Maynooth laboratory, we have modeled the experience of “insight”
provided by metaphor and analogy
--- We are presently working with a procedure (the Relational Evaluation Procedure or REP) that will allow us to examine
multiple types of relations in the context of analogy and metaphor (e.g., unidirectional relations)