Joumal of ExperimenralPsychology:
Human PereDtion and Perfomane
1982,Voi. 8, ir,ro.5. 651-663
Copyrieht l9E2 by thc Amaim
Psychologi€l Aswiation, Inc.
0096-I 523lE2l0805465I $00.75
Divided Attention Abilities in Young and Old Adults
Benjamin L. Sombergand Timothy A. Salthouse
University of Missouri-Columbia
I
I
I
I
The literature on divided attention and adult aging has not taken agedifferences
in single-task performance into account, and it has not been able to measure
divided attention independently of resource allocation strategies.Two experimentsare reportedthat,controlledfor thesefactors.In the first experiment,young
( I 8-23 years)and old (57-76 years)subjectsmade respon3es
to two simultaneous
visual displays. Stimulus durations were manipulated to equate single-task performances,and acrossdifferent conditions subjectswere induced to vary the way
in which they allocatedresourcesbetweenthe two displays.In the secondexperiment, respons€time was the dependent variable; dual-task scores were assessedrelative to each subject's single-task scores.No significant age difference
in divided attention ability independent of single-taskperformance level was
found in either experiment.The existing literature must be reexaminedin light
of theseissues.
Whether older adults are less capable of
dividing their attention among two or more
simultaneoustasksthan their youngercounterparts seemsto be a question that would
inspire little controversy.Craik (1977) summarized the prevalentposition on this issue:
One of the clearestresultsin the experimentalpsychology ofaging is the finding that older subjectsare more
penalizedwhen they must divide their attention, either
betweentwo input sources,input and holding, or holding and responding.(p. 39I )
Although the majority of the researchers
who have investigated this problem have
concludedthat there is a significantageeffect
on performance of divided attention tasks,
many of their findings may have alternative
interpretations. A closer inspection of the
available literature suggeststhat it is beset
with problemsnumerousenoughto prohibit
definitive conclusions.
The first difficulty in this researchlies in
the restrictednumber of experimentalparadigms that have been used to investigatedivided attention effects. Dichotic listenins
This researchwas supported by National Institute on
Aging Grant ROl AGO0694 awarded to Timothy A.
Salthouse.
We would like to thank Harris Cooper, Donald Kausler. John Mueller. and Marlin Thomas for comments
on an earlier draft of the article.
Requestsfor reprints should be sent to Benjamin L.
Somberg, who is now at GTE Laboratories, 40 Sylvan
Road. Waltham. Massachusetts02254.
tasks comprisb a rather substantialsegment
ofthe researchintended to discoverwhether
an ageeffectexistson divided attention tasks.
In the dichotic listening procedure,subjects
are presented with two simultaneous auditory messages-oneto eachear. The stimuli
typically consist of strings of digits or monosyllabicwords. The task of the subjectis to
reproduce the two messages.Most researchersreport that there is an overall agerelated impairment on this task. However,
despitethe frequency with which the paradigm has been studied, there does not exist
a consensusabout what processesare responsiblefor this effect.
The initial researchin this area, from the
laboratoriesof Inglis and his colleagues(Inglis & Ankus, 1965; Inglis & Caird, 1963;
Mackay & Inglis, 1963), suggestedthat the
locus of the ageeffect was a short-terrn memory deficiencyof older adults. Using Broadbent's (1958) model of information processing, they argued that if it was necessaryto
hold the messagefrom one ear in a storage
systemwhile the messagefrom the other ear
was being output, a memory deficit would
showup in impaired performanceon the second earreported.Inglis'swork confirmedthis
hypothesisby showing that the performance
of older adults wasworsethan that of young
adults on the secondchannel reported, but
essentiallyequivalenton the first channelreported.
652
BENJAMIN L. SOMBERG AND TIMOTHY A. SALTHOUSE
Craik (1965) reasonedthat if no instructions were given to the subjects regarding
which channel to reproducefirst, as was the
casein the studiesby Inglisand Caird (1963)
and Mackayand Inglis( 1963),it is likely that
subjectswould commencewith the channel
that was easier for them to recall. The superiority of a given channelmay be the cause
rather than the effect of its prior reproduction. Craik found that when the recall order
wasspecifiedto the subjectsthere wasan age
declineon both channels,althoughgenerally
there was a larger decrementon the second
channel.
Clark and Knowles (1973) arguedthat the
age deficit in dichotic listening is due to a
perceptualimpairment. They suggestedthat
the critical comparison is not perforrnance
on the first channel versusperformanceon
the second,but how well subjectscan reproduce the messagein the left ear in relation
to their ability to reproducethe messagein
the right ear. The right ear is typically the
dominant one, and stimuli presentedto it
tend to be more accuratelyperceived,Any
perceptualimpairments would be magnified
in the nondominant ear regardlessof whether
its messagewas reported first or second.
Clark and Knowles systematically varied
whether the right or left ear was to be the
first channel reported. They found that performancewas generallylower on the second
channel, but this was equally true of both
young and old adults. Performancewas also
lower in the left ear than in the right ear, and
this differencedid increaseas a function of
the age of the subjects.These data are supportive ofthe hypothesisthat perceptualprocessesmay be responsiblefor the age effect
in dichotic listening tasks. Other studies,
however(e.g.,Craik, 1965;Inglis & Ankus,
1965) also manipulated which ear should
be reported first and did find significant
Age x Order interactions.
It is unfortunate that in the examination
of adult age differencesin divided attention
ability there has been such an overemphasis
on a single methodology-particularly one
in which there is so much ambiguity about
the nature ofthe ageeffect.For a convincing
argument to be made that older adults are
lesscapableof dividing their attention than
younger adults, a wider range of paradigms
must be employed.
A secondissuethat makesthe existingliterature on aging and divided attention less
than conclusiveis that on a large variety of
laboratory tasks, younger adults perform
more capablythan older adults. This is particulary true in tasks with a critical speed
element, where sizeableage differencesare
almost invariably reported. Consequently,
there may be a problem interpreting many
divided attention results.In many dual-task
studieswhere subjectsare given two or more
tasks to perform simultaneously,we would
expect to find significant age effects under
single-taskcontrol conditions. Given such
initial discrepancies,how much of an age
differencein performanceshould we expect
under dual-taskconditions?
This problem has severalpotential solutions, but at the presenttime no study has
adequatelyresolvedit. An experiment by
Lipps Birch (1978)offeredone approachto
this issue. Her study examined age differencesamongchildren,but the theoreticalissuesare similar. Lipps Birch found that the
older children were originally more competent on each of the measuresunder singletaskconditions.Shethen gavepracticeto the
youngerchildren until their performancewas
comparableto that of the older children.
When the subjectswere put back into dualtask conditions, no age differenceswere
found. This study seeminglyarguesagainst
any age-relateddivided attention effects in
children when differencesin single-taskperformanceare controlled for. It is likely, however, that the initial practice given to the
younger subjects made the experimental
tasksmore automatic(i.e.,lessattention-demanding) for them than they were for the
older subjects.As increasedautomaticity facilitatestime-sharing,the practicetrials may
have given too much of an advantageto the
younger children.
A secondsolution to this problem hasbeen
to match subjects on single-taskperformance. Although older adults may tend to
be inferior to younger adults on some particular measure,it is frequently possibleto
find a selectedsampleof older subjeqtswho
can perform at the level of a sampleof younger subjects.This was the approachtaken in
a dichotic listening study by Parkinson, Lindholm, and Urell (1980).They reasonedthat
if the age effect in dichotic listening perfor-
manc€ is a shqt-la
non,-.,! slrould be g
agedifference rna
\
oard dtgrtsn tcq
T
tnat when ilre1 aco
enectand the Ate x
ond ear) interrrion ,
ever, the problem wid
rng proceduresis
rha
;ects who are atypi€l
lupe.norotaer suLjet
Jects)on thor variibtc
..o-ul.r... That therc
is
.oo betweenthe
Dgir
WechslerAdulr Iniet
and rhe full_rale
W
197111gn6,suppon
lo
It ma!. also be poes
rssueofage differences
rh_ro_ugh
rhe r1.p
Tn:"
emptol.ed.
Dffercncesi
oecontrolledfor b1.con
rore underexperimen
scoreundercontrol
con
rnan.testingfor an
abno
<rt\.toed
altention effect
an-vage-relabdr
.amin1
lect relativeto each
agr1
under single_hskcondid
penormed this tlpe
of e
rnrsexperimentperfornx
:!lh
fpararety and simu
tou.nd.the
same panern
r+rtnabsoluteand
relauv
easrer.ofthe two tasks
tt
rralll larger loss of
speed
rentrondemandsfior
the r
for rhe y.oungb
:l:i"_**
urrrerence
wasfioundin th
:T.rem:nr on rhe morc |
rhissrudyi
::l,L?rch
.in.
\r'hether
relative
or ab,solu
:.1:* "1. no orherinsrar
har.ehe
:!pT. 9f anaty,ses
,TIibtq rhardifferenrpan
tdent n'ith the two
measu
requrred before it
can be
sotute or relativeanalrses
pnate.
,- l:.h3* the ideat sotutio
find siruations
1_merelv-.t9
ditr-erengg
on sins
+:,:c.:
p
difficutrro :
.quire
.iy-r
when it is
done the hsks us
DIVIDED ATTENTION ABILITIES
mance is a short-terrn memory phenomenon, it should be possibleto eliminate the
agedifferenceby matchingsubjectson a standard digit-spantest.Theseinvestigatorsfound
that when they accomplishedthis, the age
effectand the Age X Order (first ear vs. second ear) interaction were eliminated. However,the problem with thesetypes of matching proceduresis that they may selectsubjects who are atypical of their agegroup (i.e.,
superiorolder subjectsor inferior young subjects)on thosevariableswith which digit span
covaries.That there is a correlation ofabout
.66 betweenthe Digit Span subscaleof the
Wechsler Adult Intelligence Scale (WAIS)
and the full-scale WAIS IQ (Matarazzo,
1972)lendssupportto this line of reasoning.
It rnay also be possibleto deal with the
issueof agedifferencesin single-taskperformancethrough the type of statisticalanalyses
employed.Differencesin baselinemay often
be controlled for by comparingeachperson's
scoreunder experimentalconditions to their
scoreunder control conditions. Thus, rather
than testing for an absoluteagedifferencein
divided attention effects,this procedure examines any age-relateddivided attention effect relativeto eachagegroup'sperformance
under single-taskconditions.Talland (1962)
performed this type of analysis.Subjectsin
this experimentperformedtwo manual tasks
both separatelyand simultaneously.Talland
found the same pattern of age decrements
with absoluteand relative measures.On the
easierof the two tasks there was a substantially larger loss of speeddue to divided attention demandsfor the older subjectsthan
there was for the young, but virtually no age
differencewas found in the divided attention
decrement on the more complicated tasks.
Although in this study it did not matter
whetherrelativeor absolutescoreswereused,
there are no other instancesin which both
typesof analyseshave been done, and it is
possiblethat different patterns could be evident with the two measures.More data are
required before it can be said whether absolute or relative analysesare most appropriate.
Perhapsthe ideal solution to this problem
is merelyto find situationsin which there is
no age difference on single-taskmeasures.
This is quite difficult to accomplish,and
when it is done the tasksusually turn out to
653
be ones with a high degreeof automaticity,
they thus show no time-sharingeffect in either age group. It may be possibleto find
tasks that have some dimension (e.g., stimulus duration) that may be manipulated to
artificially equate the performance of subjects of different ages.This strategy is employed in the presentarticle.
The third difficulty with the current literature on age effectsin divided attention is
closely tied to the baseline differencesjust
discussed.A number of studiesnot only fail
to addressagedifferencesin single-taskmeasuresbut do not even collect data under single-taskconditions. It was previously argued
that even when thesedata are available,assessingdivided attention performance may
be difficult; without suchdata the task is impossible.For example,a study by Broadbent
and Heron (1962) had young and old subjectsperforrna digit cancellationtaskthat on
some trials was combined with an auditory
monitoring task. Becausethis experiment
had trials in which the cancellationtask was
performed alone, it is possibleto determine
the effect on that task of having to perform
a concurrentmonitoring task. However,as
the monitoring task was never performed
aloneit is impossibleto determinewhat type
of divided attention cost there was on that
task. That the young subjectswere better on
the monitoring task under dual-taskconditions than the older subjectsmay havebeen
due to their ability to divide their attention
between the cancellationand monitoring
tasks.Alternatively,it may havebeendue to
their superiorityon monitoring tasksunder
any conditions.
Subjects in an experiment by Kirchner
( 1958)madekey presses
in response
to a pattern of rapidly moving lights. In different
conditions responseswere required to the
positionof the currentlight ("no-back2'condition), the positionof the immediatelypreceding light ("one-back" condition), or the
second("two-back") or third ("three-back")
precedinglight. Viewing this as a dual-task
situation, the subject simultaneouslyhad to
(a) respond to the appropriate light and (b)
remember the previous sequenceof lights.
The "no-back" condition provided a measure for the respondingtask without simultaneously having to remember light sequences,but there was no comparablecon-
t
654
BENJAMIN L. SOMBERG AND TIMOTHY A. SALTHOUSE
POC; differencesin divided atdition for the memory component. It is on the same
are reflected by the existence
ability
tention
'unoei :'one-back,"
unclear whether tt. inipuii.d'performance
similar arguments have
Poc^s.
separate
of
of the older subjects
..two-back,,,and ..tnrJe-tacti conditions is been developedfor the use of ReceiverOp
as a method of sepaa consequenceof poofiiuiO"a utt""tion abil- erating Chaiacteristics
bias in Signal
response
from
pt"p"r -".
light ratinisensitivity
ity or merely inabilitylo i"""ii '-:-Theory (Green & Swets' 1966)'
Deteition
sequences.
goal of^the presentleryarch .wasto
A final issuethat has not been adequately Another
analvsisto the investigation
dealt with in the uuuiiuUf" fiterature is that apply tfris type.of
n"is in divided attention
ug.^aiff"t
Ai"iO"a attention ;i;i"lt
it is necessary,o -r*oi"
performance'
inA.p""J""tfy of resourceallocation strateTo summarize, the following factors have
gies. The finding tfrui of,O"i adults perform
of the available
conclusiveness
atalevelbelowthatofyoungeradultsonone reducedthe
attention: (a)
divided
and
aging
on
ofseveral concurrent tasksmay be evidenci research
to aiuiA. overemphasis-ona singJeexperimental..parthat the older adults "il'L.tlUi.
An "Ait*i ?U) fuitutt lo aicount for age differtasks.
their attention u-ong-ifr" uurio,r.
performance;(c).failure
alternativeinterpretationisthatperformance .n"i. io iittgle-tast
measureson all tasks;
was lower on one tu.t G"-urr." tire older sub- to report single-task
divided attention into-measure
fuit,tt
jects allocat.O u ,-uitJi ptopo.tio" of tn"i. iO)
resourceallocation strategies'
limited attentionatresouriesio that tast ttrun dependentof.
is intended to specificallyaddress
did the young ,rrUjects.io decide which of This study
issues'Ifage differencesin ditheseexplanationsis Correct,it is necessary each-ofthese
ability are found when the
attention
per"vided
to examinethe way in which a subject's
have been considered'the posformance on one tu.t uutiit as a function of abovefactors
(1977)willhaverethe subject's performance on concurrent itionexpressedbycraik Failure to find an
support'
zubstantial
ceived
+^arzo
suggestthat cr3rk's ul:"lto-"
Norman and Bobrow (1975) have intro- age effe9twill
strong and that further tnperhaps
too
w]as
duced Performance 5p"i"ii"g'C["iu"t.ti.needed.
n"ticte-rorltiitvpe of anal- vestigationon this issueis
iliFoat;;-a
work is to
current
the
goal
of
A-secondary
ysis.In a dual-tasksiiuation, perlormanceon
tl?t may
procedures
methodological
perestablish
Task X may be pfott"a u. u iunction of
of
investigation
thenot only in
formance on Task Y. As more resourcesare be usefut
of
range
a
for
also
but
uuaii"ure ror ua"ii "g. differenc-es
allocated to one tark; [;;;;
comparisons'
the other. Thus there i;;d; i; be a negative between-subject
correlation between performance levels on
ExPeriment I
the two tasks'
'
Sperling(1978;Sperling&Melchner,1978)
rr rnwas.to
Experiment rI r'ac
purpose ofc c--^-i*or+
The
empiricat
OeriueO
(1980)
il;;;
Kinchla
and
ta"- vestigateadllt age differencesin divided atpocs by inducing,G;il6r"ryttt.
performancein a situation that contive emphasisgiven to i""t of two simulta- tenti6n
problems discussedabove' The
neoustasks.fhese furiciont piouiO" u rrr"ftt trolled ihe
the simultaneous tachistoinvolved
perfoi- tasks
way of evaluating &i[J;A'*tion
of two independent vipresentation
t-pi"
iemance, because as the subject ct arrge,
to
is posslUfeio suai dijptays. The subjects were required
source allocation r,r*gt.fi
a
specified
of
absence
presence
or
o' on" task rel- indicatethe
measurelossin p.rfo-in""
- rfor-targetoneachdisplay..:
ativetothecorrespondinggaininpe
fithough it hal been found that there is
manceontheother.Whereastheco-pu.islt
age difference in the ability to
of single Aata poirrts aoes noi uUo* ifiuiO"O a significint
iaititirv tu"tisioscopicallypresented stimuli
attention effects to be separated from rll
g4\t"n, Hamiin, & Breitmeyer, 1970;
sourceallocation strategies,if the entire POb iit-,
Till' it Williams, 1978; Walsh' WilWiftn,
are
facton
functions are analyzed,the two
that
allocation liams, & Hertzog' 1979),it is also true
separable.Cfranges-in'r.*'r""
is
directly
task
of
type
ott-thit
strategiesmay be "i.*.O * aitr"r"nt points performance
EDNJ.
related to expclrn
lhe perfonnane cr
be equaredby alb
tongerstimulus prr
ger subjecs. Spocif
were determined ir
Ject such that ther.
curate under singir
gp9nt aCedifferenc
gfllons therefore co
ollleren@sin inidal
. Another featureo
It lnvolved the cons
ror_eachsubject.Sr
under different con<
pnasrsgiven to each
I
attentionability cou
oependentofallocat
Method
subj:gtt: Sixreen yora
range.I g_23)and t 6
olrer,
) /_ /o) eachpaniciparcd
ro
y9y.nC
were lncq
_sgbjecb
ar
_rne^University9f f,{isro
pated
lor extra coursccrrl
,.".$5. The etdcrty.qr
:,-,lp
jlu]9l"
uniqv.irq
:oT'n
) panrcipation,
ano
and rbq.
Noneof rhesuticr
1111se.
orhercarefrititi
::T*
9.
t temales.
in each age gruq
.r.
the^! were in gooO treali,.
Some pcychometric .tarr
Jeq al the conclusion ofrhcr
o"
Tgr5 on rhe Dgir Sy
roioar
l.T,9l;2a1^{er.o
(30) =
6.6e.p <
Tl':9tr-,
LheVocabutarysubtd
olr
:r
a
lad -me:tnrore of 2 | .6. r
lhr older adults was
Jl.j
fiJ{jl=3.53.p<.01.
lp@ralus. The subjcrrs
r
,
rumrnaled
only by a sma.ll.d
prescntedon a Tekronir
6O6
in :-n1 ^ofthem. They rtspo
,_tk.roushbunon
rctq
i.ll?
to
cach other on a board.
A pf
*as used to gener:rE
lrtfr
r
lnatlzc s_u
bjects. r€spons6!
rr.yedure.
Each fial coq
_
prescntation
of two risual an
h"d ro indicare thc prcr
:.o111r
clnad
tarBet. The computcr
radl
a larget was to appca,
:':-not
o
ocrcrmrled
independcnrir fc
ence and absence were
equalh
trne arra1.conaincd four
r.i.
corncn oflan imaginar]
il.7" )
DIVIDED ATTENTION ABILITIES
655
screen.The targetwasa small (.2') line extendingfrom
the vertexofone ofthe four Xs. The targetcould appear
on any one of the four Xs and could point in any of
four directions(up, down, left, right). On a trial in which
the target wasto be presentit was randomly determined,
with equal probabilities,in which of the 16 possiblelocations it would appear.The other array contained four
(.3' X .3") plus signs (+) located at the corners of an
imaginary( l. I " X 1.4o)rectangleon the screen.The target in this array was also a small (.1") line extending
from the vertex of one of the four pluses.The target
could appearon any of the four plusesand could point
in any of four directions (northeast,southeast,southwest,northwest).It wasequallylikely that a targetwould
appearin eachof the I 6 possiblelocationson this array.
The two displayswereconcentricand were centeredon
the screen.As the pluseswerelocatedmore foveallythan
the Xs, they weremade smallerin an attempt to equate
the difficulty of locatingthe targetson the two displays.
Pilot data indicated thar this was successfullyaccomplished.
The subjectsusedthe left-hand keyboardto indicate
their responsesto the plus array. On each trial they
pressedthe 9 key on the keyboardif they thought that
Method
they sawa targetand the 7 key ifthey did not. The righthand keyboardwas used to indicate their responsesto
subjects. sixteen young adults (mean age, 19.8; the X array.The 9 key indicatedthe presenceofa target,
range,I 8-23) and I 6 older adults(meanage,65.I ; range, and the 7 key indicated that the target was absent.A
57-76)eachparticipatedin a single90-min. session.The forced-choiceprocedurewas used in this task to miniyoung subjectswere Introductory Psychologystudents
mize any respons€bias efects.
at the University of Missouri-Columbia who particiThe subjects were first given general instructions
patedfor extra coumecredit but were also paid bonuses
about the natureofthe task,and the displaysweredemo[ up to $5. The elderly subjectswere membersof the onstratedto them. This was followed by a block of 50
Columbiacommunity. They werepaid $7 for their travel practicetrials in which the arrays were simultaneously
and participation,and they also could earn up to $5 in
presenledfor one second,and the subjectshad to inbonuses.None of the subjectswere residentsof nursing dicatethe presenceor absenceofa targeton eachdisplay.
homesor other care facilities.There were 5 males and
A trackingprocedurewasthen employedto determine
I I femalesin eachagegroup. All subjectsreportedthat
for eachsubjectwhat stimulus duration would result in
they were in good health.
when full attention wasgiven
an accuracyof 8o9o-907o
Somepsychometricdata werecollectedlor eachsubject at the conclusionof the experimentalsession.Mean to one of the two displays.For thesetrials the subjects
had to respondto one ofthe displays,and the stimulus
raw scoreson the Digit Symbol subtestof the WAIS
varied until the appropriate value was deterduration
were 67.2 and 44.0 for the young and old subjects,reFor
mined.
the fint block of this procedure,half of the
(30) = 6.69,p < .001.On the final 20 items
spectively,
subjectsin each age group respondedonly to the Xs.
of the Vocabularysubtestof the WAIS, the youngadults
The remainingsubjectsbeganby respondingonly to the
had a mean scoreof 21.6, whereasthe mean scorefor plus
array.The initial stimulusduration for eachgubject
=
(maximum
the older adults was 32.3
score 40),
was
800
msec.For the following 40 trials eachtime that
(30):3.53,p<.0r.
the subject was correct the stimulus duration was deApparatus. The subjectssat in a darkenedroom, ilcreasedby l0 msec for the next trial. Each time the
luminated only by a small, shadedlamp. Stimuli were
was incorrect the stimulus duration was inpresentedon a Tektronix 606 (P-3I phosphor)monitor subject
l0 msecfor the next trial. Beginningwith the
creased
by
in front of them. They respondedby pressingbuttons
4lst trial the subject'saccuracyover the preceding20
on two lO-keypushbuttontelephonepadsmounted next
the
to eachother on a board. A PDP-I l/03 laboratorycom- trials wascalculated.Ifthis accuracywas 807o-907o,
remained
the
for
the
following
trial
stimulus
duration
puter was used to generatestimuli and to record and
sameas the previousone. Ifthe accuracyfor the last 20
analyzesubjects'responses.
Procedure. Each trial consistedofthe simultaneous trials was below 807o,the stimulus duration was inpresentationof two visual arrays. For each array the creasedby l0 msec for the next trial. Iflhe calculated
subjecthad to indicatethe presenceor absenceofa spe- accuracywas greater than 9OVo,the stimulus duration
by l0 msec.For eachsubsequenttrial this
olied target.The computer ftrndomly determinedwhether wasdecreased
or not a targetwas to app€aron eachdisptay.This was procedurewas repeatedwith eachcalculation basedon
determined independentlyfor each array; target pres- the 20 immediately precedingtrials. When l0 consecutivecalculationsresultedin accuracyof 807o-9OVo,the
enceand absencewere equally likely events.
One anay containedfour (.4" X .4") Xs locatedat the block ended. The averagestimulus duration over the
(-orne'sof an imaginary (1.7' X 2.4") rectangleon the final 20 trials was used as the estimate for that block.
related to exposureduration. Consequently,
the performanceof different agegtroupsmay
be equatedby allowing the older subjectsa
longer stimulus presentationthan the younger subjects.Specifically,exposuredurations
were determined individually for each subject such that they would be 80Vo-90Voaccurate under single-taskconditions. Subsequent age differencesunder dual-task conditions therefore could not be attributed to
differencesin initial performancelevel.
Another featureof the presentstudy is that
it involved the construction of entire POCs
for each subject. Subjects were instructed,
under different conditions, to vary the emphasisgiven to eachof the two tasks.Divided
attention ability could then be examinedindependentof allocation strategies.
656
BENJAMIN L. SOMBERG AND TIMOTHY A. SALTHOUSE
At the completion of this block, eachsubjectperformed
a secondtracking block with the alternate display from
the one usedfor the first block. As in the first block, the
initial stimulus duration was set at 800 msec, and the
block continued until l0 consecutivecalculationsofthe
accuracyof the 20 precedingtrials yielded a value of
SOVo-909o.
At the conclusionofthat block, a rhird block
was run using the final duration estimatefrom the second block as the initial stimulus duration. The final
block was run using the display type and duration estimate from the first block. This procedure assuredthat
the final duration estimateswere relatively insensitive
to the value chosen for the initial exposuredurations
(i.e.,800 msec).
At the conclusionofthe four trackingblocks,the stimulus duration estimatesderived from the final block of
each stimulus type were averaged.The resulting value
was usedas the stimulus duration for the experimental
trials for that subject.
In the experimentaltrials the subjectsagain were exposedto both the plus display and the X display, and
they had to respond to both. There were five experimental conditions,differing in the amount of emphasis
that shouldbe givento eachdisplay.In order to facilitate
their resourceallocation, bonus money for correct responseswas offered in proportion to the desired emphasisfor each display. In one condition (0/100) the
subjectswereaskedto devotetheir total attention to the
Xs and none of their attention to the pluses,and in
anothercondition ( I 00/0) they wereaskedto attendonly
to the plusesand to ignore the Xs. In both theseconditions they were paid l0 for each correct responsefor
the attendeddisplay and nothing fiorthe display not attended. Although all conditions required respons€sto
both displayson everytrial. in thesetwo conditions the
subjectscould merely guessfor the unattendedstimuli.
In intermediate conditions (70/30, 50/50, 30/70) the
subjectswere askedto divide their resourcesbetweenthe
displaysand were rewardedaccordingly(e.g.,in the 70l
30 condition they were paid .70 for each correct plus
and .30 for eachcorrect X). There were l0 practiceand
l0O experimental trials in each block; thus, subjects
could earn a maximum of $l for eachblock. For half
ofthe subjectsin eachgroup the order ofthe conditions
was l@/0, 70/30, 50/50. 30/70,0/10O.The other half
of each group received the conditions in the reverse
order.
Resultsand Discussion
Duration threshold data. The results of
yielded stimulus duthe tracking prdure
rations of 428.1msecfor the young subjecrs
and 644.4 msec for the older subjects.The
^70
5
F
u
G
E
c
o
ieo
g
a
E
U
g
4
0
5
0
trittut ?tRFORr^rt
TAS|(
|
6
0
?
(+}
PERCEIIT
CORRECT
0
t
!
Figure I . Accuracy of detecting targetsin an array of Xs (ordinate) as a function of rcurrl.
targetsin an array ofpluses (abcissa):Experiment i.
l
of,&rccring
I
Figure 2. Sanplc g
when full atrcndro i
is divided beturccoI
difference between I
icant, (30) : 2.65,t
Accuracydua. E
accuracy PerCenret
each of five experir
means for lhe subjc
lustratedin Figurc lreveals nearly overla
two age groups.
A separateanalys
was perforrned on er
of age and five exF
factors.For each ta
effect for condition: I
105.20,p<.01;Xa
p < .01. For neitlrcr
icant Qt < .05) effecr
30): l.l8;Xarray,
Condition X Age: pl
1.79:'X array: F(4. I
indicate that the perft
vary with the insrn
questedof them, ant
tions the older subjc
the younger subjects
ysis indicates thar wirl
perfonnan@ conlrcll
difference on dual-b!
DIVIDED ATTENTION ABILITIES
MAXIMUiIPERfORiiIAI{CE
TASKI
657
OPTIMUM
OIVIDEDATTEI{TIOI{
PEBFORMAI{GE
Drvr0ED
ATTET{Tt0it
c0sr
MITIIMUM
PERFOBMAI{CE
TASKI
FUI{CTIOilAtPEFfOBMA'{CE
BEGIOIII
MIIIIMUMPERFORMAI{CE
TASK2
TASK2
MAXIMUMPERFORMAI{8E
TASK2
Figure 2. Sampleperformanceoperatingcharacteristics.(Minimum and maximum performanceoccur
when full attention is devotedto one display.Intermediatepoints representconditionsin which attention
is divided betweenthe two displays).
differencebetween these means was significant,l(30) : 2.65,p < .05.
Accuracydata. Eachsubjectproducedan
accuracy percentagefor both tasks under
each of five experimental conditions. The
means for the subjectsin each group are illustratedin Figure l. Inspectionof this figure
revealsnearly overlapping functions for the
two agegroups.
A separateanalysis of variance (nNove)
was performed on each task with two levels
of age and five experimental conditions as
factors.For each task there was a significant
effect for condition: plus array, F(4, 120) :
105.20,p < .01; X anay,F(4, 120): 109.57,
p < .01. For neither task was there a significant (p < .05) effectof age:plus array,F(1,
30) : 1.18;X array,l'(1, 30) : 2.72,or for
Condition X Age: plus array, F(4, 120):
1.79;X anay: F(4, 120) < l. Theseresults
indicate that the performanceof subjectsdid
vary with the instructional emphasis requestedof them, and that under all conditions the older subjectsperformed as well as
the younger subjects.This traditional analysis indicatesthat with individual single-task
performancecontrolled for, there was no age
differenceon dual-task performance.
It hasbeen suggested(e.g.,Duncan, 1980)
that trials in which atarget is presenton two
simultaneousdisplaysare considerablymore
difficult than trials in which a target occurs
on only one of the two displays. However
Salthouseand Somberg(1982) used tasks
which were similar to those in the present
experiment and found that trial type (i.e.,
Signal,Array l/Signal, Array 2; Signal,Array
l/No Signal, Array 2; No Signal, Array l/
Signal, Anay 2; No Signal, Array l/No Signal, Array 2) did not interact with either age
or practice.Therefore,it was consideredunnecessaryto complicate the analysesin the
presentresearchby including this additional
factor.
POC analysis. POC analyseswere performed to examine divided attention performance independent of resource allocation
strategies.As each subject had a score on
eachtask under eachof five conditions. POC
functions for individual subjectssimilar to
the group function in Figure I could be constructed.
Figure 2 illustrates the manner in which
the analyseswereconducted.In this example
the five plotted points indicate the subject's
performance in the five experimental con-
658
BENJAMIN L. SOMBERG AND TIMOTHY A. SALTHOUSE
ditions. The Functional PerformanceRegion
(FPR) is the rectangulararea that is delineatedby the maximum performanceachieved
by the subjecton eachtask (i.e.,performance
under 100/0and 0/100 conditions).It is that
regionthat definesthe potential performance
levelsofthe subject.Ifthe subjecthas a perfect ability to divide attention between the
two tasks, performance on the three intermediary conditions (30170, 50/50, 70/30\
should lie on the point at the upper right
corner of the FPR. To the extent that there
is a cost for dividing one's attention, performanceon thoseconditions will be below and
to the left of that corner. Thus, the size of
the region marked Divided Attention Cost
(DAC) is inversely related to one's divided
attention ability. The areasof the DAC and
FPR were calculatedfor each subject.
The mean divided attention costs,in units
of FPR, were 577.8 for the young subjects
and 565.4for the elderly,respectively.This
differencedid not approachstatisticalsignificance,(30) : .69.The duration adjustment
procedurewas intended to assurethat performance of all subjectswas comparablein
the I 00/0 and 0/ I 00 conditions.If successful,
one would not expect age differencesin the
sizeof the FPR. This was the case:The FPR
valueswere 1,269.7units for the young subjects and 1,330.7 for the older subjects,
(30) : .32. It shouldthereforemake no differencewhether DACs are compared in absolutetermsor relativeto eachsubject'sFPR.
To confirm this, relative DACs were calculated by dividing each subject'sDAC by his
or her FPR. The mean relative divided attention costswere .47 for both young and old
subjects.
The major conclusion from this experiment is that there was no differencein the
ability of young and old adultsto divide their
attention between twb simultaneous tasks.
Figure I suggestsa very high degreeofoverlap betweenthe functions for the two groups,
and indeed, no significant age differences
were found in two separateanalyses.
paradigm in which speedof responserather
than accuracywas the variableofinterest. It
was arguedthat a criterion for an adequate
test ofage effectson a divided attention task
is that agedifferencesin performanceunder
single-taskconditions must be controlled.
The first experimentproducedthe result that
when such differenceswere equated there
was no age differenceunder dual-task conditions. In the tachistoscopictask of Experiment I performanceof the two age groups
was equated by varying exposureduration.
This sort of manipulation is obviously impossiblefor reaction time tasks.However,in
reaction time tasks a way of controlling for
theseinitial differencesis to useeachsubject's
single-taskperformance as a base and calculate divided attention effectsas a ratio of
dual-taskperformance.This allowsjudgment
of each subject'sdual-task performancerelative to his or her own baseline.
Informal pilot data revealedthat on the
tasks chosenfor this experiment it was virtually impossiblefor a subject to systematically vary the relative emphasisgiven to the
two tasks.Thereforeonly one dual-taskcondition could be run, and the subjectswere
given no instruction about how much importance should be attached to each task.
This usually createsa problem of not being
able to separatedifferencesin divided attention ability from differencesin resourceallocation strategies.This problem is particularly severewhen the two dependent variableshave dissimilarscales.If, for example,
divided attention costs older subjects two
units more than the young on Task A, but
coststhe young one unit more than the old
on Task B, it is fair to saythat the older adults
are penalizedmore by the demandsof divided attention only if a unit on Task A is
worth more than iralf a unit on Task B. However, since the present experiment usesthe
same real-time measurementscalefor each
of the two concurrent tasks,it is reasonable
to assumethat any loss in performance on
one task will be compensatedfor by an equal
gain in performanceon the other task.
Experiment 2
The purposeofthe secondexperimentwas
to determine whether the findings of Experiment I would generalizeto an experimental
60-82) paniciparcd ra
sion. The subjetts srrt
tion as in Experimcat
that experimenr-Old.t
travel and partictpertoa.
exlra course credrt. Er
l0 females.All subccrs
health,and atl nae r{
Mean rarl scon6 oo ,
were 7l.l for thc y-or.r
adults,(30) = 6.rt3.p <
20 items on rhe WAIS
and 29.5 for the youq
(30):3.51,p<.Ot
Apparatus. Tlrc ag
identical to thal in rbc t
visual stimuli were dr
Model 13l lA screcnet
nals through a scr of Kr
Procedure.Twodift
(RT) and repetirive\e1i
separately or corrurre!
ready aprr'ared on rhc r
at a random time rrthir
ditory signal was prerc!
the signal by depressrrg
with the index fingrr of
sistedof I 0 such tnat! T
digit nurnber appearirg
then replacedby rlr 6n
served as a cue for thc
sequenceofdigits on thc
of the right hand. As O
was replaced on thc rrr
quence.Immediatell rp
the first digit in rhe:cgu
and the subjectrcpcarcd
of four runs through th
three were used in rhc el
eachblock. The amount
sponseswas recordedas t
different sequencessrrc !
and 7614853),
but ab
quencethroughout thc r
In the dual-task cond
the digit sequence*irh d
to the auditory signalsr
mental sessionbeganrrrl
with half of the subiccr
with the key task and he
This was followed b1 fa
two more blocks of cact
data revealeda sizeat{ca
and secondkey blah Tl
consideredpractice. ad
both the fint and finat h
subsequentanal)scs
Resultsand Disctus
Method
t,
Subjects. Sixteen young adults (mean age, 18.6;
range, I 8-20) and I 6 older adults (mean age, 68.7; range,
The mean reacti
and old subjec.tsw
msec, respectivel-vi
DIVIDED ATTENTION ABILITIES
!
I
60-82) participatedin a single l-hr. experimentalsession. The subjectswere chosenfrom the samepopulation as in Experiment 1, but none had participatedin
that experiment.Older subjectswere paid $5 for their
traveland participation,whereasyoung subjectsreceived
extra cours€credit. Each group contained 6 males and
l0 females.All subjectsreportedthat they were in good
health, and all were right-handed.
Mean raw scoreson the WAIS Digit Symbol subtest
were 7l.l for the young adults and 44.4 for the older
adults,t(30) : 6.43,p < .001.Mean scoresfor the final
20 items on the WAIS Vocabularly subtestwere 19.5
and 29.5 for the young and old subjects,respectively;
( 3 0 ) : 3 . 5 1 p, < . 0 1 .
Apparatus. The apparatusin this experiment was
identical to that in the previousexperimentexceptthat
visual stimuli were displayed on a Hewlett-Packard
Model l3l lA screenand subjectsreceivedauditory signals through a set ofKoss PRO/4AA headphones.
Procedure. Two differenttasks.manual reactiontime
(RT) and repetitivekeying (key), were perfbrmedeither
separatelyor concurrently. In the RT task, the word
ready apryaredon the screento initiate eachtrial, and
at a random tim€ within a 30-secinterval a single auditory signal was presented.The subject respondedto
the signal by depressinga button on the left keyboard
with the index finger of the left hand. Each block consistedof I 0 suchtrials. The key trials beganwith a sevendigit number appearingon the screenfor 3 sec.It was
then replacedby the firsr digit in the sequence,which
servedas a cue for the subject ro begin entering the
sequenceofdigits on the right keyboardwith the fingers
of the right hand. As the subject enteredeach dig,it,it
was rcplacedon the screenby the next digit in the sequence.Immediatelyupon completionof the sequence
the first digit in the sequencereappearedon the screen,
and the subjectrepeatedthe sequence.A trial consisted
of four runs through the sequence,
but only the final
three were usedin the analyses.There were l0 trials in
eachblock. The amount of time betweensuccessive
responseswasrecordedasthe interkey interval (IKI). Two
differentsequences
wereusedin the experiment(2486359
and 7614853).but each subjectretainedthe samesequencethroughoutthe session.
In the dual-task condition, the subject had to enter
the digit sequencewith the right hand while responding
to the auditory signalswith the left hand. The experimental s€ssionbeganwith two blocksoleach task alone,
with half of the subjectsin each age group beginning
with the key task and halfbeginning with the RT task.
This was followed by four dual-taskblocks and finally
two more blocks of each task alone. Inspection of the
data revealeda sizeablepracticeeffectbetweenthe first
and secondkey blocks.Thereforethe fint key block was
considered practice, and to retain counterbalancing,
both the fint and final key blocks were excludedfrom
subsequentanalyses.
Resultsand Discussion
The mean reaction times for the young
and old subjects were 231 msec and 260
msec, respectively,in the single-taskcondi-
659
tion, and 401 msec and 509 msec, respectively, in the dual-taskcondition. For the repetitive keying task the mean interkey intervals were 254 msec and 403 msec in the
single-taskcondition for the young and old
subjects,respectively.The comparablevalues
in the dual-taskcondition were281 msecand
450 msec,respectively.Identical 2 (age)x 2
(condition) ANovAs were perfonned on the
data from the two tasks. For the RT task,
there were significant(p < .05) effectsof age,
F(1, 30) : l4.l l, condition, F(1, 30) :
257.67,and Age X Condition,F(1, 30):
9.36. For the key task, there were also significant (p < .05) effectsof age,F(1, 30):
25.47,condition,F(1, 30) :75.25, and Age
X Condition,F(1,30):4.27. The absolute
amount by which the older subjects were
slowed in dual-task conditions was greater
than the amount by which the youngersubjects were slowed.This analysis,which is the
one traditionally usedto arguefor age-related
divided attention effects,in no way takesage
differencesunder single-taskconditions into
account.
Figure 3 depictsthe relationship between
performanceon the two tasks for each age
group. It is possibleto calculate the arithmetic mean of the divided attention cost as
7z([RTdual - RT alone] + UKI dual - IKI
alonel). In terms of Figure 3, this may be
representedas the mean length of the two
dashedlines extending from the point that
represents single-task performance to the
horizontal and vertical coordinates of the
point that representsdual-taskperfbrmance.
Woodworth ( 1938)arguedthat the geometric
mean may be a more meaningful way to
combine dual-task scores.He pointed out
that if an individual hasno divided attention
ability at all, it would be possibleto completely ignore one task and to perform the
other task at a level equal to the single-task
performance.The resulting combined dualtask measure(using an arithmetic computation) would be equal to half of the average
of the two taskswhen they were performed
alone, and this would not accuratelyrepresent the subject'slack of time-sharing proficiency. Therefore, Woodworth reasoned,
geometricmeansmay be a more reasonable
way of expressingcombined dual-task performance.In terms of the presentexperiment
660
BENJAMIN L. SOMBERG AND TIMOTHY A. SALTHOUSE
this may be expressedas ([RT dual - RT
alonel2+ UKI dual - IKI alonel2)r/2.These
valuesare depictedasthe solid diagonallines
in Figure 3. Both divided attention cost measureswere calculated for each subject. The
mean valuesfor the young and old subjects
were 98 and 146, respectively,for the arithmetic computation, and 172 and 257, rcspectively, for the geometric computation.
The agedifferenceswere significant for both
the arithmetic and the geometric values,
(30) : 2.09,p < .05,and t(30) : 2.03,p <
.05, respectively.
To take initial age differenceson singletask measuresinto account,it is necessaryto
look at divided attention costsrelativeto single-taskperformance. The arithmetic computation for the averagesingle-taskperformanceis 7z(RTalone + IKI alone).In Figure
3 these values are representedas the mean
length ofthe horizontal and vertical dashed
lines extending from the axesto the points
indicating single-taskperformance.The geometric respresentationof single-taskperformanceis ([RT alone])2+ UKI alonel2)r/2;
it
is illustrated in Figure 3 as the length of the
diagonal lines connecting the origin to the
points indicating single-task performance.
Relative divided attention costs were then
calculatedas ratios ofeach subject'sdivided
attention coststo the averagesingle-taskperformance for that subject. The mean arithmetic ratios were .41 and .44 for the young
and old subjects, respectively. Mean geometric ratios were .52 for the young subjects
and .56 for the old subjects.In neither case
did the agedifference approach statistical significance,(30) < I in both cases.
In this experiment the absolute cost for
divided attention was larger for the old subjects than for the young. When initial differencesin single-taskperformancewere controlled by examining dual-taskperformance
relative to each subject'ssingle-taskperformance,however,there wasno agedifference.
The similarity of the relative divided attention cost measuresin the two age gfoups is
quite consistentwith the results of the first
experiment and provides some generalizability of those findings.
General Discussion
This study has argued that a significant
amount of the literature that has purported
to havedemonstratedan adult agedifference
in the ability of individuals to divide their
attention among two or more simultaneous
taskshasfailed to take severalimportant considerationsinto account. The presentexper-
0tD
o
Y0ultG
E SIt{GLE-TASK
a
r OUAT.TASK
/
tttEAil0tvl0E0ATTEitTt0il
c0sT
MEAII
SIlIGTE.TASK
PCBf
OBIIIAI{CE
t00
200
300
400
REACT|0ilTIME(mec)
500
Figure 3. Mean responsetimes for reaction time and key tasks under single- and dual-task conditions:
Experiment2.
A
a
iments were desig
and therefore it is
the problems *err
fioq it w:rs sug
an rncreasein th
paradigms used !
effects. Although
this goal completr
simultaneous visrl
iment I are uniqu
age differences an
oftasks that haveI
ond experiment w
generalizethe findi
to a situation in wl
being measured.
The secondissu
studieshavefound
task measures, n
manc€ difficult ro
periment in this stu
procedure was usc
tions for each sub
comparable perfor
task conditions. A
condition (83.6%
84.3Vofor the oldcr
condition (85.1%
85.3Vofor the older r
cessofthis procedl
. In making comp
n essentialthat dl
the samehsk. h mi
older subjects a k
than younger subjo
the task bnd rhus in,
of POCs.Atthoqh
ot lnterpretation. tr
did not change tbe
lished comparablet
vided attention fi
The results of rhi r
it is becaus€of a hf
manipulation tbar r
literature |6s srT*r
lessable to divile t
ger adults Thc hcr
ment, in whic-bhrqt
trolled by exaninin
duced similar reorl
this conclusionAnother isq.re rh
many studies do nc
DIVIDED ATTENTION ABILITIES
661
iments were designedto addresstheseissues, performancedata.The 100/0and 0/100 conand thereforeit is necessaryto seehow well ditions fulfill this requirementin Experiment
the problems were overcome.
l. Although theseconditions did require reFirst, it was suggestedthat there should be sponsesfor both arrays,the subjectswere inan increasein the number of experimental structedto completelyignoreone array. That
paradigms used to study divided attention accuracy on the "unattended" display was
effects.Although it is impossible to satisfy at chancelevel (50.2Voacrossboth agesand
this goal completely in a single study, the conditions) indicatesthat the subjectsprobsimultaneousvisual displaysused in Exper- ably did not allocateany effort to thosestimiment I are unique in the researchof adult uli. In Experiment 2 there were conditions
age differencesand contribute to the range in which eachtask was performed alone.
oftasks that havebeeninvestigated.The secThe final issuediscussedwas the necessity
ond experiment was designedspecificallyto to separateeffectsdue to divided attention
generalizethe findingsof the first experiment limitations from effectsdue to resourcealto a situation in which speedof responsewas location strategies.The technique of exambeing measured.
ining entire trade-off functions rather than
The secondissuediscussedwas that most singledata points eliminated this ambiguity
studieshavefound an agedifferenceon single in the first experiment.In the secondexpertask measures,making dual-task perfor- iment the comparableunit assumptionsugmance difficult to interpret. In the first ex- gestedthat ifa subjectsacrificedperformance
periment in this study a duration adjustment on one task, it would be directly compenprocedure was used to find stimulus dura- satedfor by an equal gain in the other
task.
tions fior each subject that would produce Thus, regardlessof the way in which a subject
compa.rableperformance levels for single- allocatedresourcesbetweenthe two tasks.the
task conditions. Accuracies for the 100/0 mean performancewould be the same.One
condition (83.6Vofor the young subjects; way of evaluatingthis assumptionis to com84.3%for the older subjects)and the 0/100 parethe variancesof the two tasks.If the two
condition (85.l%ofor the young subjects; tasks really do have comparable scalesof
85.3%for the older subjects)attestto the suc- performance,then it would be expectedthat
cessof this procedure.
their variancesweresimilar. The pooledvari. In making comparisonsamong pOCs, it ancesfor the key and RT tasks in this exis essentialthat all subjectsperform exactly periment were 7,156.4and 7,410.5,respecthe sametask. It might be arguedthat giving tively. The parity ofthesevaluesis supportive
older subjects a longer stimulus duration of the comparable unit assumption, and it
than younger subjectsaltered the nature of therefore seemsjustifiable to use the mean
the task and thus invalidatesthe comparison of the two tasks to representdual-task perof POCs.Although this is probably a matter formance.
of interpretation, we feel that this procedure
As the abovediscussiondemonstrates.the
did not change the task but merely estab- two studiesreported here substantiallymeet
lished comparablebaselinesfrom w-hichdi- the suggestedcriteria for divided attention
vided attention effectscould be measured. research.The first experiment was designed
The results of this experiment suggestthat specifically to address the difficulties that
it is becauseof a failure to make this sort of have plagued previous research.In Experimanipulation that so much of the previous ment 2, someassumptionswererequired,but
literature has suggestedthat older adults are the similarity of the results in the two exlessable to divide their attention than youn- periments suggeststhat they are quite reager adults. The fact that the secondexperi- sonableones.It may thus be concluded
that
ment, in which baselinedifferenceswereton- in thesepain oftasks, at least,no age-related
t-rolle$by examining relativeageeffects,pro- divided attention effectwas found. The sugduced similar results providei support'for gestionis that there are other processes
(e.g.,
this conclusion.
memory or perceptualimpairments) that are
Another issue that was raised was that responsiblefor the poorer performance of
many studiesdo hot even collect single-task older subjectson individual tasks.However,
662
BENJAMIN L. SOMBERG AND TIMOTHY A. SALTHOUSE
given these initial performance discrepancies, older individuals are just as capableas
younger adults at processingadditional information. Given the conclusionof much of
the earlier work, the results of theseexperiments are rather surprising.The presentrequite clear,and the earsultsare nevertheless
lier conclusionswere basedon researchthat
is difficult to evaluatebecauseof the uncontrolled factors describedearlier.
It was pointed out that one way that age
differencesin single-taskperformance have
previouslybeen controlled for, as in a study
by Parkinson,Lindholm, and Urell (1980),
was to match subjectson single-taskmeasures.It was arguedthat this proceduremay
have selectedsubjectswho were atypical of
their age group, and it may thus have obscuredreal age differencesin divided attention ability. In the present study subjects
were not selectedon any psychometricvariable; however,it was found that in both experimentsthe older subjectshad significantly
higher WAIS vocabulary scores than the
youngersubjects.Although it is possiblethat
this differencemay partially account for the
finding of no agOdifferencein divided attention performance,two factors make this an
unlikely hypothesis.First, the subjectsin this
experimentdemonstratedthe typical pattern
of performanceon single-taskmeasures.The
older subjectswere slower on the reaction
time tasks and required a longer tachistoscopicdurationto achievean accuracycomparableto that ofthe youngersubjects.It was
only on dual-task performance relative to
these baselines that the age effect disappeared.Second,it was found that the WAIS
Vocabulary scores were uncorrelated with
divided attention measures.In the first experiment the correlation betweenWAIS Vocabularyscoreand divided attention costwas
.12 for the young subjectsand -.11 for the
old subjects.In the second experiment the
correlationbetweenthe vocabularyscoreand
the relativedivided attention cost was .18 for
the young and -.15 for the old.
A secondmajor contribution of this study
is
. methodological.Although the arguments
herehavebeencouchedin terms of adult age
differences,they may be applied to any individual difference comparisons.There are
many practical as well as theoretical needs
for measuresof divided attention ability, and
the methodology describedhere provides a
way of obtaining that information.
The questionof controlling for differences
under single-taskis of particular relevance
when one is concernedwith measuringbetween-groupdifferencesin divided attention
ability. Either the experimental conditions
must be manipulated so that the initial performance under single-taskconditions is
equivalent for all groups, or some type of
data transformation (e.g.,the use of relative
scores)must be employed.In the absenceof
these, between-groupcomparisons may be
meaningless.
The difficulties associatedwith the confounding of divided attention limitations
is an issue
with resourceallocationstrategies
any time that divided attention performance
is assessed.
The method that is usually employed is to collect only a single data point
for each subjectunder dual-taskconditions.
In this situation it is often uncertain whether
two points differ becauseof divided attention
limitations or whether they are actually two
pointson the samePOC function.Only when
enough data have been collectedso that the
entire POC function may be constructedcan
differencesbe unambiguouslyinterpreted.
References
Broadbent,D. E. Perceptionand communication.London: PergamonPress,1958.
Broadbent,D. 8., & Heron, A. Effectsof a subsidiary
task on performanceinvolving immediate memory
by younger and older men. British Journal of Psychology,1962,53, 189-198.
Clark, L. E., & Knowles,J. B. Agedifferencesin dichotic
listeningperformance.Journal of Gerontology,1973,
28, t73-t'18.
Craik, F. I. M. The nature of the agedecrementin performanceon dichotic lislening tasks.Quarterly Journal of ExperimentalPsychology,1965,I 7, 22'7-24O.
Craik, F. I. M. Age differencesin human memory. In
J. E. Birren & K. W. Schaie(Eds.),Handbook of the
psychologyof aging.New York: Van Nostrand Reinhold,1977.
Duncan, J. The locus of interferencein the perception
of simultaneousstimuli. PsychologicalReview, 1980,
87,2'72-300.
Eriksen,C. W., Hamlin, R. M., & Breitmeyer,R. G.
Temporal factors in visual perception as related to
aging.Percepti on & Psychophysics, 197O, 7, 35 4-3 56.
Green,D. M., & Swets,J. A. Signal detectiontheoryand
psychophysics.
New York: Wiley, 1966.
Inglis, J., & Ankus, M. N. Effectsof ageon short-term
storage and serial ro
Psychology,tqis. -{6
Inglis, J., & Caird_$' I
responscsto srmufu
Joumal of Psrclxlqr
Kinchla, R. A. Tbc nrr
Nickerson (Fdr .|r:,
Hillsdah. \J: Erbr
Kirchoer. l['. l( ,{1r d
of raprdl,r rrr-A
r
moud fu.
t1
LippsBuct Ln*-.
diftrcoctr tn ue$
Etqvtnltlsol ChM P
Mrler.H {.&lnjl:r.
audlrr 11rqr ir\a
2m.
tt&ra.,
D t.r.'qrr
of d
rftrrr
Wilir.
tt?l
Normeo,D-A-lh
rwunD-l11
7,4441.
Parkinrca. S. R- f-dichoric nrory.
d
tolog.. 19fi. lS.n-a
Salthousc.T. A.. lt Sod
DIVIDED ATTENTION ABILITIES
storage and serial rote learning. British Journal of
Psychology,
1965,56, 183-195.
Inglis, J., & Caird, W. K. Age diffrjrencesin successive
responsesto simultaneous stimulation. Canadian
Journalof Psychology,1963,17,98-105.
Kinchla, R. A. The measurementof attention. In R. S.
Nickerson (8d,.), Attention and performance VIIL
Hillsdale.N.J.: Erlbaum. 1980.
Kirchner, W. K. Age differencesin short-term retention
of rapidly changing information. Journal of Experimental Psychology,1958,55, 352-358.
Lipps Birch, L. Baselinedifferences,attention, and age
differencesin time-sharing perfiormance.Journal of
ExperimentalChild Psychology,1978, 25, 505-5 I 3.
Mackay, H. A., & Inglis, J. Effectof ageon a short-term
auditory storageprocess.Gerontologia,1963,8, 193200.
Malarazzo, J. D. Wechsler'smeasurementand appraisal
of adult intelligence. Baltimore, Md.: Williams &
Wilkins.1972.
Norman. D. A., & Bobrow, D. G. On dataJimited and
resource-limitodproc€sses.
CognitivePsychology, 1975,
7. 44-64.
Parkinson.S. R., Lindholm, J. M., & Urell, T. Aging,
dichoric memory, and digit span.Journal of Gerontologr'. 1980.J5, 87-95.
Salrhousc.T. A., & Somberg,B. L. Skilledperformance:
663
The effectsofadult ageand experienceon elementary
process.Journal of Experimental Psychology:Gmeral,
t982, IIt, t76-207.
Sperling, G. The attention operating characteristic: Examplesfrom visual search.Science,1978,202,31531 8 .
Sperling, G., & Melchner, M. J. Visual search, visual
attention, and the attention operating characteristic.
In J. Requin (El.l, Anention and performance VII.
Hillsdale, N.J., Erlbaum, 1978.
Talland, G. A. The effects of age on sp€ed of simple
manual skill. Joumal of Genetic Psychology, 1962,
100,69-76.
Walsh, D. A., Till, R. 8., & Williams, M. V. Age difference in perceptual processing:A monoptic back.
ward masking investigation. Journal of Experimental
Psychology: Human Perception and Performance,
1978,4,232-243.
Walsh, D. A., Williams, M. V., & Hertzog, C. K. Agerelated differencesin two stagesof central-perceptual
processes:The efects of short duration targets and
criterion differences.Journal of Gerontology, 1979,34,
234-24r.
Woodworth, R.S. Experimental psychology.New York:
Holt. 1938.
ReceivedDecember l, 1981 r