J ournol of Gerontolog\ : P SYCH OLOG I C AL SC IENC ES
1995, Vol. 508. No. 4. Pl93-P201
C opyright 1995 b '- The Gerontolo{ical Socieu^of Amerita
DifferentialAge-Related
Influenceson Memory
for Verbal-SymbolicInformation
and Visual-Spatiallnformation?
Timothy A. Salthouse
School of Psychology,GeorgiaInstituteof Technology.
Adults from a wide range of ageswere administeredparallel versionsof three memory tasks designedto involve
verbal-symbolicinformation or vkual-spatial information. Several analytical procedures were used to determine
whether there wereselectiveage-rehted effects on measuresrfrecting spatial information processingcomparedto
those r$ecting verbal-symbolicinformation processing.The resultsprovided little supportfor this hypothesisand
insteadwereconsistentwith the existenceof a commonage-related
factor that contributesto the agedffirences in
manJ cognitive measures.
KEY issue in researchon cognitive aging is the existence of selectiveor differential deficits in which the
age-relatedeffects on some cognitive measuresare greater
than those on other measures.As an example, it is sometimes claimed that the age-relatedeffects are larger on
measuresof performancefrom tasksinvolving spatialinformation than on measuresof performancefrom tasksinvolving verbal information. If this characterizationis valid, then
(e.g., Goldstein&
it would be consistent
with the hypothesis
Shelly, l981; Klisz, 1978)thatthe right hemisphere
is more
sensitiveto age-relatedeffectsthan the left becausethe right
hemisphereis often postulatedto be more specializedfor
visual-spatialprocessingthan the left hemisphere.Findings
of this type are potentially important becausethey might
allow closer linkagesbetweenbrain and behaviorthan has
beenpossiblein the past.
Although resultssuggestiveof a differentialdeficit can be
quite informative, previous comparisonsof age differences
in verbaland spatialinformationprocessinghavebeenweak
for at leastthree reasons.First, severalof the comparisons
have involved a confounding of type of information with
other factors becausethe measureshave often been derived
from tasksthat differ in numerousdimensionsin addition to
the type of information. For example, among the contrasts
that have been cited as indicative of differential age-related
effects on verbal and spatial processing are letter span vs
perceptualclosuretasks(Kinsbourne,1974),and the WAISR Verbal vs the WAIS-R Performancescales(Klisz, 1978).
Different patterns of age trends on these tests have led to
speculationsof differential or selective aging of verbal and
spatial information processing and also of left and right
hemisphericfunctioning(e.g., Klisz, 1978).However, tests
such as theseoften differ in the role of previous knowledge,
in the familiarity or amount of experiencewith the stimulus
materials, in the presenceof time limits, in the mode of
response(e.g., oral or manual), and possiblyin the amount
of current processingrequired to produce a response.Because of these confoundings, it is difficult to determine
whetherit is the type of informationor theseother aspectsof
the task that contributeto any differential age relationsthat
might exist.
Several studieshave been reportedin which age differences have been examined with verbal and spatial tasks
designedto be equivalentexceptfor the type of information.
Unfortunately, the results from those studies have been
mixed. For example, there are some reports of equivalent
agedifferencesfor nearlyparalleltasksinvolving verbaland
spatial information (Schear& Nebes, 1980; Shelton, Parsons,& Leber, I 982), but thereis at leastone reportof larger
agedifferenceswith a visual-spatialtaskthan with a presumably comparabletask involving verbal information (Tubi &
Calev, 1989).
A secondlimitation of much of the earlierresearchinvestigating age-relatedeffects on tasks involving verbal and
spatial information is that the assessment
of the constructs
has been narrow, and hence construct validity may have
beenweak. That is, many of the contrastsin previousstudies
have been at the level of single measuresrather than at the
more meaningfullevel of theoreticalconstructs.Any given
measureis only an indirect, and henceconfounded,indicator
of the relevant theoreticalconstruct becauseit is also influencedby the specific methods, materials,and measures
usedin the assessment
of the construct.Becausetheseother
influencescannot be separatedfrom the theoretical construct
unless converging operationsand multiple indicators are
available, it is generally desirableto rely on two or more
measuresto attempt to unconfound the assessmentof the
constructfrom its specific operationalization.Reliance on
multiple measuresalso has the advantageof broadeningthe
assessment
of the construct,which will likely increasethe
generalityof the results.
The third limitation of much of the past researchcomparing age-relatedeffects on verbal and spatial tasks is that the
comparisonshave relied on a single analytical procedure,
namely, analysisof variancefocusing on the Age x Task
interaction. This is unfortunate becausenumerous concerns
Pl93
Pt94
SALTHOUSE
have been raised about the meaningfulnessof statistical
and cognitive variablesis sharedwith measuresof processinteractions(e.g., Salthouse,1991),especiallywhen there
ing speed(e.g., Salthouse,1992, 1993a,1994a),the datain
are agedifferencesin the baselinelevel ofperformance, and
this study were examined to determine whether similar
when little information is available about the form of the
relationswould be evident with new combinationsof speed
construct-variable
relationor the discriminatingpower (e.g.,
and memory variables.
reliability, rangein the measurementscale,cf., Chapman&
To summarize, adults from a wide range of ages were
Chapman,1973)of the variables.
administeredparallelversionsof threememorytasksinvolvAt leasttwo alternativeanalyticalprocedurescan be used
ing verbal-symbolicor visual-spatialinformation. The prito investigatedifferentialage-relatedeffects.One consistsof
mary questionwas whether,acrossseveraldifferent analyticonducting the analysis of variance (ANOVA) on scores
cal proceduresand measures,there is evidenceof greater
expressedin standarddeviationunits derivedeither from the
age-relatedinfluenceson measuresbasedon visual-spatial
data of one of the age groups or from the data of a separate
information processingthan on measuresbasedon verbalreferencesample.In additionto minimizing across-measure symbolic information processing. In order to provide a
differencesin variability, this form of standardizationhas
separatesample for standardizationpulposes, 100 college
the advantageofproviding a meaningfulindex ofthe degree
studentsalso performed the same battery of tests.
of overlap of the scores in a relevant distribution. An
exampledescribedin Salthouse(1991, pp. 298-299) illusMnruoo
trates that outcomesof the analysescan be quite different
when this type of comparisonis used insteadof the tradiSubjects.- Demographiccharacteristicsof the student
tional analysisof varianceon the original scores.
and adult samples,with the latterdivided into double-decade
A secondalternativemethod of investigatingdifferential
categories,are summarizedin Table l. It is apparentthat the
or selective age-relatedeffects involves examining the
researchparticipantshad an averageof I to 2.5 years of
amountof independentage-relatedvariancein eachvariable
educationbeyond high school and that their averagehealth
to determineif the variableshave distinct and unique agerating was in the very good range.Although thesecharacterrelatedinfluences(Salthouse& Coon, 1994).That is, rather
istics are typical of participantsin researchstudiesof this
than focusing on the size of the age differencesin either
type, both samples are undoubtedly positively biased in
original or standardizedunits, this type of analysisis coneducationallevel, and possibly health status,relative to the
cernedwith the extentto which the age-relatedinfluencesin
generalpopulation.
one variableare independentofthose in anothervariable. In
particular, if there are significantincrementsin the proporProcedure. - In addition to the tasks describedbelow,
tion of variance(i.e., R' ) associatedwith age in the spatial
most of the participantsin this project also performedtwo
measuresafter the variance in the verbal measureswas
tasks that are not discussed in this report. One was a
controlled,then one could infer that at leastsomeof the agecomputer-administered
versionof the Trail Making test, and
relatedeffects in the spatial measureswere independentof
the other was a computer-administered
associativelearning
(i.e., uniqueanddistinctfrom) thosein the verbalmeasures. test. Resultsfrom these tests will be combined with addiAn additional advantageof regression-based
measuresis
tional data from other studies and described in separate
that becausethey provide estimatesof both the unique and
reports.
the sharedage-relatedvariance,the relativeamountsofeach
The first four tests were administeredwith paper-andtype of influence can be determined. Information of this
pencil proceduresand were designed to assesssensorynatureis valuablein providing quantitativeestimatesof the
motor speed (Boxes, Digit Copying) and perceptualcontribution of each kind of influence, instead of simply
comparison speed (Letter Comparison, Pattern
indicating whether either effect is significantly different
Comparison).Each test consistedof a page of instructions
from zero.
and examplesand a single test page. The task in the Boxes
The primary goal of the presentstudy was to investigate
test was to createas many boxesas possibleby drawing the
the relation between age and measuresof performance in
verbal and spatialmemory taskswithout the problemsidentified above.Nearly parallelverbaland spatialmemory tasks
were designed, and the data were analyzed with several
Table l. DemographicCharacteristicsof Samples,by Age Groups
differentanalyticalprocedures,includingANOVAs on origStudents
I 8-39
40-59
60-88
inal and standardizedscores,and analysesofthe amountsof
(n : 100)
(n : 64)
(n : 55)
(n : 54't
age-relatedvariancein eachvariable that was independentof
Age
2O.l
25.3
4'1.9
70.4
the variancein the other variable. In addition, multivariate
( 5 .I )
(1.7)
(6.5)
(6.4)
analyseswere conductedto examinerelationsat the level of
7o
Female
33.0
5
l
.
6
6
0
.
0
50.0
theoretical constructsrather than only at the level of observEducation
13.9
t4.3
14.5
13.1
able or manifestvariables.
(2.s)
(1.2\
(I .9)
(2.'t)
A secondarygoal of the current project was to examine the
Health
2.O
2
.
O
2
.
3
2.2
influence of processingspeedon the age differences in tasks
(
0
.
8
)
(
t.0)
(0.e)
(
l
.0)
with different kinds of information and processingrequirements. Because previous studies have found that a large
Note. Education refers to years of formal education completed, and
proportion of the age-relatedvariancein a variety of memory
healthrefersto self-ratingon a 5-point scale(l : Excellent,5 : Poor).
VERBAL AND SPATIAL INFORMATION
fourth line in three-sided figures. The task in the Digit
Copying test was to copy digits from an upper squareinto a
lower square. In both of these tests the measureof performance was the number of items completedin 30 sec. The
Letter Comparisontest consistedof 2l pairsof letterswhich
were to be classifiedas sameor different by writing an S or a
D on the line between the members of the pair. The Pattern
Comparisontest consistedof 30 pairs of line patternswhich
were to be classified as sameor different by writing an S or a
D on the line between the members of the pair. In both of
theselattertests,in order to adjustfor guessing,the measure
of performance was the number of correct responsesminus
the numberof incorrectresponseswritten in 30 sec.
The two reaction time tasks involved physical identity
judgments with respect to a pair of letters or a pair of
unfamiliar symbols.The symbolswere similar in size to the
lettersbecauseboth setsof stimuli were constructedfrom an
8 x 8 matrix. The stimuluspair was presentedin a vertical
arrangementin the middle of the computerdisplay, together
with a reminder that the "/" key was to be pressedfor
same decisions and the "2" key was to be pressedfor
different decisions. An initial practice block of 18 trials
was administeredwith each type of stimuli, followed by
blocks of 45 trials eachwith letters,symbols, symbols, and
letters.Both accuracyand reactiontime (RT) in msec were
recorded,but becausemean accuracywas greaterthan96Vo
with both typesof stimuli, only the RT resultswere analyzed
further.
Figure I is a schematic illustration of the sequenceof
eventsin the two versionsof the threememorytasks.Eachof
these tasks was administeredwith two blocks of practice
trials, a block of trials in the verbalcondition, two blocks of
trials in the spatialcondition, and a final block of trials in the
verbalcondition.
Trials in the Matrix Memory task were precededby a
reminder on the screenconcerningthe type of information
(identitiesor positions of the target items) that was to be
remembered.The matrix stimulus, consistingof 7 shaded
lettersin an arrayof25 letters,was then presentedfor 3 sec.
This was followed by the responsedisplay, which consisted
ofseven horizontallines in the verbalcondition, and a blank
matrix in the spatialcondition. Responseswere enteredby
typing the lettersin the verbal condition and by using arrow
keys, followed by pressingthe spacebar, to indicatetarget
positions in the spatial condition. In both conditions the
participants were required to produce seven unique responseson each trial to eliminate biasesagainstguessing.
Each experimentalblock consistedof five trials, and there
were two trials in eachpracticeblock.
Trials in the Element Memory task consistedof a 2-sec
presentationof a complex stimulus, a .5-secinterval with a
blank display, and then the presentationof a probestimulus.
The task was to decide whether the probe stimulus had
previously been an element in the complex stimulus. The
complex stimuli in the verbal condition consistedof seven
randomly selected letters, and the probe stimulus was a
single letter. The complex stimuli in the spatial condition
consisted of seven connected line segmentsforming a pattern, and the probe stimulus was a single line segment.In
both conditions the probe stimulus was an element of the
Pl95
MatrixMemory
FT;FEfn
l r t t r l^ l F l r I
l Y l r l u l zf N l
lEfol|i:Ffn
lTlElrfHTF-l
ElementMemory
GBEAFHD
KeepingTrack
Figure l. The stimuli and trial structurefor the three pairs of memory
tasks.
complex stimulus(i.e., one of the sevenletters,or one of the
line segmentsin the samepositionin the array)on 507oof the
trials. Responseswere communciatedby pressingthe "/"
key for YES, and pressing the "2" key for NO. Each
practice block contained five trials, and the experimental
blocks eachcontained50 trials.
The Keeping Track tasks were basedon those described
by Salthouse,Babcock,and Shaw ( I 99 I ). In eachcondition,
a trial beganwith the presentationof an initial value (i.e., a
digit, or an asteriskin a particularspatiallocation)in eachof
two squares.A seriesof transformationswas then presented
requiring the subject to update the status of the relevant
variable by carrying out the specified arithmetic operation.
or by repositioningthe asteriskaccordingto the length and
direction of the displayedarrow. Finally, a question mark
appearedin one of the squares indicating that the current
value of that variable should be reported. The response
consisted of typing the number in the numeric (verbalsymbolic) condition and using arrow keys to position a
cursor in the final location of the target in the spatial
condition. The measureof performance was the percentage
error in the verbal-symbolic condition and an index of error
magnitude in the spatial condition (in the form of the mean
distancein a Cartesiancoordinate systembetweenthe cursor
Pl96
SALTHOUSE
position and the correcttargetposition). The stimulusdurations, .75 sec per display for numeric information and 1.5
sec per display for spatial information, were selected from
the resultsof the Salthouseet al. (1991) studiesto yield
moderatelevels of accuracyin both young and old adults.
Practiceblocks consistedofthree trials each.and therewere
l5 trials in eachexperimentalblock.
RnsuLrs
Students
Means, standard deviations, and estimates of reliability
(obtained by boosting the correlation between the measures
from the two trial blocks by the Spearman-Brownformula) for
the studentdataarepresentedin Table 2. It can be seenthat the
reliabilities are generally in the moderate range, with the
lowest value of .60 for the matrix memory-spatialvariable.
Relationsamong the variableswere examinedwith confirmatory factor analyses (J<ireskog& Sorbom, 1993).
Based on earlier research (Earles & Salthouse, 1995;
Salthouse,I 993b), the speedmeasureswere hypothesizedto
form three factors: Boxes and Digit Copying comprising a
Sensory-MotorSpeedfactor; Letter Comparisonand Pattem
Comparison comprising a PerceptualComparison Speed
factor;and ReactionTime-Letters and ReactionTime-Symbols comprising a Reaction Time Speed factor. The confirmatory factor analysis indicated that the hypothesized
structureprovided a good fit to the data (i.e., 1, (df : 6, l/
: 1 0 0 ) : 4 . 5 6 ;R M R : . 0 2 6 ; G F I : . 9 8 ; A G F I : . 9 5 ) .
Correlations between the factors were .73 (St : .13)
betweenSensory-MotorSpeedand PerceptualComparison
Speed;.33 (SE : .l l) betweenSensory-MotorSpeedand
ReactionTime Speed;and .48 (SE : . l2) betweenPerceptual ComparisonSpeedand ReactionTime Speed.
Two memory factorscorrespondingto verbal(i.e., matrix
memory-verbal, element memory-verbal, and keeping
track-verbal) and spatial(i.e., matrix memory-spatial,element memory-spatial, and keeping track-spatial)information were hypothesizedfor the memory variables.The con-
firmatory factor analysis indicated that the model provided
an adequatefit to the data (i.e., X'(df : 8, N : 100) =
20.29; RMR : .057; GFI : .94; AGFI : .85). The
correlationbetween the verbal and soatial factors was .78
(SE: .ll).
Adults
Table 3 contains means, standarddeviations, estimated
reliabilities, and the proportionsof varianceassociatedwith
linear and quadratic age-related trends for the six speed
variables and the six memory variables. Notice that all
reliabilitieswere greaterthan .8 and that, althoughthe linear
age relations were significant in all variables, none of the
quadratic(age-squared)age trendswas significantlygreater
than zero.
The memory variablesfrom each set of parallel taskswere
subjectedto an Age (18-39, 40-59,60-88) by Condition
(verbal,spatial)analysisof variance.Although the age(i.e.,
F's > 10.0)and condition(i.e., F's > 143.0)effectswere
significantin eachanalysis,the Age x Conditioninteraction
was significant only in the contrast of the keeping trackverbaland keepingtrack-spatialmeasures[i.e., F(2,170) :
12.42, MS. : 0.711. This interactionin this case was
attributable to a greater increase of errors with age in the
spatialmeasure(i.e., meansof 2.16,3.75, and3.92for the
young, middle, and old groups, respectively)than in the
verbal measure(i.e., meansof .50, .59, and .70, for the
young, middle, and old groups,respectively).Note, however, that the different scalesfor the two measuresmakethis
interactiondiflicult to interpret.
Scoresofthe adultson eachvariablewere convertedto zscoresfrom the studentdistributionto providecommonunits
for comparison.The meansin each of three age groups are
plotted in Figure 2 for the six memory variables.The same
type of Age x Condition analysisof varianceconductedon
the original scoreswas also conductedon thesestandardized
T"bl" 3. Ch"r".,"d
Proportion of
Variance
Table2. Characteristics
of Variables
in StudentSample(n : 100)
Variable
Variable
Boxes
Digit Copying
Letter Comparison
Pattern Comparison
Reaction Time-Letters
ReactionTime-Symbols
MMV
MMS
EMV
EMS
KTV
KTS
Mean
52.3
56.1
12.0
20.5
554
5'79
6.28
5.t2
86.0
I t.7
0.38
2.61
SD
Est. Rel
15.8
8.4
2.8
J.O
86
85
0.49
0.78
9.0
12.6
0.18
0.86
- t-)
.88
.t)
.60
.78
.89
.68
.82
Nores. MMV : matrix memory-verbal; MMS = matrix memoryspatial; EMV : element memory-verbal; EMS : element memoryspatial; KTV : keeping track-verbal; KTS : keeping track-spatial.
Maximum scores in the memory tasks were 7.0 in the MMV and MMS
tasks, 100.0in the EMV and EMS tasks,and 0.0 (no errors)in the KTV and
KTS tasks.
Boxes
Digit Copying
Letter Comparison
Pattern Comparison
Reaction Time-Lrtters
ReactionTime-Symbols
MMV
MMS
EMV
EMS
KTV
KTS
Mean
50.8
50.9
9.5
15.6
748
'792
1 5 t.
I 1.5
3.6
4.4
214
229
4. 9 8
I .38
3.'7t l.16
79.7
10.6
68.5 t2.l
0.59 0.22
3.44
1.37
Est.Rel.
.88
.91
.88
.88
.80
.84
.84
.91
Age
. I 13*
.199*
.246+
.335+
.251*
.256*
.27'7*
.402*
.087+
.070*
. l70x
.142*
Age'
.003
.004
.001
.002
.000
.000
.001
.001
.001
.001
.009
.015
Nores. MMV : matrix memory-verbal; MMS : matrix memoryspatial; EMV = element memory-verbal; EMS : element memoryspatial; KTV : keeping track-verbal; KTS : keeping track-spatial.
Maximum scores in the memory tasks were 7.0 in the MMV and MMS
tasks, 100.0in the EMV and EMS tasks,and 0.0 (no errors)in the KTV and
KTS tasks.
+ p( . 0 1 .
VERBAL AND SPATIAL INFORMATION
MMV
..€F
MMS
-6-
EMV
EMS
o
o
o
KTv
-D-
at
KTS
X
- z
c
o
E
f
U)
€ L
n
GhronologicalAge
Figure 2. Means at three age levels for the six memory measures
expressed in standard deviations from the distribution of l0O college
students. MMV : matrix memory-verbal; MMS : matrix memoryspatial; EMV : element memory-verbal; EMS = element memoryspatial; KTV : keeping track-verbal; and KTS : keeping track-spatial.
scores.The age effectsin theseanalyseswere all significant
(i.e., F's > 10.30),and the conditioneffectswere significant in the matrix memory (F' : 3l .65) and elementmemory (F : 27.07) tasks.Only one Age x Condition interaction was significant,and that involved the contrastof the two
matrix memory measures[i.e., F(2,170) : 6.67, MS, :
2.14). Examinationof Figure 2 revealsthat the interaction
was attributable to larger age effects for the verbal measure
than for the spatialmeasure.
It appearsfrom the results describedabove and from the
datasummarizedin Table 3 and in Figure 2 that the verbal and
spatial memory variablesdiffer with respectto the magnitude
of their relations with age. However, it is important to note
that the rankings of the measureswith respectto degree of
age-sensitivitydiffer acrosstypes of comparisons.For example, the Age x Condition interaction on the raw scores
suggestedthat the keeping track-spatial measurewas more
age-sensitivethan the keeping track-verbal measure,but the
sametype of analysison the standardizedscoresrevealedno
evidenceof differential effects on those variables: instead. it
suggestedthat the matrix memory-verbal measurewas more
age-sensitivethan the matrix memory-spatial measure.Furthermore, the age sensitivity of the matrix memory-spatial
measurewas larger than that of the matrix memory-verbal
measurein the amount of varianceassociatedwith age (Table
3), but the ranking of the measureswas reversedin the
comparisonin standarddeviationunits (Figure2).
The next set of analysesexamined the amount of agerelated variance in each memory measurethat was independent of, or distinct from, the age-relatedvariance in the other
measurefrom the sametype of task. Hierarchical regression
analyses were used for this purpose by determining the
increment in the R' associatedwith age after control of the
variable from the parallel memory task. The residual proportions of age-relatedvariancewere .ll0 @ <.01) for the
Pl91
matrix memory-spatial measurel .017 for the matrix
memory-verbal measure; .020 for the element memoryspatial measure;.035 (p < .01) for the element memoryverbal measure;.051 (p <.01) for the keepingtrack-verbal
measure; and .029 (p < .01) for the keeping track-spatial
measure.
Although two of the spatial tasks have a significant
amountof unique age-relatedvariance,this is also the case
for two of the verbal tasks. Therefore, while it is true that not
all of the age-relatedvariance in these measuresis shared
with the measurefrom the parallel task, this pattern occurs
for measuresfrom tasksinvolving verbal-symbolicinformation as well as for measuresfrom tasks involving visualspatial information.
The results from the analysesdescribed thus far provide
little support for the hypothesis that age-relateddeficits are
greaterin tasks involving spatial information than in those
involving verbal information. However, as mentionedearlier, resultsfrom singlemeasuresmay have limited generalizability becauseof the large specific influencesassociated
with particularmethods,materials,and typesof assessment.
For the next set of analyses,therefore, the variableswere
aggregatedto examineinfluencesat the constructlevel rather
than at the level of single measures.
As with the studentdata, three speedfactors were hypothesized, and a confirmatory factor analysisrevealeda good fit
of the datato this structure(i.e., X'(df : 6, N : 113) :
1 3 . 6 5 ;R M R : . 0 2 9 ;G F I : . 9 7 ; A G F I : . 9 1 ) . C o r r e l a tions between the factors were .84 (SE : .05) between
Sensory-MotorSpeed and PerceptualComparison Speed;
.44 (SE : .07) betweenSensory-MotorSpeedand Reaction
Time Speed;and .61 (SE : .06) betweenPerceptualComparisonSpeedand ReactionTime Speed.
Separateverbal and spatialfactorswere hypothesizedfor
the memory variables,and the confirmatoryfactor analysis
revealedthat this structureprovided a moderatefit to the data
( i . e . ,X ' ( d f : 8 , N : 1 7 3 ): 3 9 . 5 6 ;R M R : . 0 4 7 ; G F I :
.94; AGFI : .84). However, the correlationbetween the
verbaland spatialfactorswas .98 (SE : .04), which was not
significantlydifferent from 1.0. There is thus no evidence
for the existenceof distinct memory factorscorrespondingto
verbal and spatial information in the data from the adult
sample.
A completemeasurementmodel was examinedwith unitary speed and memory latent constructs. Correlated residuals were allowed between the Boxes and Digit Copying
measuresand between the two reaction time measures to
accommodatethe existenceof distinct speedfactors. (The
error covariance between the letter comparison and pattern
comparisonmeasureswas not significantly different from
zero and hence was omitted from the model.) This model
providedan adequatefit to the data(i.e., X'(df : 51, N :
1 7 3 ) : 1 1 4 . 8 6 ; R M R : . 0 5 3 ; G F :f . 9 1 ; A G F I : . 8 6 ) ,
and the correlation between the speed and memory factors
was .75 (SE : .05).
Analyses were also conducted to determine whether the
same measurement model fit the data from the student
sample.Significantreductionsin 12 valueswere obtainedin
the two-sample analysis when factor correlations, factor
loadings, and error variances were allowed to differ across
Pl98
SALTHOUSE
samples,either alone or in combination. It can therefore be
infened that all aspectsof the factor structure differ across
the two groups. In other words, the relations between factors, the relations of variables to factors, and the residual
variances of the variables all differed in the best-fitting
models for the student and adult data. Perhaps of greatest
interest, the conelation between the speed and memory
factorswas .38 (SE : .12) in the studentdatabut .75 (SE =
.05) in the adult data.
A structural model for the adult data was created by
adding age to the model, with paths from age to speed and
froln age to memory. The fit of this model was adequate
(i.e., X' (df : 6l, N : 173) : 146.56:RMR : .055;GFI
: .89; AGFI : .83), and the structuralcoefficientsfor the
path model are portrayed in Figure 3.
Although the direct path between age and the memory
latent construct was -.23, the correlation between age and
the memory construct was -.64. Becausethe total agerelatedvarianceon the memory constructwas .410 (i.e., .64'), whereasthe amount of age-relatedvariance was only
.053 (i.e., -.23' ) when speedwas considered,theseresults
suggestthat a large proportion(i.e., [.410-.053]/.410 x
100 : 87.l%o)of the age-relatedvarianceon the memory
construct was sharedwith measuresof speed.
The proportions of sharedage-relatedvariance were also
examined for each combination of speedand memory variables.That is, for each variable the proportion of age-related
variance shared with every other variable was estimated by
determining the R' for age, determining the increment in R2
for age after control of the other variable, and then subtract-
Figure 3. Coefficients for the structural model illustrating relations
among age, speed, and memory constructs. DCopy : Digit Copying,
LetCom = Letter Comparison, PatCom = Pattem Comparison, RTL =
Reaction Time-lrtters, RTS : Reaction Time-Symbols, MMV : matrix
memory-verbal, MMS : matrix memory-spatial, EMV = element
memory-verbal, EMS : element memory-spatial, KTV : keeping trackverbal, and KTS : keeping track-spatial.
ing the latter from the former, and dividing the difference by
the R' for age. The resultingvalues, which are presentedin
Table 4, reflect the proportion of age-relatedvariance in the
target (column) variable that is shared with the controlled
(row) variable. The matrix is not symmetric because the
amount of age-relatedvariance, and consequently the proportion of age-relatedvariancethat is shared,is not necessarily equivalent for the two variables in each pair.
The entries in Table 4have considerablevariability, with
a range from . 16, for the proportion of reaction time-symbol
age-related variance shared with the element memoryverbal measure,to .99, for the proportion of reaction timesymbol age-relatedvariance shared with the reaction timeverbal measure. An analysis of variance revealed that the
meansfor the four quadrantsin Table 4 (i.e., top left speedspeed : .6671,top right, memory-speed : .584; bottom
feft, speed-memory : .452; and bottom right, memorymemory : .696) were significantlydifferent, F(3,128) :
ll .09, MS" : .036. Bonferroni contrastsindicatedthat the
mean for the age-related variance in the speed variables
sharedwith the memory variables(i.e., .452) was significantly lower than the other values, but that there were no
differences between the means for the memory age-related
variancesharedwith othermemory variables(i.e., .696) and
for the memory age-related variance shared with speed
variables(i.e., .584). At leastbasedon theseresults,therefore, it appearsthat speedis a more important influence on
the age-relatedvariance in the memory measuresthan vice
versa.In fact, the lack ofa significantdifferencebetweenthe
memory-memory and memory-speed values suggeststhat,
on the average, there is nearly as much overlap of agerelated variance between any particular memory variable
and any particular speed variable as between two memory
variables.
An analysis of variance was also conducted on the data
from the memory-memory pairs to determine whether the
proportion of shared age-relatedvariance varied according
to the typ€ of information in the task. The means for the
proportions of age-related variance were: for the spatial
memory measures, .677 shared with other spatial memory
measures,and .696 shared with verbal memory measures;
and for the verbal measures, .738 shared with other verbal
memory measures and .680 shared with spatial memory
measures.An analysisof variance revealedthat there was no
significantdifferenceamong theseconditions(i.e., F(3,26)
: .09, MS. : .005).
Estimatesof the proportionsof sharedage-relatedvariance
can be convertedinto a type of correlation coefficient, which
has been termed the quasi-partial correlation (Salthouse,
1994b). The quasi-partial correlation reflects the average
proportion of age-relatedvariancethat is sharedbetweentwo
variables.It is computedby determiningthe squareroot of the
product of the two proportions for a given pair of variables
and then taking the squareroot ofthat value. For example,the
proportionsfor the MMV and MMS variablesin Table 4 were
.93 and .74, which correspondto a quasi-partialcorrelationof
.91. The mean of the 66 quasi-partial correlations resulting
from thesecomputationswas .75.
Two exploratory factor analyseswere next conducted on
the matrix of quasi-partial correlations with both one-factor
,\
I
1
I
I
VERBAL AND SPATIAL INFORMATION
Pl99
Table4. Proportionsof SharedAge-Related
Variance
Criterion Variable
Controlled
Variable
I
I Boxes
2 DCopy
3 LetCom
4 PatCom
5 RTL
6 RTS
7 MMV
8 MMS
9 EMV
IO EMS
II KTV
12 KTS
t2
l1
X
.99
.82
.99
.'79
.71
.6'7
.75
.27
.45
.58
.61
.75
X
-82
.95
.59
.54
.54
.u
.26
.26
.54
.45
.43
.70
X
.84
.58
.63
.56
.65
.31
.39
.25
.46
.67
.67
X
.47
.51
.52
.70
.26
.31
.46
.3'7
.41
.49
.58
.62
X
.99
.65
.76
.22
-23
.47
.40
.35
.43
.61
.64
.99
X
.f9
.69
.16
.19
.49
.37
.30
.40
.50
.61
.59
.55
X
.93
.50
.29
.59
.46
.70
.69
.76
.99
.'74
.68
.93
.94
.72
X
.90
-92
-)+
.JJ
.42
.60
.50
.47
.74
X
.29
.JZ
.5'7
.41
.56
.'77
.84
.5'7
.44
.99
.97
X
.58
.90
.74
A1
.64
.56
.83
.67
.51
.60
.79
.65
.tl
.@
.87
.99
.58
.45
.81
.93
.51
.54
.'t9
X
x
.71
Notes. DCopy : Digit Copying; Letcom : Letter Comparison; Patcom : Pattern Comparison; RTL = reaction
time-letters; RTS = reaction trmesymbols; MMV : matrix memory-verbal; MMS : matrix memory-spatial; EMV = element memory-verbal;
EMS : element memory-spatial; KTV :
keeping track-verbal; and KTS : keeping track-spatial.
and two-factorsolutions.The resultswith both solutionsare
summarizedin Table 5, whereit can be seenthat 77Voof the
variance is accountedfor by a single factor solution and that
a second factor accounts for an additional 1.5Vo of the
variance. The correlation between the two factors was .69,
and inspection of the rightmost two columns reveals that
when two factors are specified, the pattern of loadings for the
two factors correspondsto the speedand memory variables.
Two important points should be noted about the data in
Table 5. The first is that the communalitiesareall quite high,
with an averageof 777o of the age-relatedvariance in these
variables accounted for by a single factor and nearly 85Zo
accountedfor by two factors. Becausethereis little residual
variance in the variables that is available to be associated
with other factors, any more specific factors that might exist
could presumablyaccountfor relativelysmall proportionsof
the age-relatedvariancein thesevariables.
The second point to note from Table 5 is that the .69
correlationbetweenthe two factors indicatesthat even when a
second factor is extracted, it is highly related to the first
factor. This finding is consistent with the high correlation
(.75) betweenthe speedand memory factors in the measurement model and with the large path coefficient (.59) between
theseconstructsin the structuralmodel portrayedin Figure 3.
Drscussrox
1
l
Before discussingthe implicationsof the current results,
limitations of the presentstudy should be mentioned. One
limitation is that despite the intentions, the measuresselected to involve different types of information may not have
been completely parallel or equivalent in all respectsexcept
for type of information. For example, the keeping traclispatial measurerequired two-dimensional transformations,
whereasthe addition and subtraction operationsin the keeping track-verbal task were in a single dimension. Differences in the processing requirements may therefore have
influencedthe age relations in the verbal and spatial versions
of each task. Furthermore, although performance in both
tasks was assessedin terms of errors, the sensitivity of the
distance-from-the-targetmeasurein the spatial task may not
Table 5. Community Estimates and Loadings From Exploratory
Factor Analyses With I and 2 Factors
Communality
Variable
l-Factor
2-Factors
Boxes
Digit Copying
Letter Comparison
Pattern Comparison
Reaction Time-lrtters
Reaction Time-Symbols
MMV
MMS
EMV
EMS
KTV
KTS
.'741
.7'7|
.746
.831
.'773
.'738
.812
.807
.675
.'7t9
.utt,
.'772
9.237
.'770
.823
.862
.831
.853
.836
.829
.852
.832
.881
.801
.9G04
.841
0.907
.845
.69
Eigenvalue
Cumulative proportion of variance
Correlation between factors
Loadings
.906
.92'1
.910
.901
.909
.909
.76t
.774
.591
.679
.768
.116
.669
.680
.671
.7't3
.700
.662
.904
.884
.935
.891
.936
.910
Nores. MMV : matrix memory-verbal; MMS : matrix memoryspatial; EMV = element memory-verbal; EMS : element memory_
spatial; KTV : keeping track-verbal; and KTS : keeping track-spatial.
have been equivalent to the percentageerror measurein the
verbal-symbolictask. A secondpossible limitation of the
current study is that the influence of type of information
being processedmay have been low relative to other in'
fluences,such as the type, or amount, of requiredprocessing. To the extent that this was the case, the systematic
variance attributable to type of information may have been
obscured by the variance associated with other factors.
These and other as yet unknown characteristicscould have
contributed to the weak differentiation of verbal and spatial
tasks in the present study. Nevertheless, it is important to
emphasizethat: (a) the memory measuresin this study were
all quite reliable; (b) the averagescoreswere generally in the
middle of the measurementrange; and (c) the two memory
constructs were moderately differentiated in the data from
the studentsample. These features,together with the plausi-
P200
SALTHOUSE
bility of the verbal-spatialdistinction in the tasksportrayed
in Figure l, suggestthat the cunent studyprovidesa reasonable, albeit not definitive, test of the hypothesisof differential aging of verbal and spatialinformation processing.
The first major conclusionfrom the resultsof this study is
that there is little evidencefor selectivity or specificity of
age-related effects on verbal and spatial memory tasks.
Among the findings leading to this inferenceare the mixed
resultsfrom the comparisonof individual pairs of measures.
Specifically,the patternwith respectto which memberof the
pair exhibitedthe largerdegreeof age-sensitivitywas inconsistentacrosscontrastsoforiginal scores,of scoresin young
adult standarddeviation units and of proportions of unique
age-relatedvariance.The factor analysisresultsalso fail to
support the hypothesis of differential age-relatedinfluences
on tasks involving visual-spatial compared to verbalsymbolic information. That is, the .98 correlationbetween
the verbal and spatial factors in the confirmatory factor
analysisindicatesthat the measuresare not organizedinto
distinctfactorsin the datafrom the adult sample.The similar
values of sharedage-relatedvariance for combinationsof
memory measuresinvolving the same (i.e., meansof .677
and .738) and different (i.e. , meansof .646 and .680) types
of information in Table 4 are also inconsistentwith the
hypothesisof selectiveor differentialage-relatedinfluences.
The secondmajor conclusion from the presentstudy is
that a large proportion of the age-relatedvariance in the
memory measuresis sharedwith that in the speedmeasures.
This conclusion derives from several sourcesof evidence
obtainedfrom different types of analyticalprocedures.For
example, one relevantresult is the .75 correlationbetween
the speedand memory factorsin the measurementmodel fit
to the datafrom the adult sample.Not only is this value high
in absoluteterms but it is also significantlygreaterthan the
correlation(.38) obtainedin the model fit to the studentdata.
The largercorrelationin the age-heterogeneous
samplesuggeststhat the speedand memory constructsmay be becoming more similar, or lessdifferentiated,with increasedage.
A secondtype ofrelevant evidencederivesfrom analyses
of the proportions of age-related variance shared across
combinationsof memory and speedvariables.An averageof
over 597o of the age-related variance in the six memory
variablesand the six speedvariableswas sharedwith the
other variables(Table 4), and a single factor accountedfor
77Voof theage-relatedvariancein the I 2 variables(Table5).
Finally, the structuralmodel summarizedin Figure 3 led to
estimatesthat 87. l7o of the age-relatedvariance in the
memory factor was sharedwith the speedfactor.
The precedingresultsall seemto convergeon the conclusion that there is considerablecommonality in the agerelated influences on variables representing memory functioning and variables reflecting speed of processing.
Although the directionof the causallinkageis ambiguouson
the basis of the results summarizedabove, two additional
results suggest that speed may be the more fundamental
construct. First, when the structuralmodel was altered to
reversethe direction of the path betweenspeedand memory,
it led to estimatesthat 72.8Voof the age-relatedvariance in
the speedconstruct was shared with the memory construct,
which is less than the 87. l7o of ase-relatedvariancein the
memory construct that was sharedwith the speedconstruct.
And second, analyses of the proportions of shared agerelated variance in pairs of individual variablesrevealedthat
the average amount of age-relatedvariance in the memory
variablessharedwith speedvariables(i.e., .584) was significantly greater than the average amount of age-relatedvariance in the speed variables that was shared with memory
variables(i.e., .452).
The presentresults are therefore consistentwith previous
speculations(e.g., Salthouse,1992)that the speedmeasures
reflect how quickly many different types of processing
operations can be executed. Furthermore, as processing
speed slows down with increasedage, it may become a
central factor influencing the age-relatedeffects in a variety
ofdifferent cognitivetasks,includingthoseinvolving different types of information.
In summary, the resultsof this study do not supportthe
hypothesisthat age-relatedeffects are selectivelyor differentially greaterfor memory tasksinvolving spatialinformation compared to those involving verbal information. Instead, the results suggestthat there may be considerable
commonalitiesin the age-relatedinfluencesacrossdifferent
types of memory measures,as well as across measures
reflectingspeedofprocessing.Futureresearchshouldtherefore not only continue to seek evidencefor differential or
selectiveage-relatedinfluencesbut in addition shouldinvestigate reasonsfor the common age-relatedinfluencesthat
also appearto exist.
ACKNowLEDCMENTS
This researchwas supportedby NIA grant R37 AG-06826. I would like
to acknowledgethe valuable assistanceof M. Bridges, J. Crawford, A.
May, R. Murray, N. Riecke,and B. Smith in recruitingand testingresearch
participants.
Address correspondenceto Dr. Timothy A. Salthouse,School of Psychology, Georgia lnstituteofTechnology, Atlanta, GA 30332-0170.
RspennNcEs
Chapman,L. J., & Chapman,J. P. ( 1973).Problemsin the measurementof
cognitive deficit. P sycholo gical B ul let in, 70, 380-385.
Earles, J. L., & Salthouse,T. A. (1995). Interrelationsof age, health,
and speed. Journal of Gerontology: Psychological Sciences, 508,
P33-P41.
Goldstein, G., & Shelly, C. (1981). Does the right hemisphereage more
rapidly than the left'! J ournal of C linical N europsychology, 3, 65-78.
Jdreskog,K. G., & Stirbom, D. ( I 993). LI SREL8 : U ser' s referenceguide.
Chicago:Scientific Software lnternational.
Kinsboume, M. (1974). Cognitivedeficit and the aging brain: A behavioral
analysis. International Journal of Aging and Human Development, 5,
4l-49.
Klisz, D. (1978). Neuropsychologicalevaluation in older persons.ln M.
Storandt, I. Siegler, & M. Elias (Eds.),Tfte clinical psychology of
aging,(pp.7l-95). New York: Plenum.
Salthouse,T. A. (1991). Theoretical perspectiveson cognitive aginq.
Hillsdale, NJ: LawrenceErlbaum.
Salthouse,T. A. (1992). Mechanismsof age-cognitionrelations in adultftood. Hillsdale, NJ: LawrenceErlbaum.
Salthouse,T. A. (1993a). Speed mediation of adult age differences in
cognition. DevelopmentalP sycholo gy, 29,'722-738.
Salthouse,T. A. (1993b). Speedand knowledge as determinantsof adult
age differences in verbal tasks. Journal ofGerontology: Psychological
Sciences.48.P29-P36.
Salthouse,T. A. ( I 994a). The natureof the influenceof speedon adult age
differencesin cognition. Developmental P sychology, 20, 240-259.
Salthouse,T. A. ( 1994b). How many causesare thereof aging-relateddecrements in cognitive functioning? DevelopmentalReview, 14, 413-43'7.
VERBAL AND SPATIAL INFORMATION
S a l t h o u s eT, . A . , B a b c o c k ,R . L . , & S h a w , R . J . ( 1 9 9 1 ) .E f f e c t s o f a d u l t
age.on structural and operational capacities of working memory. psychologt and Aging, 6, ll8-121.
Salthouse,T. A., & Coon, V. E. (1994). Interpretationof differential
deficits: The caseof aging and mental arithmetic. "/ournal of Experimentcl Psychology:Learning, Memory and Cognition, 20, ll72_llg2.
Schear, J. M., & Nebes, R. D. (1990). Memory for verbal and spatial
information as a function of age. Experimental Aging Researih, 6,
2 ' 7t - 2 8 2 .
P20l
Shelton, M. D., Parsons,O. A., & Leber, W. R. (19g2). Verbal and
visuospatial performance and aging: A neuropsychological approach.
Journal of Gerontology, 37 , 336-341 .
Tubi, N., & Calev, A. (1989). Verbal and visuosparialrecall by younger
and older subjects: Use of matched tasks. psycftalogy and Aging,-4,
493-495.
ReceivedJune2l,1994
Accepted September26, 1994