PsychonomicBulletin & Raiqt
r996, 5 (1),5670
Is agekinder to femalesthan to males?
FLITABE IH J. MEINZ and TIMOIHY A SALiTHOUSE
Geuryia Insntuuof Tuh,rwlqg, Atlanta., GwVi/a
Psychonomic Bulletin & Review
1998,
s (1),s6-70
Is age kinder to femalesthan to males?
EUZABETH J. MEINZ and TIMOTHY A. SALITHOUSE
Georgia Institute of Technolngy,AAnntn, Georgia
To investigate possible gender differences in the patterns of age-cognition relations, a metaanalysirswas conducted on the data from 25 studies (5,201 participants) by the second author and
his collaborators. Gender and age differences consistent with prior reports were found, but there
were few measures on which the age x gender interactions were significant. Although the mqiority
of neuroimaging studies investigating gender differences in age-related atrophy and functional de
cline report greater age-related differences in males, the only significant interactions on the cognitive measures in this study (on mea^suresof speed and reasoning) were in the direction of lesser agerelated declines for males than for females.
Age-related differenceshave been reported on various
measuresof cognitive functioning, frequently in the direction of lower levels of perfornance with increasedage
(for reviews,seeBlanchard-Fields& Hess, 1996,and Craik
& Salthouse,1992). Gender-cognition relations are also
well documented, with higher performance in males on
some measures,such as spatial abilities, and higher performancein femaleson others,suchas verbal abilities (for
reviews,seeHalpern, 1992,andMaccoby & Jacklin,1974).
Although this pattern of gender--cognitionrelations reportedly remainsrelatively stableacrossage (Schaie, 1996F
for example,the male advantagein spatial abilities is evidentthroughoutthelifespan(Willis & Schaie,l988fthe
interrelations among age, gender,and cognition have received little attention. In other words, although the qualitative pattern of gender diffbrences in performance may
remain stableacrossthe lifespan,it is not yet clear whether
the quantitative relations between age and cognitive performance are the samefor femalesand males.An interaction ofage and gendermight be expectedin light ofrecent
reports of gender differencesin age-relateddecline on
some measuresof brain morphology and functioning.
For example,Gur et al. (1991) reportedthat, although
men and women showedsimilar patternsof age-related
decline in brain volume, there was a significant interaction of age and gender on another measureof brain atrophy, cerebral spinal fluid volume (CSF). Men showed
greaterage-relatedincreasesin CSF than did women, suggesting more pronounced atrophy with advancing age in
malesthan in females.Interactionsof ageand genderwere
also found in ratios of ventricular to sulcal volume because
therewas no gendereffect in the rate ofventricular atrophy,
but men had greater age-relatedsulcal atrophy than did
This researchwas supportedby NIA Training Grant AG00l75 to
E.J.M. and NIA Grant R37 6826 to T.A.S. The authors gratefully acknowledge the helpful comments of Paul Verhaeghen,Diane Halpem,
David Lane, Lars-Gtiran Nilsson, and one anonymousreviewer. Correspondenceshould be addressedto E. J. Meinz, School of Psychology,
Georgia Institute of Technology, Atlanta, GA 30309 (e-mail: gt4956b
@acme.gatech.edu).
Copyright 1998PsychonomicSociety,Inc.
women. There were also gender differences in the hemispheric patterns of age-related atrophy; women showed
similar atrophy acrossthe hemispheres,but men showed
greater atrophy in the left ventricles than in the right ventricles. The first two findings, that of larger increaseswith
agein maleson overallCSF and sulcalCSF,suggestgreater
age-relatedbrain ahophy in men than in women, and the
third finding indicatesthat malesshow greaterage-related
decline in left hemispherestructuresbut not right hemispherestructures.
Kaye, DeCarli, Luxenberg,and Rapoport (1992) also
found that the onset oflarge age-relateddeclines in brain
volume occurred earlier in men (beginning in the 1950s)
than in women (beginningin the 1960s).Additionally, gender differencesin the relation ofage to brain changeshave
beenfound with femalesexhibiting smaller declinesin the
caudate nucleus and putamen (Raz, Torres, & Acker,
1995),and in whole brain, frontal, and temporal lobe volume (Cowell et al., 1994; Murphy et al., 1996). However,
males showedless age-relatedloss of hippocampaland
parietal volume (Murphy et al., 1996).
A few studieshave usedmeasuresof brain functioning
rather than structurewhen investigatingage-gender relations. In measuresof slow electrophysiological activity,
Duffy, McAnulty, and Albert (1993) found greateragerelated activity reductionsin men than in women. Gender
differencesin the relation of age to measuresof brain metabolism havealso beenreported, with greatermetabolic declinesin the left hemispheresthan in the right hemispheres
of men, but either greater metabolic decline in the right
hemispheresor equal metabolic decline acrossthe hemispheresin women(Murphy et al., 1996).Theseresultsare
consistentwith the structuralfindings of Gur et al. ( l99l ),
and they extendthem with functional evidence.
Although the precedingstudiesprovide fairly strongevidence for age-gender interactions in brain structure and
function, severalauthorshave noted that it is desirableto
determine whether similar interactions exist with behavioral data(e.g.,Cowell et al., 1994;Gur et al., I 991; Raz
et al., 1995).Isolatedstudieshaveinvestigatedthe effects
of gender on age-cognition relations with specific cogni-
56
IS AGE KINDER?
Table I
oflncluded
57
Studles
Age
Males
4
)
6
7
8
9
t0
ll
l2
l3
l4
l5
l6
l7
l8
19
20
1t
))
z)
24
25
26
202
218
195
128
221
228
223
100
305
240
l?{
246
258
I ts
I l'l
t78
227
t11
l3l
259
l9'l
139
zJ)
242
383
2,411
59
9l
86
43
86
t25
ll0
53
124
8'l
60
95
102
8l
57
6l
lil
94
80
95
83
58
108
r02
186
1,129
Females
M
143
127
109
85
54.6
53.3
5l.5
46.1
48.5
50.3
49.2
45.0
51.1
48.l
51.5
46.6
48.7
46.5
49.0
49.0
50.7
46.9
46.1
51.4
49.6
47.7
48.5
49.1
46.0
36.8
r35
103
ll3
47
181
153
65
151
r56
92
60
lt7
l16
139
5l
164
lt4
'tl
125
140
t97
t,282
S
D
Citation
18.3
14.0
15.7
15.6
17.4
17.7
17.6
l'7.4
16.6
15.9
16.5
t7.0
17.0
19.6
16.4
20.1
r 6.8
16.4
20.0
18.4
17.4
16.4
17.3
t7.5
l6.8
19.6
tive or neurbpsychologicaltests. For instance, in one of
three verbal memory tests,Larrabeeand Crook (1993)
found a significant interactionofage and gender,suchthat
females showed less age-relateddecline than did males,
but, in an earlier study, no significant interactions were
found on thesetypes oftests (Larrabee, Trahan. Curtiss.
& Levin, 1988).In a studyby Elias, Robbins,Walter,and
Schultz(1993), a largebattery ofneuropsychologicaltests
was administered,but a significant interaction of age and
gender was found only on the Halstead-Reitan category
subtest.lnteractions ofage and genderwere testedin a study
of gender-relatedepisodic and semantic memory diffe;encesby Herlitz, Nilsson, and Bdckman (1997\, but no
significant interactionswere found.
Schaie (1996) summarizedthe findings of the Seattle
longitudinal study,stating
age-difference
patternsappeargenerallyinvariantacross
sexwithindomains(albeitthereis strongevidence
for overall genderdifferences
in levelofperformance),
but it is not
a foregoneconclusionthat suchinvarianceholds for all
abilitiesor for all markersof a givenability.(p. 83)
This statementimplies that more thorough investigations
of the interrelationsamong age,gender,and cognitiveperformance are needed.
The purposeofthe presentstudy was to conduct such a
systematicinvestigation using meta-analyticprocedures.
However, becausethe requisite information for a metaanalysisof this type (i.e., effect sizeestimatesfor the age
X gender interaction) is seldom reported, the present
Hambrick,Salthouse,
&Meinz (1997,Studyl)
Hambrick,Salthouse,
& Meiw (1997, Study2)
Hambrick,Salthouse,
& Meinz(1997,Study3)
(1998)
Meinz& Salthouse
(1991,Study1)
Salthouse
(1991,Study2)
Salthouse
(1991,Study3)
Salthouse
(1992)
Salthouse
(1993)
Salthouse
(1994a,Studyl)
Salthouse
(1994a,Study2)
Salthouse
(1994b,Studyl)
Salthouse
(1994b,Study2)
Salthouse
(1995a)
Salthouse
(1995b,Srudy2)
Salthouse
( I 996)
Salthouse
Salthouse
& Babcock(1991,
Srudyl)
Salthouse
(1991,Study2)
& Babcock
Salthouse,
Fristoe,Lineweaver,
& Coon(1995)
Salthouse,
Fristoe,& Rhee(1996)
Salthouse,
Hancock,Meira, & Hambrick(1996)
Salthouse,
Kausler,& Saults(1988,Studyl)
Salthouse,
Kausler,& Saults(1988,Study2)
Salthouse
& Meinz(1995)
Salthouse
& Mitchell(1990)
Woodcock-Johnson
standardizationdata
analyses
arebasedon rawdatafrom studiesconducted
bv
T.A. Salthouse
andhiscollaborators
overthepastI 0 years.
METHOD
Selection ofStudies
The primary analysisfocusedon raw datafrom 25 cross-sectional
studiesconductedby T. A. Salthouseand his collaborators.All of
these sfudiesincludedboth male and femaleparticipantsof a continuous age range.Participantsin all ofthe studieswere recruited
through newspaperads, community groups, or personalcontacts,
and no specialappealsweremadein the studiesto eitherfemalesor
males.To the bestofour knowledge,noneofthese individualsparticipatedin morethanoneofthe includedstudies.performancemeasuresiiom all ofthe cognitive variablesin thesestudieswere included. There were a total of 5,201 participantsand 74 diferent
variablesacrossthesestudies.Although this wasnot a typical metaanalysis(for the reasonsdescribedearlier),the resultsofthe analyses should be broadly generalizablebecausethe age relations in
thesedatahavebeenfound to be similar to thosereportedbv other
investigators
(i.e.,Verhaeghen
& Salthouse,
1997).
Data from the most recent standardizationof the Woodcock_
Johnsontest battery(Woodcock& Johnson,1989,1990)were also
includedin a separateanalysisto be comparedwith the resultsofthe
primary analysis.I Data on 35 variablesfrom 2,41I participantsbetween 18 and 94 yearsofage were availablein this data set.Characteristicsofthe participantsfrom eachofthe studies,includingage
and genderdistributions,arereportedin Table l.
Classification of the Tasks
All taskswereclassifiedasoneofeight types:speed,spatial,primary memory/workingmemory, episodic memory, reasoning,
knowledge,verbalfluency,andmiscellaneous.
Briefdescriptions;f
all tasksincludedin the meta-analysisare listed by task classifica-
MEINZ AND SALTHOUSE
58
to
Gender
&
Io r o
:
f.o
lr
n
-t -.0 -.6 -.4 -2
2 .1 .6 .8 t.0
to
Age
c)
9.0
E
f.o
lt
n
-1 -.8 -.8
10
-2
2
.1
.6
.E 1 . 0
GenderDtfference
80
9
g.0
E.o
lr
n
-1 -.8 -.6 -.4 -2 0
2
.1 .6 .8 1.0
Figure 1. Gender, age,and gender differcnce in age-correlation
effect-sizefrequenciesfor ell variables in the primary studies.
Note that the observationsare not independentbecausethe correlations are based on individual variables, and, thus, several
mersunesof the ssme task classification may be included from
a single study. A total of 240 correlations are included in eech
distribution.
tion in AppendixA. Theseclassificationswerechosenbecausethey
representdistinct categoriesofcognitive taskswith differing levels
of age-performancerelations.Although classificationby type of
stimulusmaterial(i.e., verbal, spatial,numerical,etc.) is also a viable option (becausegenderdifferenceshave been studiedusing
theseclassifications),it wasdecidedthat this levelofanalysismight
be more appropriatefor homogeneityanalysesif, in fact, heterogeneitywithin the classificationswas found.
dataweregroupedin
The variablesfiom the Woodcock-Johnson
a similar mannerandaccordingto the abilitiesthat the tasksarepurDortedto measurewheneverpossible.
Procedure
Effect sizes.Following the suggestionofRosenthal (1991),the
Pearsonproduct-momentcorrelationcoefficient,r, was chosenasa
measureofeffect sizein theseanalyses.All variableswerescaledso
that higher scoresindicatedhigher levelsofperformance, and, therefore, negativecorrelationswith age indicatedage-relateddecline'
Genderwascodedsuchthat maleswereassignedscoresofO andfepositivegenderefmaleswere assignedscoresof l; consequently,
fectsindicatehigherlevelsofperformancein females,andnegative
gendereffectsindicatehigher levelsofperformancein males.
To expressthe age x genderinteractionas an effect size, a differencescorewas computedbetweenthe age effect sizes for males
and for females by first converting eachage correlation to a Fisher
z, subtractingthesezs, and then convertingthe z6.i6backto an r6i6.2
Effect sizesfor femalesweresubtractedfrom thosefor males;thus,
a negativevalue indicatesgreaterage-relateddeclinein malesthan
in females,anda positivevalueindicatesgreaterage-relateddecline
in femalesthan in males.To illustrate,if the ageeffecton a variable
was -.6 for malesand - .4 for females,166 is equalto the z transformationof the male effect size(-.69) minusthe z transformation
of the femaleeffectsize( - .42),which is equalto - .27.When convertedback into the correlation metric, this negativedifference score
(ra;6: -.26) indicatesgreaterdeclineamongmalesthan females'3
For illustrativepurposes,the distributionsofeflect sizesfor gender,
age,andthegenderdifferencein ageconelationsacrossall variables
(N :240) areshownin Figure I . It shouldbe notedthat the rangeof
effectsizesin thesehistogramsis relativelynarrowandthat only the
distributionof effect sizesfor agewas not centeredat zero.
thatwould resultfrom
In orderto avoidthe stochasticdependence
using severalindicesof one cognitiveability from the samestudy
and the sameparticipants,a singleeffect sizemeasurewas derived
for eachstudy in a given task classification.For instance,for each
studythat includedtasksclassifiedasmeasuringspeed,only oneeffect sizewasusedin the analyses,regardlessof whetherthe particular study included I or l0 speedvariables.To obtain this average
effectsize,theeffectsizesfor eachtaskweretransformedvia Fisher's
r-to-z transformation,and then the zs were averagedto yield a z*-*".
Finally,thez**" wasconvertedbacktoaneffectsizemeasure(r*op.
The averageefFectsizesfor the studiesin the primary analysisare
listed in AppendixB.
Etfect size estimation. Mean weightedeffect sizes (r+) were
computedacrossstudiesfor gender,age,and the age x genderinteractionterm for eachtask classification.Theseweightedcoefficientsandtheir 950%confidenceintervalswerecomputedby assigning weightsto eachstudyon the basisof samplesizes,asdescribed
in Rosenthal(1991)and in HedgesandOlkin (1985).Weightedeffect sizeestimatesandconfidenceintervalsfor eachtaskclassification in the primary analysisarepresentedin Table2.
Testsfor homogeneityof effect sizeswithin the task classifications werealsoperlormedby computingQ1, a measureof total homogeneity,andcomparingthis valueto the critical valueestablished
by the chi-squaredistribution.Ifthe effect sizeestimatesare found
(i.e., ifpl exceedsthe critical value),moderato be heterogeneous
tors may be tested.Successfulmoderatorsare thosethat yield hobetweenwithin-groupvariance,pvy,andheterogeneous
mogeneous
group variance,Qs. For instance,if p1 revealsheterogeneityin the
measureof speed,the taskmediummay be identified as a potential
and
moderator,andthe group may be divided into paper-and-pencil
computerizedtasks.The goal of subsequentanalyseswould be to
dernonstratethat therewassignificant variancebetweenthe moderator
groups(p/, but therewas little variancewithin eachof the groups
(Q* for paper-and-penciltasks and p1ryfor computerizedtasks).
WoodcockJohnson Analyses
Age, gender,andage x genderinteractioneffectsizeswerecomputed for eachtask in the Woodcock-Johnsondata. Theseeffect
sizeswereaveragedwithin taskclassificationsto producemeaneffect sizesthat could be comparedwith the weightedmeaneffect
IS AGE KINDER?
59
Table 2
Mean Weighted Correlatlons,95To Confldence Intervalg and Homogeneity Stetistics
for Thsk Classiflcatlons in the Primarv Anelvsis
TaskType
Eflect
N
r+
LLNL
Q,
4,99
Speed
Gender
Age
Male
Female
Gender difierence
2,585
Spatial
Gender
Age
Male
Female
Gender difference
Primary/Workingmemory
l4
2.632
Gender
Age
Male
Female
Gender difference
Episodicmemory
12
-.18
-.36
-.35
-.36
.01
-.22/-.14
-.39t-.32
- .381-.32
-.401-.33
-.02/.05
11.7
6
24.83t
57.00*
28.50*
50.45r
M>F
Y>O
-.07
-.39
-.39
-.39
.00
-.1l/-.03
-.42/-.35
-.42/-.36
-.42/-.36
-.04/.04
4.46
33.61|
44.90r
46.73*
40.43*
M>F
Y>O
.03
-.35
-.36
-.0|.07
- .39t-.32
-.39t-.t2
- .4lt- .i4
-.02t.06
20.42*
26.1V
39-44'
32.40*
26.27'
n.s.
Y>O
- -09t-.02
- . 4 1 / -. 3 5
-.38t*.22
-.44/-.38
.03/.1
l
10.73
77.44t
148.60*
48.99*
'16.5t+
M>F
Y>O
.02
2,97|
Gender
Age
Male
Female
Gender difference
-.05
-.38
-.J)
-.41
.07
r,380
Gender
Age
Male
Female
Gender difference
Verbal fluency
29.28
80.68r
l3l.l0r
82.60r
7f .61;
-)t
l2
Knowledge
.02/.0'l
-.53/-.48
* .51/-.47
-.54/-.50
.02/.07
2,460
Gender
Age
Male
Female
Gender difference
Reasoning
.04
-.50
-.49
-.52
.05
-.04
n.s.
n.s.
n.s.
Fv)o)Mv)o
n,s.
o>Y
.23
.00
l.7t
2l .00*
36.'t51
10.83
18.59r
.M
-.03
-.o4
-.02
-.02
- .04/.t2
-. l 1i.05
-.t2/.04
-. I 0/-.06
-. l 0/.07
4.21
1937r
46.95*
I 1.83i
18.49*
n.s.
n.s.
.0'l
-.39
-.38
-.40
.03
.0r/.13
- .44t- .J4
- .431-.t2
- . 4 4 1 .-3 5
-.03/.09
3.13
6'7.15r
86.92*
56.92x
1 0 . 2+1
F>M
Y>O
-z)
598
Miscellaneous
Fy)o)My)o
-.09/.02
.t8t.28
.18/.28
.r8/.28
-.05/.05
.z)
Gender
Age
Male
Female
Gender difference
F>M
Y>O
n.s.
n.s.
I,082
Gender
Age
Male
Female
Gender difference
n.s.
Note-Task type is the task classification formed fiom variables believed to measurea common cognitive abllity; ellect refersto the type of effect
reported; *, number of studies included in analysis; N, number of participants included in analysis; r+, mean weighted correlation; LLfuL, lower
and upper limits of 95% confidence interval; Qr,total homogeneity; n.s., nonsignificant; F>M, significantly higher performance in females than
in males; M>F, significantly higher perfiormancein males than in females; Y>O, significantly higher performance in young participants than in
older participants; F )o)M">o,
significantly greater age-relateddecline in females than in males. *p. .05 (nonhomogeneity).
sizesin the primary dataset.The Woodcock-Johnson
task classificationsand their meaneffect sizescan be found in Table3.
RESULTS
Gender Effects
As shownin Table2, significant genderdifferenceswere
found in all task classificationsexceptepisodicmemory
knowledge, and verbal fluency in the primary sample,
with females outperforming males on tasks of speed(r :
+.05) and in the miscellaneouscategory(r: +.07). There
was a small advantagein females for tests of verbal fluency (r: +.04), but this differencewas not statistically
significant becausethe 95% confidence interval included
zero. Males outperformedfemaleson spatial tasks (r :
-.18), primary memory/workingmemory tasks(r: -.07),
60
METNZAND SALTHOUSE
Ibble 3
Age and Gender Relations on the WoodcockJohnson Test Battery
Age Effect (r)
TaskType
Speed
PrimaryAVorking memory
Episodic memory
Reasoning
General knowledge
Vocabulary
Auditory processing
Visual processing
Quantitative
Miscellaneous
2
)
2
J
J
J
2
2
l3
Gender r
Age r
Male
Female
.08
.00
.04
-.01
-.04
.05
.03
.09
-.t5
.10
-.62
-.51
-.34
-.55
-.50
-.32
-.14
-.64
-.28
-.49
-.49
-.24
-.04
-.43
-.47
-.30
-.25
-.J
I
-.) t
-.49
-.28
-.08
-.45
-.49
-.30
-.26
-.JU
-.5I
-.29
-.30
Gender Difference
in Aee r
.18
-.0'7
-.09
-.01
-.09
-.10
-.09
-.06
.01
-.06
Note-Task type is the task classification formed from variables believed to measurea comrnon cognitive
component;j, number ofvariables contributing to efiect size; gender r, point-biserial correlation with gender
(male : 0, female : I ); age r, correlation with chronological age; age r for maleVfemales,seeTable I for number ofmale and female participants; gender difference in age r, age x gender interaction term expressedas
r"r" (male) - r"r" (female).
and testsof reasoning(r: -.08). In general,with the exception of the measuresof spatial abilities, the gendereffects were quite small.
into groups by the type of material used: verbal, spatial,
numerical, or symbolic (e.g.,the unfamiliar symbols used
in associativelearning tasks). Finally, spatial tasks were
divided into computerizedand noncomputerizedtasks.In
Age Effects
order to compute between-grouphomogeneity statistics,
Becauseall tasks were scored such that higher scores p", independentmoderator classesmust be formed to
indicate higher levels ofperformance, negativeage corre- avoid comparingclassescontainingthe sameparticipants.
lations reflect lower scoresassociatedwith increasedage For example,ifa study containsboth paper-and-pencil
on a task. As shown in Table 2, in the primary data set,there and reactiontime measruesof speed,then either the paperwere significant negativeagecorrelationson tasksofspeed and-pencil measureor the reaction time measuremay be
(r: -.50), spatialabilities(r: -.36), primary memory/ included in the moderator analyses,but not both. The asworking memory(r: -.39), episodicmemory(r: -.35),
signment of each task classification to a moderator class
reasoning(r : *.38), and on the miscellaneoustasks(r:
was random whenevermore than one classof task existed
-.39), and there were significantly positive effects of age within a study.
on the measuresof knowledge(r : +.23). There was no
Table 4 presentsthe gender,age, and difference score
significant effect ofage on verbal fluency.
effect sizesand the Qr or Qw statisticsfor eachtask classification or moderatorclass for episodic memory, priAge x Gender Interactions
mary memory/working memory reasoning,spatial abilities,
As noted above,the differencescoresthat serveas efTect and speed(no moderatorsofknowledge, verbal fluency,
size indicatorsof gender X age interactionswere com- or miscellaneoustasks were identified and tested).For
puted by subtracting female age correlations from male none of these potential moderators was the desired outage correlations; thus, negative difference score effect come of between-groupheterogeneityand within-group
sizesindicate larger age-relateddecline in males,and pos- homogeneity obtained;thus, there is little convincing evitive difference score effect sizes indicate larger age- idencethat the effect sizesvaried systematicallyas a funcrelateddecline in females.Table 2 revealssignificant gen- tion of thesefactors.At the presenttime, therefore,we are
der differences in the magnitude of age effects on mea- unableto accountfor the significant variability within the
suresofspeed (r: +.05) and reasoning(r: +.07), in both various task classifications.Nevertheless,it should be
casesin the direction of somewhatsmaller age-relatedde- noted that the confidence intervals for most of the effect
cline in malesthan in females.
size estimateswere fairly narrow, and, thus, the withincategory variability does not preclude conclusionsabout
Moderator Variable Analyses
overall effect sizesin the various classifications.
Becausealmost all of the Q.. statisticswere significant
One point of interest revealedin the moderator analy(seeTable 2), indicating heterogeneityof variancewithin sesis that the gender effects on some of the task classifitask classifications,severalanalyseswere performed in an cations seemto be a result ofdifferential genderrelations
attempt to identiff moderatorsof the effects within each acrossmoderator classes.For instance,closer inspection
classification. For instance,speedtaskswere divided into of the nonsignificant gender effect on episodic memory
two groups: paper-and-penciltasks and reaction time tasks revealsthat there were significant gender effects in
(computerized) tasks. Episodic memory, primary mem- the direction of higher performance in females on verbal
ory/working memory, and reasoning tasks were divided episodic tasks,but nonsignificant gendereffects on sym-
IS AGE KINDER?
6l
Thble 4
Gender, Age, and Interactlon Eifect Sizesand Homogeneity Statistics
According to Trsk Clesslflcation end Moderetor Classes
Gender
TaskClassification
Speed
ModeratorClass
Reaction time/Computerized
Paper-and-pencil
Spatial
Reaction time/Computerized
Paper-and-pencil
PrimaryAil'orkingmemory
Numerical
Spatial
Verbal
Episodicmemory
Symbolic
Verbal
Reasoning
Numerical
Spatial
Verbal
r+
24
8
l6
tl
J
8
t4
2
8
12
7
5
l2
2
)
)
.04*
.03
.06r
-.18r
-.17*
-.lgr
-.07*
-.13r
-. l6r
-.03
.03
.00
.09*
-.05r
.00
-.06
-.04
Age
Qror Qw
29.28
19.76t
18.03
11.76
t1<
9.2s
4.46
3.79
0.37
4.51
20.42*
t2.04
2.10*
10.73
2.21
6.26
2.86
-.50r
-.49*
-.51*
-.36*
- . 3 5r
-.3' 7;
-.39*
-.35*
-.4Jt
-.37*
-.35*
-.31r
-.40+
-.39t
-.36i
-.37'
-.42*
O- or O-,
80.68i
13.89
60.81|
24.83+
7.58*
20.02t
33.61*
3.94
0.05
26.86*
26.r1,
3.90
l5.41|
74.44*
2-86
33.48r
44.66'
Gender
Diffr+
.05r
-.01
.08r
.01
.15t
.00
.00
.03
-.04
.00
.02
.04
-.01
.071
. lg *
.u)
.08r
Q, or O*
73.61r
t 6.28r
43.01r
50.45r
5.98
44.84*
40.43r
25.57*
3.5r
36.98*
26.27t
23.24i
6.50
76.5f
23.80*
47.99*
6.44
Note-t, number of studies included in analysis; r+, averageweighted correlation;
Qr or Qw,total or within-class homogeneity.e1 is reported for
the task-classifications(i.e., speed, spatial, primary memory/working memory episodic memory and reasoning), and g; is reportLd foithe moderator classes(i.e., reaction time/computerized, paper-and-pencil,numerical, spatial, verbal, and symbolic). tp . .OSinonho-ogeneity).
bolic episodic memory tasks. Similarly, the male advantage on primary memory tasks seemsto be a result of
higher levels of performanceon numerical and spatial
tasks,but not on verbal primary memory/working memory
tasks.Also, the female advantageon speedtasks appears
to be a result of superior female performance on paperand-pencil tasks,but not on reaction time speedtasks.
With regard to the gender differences in age effects, differencesin the direction of lesserage-relateddecline in
males were more pronouncedon verbal and numerical
(relative to spatial) reasoningtasks,on computerized(relative to paper-and-pencil)spatialtasks,and on paper-andpencil (relative to reaction time) speed tasks. However,
theseresultsshouldbe interpretedwith cautionbecauseof
the random assignmentto moderatorclassesand the small
number of studieswithin the moderator classes.
In order to examinethis possibility,I'tests of equalvariancewere performed to comparethe variancesof age and
ofthe dependentvariablesacrossgender.Becausethe raw
dependentvariableswere in different units, they were converted to a common metric within each studv bv transforming them into standardscoresbasedon the distribution of both males and females.The standardscoreswere
then averagedto form one index ofthe dependentvariable
for each task classification within a given study.
Sigaificant F tests would indicate that the variance of
femalesand males differed, and, if this were the case,the
gender differences in age correlations might be partially
attributable to a smaller range of variation in one gender
than in the other. Howeveq the results of these analyses,
summarized in Appendix C, indicate that there were no
significant genderdifferencesin the agevariances,and there
were significant gender differences in only 3 of 89 comVariance Analyses
parisons(one each in the speed,episodic memory and
The correlation coefficients (r) usedhereas an index of miscellaneouscategories)on the dependentvariables.The
effect size are a function of the slope of the linear rela- small number of significant F testsmakesit unlikely that
tionship between age and performance on the dependent the pattern of results reported in this study were attributvariable, the variance ofage, and the variance ofthe de- able to genderdifferencesin the amount of variation in eipendent variable. It is therefore possible that gender dif- ther age or the dependentvariable. Theseresults are also
ferencesin the varianceeither on the dependentvariables of substantiveinterestbecausethey imply that, at least in
or on age could distort the comparisonsofthe age corre- these samples,adult males and females do not differ siglations betweenmales and females. In other words, gen- nificantly with respectto interindividual variability on
der differences in the age correlation on a given depen- measuresof cognitiveabilities.
dent variable may be partially attributable to greater age
variation or to greater variation in the dependentvariable Summary
in one gender relative to the other. In fact, male children
The patternof gender--cognitionrelationsin the primary
havebeen found to be more variable than female children datasetwas mixed, and, with the exceptionof the spatial
in numericaland spatialabilities(seeMaccoby& Jacklin, measure,the effect sizes were quite small. There were
1974.for a review).
largeageeffectson most of the variables,with negativeef-
62
MEINZ AND SALTHOUSE
fects of age on measuresof speed, spatial abilities, primary memory/working memory and episodic memory
but positive effects ofage on knowledge, and no age effectson verbal fluency.Small interactiveeffectsofage and
genderwere found, with a significant advantagefor males
in speedand reasoning.The results ofvariance analyses
suggestedthat it is unlikely that the results were influencedby gender differencesin the variance ofage or the
variance of the dependentvariables.
The Woodcock-Johnsondata were analvzed for comparisonwith the primary dataset.No confiience intervals
could be establishedin this data set becauseeachparticipant performed every task in the battery and, thus, only
the magnitude of the effects can be examined.Inspection
ofTable 3 reveals that gender effects in these data were
small with the exception of a moderate gender effect in
quantitative abilities, in the direction ofhigher levels of
performancein males.The agetrendswere similar to those
in the primary data set in that there were moderatelynegative effects of age on all measuresexcept vocabulary.
Surprisingly, the effects of age on knowledge in this data
setwere negative.As in the primary dataset,the direction
of the interaction effects were mixed. Greaterage-related
decline was evident among females on the measureof
speedand among males on severalof the other measures.
DISCUSSION
Although considerableeffort has been devotedto documenting gender differences in the relations of age to
brain morphology and functioning, few studieshave confirmed thesefindings with behavioral data.A meta-analysis
was therefore performed on 25 data sets encompassinga
wide variety of cognitive taskswith the goal of examining
age X gender interactions on cognitive measures.The
main effectsof age and genderwere also examinedfor the
purpose of comparison with the literature on gendercognition and age-+ognition relations. Becausemost of
the analysesrevealedevidenceofheterogeneity, it is possible that sometrue effects were concealed.Nevertheless,
the confidence intervals for estimated effect sizes were
relatively narrow and, thus, the results can be considered
reasonablyprecise.
The gender--cognitionrelations found in the primary
datasetwere, for the most part, similar to thosefrequently
reportedin the literature. It shouldbe noted that, while the
present analysis revealedmostly male advantages,some
measuresat which'femaleshave been reportedto excel
(notably reading comprehensionand writing) were not included(Halpern, 1992).
Males outperformed females in spatial abilities, which
is consistentwith the results of a meta-analyticreview by
Linn and Petersen(1986) and otherreviews(e.g.,Schaie,
1994). We found no significant gender differences on
measuresof verbal fluency or knowledge. Prior studies
have been mixed on this issue because,while there have
beensomereports of a female advantagein generalverbal
abilities(Hyde & Linn, 1986;Schaie,1994),otheranaly-
seshaverevealeda female advantageon testsofanagram
solution, general verbal abilities, and analogies, but no
difference in vocabulary (Hyde & Linn, 1986) or general
knowledge (Herlitz et al., 1997).In this study, females
were faster on tests of speed, but there was a male advantage in reasoning.The latter result is inconsistentwith the
work of Schaie (1994), in which a female advantagein
testsof inductive reasoningwas noted.
One inconsistencyconcernsthe gendereffectsreported
on episodicandprimary memory tasks.Herlitz et al. (1997)
investigatedgenderdifferenceson episodic,semantic,and
primary memory tasks.While they also reported no gender differencesin semanticmemory (which included tasks
similar to those in the knowledge classification), they reported sizablegender differences on severalepisodic memory tasks,in the direction ofhigher levels ofperforrnance
in females,and no genderdifferenceson testsof primary
memory. On closer inspection,however,the apparentdiscrepanciesbetweenthe Herlitz et al. study and the present
analysesare most likely attributable to the types of tasks
used, becausethc moderatoranalysesrevealedthat, when
consideringonly verbal tasksas in the Herlitz et al. study,
there was a female advantageon episodic memory tasks,
and there were no gender differences in primary memorylworking memory tasks.Theseresults suggestthat the
type of stimulus material (e.g., verbal, spatial) should always be consideredwhen investigatinggenderdifferences
in broad task classifications.
The gender relations in the Woodcock-Johnson data,
for the majority of measures,were consistentwith thosein
the primary data set. One exception is in the case of vocabulary where females performed at higher levels than
males did. The largestrelation with genderin this data set
was for the quantitative measures.Superior quantitative
performancein maleshasbeenfound by someresearchers
(Hyde, l98l; Schaie,1994),but not others(Hyde, Fennema,& Lamon, 1990).It is possiblethat theseinconsistenciesare simply the result of the magnitudeof the gender
effects;the effectsare fairly small and greaterthanr: . I on
only two measures(spatial and quantitative).
The age-relatedeffects in this study were similar to
those reported in other cross-sectionalstudiesin that substantial negative relations of age on measuresof speed,
primary memory/workingmemory episodicmemory spatial abilities, and reasoningwere found (seeBlanchardFields & Hess, 1996,and Craik & Salthouse,1992,for reviews). It remainsto be seenwhether comparablepatterns
would be evident in longitudinal designs,and resultsfrom
more extensivesequentialdesigas are neededbefore the
source(i.e., endogenousor exogenous)of theseeffects
can be identified.
Discrepanciesexisted between the Salthouseand
Woodcock-Johnsondata sets in terms of the relation of
age to measuresof knowledge. The data from Salthouse
and collaboratorssuggestmoderatepositive ageeffectson
measuresof knowledge (generalknowledge and vocabulary). Although there was a small positive age relation on
the two vocabulary measuresin the Woodcock-Johnson
IS AGE KINDER?
data, there was a negative relation between age and other
measuresofknowledge (generalknowledge:socialstudies,
science,and humanities). It is not clear whether the different patternsof agerelations on measuresof knowledge
are due to differences in the nature of the samples (i.e.,
conveniencesamplesvs. representativesamples)or to other
factors.One factor that could be involved is amount ofeducation. While education level is very similar acrossage
in the studiesby Salthouseand his collaborators,the education level in the Woodcock-Johnsondatadecreaseswith
age,reflecting population trends. However,this factor is not
responsible for all of the differences in age-knowledge
relations between the data sets,becausea negative ageknowledge relation remains after statistical control of
educationin the Woodcock-Johnsondata set.
The findings of primary interest in this study are those
concerning the interactions ofage and gender.Although
the majority of results from studiesbasedon neuroanatomical data suggest greater age-related atrophy and functional loss in males than in females, the only significant
interactions found in the analysisof the primary data set
were in the direction of greaterage-relateddecline among
femalesthan among males.Thesedataare consistentwith
Schaie's(1994) idea that, if anything, women decline earlier on some fluid abilities, such as perceptualspeedand
reasoning.Schaiealso suggestedthat women show decline earlier on verbal fluency,and men on crystallized
abilities, but thesespeculationswere not supportedin the
presentanalyses.Moderator variable analysessuggested
that genderdifferencesin age-rognition relations may be
related to the type of stimulus material (e.g., verbal, spatial) or method of task administration(e.g., computerized/
reactiontime, paper-and-pencil),
but more investigation
is neededbefore strong conclusionscan be drawn. The
pattern of resultsin the Woodcock-Johnsondatawas mixed,
but, again, almost all of the effects were very small.
In conclusion,the resultsofthese analysesrevealthat
there are large (i.e., between l2oh and 25o/oof the total
variance)effectsrelatedto age,small (i.e.,at most 3%oof
the variance)effects associatedwith gender,and small
to nonexistent(i.e., less than l%o)effects associatedwith
interactions of age and gender on measuresof cognitive
performance.
As discussedin the introduction, evidencefor differential patterns of aging between men and women has been
reported with someneuroanatomicalmeasures.However,
theseresults have been somewhatinconsistentwith respectto which genderexhibited greaterage-relateddecline, and, apparently,no meta-analyticintegrationsof the
neuroanatomicalfindings havebeenpublished.The most
convincingresolutionofthe discrepancies
betweenstudies with neuroanatomicalvariables and with cognitive
variableswill likely require studieswith moderatelylarge
samplesof femalesand males from a wide agerangewho
eachcontribute both neuroanatomicaland cosnitive
data.
-evidence
Unless,and until, such studiesyield strong
of
significant age X gender interactions,any speculation
about the basis for differential
aging among
females
63
and
males seems premature.
REFERENCES
BrrNcnrno-Frnros, F., & Hnss, T. M. (Eos.) (1996). Perspectives
on cognitive change in adulthood and aging. New York: McGrawHill.
CowsLL, P.E., Tunrrsxl B. I., Gun, R. C., GnossulN, R. I., Snrlspl,
D. L., & Gun, R. E. (1994). Sex differences in aging of the human
frontal and temporal lobes. "/ournal of Neurcsc ience, | 4, 47 4847 55.
Curx, F. I. M., & SALrHousE, T. A. (Eds.) (1992). The handbook of
aging and cognilroz. Hillsdale, NJ: Erlbaum.
Durry, F. H., McANuLr1 G. B., & Arnnnr, M. S. (1993). The pattern
of age-relateddifferences in electrophysiological activity of healthy
males and females.Neurcbiologt of Aging,14,73-84.
Elres, M. F., RonnrNs,M. A., Welren, L. J., & Scsulrz, N. R. (1993).
The influence of gender and age on Halstead-ReitanNeuropsychological Test performance. Journal of Gerontology : Psychological Sciences. 4E. P278-P281.
GuR, R. C., Mozlrv, P.D., REsNrcK,S. M., Gorrl-rrs, G.E,.. KosN. M..
ZMrusRMA,N,
R., HEnulN, G., Arlls, S.,GnosslrlN, R., BSRRETTA,,
D.,
Enwrx, R., & Gun, R. E. (1991). Gender differencesin age ef,fecton
brain atrophy measured by magnetic resonance imaging. Proceedings
of the N at io nal A cademy of Sci ences, tE, 2845 -2849HrrpenN, D. F. (1992). Sexdillerences in cognitive abilires. Hillsdale,
NJ: Erlbaum.
Hrrr.rsnrcx,D. 2Z.,Slr-rHousn, T. A., & MErNz, E. J. (1997). Predictors
of ctrtssword puzzle performance and moderators of age--cognition re/ations. Manuscript submitted for publication.
Hroces, L. V., & Olxrn, L ( 1985).SratlsrrZalmethodsfor meta-analysis.
Orlando, FL: Academic Press.
Hpnr-rrz, A., NrLssoN,L.-G., & BAcxunN, L. (1997). Genderdifferencesin episodic memory. Memory & Cognition,2S, 801-8 I l .
HvoB, J. S. ( 198I ). How largeare cognitive genderdifferences?American
Psychologist,36, 892-901.
Hyon, J. S., FpNNrlaa,8., & Leuox, S. J. ( 1990).Genderdifferencesin
mathematics performance: A meta-analysis. Psychological Bulle tin,
107.139-155.
HvnB, J. S., & LINN, M. C. (1986). Thepsychologt ofgender: Advances
through meta-analysis. Baltimore: Johns Hopkins University Press.
Kavr, J. A., DEClnu, C., LuxrNmnc, J. S., & Rnpoponr,S. L (1992).
The significance of age-relatedenlargementofthe cerebral ventricles
in healthy men and women measuredby quantitative computed x-ray
tomography. Jou rna I of t he A meric an Geri atric s Society, 40, 225 -21I .
Lenmsne, G. J., & CRooK,T. H. ( I993). Do men show more rapid ageassociateddecline in simulated everyday verbal memory than do
women? Psychologt & Aging, 8, 68-7 1.
L e n n l n n e , G . J . , T u n e N , D . 8 . , C u n r r s s ,G . , & L E v r x , H . S . ( t 9 8 8 ) .
Normative data for the verbal selective reminding test. Neuropsychologt,2, 173-182.
LINN, M. C., & Prrensrr, A. C. (1986).A meta-analysis
of genderdifferencesin spatial abilities: Implications for mathematicsand science
achievement.In J. S. Hyde & M. C. Linn, (Eds.), Thepsychologt of
gender: Advancesthrough meta,analysis(pp. 67-l0l). Baltimore:
Johns Hopkins University Press.
Meccony, E. 8., & hcxln r, C. N. ( I 974). Thepsychology of sex diferences.Stanford, CA: Stanford University Press.
MstNz, E. J., & SelrnousE, T. A. (1998).The effectsofage and experience on memory for visually presentedmusic. Joarnal of Gerontologt: Psychological Sciences,53B, P60-P69.
MunpHy, D. G. M., DEClnlr, C., McINrosH, A. R., DALY,E., MENrrs,
M. J., hnrruNr, P., SzczEplNlx, J., Scnaplno, M. B., Guov, C. J.,
HoRwrrz, B., & Rapoponr, S. I. (1996). Sex differencesin human
brain morphometry and metabolism: An in vivo quantitativemagnetic
resonanceimaging and positron emissiontomography study on the effect of aging.Archives of General Psychiatry,53, 585-594.
Rlz, N., Tonnrs, I. J., & AcKER,J. D. (1995).Age, gender,and hemispheric differences in human striatum: A quantitative review and new
64
MEINZ AND SALIHOUSE
data from in vivo MRI morphometry. Neurobiologlt of Learning &
Memory,63,133-142.
RossNtnrr, R. (1991). Meta-analytic proceduresfor social research.
Beverly Hills, CA: Sage.
SeLnrousE, T. A. (1991). Mediation of adult age differences in cognition by reductions in working memory and speedofprocessing. Psychological Science,2, l'l 9- 183.
SII-rHouss, T. A. (1992). Influence ofprocessing speedon adult agedifferencesin working memory. Acta Psychologica, Tg, 155-170.
Sllrnousr, T. A. (1993). Speedmediation ofadult age differences in
cognition. Developmental Psychology, 29, 722-7 38.
SlI-rnousr, T. A. (1994a). Aging associations:Influence ofspeed on
adult age differences in associative learning. Journal of Experimental
Psychologt: Learning, Memory, & Cognition,20, 1486-1503.
SllrHouss, T. A. (1994b). The nature ofthe influence ofspeed on
adult age differences in cognition. Developmental Psychology,30,
240-259.
Slltuousn, T. A. (1995a). Differential age-relatedinfluences on memory for verbal-symbolic information and visual-spatial information? Journal of Gerontology: Psychological Sciences,508, Pl 93P201.
Salrnousn, T. A. (1995b). Influence ofprocessing speed on adult
age differences in learning. SwissJournal of Psychology,54, 102fi2.
SALrHousE,T. A. ( I 996). Generaland specific speedmediation ofadult
age differences in memory. Journal of Gerontologlt: Psychological
Sciences,5lB. P30-P42.
Sllrnouss, T. A., & Bucocr, R. L. (1991). Decomposingadult age
differences in cognition. DevelopmentPsychology, 27,'7 63-'176.
SelrsousB, T. A., FRrsroE,N. M., LrNeweevpn,T. T., & Coox. V. E.
(1995). Aging of aftention: Does the ability to divide decline?Memory & Cognition, 23, 59-71.
SlrrnousE, T. A., FRrsroE,N. M., & RHEr, S. H. (1996). How localized are age-relatedeffects on neuropsychologicalmeasures?Neuropsychology, 10, 27 2-285.
Srr-rHouse, T. A., HaNcocx, H. E., MsrNz, E. J., & HnMsnrcK,D.Z.
(1996). The influence ofvisual acuity on age{ognition relations.
Journa I of Gercnto loglt : Psychological Science, 51B, P3 I 7-330.
Sllrnousp, T. A., KAUSLER,D. H., & Slurrs, J. S. (1988). Investigation ofstudent status, background variables, and the feasibility ofstandard tasks in cognitive aging research.Psychologt & Aging,3,29-37.
Snrsousp, T. A., & MrrNz, E. J. (1995). Aging, inhibition, working
memory and speed.Journal of Gercntologt: Psychological Sciences,
50B, P297-P306.
SrlrHouss, T. A., & Mrrcunr-r-, D. R. D. (1990). Effects of age and
naturally occurring experience on spatial visualization performance.
Developmental Psychology, 26, 845-854.
Scslrc, K. W. (1994). The course of adult intellectual development.
American Psychologist, 49,304-3 13.
ScHAIE,K. W. (1996). Intellecual developmentin adulthood: TheSeattle longitudinal study. Canbidge: Cambridge University Press.
VERHAEGHEN,
P., & Slr-rnousE, T. A. (1997). Meta-analyses of
age--cognition relations in adulthood: Estimates of linear and nonlinear age effects and structural models. Psychological Bulletin,122,
23t-249.
WIr-rrs, S., & ScHers, K. W. (1988). Gender differences in spatial abilities in old age: Longitudinal and intervention findings. SexRoles, 18,
l 89-203.
Wooococr, R. W., & JouNsoN, M. B. (1989, 1990). Woodcock-Johnson
Psycho-Educational Battery-Revised.Allen, TX: DLM.
NOTES
l . We would like to thank Richard W. Woodcock for generously sharing thesedata.
2. The ro,uas used to expressgroup differencescould be convertedto
I values by way of the formula 12 : lt2/Q2+df ).
3. Note that tests ofthe age X gender interaction are also possible by
contrasting the 957oconfidence intervals for the effect size estimatesof
the age relations for females and males. However,in order to expressthe
age, gender,and age X gendereffect sizesin a common metric, we focus
on the analysesofinteractions in terms ofcorrelational units.
APPENDIXA
Task Descriptions
Task
Speed
Letter comparison
Patterncomparison
Number comparison
Finding,4s
Digit symbol substitution
Number matching
Patternmatching
Letter completion
Number completion
Number transformation
Letter transformation
Digit copy
Number copy
Letter copy
Symbol copy
Description
Write the letter S if pairs of letters are physically identical; otherwise,
write the letter D.
Write the letter S if pairs of line patternsare physically identical;
otherwise, write the letter D.
Indicate with a wrinen response*trich pairs of numbers are physically
identical.
Mark with a written responsethe items that contain the letter A.
Write the symbol associatedwith a digit according to a code table.
Mark with a written responsewhich of severalalternative setsof
numbers matchesthe target numbers.
Mark with a wdtten responsewhich of severalalternative sets
of patternsmatchesthe target pattern.
Write the letter that is missing from one of two pairs of letters.
Write the digit that is missing from one of two pairs of numbers.
Write the digit that would result by adding either I or 2 to the
displayeddigit.
Write the letter that would result by going one or two letters forward
in the alphabetic sequence.
Copy the displayed digit in the spaceimmediately below
or adjacentto it.
Copy the displayednumber in the spaceimmediately below
or adjacentto it.
Copy the displayed letter in the spaceimmediately below
or adjacentto it.
Copy the displayed symbol in the spaceimmediately below
or adjacentto it.
IS AGE KINDER?
Word copy
Boxes
Line marking
Counting backwards
Naming time
Trail making
Digit digit
Digit symbol
Memory search
Duration threshold
Spatial
Cube assembly
Paperfolding
Spatial rotation
Block design
Object assembly
Surfacedevelopment
Perceptualclosure
Primary/Working Memory
Sentencespan
Computation span
Word span
Digit span
Letter memory
Keeping track (numeric)
Keeping hack (spatial)
Matrix memory (verbal)
Matrix memory (spatial)
Element memory (verbal)
Element memory (spatial)
N-back
Episodic Memory
Associative memory
Copy the displayed word in the space immediately below
or adjacentto it.
Draw a line to make a three-sided figure into a square or box.
Draw vertical lines through short horizontal lines, or draw horizontal
lines through short vertical lines.
Count backwardsby a designatedquantity (e.g., ones, threes,etc.).
Read visually presentedwords.
Connect haphazardlypositioned circles according to numerical or
alternating numerical and alphabetical sequences.
Pressone key iftwo digits are identical; otherwise, pressa different
key (speed).
Pressone key ifa digit and symbol match according to a code table;
otherwise, pressa different key (speed).
Classifu a probe item with respectto whether it was (by pressing one
key) or was not (by pressing another key) in the previously presented
memory set (speed)(Sternbergmemory searchtask).
Perform three simple arithmetic operations (psychophysically determined minimum time needed).
Decide whether arrows on two of six connectedboxes would point at one
another ifthe boxes were assembledinto a cube (speed and accuracy).
Determine whether a particular pattern of holes would result if a
piece ofpaper were folded in the designatedfashion and a hole
punched in the specified location (speedand accuracy).
Identifu by a written responsewhich ofseveral alternative patternsis
physically identical to the target pattern but merely rotated in the
sameplane (speedand accuracy).
Assemble a set of blocks to produce a pattern that matchesthe target
pattern (speed).
Rearrangea set ofpuzzle pieces to form a complete object (speed).
Determine the correspondencebetweendesignatedpositions across
two- and three-dimensionaldrawings of the sameobject (speedand
accuracy).
Identi$ objects from degradedpictures (speedand accuracy).
Rememberthe last word in each sentencewhile also answering questions
about the sentences(accuracy).
Rememberthe last digit in eacharithmetic problem while also solving
the arithmetic problem (accuracy).
Rememberas many words in sequenceas possible (accuracy).
Remember as many digits in sequenceas possible (accuracy).
Remembera sequenceofthree, five, or sevenletters (accuracy).
Remember the latest value of a set of digits that are repeatedly updated
by meansof arithmetic operations(accuracy).
Rememberthe latest position of a set of asterisksthat are repeatedly
displacedby meansof arrows of varying length and orientation
(accuracy).
Rememberthe identities of 7 target letters from a set of 25 letters
(accuracy).
Rememberthe positionsof 7 targetlocationsfrom a set of 25 positions
(accuracy).
Decide whether a single letter was presentedin an earlier memory set
(accuracy).
Decide whether a single line segmentwas presentedin an earlier line
pattern (accuracy).
Report the digit N items back in a sequence(where N : 0, 1, or 2)
(accuracy).
Recognizewhether elementsof a particular pair (i.e., digits, letters,
or symbols) were previously paired together when either member of
the pair was last presented(accuracy).
65
66
MEINZ AND SAIJTHOUSE
APPEIIDD( A (Continued)
Task
Associative learning
Name-number association
Paired associate
Free recall
RAWT
(Rey auditory verbal learning test)
Activitymemory
Frequencyjudgment
Maze learning
Temporal memory
Reasoning
Raven'sprogressivematrices
Number seriescompletion
Integrative reasoning
Geometric analogies
PMA reasoning
Shipley abstraction
Cattell's matrices
Letter sets
Knowledge
Vocabulary
Shipley vocabulary
General information
Verbal Fluency
Letter fluency
Sequentialassociates
Make words
Miscellaneous
Arithmetic computation
Sentencecomprehension
Stroop interference(color)
Stroop interference(position)
Stroop interference (number)
Anagram
Description
Select the secondmember of a pair of elements(i.e., digits, letters, or
symbols) when presented with the first member of the pair (accuracy).
Remember the pairings between numbers and names (accuracy).
Recall the second member of a pair of words when presented with the
first member of the pair (accuracy).
See or hear a list of l2-20 unrelated words and immediately attempt
to recall them in any order (accuracy).
Recall a set of l5 unrelatedwords with repeatedstudy-recall
presentations (accuracy).
Recognizethe order in which a set of activities was performed
in a session(accuracy).
Decide the frequency with which words were presented (accuracy).
Make choice decisions acrosssuccessivetrials in a spatial maze
(accuracy).
Reproduce the temporal order in which I 6 words appeared
(accuracy).
Selectthe best completion of the missing cell from a matrix
containing geometric patterns(accuracy).
Identiff the best continuation ofa seriesofnumbers (accuracy).
Decide the truth of a statementpertaining to two elementsgiven a set
of premisesdescribing relations among premises(accuracy).
Decide which ofseveral alternativesrepresentsthe best D term for
A:B is to C:D1,D2,D3, with stimuli consisting of geometric patterns
(accuracy).
Complete a seriesof numerical items (accuracy).
Complete a seriesof numerical and verbal items (accuracy).
Select the best item for completion of the missing cell from a matrix
containing geometric patterns(accuracy).
Decide which set ofletters does not fit with the others (accuracy).
Selectwhich alternative is the best synonym or anton)rmofthe target
word (aecuracy).
Selectwhich alternative is the best synonym or antonym ofthe target
word (accuracy).
Selectwhich alternative is the best answerto questionson a variety of
topics ranging from American History to Geography,Mythology,
Music, and World Literature (accuracy).
Generatewords that begin with a specified letter (number generated
within a specifedamount of time).
Generatewords that are associatesof the preceding item (number
generatedwithin a specified amount of time).
Create as many words as possible from a set of letters (number
createdwithin a specified amount of time).
Carry out elementary arithmetic operations(speedand accuracy).
Answer simple verbal comprehensionquestions(speedand
accuracy).
Name ink colors when the targetsare words of diferent colors
(speed).
Name target positions when the targets are words indicating different
positions (speed).
Name the number of elementswhen the elementsare digits
representingdifferent numbers (speed).
Rearrangea set of letters to make a word (number completed).
IS AGE KINDER?
APPENDD(B
Age and Gender Reladons from Individual
Studies ln the Primary Analysis
Study
i
Gender r
I
2
3
2
-.10
.07
.05
.05
.06
.09
.05
.03
.00
-.01
.15
.03
-.05
.00
.t2
.03
J
4
)
6
7
8
9
l0
ll
t2
I J
t4
l5
l6
l8
l9
20
2l
22
23
24
25
z
4
J
3
J
J
0
6
7
6
8
7
7
0
.t I
J
6
7
4
2
z
8
z
.01
-.01
.ll
.18
.10
-.11
.15
3
6
7
9
1
2
13
15
2
0
2
2
24
2
5
l
2
2
l
2
I
1
3
2
I
2
-.20
-.t4
-.08
-.22
-.24
-.13
-.18
-.14
-.10
5
6
7
8
t4
t5
l6
17
l8
l9
2l
22
2
2
-.09
-.01
-.0'7
.04
-.06
-.09
-.07
ZJ
24
z
z
6
z
I
4
4
I
J
z
z
z
-
-JZ
-.18
-.ll
-.01
-.05
-.06
-.08
-.12
-.ll
Age Correlation (r) Gender Difference
Female
in Age r
Age r
Male
Speed Tasks
-.48
-.51
-.62
-.47
-.56
-.64
-.55
-.60
-.45
-.52
-.39
-.49
-.35
-.50
-.52
-.58
-.60
-.59
-.56
*.46
-.52
-.53
-.54
-.24
-.48
-.52
-.01
.00
-.t7
.02
.05
-.53
-.50
-.58
-.54
-.56
-.52
-.40
-.60
-.60
-.50
-.18
-.47
-.50
-.56
-.48
-.58
-.72
-.55
-.61
-.54
-.49
-.38
-.49
-.40
-.48
-.55
-.58
-.64
-.68
-.58
-.50
-.46
-.48
-.49
-.34
SpatialTasks
-.38
- -35
-.17
-.18
-.34
-.4t
-.39
-.32
-.38
-.35
-.28
-.19
-.43
-.53
-.46
-.47
-.25
-.24
-.59
-.48
-.34
-.25
- .43
-.13
-.26
-.43
-.39
-.33
- .4r
-.46
-.23
-.39
-.41
.09
-.05
-.17
.13
.04
.15
-.15
-.01
-.02
-.26
.t7
- .45
-.44
-.42
- .34
-.47
-.25
-.62
-.39
-.41
-.48
-.2s
-.41
-.38
-. 15
-.09
-.05
.01
-.01
.09
-.tJ
-.46
-.54
-.59
-.56
-.54
-.35
-.56
-.41
-.50
Primary/WorkingMemory
-.48
-.53
-.46
-.48
-.42
-.41
-.34
-.35
-.43
-.39
-.32
-.43
-.54
-.47
-.44
-.51
-.JJ
-.23
-.36
-.28
-.26
-.28
-.42
-.43
-.41
- .44
-.16
- .17
.23
-.01
.10
.23
-.09
-.05
-.01
.14
-.06
.07
.00
.16
.20
.09
.13
-.19
-.17
-.01
.17
-.Ll
.21
-.16
.20
-zJ
-.05
-.02
-.07
-.01
67
68
MEINZ AND SAIJIHOUSE
Study
9
l0
ll
t2
l3
t4
15
l6
20
2l
22
23
2
2
2
2
I
I
2
2
2
I
r
2
2
3
5
6
2
l
2
2
t
z
9
3
2
I
I
2
2
t
2
13
20
2
3
2
5
J
J
APPEITIDD(B (Condnued)
Age Correlation(r) GenderDifference
Genderr
Male
Female
in Ase r
EpisodicMemory
-.40
-.38
-.43
.08
.06
-.01
-.35
-.39
-.32
-.09
-.25
-.15
-.35
.11
.21
-.27
.
2
4
-.31
.08
.07
-.27
-.22
-.31
.04
.10
-.09
-.41
-.31
-.49
.21
-.04
-.33
-.40
-.30
-.ll
-.52
-.50
-.57
.t7
.09
-.50
-.55
-.48
-.10
.t2
-.19
-.33
-.38
-.35
-.03
-.23
-.28
-.19
-.09
.04
-.29
-.32
-.05
.05
-.16
.04
.00
-.11
-.03
-.13
-.02
-.16
-.05
.00
-.33
-.03
2
2
2
-.04
-.08
-.0'l
-.03
.02
.01
-.06
I
2
l6
2
-.01
-.02
.17
I
l6
t7
l8
24
I
I
2
2
I
2
J
l6
l9
20
2l
z
I
Reasoning
-.42
-.33
-.50
-.58
-.67
-.65
-.35
-.35
-.42
-.38
-.46
-.49
-.35
-.22
-.Jt
-.40
-.09
.t4
-.45
-.46
-.36
-.37
-.26
-.18
Knowledge
.27
.31
.t2
.05
.13
.13
.t7
.25
.52
.59
.25
.19
.23
.19
VerbalFluency
-.ll
-.06
-.18
-.33
.24
.34
-.44
-.44
-.68
-.35
-.47
-.41
-.44
-.31
-.25
-.44
-.36
-.33
.13
-.19
.05
.00
. lI
-.10
.25
-.10
.38
-.03
-.01
.16
.2'l
.16
.t2
.13
.39
.28
.26
.04
-.t2
.01
.12
.26
-.10
-.07
-. 13
-.10
.19
.08
-.24
.16
.00
-.58
-.51
-.51
-.32
.10
-.08
-.09
.16
.03
Miscellaneous
J
.01
.03
.05
.10
.16
.04
-.60
-.55
-.46
-.3I
.10
-.63
-.58
-.38
*.30
Note-For study numbers, see Table 1;7, number of variables contributing to effect size;
gender r, point-biserial correlation with gender (male = 0, female : l); age r, correlation
with chronological age; age r for males/females,seeTable I for number of male and female
participants; gender difference in age r, age X gender interaction term expressedas
r"r"(male) - r"r"(female).
IS AGE KINDER?
APPEI\DD(C
Standard Deviations for Age and
Thsk Classilication According to Gender
Age
Dependent Variable
SD
SD
Study
I
2
J
4
5
6
7
8
9
l0
ll
t2
l3
t4
l5
l6
l8
l9
20
2l
22
23
24
25
5
6
7
9
t2
l3
l5
20
22
24
25
5
6
7
8
t4
15
l6
t7
18
l9
2l
22
1A
df
Male
Speed
t 7. 3 t
14.17
t6.t2
14.65
16.57
r7.2t
t 7. t 4
t6.4t
16.33
15.89
15.82
t6.49
16.48
18.99
t6.79
19.89
t6.22
t9.46
I8.89
t6.17
t6.07
t 7. 2 1
16.39
16.44
1.32
l.l4
t.l I
t.4l
1.28
l.l0
t.l2
t.26
1.09
1.03
I .18
r.r4
l.14
I .15
l.l6
1.0' 1
t.07
l.l I
r.r4
1.33
l .l0
|.02
1.33
L06
0.76
0.85
0.91
0.82
0.89
0.89
0.89
0.78
0.59
0.84
0.54
0.81
0.66
0.95
0.80
0.60
0.84
0.77
0.66
0.81
1.08
0.80
0.73
0.85
0.87
0.94
0.90
0.73
0.91
0.93
0.90
0.98
0.63
0.75
0.55
0.80
0.7r
0.85
0.74
0.55
0.94
0.82
0.75
0.82
0.68
0.99
0.74
0.83
1.33
t.22
1.02
t.26
1.03
1.09
1.03
1.57
l.16
1.28
1.03
1.02
t.t1
r.27
l.l8
t.t7
r.25
l.l4
1.32
t.02
2.52*
1.55
r.02
1.05
Spatial
t6.t2
1 7. 2 t
17.t4
16.33
t6.49
16.48
16.79
18.89
16.07
16.39
16.44
I .l I
Ll0
t.t2
1.09
1.14
l.l4
1.16
r .1 4
l.l0
1.33
1.06
1.00
0.81
0.74
1.00
0.90
1.00
1.00
0.83
0.78
L05
0.95
0.95
0.74
0.65
1.04
0.73
1.04
0.98
0.87
0.'76
0.88
0.85
l.l l
l.l9
r.27
l.08
1.50
1.08
1.04
L l0
r.07
1.41
t.24
PrimaryAV'orking
Memory
18.72 16.57 1.28
18.06 17.21 1.10
18.1l
t7.r4
r.12
18.44 t6.4r
r.26
20.40
18.99 1 l5
ts.62
16.79 1.16
20.60 19.89 1.07
18.47 1s.12 r.49
t6.79
16.22 r.07
20.47
19.46
r. r l
18.67 16.17 1.33
1 6 . 8 s 1 6 . 0 7 1 .l 0
t7.4t
r7.2r
1.02
18.89 16.39 1.33
0.94
0.91
0.94
0.92
0.39
0.89
l.16
0.92
0.77
0.97
0.78
0.81
0.82
0.92
0.83
0.90
0.86
0.87
0.35
0.78
0.91
0.73
0.81
1.04
0.79
0.82
0.82
0.85
r.28
1.01
l .l 9
r.t2
|.22
|.28
1.63
1.59
l.l0
1.14
t.02
t.02
r.02
t.t1
(58,142)
(90,126)
(85,108)
(42,84)
(85,134)
19.90
13.30
15.33
17.42
t8.72
(r24,r02) 1 8 . 0 6
( l 0 e , ll 2 )
1 8 . I1
(s2,46)
18.44
( 123,180)
r7.09
(85,l 53)
16.t4
(5e,64)
r7.20
(e4,l 50)
r7.63
( 1 0 1 , 1 5 5 ) 17.64
(79,92)
20.40
(5e,56)
15.62
(60,1l6)
20.60
( e 3 ,3
18 )
r7.69
(7e,50)
20.47
( 163,94) r 7 . 6 7
(82,1 3) 18.67
(s7,70)
16.85
(107,r24) t't.41
(l0l,l0e) 18.89
( 185,1e6)
16.93
(85,108)
(124,r02)
( 1 0 9 ,l12 )
( 123,180)
(e4,l 50)
(101,15s)
15.33
18.06
l 8 . lI
17.09
t7.63
t7.64
(se,s6) 15.62
( 163,e4) 17.67
(s7,70) 1 6 . 8 5
( 1 0 1 , 1 0 e )1 8 . 8 9
(l 85,196) 1 6 . 9 3
(85,134)
(124,t02)
( 1 0 9 ,1r 2 )
(s2,46)
(79,92)
(59,s6)
(60,116)
(l l0,l 15)
(93,138)
(79,50)
(82,113)
( 5 7, 7 0 )
(r07,124)
(101,109)
69
MEINZ AND SALTHOUSE
APPEIIDD( C (Conttnued)
DependentVariable
SD
study
df
EpisodicMernory
17.09
16.33 I .09
16.t4
15.89 1.03
17.20 15.82 l.l8
1 7. 6 3
t6.49
t.t4
17.64 16.48 l.l4
20.40
18.99 l.l5
15.62 16.79 1.16
20.60
19.89 t.07
t 7. 6 7
18.89 I .14
18.67 t6.17
1.33
16.85 16.07 l.l0
t7.4r
t7.2t
l.o2
0.89
0.83
0.86
0.75
1.03
0.94
0.93
0.87
0.88
1 .l 4
0.69
0.68
0.95
0.83
0.89
0.75
0.98
L08
0.83
0.88
0.89
0.82
0.70
0.68
t.t4
l.0l
t.07
r.02
t.r2
r.32
r.26
1.04
t.02
1.95*
r.02
1.00
( 1 0 9 ,l12 )
( 123,180)
(e4,l 50)
( l 0 l ,I 5 5 )
( 163,e4)
(107,r24)
( 185,196)
19.90
13.30
15.33
r 8.73
18.06
l 8 . l1
r 7. 0 9
t 7. 6 3
t 7. 6 4
t 7. 6 7
t 7. 4 1
16.93
Reasoning
r 7. 3 t
1. 3 2
r4.r7
1.14
r6.t2
l.1l
16.57 I .28
t 7. 2 1
l .10
r 7. r 4
t.t2
16.33 1.09
t6.49
r.t4
16.48 | .14
I8.89
t.t4
t 7. 2 1
1.02
\6.44
L06
0.90
0.95
L00
0.84
0.81
0.74
t.o7
0.95
l.0l
0.98
0.90
0.91
0.92
0.80
0.99
0.91
0.77
0.66
1.12
0.78
1.00
1.02
0.84
0.92
1.05
r.42
l.0l
1.18
l. I I
t.27
I .08
1.48
r.02
L09
l.l3
l.0l
(58,142)
(90,t26)
(85,108)
(60,I l 6)
(7e,50)
(163,e4)
(82,I l 3)
19.90
13.30
15.33
20.60
20.47
r 7. 6 7
18.67
Knowledge
17.3t
t.32
r4.r7
1.14
16.t2 1.1I
19.89 1.07
19.46 l.1l
18.89 | .14
r6.l't
1.33
0.94
0.82
0.83
0.98
1.03
0.99
0.99
0.96
0.96
0.93
I .01
0.98
I .01
0.99
1.04
1.35
1.26
I .07
Llo
I .05
1.02
VerbalFluency
(58,142) 19.90 t'1.3r r.32
(e0,126) 1 3 . 3 0 r 4 . r 7
l.l4
(60,I I 6)
20.60
19.89 t.O1
1.02
0.82
0.80
0.87
0.94
l .08
1.35
t.32
r.82
Miscellaneous
19.90 17.3r
r.32
13.30 r4.r7
r.t4
20.60
19.89 1.07
18.47 ts.t2
1.49
16.79 16.22 r.07
18.89 16.39 1.33
1.04
0.91
t.t4
0.95
0.85
0.61
0.99
1.05
0.94
0.88
0.95
0.84
r.r2
I .33
r.46
L16
1.26
l.9l,r.
9
l0
ll
l2
l3
t4
l5
l6
20
2l
22
23
(123,l 80)
(85,r 53)
(5e,64)
(e4,l 50)
(r01,155)
(79,92)
(5e,s6)
(60,1l 6)
( l 63,e4)
(82,1
l3)
(57,70)
(107,r24)
I
2
(58,142)
(90,126)
(85,108)
(85,l 34)
J
5
6
7
9
l2
l3
20
z3
25
I
2
-t
l6
l9
20
2l
I
2
l6
I
2
16
t7
l8
AA
Male
(r24,r02)
(58,142)
(90,1
26)
(60,1l 6)
( l 1 0 , l1s )
(93,138)
(101,10e)
Note-The F testswere computed by dividing the larger variance value (females or
males) by the smaller variance vahte. df = (number of males - l, number of fem a l e s- l ) . * p . . 0 5 .
(ManuscriptreceivedJanuary3l, 199'l;
revisionaccepted
for publicationSeptember
9, I 997.)