Pl: FCH/SPH
KB008-12
P2: FCH
KB008-Staudinger
August12,2002
of Aging'
12 Interrelations
Knowledge,and Cognitive
Performance
TimothY A. Salthouse
VA, U'S'A'
Universityof Virginia,Charlottesville,
Abstract
the relationbetweenage
This chapterreviewsthe researchliteratureconcemedwith
betweenageandcogand knowledgeand with the role of knowledgeon the relations
of experience
thataccumulation
Althoughit is generallyassumed
nitiveperformance.
withageleadstogreaterquantityorqualityofknowledge,theempiricalevidencefrom
large-scalestudieswithrepresentativesamplesofresearchparticipantsSuSgeststhat
age 18 to aboutage40 or
wh]le thereappearsto be un increasein knowledgefrom
of eitherstabilityor decline'
50, the dominanttrendin lateryearsof adulthoodis one
Amongthehypothesesdiscussedtoaccountforthelackofcontinuousgrowthinknowlgains'an asymptoteon
edgearegenerationalconfoundsin education,lossesoffsetting
of one'sknowledge'Each
""po.ur" to new information,andincreasedspecialization
hypothesishassomeplausibility,butitisconcludedthatthereasonsforthefailure
tofindcontinuousage-relatedincreasesinknowledgearestillnotwellunderstood.
relationsbetweenageand
severalconceptualmodelsof the role of knowledgeon the
cognitionarediscussed,includingmoderation,mediation'andmigration'Becauseinteractionsofageandknowledgehanebe"ninconsistent,andbecauseStatisticalcontrol
ofknowledgetendstoincreaseratherthandecreasethenegativerelationsbetween
ageand,n"uru."rofcognitiveperformance,theavailableempiricalevidenceseemsto
on cognitionappearto be
favorthemigrationinterpretation.That is, age-relatedeffects
reducedamongpeoplewithhighlevelsofknowledgebecausepeopletendto..migrate''
into higherknowledgegroupswith increasingage'
different agetrendsdatesback
The realizationthat differentcognitivevariableshave
totheearliestempiricalstudiesonaging.Forexample,severalstudiespublishedinthe
measuresof memory and reasoning
1920sfound larger age-relateddifferencesfor
(e.g',Foster& Taylor' 1920;
thanfor measuresof vocabularyandgeneralinformation
beenusedto characterize
have
labels
Hollinsworth,1927;Willougniv, DtT)' Many
26s
P1: FCH/SPH
K8008-12
P2: FCH
August12,2002
KB008-Staudinger
Understanding H uman D eveIoPment
266
0.2
Generalinformation
E
o
o
a
-o.2
E
o
E
a
o
(\t
o
q)
..\
\ l^**-*f/\t \4-';/'
\1----.J
0.0
Vocabulary(opposites)
\
-0.4
\-\.,
\t--\-..--
-uo
Analogies
-0.8
-1.0
IE
Age
Chronological
Figure 12.1
Sample
Age Relations on Three Different Cognitive Variables in a
of 678 adults'
DatafromtheJonesandConrad(1933)study'
on stored information'
the two types of cognition (such as new learning versus reliance
and mechanics
abilities,
educativeversusreproductiveabilities,fluid versuscrystallized
descriptive
most
the
be
may
versus pragmatics), but the terms process and product
assessment,
of
time
at
the
because the first term refers to the efficiency of processing
andthesecondtermref-erstothecumulativeproductsofpastprocessing(seealso
P. Baltes, this volume; Wellman, this volume)'
by Jones and conrad
The different age trends can be illustrated with data collected
(1933)onthreesubtestsfromtheArmyAlphatestbattery(seeFigure12.1).(The
performance in standard
vertical axis in most of the figures in this chapter represents
and studies' sometimes
variables
score or z-score units to facilitate comparison across
as in this figure, and
distribution
a sample of young adults will be used as the reference
In either case'
distribution'
reference
the
other times the entire sample will be used as
and
distribution'
relevant
the
of
the units along the verticafaxis are standard deviations
the
to
relative
difference
age
the
of
thus the scale is infbrmative about the magnitude
distribution of scores in the reference sample')
55 years of age the measures
Notice in Figure 12. I that across the range from 20 to
but that increased age
stable
remain
to
tend
of general information and vocabulary
same general pattem
This
measure.
is associated with lower scores on the analogies
different types of
that
accepted
widely
now
has been reported many times, and it is
trends'
age
cognitive measures can have quite different
Therearetwomajorimplicationsofthesefindings.Thefirstimplicationisthat
age relations, no single number
because the constituent variables exhibit different
ability across all of adulthood'
cognitive
will be meaningful as an index of overall
Thatis,ifcognitiveabilityiscomposedofbothproductandprocessaspects,then
Pl: FCH/SPH
K8008-12
P2: FCH
KB008-staudinger
August 12' 2002
Interrelations
of AginT, Knowledge, and Cognitive Performance
261
Jones and
it is not meaningful to refer to a single age trend in cognitive functioning'
Conrad(1933)expressedthispointelegantlybystatingthattheolderadult
than do young
derives more intellectual power from accumulated stocks of information
adults.
are at least
The second major implication of the different age trends is that there
ageaging-the
cognitive
of
field
the
in
explain
to
phenomena
two quite distinct
or increase with
stability
either
and
cognition
of
process
aspects
in
related decline
aging research has
age in product aspects of cognition. The majority of cognitive
in remediating
interest
of
an
part
because
in
cognition,
of
focused on pro.rr. aspects
focus on two
I
chapter'
in
this
However,
decline.
cognitive
or prevenung age-related
The first
cognition.
of
aspects
knowledge
or
product
and
aging
with
issuesconcemed
knowledge'
of
quantity
of
the
measures
and
age
between
relation
issue concerns the
presumably derives from
which might be expectedto be positive becauseknowledge
with advancing
experienceand experienceis often assumedto increasecontinuously
relations between age
age. The second issue is the role played by knowledge on the
greater knowledge to offset
and cognitive performance and particularly the ability of
theconsequencesofage-relateddeclineinprocessaspectsofcognition.Thisissue
(1990) selective optimization
is particularly relevant to the P. Baltes and M. Baltes
that optimizing one's
with compensatron(SOC) framework becauseof the possibility
that may be occurring in
knowledge in select domains might compensatefor declines
carstensen' this volume;
other aspectsof cognitive functioning (see also M. Baltes &
P. Baltes, this volume; cf. Lemer, Dowling, & Roth' this volume)'
of the RoIeof Knowledgeon
Conceptualizn'tions
Age- Cognition Relations
J
alternativepossibilities
Before reviewing the relevantliterature,it is important to clarify
fortheroleofknowledgeonagedifferencesincognitiveperformanceandwaysthat
little researchfocusing
they might be distinguished.Becausethere has been relatively
most of the conperformance,
cognitive
and
on the interrelations of age, knowledge,
ofresearch on
context
in
the
discussed
ceptualizationsto be describedwere originally
models' all of
or
possibilities,
Five
relations.
the role of experienceon age-cognition
are
performance'
cognitive
on
knowledge
of
which assumethat there are positive effects
to
dilTerent
correspond
figure
each
in
functions
(The
three
in Figure 12.2.
'l----fu',tesented
are (A) moderation'
lJ"iNnowledgel) The five models and their main characreristics
(C) decline' negeffects;
age
no
(B)
stability,
interaction between age and knowledge;
between age and
relation
negative
but
effects
no
age
ative age effects; (D)Lediation,
positive relation between age
knowledge; and (E) migration, negative age effects but
andknowledge.Thisclassificationschemeisnotexhaustivebecausetherecouldbe
relations between age and
positive age eff-ects,as well as different combinations of
five models portrayed in this figure
knowledge with stability or decline. However, the
possibilities to account for
appear to be the most plausible and theoretically interesting
theroleofknowledgeonage.cognitionrelationsatthecurrenttime.
Pl: FCH/SPH
K8008-12
P2:FCH
---1V-/h
KB008-staudinger
268
August12,2002
Understanding Human Development
Knowledge
d)
o
C
6
E
!
il)
o
.z
g
o
ChronologicalAge
(B)
Knowledge
o
o
o
o
c
6
E
F
E
6
High
E
Mod.
o
Low
o
!
o
0)
.z
.z
=
c
o
C)
O
ChronologicalAge
ChronologicalAge
Knowledge
o
o
o
o
C
o
E
E
o
o
E
o
o
0)
o
o
c
o)
o
!
Ch16nologicalAge
Figure 12.2
Chronological
Age
SchematicIllustrationsof Five AlternativeModels of the Joint Effects of Age and Experienceor Knowledgeon Cognitive Performance.
PanelA representsthe moderationmodel.This is the view that knowledgemoderatesthe relationsbetweenageand cognitiveperformance,suchthat thereare large
negativeagerelationsamongpeoplewith little relevantknowledgebut small to nonexistentagerelationsamongpeoplewith greateramountsof relevantknowledge.Some
versionof the moderationmodelis the most frequentlymentionedpossibilityin the
researchliteratureconcernedwith experience,andthe searchfor the predictedage-byexperience(or age-by-knowledge)
interactionhasbeeneitheran explicit or an implicit
goal in many studies.
Pl: FCH/SPH
K8008-12
P2: FCH
KB0O8-Staudinger August12,2002
Interrelations
of Aging, Knowledge, and Cognitive Performance
269
There are some cases,as with certain measures of reading or aspects of language
comprehension or production, in which an interaction may occur that is attributable to
an absolute or a functional measurement ceiling, in which almost everyone eventually
achieves nearly the same high level of proficiency. However, because of the restricted
range of assessment,the absence of age differences in these situations may be of only
limited interest. That is, if the criterion activity can be mastered by virtually everyone,
then there will likely be little or no sensitivity of the performance measuresto variables
such as age or additional knowledge. As an illustration, consider possible criterion
variables reflecting proficiency in aspectsof walking or talking. Because these tasks
are masteredby most normal people, the range of variability would likely be highly
restricted, and consequently it may be unrealistic to expect them to have relations
with other variables. The moderation interpretation is more interesting when one can
be confident that the interactions are not attributable to measurement artifacts, and yet
peoplewith high levelsof knowledge still have smaller age-relateddeclinesthan people
with lower levels of knowledge.
Panel B represents positive knowledge effects but no age effects (stability), and
Panel C represents positive knowledge effects with negative age effects (decline). In
both cases, the age and knowledge effects are additive, and it is assumed that there is
no relation between age and amount of knowledge. The same age trends are therefore
expected at each level of knowledge.
Panel D represents a mediation model in which lack of knowledge is presumed
to be responsiblefor the observedage-relateddecline. The flat lines signify that there
is no relation between age and cognitive performance at any level of knowledge, but
the ellipse indicates that increased age is typically associatedwith lower levels of
knowledge within the population. When the emphasis is on experience instead of
knowledge,this interpretationis known asthe disuseperspectivebecauseit is frequently
postulated that disuse functions as a mediator of age-relateddecline, as indicated by
"use it or lose it." The model in Panel D predicts main effects of
the admonition to
experience(or knowledge)but no age-by-experience(or age-by-knowledge)interaction
and a negative relation between age and experience (or there would be no evidence of
disuse).Furthermore,this model implies that if a measureof the amount of experience
(or knowledge) is statisticallycontrolled, then the age effectson measuresof cognitive
performance should be substantially reduced. That is, if there is no disuse (in this
case, because of statistical control), then little or no age-relateddecline would be
expected. According to the mediation model, therefore, negative age-cognition relations
are attributable to a negative relation between age and amount of relevant experience
(or knowledge).
The mediation model differs from the other models under consideration because
it assertsthat Iack of knowledge (or experience) mediates age-related decline observed
in certain measures of cognitive performance. However, it should be noted that this
interpretation is plausible only if there is a negative relation between age and relevant
knowledge (or experience).
Pl: FCH/SPH
KB008-12
P2: FCH
KBO08-Staudinger August12,2002
270
Understanding Human D evelopment
Panel (E) represents what can be termed a migration model. This model predicts
a main effect of knowledge but with no interaction of age and knowledge, implying
that the effects of knowledge are similar at all ages (or equivalently, that the age effects are similar across all levels of knowledge). The distinguishing feature of this
model is that there is frequently a positive relation between age and knowledge (represented by the ellipse superimposed on the functions for each level of knowledge),
as if with advancing age there is a "migration" of many of the individuals from lower
to higher levels of knowledge. Because knowledge is positively related to many aspects of cognitive performance, no overall relation may be evident between age and
measures of cognitive perfoffnance if the benefits of greater knowledge function to
counteract the effects of any age-related declines that might have occurred in basic
aspectsof cognitive functioning. The migration model thereforeimplies that statistical
control of knowledge would result in an increasein negativeage differencesbecause
the greater knowledge effectively serves to offset any declines in basic processing efficiency that the individuals may be experiencing. In other words, the migration model
attributes the absence of strong negative age relations in certain measures of cognitive performance to positive age-knowledgerelations.This model may be the closest
to what P. Baltes and M. Baltes (1990) refer to as optimization with compensation if
the migration into higher knowledge levels is viewed as optimization and the greater
knowledge is interpreted as compensating for declines in other aspectsof cognitive
functioning.
The mediation and migration models are similar in that in neither case is an interaction predicted betweenage and knowledge on cognitive performance,but instead
there is either a negative (mediation) or a positive (migration) relation between age
and knowledge that servesto alter the age-performance relations. Howeveq the prediction from the migration interpretation is opposite that of the mediation interpretation
becausethe age-knowledgerelation is positive rather than negative,and thus control
of knowledge should increase,rather than decrease,the magnitude of the negative
age differences in measures of cognitive performance. Stated somewhat differently, in
mediation the presence of a negative age relation is caused by a lack of knowledge,
whereas in migration the absence of a negative age relation is caused by an abundance
of knowledge.
The moderation and migration models are also similar in that both models predict
high levels of cognitive performance among older adults with high levels of relevant
knowledge. However, the underlying mechanisms in the two cases are quite different. In the moderation model, knowledge serves as a moderator of the age-cognition
relations becausethe interaction indicates that people with high levels of knowledge
exhibit smaller age-related declines in cognitive performance than people with low
levels of knowledge. This model therefore implies that the impact of aging on certain
aspects of cognitive functioning can be altered as a function of the amount of relevant knowledge. In contrast, the migration model attributes high cognitive functioning
in certain samples of older adults to their higher levels of knowledge offsetting the
Pl: FCH/SPH
K8008-12
P2: FCH
KBo08-Staudinger
August12,2002
Intenelations
of Aging, Knowledge, and Cognitive Performance
Z'/l
consequences of declines in basic abilities. Unlike the moderation model, in the migration model knowledge does not have a moderating or interactive role on the agecognition relations, but rather age and knowledge are "confounded" such that increased
age tends to be associated with higher levels of knowledge.
To summarize, the key features of the five perspectives, all of which assume positive effects of knowledge on measures of cognitive performance, are as follows:
(A) moderation-interaction of age and knowledge, smaller age relations at high
levels of knowledge; (B) stability-no age effects, no age-knowledge relation; (C)
decline-negative age effects, no age-knowledge relation; (D) mediation-negative
age-experience (age-knowledge) relations, control of experience (knowledge) reduces
negative age differences; and (E) migration-positive
of knowledge increases negative age differences.
age-knowledge relations, control
In all except the moderation model, the effectsof age and knowledge on cognitive
performance are additive, and therefore to account for the age relations on measures
of cognitive performance it is only necessaryto explain the main effects of age and
knowledge and any relation that might exist between age and knowledge. However,
if the evidence were to support the moderation model, it would also be necessaryto
explain why the age-related effects differ as a function of amount of knowledge.
Age - Know ledge Relations
Before examining the researchresults concerning age and knowledge, it is useful to
briefly describe how knowledge has been assessedin studies of aging and cognition.
The most common method is by various tests of word knowledge, or vocabulary, in
which the examinee is required to either produce or select an appropriate definition
of a target word such as pofiirtcatu or virulent. When the questionsrefer to different
types of information, the tests are assumed to assess general information. A number of specialized knowledge tests have also been used in which the questions refer
to particular domains of knowledge, such as science, social science, or humanities.
Examples of questionsin these specializedtests are "What is the function of the kid"What
ney?"
is a regressive tax?" and "Who was the composer of the Brandenburg
Concertos?"
Many studies investigating age-relatedeffects in cognition have administered some
type of vocabulary test in small conveniencesamples. The typical finding in these
studies is that the older adults in the sample tend to have higher vocabulary scores than
the young adults. Although it is tempting to infer from these results that knowledge
increasescontinuously with age, this conclusion may not be warranted becausethe
samples in these studies are usually small and may not be representative of the general
population becausethey often include only young college students and highly motivated
older adults.
Problems of unrepresentative samples also plague studies involving age comparisons with other measuresof knowledge. For example, Stanovich, West, and Harrison
Pl: FCH/SPH
K8008-12
P2: FCH
KB0O8-Staudinger August12,2002
212
Understanding Human Development
(1995) reported that older adults were higher in measuresof general knowledge (based
on tests involving the recognition of authors, magazine titles, and so on) than college
students. However, 737o of the older adults in their sample had college degrees, and
45Vo had advanced graduate degrees, and thus the two groups were almost certainly
not equally representative of their respective age cohorts.
In a recent report, Ackerman and Rolfhus (1999) compared college students and
adults between 30 and 59 years of age on several specially constructed knowledge
tests. The adults had higher knowledge scores than the students, but within the adult
sample the age correlations for the tests ranged from -.13 to .19 with a median of
-.015. Their resultsthus provide no evidenceof greaterknowledge with increasedage
among the nonstudentadults in the sample.A small positive correlation of .19 between
age and a composite measure of knowledge based on these same tests was reported in
a later study by Ackerman (2000). However, the sample in that study was relatively
young (mean age of 34.2 years) and unrepresentative of the population with respect to
education(all participantshad acquireda bachelor'sdegree,and 25Vohadalso received
advanced degrees).
One project with relatively large samplesis Schaie's (1996) SeattleLongitudinal
Study. Results from two vocabulary measuresadministered to new cross-sectional samplesinthe 1984andl99l wavesof thatprojectrevealedanincreaseintheaveragescore
from the twenties to the forties, then a period of stability, and a decreasearound age 70.
The most informative data on age relations are those derived from nationally
representative samples, typically stratified on age, gender, ethnic group, educational
attainment, and geographical region. The normative samples for standardized tests
of cognitive ability possess these characteristics, and therefore they are particularly
valuable for determining true age trends in relevant variables.
Figure 12.3 illustrates the age relations for the information and vocabulary subtest scores from the normative samplesin the 1955, 1981, and 1997 revisions of the
Wechsler battery. Notice that the mean scores increase from age 20 to about age 40
or 50 and decreasebeyond age 50. The Kaufman Adolescent and Adult Intelligence
Test (Kaufman & Horn, 1996) is another standardized test with norms from a large nationally representative sample. Figure 12.4 portrays the age trends on two measures of
vocabulary from that test battery, double meanings and definitions. Once again, there is
a slight increaseto about age 50 followed by a decline. Finally, the Woodcock-Johnson
Psycho-EducationalBattery-Revised (1989, 1990) is unique becauseit includes three
tests of specialized knowledge (in the topics of science,social science,and humanities). Figure 12.5 portrays the age trends for the three measures of knowledge in the
normative sample for this battery.
The mean scores in all of these data sets exhibit a similar pattern-namely, an
increase from the twenties to the forties or fifties, followed by a gradual decline totaling between one-half and one standard deviation from about age 50 to 80. Although
not represented in these data, results from studies focusing on adults above the age of
70 indicate that there may be an acceleration in the decline of measures of vocabulary
Pl: FCH/SPH P2:FCH
---\P-aK
K8008-12 KB008-StaudingerAugustl2.2002
Interrelations of Aging, Knowledge, and Cognitive performance
Zj3
A. Vocabulary
1.0
z\
E
8 o.s
r
a
E
(E
o
o
6
\
.//
\-
-.--* ---
^ .^U
U
.{
(\I
I
o
C\
o
o)
--.- WAIS(195s)
---v- WAIS-R(1981)
--{ - WAIS-lll(1997)
-0.5
- t.u
40
50
60
70
ChronologicalAge
B. Information
g
O
o
U)
n
- ' -c
E
(E
g
U)
s
\
0.0
\._,
C\l
I
o
(\l
or
-.- wAls(195s)
---q- wAts-R(1981)
--{ - wAts-n(1997)
-u'3
-1.0
20
30
40
50
60
70
80
90
Chronological
Age
Figure 12.3
Age Relationson theVocabulary(A) andInformation(B) Variables
from the Normative Samplesin Three Versionsof the Wechsler
Battery.
Datafrom Wechsler(1955,1981,1997)
Pl: FCH/SPH P2:FCH
--tp- ,-k
K8008-12 KBO08-StaudingerAugust12,2002
a1 A
Understanding Human Development
1.00
o
u./c
a
0.50
Y
Definitions
(u
o.25
o
a
0.00
o
o
b-
-o.25
-0.50
.E
(tt
-0.75
f,
ut
-1.00
20
30
40
50
60
70
80
90
Chronological
Age
Figure 12.4
Age relationson VariablesAssessingWord Knowledgefrom the
NormativeSample(N : 1,500)in the KaufmanAdult Intelligence
Test.
Data from Kaufmanand Horn (1996).
-a-f---+
--o
Scien@ (r = -.24)
Social siene (r= .02)
Humanities(r = -.'lE)
0.5
l
.9
.g 0.0
o
o
(u
-o.s
E
(l'
a
-1.0
20
30
40
50
60
70
80
90
Chronological
Age
Figure 12.5
Age Relationson ThreeTestsof Specialized
Knowledgefor 1,184
Individualswith More ThanNine Yearsof Educationfrom the NormativeSamplein the Woodcock-Johnson
Testof CognitiveAbilities.
Bars aroundeachpoint are standardenors. Data from Salthouse(l 998).
Pl: FCH/SPH
KB008-12
P2: FCH
KB008-Staudinger
August12,2002
Interrelations of Aging, Knowledge, and Cognitive Performance
2j5
andknowledgeat olderages(e.g.,Hultsch,Hertzog,Dixon, & Small, 1998;Lindenberger& Reischies,1999).
Assessmentsof other types of knowledgereveal broadly similar pattems.For
example,Demmingand Pressey(1957)developedthreetestsof what they termed
practical knowledge.The tests,and samplequestionsin each,wereas follows.
Yellowpages Where in the yellow pagesof the telephonedirectory would you
look if you wantedto buy an Airedale-under heatingequipment,kennels,shoe
stores,real estate,dairy equipment?
lzgal terms A documentcontrollingdispositionof one'spropertyat deathis
calleda bond,title, contract,will, equity.
Occupations The personto baptizea babyis a naturalist,notarypublic, nurseryman,magistrate,
clergyman.
These tests were administered to two nonrepresentative samples-namely, inmates in
a penitentiary and students attending evening classes. In both samples, the average
scores increased to about age40, but very few adults in either sample were above the
age of 50, and thus there was no information in this study about the trend at older ages.
However, a later study by Gardner and Monge (1977) found that the age funcrions
for scores on specially constructed tests of knowledge of transportation, finance, and
disease peaked in the fifties and then declined, which is the same pattem found with
measures of vocabulary and general information.
Tacit knowledgerelevantto managementwas recently assessed
by Colonia-Willner
(1998) in a unique sample of 157 midlevel bank managersbetween 24 and,56 years
of age. The measuresof tacit knowledge in this study were based on ratings of the
desirability of alternativeresponsesto various scenarios,which were scored in terms
of how closely an individual's ratings resembledthe ratings of "experts." Becausethe
scores were deviations, lower values correspond to better performance (smaller deviations represent ratings that are closer to the expert ratings). All correlations between
the tacit knowledge measures and age were positive, with a .28 correlation between
age and the overall tacit knowledge score, indicating that increased age was associated
with less tacit knowledge (a larger discrepancy from the experts).
A third example of a specializedtype of knowledge concernsthe concept of wisdom, which is often assumed to be a domain in which older adults have an advantage
over young adults. Unfortunately, this assumption has been difficult to evaluate because of a lack ofconsensus about how wisdom might be assessed.One approach to
the assessmentof wisdom-related knowledge has involved tests of the comprehension
or interpretation of proverbs because they are often considered to reflect fundamental
principles of human existence. However, research on proverb interpretation has revealed either a pattem of stability across the adult years (e.g., Sorenson, 1938) or a
pattern of age-related declines (e.g., Aftanas & Royce, 1969; Albert, Duffy, & Naeser,
1987; Bromley, 1957; Hamsher & Benton, 1978). Baltes and his colleaguesadopted
Pl: FCH/SPH
KB008-12
P2: FCH
KB008-Staudinger
276
Augustl2,2002
Understanding Human D evelopment
another approach to the assessmentof wisdom by proposing that an individual's level
of wisdom could be assessedby his or her responsesto selected life dilemmas, scored
according to several theoretically determined criteria. This method of assessmenthas
been used in a number of independent studies, and in none ofthem have there been significant age differences favoring older adults (Baltes, Staudinger, Maercker, & Smith,
1995; Smith & Baltes, 1990; Staudinger,1999).
To many researchers working in the area of aging and cognition, a discovery
of stability across most of the adult years is a welcome finding, particularly when
contrasted with the widespread declines found in variables reflecting process aspects
of cognition. Nevertheless, the absence of more dramatic and sustained increases in
knowledge is puzzling becauseincreasedage is presumably associatedwith greater
cumulative exposure and opportunities to acquire information, and thus one might
expect knowledge to increasecontinuously acrossmost, if not all, of the adult years.r
Several hypotheses have been proposed to account for the lack of continuous agerelated increasesin knowledge, but it should be recognizedat the outset that they are
probably not mutually exclusiveand almost certainly are not exhaustiveof the possible
explanationsfor this phenomenon.
One interpretation of the failure to find age-related increases in knowledge is that
there have been generational shifts in the average amount ofeducation, and amount of
education is usually positively correlated with knowledge. It has therefore been argued
that age trends in knowledge might be altered if adjustments were made for the age
differencesin either the quantity or quality ofeducation. There aretwo issuesto consider
when evaluatingthis interpretation-what is actually achievedby control of education,
and the empirical consequenceson age-knowledgerelations of controlling amount of
education. With respectto the first issue,there are at least three possibilities for how
education might be related to level of knowledge acrossadulthood. One is that education
has a direct influence on amount ofknowledge by the increased opportunities to acquire
knowledge during the period of formal schooling.A secondpossibility is that education
hasan indirecteffect on level of knowledgeby stimulatingknowledge-seekingactivities
at later periods in one's life. Finally, it is also possible that amount of education is a
proxy for general intellectual ability in that greater amounts of intellectual ability are
presumably neededto gain accessto higher levels of education.To the extent that this
latter possibility is valid, then the results of controlling for amount of education may
not be readily interpretable because education may actually be a surrogate measure of
the efficiency with which knowledge can be acquired. At the current time, there is little
relevantevidencethat would allow thesepossibilitiesto be unequivocallydistinguished,
and it is likely that severalof them are operating simultaneously.
Fortunately, at least some empirical results suggestthat difficulties ofinterpretation
may not be serious because control of education often has little effect on the relations
between age and knowledge. For example, this is true in the nationally representative
sample used in the norms for the Woodcock-Johnson Psycho-Educational Battery, even
though the correlations between amount of education and the measures of knowledge
Pl: FCH/SPH
K8008-12
P2: FCH
KBO08-Staudinger August12,2002
Interrelations
of Aging, Knowledge, and Cognitive Performance
217
were in the .4 to .5 range. That is, the education-partialed correlations between age
and knowledge were nearly identical to the unpartialed correlations (science, partial,
r : -.26, original r : -.24; social science, partial r : .02, original r : .03; and
humanities,partial r - -.18, original r : -.19).
A second interpretation ofthe lack ofconsistent age-relatedincreasesin knowledge
is that losses offset gains, in the sense that the acquisition of new information may be
accompanied by the forgetting of old information. The combination of the two opposing
processescould lead to stability in measures of overall knowledge if the effects of the
two processesare close to equilibrium. Unfortunately, this interpretationis difficult to
investigatewithout detailed assessmentsover time of the same types of information in
the same individuals.
A third interpretation is that after a certain point, increased age may be associated
with progressively more restricted exposure to new and different experiences.This view
suggests that there may be something analogous to an asymptote, perhaps around the
period of middle adulthood, on opportunities to acquire new information. That is, some
time after formal schooling is completed, there may be a decrease in the diversity of
one's experiences,and consequentlyexposureto new types of information may become
progressively more limited. This interpretation therefore implies that the reason for the
lack of continuous increasesin knowledge with advancing age is not becauseof any
factors operating within the individual but rather becauseof a gradual constriction in the
rangeofnovel experiencesas peoplegrow older. It should be possibleto investigatethis
interpretation by attempting to relate measuresof the diversity or novelty of information
exposureto measuresof available knowledge, but researchof this type has apparently
not yet been conducted.Perhapsthe closestresearchis the Kohn and Schooler project
(e.g., Kohn & Schooler, 1983; Schooler, Mulatu, & Oates, 1999), in which relations
between substantivecomplexity of work and measuresof cognitive functioning were
examined, but their assessmentsdid not include measuresof knowledge.
A fourth interpretation ofthe lack ofcontinuous age-relatedincreasesin knowledge
is based on the idea that one's knowledge becomes progressively more specialized with
age and experience. That is, tests designed for the general population are necessarily
broad and relatively superficial rather than highly specific and detailed, or else they
would have very limited applicability. However, when an individual pursues vocational
and avocationalinterests,much of his or her knowledge is likely to becomeincreasingly
more specialized,and these broad tests would probably fail to detect any increasesin
specialized knowledge that may be occurring. The idea that it may not be possible
for a single test to assessall facets of one's knowledge because of its specialized
and idiosyncratic nature has been recognized by several writers, including recently by
Ackerman (1996) and Cattell (1972, 1998).
The specialized-knowledge interpretation is challenging to investigate because
on the one hand, it is assumed that if tests could be specialized to the individual,
then they might exhibit increases with age, but on the other hand, if the tests are too
highly specialized,then it may be impossible to make any across-personcomparisons.
Pl: FCH/SPH
K8008-12
P2: FCH
KBOO8-Staudinger August12,2002
278
Understanding H uman D evelopment
Nevertheless,
two typesofevidenceseemrelevantto the selectivespecializationinterpretation.The first consistsof theexaminationof relationsbetweenageandspecialized
knowledgetestsamonghigh performers,or experts,within a givendomain.That is, in
somestudies,the individualswereselectedspecificallybecausetheyhad considerable
experiencewith the relevantactivity,andconsequentlytheycan be consideredexperts
in that domain.
Thereare only a few studiesof this type, andthe samplesare typically small,but
theresultsarenevertheless
informative.In anearlystudyof 20 bridgeplayers,chamess
( 1979)found a smallpositivecorrelationbetweenageandscoreon a bridgeknowledge
test (r:.16), but in a later study(Chamess,1983)with 45 bridgeplayersranging
from 2l to 71 yearsof age,the age-knowledge
correlationwas -.40. pfau andMurphy
(1988)reporteda correlationof -.12 betweenage(acrossa rangefrom 17 to 75) and
chessknowledgein a sampleof 59 chessplayers.Finally,Meinz andSalthouse
( 1998)
administered
a testof musicalnotationknowledgeto 128 individualswho variedin
amountof musicalexperienceandfounda correlationof - . I 7 betweenageandmusical
knowledge.The correlationsbetweenageandthe measures
of knowledgearetherefore
small and are more often negativethanpositive.Furthermore,it is importantto point
out that the failureto find strongpositivecorrelationsis not becausethe knowledge
testswerenot valid in the domainsbecausethe correlations
betweenthe knowledge
scoresand measures
of skill or performance
in the domainwerequite high, ranging
from .63 to .86.
The apparentimplicationof theseresultsis thathigh levelsof specialization
can
be achievedin a domainwithout evidenceof age-relatedincreasesin relevantknowledge.This is surprisingbecauseone of the factorsthat might havebeenpresumedto
contributeto the older adults'expertiseandhigh performancein the domainin greater
knowledge.
A secondmethodof investigating
the specialized-knowledge
interpretation
consistsof examiningknowledgein topicsgroupedaccordingto the individual'sselfratedinterest.The rationaleis that if as peoplegrow older they tend to specializein
a relativelysmall numberof topics,then the relationsbetweenage and knowledge
would be expectedto be more positivefor topics that the individualsrate as high in
interest.
This prediction was examinedin two recent studies in my laboratory (Study
3 in Hambrick,Salthouse,& Meinz, 1999,andan unpublishedstudy).Individuals
in both studieswererecruitedto participatein a projectinvestigating
determinants
of
crossword-puzzle
skill, and the quantityof knowledgein differenttopic areaswashypothesized
to be onepotentialdeterminant
of puzzle-solving
ability.The participants
in the studieswere first askedto rate (on a five-point scale)their levels of interest
in 10 different topics (American history American literature,art, geography,music,
mythology,science,sports,world history, and world literature),and then their level
of knowledgein eachareawas assessed
with both multiple-choiceand short-answer
recallquestions.
Pl: FCH/SPH
K8008-12
P2: FCH
KBO08-Staudinger Augustl2,2002
Interrelations
of Aging, Knowledge, and Cognitive Performance
219
If selective specialization is operating, then we would expect main effects of interest (higher levels of interest associated with greater amounts of knowledge), and,
most important, an interaction of age and interest in the direction of larger age-related
increases in knowledge for high-interest topics.
The first study involved 195 adults between 18 and 87 years of age. Figure 12.6
portrays the age relations in this study, with the different panels containing data from
the different topics. The three lines in each panel correspond to the average age function
across the entire sample, and separate lines represent the functions for the subsets of
individuals rating their self-interestin the topic as high or as low. It is apparent that
despite main effects of interest in each topic (higher interest associatedwith greater
knowledge), the three setsofage relations were nearly parallel. Furthermore, regression
analyses in which age and interest were both treated as continuous variables revealed
that none of the cross-product interactions between age and self-rated interest were
statistically significant. The results of this study are therefore inconsistent with the
selective specialization interpretation because they suggest that the relation between
age and knowledge was the same regardless of the level of self-reported interest.
The second study involved 206 adults between l8 and 84 years ofage. This was
a replication of the previous study except that now the participants were also asked
to rate their interests in the 10 topics for each decade of their life from age 15 to the
current time. We hypothesized that there might be a different outcome with measures
of cumulative or peak interest instead of current interest, but in fact the patterns for
each type of interest measurewere actually quite similar and nearly identical to that
found in the first study. Scores on the knowledge tests were greater when self-rated
interest was high, but there were no interactionsof age and self-rated interest. Once
again, then, there was no trend for greater age-related increasesin knowledge for topics
in which the individual reports the highest levels of interest.
The results of thesetwo studies,along with those from the studiesinvolving people with specific types of experience,are obviously not consistentwith the specialized
knowledge interpretationof the lack of continuousage-relatedincreasesin knowledge.
Because the differential education interpretation does not appear plausible and there is
no evidence relevant to the gain-loss equilibrium and the restricted exposure interpretations, there is not yet a convincing resolution of the puzzle of why knowledge does
not increasecontinuously with advancingage.
Role of Knowledge on Cognitive Functioning
Although the evidenceis mixed with respectto a positive relation betweenage and
knowledge,thereare severalreasonsfor expectingpositiverelationsbetweenknowledgeand cognitiveperformance.That is, prior researchsuggestsat leastfive possible
consequences
of greaterknowledgeon aspectsof cognitiveperformance:( I ) enhanced
memoryandothercognitiveprocesses
throughricherandmoreelaborateencodingand
moreeffectiveretrievalcuesbecauseof thepresenceof an organizationalstructureand
Pl: FCH/SPH
KB008-12
P2: FCH
/f..
KBO08-staudinger Augustl2,2002
280
Understanding Human D evelopment
Y
;
i i l
-
o
e
o
N
i
t l
l r
l l
, l
; l
.c.
(I,
o
E
o
r
t
a
:
: g B
! u
o)
o
-
@
a
* o
N
@
O
*
O
N
-
O
@
N
@
n
t
O
N
-
d
-
O
o
x
d
a
6 E
6
l
(!
-/i
o
-_
c
'=
E
@
N
@
@
h
$
O
N
F
O
O
N
@
6
+
o
N
F
--':z
F.E.;
Y
I
o
o
o
o)
q <
@
N
@
6
$
O
N
F
.O
o )
O
-
=
a
:
- . = 5
b i s E
€ E E ;
x EaH
€ ' F
^
v
o
o
o
o
< = !
o
:;
ao g
i t :
>
E r o
i T sg ,
6
tr 4*
V !E:.
o E '-{,
E O F a X
4
E E i
d
O
@ N ( O 6 . + O N F O
N
@
O
+
O
N
-
O
@
N
@
n
*
t
O
,
N
r
O
6
. Y O o
c
o r F
FCE
'6
E
f
O
N
@
b
+
O
N
-
O
N
c O t - @ ( ) . + o N F O
d
5
€6 ?0 .t g
r
p
:
'-i
s
q E
-i 6
H !
F
5
o
a
Ho
> -
F.9
a Eit
( . ) _ - o *
E
? . =
S S E q
v
N
13?
@ -cc )
3
t
O
6
-
# iyY *
=
9 E
o J= E
o
.
* ' ; 6 =3 ;
' t :r 6RgHE9
O 5
o
o)
o
6
9 E
b'E
e
F
o o
9
o
g,h
o
3
A
N e
a
\o
ci
0)
bo
?:
A
q i
Pl: FCH/SPH
K8008-12
Y2:FCH
KB008-Staudinger
August12,2002
Interrelations of Aging, Knowledge, and Cognitive Performance
281
the ability to chunkinformationin termsof familiar groupings;(2) facilitatedsolutions
of problemsby quickeraccessto relevantinformation(suchaspattemrecognition)and
betterorganizedrepresentations
of the problem,which may allow a greateramountof
information to be consideredwhen making decisions;(3) more accurateprediction
of future outcomesby capitalizingon expectationsabouteventsthat generallyoccur;
(4) more efficient modesof functioning,such as relianceon knowledge-based
direct
retrievalor compiledalgorithmsinsteadof slow,controlledprocesses;
and (5) reduced
processingrequirements(suchasattentionandworking memory)whenprior solutions
for familiar problemscanbe easilyaccessedandthe needfor resource-intensive
computationcan be minimized.Moreover,in somecases,a greateramountof knowledge
hasbeenfound to be associatedwith performancein cognitivetasksthat is equivalent
to (Walker,1987)or evengreaterthan(Chi, 1987)thatof individualswith higherlevels
of basiccognitiveabilities.
positiverelationsbetweenknowledgeandmeasures
In light ofthe expected
ofcognitive performance,it is meaningfulto considertherole of knowledgeon age-cognition
relations.Relevantresearchfrom my laboratorywill thereforebe discussedto attempt
to determinewhich of thepossibilitiesportrayedin Figure12.2is mostplausible.
I refer to the first study,an unpublishedstudy conductedwith a former graduate
student,ElizabethMeinz,asthe City MemoryStudy.In this study,we testedmemory
with namesof streets,shoppingdistricts,residentialareas,andotherinformationrelevantto oneof two cities.Ttvomemorylistswerepresented,
eachcontainingeightitems
from Atlanta,thecity of residence,
andeightitemsfrom St.Louis,whichwaspresumThe itemswerepresented
sequentially
and
ably unfamiliarto mostof theparticipants.
werefollowedimmediatelyby writtenrecall.
The samplein this studyconsistedof 125adultsrangingfrom 18 to 83 yearsof
age.The participantshad lived in Atlanta between0 and 68 years,with an average
but we assumedthat therewas
of 19 years.No testof knowledgewas administered,
greaterknowledgefor items from the more familiar city and possibly more so with
increased
agebecause
of a correlationof .38betweenageandthenumberof yearsthe
individual reportedhaving lived in Atlanta.
As expected,moreitemswerecorrectlyrecalledfrom thefamiliar city (mean59Vo
correct)than from the unfamiliar city (mean23Voconect).If knowledgecontributes
to the preservationof memory performanceacrossadulthood,an age-by-familiarity
interactionmight havebeenpredictedin the directionof smallerage-relateddeclines
for familiar items.However,this wasnot found,and insteadtherewerenearlyparallel
for familiar (agecorrelation: -.25) and unfamiliar (agecorreage-relateddecreases
-.24)
items.In otherwords,theresultsindicatethatalthoughfamiliaritywas
lation:
beneficial(becauseof the main effect of the city on memoryperformance),it was not
differentially so for older adults.
Thesefindingsareobviouslyinconsistentwith the moderationhypothesisbecause
decline
therewasno interactionof ageandfamiliarity or knowledge,andtheage-related
was similar for the high-knowledgeand low-knowledgematerials.Becauseno direct
Pl: FCWSPH
K8008-12
P2: FCH
KB008-Staudinger
282
August12,2002
UnderstandinI Human D evelopment
measureof knowledgewasavailable,it wasnot possibleto examinethe age-knowledge
relation or the effects of statisticalcontrol of knowledgeon the age relations,and
consequentlytheresultsmay or maynot be consistentwith themigrationinterpretation.
A secondstudy,alsoconducted
with ElizabethMeinz (Meinz& Salthouse,
1998),
involvedtwo comparisonsof the role of knowledgeon the relationsbetweenageand
memory.The participantsin this studywererecruitedto representconsiderablevariation in musicalexperience,and the amountof experiencerangedfrom noneto many
yerusasa professional
musician.Onecomparison
in this studyconsisted
of a contrast
betweenmemory for musicalmaterial(notesfrom a melody on a staff) and for nonmusical material (symbolson concentriccircles).The assumptionwas that most of
the participantswould be very familiar with the musicalstimuli but would havelittle
or no familiarity with the nonmusicalstimuli. Following the samerationaleas in the
City Memory Study,therefore,it waspredictedthat the age-relateddeclinesmight be
much smaller for the more familiar material.However,this prediction was not confirmedbecausethe agetrendsweresimilarfor the musicstimuli (r - -.31) and for
the nonmusicstimuli(r : -.35).
( 1998)studyconsisted
Thesecondcomparison
in theMeinzandSalthouse
of contrastingthe age relationsfor the music stimuli amongpeoplewith different amounts
of musicalknowledge.Levelof musicalknowledgewasassessed
by a testof musical
notationinformation that was administeredto all participantsin the study.Although
thereweremaineffectsof age(negative)andof knowledge(positive)on themeasureof
music memoryperformance,therewas no interactionof ageand knowledge.Instead,
(r : -.67), medium-knowledge
theagerelationsweresimilarfor thehigh-knowledge
-.38)
-.52)
(r:
(r :
andlow-knowledge
individuals.
The lackof an interaction
is
with themoderationview.However,therewasalsolittle evidence
clearlyinconsistent
for themigrationinterpretationin this studybecauseeventhoughtherewasa high positive correlation(r : .8I ) betweentheknowledgescoreandthemeasureof musicmemory performance,therelationbetweenageandknowledgewasactuallyslightly negative
(r : - .17) ratherthanpositive.At leastin this study,therefore,therewasno evidence
for a migrationto higherlevelsof domain-specificknowledgewith increasedage.
The next set of studiesinvolveda varietyof verbalfluencytasksin which the
individual was allowed a limited time to generateas many words as possiblethat
satisfiedspecificcriteria,suchas beginningwith a particularletteror being a member
of a particularcategory.Becausetheitemsin thesetasksarewords,relevantknowledge
can be assessed
with a testof vocabulary.
The outcomesof particularinterestin thesecomparisonswere the age-by-knowledgeinteraction(aspredictedby themoderationhypothesis)andtheamountofincrease
in the negativeage relation after control of the measureof knowledge(as predicted
by the migration hypothesis).Data were availablefrom four separatestudy and task
combinationsinvolving differentfluencytasksor samplesof researchparticipants.In
eachcase,thesamplesconsisted
ofbetween120and250adultsspanninga widerange
of ages.
Pl: FCH/SPH
KB008-12
P2: FCH
KBO08-Staudinger August12,2002
Interrelations of Aging, Knowledge, and Cognitive Performance
283
Table 12.1
Au:; Please
provloe
TT.
Proportion of Variance
Study Variable
Age* Knowledge
Age
Age.Knowledge
h
LIf"
002
.038
.100
lxr
lqu'
i,
llot*. 1t .\,,.0,,
.\h
\
.v
vi'
v'
1u''L't
.{''
Salthouse,Fristoe,and Rheo
(1996),N = 259
Letter fluency
Salthouse,Toth, Hancock,and
Woodard(1997),N: 124
Letter fluency
CategoryFluency
Hambrick, Salthouse,and Meinz
(1999),Study 1, 1,1: 201
Letter fluency
( 1993)
Salthouse
Study l, N = 100young,100old
Nouns
S-words
Anagrams
Word switch
Study2, N = 77 young,77 old
Word beginnings
Word endings
Make words
Scrabble
.5
.,.}t
c)
oJ ,,".c\1
.004
.002
.016
.018
.067
.072
.000
.058
.204
.020
.000
.000
.000
.067
.080
.o25
.316
.ll2
.145
.080
.4t2
.001
.011
.006
.040
.t20
.2r3
.265
.055
.1'75
.283
.343
.094
The top portion of Table l2.l summarizesthe proportionsof varianceassociated
with the age-by-knowledgeinteractionand the age-relatedvariancebefore and after
control of the measureof knowledge.The absenceof any interactionsof age and
It can alsobe seen
with the moderationhypothesis.
knowledgeis clearlyinconsistent
that in eachcasestatisticalcontrolof the knowledgemeasuresincreasedthe magnitude
of the age-relatedvarianceby making the age relationsmore negative.This result,
with greater
togetherwith thefindingthatin thesesamplesincreasedagewasassociated
knowledge,is consistentwith themigrationinterpretation.In otherwords,thesefindings
suggestthat knowledgedoesnot appearto moderatethe negativerelationsbetweenage
and various measuresof cognitive performancebut rather that the negativeeffects
with
with increasing agecanbe attenuatedwhenincreasedageis associated
associated
greateramountsof relevantknowledge.
The sametypesof analyseswerealsoconductedon datafrom eightdifferentverbal
( 1993).Therewere100young
in two studiesin a projectby Salthouse
tasksinvestigated
adultsand 100olderadultsin Study1 and77 youngadultsand77 olderadultsin Study2.
Table12.l inthesame
thesestudiesarereportedinthebottomportionof
Theresultsof
manneraswith thefluencyvariables.As canbe seen,thesevariablesexhibitedthesame
pattem as the fluency variables.Only one of the eight possibleinteractionsbetween
t
a,u\u-F.,--r'\
"
\,1
\oV\
O
-
\
"N'\J
i,h'
I o'" .-/
'w"
P1: FCH/SPH
KB008-12
P2: FCH
KB0o8-Staudinger
August12,2002
284
Understanding Human D evelopment
age and knowledge was statistically significant (the Scrabble variable in Study 2),
and therefore the results are inconsistent with the moderation hypothesis. In contrast,
the relation between age and knowledge was positive in each study, and the negative
effects of age on the performance variables increased after control of knowledge, both
findings consistent with the predictions from the migration hypothesis.
Another type of analysis is also relevant to the role of knowledge on age-cognition
relations. This analysis examines two possibilities for how knowledge might contribute
to age-related effects on cognitive performance. First, older adults could have a higher
quantity of knowledge than young adults but have the same pattem of utilization of
knowledge. And second, compared to young adults, older adults might exhibit a greater
reliance on knowledge relative to other determinants of performance.
The distinction between these two altematives can be expressedin terms of a
regression equation,
P:a(K)+b(A),
whereP is performanceon thecriteriontask,K is knowledge,andA is generalabilities
(or processaspects
ofcognition).
The interestingquestionwithin this frameworkis, Wherearethe ageeffects?The
levelof relevantabilities,A, is likely to be lower for older adultsbecauseit reflects
processaspects
of the
of cognition,whereasK may be higherfor olderadultsbecause
evidenceof increasesin vocabularyknowledge,at leastfrom the twentiesto the fifties.
Finally, agedifferencesmay be evidenton a and b, which would indicatedifferential
relianceon the performancedeterminants,as though there is a shift with age in the
methodusedto perform the task.In particular,older adultsmight havelower values
of b (reflectingreducedrelianceon processabilities)but highervaluesof a (reflecting
greaterrelianceon knowledge).
If measuresof P, K, and A are
How can thesepossibilitiesbe investigated?
available,then regressionequationscan be constructed,and the resultingparameters
can be comparedacrossindividualsof differentages.Ideally,analysesof this type
would be carriedout at the level of individuals.However,this would not only require that eachindividual perform severaldifferent criterion tasks,but different sets
ofpredictors for eachtask would alsobe neededto ensurevariationin both predictors
analysescould be conducted.Data suchas theseat
and criteriasuchthat regression
the level of individualsare not yet available,but relevantdata at the group level are
containedin the two studiesin the Salthouse(1993)project.Knowledgecan be aswith measures
sessedwith testsof vocabulary,other typesof ability can be assessed
of perceptualspeed,and scoreson the simple verbal taskscan serveas the criterion
variables.
ofolder adultsin knowledge(olderadults
The analysesrevealeda slight advantage
were superiorto young adultsin K), a largeadvantageof young adultsin perceptual
speed(youngadultsweresuperiorto older adultsin A), andrelativelysmall agedifferencesin the criterion tasks(the two groupswere nearlyequivalentin P). Of greatest
Pl: FCH/SPH
I K8008-12
P2: FCH
KB008-Staudinger
August12,2002
Interrelations
of Aging, Knowledge, and Cognitive Performance
285
interest were the regression coefficients. No significant differences were apparent in
the regression parameters across the two age groups. In other words, although there
were significant age differences in the level of knowledge and in the level of speed,
there was no age difference in the relation of these variables to the criterion measures.
The findings therefore imply that there was a similar weighting, or "importance;' of
knowledge and ability in the samplesof young and old adults.
These results led to the conclusion that'Age differencesare reduced in tasks with
moderate to large knowledge involvement not becauseof changesin the predictive value
of different factors at different ages but because the average level of one performance
determinant (knowledge) tends to increase with age at the same time that the average
level of the other performance determinant (speed) tends to decrease" (Salthouse, I 993,
p. 35). This conclusion is obviously more consistentwith the migration interpretation
than with the moderation interpretation. It therefore appears that greater knowledge
does not alter the relation between age and relevant measuresofcognitive performance
when people of different ages have similar levels of knowledge but rather that the
consequencesof age-relateddeclines in other cognitive abilities can be minimized
when increased age is associatedwith greater amounts of knowledge. Knowledge might
therefore function as an attenuating factor in age-cognition relations, but only when
it is positively correlated (or confounded) with increased age (see also Staudinger &
Lindenberger,this volume).
Summary
First,although
mostresearch
I will finishwith two simpleconclusions.
concerned
with aging and cognition has focused on processaspectsof cognition, I believe I have
shown that there are many important and interesting issues related to product aspects
of cognition. And second, although it is still not clear why it has been difficult to
find evidence of continuous age-related increases in knowledge in moderately large
representative samples, there is evidence that when older adults have high levels of
knowledge, their performance in tasks where that knowledge is relevant can equal or
possibly even surpass the performance of young adults.
Note
l. As an aside,it is interestingto note that the relation of cumulativeexperienceto testsof knowledgehas
beenrecognizedfor decades.In fact, Conrad(1930, p. 594) suggestedthat becauseof greaterexposure
opportunitieswith advancingage,stability of performanceon most knowledgetestsactually implied the
presenceof decline.Furthermore,becauseof the presumedconfoundingof age and experience,Conrad
However,this proposal
recommendedagainstthe inclusionof knowledgetestsin cognitiveassessments.
wascounteredby Sorenson( 1933),who claimedthat informationandknowledgetestsshouldnot only be
included in cognitive test batteriesbut in fact should be emphasized.His argumentwas that knowledge
testswere actually fairer to older adultsthan othertypesof cognitivetestsbecausethey did not introduce
a penaltyfor disuse,as he suggestedwas the casefor other typesoftests.
Pl: FCH/SPH
KB008-12
P2: FCH
KB008-staudinger
286
/t-.
Augustl2,2002
Understqnding Human Development
References
Ackerman, P L. (1996). A theory of adult intellectual development:Process,personality,interests,and
knowledge.I ntelligence,22, 227157 .
Ackerman, P L. (2000). Domain-specificknowledge as the "dark matter" of adult intelligence: Gf/Gc,
personality,and interestcorrelates.Journal of Gerontology:PsychologicalSciences,558, P69-P84.
Ackerman,P.L., & Rolfhus,E. L. (1999).The locusofadult intelligence:Knowledge,abilities,andnonability
traits.Psychologyand Aging, 14,314-330.
Aftanas,M. S., & Royce,J. R. (1969).Analysisof brain damagetestsadministeredto normal subjectswith
factor scorecomparisonsacrossages.Muhivariate BehavioralResearch,4,459481 .
Albert, M. S., Duffy, F. H., & Naeser,M. (1987). Nonlinear changesin cognition with age and their neuropsychologiccorrelates.CanadianJoumal of Psychology,41, l4l-157.
Baltes,P.B., & Baltes,M. M. (1990).Psychologicalperspectives
on successfulaging:The modelofselective
optimizationwith compensation.In P B. Baltes& M. M. Baltes(Eds.),Saccessfulaging: Perspectives
from the behavioralsciences(pp.1-34). Cambridge:CambridgeUniversity Press.
Baltes,P.B., Staudinger,U. M., Maercker,A., & Smith,J. ( 1995).Peoplenominatedas wise: A comparative
study of wisdom-relatedknowledge.Psychologyand Aging, 10,155-166.
Bromley,D. B. ( 1975).Someeffectsof ageon the quality of intellectualoutput.Journal of Gerontology,12,
3 I 8-323.
Cattell, R. B. (1972).Abilities: Their structure,Browthand action. Boston:Houghton-Mifflin.
Cattell,R. B. (1998).Where is intelligence?
Someanswersfrom the triadic theory.In J. J. McArdle &
R. W Woodcock(Eds.),Human cognitiveabilities in theory and practice. (pp. 29-38). Malrwah,NJ:
Erlbaum.
Charness,N. ( | 979). Componentsof skill in bridge.CanadianJournal of Psychology,33, l-16.
Charness,N. (1983). Age, skill, and bridge bidding: A chronometricanalysis.Journnl ofVerbal l*arning
and VerbalBehavior.22.4O6416.
Chi, M. T. H. (1978). Knowledge structuresand memory development.In R. Siegler (Ed.), Children's
thinking: What develops?(pp. 73-96). Hillsdale,NJ: Erlbaum.
Colonia-Willner,R. (1998). Practicalintelligenceat work: Relationshipbetweenaging and cognitive efficiency amongmanagersin a bank environment.Psychologyand Aging, I 3,45-5'7.
Conrad,H. S. (1930).Generalinformation,intelligence,and the declineofintelligence. Journal ofApplied
Psychology, I 4, 592-599.
Demming,J. A., & Pressey,S. L. ( I 957).Testsindigenousto the adult andolder years.Journal of Counseling
Psychology, 4, | 44-l 48.
Foster,J. C., & Taylor,G. A. (1920).The applicabilityof mentalteststo personsover 50. Journal of Applied
Psychology,4,39-58.
Gardner,E. F., & Monge, R. H. (1977). Adult age differencesin cognitive abilities and educationalbackground. Experimental A ging Research, 3, 337-383.
Hambrick, D. 2., Salthouse,T. A., & Meinz, E. J. (1999). Predictorsof crosswordpuzzle proficiency and
moderatorsof age-cognitionrelations.Journal of ExperimentalPsychology:General,I 28, 13l-164.
Hamsher,K. S., & Benton,A. L. (1978).Interactiveeffectsofage and cerebraldiseaseon cognitiveperformance.Joumal of Neurology,217, 195-200.
Hollingworth, H.L. (1927).Mental growth and decline.New York: Appleton.
Hultsch,D. F., Henzog,C., Dixon, R. A., & Small,B. J. (1998).Memorychangein the aged.New York:
CambridgeUniversity Press.
Jones,H. 8., & Conrad,H. (1933).The growth anddeclineof intelligence:A studyof a homogeneousgroup
betweenthe agesof l0 and 60. GeneticPsychologicalMonographs,13,223-298.
Kaufman, A. S., & Horn, J. L. (1996). Age changeson tests of fluid and crystallizedability for women
and men on the Kaufman Adolescentand Adult IntelligenceTest (KAIT) at ages 17-94. Archivesof
Clinical Neuropsychology,I l, 97-121.
Pl: FCH/SPH
KB008-12
P2: FCH
KB008-Staudinger
August12,2002
Interrelations of Aging, Knowledge, and Cognitive Performance
287
Kohn, M. L., & Schooler,C. (1983). Workand personallry.Norwood, NJ: Ablex.
Lindenberger,U., & Reisches,F. M. (1999).Limits and potentialsof intellectualfunctioning in old age.In
P B. Baltes& K. U. Mayer (Eds.),Ifte BerlinAging Study:AgingfromT} to 100.New York: Cambridge
University Press.
Meinz, E. J., & Salthouse,T. A. (1998).The effectsof ageand experienceon memoryfor visually presented
mtsic. Journal of Gerontology:PsychologicalSciences,538, P60-P69.
Pfau, H. D., & Murphy, M. D. (1998). Role of verbal knowledge in chessskill. American Journal of
Psychology, I 0I,'l 3-86.
Salthouse,T. A. (1993). Speedand knowledgeas determinantsof adult age differencesin verbal tasks.
Joumal of Gerontology: PsychoIogical Sciences, 48, P29-P36.
Salthouse,T. A. (1998). Independenceofage-relatedinfluenceson cognitive abilities acrossthe life span.
DevelopmentalPsychology,34, 851-864.
Salthouse,T. A., Fristoe,N., & Rhee,S. H. (1996). How localizedare age-relatedeffectson neuropsychological measures?
Neuropsychology,
10, 272-285.
Salthouse,T. A., Toth, J. P, Hancock, H. E., & Woodard,J.L. (1991\. Controlled and automaticforms
of memory and attention: Processpurity and the uniquenessof age-relatedinfluences.Journal of
Gemntology: PsychologicalSciences, 52B, P2l6-P228.
Schaie,K. W. (1996). Intellectual developmentin aduhhood: The SeauleInngitudinal Sraly. New York:
CambridgeUniversity Press.
Schooler,C., Mulatu, M. S., & Oates,G. (1999).The continuingeffectsof substantivelycomplex work on
the intellectualfunctioning of older workers.Psychologyand Aging, 14,483-506.
Smith, J., & Baltes,P B. (1990). Wisdom-relatedknowledge:Age/cohortdifferencesin responseto lifeplanningproblems.DevelopmentulPsychology,26, 494-505.
Sorenson,H. (1933). Mental ability over a wide rangeof adult ages.Jounnl of Applied Psychology,17,
't29-:7
41.
Sorenson,H. (1938).Aduh abiliries. Minneapolis:University of MinnesotaPress.
Stanovich,K. E., West,R. F., & Harrison,M. R. (1995).Knowledgegrowth and maintenanceacrossthe life
span:The role of print exposure.DevelopmentalPsychology,3 I , 8ll-826.
Staudinger,U. M. ( I 999). Older and wiser?Integratingresultson the relationshipbetweenageand wisdomrelatedperformance.InternationalJournal of BehavioralDevelopment,23,641464.
Walker, C. H. (1987). Relative importanceof domain knowledgeand overall aptitude on acquisition of
domain-relatedinformation.Cognitionand Instruction,4, 2542.
Wechsler,D. (1955).Manualfor the WechslerAduh IntelligenceScale.New York: PsychologicalCorporation.
Wechsler,D. ( I 98 l). WAIS-RManual: WechslerAduh IntelligenceScale-Revised.
New York: Psychological
Corporation.
Wechsler,D. ( 1997).WAIS-lll: WechslerAdult IntelligenceScale,Third Edilion. Administrationand Scoring
Manual. SanAntonio: PsychologicalCorporation.
Willoughby, R. R. (1927).Family similaritiesin mentaltest abilities(with a note on the grcwth and decline
of theseabilities). GeneticPsychologicalMonograph,2,235-277.
Woodcock,R. W, & Johnson,M. B. (1989, 1990).Woodcock-Johnson
Psycho-Educational
Battery-Revised.
Allen, TX: DLM.