Memory & Cognition
t984, 12 (6), 613-620
Division of attention: Age differences
on a visually presentedmemory task
TIMOTHY A. SALTHOUSE. JANICE DAVENPORT ROGAN. And KENNETH A. PRILL
IJniversity
of Missouri,Columbia,Missouri
.
I
I
I
I
Young and old adults were comparedin their efficiency of remembering concurrently presented
seriesofletters and digits in three separateexperiments.Instructions and payoffsto vary attentional emphasis acrossthe two types of material in different conditions allowed the examination
ofattention-operating characteristics in the two age groups. Strategy-independent measures derived from these attention-operatingcharacteristicsrevealedthat older adults exhibited greater
performance deficits than young adults when dividing their attention between the two tasks,
even though dual-task difficulty was individually adjusted for each subject.It was concludedthat
either the total amount of attention available for distribution or the efficiency of its allocation
decreasedwith age even though the ability to vary one's attention between concurrent tasks in
responseto instructions and payoffs remained intact.
Difficulties in dividing one's attentionacrosstwo or
more activities have beenpostulatedto be responsiblefor
many of the perceptual, cognitive, and motor deficiencies observedwith increasedage. For example,Wright
(1981)assertedthat "one of the most replicablefindings
about short-termmemory changeswith increasingage is
that older adults' performanceis affected more adversely
by divided attention conditions than is that of younger
adults" (p. 605). Burke and Light (1981) and Craik
(197'l) have drawn similar conclusionsbasedon extensive reviews of the literatureon memory and aging. Indeed,severalstudies(e.g., Caird, 1966;Inglis & Ankus,
1965; Inglis & Caird, 1963; Parkinson, Lindholm, &
Urell, 1980)have reported that older adults generally exhibit greater performanceimpairmentsthan young adults
when required to divide their attention betweentwo concurrent tasks in dichotic-listeningsituations.
However, we believethat at leastthreeproblemshamper the interpretation of these divided-attention studies:
lack of control over the individual's relative emphasison
one task or the other, unknown resourcerequirementsfor
each task, and uncontrolled age differences on each task
when performed in isolation. With respect to the first
problem, one cannothope to quantify the dual-taskdecrement if the magnitude of the decrement varies with
differential emphasis on the two tasks; for example, a
small decrement might result with heavy emphasis on
Task I and light emphasison Task 2, but a large decrement might be obtainedwhen the tasksreceive equal emphasis.The secondproblem relatesto the fact that Task I
This research was supported by a Research Council grant from the
GraduateSchool, University ofMissouri, and ResearchCareer Development Award N.I.A. I K04 AG00I46-01A1 to T.A.S. We would like
to thank the anonymous reviewers for helpful suggestionson an earlier
draft of this article. Please address reprint requests to T. Salthouse,
Department of Psychology, University of Missouri, Columbia, MO
65211.
may require, say, 5% of the total attentionalcapacityto
producea unit increasein performance,whereasTask 2
may require orlJyl% of the capacity to achievecomparable performance improvement. Because performance
generally varies across individuals on both concurrent
tasks, only qualitative comparisonsof the severity of
divided-attention impairment have been possible in the
earlier studies.
With respectto the third problem, the addedcomplexity posedby the division of attention may have different
effectsdependingupon the proficiency with which the subjects handlethe tasksin single, focused-attention,
conditions. If different individuals perform at varying levels
in single-taskconditions, it is likely that they differ in the
proportion by which task difficulty is increasedby the requirementof having to perform two taskssimultaneously.
As a consequence,many divided-attention comparisons
in the past may have been confounded with overall level
of difficulty such that the poorer-performing individuals
in the single task experienceda greater increment in overall difficulty in the divided-attention conditions than the
better-performing individuals becausethey were already
operating closer to their performance limits.
The first two of these problems seem resolvable with
a modification of a procedure introduced by Kinchla
(1980)and Sperling(1978; Sperling& Melchner, 1978).
Their method is to obtain data across several dual-task
conditions, with each condition involving different relative emphases on the two tasks. In this manner, an
attention-operating characteristic (AOC) can be constructed in which performance on Task I is represented
along the ordinate and performance on Task 2 is
representedalong the abscissa.A given point on the AOC
signihesa particular combinationof Task I emphasisand
Task 2 emphasis,but the complete function indicatesthe
overall, emphasis-independent,divided-attention effect.
Moreover. becausethe axesof the AOC are scaledin units
of performance on each task, one can directly compare
613
Copyright 1985 PsychonomicSociety, Inc.
614
SALTHOUSE,ROGAN, AND PRILL
the effects of performance change in Task
I in units of
Task 2, thereby solving theproblEm ofunkno*n
resource
requirements.
Our modification to the AOC procedure is to
use the
area above the AOC as a measure of divided_attention
attentiontask was to rememberas many items
as possi_
ble from the two channelswhen they *"i"
f..r"nt"O .on_
presentingthe stimuti one pair ar a
:lli.",lf
.j"stld q{
ttme, all the
stimuli were presentedsimultaneouslyto
facilitate resourceallocation acrosschannels;
Gut ir, m.r"
Sarthouse,
re82).rhe reasoningshouldhavebeenmore leeway
:::::J:1.
:"T?:..e.a
for participantsioOistrib_
rs rnat pertect division of attentionwould be
manifestei ute their attentionthan with paced
siquential presentation.
c.onsistingof a single point correspondingro
A.limited presentationtime (3 ,..) *u, employed
ll^"1_loc
the lntersectionof the lines representingmaximu.
to
minimize organizarional factors that might f,uri
i.riJJt"J
formance on eachtask. This puit".n *ouid ,ho*
that per_ in the transfer of information into
toni+erm memory.
formance on each task is unaffectedby Oemanas
for per_
To someresearchers,it might ,".rn i.unga to use
formance on the other. Such an eOi would
the
"divided
encompass term
attention" in the presentconiext. Our ar_
the entire area ofthe dual-taskspace,and
thereforethe
gument.forthe presentusageis is follows:
Somethingis
divided-attentioncost would be b. Less+han_maximum
responsiblefor effective performance on both single
ind
performance on one or both tasks would result
in AOCs
dual tasks;that somethinghasclear limitations in
ti'at per_
below and to the left of the optimum point,
and thus the
formancecannotbe infinitely increased;and whatever
it
area above the AOC can be interpreteh as a reflection
of
is.caneasily be allocatedo, diuidedacrossdistinct
tasks.
the costs of divided attention.
Theseare all characteristicscommonly anributed
to the
The problem ofuncontrolled age differences in
single_ conceptofattention, and the fact that
tire taskshad a du_
task conditions of dichoticJistening experiments
was dis_ ration of 3 sec meansmerely,that there
was ample time
cussedby Parkinsonet al. (19g0), who pointed
out that
for the operationof all relevanrprocesses.It is true
that
many studiesreported trends for older indlividuals
to have thesecharacteristicsalso apply to structural
concepts
such
smaller
spans
than their young counterparts. In
T"--9ry
as memory capacityor the numb€r of slots available
in
a study of their own, parkinson et al. fo-undthat
dichotic_ some finite. storage system. and the
present research
listening differencesbetween their youn!
unO ota ug"
results might as easily be interpreted;ith a structural
groups disappearedwhen subjectswere mitched
on difit
m:qphor. However, becauseof our interestin the process
"b"
span and screenedfor h_ea_ring
deficits. Thus, it rnuy
of alteringemphasisfrom one task ro another.and
in inthat slightperformance differencesare enlarged
when the
terpretingthe resultingfunctionsas reflecrionsof differen_
number of mental operationsrequired to periorm
the task
tral allocationof a flexible_capaciq.u.eprefer a more
dy_
is increased, as is certainly the case when two tasks
are
namicconceptualization
of resourie limitarionsto account
performed concurrentlv.
for the findings.
An even more impreisive demonstrationof the
impor_
tance of controlling for single_taskperformance
when
making comparisonsin dual-tasksituationswas
EXPERIME\T
I
evident
in a recent experimentby Sombergand Salthouse(lgg2).
Method
Here, the samepattern of attentio=n
allocation was found
Subjects.
Twenty-four
ln young and old adults with two concurrent perceptual
collegestudenrs
I meanase= | E.9r ears,
range=l8 to 22years)
and24olderadulrs
rmean
discriminationtasksafter stimulusdurationsweie
ole=69 5 1ears.
adjusted *".9:=.5?,982.years)
participared
in a srngle
.e.ion nf approxi_
yr,"tgequivalentperformanceacrossagegroups
mately 1.5 h. There were ll males and l-1
in the
1-"
iemalcs In each group.
slngle-taskconditions.Theseresultsarelspeciaily
con_ l^.-:lT,"l finding in rhe psychologrcal trrerarure on'agrng rs that
vincing becausethe equatingtechniqueuuold"a ..select lncreased age rs associated with poorer perlbrmant-e on
speeded
measures, but is either unrelated to or gr.srtrrelr
gloup" interpretations that might beapplied to the par_
correlated nrth
abitity..
The preseni sampte of sublecrs *ere
kinson etal. (1980) result deJcribeaabove. Similarly,
T:1:T1"f,_y.rbal
conslsrent
wlth these trends in that rhe rcxrng sub.Jects
irad hrgher
although the two groups were identical in their dual_task scores on
the s@-based Wechsler .lOuh lnreihgence
Sca.le(WAIS)
performance, they still exhibited typical age trends
digit symbolsubstitution
in the
test.[65 2 r, .r: i,-ii+Or=-A.tS,
p .
duration required for a fix-edlevel ofpe.ceptual accuracy .00011,
but lowerscoreson rhl Nelson-Denny
normC Vocabu_
laryTest[20.8vs. 23.1;t(4g;=2 93.p < .05f m. i"n", perhaps
with each component task. The uppu..nti.plication
,
is
duein,partto a greateraverage
that age differences found in othei studies may
numberof yearsof education
in
not be
theoldersampte
,rGr=i.zo,ii.".doo]f . Thecur[16.4vs.13.4':
solely attributable to the unique requirements of having
rent samplescan therefore be consideredrepresentative
of their
to divide one's attentionbetweentwo concurrentactivities. respectrve
populations,at leastin termsofthe abovemeasures.
InIt secmeddesirable to extend the Somberg and Salthouse deed, if anything, the older
were superior
to the young
'
subjectson the dimensionof-subjgcts
(1982) procedure to a more demandingrn"rnory
veibal ability.
task that
Apparatus. Stimuli were presentedon a computer-controlled
might be expectedto involve greater atounts of pro".rr_
monirorposirionedin fronr ofthe seatedsublect.A
ing over.a longer period of time than the percepiual
dis_ l:lil9i:pl"y
laDoratory
computerwasusedto generaterandomlyorderedseries
crimination task. The presentexperimentstherifore
em_ of letters (all consonants)and dilits (0_9)and to i..oiO and ana_
ployed a vlsual concurrent-memorytask with two
lyze responses.
distinct
Procedure. The subjectswere first given instructionsatrrur
setsof material, eachconstitutinga ..channel." One
rhe
set general
natureof the task, and memory spanfor both drsrr: and
of material consistedof letters and the other of digits
to
A list of materiaiconsirrea.tli;.;;;l
facilitate "channel" categorization, and the divided_ letterswas thenassessed.
column of randomlygenerateditems with the con\rrarnl
lhtrt rhe
I
I
t
r
I
DIVIDED ATTENTION
same item could not occur in consecutive positions. A trial cons i s t e do f p r e s e n t a t i o n o f t h e l i s t f o r I t e c , i f t e r w h i c h t h e s u b j e c t
attempted to orally reproduce the items in their proper (top-tobonom) sequence.The responseswere keyed into the computer by
the experimenter, and a correct trial was defined as all items being
rcported in the correct sequence.The number of items started with
three and was increased by one with two correct reproductions of
the sequenceuntil four of five trials were incorrect, at which time
the span was identified as the previous sequencelength. The maximum sequencelength correctly reproduced in each oftwo separate
trials therefore defined the span. Two blocks of letters and two
blocks ofdigits were presentedin a counterbalancedorder, and the
averageof the two assessmentsserved as the memory span for each
type of material.
I n t h e d u a l - t a s kt r i a l s , b o t h a s e r i e so f d i g i t s a n d a s e r i e so f l e t ters were presented simultaneously for 3 sec, with subjects being
required to respond, again orally, to both. The two setsofmaterial
were arranged in two columns horizontally separatedby approximately 5'of visual angle, with the material to be reported first always on the left. The number of items presented in the dual-task
conditions was 75% (truncated to the nearest integer) of the individual's span length for each type of material, that is, 75% span
length of digits and'15% span length of letters, yielding a total of
I 50 % of the average of the spans for the two setsof material. (The
150% value was chosen to ensure that the composite task requirements exceededa subject's capacity, but was not so overwhelming
that it made the task too frustrating.)
Five experimental conditions were distinguished by the emphasis (manipulated by payoffs of 0Q to 40 per correct response)the
subjects were to give to each of the two memory tasks: 0/4, l/3,
212,311, and 4/0. Responseswere always required to both series,
but in the 0/4 and 4/0 emphasisconditions, random guessesfor the
unattended series would have been sufficient.
Therc were two blocks of trials, with each block containins five
subblocks of l0 trials for each emphasis condition. The orJer of
emphasisconditions (Ol4 to 4lO vs. 4/0 to 0/4) was counterbalanced
across subjects. One-half of the subjects reported digits first and
letters second for the first block, and the reverse for the second
block. The remaining subjects reported letters first and digits second in the first block, and the opposite in the second block.
Results
The young adultshad slightly higher letter spans[5.88
v s . 5 . 5 2 ; t ( 4 6 ) : 1 . 6 9 1 a n d d i g i t s p a n sU . 2 7 v s . 6 . 9 6 ;
(40) = l. l6l thanthe older adults,but the differencewas
not significantwith either type of material. The percentagesof correct responsesin the dual-taskconditions were
subjectedto a2 (age)x2 (material)x2 (order)x5 (emphasis)analysisof variance.Age was a between-groups
variable, and all other variableswere within groups. Two
additional analyseswere also conducted. In one, the
criterion for scoringrecall attemptswas relaxedto count
an item as correct whetheror not it was in the proper serial
position. This "free-recall" analysisyielded resultsnearly
identical to those reported below, but with a slightly
smallerorder effect due to the secondorder's benefiting
more from the relaxedcriterion. In anotheranalysis,the
possibilityof differencesbetweenthe trial-block sequences
(i.e., performanceon the first trial block vs. performance
on the second)was examined,but no main effectsor interactionswere found, so that datawere collapsedacross
this variable.
All the main effectsin the initial analysiswere significant: age-older individualsperformedat a lower over-
6I5
all levelthanyoungindividuals[F(l,46):5.35, p < .05];
material-letters were more difficult to recall than digits
lF(I,46):12.65. p < .0011;order-the secondseries
reportedgave the subjectsmore difficulty than the first
series [F(1,461:274.t0, p < .0001]; and, finally,
emphasis-the subjectswere able to shift their attention
from one task to the other [F(4,184):520.72, p <
.00011.The agedifferencesmust be qualified, however,
by the presenceof a significantage x order interaction
in Figure 1, both
lF(I,46):5. 12,p < .051.As illustrated
agegroupswere worse at recallingthe secondseries,but
the decreasein performancewas greaterfor the older individuals. It also can be seenby the parallel curves that
both agegroupswere comparablein their abilitiesto shift
attentionfrom one task to the other; similarity in the trends
for both age groupsis also indicatedby the lack ofa significant age x emphasisinteraction.Only two other interactions were significant: order X emphasis
of moreexrreme
lF(4,184):159.26,p < .00011,because
scoresfor the first seriesrecalledat both low and high
emphasis conditions, and material x emphasis
scores
lF(4, 184): 4.16, p < .0051,becauselow-emphasis
tendedto drop lower for letters than for digits.
Becausethe interactionbetweenageand order was significant, a closerexaminationof the differenceswas made
by separateanalyseson eachorder. The slight difference
in favor of young adultson the first set recalledwas not
significant,and no differencewas found with respectto
materialon this set. The subjectsdid alter their attentional
emphasison the first-recalledmaterial, however, because
the emphasismain effect was significant [F(4,184) :
6 2 1 . 6 8p
, <.00011.
Performanceof young and old adultsdivergedsignificantlyon the secondseriesreported[F(1,46):6.90, p <
.051.The materialmain effect was also significantin the
lI
o
o
o 6 0
o
o
G
c
o
9 4 0
o
o-
tsl
Recalled
t
t
t
I
L
0
I
I
1
L
L
2
i
3
,
4
EmphasisCondition
Figure l. Percentage correct across emphasis conditions for tasks
recalled first and tasks recalled second, collapsed across type of
material, Experiment l. A given point represents the average of96{)
trials (20 observations for each of 24 subjects for both letter and
digit material). The emphasis conditions are designated in terms of
the payoff received for correct performance with that set of material.
616
SALTHOUSE,ROGAN, AND PRILL
secondseries[F(1,46): 11.40,p < .005], reflectingthe
fact that letters were more difficult than digits when they
were reported second.The emphasismain effect continued
to be significantin the secondseries[F(4,184):182.68,
p < .00011,as was an age x emphasisinteraction resulting from the elderly group's lower performance on all
but the lowest emphasisconditions in the material reported
second[F(4,184):3.28, p < .05].
It is interesting to note that performance was above
chance(i.e., l0% for digits and 5% for letters) even in
the O-emphasisconditions. This appearedto be attributable to a tendencyamong many subjectsto remember the
first or the last item of the unattended(nonemphasized)
set, in addition to as many items as possible from the attended set.
Divided-attention cost was first computed for each individual in a dual-task space with coordinates ranging
from0% to IOO%on eachtask axis. The mean dividedattention cost regions, that is, the areas above the AOCs
(seeFigure 2), were .320 for young adultsand .388 for
older adults It(461:2.46, p < .051.
Relative divided-attention cost was also computed on
the basis of each individual's functional performance
region (Somberg& Salthouse,1982). In this analysis,the
dual-task spaceis restricted to the region defined by the
minimum and maximum performance levels actually obtained on each task, rather than by theoretical limits of
O% and 100%, thus taking into considerationeach individual's actual range of performance. Age-related decrements in performance comparableto those obtained with
the absolute divided-attention cost measurewere observed
with theserelative measureslt(461-2.69, p < .051,with
means of .270 for young and .355 for old.
o
o
o
@ o
o
u
4
u
3E
c
o
I
Q)
20
40
60
80
(Percenlage
Correct)
Letters
100
Figure 2. Empirically derivedattentionoperatingcharacteristics
(AOCs) for young and old adults, Experiment l Each point
representsperformance in one emphasiscondition (20 observations
for each of 24 subjects).For example,the point in the upper left
on both functions rrpresentsperformanceon the digit task (ordinate)
and the letter task (abscissa)when subjectswere instructed to place
lfi)7o emphasison the digits and 0Voemphasison the letters.
Discussion
Older adults were found to perform lesseffectively than
young adults across several different methods of measuring the divided-attentiondecrement.This finding may
have to be qualified somewhat,however, becauseit was
found that older adults performed nearly as well as young
adults on the first set recalled, but the differencesbetween
groups becameapparenton the secondset recalled in the
form of an age main effect and an age x emphasisinteraction. Severaldichotic-listeningstudieshave reporteda
similar result(althoughalwaysat a single,unknown,emphasiscondition), and two common interpretationshave
beenthat older adults are more susceptibleto either spontaneousdecayof second-recalleditems during the interval when first-set items are being recalled, or to response
interference effects produced b1' the recall of first-set items
(Craik, 1977). If either of tlresemechanismswere responsible for the presentdivided-anentionresults,it could be
argued that structural. rather than capacity, limitations
(Salthouse,1982) were responsiblefor the presentage
differences.Experimens 2 arrd3'*'ere thereforedesigned
to allow theseinterpretationsto be directly investigated
and to determine whether thel could account for the
presentage differencesin divided-anentionability.
EXPERIMENT
2
Experiment 2 was similar to Experiment l, but only a
singleresponsewas requiredto eachsetof material.One
position in each iuray wils cued at the time of the response,
and the task was simply to identify the cued item. This
manipulation greatly reduced the responserequirements
of the task, and, therefore. if response interference or
spontaneousdecay is the primary mechanism responsible for age differencesin divided attention,the differences
should have been minimized or eliminated.
Method
Subjects.Sixteencollegesodents(meanage:18.6 years,
range= I 8 to 22years)andl6 olderadults(meanage: 70.I years,
range:60to E4years)participated
in a singlesession
of approximatelyI .5 h. Therewere5 malesand1l femalesin theyounggroup
and4 malesand12females
in theoldgroup.Meanyearsof education were13.4for theyoungand17.l for theold [t(30):5.29,
p < .0011.
Meanscores
on theWAISdigitsymboltestwere67.I
for theyoungand216.6
for theold [t(30):4.89,p < .001].In both
mqsures,thecurrentsamples
weresimilarto thoseof ExperimentI
with commonlyreportedtrends.Noneof the subandconsistent
jectshadparticipated
in the previousexperiment.
Procedure.Thegeneralprocedure,
apparatus,
andmostof the
l.
specificdetailswerethesameasthosedescribed
for Experiment
Therewerethreemajordifferences.
Thefirst differencewasthat,
in Experiment
2, eachsetofmaterialin bothsingle-anddual-task
involveda singleresponse,
ofthe identityof
conditions
consisting
theitemcuedby a setof questionmarksin thepositionoccupied
by thetargetin thestimulusarray.Theremainingitemsin thearraywereindicated
with dashes,
andbelowthearraywasthequestion "Which letter?"or "Which digit?" to remindthesubjectof
the materialhe or shewasto supplyby pressingtheappropriate
keyon thekeyboard.Thelocationofthe probein thestimulusar-
I
I
I
DIVIDED ATTENTION
I
ray wasvariedrandomlvacrosssequence
positionswith the re_
strictionthatontheaveriseeachporitlon
,uJuiJ[J p.j* "quurfy
often.A response
*u. ,"oiir"d roi.n..y
f.i!;"";; fiJity a
on.a givenbtockof duaj_msk
oiurr.iiJol"o-.;;,J*:l andsuess.
terter
probeswasconstant,
butthiso.0".,u. uulun"laii.ir", i.,ur
uro"t,
tor eachsubiect.
Theseconiprocedural
difference
fromthepreviousexperiment
initialsingle_rask
spans
werederermined
Ltt-rlS"
witha cnterion
ot tourcorrecrresponses
ourof five ir.rcu; ;;;;i; i*l outor rru.
to minimize
thecontribution-of
cnun.e*itr,;;lr?r;;;;
response
Thethird modification
of rn. pi.ui*.-.*pt irn"n, *u,
f.,jltl],
anrncrease
from l0 to 20 trialsperemphasis
.onAiiiin p". uro"t,
ror a totalof 200 trials.
Results
The age differences were significant
o
@<)
o g
5
o
q
o
o
in
: 6.56,old= 5.o: itro.l: j.ij,'fthe lener_span
task.
[young
;
. .0s1,uu,
ngt.illhl9igit-span
task[y^oung
= i .{s, ,tdJ .al; t(30)
j_,t.ol. Thepercentag.,
or.o.i..t ..;p";r;, inthedual_
task conditionswere subjected,o
un ug. x mate.lat x
order x emphasisanalyiis of variancE.
the'foilowing
main effects were signihcun,,
ug"_young'riU;""t, t uO
jaer
higher
(Percentagerscorrect)
Figure 4. Empiricallv_derived AOCs
for young and old adults,
Experiment 2. Each ooint represents
pe.forman"E in one emphasis
condition(40
observation.
ri. .""r,-Jiiffiil;;l
scores
rhan
subjeciifFiiJAj=;.21, p (
.0051;order-thetrrst,..ie, ,eported
r,ujiigr,". ,"or", similar age differences.on both orders
and the absence
lF(I,30): 26.20,p < .000rI ; ";J ;ilili;:rcor",
in_ of a significant interaction U"t*".n ug"lni order, sug_
creased
withanentionat
empirasis
fF(4,it6;=;60.90,p geststhat responseinterference effects i" l"r, pronounced
< .00011.
The only significantrnt;r;";i;;u,
than in Experiment -.
The AOCs derive_dfrom these data are illustrated
in
effects were more p.onoun""d-at higher
Figure 4. As implied by Figure 4,
;il;;"I,,
attentional emphases.
had a sig_
nificantty
The significantrrendsare illustrated
-highei absolute"divtd;:;;;
cosr than
.
in Figure 3. No_ young adults[.339 vs. .240; t(301:3.
16,
:
.005].
The
;
o.0.."n".t. ftiro.run." age differenceswere in the expecteddireciion
ll.,l1"l: g:rqitethesignificunt
but
did
not
marerial*u, rnu"t'.jorrr to that achievestatisticalsignificance
:1.I.j""-d-reported
*ith th";;;sure or.etu_
ur
urcrlrst_reponed
materialthanwasthecasein Experi_ tive divided-aftention
cost based on each individual,s
mentI (cf. Figurel). Thistrend,togetheiwitt
.ougtty functional performanceregion
[.3g2 "r. .:jO; (30) =
1.65,.15>p>.101.
emphasis
rr(qnq:i.t
1.9:I
i"9
tng that
the order
,oo
I
H' |
I
""
"" ":=t4"
,.;/4.;il_,*i
o"\1, od
' .
p*f
,/'
l
,.f i'..
r
!---
indicat__
r.,n"J[o I
I
8or
C I
? uolE
b",*""n
, ;'.'.i;i,
'./,,,'
/'/
I
l
I
I
I
r
'
Il
I
l
Discussion
The major findins of Experiment
2 was that the age
differences in dividid-aaention cost
were still evident
when the memory tasks are modified
to _inirr#" response
interference. However, the age
differences in the rela_
tive divided-attentioncost measure
were not "a
significant,
and the datain Fieure 3.indicatetfr",f,"i.
*",
tenOency,
atbeit not sratisd;atly signincanr,-foi'irJ
uJ."iin"r"n.",
to.be.morepronounced-o"rh; ;;;j;;"":iJ
It is thereforestill possible,o ".gu" ,hui-ril-"-ofmateriar.
the age
differencesin divided attention'fouiJi"
t_p"rfment I
wer^e_mediated
by greatersusceptibilityto ,.'spons.
in_
terlerence_or
spontaneous
decaywith in..easeaage.Ex_
periment3 wasconsequently
designeJio-p.oiia" uaai_
tional eviclencerelevant," ,f,t i"i".pr*,i"".
..*,*,*]I"Jlj*x"ir.r,asks
recattedfirsr and tasks recailed
;r;;,;;;;;.*s
ryp€ of
material, Experiment2. A civen po,n,
."p.ount".ti"lJera!e or r,Zm
trials (40 observationsfoieach'of
lo *ui*i.
i". tii*,etrer and
digif material). The emphasisconditions
are designared
in termsof
the payoff receivedfor cbrrcct p",f"r";;;;i'if;",'#ji,n","ri"r.
EXPERIMENT 3
In an aftemptto completelyeliminate
response
^
inter_
fglengeanddecayeffecti ro, p..fo.manl'in[J.
aiuia"a_
attentionconditions,thememory_span
taskswerefurther
modifiedin two respects.One."Ain""tio"
"Jnrirt"O of
618
't"t.'1,l
SALTHOUSE.ROGAN.AND PRILL
re-presentingall items in the array exceptthe cued item
at the time of response.It was believed that this would
reducethe necessityof cycling through one's memory to
locate the probe item at the time of recall, thereby
minimizing the possibilityof interferenceand shortening
the time to generatea response.The secondmodification
was to requesta responsefrom only one set of material
in the dual-taskconditions.That is, althoughboth letter
and digit arrays were always presented,on a given trial
the subjects were queried about only one (randomly
selected)array. Becauseonly a single item was to be
reported, there was no possibility of the recall of earlier
items interfering with the recall of subsequentitems, or
of information to be reportedseconddecayingduring the
reporting of information from the first series.
_=::qr_
8{J I
I
o
o
\il
6()
a
O
o
o__
{
I
lc
L- - l
'tu'"uT3fl5,.u"o""'t'
Method
Figure 5. Empiricalll derived AOCs for young and old adults,
Subjects.Sixteencollegestudents(meanage:19.1 years,
performance in one emphasis
(mean
range:18to22years)
age:66.6years, Experiment(303. Each point represents
and16olderadults
obsenations for each of 16 subjects).
range:62to 77 years)participated
ofapproxi- condition
in a singlesession
mately1.5h. Therewere6 malesand l0 females
in theyoung
group,and4 malesandl2 females
in theoldergroup.Meanyears Discussion
of education
were 13.6for the youngand l5.l for the old
The major finding of Erperiment 3 was that the age
were66.3
symbol
scores
It(30):1.63,.15> p > . l0l. Meandigit
in divided-attentioncost were replicatedin a
differences
p < .0001].These
for theyoungand42.9for theold[t1:O;=6.33,
results,
similarto thoseof Experiments
I and2, againsuggest
that task with virtually no opportunity for responseinterferpopula- encebecauseonly a singleresponsewas requiredon each
oftheir respective
the currentsampleswererepresentative
in eitherof the preced- trial. Moreover, in the presentexperiment,the agediffertions. Noneof the subjectshadparticipated
ing experiments.
enceswere significant in both the absoluteand relative
Procedure.Most of the proceduraldetailsweresimilarto those
measures
of divided-attention
cost.despitelower statistiThe major modificationswere:
of the precedingtwo experiments.
power
l. due to a smaller
cal
than
that
in
Experiment
addingthe identitiesof the noncueditemswhenpromptingfor the
numberofsubjectsper agegroup. It canthereforebe conrecall response;requestinga responsefrom only one of a trial's
two arraysin the dual-taskconditions;and increasingthe number cluded that the age differencesin divided-attentionabiloftrials per block to 150,for a total of 300 acrossthe two blocks,
ity with two concurrentmemon tasksare not attributafor the lossofdata from the second-reported ble simply to greater output interferenceon the part of
to partially comp€nsate
array. Determinationof which array to probeon a specificdualolder adults.
tasktrial was random,with the restrictionthat, on the average,the
letter and digit arrayswould receivean equal numberof probes
with eachattentional
emphasis.
GENERAL DISCUSSION
A primary focusof the presentexperimentswas the use
Results
The age differenceswere not statisticallysignificant of the AOC analysis,as first usedby Sombergand Salthouse(1982)with respectto divided attentionand aging,
with eitherthe digit-spantask [young : 8.16, old : 8.00;
tasks.
in relatively demandingconcurrent-memory-span
(30) < l.0l of the letter-spantask [young : 6.81, old
: 6.31; t(30) : 1.23, p > .501.The only significant Both the relative (Experiments I and 3) and absolute(Experimentsl-3) measuresof divided attentioncost indicated
effect in the analysis of variance on the percentagecorrect responsesin the dual-taskconditionswas emphasis that older adultswere more penalizedthan young adults
lF(4,120) :268.20, p < .00011.The ageeffectin per- by the divided-attention requirement, even after the
difficulty of the concurrenttaskswas adjustedto the same
centagecorrect in the dual-taskconditionswas in the exp e c t e dd i r e c t i o n( y o u n g : ' 7 0 . 1 % , o l d : 6 ' 7 . ? % ) , b u t proportional level for each individual subject.
Obtaining roughly equivalentagedifferencesin dividedfailed to reachan acceptablelevel ofstatistical significance
:
attention
costs across the three experimentsnot only
.
2
5
1
.
>
l
.
2
O
,
P
lF(1,30)
the reliability of the basicphenomenon,but
Despitethe similar overall level of performancein the demonstrates
also
suggests
that the locus ofthe age differenceis in the
revealed
an
age
deficit
two age groups, the AOCs still
initial stageof registrationor encodingof the informain divided-attention costs. The data are illustrated in
Figure 5, in which it can be seenthat older adults had tion. This inferenceis basedon the nearly identicalage
trends when the potential for storagedecay or response
higher divided-attentioncoststhan young adults[.218 vs.
were
interferencewas systematicallyreducedfrom Experiments
The
differences
age
.142; t(30):2.20, p < .051.
I to 2 to 3. Furthermore, the useof measuresderived from
measure
based
on
relative
cost
with
the
significant
also
individually determinedfunctional performanceregions AOCs, which representdual-taskperformanceacrossa
rangeof emphaseson the two tasks,indicatesthat the age
[ . 3 3 ] v s . . 1 9 9 ;t ( 3 0 ; : 2 . 6 1 , P < . 0 5 ] .
1
DIVIDED ATTENTION
I
differencesare not anributableto differential biases
or
strategiesfavoring one task over the other.
It thereforeseemsreasonableto concludethat the
age
differencesin the dual-taskconditionsare causedby
agi_
related limitations in successfully,
encodingitems when
two simultaneoussetsof material u.. p."rJnt"d. A
rela_
tively uninterestinginterpretationof this finding might
be
that it is causedby slower shiftsof fixation from one
task
(e.g., digits)to the other (e.g., lefters)in older
adultsthan
yo,ungadults. Although we cannot unequivocally
re_
in
ject this possibility. it is highly unlikely
that it could ac_
count for more than a small proportionof the age
differences.because
eye movementsof 5. (the spatiallsepara_
tion berween the arrays) typically'r"quir" less
than
a5
in young adults(Saitirour"& filir, 1980),and
31e.c
probably not much more in older adults. Atthese
rates,
a large number of redistributions of fixation could occur
within a very small fraction of the 3_secexposure
rime.
Our interpretationof the age differencesin the dual_
task conditionsis that they are causedby an age_related
reductionin a dynamic rather than a structura'lform
of
attentionalcapacity.This type of capacitymay simply be
equivalentto the rate of performing mlntal'operations
(Salthouse,1982), or it may be anal6gousto what
Craik
and Byrd (1982)termed..mentalene.gy... In eithercase,
however, the fact that fewer total itemi were reported in
the dual-taskconditionsthan the averageof the spansin
the single tasks (see Table l, as well is similar results
by Inglis & Ankus, 1965,and Inglis & Caird, 1963,with
dichotic listening tasks) suggestithat perfoimancewas
not limited by purely structuralfactors(e.g., number
of
slots). Instead,performanceappearsto bJrestricted by
more active processes,such as the initial allocation, or
subsequent
redistribution,coordination,and monitoring,
of capacity{emanding encodingoperationsacrossthe two
concurrenttask. Either the amountof the resourcesavailable for theseactivities or the efficiency with which they
are allocated to the various processing componentsap_
pears to decreasewith increasingage.
Despitelessefficient divided-atteniionperformancein
older adults than in young adults, the two age groups
.-c1rr'd I() allocateattentionacrossconditionsln
a simi_
hm of {reragettilt:t-t"lno Duat-Task
performance
loF
r-.:'.;:
Si:ic
l..r
i r:r:::e
Young
old
Average Number
of Items Recalled in
Dual-Task Conditions
Sl:: In
( ,:\:.!:,\n\
:l
l
n Ji
o:..
5 . 8I
5.25
Err-:-<:r l
Young
old
.
.rg
/-l-r
6.51
Expennr:r :
7.49
'7.16
6 -10
{
1 r
l r
6.{?
619
lar fashion,as indicatedby the comparabletrend of em_
phasisvariationsin all figures. This finding is important
in that it suggeststhat the ability to distribut-eone,i atten_
tion across two concurrent activities is relatively un_
affected by increasedage. There may be less attentional
capacityavailablefor distribution, or more overheadmay
be required to monitor the distribution of attention, but
the effectivenessof actualattentionallocationamong con_
current activities does not seem to be reduced between
2O and70 yearsofage. In this respect,then, characteriz_
ing the difficulty simply as poo.e. division of attention
may be misleading becauseyoung and old adults appear
to be equally proficient in the aciual partitioning of
the
availableattentionacrosstasks in responseto tlie vary_
ing emphasisconditions.
Finding older adults to be more disadvantagedthan
young adults when required to divide their attenti,onis
in_
consistentwith Somberg_and
Salthouse's(19g2) finding
of no divided-attentiondifferencesacrossagegroupscom_
parable.tothose employed here. The appaien-tcontradic_
tion in the patternof resultsmay be attiibutableto differ_
encesin the complexity of the tasks employed in the two
studies.The Sombergand Salthouseexperimentusedtwo
perceptualdiscriminationtasksthat seemto have involved
minimal.processingof information, when processingof
information is definedas the hypothesizednurnberof men_
tal.operationsperformed. The discrimination task required
subjectsmerely to detect and respond to the presenceof
a target. When two discrimination tasks were performed
concurrently, the number of mental operationsincreased,
but the greater demandswere still apparently within the
capabilityof both age groups. The iurrent experiments
usedmemory-spantasksin which the individuil was re_
qrriredto identi$, remember,and then reproduceeither
all, or a specifiedmember, of a seriesof letiers and digits.
Perhapsbecauseofthis addedcomplexity, agediffereices
were evident in the costs of dividing ittention between
two concurrentactivities. In other words, the explana_
tion that may account for the apparent discrepaniy be_
tweenthe current findings and thoie of SombergandSalt_
"of
house is simply that the larger the number
mental
op-erationsto be performed, the larger is the absoluteage
differencebetweenyoung and older adults. Wright (l9g'i)
cameto a similar conclusionwhen older adultsperformed
worse than young adults on a complex singljtask, and
interpretationshave been presentJdpreviously
3nal_oqgus
by Salthouse(1982, in press).
To summarize, older adults are penalized more than
young adults by the requirement oldividing their atten_
tion betweentwo concurrent tasks even when-thedifficulty
of the dual-task situation is the same frxed percentageoi
single-task performance for each individual. Howdver,
because_an
earlier experiment with a simpler set of tasks
revealed no age differences in divided_aitentionabilitv.
and becausethe present AOC analysesrevealed similai
capabilitiesof dividing the availableattentionalresources,
we suspectthat the age-associated
problem is not due sim_
ply to allocationof attention to alternative..channels,'
620
SALTHOUSE,ROGAN, AND PRILL
but, rather, is a problem in dealing with increasedcomplexity of the total situation. Age differences may be
present whenever composite task difficulty or demands
upon processingcapacity are great, and although dividedattention tasksoften involve high levels of difficulty, they
do not necessarilydo so, and there are many single-task
situations in which the level of task difficulty is high. Future research systematically analyzing the effects of additional mental operations (task difficulty) on the singletask and divided-attention performance of adults of varying ages would be desirable.
REFERENCES
Therole
andaging:
Bunxn,D. M., a LrcHt,L. L. (1981).
Memory
of retrieval processes. Psychological Bulletin, 90, 5 l3-546.
CerRD,W. K. (1966). Aging and short-term memory. Joumal of Gerontolosy, 21, 295-299.
CnrIx, F. I. M. (1977). Age differencesin human memory. In J. E.
Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging. New York: Van Nostrand Reinhold.
Cnerx, F. I. M., e Byno, M. (1982).Aging and cognitivedeficits:The
rate of attentionalresources.In F. I. M. Craik & S. Trehub (Eds.),
Aging and cognitive processes. New York: Plenum.
Iuclrs, J., a ANxus, M. N. (1965). Effects of age on short-term storage
and serial rote learning. British J ourrnl of Psychology, 56, I 83- 195.
and serial rote learning. British Journal of Ps,v-chology,
56, I 83-195.
I N c L I S ,J . , a C e t n o , W . K . ( 1 9 6 3 ) . A g e d i f f e r e n c e si n s u c c e s s i v e
responsesto simultaneousstimulation. Canadian Journal of Psychology,17,98-105.
KrNcsr-e, R. A. (1980).The measurementof attention.ln R. S. Nickerson (Ed.), Attention and performance 12111.
Hillsdale, NJ: Erlbaum.
P a n x r n s o N ,S . R . , L r r . r o H o r - r Jr r., M . , a U n r u , T . ( 1 9 8 0 ) .A g i n g ,
dichotic memory, and digit span.Journal of Gerontology,35, 87-95.
SerrHousB, T. A. (1982). Adult cognition: An experimentalpsychology of human aging. New York: Springer-Verlag.
SrLtHousr, T. A. (in press). Speedof behavior and its implications
for cognition. In J. E. Birren & K. W. Schaie(Eds.), Handbookof
the psychology of aging (2nd ed.). New York: Van Nostrand Reinhold.
S,rlrHouse, T. A., a Elr-rs, C. L. (1980). Determinantsof eye fixation duration. Ameican J ournal of Psychology, 93, 2O'7-234.
SowsBnc, B. L., a Selrsouse, T. A. (1982). Divided attentionabilities in young and old adults. Joumal of Experimental Psychology:
Human Perception ancl Performance, E, 651-663.
Snnnr-rNc,G. ( I 978). The attentionoperatingcharacteristic
: Examples
from visual search.Science,202, 315-318.
SesnLtNc, G., a MsI-cFrNBn,M. J. (1978). Visual search,visual attention, and the attention operating characteristic. In J. Requin (Ed.),
Attention and performance 211. Hillsdale, NJ: Erlbaum.
WrucHr, R. E. (1981). Aging, divided attention,and processingcapacity. Journal of Gerontology,36, 605-614.
I
t
I
(Manuscript receivedJanuary 3, 1984;
revision accepted for publication August 2, 1984.)
I
t