J ournal of Gerontology : PSYCH OLOCICAL
1 9 9 6 ,V o l . 5 l B , N o . 6 , P 3 l 7 - P 3 3 0
1996 by The Gerontological
Seiety of Arcrica
of Age, VisualAcurty,
Timothy A. Salthouse,Holly E. Hancock,ElizabethJ. Meinz, and David Z. Hambrick
School of psychology, Georgia Institute of Technology.
It hasrecently beensuggestedthot
on ,nan! meanr.resof cognitive
Ihrge Foportion of the age-relatedinfluences
functioning h mediatedthrough a singk commonJactor. Thi hypothesis brrn suppoied by the disioviry that
much of the age'relaredvariancein diflerent cognitivemeasuresis shared,and is not iiitinct or-independent.Thr1yearlier resultswerereplicatedin thisproiect, and il wasalsodiscoveredthat measuresof corrected visual acuityai
processingspeedshareavery-largeproportion of the age-relatedvariancein mcasuresof workingmemory,
learning' and conceptidentification. The apparentimptication is that the commonfacir that af,pearsn iontribute
age'relateddffirences in ma-ny.cognitivemeasuresis quite broad and may refuit a relativefy'generalreduction in
central nemoussystemfunctioning.
research has established that age-related inr\ fluences on many different cognitive variables are not
independent,but instead thatl0%oor more ofthe age-related
variance is shared with other variables (Salthouse, l99}b,
1994b, 1996a). This finding suggeststhat adult age differences in cognition are not exclusively attributable to taskspecific processesbut instead are determined at least partially by broader or more general factors. Variations in these
broad factors are unlikely to be responsible for all of the
observed age differences in cognitive functioning, but it is
important to assessand understandthe contribution of anv
general factors that might exist because the role of morl
specific factors cannot be accurately determined unless the
general influences are first taken into consideration.
One approachthat can be used to investigatethe nature of
the hypothesizedcommon or general factor is to determine
which variables "age together" in the sensethat they share
large portions of their age-relatedvariance. That is, to the
extent that a variable is found to have considerableoverlap
of its age-relatedvariance with the age-relatedvariance in
other variables, then it can be inferred to be either a causeor
a consequenceof the hypothesizedcommon factor.
For example, a number of studies have examined measures of how quickly simple comparison or substitution
operations can be executed. Nearly everyone achievesperfect accuracy in these tasks if enough time is allowed, and
thus performanceis usually assessedin terms of how quickly
the tasks can be completed. Because measuresof performance in tasks of this type have been found to share'75Voor
more of the age-relatedvariance from a variety of cognitive
measures, speed of processing has been postulated to be
centrally involved in the hypothesized common factor
(Salthouse,1993, 1994b, 1994c, 1996a,I996b).
Recently,however,Lindenbergerand Baltes(1994) have
reported that measuresof sensory ability also shared large
proportions of age-relatedvariancewith severalmeasuresof
cognitive functioning in a sample of adults between 70 and
103 years of age. Furthermore, the relations they reported
were apparently not merely a consequenceof difficulty
registeringthe stimuli, becausesimilar patternswere evident
with measuresfrom three different sensory modalities vision, hearing, and balance. Among the possibilities discussedby these authors was that the sensory and cognitive
measureswere related becausethey were both indicators of
the hypothesizedcommon factor that has been postulatedto
contribute to adult age differences in many measures of
cognitive functioning.
The goal of the current project was to investigatethe role
of sensory ability on age-cognition relations in healthy
samples of adults from a younger age range - between
approximately 18 and 80 years of age - than the sample
studied by Lindenberger and Baltes (1994). (SeeAppendix,
Note I ). Only measuresof near visual acuity were examined
becauseLindenberger and Baltes (1994) found similar relations with measuresfrom each sensorymodality, and vision
is the easiestsensory modality to assess.Moreover, visual
acuity was assessedwhile the individuals were wearing their
normal corrective lenses, becauseLindenberger and Baltes
(1994) found that this measure exhibited stiong relations
both with age and with measuresof cognitive functioning.
Note that, becausevision is assessedwhen the research
participantswere wearing corrective lenses,everyonemight
have been expected to have close to optimum acuity if the
optical corrections were fully effective in remediating any
visual defects. However, the research literature contains
many reports of age-related declines in corrected visual
acuity (e.g., Burg, 1966; Chapanis, 1950; Fozard, 1990;
Gittings & Fozard, 1986; Pitts, 1982). There is some difference of opinion as to the primary factors responsiblefor the
age-relatedacuity loss, becauseKline and Schieber(19S5,
p. 3l) claim that "Much of the slight to moderate loss in
static visual acuity accompanying normal aging appearsto
be due to changesin the optic media of the eye," whereas
Weale (1982, p. 167) suggeststhat optical factors are responsiblefor only some of the declinesin visual acuity, with
the rest attributable to loss of neural cells. When acuity is
assessedat relatively close viewing distances, as was the
casein the presentstudies, reductionsin the effectivenessof
accommodation probably also contribute to negative relations between age and visual acuity becauseof a decreased
ability to focus on near objects. Regardlessofthe reasonsfor
the age-related declines in corrected near visual acuity,
however, the visual acuity measureis of interestif it is also
relatedto measuresof cognitive functioning becauseit might
then be another reflection of the hypothesized common
The primary analytical strategy in this project involved
partitioning the varianceamong age. vision. and cognitive
variables to determine how much variance is shared in
various combinations. The goal was to find out which
variables "age together" by, in effect, examining the correlations between the age-related effects on different variables. That is, the age-relatedeffectscan be expressedas the
squareof the correlation(i.e., the covariance),and then the
degree of independenceof the relations between age and
different variables can be examined by inspection of the
overlap of the age-variablecovariances.
Commonalityanalysis(Pedhazur,1982)was the principal
method used to accomplish the variance partitioning. When
there are two predictors(e.9., age and vision) of a measure
of cognitive functioning, three variance proportions are of
interestin commonality analysis.Two of theseproportions
representunique contributionsof age and of vision, respectively. They can be computedwith hierarchicalregression
proceduresand correspondto the increment in Rt associated
with one predictor variable after the variance in the other
predictorvariablehas beencontrolled.The estimatestherefore representthe variance in the criterion variable associated with one predictor that is independent of the other
predictor. These unique variance estimateswould be expected to be high if most of the influences of the predictor
were distinct from the other predictor, but they would be
expectedto be low if most of the influenceswere shared.The
third variance estimate representsthe common variance in
the criterion variable that is sharedbetween the two predictors, and is not unique to either. It is computed by subtracting the estimateof the unique contribution of a predictor on
the criterion variable from the total effects ofthe predictor on
that variable. In the current context, this estimate of shared
variancecan be interpretedas the contribution ofthe hypothesizedcommon or generalfactor on the age-relatedeffects in
the criterion variables.
An extension of commonality analysis proposed by
Salthouse(1992b, 1994b) was also used to expressthe ratio
of shared to total age-related variance in the form of a
correlation coefficient. The traditional Pearson productmoment correlation reflects the square root of the ratio of
sharedto total variance for all of the variance in the variables, and the partial correlation controlling for age conesponds to the square root of the shared to total ageindependent variance in the variables. In contrast, the
quasi-partial correlation is the squareroot of the ratio of the
sharedto total age-related variance. It will be high if much
of the relation betweenthe variablesis becauseof a common
factor associatedwith both variablesand with age, and it will
be low if most of the age-relatedinfluences are unique.
Commonality and quasi-partial correlation analyseswere
conducted both with cognitive measures and with speed
measures as the criterion variables. Speed measures are
interestingbecauseprevious researchhasrevealedthat speed
measuressharea large proportion of the age-relatedvariance
with many cognitive measures(e.g., Bors & Fonin, 1995;
Bryan & Luszcz, 1996;'Graf & Uttl, 1995;Hertzog, 1989;
Lindenberger,Mayr, & Kliegl, 1993;Nettelbeck& Rabbitt,
1992; Salthouse, 1992a, 1993, 1994a, 1994c, 1996a,
1996b;Schaie,1989, 1990).
Analyses from three separatedata setsare reported in this
article. Two data setswere from studiesconductedfor other
purposes, but some of those data were amenable to the
presentanalysesbecausethe participantsspanneda wide age
rangeand measuresof visual acuity and speedwere obtained
from every participant. The other tasks in thesestudieswere
not traditional cognitive tasks and thus only the speedmeasuresfrom those studiesare reported here. The third data set
is from a new study with the samefour speedmeasuresas in
StudiesI and2 and also threemeasuresof working memory
and measuresfrom an associativelearning task and from a
cation task.
Studies I and 2
The purposeof the analysesin the initial two studies was
to investigatethe role of vision on the relations between age
and relatively simple measuresof processingspeed. Of
particular interest was whether strong negative relations
betweenage and correctednear visual acuity would be found
in samplesof healthy adultsbetweenapproximatelyl8 and
80 years of age and the degree to which the age-related
variance in the measuresof processing speed was shared
with the age-relatedvariancein the vision measure.
Subjects.- Participantsin thesestudiesconsistedof 77
and 127adults,respectively,in StudiesI and 2. Descriptive
characteristicsof the participantsare summarizedin Table l,
where it can be seen that nearly all of them reported themselvesto be in good to excellent health. (More details about
the participantsare provided in the complete reports of these
studies; Salthouse, Hambrick, Lukas, & Dell, in press;
Meinz & Salthouse,1996).
Procedure. - Visual acuity was assessedby means of
a near-vision eye chart held at a distance of approximately
30 cm in a room with normal (uncontrolled) ambient illumination. The chart (Scalae Typographicae Birkhauseri,
Birkhauser Verlag, Basel) containedboth Landolt C and
two-digit number stimuli in 10 different font sizes conespondingto Snellenacuity ratiosof .l to 1.0. The assessment
consistedof asking researchparticipantsto read the numbers
or state the direction of the gap in the C with each type of
stimulus, first with the left eye covered and then again with
the right eye covered. The Snellenratio correspondingto the
smallest font size at which this could be accomplishedwith
fewer than two errorsout of the 8 to l6 items at eachfont size
was identified as the visual acuity estimate.Participantsused
any correctivelensesthey had availableduring the testing,
Table l. Characteristicsof Participantsin Studies I and 2
Study I
Health Rating I
Health Rating 2
Health Satisfaction
Head Injury
Vision - Right Eye
Vision - Left Eye
Synonym Vocabulary
Antonym Vocabulary
Letter Comparison
Pattem Comparison
Age r
-. l0
. 6 1+
l)_ /
0 . ll
Age r
). t
[email protected]
-. ll
Education is number of yearsof formal educationcompleted, and Health Rating, Health Satisfaction, and Health-RelatedLimitations are ratings on a
5-point scale where lower numbers indicate better health. Responses to the Cardiovascular Surgery, Hypertension Medications, Head Injury, and
Neurological Treatment items were Yes/No, and thus the meansconespond to the proportion of individuals reporting a positive response. Vision scores are
the averageof the Snellen ratios for the number and Landolt C stimuli. Scoresin the Vocabulary and PerceptualSpeedtests are number of items correct, and
scores in the Reaction Time tasks are in msec.
*P< .ol.
but we have no informationaboutthe recency,or accuracy,
of their optical correction.
Although this particular visual acuity test has not been
widely used in the United States,it has severaladvantages
for the currentpurposes.First, and most important, the test
is from the same set of acuity tables used by Lindenberger
and Baltes (1994) and Baltes and Lindenberger(1995, in
press), and therefore we can examine the replicability of
their results with a very similar assessmentinstrument.
Second,unlike many acuity tests,two typesof stimulus(2digit numbersand Landolt C) are presented,and thereforeit
is possible to determine whether the results are specific to a
particulartype of stimulus.Third, the stimuli are calibrated
in equal Snellenratiosfrom 0.1 to 1.0 in stepsof 0.1, and
thus there is a wide range of sensitivity within the normal
population. And fourth, the acuity estimatesfrom this test
were found to correlate .91 with the estimatesfrom a more
traditionalvisual acuity test (i.e., the LighthouseNear Visual Acuity Test, Modified ETDRS with Sloan Letters)in a
sampleof 19 individuals.
Two of the speedtaskswere administeredwith paper-andpencil procedures. The letter comparison task consisted of
the presentationof pairs of three, six, or nine letters, with
approximately half of the pairs differing in the identity of
one letter. The participant was instructed to write an "S"
(for same) or a "D" (for different) on a line between the
numbers of the pair and to work as many of the items as
possiblewithin 30 sec. The pattern comparisontest was very
similar except that the pairs consistedof patterns composed
of three, six, or nine line segments.Each test beganwith a
pagecontaining severalsampleitems, and then was administered in two separatelytimed (30 sec) sections.The score in
each section was the number of items marked correctly
minus the number of items marked incorrectly, and the
averageof the two scoresservedas the primary performance
The digit-digit and digit-symbol reaction time tasks were
administeredon computers. Trials in each task consistedof
the presentationof a code table at the top of the computer
screenand a pair of probe items in the middle of the screen.
In the digit-digit task, the code table contained nine pairs of
identical digits and hence was superfluous, but in the digitsymbol task, it contained nine digits each paired with a
different symbol. Probe items consistedof pairs of digits in
the digitdigit task and pairs of a digit and a symbol in the
digit-symbol task. Researchparticipants were instructed to
press the " 1" key on the keyboard if the members of the
probepair were the same(i.e., either physically identical in
the digitdigit task or associationallyequivalent in the digitsymbol task), and to press the "2" key on the keyboard if
the membersof the pair were different. A practice block of
l8 trials precededthe experimental block of90 trials in each
task. Because accuracy averaged over 95Vo, the median
reaction time servedas the primary measureof performance
in thesetasks.
No constraintson viewing distance were imposed in any
of the tasks. However, the visual anglesat a viewing distanceof 45 cm were approximately two degreesfor the letter
comparison and pattern comparison stimuli, and four degreesfor the digit-digit and digit-symbol stimuli.
The visual acuity scoreswith the Landolt C and with the
two-digit number stimuli were highly correlated with one
another (i.e., r's > .7), and thus the averageof the two
scores was used as the visual acuity estimate for each eye.
The vision scoresacrossthe two eyes were also moderately
to highly correlatedwith one another(r : .82 in Study l, r
: .49 in Study 2), and thus the averageacrossthe two eyes
was used as a compositevision score (seeAppendix, Note
2). Estimatedreliability of the compositevision scorewas
computed by determining the partial correlation betweenthe
scoresfor the two eyescontrolling for age and then boosting
that value by the Spearman-Brown formula. The resulting
estimateswere .87 in Study I and .59 in Study 2. Because
the results of the analysesreported below were very similar
with the visual acuity scorein eacheye servingas the vision
measure, the aggregation across eyes primarily serves to
increasethe reliability of the vision measure(seeAppendix,
Note 3).
Figure I portrays the relations between age and the composite vision measurein Studies I and 2. It is apparentthat
there were strong negative age relations on the corrected
near-vision acuity measurein samplesranging from l8 to 80
yearsof age.
Regression analyses revealed that the quadratic (agesquared)term was significant in both Study I and Study 2
and was responsiblefor an additional6.6Voof the variancein
Study I and an additional 3.0Voof the variance in Study 2.
Separateanalyses on the subgroups above and below the
median age indicated that the nonlinear effects were attributable to a smaller age relation at older ages. Neither the
gendermain effect nor the interaction of Age X Gender was
significantin either study.
The influence of health measureson the relation between
age and vision was examined by conducting a principal
componentsanalysison the eight health measures(seeTable
l), and then controlling the variancein the componentscores
before examining the relationship between age and vision.
The principal componentsanalysisof the health variables in
Study I indicated that two components had eigenvalues
greater than 1.0. The first component had high loadings on
all health variables except for reports of head injury and of
treatment for neurological disorders and was correlated .29
with age. The secondcomponent had high loadings on the
head injury and neurological treatment variables and was
correlated -. 14 with age. The R2 associatedwith age in
predictionof the compositevision index was .379, and this
was reducedto .293 after control of factor I and to .376 after
control of factor 2.
In Study 2 the principal components analysis revealed
three componentswith eigenvaluesgreater than l. The first
had major loadings from the self-rated heath variables, the
second had a high loading from the cardiovascular surgery
variable, and the third had a high loading from the report of
neurological treatmentvariable. Correlationsof age with the
componentswere .05, -.01, and -. 10, respectively.The
age-related variance in the composite vision measure was
.510, and it was reducedto .507 after control of the first
component;it was reducedto .509 after control ofthe second
Y = .901- .0080q,12=.379
a 0.6
0.8 . \ - . .
\ . . ..
\ \.
. . \ \ . .. .
- . \
y - 1 . t.049- .010(X),r2
Figure l. Relation between composite visual acuity score and age in Studies I and 2. Each point representsa different individual.
component, and it increasedto .516 aftercontrol of the third
The results of the analysesjust describedsuggestthat the
observedrelations between age and vision are not mediated
by poorer health, at least as health is assessedwith the
relatively crude self-report measuresin these studies. Similar analyseswith control ofthe variable ofyears ofeducation
also resulted in little reduction of the age-ielatedvariance in
the composite vision measure.That is, after the amount of
educationvariable was statistically controlled, the R, for age
was reducedfrom .379 to .37| in Study I , and from .5 10io
.479 in Study 2.
The initial analysis on the speed measuresconsisted of
computing correlations, partial correlations, and quasi_
partial correlations between pairs of speed measures. In
Study I the absolute magnitude of the correlations ranged
from .34 to .65, the rangefor the partial correlationswas .24
to .59, and the range for the quasi-partial correlations was
.71 to .86. In Study 2 the rangesof the absolutevalueswere
.27 to .63 for the correlations, .10 to .54 for the partial
correlations,and .58 to .91 forthe quasi-partialcorrelations.
The relatively large values of the quasi-partial correlations
indicate that a substantialproportion of the sharedvariance
between pairs of speed variables was also related to age.
However, it should be noted that one reasonthe quasi-partial
correlationsare larger than the other correlationsis that all of
the age-relatedvariance was reliable, whereas some of the
total variance and of the age-independentvariance was due
to error and hencewas not systematic.
A composite speed index was created by subtracting the
average z-score for the digitdigit reaction time and digir
symbol reactiontime measures(r : .65 in Study I and r =
.50 in Study 2) from the average z-score for the letter
comparison and pattern comparison measures(r : .49 in
Study I and r : .63 in Study 2). Note that the subtraction
reflects the fact that the reaction time measuresare scaledin
time per item, whereasthe comparison measuresare scaled
in items per time. This composite speedindex served as an
additional speedmeasurein the subsequentanalyses.
Influence of Vision on Age-SpeedRelations
Tests for the Age x Vision interactionwere conductedby
enteringthe cross-productterm after the ageand vision term's
in the multiple regression equations with the five speed
measuresas criterion variables. Only one of the interaction
tgrms (i.e., on Digit Symbol ReactionTime in Study 2) was
significantat the specified(cr : .Ol) significancelevel, and
therefore there is little evidence that the relations between
vision and speedvaried as a function of age.
Table 2 contains commonality estimates of the proportions of variance in the speed measures associated with
different predictors. Note that the proportion of variance in
the speedmeasuresunique to vision was near zero for all five
speed measures. This indicates that there was little ageindependentrelation between vision and speed and implies
that almost all of the relation between vision and speedwas
attributable to the age variation.
The estimatesof the variance unique to age ranged from
46 to 66Voof the total age-relatedvariance in Study l, and
from 38 to6l%o in Study 2.The percentageofthe total agerelated variance common to vision averaged43Vo instudy I
and 53Vo in Study 2. It can therefore be concluded tirat
approximately half of the age-relatedvariance in the current
speed measures is shared with a measure of near-vision
Although only about 5O7oof the age-relatedinfluenceson
speedwere sharedwith the vision measure,it is nevertheless
important to note that there were strong negative relations
betweenage and correctednear-visual acuity. and moderate
correlations between the visual acuity meisure and speed
measuresobtained under high visibility conditions. The next
study was thereforeconductedto determinewhether similar,
or possibly even larger, relations of vision would be evident
with measuresfrom higher-order cognitive tasks.
Study 3
Previousresearchhas indicated that processingspeedis a
major factor in the age relations on any cognitive measures
(e.g., Bors & Forrin, 1995;Bryan &Luszcz,1996; Graf &
Uttl, 1995; Hertzog, 1989; Lindenberger,Mayr, & Kliegl,
1993; Nettelbeck & Rabbitt, t992; Salrhouse,1992a, 19i3,
1994a, 1994c, 1996a,I 996b; Schaie, I 989, I 990). The question of interest in this study was whether the age-reiated
variancethat is sharedwith speedis uniqueor whetherit is also
sharedwith measuresof vision. If the laffer is the case, this
would suggestthat a common factor reflectingrelatively broad
centralnervoussystemfunctioning may be responsiblefor the
mediationof age-relatedcognitive differences.
The tasks administeredin this study consistedof the same
four speed tasks used in Studies I and Z and, in addition.
three working memory tasks and two tasksassessinghigherorder cognitive functioning. Two of the working memory
tasks, reading span and computation span, have been used
in several previous studies (Salthouse & Coon, 1994;
Salthouse& Meinz, 1995). The zback task was basedon a
task originally described by Kay (in Welford, 1958) and
Kirchner (1958). It consistedof the presentationof a series
of randomly selected digits with the participant asked to
report the digits n back in the sequence.Values of n equal to
0, l, and 2 were usedin this study.
The two higher-order cognitive tasks were associative
learning (Salthouse, 1994a) and a computer-administered
version of the Wisconsin Card Sorting Test (WCST; Heaton,
Chelune, Talley, Kay, & Curtiss, 1993). These particular
cognitive tasks are of special interest because both yield
measuresofperseveration responsesthat have been found to
increase in frequency with increasing age. Although the
increase in perseveration responses with increased age
seemswell established,particularly for the WCST, there are
two important questions about this phenomenon. First, are
perseveration measuresfrom different tasks highly conelated, as would be expected if they reflect a common construct?And second, are the age differences in perseveration
responsesmediated by age-related differences in working
memory, as might be expected if they are attributable to a
failure to effectively process feedback information (cf.,
Salthouse, 1994a)2 It should be possible to answer these
Table2. CommonalityEstimatesfor SpeedMeasures,StudiesI and2
Unique to
Unique to
Common to
Age & Vision
Letter Comparison
.1 3 1
.1 2 8
.1 3 0
.1 3 0
.1 3 3
.1 7 8
.1 3 6
Pattern Comparison
DigirDigit ReactionTime
Digit-Symbol Reaction Time
Study2(n: 127)
Letter Comparison
Pattem Comparison
Digit-Digit ReactionTime
Digit-symbol ReactionTime
questions with data from a study in which the participants
performed a battery of working memory and associative
learning tasks in addition to the WCST.
The data in this study were examined with two sets of
commonality analyses.The first setof analyseswas identical
to those in Studies I and2, with age and vision as predictors
of the speedmeasures.The secondset of analysesinvolved
threepredictors(i.e., age,vision, and speed)ofthe working
memory and cognitive measures.The goal in theseanalyses
was to determine whether the age-related variance shared
with speed and cognition was the same as the age-related
variance shared with vision and cognition. If so, then this
result would be consistentwith the common factor interpretation. If not, then separatespeed and vision influences on
the age differences in cognition would presumably need to
be postulated.
Subjects.- The sample consistedof 197 adults between
l8 and g2years ofage. None ofthe individualshad partici-
pated in either of the previous studies.Descriptive characteristics of the participants are summarizedin Table 3, where it
can be seenthat most of the participantsreportedthemselves
to be in good to excellent health, and had attendedcollege
for an averageoftwo to three years.
Procedure. - All participants performed the following
sequenceof tasks in a single sessionof approximately two
hours. The tasks included letter compiuison' pattern comparison, synonym vocabulary, antonym vocabulary, digitdigit reaction time and digirsymbol reaction time (in counterbalancedorder), sentencespan, computation span, nback
with n equal to 0, l, and 2 (in counterbalancedorder),
WCST, and associativelearning.
pattern comparison, digit-digit
The letter
reaction time, and digitsymbol reaction time tasks were
identical to those administeredin Studies I and 2. The same
vocabulary tests from the earlier studies were also used in
this study and consisted of 10 four-alternative multiple
choice items for both the synonym and antonym tests.
The reading spanand computation spantasks were identi-
Table 3. Characteristics of Participants in Studv 3
Age Group
I 8-39
Age r
30.0 (6.4)
r4.8 (2.3)
2.0 (0.e)
2.2 (0.9)
2.3 (0.8)
1.5 (1.0)
0 . 0 3( 0 . 1 7 )
0 . 0 4( 0 . 2 1 )
0.69 (0.22)
s.3 (2.8)
4.8 (3.0)
s0.5 (6.0)
ts.t (2.3)
2.1 (0.9)
2.3 (0.8)
2-4 (O.7)
1.7 (0.8)
0 . 0 1( 0 . 1 2 )
0 . 1 6( 0 . 3 7 )
0.t2 (0.32)
6.8 (3.0)
6.1 (3.4)
69.8 (7.0)
14.8 (3.2)
2.0 (0.9)
2.3 (0.8)
2.3 (0.8)
1.8 (0.9)
0 . 1 3( 0 . 3 4 )
0. l l (0.32)
0 . 3 5( 0 .l s )
0 . 3 4( 0 . 1 6 )
1.3 (2.9)
6 .r ( 3 . 3 )
-. l5
Health Rating I
Health Rating 2
Health Satisfaction
Cardiovascular Surgery
Head Injury
Neurological Treatment
Visual Acuity - Right Eye
Visual Acuity - Left Eye
Synonym Vocabulary
Antonym Vocabulary
Nole.'Educationis numberof yearsof formal educationcompleted,and HealthRating, HealthSatisfaction,and Health-RelatedLimitationsareratingson a
5-point scale where lower numbers indicate better health. Responsesto the CardiovascularSurgery, Hypertension Medications, Head lnjury, and
Neurological Treatment items were Yes/No, and thus the meansconespond to the proportion of individuals reporting a positive response.Vision icores are
the averageof the Snellen ratios for the number and Landolt C stimuli. Scoresin the Vocabulary tests are number of items correct.
*p < .ol.
cal to those used in earlier studiesby Salthouseand Coon
(1994) and Salthouseand Meinz (1995). Eachconsistedof a
practice set of trials followed by two experimental blocks
with different items in each set. Trials in the reading span
task involved the presentationof a short sentenceaccompanied by a question and three alternative answers. The researchparticipant was instructedto usethe arrow keys on the
keyboard to position an arow in front of the correct answer
to the questionwhile also rememberingthe last word in the
sentence.After selecting the answer to the questions, a
prompt appeared, requesting the participant to recall the
targetwords by typing them on the keyboard. The number of
sentences (and to-be-remembered words) increased to a
maximum of nine as long as the participant was correct on
both the comprehensionquestion and the recall on at least
two of the three trials at eachlist length. The spanestimate
was the largestnumberof items at which the participantwas
correct on both the comprehensionand the recall on at least
two of three trials. The computation span task was very
similar to the reading spantask, except that it consistedof
arithmetic problems insteadof sentences,and the items to be
rememberedwere digits insteadof words.
The nback task involved the presentationof a sequenceof
l0 to 12 (i.e., n + 10)digits on the computerscreenwith the
participantinstructedto type the digit 0, l, or 2 itemsback in
the sequence.Each digit appearedfor 1.5 sec, and the
appropriateresponsehad to be enteredwithin that interval to
be counted as correct. Participantsreceived practice in each
of the three conditions (i.e., n : 0, l, and 2) before
performing a total of six trials in each condition, with the
conditionspresentedin a counterbalanced
order (i.e., 0-l-22-l-O). The n : 0 condition was primarily a control condition becausethere was no storagerequirementwhen the digit
to be typed was currently on the screen.Performancecould
be less than maximum (l00Vo) in this condition becauseof
confusionaboutthe instructionsand/ordifficulty in locating
the responsekeys and respondingwithin the 1.5-secinterval. The influence of these factors on performance in the
otherconditionswasexaminedby computingthe residualsin
p r e d i c t i o n otfh e n : I a n dn : 2
regressionequation after controlling for the n : 0 score.
However, becausethese residualswere highly correlated
with the raw scores(i.e., .83 for n : I and .90 for n : 2),
only the raw scoreswere usedin subsequentanalyses.
A computer-administered
version of the Wisconsin Card
SortingTest was usedto presentthe WCST. (The computer
program was developedby John L. Woodard, who kindly
allowed us to use it in this study). The standardversionof this
testconsistsof a setof four stimuluscardsand I 28 response
cards, which are to be sorted into the appropriate stimulus
categoryaccordingto principles(i.e., on the basisof color,
form, and number) that had to be discovered, and which
changedthroughout the test. Insteadofpresenting the stimulus and response items as cards, in the computeradministered version they were displayed as boxes on the
computer screen.A responsecard was sorted into the appropriate category by typing a number from I to 4 corresponding to the stimulus item below which the response card
should be placed. The responsecard then appearedunderneath the stimulus card and both auditory (i.e., tones of
different frequencies) and visual (i.e., "Right" or
feedback was presented. The two measuresof
primary interest in this test were the number of categories
(out of a maximum of six) successfullycompleted,and the
percentageof perseverativeerrors in which the participant
continued to respondto a previous category after the sorting
principle had changed.
The associativelearning task was very similar to the tasks
described by Salthouse (1994a). An initial practice block
with two pairs of stimuli was presented, followed by two
blocks of six trials each with six symbol pairs as stimuli.
Trials in the task consisted of the presentation of a single
stimulus item on the left of the screen and a column of six
responseitems on the right of the screen.The responsewas
selectedby using arrow keys to position an ilrow in front of
the designated response item, after which feedback was
presented in the form of an auditory signal and visual
highlighting of the correct response term. A variety of
detailed performancemeasurescan be derived from this task
(seeSalthouse,1994a),but the two of primary interestin this
study were the percentage of correct responses and the
percentageof perseveration responsesin which the same
incorrect responseto a stimulus was repeatedon successive
As in Studies I and 2, participants viewed the stimuli
without constraints;thereforeviewing distancewas not controlled. However, visual angles for the target stimuli at a
viewing distanceof 45 cm were approximately 4 degreesfor
the items in the digit-digit, digit-symbol, and associative
learning tasks, 2 degreesfor the charactersin the reading
spanand computation spantasks, l4 degreesfor the digits in
the nback task, and 6 degreesfor the individual symbols and
"cards" in the computer-administered
24 degrees for the
Age Relations
Means, standard deviations, age correlations and estimated reliabilities of the performancemeasuresare summa-
rized in Table 4. All variables were significantly related to
age except for the WCST perseverativeerror measure, and
the reliability estimateswere all in the moderaterangeexcept
for the associative learning perseverative erors measure.
Becausethe WCST was administeredonly once, no reliability estimates could be computed for the measures in this
The age relationships on the measuresof performance in
the associative learning task were similar to those from two
studiesreported in Salthouse(1994a) where the age correlations were -.41 and -.30 for the percentagecorrect measure
and .36 and .20 for the perseverationelror measure.The age
relations on the WCST measuresin this study were smaller
than those in a recent study (Salthouse, Fristoe, & Rhee,
1996)using the traditional card version ofthe test. That is, in
the earlier study the age correlations were -.41 for the
number of categoriesmeasureand .43 for the perseveration
measure.It is not clear whether the smaller age relations in
the current study are attributable to differences in the format
ofthe test, to sample differences, or to factors related to the
other tests administeredprior to the WCST.
Other measuresfrom the associativelearning task and the
WCST were also examined. Unlike earlier studies
(Salthouse, 1994a), the measureof percentageforgetting in
associativelearning had a low (r : .10) and nonsignificant correlation with age in this sample. The percentageof
conceptuallevel responsesin the WCST had a correlation of
-.20 with age, but it was largely redundant with the other
WCST measures because it was correlated .91 with the
number of categoriesmeasureand -.80 with the percentage
of perseverationerrors measure.
The correlation betweenage and the WCST perseveration
Table 4. PerformanceMeasuresin Studv 3
l 8-39
Age r
Letter Comparison
Pattem Comparison
10.5 (2.9)
1 8 l. ( 3 . 6 )
8.8 (2.9)
1 5 . s ( 3 .l )
'7.3 (2.6)
13.2 (3.6)
Reaction Time
698 (148)
l30l (317)
786 (l9l)
rs72 (329)
852 (184)
l80s (476)
3.9 (2.2)
2.6 (r.4)
3.7 (2.0)
2.3 (l.l)
3.0 (2.2)
2.0 (1.2)
8 4 . 7( 1 8 . s )
66.8 (33.7)
42.r (29.0)
s'7.r (28.'t)
34.9 (22.7)
6t.3 (21.9)
5 1 . 9( 3 1 . 5 )
32.4 (22.6)
Associative Leaming
7o PerseverationError
4 0 . 3( 1 6 . 3 )
13.4 (8.4)
30.9 (13.4)
18.6 (e.6)
3 0 . 3( 1 s . 5 )
l7.7 (8.5)
Wisconsin Card Sorting Test
Number of Categories
7o PerseverationError
4.o (2.2)
20.3 (12.6\
3.6 (2.2)
2 0 . 0( 1 1 . 2 )
2.7 (2.t)
24.0 (r2.O)
Working Memory
Computation Span
Reading Span
"Estimatedreliability is computed by boosting the partial conelation (controlling for age) between the scoreson the two administrations of the test by the
Spearman-Brown formula.
*p <.01.
enor measure was small (r : .09), and the perseverative
measures from the associative learning and WCST tasks
were weakly related to each other (r : .16). The low
correlation between the two perseveration measures provides little evidencefor a common perseverationconstruct.
z-scores for the four working memory measures. In both
cases, higher scores in the composite measures corre_
spondedto better performance.
Hierarchical regression analyses were next conducted
with the working memory measuresas the criterion variables
and age and the composite speedmeasureas predictors. The
age-relatedvariance in the working memory measureswas
significantly greaterthan zero for all measureswhen age was
the only predictor, but it did not differ significantlf from
zero when age was consideredafter control of the composite
speedmeasure.For example, the proportions of age-related
variance for the composite working memory measurewere
.084 for age alone, and .000 for the increment in R2associated with age after control of the speedindex. These results
are very similar to thosefrom severalearlier studiesin which
processingspeedhas been postulatedto mediate age-related
influences on untimed measuresof working memory (e.g.,
Salthouse, 1992a; l994d; Salthouse & Babcock, t99t;
Salthouse& Meinz, 1995).
Table 6 contains the results of hierarchical regression
analyseswith cognitive measuresas the criterion variables
and age and the composite speed and composite working
memory measuresas predictor variables. Notice that there
was substantialreduction in the age-relatedvariance ofeach
criterion variable after control of the composite working
SpeedandWorking Memory
Table 5 contains correlations, partial conelations, and
quasi-partialcorrelationsfor the speedand working memory
measures. In all cases, the product-moment correlations,
representing shared total variance, were somewhat higher
than the partial correlations, representing shared igeindependentvariance. The quasi-partial correlations, repielgnling shared age-relatedvariance, were consistently the
highest. This pattern indicates that the variables shared a
large proportion of their age-related variance, but much
smaller proportions of their total variance or of their ageindependentvariance.
Composite speed and working memory variables were
formed for later analyses.The compositespeedmeasurewas
created by subtracting the average of the z-scores for the
digirdigit reaction time and digit-symbol reaction time measuresfrom the averageof the z-scoresfor the letter comparison and pattern comparisonmeasures.The compositeworklng memory measure was formed from the average of the
Table 5. Correlations Between SpeedMeasuresand Between Working Memory Measures,
Study 3
Digit-Digit-l-etter Comparison
Digit-Digit-Pattem Comparison
Digit-Symbol-L,etter Comparison
Digit-Symbol-Pattem Comparison
[ftter Comparison-Pattern Comparison
Computation Span-Reading Span
Computation Span-Nback- l.
Computation Span-Mack- I
Reading Span-Mack-l
Reading Span-Mack-2
Table 6. Increment in R'Associated with SuccessivePredictors in Hierarchical RegressionAnalyses, Study
Associative [,eamine
7o Conecl
7o Perseveration
No. of Categories
7o Perseveration
.l 3 l *
*p < .ol.
memory measure. However, the reduction of age-related
variancewas also considerableafter control of the composite
speed measure, and the residual age-related variance was
nearly the same after control of only the speed measureas
after control of both the speed and working memory measures. The finding that a large proportion of the age-related
variance in measuresof working memory and higher-order
cognitive functioning from tasks without time limits is
shared with simple measures of processing efficiency is
consistentwith the resultsof numerousrecentstudies(e.g.,
Lindenberger et al., 1993; Salthouse, 1992a, 1993, 1994a,
when age was the only predictor and was .385 after control
of the self-rating component, .327 after control of the cardiovascularcomponent, and .299 after control of the neurological component. There was some reduction in the relations
between age and vision after control of the health variables,
particularly after control of the variance in measures of
reports of head injury and treatment of neurological disorder. However, this is a somewhatdifferent pattern than that
observedin Study I and may simply reflect sampling variation. There was little reduction of the age-relatedvariance in
the compositevision measureafter control of the number of
yearsof education(i.e., from .405 to .401).
The estimatedreliability of the compositevision measure,
computed in the same manner describedin Studies I andZ,
was .80. The relationbetweenage and the compositevision
measureis portrayed in Figure 2, where it can be seen that
the parametersof the regressionequation were very similar
to those in Studies I and 2. Tests of the quadratic age trend,
of the gender main effect and of the Age x Gender interaction indicated that none were significant.
The influence of health measureson the age-vision relations was examinedin the samemanneras in Studies I and2.
The principal componentsanalysis on the eight health measures yielded three components with eigenvalues greater
than 1.0. The components,definedin terms of the variables
with the highest loadings (and the correlationsof the components with age), were: self-ratings (.14), cardiovascular
(.41), and neurological (-.14). The age-relatedvariance
(i.e., R'associatedwith age)in the vision measurewas .405
Influence of Vision on Age-SpeedRelations
Tests were conductedfor the interaction of Age x Vision
on the speedvariables, but the interaction was not significant
for any speed measure. As in Studies I and 2, therefore,
there is little evidence that the relation between vision and
speedvaries as a function of age.
Table 7 contains the commonality estimatesfor the speed
criterion measures.Note that there was relatively little variance sharedbetween vision and speedthat was independent
of age, but that about one third of the total age-related
variancein speedwas independentof vision. The estimates
of the common influencein this study were somewhatlarger
than those in the earlier studies, but the overall pattern is
generallysimilar to that in StudiesI and2.
n o
^ a
a \
a a
o a
. . t
a a
o \aaa
a aaal
- a
\ a - a
a ooa-
.a .
. \
a a - a
a a
- o
aa |}..a
a a
a o
a o
Influence of Vision and Speedon Age-Cognition Relations
Interactionsof Age x Vision and Age x Speedwere
examined in multiple regressionequationswith the working
memory and cognitive measuresas the criterion variables.
None of the interactions were significant for any of the
cognitive criterion variables, and thus there is no evidence
that the relations between vision and working memory or
betweenvision and the cognitive measuresvary according to
Table 8 contains results of the commonality estimates
with age, vision, and speed as predictors of the cognitive
measures.Notice that a very similar pattern was evident with
all measures. The unique contribution of age was quite
small, and most of the age-relatedvariance was sharedwith
both vision and speed. These results are consistent with the
earlier findings that a large percentage of the age-related
variancein measuresof working memory and of higher order
cognition is sharedwith a measureof speed(also seeTable
6). However, the previous findings are extended by the
discovery that vision is also a component of that factor. An
average of almost 89Vo of the age-related variance in the
working memory and cognitive measureswas shared with
both vision and speed.
Figure 2. Relation between composite visual score and age in Study 3.
Each point representsa different individual.
Structural Equation Model
The final analysisexamined the fit of a structural equation
model with a single common factor postulatedto mediatethe
age-relatedinfluences on all speedand cognitive variables'
Severalof the measureswere transformed (i.e., the reciprocals of the digirdigit and digirsymbol reaction time values
were multiplied by 10,000, and the nback and associative
learning percentagecorrect values were divided by l0) to
Table 7. CommonalityEstimatefor SpeedMeasures,Study 3 (n :
Unique to
Unique to
Common to
Age & Vision
. 0 r5
Digit-Digit ReactionTime
Digit-Symbol ReactionTime
. 0 t3
obtain similar variancesof the measuresfor the analysis.A
single common factor with relationsfrom age and to all of
the variables was then specified, and each variable was
examinedto determineif it had a significantrelationdirectly
from age. The only variableswith direct relationsfrom age
werethe two vision measures.Despitelittle attemptto model
relationsamong variables,exceptto allow correlatedresiduals betweenmeasuresderivedfrom the samemethods,the
model provided a moderately good {it to the data (1, 1df :
5 9 1: 1 3 5 . 0 6 , S t dR. M R : . O 7 , G F I: . 8 9 , A G F I : . 8 3 ,
CFI : .94). The model, with significantpath coefficients
expressedin standardizedform, is illustratedin Figure 3.
General Discussion
Two sets of results from the present studies were rather
surprising.The first unexpectedresultswere the strong negative relations betweenage and a measureof correctedvisual
acuity found in three independentsamples(Figures I and 2).
The secondsurprisingsetof resultswas the high proportion of
the age-relatedvariancein measuresof speed,working memory, associativelearning, and concept identification that was
sharedwith the measuresof vision (Tables 2, 7, and 8).
Although similar findings were reportedby Lindenbergerand
Baltes(1994), their samplewas composedentirely of older
adultsand eventhey suspectedthat the relationsofthe sensory
measureswould be much reducedin a younger sample. (But
note that these researchershave recently extendedtheir research to a wider age range and found similar results; cf.,
Baltesand Lindenberger, 1995,in press).Furthermore,many
investigatorshave screenedresearchparticipants for vision
but have not reportedthat large numbersof potential participants were excluded for this reason, and they still found
significant age differencesin many cognitive measures(e.g.,
Hahn & Kramer, 1995; Hartman, 1995; McCalley,
Bouwhuis,& Juola, 1995;Paul, 1996).This raisesthe posiibility that there is somethingunusualabout the current vision
assessment,and it is certainly true that corrected visual
acuity was not measuredunderoptimum conditionsbecause
therewas no control over illumination and no restraintswere
usedto ensurethat the viewing distancewas exactly 30 cm.
Nevertheless,the measureswere generallyquite reliable as
evidentin the reliability estimatesand in the strongrelations
with other variables.Moreover, the age relationswere also
apparentlynot attributableto declining healthbecausethere
was relatively little attenuationof the relationsbetweenage
and vision when measuresof self-reportedhealth were
controlled.Of course,the rangeof healthstatusexaminedin
thesestudieswas likely quite limited comparedto the general populationbecausenearly all of the participantsin the
currentstudiesreportedthemselvesto be in good to excellent
It is also important to note that there was little evidence
that the relation betweenvision and either the speedor the
cognitivemeasuresvariedas a function of age. This conclusion is admittedlybasedon acceptanceof the null hypothesis, but becausethe pattern of a nonsignificantinteraction
betweenage and vision was replicatedacrossthree or more
measuresin eachof threeindependentstudies,it can probably be treatedwith someconfidence.The relationsinvolving
vision therefore do not seem to be attributable to visual
pathologiesemergingonly at middle or late adulthood.
Becausecorrelationsamong age, measuresof cognitive
functioning, and measuresof visual functioning have been
reportedby Clark (1960) and Heron and Chown (1967), the
data from those studies were reanalyzed to estimate the
amount of age-relatedvariance in their cognitive measures
that was sharedwith the vision measures.The assessment
vision in the Clark (1960) study was in terms of "near
accommodationdistance" but specificdetailsof the stimuli
or viewing distance were not provided. Analyses of the
correlationsin the Clark sample of 102 adults between20
and 70 years of age revealedthat this vision measureshared
58.l%oof the age-relatedvariancewith the PMA Reasoning
measure and 50.77o of the age-related variance with the
PMA Spacemeasure.Heron and Chown (1961) assessed
Table8. CommonalityAnalyseson CognitiveMeasures,Study3
Criterion : Computation Span
Unique to Age
Unique to Vision
Unique to Speed
Common to Age & Vision
Common to Age & Speed
Common to Vision & Speed
Common to Age, Vision, & Speed
Criterion = Reading Span
Unique to Age
Unique to Vision
Unique to Speed
Common to Age & Vision
Common to Age & Speed
Common to Vision & Speed
Common to Age, Vision, & Speed
Criterion : Nback-l
Unique to Age
Unique to Vision
Unique to Speed
Common to Age & Vision
Common to Age & Speed
Common to Vision & Speed
Common to Age, Vision, & Speed
Criterion : Mack-2
Unique to Age
Unique to Vision
Unique to Speed
Common to Age & Vision
Common to Age & Speed
Common to Vision & Speed
Common to Age, Vision, & Speed
Criterion = Associative Leaming,
Unique to Age
Unique to Vision
Unique to Speed
Common to Age & Vision
Common to Age & Speed
Common to Vision & Speed
Common to Age, Vision, & Speed
Criterion : WCST,
Number of Categories
Unique to Age
Unique to Vision
Unique to Speed
Common to Age & Vision
Common to Age & Speed
Common to Vision & Speed
Common to Age, Vision, & Speed
.1 3 8
. ll 9
Figure 3. Structural model with a single common factor mediating the
age-relatedinfluences on all observed variables. Numbers are standardized
coefficients. The variables were: WCST-NC = number of categories
achieved in the WCST; AssocPC : percentage correct in associative
learning; NB-2 : percentagecorrect in the nback task with n : 2; NB- I :
percentage correct in the nback task with n : l; WM-N : working
memory with number stimuli (i.e., computationspan);WM-V : working
memory with verbal stimuli (i.e., reading span); DSRT : digit symbol
reaction time; DDRT : digit-digit reaction time; Patcom : pattern
comparison; lrtCom : letter comparison; Vision-R : visual acuity in
right eye; and Vision-L : visual acuity in left eye.
.1 8 0
visual acuity with Landolt C stimuli viewed at a distanceof 6
meters with both eyes (uncorrected). Analyses of the relevant correlationsrevealedthat, in their sampleof 300 males,
3l.4%o of the age-relatedvariance in the score on the Raven's ProgressiveMatrices was shared with the vision measure and that 46.OVowas shared in their sample of 240
females. Neither of these studies reported estimatesof the
reliability of the vision measures,and thereforeit is possible
that the smaller proportions of shared age-relatedvariance
than in the current studies are attributable to less reliable
assessmentof vision. It is neverthelessimportant to note that
similar results indicating that moderateto large proportions
of the age-relatedvariance in measuresof cognitive functioning are shared with measuresof visual functioning are
apparentin other data sets with different measuresof vision
and cognition.
Following Lindenberger and Baltes (1994), there seemto
be three possible interpretationsof the relation between the
vision and cognitive measures.One possibility is that visual
deficits are responsible for the cognitive deficits because
individuals with low vision may not be able to adequately
register the stimuli. Although intuitively reasonable, this
interpretationdoesnot seemplausible in the current situation
becauseall stimuli were of high contrast and relatively large
in terms of visual angle. That is, the smalleststimuli subtended visual angles of about two degrees at a typical
viewing distance and should have been easily visible by
everyone because all participants had composite vision
scoresabove zero. It is also important to note that there was
little or no relation between the vision measuresand either
the speedor the cognitive measuresthat was independentof
age (cf., Tables 2,'1 , and 8). The influenceof vision was
thereforelimited to the aspectsof the speed, working memory, and cognitive tasksthat were affected by increasedage.
A secondpossibleinterpretationofthe influenceofvision
in thesestudiesis that the visual assessment
involves cognitive processes.That is, researchparticipantsneedto comprehend instructions, to maintain concentration, and to have at
least a minimum level of motivation to perform well in the
visual acuity task. Although it is true that all these aspects
are required in the vision test, the cognitive demands are
almost certainly substantially less than those in tasks explicitly designedto assesscognitivefunctioning.Ifthe cognitive
requirementsin the vision test are responsible for the relations with the other variables, therefore, a very primitive
form of cognition must be involved. An interesting, and
testable,implication of this interpretationis that the relations
to the cognitive measures should be greatly reduced or
eliminatedif the visual assessments
could be obtainedwith
little or no cognitive involvement.
A third interpretation of the vision-cognition relations is
that the visual acuity measuresare anothermanifestationof a
common factor contributing to the age differences in many
behavioralvariables.This interpretationdiffers from the first
becauseinstead of attempting to attribute causal priority to
certain variables, all of the variables are considered to be
reflectionsof a common factor that is related to both age and
to the variables. To illustrate, although a path model with
vision postulated to mediate many of the age-related influenceson the other variables would likely provide a good
fit to the data, similar good fits would probably also occur for
models based on any variables having moderate to high
loadings on the common factor. From this perspective,
therefore, it may not be particularly productive to attempt to
specify the causal priority among the variables with the
currently availableinformation.
The key feature of this last interpretation is that the
common factor is postulated to contribute to much of the
age-relatedinfluenceson a wide rangeof cognitive measures
(e.g., Salthouse,1994b;Salthouse,
Fristoe,& Rhee, 1996).
The novel contribution of the current studies and of the
studiesby Lindenbergerand Baltes (1994) and Baltes and
Lindenberger(1995, in press)is to suggestthat fairly basic
sensory functions are also included within the common
factor. That is, the common factor appea$ to reflect primitive central nervous system functions representednot only
by measuresof processingefficiency, but also by measures
of correctedvisual acuity. However, it should be emphasized that the common factor is not responsible for all
observedage-relatedeffects becauseunique or independent
age-relatedinfluences were evident on the vision measures
in Study 3 (cf., Figure 3), and independentage-related
influenceshave been found on measuresof memory in other
studies(e.g., Salthouse
et al., 1996).
It is interestingto consider the implications of the hypothesizedcommon factor for a theorist attempting to accountfor
the age differences observed in measuresof working memory or in a cognitive task like associativelearning or concept
identification. The theorist could attempt to account for the
observedage differences in thesetasks by focusing on fairly
specific processes,such as the ability to inhibit irrelevant
information, or to shift set when receiving changing feed-
back. However, the resultsofthese and other recentstudies
indicate that there may be no significant age differences in
the to-be-explainedmeasuresafter the researchparticipants
are statistically equatedon measuresof speedor measuresof
near-visualacuity. These findings imply that the age-related
influenceson many cognitive measuresare not independent
of the age-relatedinfluenceson measuresof simple processing efficiency and of corrected visual acuity. This in turn
indicates that the age-relatedeffects on the cognitive tasks
are not restricted to processesspecific to those target tasks.
Limiting the explanatory focus to a single task may therefore
result in misleading conclusions about the factors responsible for the age-related differences in that task. If, as the
results of these and other studies seem to indicate, those
differences are not independentof the differences in other
tasks, then the theorist may simply be describing one symptom or manifestation of a much broader phenomenon by
concentratingexclusively on the results of a single task.
Although the researchdescribed above raises at least as
many questions as it answers, it does suggestan important
priority for future research.That is, a major goal should be to
explore the nature of the hypothesized common factor by
determining what other combinationsof variablesshareagerelated variance and by determining what variables have
independentage-relatedinfluences.The discovery that measuresof sensoryability appearto be involved in the common
factor also suggeststhat the range of variables to be examined should be expandedto include noncognitive variables.
This researchwas supported by NIA Grant AGR-37 6826 to Timothy A.
Salthouse. We would like to acknowledge the valuable contributions of
Paul Baltes and Ulman Lindenberger in many phasesof this project and for
supplyingcopiesof the visual acuity tests.
Address correspondenceto Dr. Timothy A. Salthouse, School of Psychology, GeorgiaInstituteofTechnology, Atlanta, GA 30332-0170.
Baltes, P. B., & Lindenberger, U. (1995). Sensorik und Intelligenz:
lntersystemische Wechselwirkungen und Veraenderungen im hohen
Alter. Akademi e-J ournal, 20-28.
Baltes, P. B., & Lindenberger, U. (in press). Emergence of a powerful
connection betweensensoryand cognitive functions acrossthe adult life
span: A new window on the study of cognitive aging? Psychology and
Bors, D. A., & Forrin, B. (1995). Age, speedof informationprocessing.
recall, and fl uid intelligence. Intelligence, 20, 229-248.
Bryan, J., & Luszcz, M. A. ( I 996). Speedof information processing as a
mediator between age and free recall performance. Psychology and
Burg, A. (1966). Visual acuity as measuredby dynamic and static tests.
Journal of Applied Psychology, 50, 460-466.
Chapanis, A. (1950). Relationships between age, visual acuity, and color
vision. Human Biology,22, l-31.
Clark, J. W. ( l960). The aging dimension: A factorial analysisof individual
differences with age on psychological and physiological measurements.
Journal of Gerontology,I 5, 183-187.
Fozard,J. L. (1990). Vision and hearingin aging. In J. E. Binen & K. W.
Schaie(Eds.), Handbookofthe psychologyofaging (3rded., pp. 150170). San Diego: Academic Press.
Gittings, N. S., & Fozard, J. L. (1986). Age changesin visual acuity.
Experimental G erontolo gy, 2 1, 423-434.
Graf, P., & Uttl, B. (1995). Component processesof memory: Changes
acrossthe adult lifespan. SwissJournal of Psychology, 54, ll3-138.
Hahn, S., & Kramer, A. F. (1995). Attentional flexibility and aging: You
don't need to be 20 years of age to split the beam. Psychology and
Asing, 10,597-609.
Hartman, M. (1995). Aging and interference: Evidence from indirect
memory tests. Psychology and Aging, 10,659-669.
Heaton, R. K., Chelune,G. J., Talley, J. L., Kay, G. G., & Curtiss, G.
(1993). Wisconsin Card Sorting Test Manual: Revised and Expanded.
Odessa, FL: Psychological AssessmentResources.
Heron, A., & Chown, S. M. (1967). Age andfunction. Boston: Little,
Hertzog, C. (1989). Influences of cognitive slowing on age differences in
intelligence. Developmental Psychology, 25 , 636-651 .
Kirchner, W. K. (1958). Age differences in short-term retention of rapidly
changing information. Journal of Experimental Psychology, 55,
Kline, D.W., & Schieber,F. (1985).Vision and aging. In J. E. Birren & K.
W. Schaie(Eds.), Handbook of the psychology ofaging (pp.296-331).
New York: Van NostrandReinhold.
Lindenberger,U., & Baltes, P. B. (1994). Sensoryfunctioning and intelligence in old age: A strong connection. Psychology and Aging, 2,
Lindenberger,U., Mayr, U., & Kliegl, R. ( 1993).Speedand intelligencein
old age.Psychologyand Aging,8,207-220.
McCalley, L. T., Bouwhuis, D. G., & Juola, J. F. (1995). Age changesin
the distribution of visual attention. Journal of Gerontology: Psychologic a l S c i e n c e s5. 0 8 . P 3 l 6 - P 3 3 1 .
Meinz, E. J., & Salthouse,T. A. (1996). Experienceandage effectsonthe
recall of visually presented music. Unpublished manuscript.
Nettelbeck,T., & Rabbitt, P. M. A. ( 1992).Aging, cognitiveperformance,
and mental speed.Intelligence, 16, 189-2O5.
Paul, S. T. (1996). Searchfor semanticinhibition failure during sentence
comprehensionby younger and older adults. Psychology and Aging, I I ,
Pedhazur, E. J. ( I 982). M ultiple regr ession in behavioral researc h : Explanation and prediction. (2nded.). New York: Holt, Rineha( & Winston.
Pitts, D. G. (1982). The effectsofaging on selectedvisual functions:Dark
adaptation,visual acuity, stereopsis,and brightnesscontrast. In R.
Sekuler, D. Kline, & K. Dismukes (Eds.), Aging and human visual
f u n c t i o n( p p . l 3 l - 1 5 9 ) . N e w Y o r k : A l a n R . L i s s .
Salthouse, T. A. (1992a). Influence of processing speed on adult age
differences in working memory. Acta Psychologica, T9, 155-170.
Salthouse,T. A. (1992b). Shifting levelsofanalysis in the investigationof
cognitive aging. Human Development, 35, 321-342.
Salthouse, T. A. (1993). Speed mediation of adult age differences in
cognition. Developmental P sychology, 29, 722-'7 38.
Salthouse,T. A. (1994a). Aging associations:Influenceof speedon adult
age differences in associative leaming. Journal of Experimental Psy'
cholo gy : Learning, M emory and C ognition, 20, 1486- 1503.
Salthouse,T. A. (1994b). How many causesare there of aging-related
decrements in cognitive functioning? Developmental Review, 14,
Salthouse,T. A. (1994c). The nature ofthe influence ofspeed on adult age
differences in cognition. Developmental P sychology, 30, 240-259.
Salthouse, T. A. (1994d). The aging of working memory. Neuropsychology,8, 535-543.
Salthouse,T. A. (1996a). General and specific speedmediation ofadult age
differences in memory. Journal Gerontology: Psychological Sciences,
5tB. P30-P42.
Salthouse, T. A. (1996b). The processing speed theory of adult age
differences in cognition. P sychological Review, I 03, 403-428.
Salthouse,T. A., & Babcock, R. L. (1991). Decomposingadult age differencesin working memory. DevelopmentalPsychology,27,'163-'776.
Salthouse,T. A., & Coon, V. E. (1994). Interpretationof differential
deficits: The caseof aging and mental arithmetic. J ournal of Experimen'
tal Psychology: Learning, Memory, and Cognition,20, 1172-1182.
Salthouse,T. A., Fristoe, N., & Rhee, S. (1996). How localizedare agerelated effects on neuropsychological measures?Neuropsychology, 10,
S a l t h o u s eT, . A . , H a m b r i c k , D . 2 . , L u k a s , K . , & D e l l , T . C . ( I n p r e s s ) .
Determinants of adult age differences in synthetic work performance.
Journal of Experimental Psychology: Applied.
Salthouse,T. A., & Meinz, E. J. (1995). Aging, inhibition, working
memory, and speed. Journal ofGerontology: Psychological Sciences,
508, P29'7-P306.
Schaie, K. W. (1989). Perceptual speedin adulthood: Cross-sectional and
longitudinal analyses.P sychology and Aging, 4, 443-453.
Schaie, K. W. ( 1990). Correction to Schaie ( 1989). Psychology and Aging,
5, l'71.
Weale, R. A. (1982). Senile ocular changes,cell death, and vision. In R.
Sekuler, D. Kline, & K. Dismukes (Eds.), Agrng and human visual
f u n c t i o n( p p . l 6 l - l 7 l ) . N e w Y o r k : A l a n R . L i s s .
Welford, A. T. (1958). Ageing and human skill. London: Oxford University Press.
ReceivedOctober I 3, ],995
Accepted April 8, 1996
l. These researchershave recently extended their sample by adding adults between 25 and 7O years of age (Baltes & Lindenberger, 1995, in press)and have found resultsquite similar to
those reported in the Lindenberger and Baltes (1994) study.
2. The correlations were much lower in Study 2 becausesix of the
participantsin this study were effectivelyblind in one eye (i.e.,
they could not report the identity of the items in the largest font
size in the vision chart). Elimination of the data from individuals with scores of zero in one eye increased the correlation
betweenthe measuresin the two eyes from .49 to .67. For these
individuals the eye with the nonzero score was used as the
compositevision score.
3. The use of the average score from the two eyes assessed
separatelyrather than the scoreof the best eye, or the score from
an assessmentwith both eyes together, probably contributed to
somewhat lower acuity estimatesthan those reported by other
investigatorsbecausethis measureemphasizesthe worst rather
than the best measureof acuity.
Related flashcards
Create Flashcards