1
Outline
1.
2.
3.
What is intelligence?
Can it be measured?
Differences in measured intelligence
a. Within group
a. Heritability
b. Twin studies
c. Manipulating environments to increase IQ
Between groups
Are intelligence tests culturally biased?
Are there multiple intelligences?
Spearman’s g
Intelligence in daily life
Stanford-Binet
b.
4.
5.
6.
7.
8.
2
Outline
1.
2.
3.
What is intelligence?
Can it be measured?
Differences in measured intelligence
a. Within group
a. Heritability
b. Twin studies
c. Manipulating environments to increase IQ
Between groups
Are there multiple intelligences?
Spearman’s g
Intelligence in daily life
Stanford-Binet
Wechsler tests
b.
4.
5.
6.
7.
8.
3
What is intelligence?
“Intelligence is a very general mental capability
that, among other things, involves the ability to
reason, plan, solve problems, think abstractly,
comprehend complex ideas, learn quickly and
learn from experience. It is not merely book
learning, a narrow academic skill, or test-taking
smarts. Rather, it reflects a broader and deeper
capability for comprehending our surroundings –
“catching on,” “making sense” of things, or “figuring
out” what to do.” (Gottfredson, 1994)
4
Can intelligence be measured?

Yes.

And intelligence
tests measure it
well.
5
Can intelligence be measured?

Psychologists
define intelligence
as “what IQ tests
measure”


This makes sense.
Intelligence tests
have been carefully
created, revised,
and improved for
100 years precisely
to do that job.
6
Can intelligence be measured?

Intelligence tests
are accurate (they
are reliable and
valid)

Intelligence tests
do not measure
creativity,
character,
personality, or
other individual
differences
7
Can intelligence be measured?

IQ scores from
tests such as the
Stanford-Binet and
the WAIS are the
best predictors we
have of a variety of
important
outcomes

This includes life
outcomes relating
to health, career,
personal
relationships, and
crime
8
Differences in measured
intelligence
a.
Within groups
a. Heritability
b. Twin studies
c. Manipulating environments to increase
IQ
b.
Between groups
9
Within group differences

Phenotypic
variation


Total variation in
the appearance of
members of a
species
Includes things that
can be “made
visible” (such as
blood types)
10
Within-group differences

Heritability

the proportion of
phenotypic
variation in a
population that is
due to genetic
variation among
individuals in that
population
11
Within-group differences

Heritability

If all environments
were to become
exactly equal for
everyone,
heritability would
be 1.0 because all
remaining
differences in IQ
would have to be
biological
12
Within-group differences

Heritability

If every person
were a clone (so
that everyone had
exactly the same
genetic material),
all remaining
differences in IQ
would have to be
environmental
(heritability would
be 0)
13
Within-group differences

Members of the
same family tend to
differ in IQ by about
12 points on
average.

Siblings have
different
environments and
also share only half
their genes (on
average)
14
Within-group differences

Adopted children
resemble their birth
mothers more than
their adoptive
mothers in
intelligence, even if
they have never
met their birth
mothers

Plomin et al.
(1997): no relation
at all after early
childhood between
the IQs of adoptive
parents and of the
children they
adopted
15
Within-group differences

McGue et al.
(1993): found an
average correlation
of zero for adoptive
siblings tested as
adults

Loehlin et al.
(1997) on Texas
Adoption Project:
shared family
environment
influences IQ only
for very young
children; genetic
effects increase
with age
16
Twin studies
1.
2.
3.
Are identical twins more similar in IQ
than fraternal twins?
Are fraternal twins more similar in IQ
than pairs of non-twin siblings?
Are non-twin siblings more similar in
IQ than unrelated children raised in
the same home?
17
Are identical twins more similar in IQ
than fraternal twins?

Plomin, DeFries,
McClearn, &
McGuffin (2001)
summarized results
of studies of more
than 10,000 pairs
of twins


The average
correlation of IQ
scores for identical
twins: .86
For the fraternal
twins: .60
18
Are fraternal twins more similar in IQ
than pairs of non-twin siblings?


Fraternal twins are
no more genetically
similar to each
other than any pair
of non-twin siblings
But fraternal twins
are more similar to
each other in IQ
than non-twin
siblings.

Why? Jensen
(1998):



Prenatal factors
such as mother’s
age, nutrition,
health
Blood antigen
incompatibilities
Obstetrical
procedures
Are non-twin siblings more similar in IQ
than unrelated children raised in the same
home?


Jensen (1998)
analyzed 27,000
sibling pairs
average IQ for
ordinary siblings
reared together
was .49


correlation of .25
for IQs of unrelated
persons raised
together when IQ
was measured
during childhood
this correlation
dropped to 0 when
the children were
tested as adults
19
20
Does this mean intelligence is
hereditary?

Intelligence is
partially inherited


About 50% of the
variability in IQ
scores can be
traced to genetic
influences
Differences in
environment quality
are more important
for younger
children than for
adolescents
21
Does this mean intelligence is
hereditary?

The nature of
nurture effect – part
of the effect of the
environment may
be genetic


Differences in
environments may
be produced by
genetic differences
E.g., an artist and
an engineer may
create very
different
environments for
their children
22
Within-group differences

Can we manipulate
environments to
increase IQ?



U.S. Government’s
Head Start
program
Milwaukee Project
Carolina
Abecedarian
project
23
Head Start program



Modest gains
Smallest gains for
children most at
risk
Gains do not
survive long after
child leaves the
program

Neisser et al.
(1996): by end of
elementary school,
no differences
between those in
vs. not in program
24
Milwaukee project


Modest gains
Effects do not last
long after exit from
program

Gains only on tests
requiring skills
taught in the
program – effects
do not generalize
to new tasks
25
Carolina Abecedarian project

Intensive
intervention
provided daily for
the first 5 years of
the child’s life


IQ scores for
intervention group
5 points higher
than for control
group at age 12 (7
years after exit
from program)
Positive effect on
failure & drop-out
rates
26
Are intelligence tests culturally
biased?

No.

Tests of widely
varying kinds (e.g.,
verbal abilities,
spatial abilities),
including those
considered most
“fair,” give the
same results.
27
Are intelligence tests culturally
biased?

Group differences
just as large on
Ravens
Progressive
Matrices as on
WAIS

IQ scores have
same utility for
prediction
regardless of race
or socio-economic
status.
28
Bias vs. Fairness

It’s important to
distinguish
between these two
concepts:


Bias
Unfairness
29
Bias vs. Fairness

A test is biased if it
gives a
systematically
wrong result when
used to predict
something.

So, an intelligence
test would be
biased if, for
example, it
underestimated
one group’s
probability of
success in a given
endeavor.
30
Bias vs. Fairness

Use of a test is
unfair if it treats
people differently


E.g., if a verbal test
probes for knowledge
acquired from
schooling, use of that
test with people who
have not had such
schooling would be
unfair
Note that the test itself
is not implicitly unfair –
but use of the test may
be unfair
31
Bias vs. Fairness

When you use an
unfair test, the
result need not be
biased. The result
may still have good
predictive value.


E.g., if you test
non-native
speakers of English
with the SAT, that
use of the test is
unfair, but not
biased
Results will predict
academic success
in English-speaking
countries.
32
Are group differences in IQ real?

Yes.


Members of all
ethnic/racial groups
are found at all
levels of IQ.
But groups vary in
where their scores
cluster (that is, in
the means).
IQ
It’s important to note that the group differences are in “central
tendency” (mean) – there is lots of overlap, and all groups are
represented at low, medium, and high levels of IQ
34
Are group differences in IQ real?

Highest IQ scores
are for Ashkenazi
Jews


Cochran et al.
(2006): medieval
social environment
for European Jews
selected for verbal &
math intelligence (but
not spatial)
Some relation to
disease genes?
35
Are group differences in IQ real?

Curves for some
Asians are
somewhat higher
than for Whites;
curves for Blacks,
Hispanics somewhat
lower than for Whites

We don’t know why
these effects are
found, but there is
much debate on this
question
36
Sources of between-group differences



Next two slides have statements from
leading scholars in the field on what we
know (and don’t know) about the
sources of between-group differences in
measured intelligence
Gottfredson (1997), Intelligence
Neisser et al. (1996), American
Psychologist
37
Sources of between-group differences
“There is no definitive answer to why IQ bell curves
differ across racial-ethnic groups. The reasons for
these IQ differences may be markedly different
from the reasons for why individuals differ among
themselves within any particular group… Most
experts believe that environment is important in
pushing the bell curves apart, but that genetics
could be involved, too.” Statement of the 52
experts, Intelligence, 1997, p.15.
38
Sources of between-group differences
“It is clear that genes make a substantial
contribution to individual differences in
intelligence test scores, at least in the White
population. The fact is, however, that the high
heritability of a trait within a given group has no
necessary implications for the source of a
difference between groups…Thus the issue
ultimately comes down to a personal judgment:
How different are the relevant life experiences of
Whites and Blacks in the United States today? At
present, this question has no scientific answer.”
Neisser et al., (1996), p.95
39
Sources of between-group differences


Gene-based
temperamental
factors?
Family size (now
decreasing in
N.A.)?

SES?


but differences
present when
SES controlled
Caste?

Lacking “effort
optimism”
(Ogbu, 1978)
40
Sources of between-group differences


Priming effects
Bargh, Chen, &
Burrows (1996)



IV: elderly
stereotype
DV: Walking speed
Primed subjects
walked more slowly

Dijksterhuis & van
Knippenberg (1998)



IV: intelligence/
stupidity traits primed
DV: performance on
a knowledge test
Trait primes affected
performance
41
Sources of between-group differences

Relevance of the
priming studies




Stereotypes may
influence those who
are stereotyped
Direct effect on test
performance?
Indirect effect on
effort optimism?
Both?
42
Sources of between-group
differences

Culture (Boykin,
1996)


Assessment itself
alienates?
American
schooling conflicts
with deep
structure of
African-American
culture?

Members of ethnic
groups might answer
some items
differently but still
correctly – some
non-standard
responses given by
(e.g.) minority
children may be
standard in their subculture.
43
More reading (1): People who argue
group differences are real:
Neisser, U. et al. (1996), Intelligence: Knowns and
Unknowns. Am. Psychologist, 51(2),77-101
Buckhalt, J.A. (2002). Learning and Individual
Differences, 13, 101-114.
Gottfredson, L.S. (1997). Intelligence, 24 (1), 13-23.
Gottfredson, L.S. (2000) Psychology, Public Policy,
& Law Special Issue, 6(1), 129 – 143.
Jensen, A.R. (2000). Psychology, Public Policy,
and Law Special Issue, 6 (1), 121-127.
44
More reading (2): People who argue
group differences are artifacts:
Chan, D., et al. (1997). J. Applied Psychology, 82
(2), 300-310.
Hale, J.B., et al. (2001). School Psychologist,
Fall,113-118.
Helms, J.E. (1997). In D.P. Flanagan, J.L.
Genshaft, & P.L. Harrison (Eds.), Contemporary
intellectual assessment: theories, tests, and
issues (517-53).
Steele, C.M. (1997 & 1998). American
Psychologist, 52 (6) 613-629 and 53 (6) 797811
45
Are there “multiple intelligences?”


No.
Intelligence is multidimensional, but all
intelligence tests,
whatever their
form, measure the
same ability.

This is true whether
tests emphasize
verbal or nonverbal skills, and
whether they
require specific
cultural knowledge
(such as
vocabulary).
46
Are there “multiple intelligences?”

Howard Gardner’s
idea: IQ tests tap
only one of many
different kinds of
intelligence (e.g.,
bodily-kinesthetic,
musical, interpersonal, etc.)


No evidence
supports this view.
Gardner’s tests
appears to
measure interest,
motivation, other
things.
Most cannot be
scored objectively.
47
Are there “multiple intelligences?”


Sternberg’s
Triarchic Theory is
just as bad as
Gardner’s.
No useful tests of
his “practical
intelligence” exist.

He is accused of
misrepresenting
data, using unclear
definitions,
obscuring his
methods, not
reporting results
fully, using
unrepresentative
samples.
(Gottfredson, 2003)
48
Are there “multiple intelligences?”

All intelligence
tests measure the
same ability

The most important
aspect of
intelligence is g –
general ability
49
Spearman’s g

Spearman (1904):

When a large
group of people are
given a variety of
ability tests, the
correlations among
the test results are
almost always
positive.
50
Spearman’s g

Spearman (1904):


That correlation is
called the positive
manifold
It is the basis for
the idea that there
is a general mental
ability (called g).
51
Spearman’s g

g is not the same as
IQ


to find g, you have to
do factor analysis
to find IQ, you use an
IQ test
Measurement error
Factor
Analysis
g
S
V
Test Score
IQ
53
Spearman’s g

g is essentially a
biological variable


all tests involving
cognitive ability
measure g to some
extent
none measure only
g: any cognitive test
has some variance
due to causes unique
to that test.
54
Spearman’s g

IQ scores reflect
individual
differences in
underlying
constructs (e.g., g
and group factors),
unique abilities,
and measurement
error.

In contrast, factors,
such as g, are
derived from
correlations and
reflect individual
differences in
underlying
constructs. Factor
scores provide best
estimates of g.
55
Does g matter?

Yes. g is a better
predictor of
educational and
work performance
than any other
measure we have.

g is the most
important
determiner of
scores on every
test of cognitive
ability (in people
who can be tested).
56
Is intelligence more than just g?

Yes. More than 70
different “group
factors” have been
identified.



Hierarchical theory: g
at top (most general)
At Level II, 8 broad
cognitive abilities (e.g.,
fluid intelligence)
At Level I, narrow
abilities (memory span,
inductive reasoning,
etc.)
57
Is intelligence important in daily
life?


Yes. (See Slide
58.)
IQ is more strongly
related to important
educational,
occupational,
economic, and
social outcomes
than any other
single variable.



Relation is strong
in education,
military training
Moderate in social
competence
Modest in lawabidingness
58
Is intelligence important in daily
life?

IQ increases in
importance as life
gets more complex
– in novel,
ambiguous,
changing,
unpredictable, or
multi-dimensional
situations.


IQ is important in
professions,
management
Less important
where only routine
decision-making,
simple problemsolving are
required
59
Is intelligence important in daily
life?

Some personality
characteristics,
talents, physical
capabilities, etc.
are also important.

But intelligence
transfers across
tasks and settings;
those other
characteristics do
so less or not at all.
IQ:
% populn
High
Risk
<70
5
Uphill Keeping Out
Yours to
Battle Up
Ahead
Lose
71-90 90-110 110-130
> 130
20
50
20
5
LF (M)
Job (M)
Divorce
Illegit (F)
Poverty
Prison (M)
Welfare (F)
Dropout
22
12
21
32
30
7
31
55
19
10
22
17
16
7
17
35
15
7
23
8
6
3
8
6
14
7
15
4
3
1
2
0.4
10
2
9
2
2
0
0
0
LF = Out of labor force > 1 month in the last year; Job = Unemployed >
1 month/year; Welfare = Chronic welfare recipient
61
Individual Intelligence Tests






administration requires advanced training
tests cover wide range of age and ability
examiner-subject rapport is important
immediate scoring of items
usually requires about one hour
allows opportunity for observation
62
2 Important Tests
Binet
 asked to identify
intellectually limited
children so they
could be removed
from the regular
classroom and put
in special
education
Wechsler
 Responded to
perceived
shortcomings of the
Binet test thirty
years later
63
Binet’s 1905 test


No intelligence tests
existed to guide
Binet and colleague
Simon
Like Spearman,
thought of
intelligence as a
general mental ability


Wanted tasks to
measure judgment,
attention, and
reasoning.
Two major concepts:
 Age differentiation
 General mental
ability.
64
Binet’s principles of test
construction:

Age differentiation:

Binet searched for
tasks that could be
completed by 2/3 to
¾ of the children in a
particular age group
& were completed by
fewer younger
children and more
older children.
65
Binet’s principles of test
construction:

General mental
ability:


Measured only the
total output on the
various tasks.
Judged value of task
in terms of its
correlation with the
combined result of all
other tasks.
66
Binet-Simon (1905):


First formal
intelligence test
30 items ordered by
difficulty

Drawbacks:



Output: labeled
children idiot,
imbecile, and moron
(these were technical
terms at that time)
Norms: produced
using only 50 children
Validity: no evidence
offered
67
Binet-Simon (1908):
Grouped items
according to age level
rather than simply
according to increasing
difficulty.
 Introduced concept of
mental age to deal with
problem of output



Increased norm
group to 203
children.
Still produced only
one score heavily
dependent on verbal,
language, and
reading abilities
68
1916 Stanford Binet scale:

Lewis Terman of
Stanford University
translated Binet test
into English and
introduced it to
America.

Terman increased
size of
standardization
sample, but included
only white nativeCalifornian children.
69
1916 Stanford Binet scale:


Introduced intelligence
quotient (IQ) concept to
show subjects’ rate of
mental development.
IQ = (MA/CA) x 100


Maximum mental age
was 19.5. Set maximum
chronological age at 16.
S-B tests used on
millions of U.S. Army
recruits starting in 1917,
after mobilization for
World War I – a rich
source of data for postwar research.
70
1937 Stanford-Binet Scale


Extended age range
down to 2 and up to 22
years, 10 months.
Some performance
items added – but 75%
of items still verbal



Scoring standards and
instructions were
improved
Standardization
sample improved to
include 3184 subjects
from 11 states.
Developed alternate
forms (Forms L & M) to
facilitate research
71
Problems with 1937 Stanford-Binet

Reliability higher for
older subjects than
for younger ones and
higher for those in
the lower IQ ranges


Scores were most
unstable for young
children with high IQ
Each age group also
had different
standard deviations
which made
interpretation difficult
72
1960 Stanford-Binet:

Used Binet’s
principles to redo
scale


Looked for tasks on
which success
becomes more likely
as age increases
Looked for tasks for
which scores
correlated with test
scores.
73
1960 Stanford-Binet:

Introduced the
deviation IQ concept.
Set mean at 100 with
SD of 16.


Could now compare
scores of one age
level with another.
Deviation IQ: score
worked out in terms
of number of SDs
above or below age
mean
74
1960 Stanford Binet
Each boundary (---) marks 1 standard deviation
75
1986 Stanford-Binet scale


Multi-dimensional,
but hierarchical –
with g at the top of
the structure
4 main factors:




Verbal reasoning
Abstract/visual
reasoning
Quantitative
reasoning
Short-term memory
76
1986 Stanford-Binet scale

The individual tests
(such as Pattern
Analysis or
Vocabulary) each
give information
about one of the
factors


The factors in turn
give information
about g.
g reflects common
variability across all
the tasks
77
2003 Stanford-Binet scale

Now 5 main
factors, each tested
in verbal &
nonverbal domains





Fluid Reasoning
Knowledge
Quantitative
Reasoning
Visual-Spatial
Processing
Working Memory
78
2003 Stanford-Binet scale

Uses routing tests to
estimate subject’s
level of ability


Verbal and nonverbal routing tests
Back to original agescale approach
(items with differing
content grouped
together on basis of
difficulty).
79
2003 Stanford-Binet scale

The S-B5 was
normed on a
stratified random
sample of 4,800
individuals (2 – 85
years of age) that
matches the 2000
U.S. Census.

Bias reviews on all
items for sex,
ethnic,
cultural/religious,
regional, and
socioeconomic
status issues.
80
Psychometric properties of 2003 S-B

Internal
consistency
reliability is .98 for
composite and .93.97 for area scores.

Some individual
test scores are
lower: .73 for
memory for objects
is the lowest.
81
Psychometric properties of 2003 S-B

Test-retest
reliabilities for
composite score
were .91 and .90
for 5 and 8-yearolds.


Factor analysis
supports the
structure of the
test.
Correlations with
other IQ tests are
generally in the
.70s and .80s
82
Psychometric properties of 2003 S-B

Convergent validity
assessed with:

Stanford-Binet
Intelligence Scale,
4th Edition, the
Stanford-Binet Form
L-M, the WoodcockJohnson III, the
WAIS-III, the WISCIII, and the WPPSI-R