You are on page 1of 14

Journal of Experimental Psychology:

Human Perception and Performance Copyright 1 y the American Psychological Association, Inc.
1988, Vol. 14, No. 2, 267-280 0096-1523/88/S00.75

Effects of Aging and Task Difficulty on Divided Attention


Performance
Joan M. McDowd and Fergus I. M. Craik
Erindale College, University of Toronto, Toronto, Canada

We report two experiments that compare the performance of young and older adults on
perceptual-motor tasks involving division of attention. Previous studies have shown older people
to be especially penalized by divided attention situations, but the generality of this finding was
recently challenged by Somberg and Salthouse (1982). The present study was conducted to
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

investigate the possibility that age differences in dual-task performance are amplified by an
This document is copyrighted by the American Psychological Association or one of its allied publishers.

increase in the difficulty of the constituent tasks, where difficulty was manipulated by varying
the central, cognitive nature of the tasks (Experiment 1) or the degree of choice involved
(Experiment 2). With the present tasks, strong evidence was found for an age-related decrement
in divided attention performance. Contrary to our original expectations, however, it does not
seem that division of attention presents some especial difficulty to older people. Rather, division
of attention is one of several equivalent ways to increase overall task complexity. In turn, age
differences are exaggerated as tasks are made more complex.

There is an extensive body of literature documenting the When two tasks must be performed concurrently, older people
changes in cognitive abilities that come with increasing age in may perform less well than their younger counterparts because
adulthood (see Kausler, 1982, and Salthouse, 1982, for re- of decreased storage capacity or less flexible deployment of
views). One such change that has been demonstrated in many control processing. A further possibility is that the processes
experimental studies is the poorer performance of older adults controlling the division of attention may require more "pro-
in situations that require division of attention (Broadbent & gram space" in older people, leaving fewer processing re-
Gregory, 1965; Broadbent & Heron, 1962; Inglis & Caird, sources to carry out the tasks themselves (Craik, 1973, 1977).
1963; Kirchner, 1958; Welford, 1958). On the basis of this However, the need for such speculations was called into
literature, Craik (1977) stated that "one of the clearest results question by the arguments and findings of Somberg and
in the experimental psychology of aging is the finding that Salthouse (1982). These authors questioned the validity of the
older adults are more penalized when they must divide their existing evidence for an age decrement in divided attention
attention" (p. 391). performance on methodological grounds and presented two
The reasons for the older person's greater difficulty in new experiments in which older adults were found to be at
performing two concurrent mental tasks are not clear. One no more of a disadvantage under divided attention conditions
suggestion is in terms of decreased "functional distance" than were younger adults. Their objections to previous work
between the regions responsible for the two sets of mental included the point that many of the studies on which the
operations in the older brain (Kinsbourne, 1980); the idea is original conclusion was based used the dichotic listening
that the necessary operations show a greater tendency to paradigm—a wider selection of tasks would be preferable.
mutual interference in older people. A second type of expla- Somberg and Salthouse also commented on a frequent failure
nation appeals to information-handling limitations in older to report single-task measures on the tasks studied and a
adults. For example, Welford (1958, 1977) has postulated a failure to take age differences in single-task performance into
decline in short-term memory capacity, "especially when account.
attention has to be switched to other data or action during The studies reported by Somberg and Salthouse (1982)
the period of retention" (Welford, 1958, p. 484). This rela- were designed to overcome these limitations. In the first
tively active view of short-term storage is similar to the notion experiment, subjects detected the presence of a target on each
of working memory, which, according to Baddeley and Hitch of two simultaneously presented visual displays. Age differ-
(1974), consists of a limited capacity "work space" that can ences on single-task detection performance were eliminated
be divided between storage and control processing demands. by adjusting the presentation duration of the visual displays
for each subject so that detection accuracy was 80%-90% on
each display. Under these conditions, no age differences were
This research was supported by a grant from the Natural Sciences
observed when attention was divided between the displays. In
and Engineering Council of Canada to F. I. M. Craik and by funds
the second experiment, speed of response rather than accuracy
from the Program in Gerontology, University of Toronto.
was examined. The tasks were (a) simple reaction time (RT)
We wish to thank Lily Moysiuk for technical help, and we are
grateful to the reviewers for useful suggestions.
to an auditory signal and (b) repetitive keying of a constant
Correspondence concerning this article should be addressed to seven-digit sequence. Subjects performed each task alone and
Joan M. McDowd, who is now at the Department of Psychology, also performed the two tasks simultaneously, entering the
University of Southern California, Los Angeles, California 90089. digit sequence on a keyboard with the right hand while

267
268 JOAN M. McDOWD AND FERGUS I. M. CRAIK

responding to the auditory signals with the left hand. For each and Salthouse might be due to differences in the complexity
task, the interaction between age and single/dual-task per- of the tasks used, with complexity defined as the number of
formance was statistically reliable, showing that the older hypothesized mental operations required. They further sug-
subjects were penalized more than were young subjects by gested that age differences emerge when the overall complexity
dual-task conditions. Previous workers have typically taken of the situation is high or when demands on processing
this type of evidence to indicate age differences in divided capacity are great—"the larger the number of mental opera-
attention. However, Somberg and Salthouse argued that be- tions to be performed, the larger is the absolute age difference
cause there are substantial age differences even under single- between young and older adults" (Salthouse et al., 1984, p.
task conditions, the proper way to evaluate the effect of 619). This line of reasoning echoes an earlier suggestion by
divided attention is to express the increases in RT associated Salthouse (1982) that division of attention does not represent
with dual-task performance as a ratio of single-task RTs. some special circumstance but is just one of many cases in
When the data of their second study were expressed in terms which overall complexity is high.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

of these relative divided attention costs, age differences in


It is therefore likely that the effects observed in many dual-task
This document is copyrighted by the American Psychological Association or one of its allied publishers.

performance disappeared. The problem of different baselines situations are no different than those that would be obtained
in developmental research is a difficult one, and the solutions when task difficulty or complexity is manipulated in a single
adopted by Somberg and Salthouse are probably as reasonable task. This argument thus suggests that the requirement of per-
forming two concurrent tasks is simply another means of increas-
as any. The ratio measure is certainly a very conservative one,
ing task complexity, and that there may not be a specific age
however, and it is possible that data transformed in this way deficit associated with divided attention, per se. (Salthouse, 1982,
overcompensate for baseline differences (Ackerman, Schnei- p. 189)
der, & Wickens, 1984).
The present studies were undertaken to gather further in-
In any event, Somberg and Salthouse's (1982) findings of
formation on age differences in performance under conditions
either small age differences or no age differences in divided
of divided attention and, in particular, to examine the notion
attention performance are clearly at odds with results of
previous studies. One possibility is that their methodological that age-related decrements in divided attention performance
increase as the difficulty and complexity of the component
improvements revealed the true state of affairs, but a second
tasks increase. In Experiment 1, difficulty was manipulated
possibility is that the tasks used by Somberg and Salthouse
required relatively shallow (Craik & Lockhart, 1972) or au- qualitatively by using tasks that involved either shallow sen-
tomatic (Hasher & Zacks, 1979) processing. The literature is sory-motor decisions or deeper cognitive decisions. In Exper-
consistent in showing (a) that age decrements in performance iment 2, task difficulty was again varied qualitatively, but
are greater in tasks requiring more complex translation proc- complexity was also varied quantitatively by using tasks re-
esses between perception and action (Welford, 1958) and (b) quiring two-choice, four-choice, or eight-choice decisions. The
that older people are particularly penalized when the com- points at issue were, first, a further exploration of the condi-
plexity involves substantial central as opposed to peripheral tions under which age differences in divided attention per-
processing (Cerella, Poon, & Williams, 1980). formance are observed; second, an investigation of the sug-
The possibility that age differences in divided attention gestion that age decrements in this situation are amplified as
performance can be affected by the tasks involved is given the complexity of the component tasks is increased; and third,
some credibility by the results of two experiments reported an examination of Salthouse's (1982) suggestion that division
by Craik (1973) and discussed by Kausler (1982). In the first of attention simply represents one case in which overall
experiment, subjects attempted to detect a faint tone on one complexity is high.
auditory channel while performing a short-term recognition
memory task for digits on a second auditory channel; the age Experiment 1
decrement in divided attention performance was slight. How-
ever, the age decrement was substantially greater in the second The purpose of Experiment 1 was to explore the notion
experiment, in which subjects performed a visual short-term suggested by Craik (1973) that division of attention is more
memory task for letters while monitoring a lengthy string of disruptive to performance in subjects of all ages when the
digits (presented auditorily) for the presence of targets, defined component tasks involve relatively deep cognitive processing
as three consecutive odd digits (e.g., 391, 537). It is at least than when they involve shallow sensory-motor processing;
reasonable to argue that the second set of tasks involved more further, that the age-related decrement in performance is
central, cognitive processing than did the first. amplified by tasks that require deeper processing. Although
Salthouse, Rogan, and Prill (1984) also reported age differ- depth of processing cannot be measured precisely at this time,
ences under conditions of divided attention when the tasks the results of many studies show strong effects of the variable
were more complex than those used by Somberg and Salt- on aspects of memory and attention (Craik & Tulving, 1975;
house (1982). In the later study, Salthouse et al. presented Dark, Johnston, Myles-Worsley, & Farah, 1985; Hyde &
subjects with two simultaneously displayed lists of letters and Jenkins, 1969; Johnston & Heinz, 1978; Treisman, 1964,
digits, respectively, for short-term recall. The difficulty level 1969). Depth of processing, therefore, seems to be a dimension
was equated across subjects by presenting each person with of some psychological importance.
strings that were 75% of that person's span for the material The present study involved two visual tasks and two audi-
(digits or letters) presented alone. Salthouse el al. suggested tory tasks. In line with previous studies (e.g., Craik & Tulving,
that the difference between their results and those of Somberg 1975), the two tasks within each modality involved the same
AGING, TASK DIFFICULTY, AND DIVIDED ATTENTION 269

stimuli and the same responses, but the qualitative nature of were targets, semirandomly positioned within each sequence. Again,
the stimulus-response (S-R) mapping differed between the the subject's task was to press a key each time a target was detected;
members of each pair of tasks, so that one task required only nontargets received no response. Reaction times on both tasks were
measured with a millisecond timer that was triggered by an electronic
superficial sensory processing, whereas the other required
signal on the second channel of the magnetic tape recorded to be
deeper, cognitive processing. To keep the exposition theoret-
coincident with the onset of the target word. The timer was stopped
ically neutral, the two members of each pair of tasks are by the subject's keypress, and the RT was recorded by the experimen-
designated easy and difficult, respectively. Subjects performed ter.
each of the four tasks alone and also in the four dual-task Visual tasks. Both visual tasks involved the same computer-gen-
combinations made possible by combining one visual with erated display. This display was a 3-cm x 7-cm rectangle presented
one auditory task (easy/visual-easy/auditory . . . difficult/ with its longer side on the horizontal axis. The rectangle was divided
visual-difficult/auditory). We predicted that the difficulty of with vertical lines into four sections of equal size, and on each trial
dual-task performance, relative to performance on the com- an alphanumeric character appeared in one of the four sections. On
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ponent tasks performed singly, would increase from the easy- one half of the trials the character was a number, and on the other
This document is copyrighted by the American Psychological Association or one of its allied publishers.

half of the trials the character was a letter of the alphabet. Of the
easy combination to the difficult-difficult combination, with
trials that were numbers, half were odd digits and half were even
the other two combinations yielding intermediate results. We
digits. Of the trials that were letters, half were vowels and half were
also predicted that this ordering of dual-task combinations consonants. The nature of the character and its location within the
would be more disruptive to the performance of older subjects rectangle varied randomly, with equal numbers of item types appear-
than to the performance of their younger counterparts. ing in each position.
For both versions of the visual task, there were four response keys.
In the easy (shallow) version, the subject pressed the key correspond-
Method ing to the position of each alphanumeric character as it appeared in
one of the four sections of the rectangle. That is, the correct response
did not depend on the type of character displayed; position was the
Subjects only important variable. The stimulus-response mapping was
straightforward; that is, the correct response to a character in the left-
Sixteen young and 16 older adults participated in the experiment.
most section of the display was to press the leftmost key, and so on.
The young adults (Jl/age = 19.4 years; SD = 1.1 year) were under-
For the difficult (deep) visual task, each of the four response keys
graduates at Erindale College, University of Toronto, and received
was assigned one of the following labels: vowel, consonant, odd digit,
course credit for their participation. The older adults (M age = 69.0
or even digit. In this case, the subject's task was to identify the
years; SD = 5.6 years), were home-living volunteers in good health
character that appeared on the screen (regardless of its position within
who came to the laboratory to be tested; they received travel expenses
the rectangle) and press the key corresponding to the appropriate
but no payment for their services. All subjects were given the Mill
category to which the character belonged. This task is referred to as
Hill vocabulary test, a six-alternative, forced-choice, synonym-rec-
the character task. For both versions of the visual task, the stimulus
ognition test with a total possible score of 20. Scores on vocabulary
display stayed on the screen until the subject made a response. A
tests hold up or even increase somewhat with increasing age (Heron
correct response caused a new character to appear immediately—
& Chown, 1967; Salthouse, 1982). Performance on such tests is
thus, the tasks were of the continuous reaction time (CRT) type. If
therefore often used to match different age groups on verbal ability,
the subject made an incorrect response, the same display was repeated
inasmuch as verbal abilities relate to other types of cognitive perform-
until a correct response was made; in this case, the subject's RT was
ance. In the present study, vocabulary scores were 12.3 (SD = 2.3)
cumulated until the correct key was pressed. Errors were thus reflected
and 15.7 (SD = 2.7) for the young and older groups, respectively—a
in lengthened RTs. In both versions, subjects performed the task
significant superiority for the older group, £(30) = 4.06. p < .01. If
continuously for a 24-s period, and the measure taken was the mean
anything, the older group had higher verbal abilities, and thus any
correct response time.
age-related deficits found in task performance are likely to underes-
timate the true state of affairs.

Procedure
Tasks
Subjects were seated in front of a computer screen and informed
Auditory tasks. On each trial of both versions of the auditory task, that they would be performing two tasks—one in response to visual
the subject monitored a list of 12 words for the presence of targets. stimuli and one in response to auditory stimuli. Because there were
In the easy (or shallow) version, 9 of the 12 words were spoken in a two tasks in each presentation mode, there were four dual-task
male voice, whereas the 3 target words were spoken in a female voice. combinations: voice-position, voice-character, category-position,
The subject's task was simply to press a response key each time a and category-character. Each subject performed one block of trials
target word was detected. The serial positions of target words were under each of the four combinations, and the order of task conditions
randomly assigned for each list, with the restriction that no two was counterbalanced across subjects.
targets were presented in adjacent serial positions. Fourteen such lists Within each block, the appropriate visual task was explained first
were prerecorded on tape at a rate of one word per 2 s and were and the subject given one practice sequence of 15 stimulus-response
presented to subjects via headphones. This easy version of the audi- (S-R) events. This practice was followed by a scored 24-s sequence
tory task is referred to as the voice task. (this sequence was of approximately equal length to the time taken
In the difficult (or deep) version, 14 lists of 12 words were also to presept one 12-word list in the auditory task). The 24-s visual
presented, but the subject's task was to detect words denoting living sequence gave the subject's score for the visual task performed alone.
things. In this category task, all words were spoken by a male speaker Instructions for the appropriate auditory task were then given, and a
at the same rate as in the easy task (one word per 2 s). As in the easy practice trial was performed. Next, the subject was given one practice
version, the words were all two-syllable nouns, and 3 of the 12 words sequence, with both tasks performed concurrently. Finally, three
270 JOAN M. McDOWD AND FERGUS I. M. CRAIK

scored dual-task sequences and two scored single-task auditory lists Reaction Times
were presented; the order of these five experimental sequences was
randomized separately for each subject in each block. Table 2 shows mean RTs for the two age groups on the
This procedure was repeated for each of the four task combination four tasks performed singly and in the appropriate combina-
blocks. A 30-s rest period was provided between each block. In the tions. Table 2 shows longer reaction times for the older group
dual-task conditions, subjects were instructed to divide their attention for tasks carried out in combination, as opposed to alone, and
equally between the two tasks. for the difficult tasks—both the task on which the RT was
measured and when that task was combined with a difficult
Results task.
Single-task performance. Focusing first on the RTs for each
Errors task performed alone, Table 2 shows that for both the auditory
and visual tasks, the more difficult (deeper) task took longer
Given the design of the visual tasks, in which a given display
to perform. Analyses of variance (ANOVAS) were carried out
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

was presented until the subject made a correct response to it,


within each modality. In the case of the visual tasks, there
This document is copyrighted by the American Psychological Association or one of its allied publishers.

the only errors recorded in the present experiment were errors


was a significant effect of difficulty, F(l, 30) = 145.3, p <
of omission (misses) in the auditory tasks. Essentially no errors
.001; an effect of age, F(l, 30) = 52.2, p < .002; and a
were made in the single-task conditions (a total of 7 errors for
significant Age x Difficulty interaction, F(l, 30) = 5.66, p <
the older group and 1 error for the young group). However,
.05. For the auditory tasks, the effect of difficulty was again
under divided attention conditions a number of errors were
reliable, F([, 30) = 360.7, p < .001; the effect of age was
made by both age groups; the group totals and percentages
marginally significant, F(l, 30) = 3.83, p < .06; but the Age
are shown in Table 1. The table shows that more errors were
x Difficulty interaction did not approach significance, F <
made in the deeper category task and that older subjects made
1.0. Some caution should be exercised in interpreting the
relatively more errors in this condition.
increase in RTs from the easy auditory task (voice) to the
Because so few errors were made on the auditory task, no
difficult auditory task (category); the nature of the tasks simply
formal statistical analysis was carried out on the error data.
necessitates a longer sample of the word in the latter case, in
However, two points may be made from the pattern of data
which it must be decided whether the word denotes a living
in Table 1. First, RTs were the major dependent variable in
thing, than in the former case, in which the sex of the speakers
the experiment, and because older subjects made more errors,
is judged. Thus, whereas we continue to refer to the category
presumably the recorded RTs for the older group underesti-
task as being more difficult than the voice task (in the sense
mate the values that would have been found if error rates had
that the category task necessitates more central processing), it
been equated across the two age groups. That is, speed-
seems certain that the increase in RT cannot all be attributed
accuracy trade-off considerations make it likely that if error
to an increased difficulty of processing in the category task.
rates had been equated between the age groups, the RTs for
However, the primary interest lies in how RTs on these tasks
older subjects would have been higher. As we shall see, this
changed when they were performed in combination.
would have further amplified the effects actually found. Sec-
Dual-task performance. There are various ways in which
ond, Table 1 shows that both older and young subjects made
the dual-task data may be displayed and analyzed. The major
few errors in the relatively shallow voices task but that the age
point at issue is whether older adults are differentially penal-
decrement was substantial in the deeper category task. This
pattern of results is exactly that discussed speculatively early
in the article, that older subjects may not be differentially Table 2
penalized in divided attention conditions when the tasks are Mean Reaction Time (in Milliseconds) as a Function of Age,
relatively automatic and require little analysis (e.g., Somberg Task, and Task Combination
& Salthouse, 1982) but that an age decrement in divided Task and group Alone Easy Difficult
attention does emerge when more cognitive tasks are used.
Visual tasks
(with voice) (with category)

Table 1 Position (easy)


Old 632 782 878
Total Errors and Percentage Errors on the Auditory Tasks
Young 409 435 450
for Young and Older Adults in Experiment 1 Character (difficult)
Young group Older group Old 1108 1462 1721
Young 728 805 967
Condition Total % Total %
Voice Auditory tasks
(with position) (with character)
Alone 0 0.0 0 0.0
Position 1 0.7 0 0.0 Voice (easy)
Character 2 1.4 4 2.8 Old 372 649 973
Category Young 318 478 589
Alone 1 0.7 7 7.3 Category (difficult)
Position 6 4.2 15 10.4 Old 734 1066 1413
Character 7 4.9 25 17.4 Young 694 799 1004
AGING, TASK DIFFICULTY, AND DIVIDED ATTENTION 271

ized in situations involving division of attention when tasks Absolute divided attention (DA) costs. In Salthouse's ter-
demanding moderate cognitive involvement are used, and minology, the absolute DA cost for a given task is simply the
further, whether the age-related decrement in division of amount by which mean RT increased from single- to dual-
attention is increased when task difficulty is increased. Figures task conditions. These difference scores are presented in Table
1 and 2 shed light on these questions by showing single- and 3. In Table 3, the target task is the one for which the measures
dual-task performance for each age group and in each of the are given, and the shared task is the task it was combined
four task combinations. Note that the single-task data are with in that condition. Table 3 shows that divided attention
necessarily repeated; for example, the values for the position costs are higher for the older group in all conditions; this
task performed alone are the same in the position-voice and contrast is equivalent to the Age x Single/Dual interaction in
position-category combinations. The figures show clearly that the previous analysis. In addition, the layout of the data in
older subjects were penalized more by the shift from single- Table 3 enables us to determine whether age differences in
to dual-task conditions and that the size of this Age x Single/ DA costs are affected by the difficulty of the task being
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Dual-Task interaction is amplified by the use of more difficult measured and also whether age differences in DA costs are
This document is copyrighted by the American Psychological Association or one of its allied publishers.

tasks. This statement is confirmed by the results of an analysis affected by combining the target task with a more difficult
of variance (ANOVA) on the combined data from the two shared task.
figures; only the major results of interest are reported. The To assess these effects, a 2 x 2 x 2 x 2 (Age x Modality x
interaction between age and single/dual-task performance was Difficulty of Target Task x Difficulty of Shared Task) ANOVA
reliable, F(l, 30) = 22.2, p < .001, showing that the require- was carried out on the data in Table 3. The four main effects
ment of divided attention was more detrimental to the per- were all highly reliable (p < .01). Of greater interest, the
formance of older adults than to that of young adults. The interactions of age with both target task difficulty, F( 1, 30) =
three-way interaction between age, single/dual, and task com- 5.07, p < .03, and with shared task difficulty, F(l, 30) = 5.64,
bination was also significant, F(3, 90) = 4.04, p < .01. The p < .03, were reliable, and neither interaction was qualified
significant three-way interaction shows that the age decrement by significant higher order interactions. Thus, we can con-
in divided attention is itself affected by the tasks involved. clude that the age decrement in divided attention performance
Both figures show a tendency for the age decrement to be (given here by DA costs) was amplified reliably both by
exacerbated as the component tasks increased in difficulty. increasing the difficulty of the task being performed and by
The four-way interaction between age, single/dual-task per- increasing the difficulty of the concurrent task.
formance, task combination, and modality, was not signifi- The data may be further condensed by calculating the
cant (F< 1.0). average DA cost for each pair of combined tasks. For example,
Factors contributing to the age decrement in divided atten- the mean cost for the position-voice combination in the
tion may be further specified by looking at divided attention young group is (26 + 160)/2 = 93 ms; for the character-voice
costs, and it is to these derived data that we now turn. combination in the young group, the average DA cost is (77

AUDITORY TASKS
1600
VOICE VOICE CATEGORY CATEGORY
I POSITION) (CHARACTER) (POSITION)
1300

1 100 •

500

SINGLE DUAL SINGLE DUAL SINGLE DUAL SINGLE DUAL

TASK CONDITION

Figure 1. Reaction times for the two auditory tasks (voice and category) as a function of age, single or
dual task condition, and task combination.
272 JOAN M. McDOWD AND FERGUS I. M. CRAIK.

VISUAL TASKS

POSITION CHARACTER CHARACTER


1700 (CATEGORY) (VOICE)

1500

— 1300
o
s
£ 1100

9
K °°
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

700
This document is copyrighted by the American Psychological Association or one of its allied publishers.

500
____ -o
-

300
SINGLE DUAL SINGLE DUAL SINGLE DUAL SINGLE DUAL

TASK CONDITION

Figure 2. Reaction times for the two visual tasks (position and character) as a function of age, single
or dual task condition, and task combination.

+ 271 )/2 = 1 7 4 ms. One can argue strongly for averaging DA baseline levels are substantially higher for the older group (see
costs over the two component tasks in that if the subject task-alone data, Table 2). Accordingly, Somberg and Salt-
chooses to concentrate more heavily on one or the other task, house (1982) suggested that relative DA cost was the more
the DA cost will simply be increased on the less attended task. appropriate measure; this figure is calculated by expressing
The subject cannot lessen the total DA cost by using a strategy. each of the absolute DA costs as a proportion of average
The means for each age group and task combination are single-task performance on the two component tasks. For
shown in Table 4 under absolute DA costs. example, single-task RTs for young subjects in the position-
In line with the original predictions, the absolute DA cost voice combination were 409 and 318 ms, respectively; there-
figures in Table 4 increase in general from the combination fore, relative DA costs were 93 / 1/2(409 + 318) = 0.26.
of the two easy tasks (position-voice) to the combination of Values for both age groups in all four task combinations are
the two difficult tasks (character-category). The older group shown in Table 4 under relative DA costs. Table 4 shows that
shows greater DA costs and, crucially, the age difference is the values for relative DA costs were substantially higher for
amplified by the increasing difficulty of task combination. An the older group than for the young group in all cases. An
ANOVA on these data yielded significant effects of age, F(l, ANOVA on the data yielded a significant effect of age, F(l, 30)
30) = 22.76, p < .001; task combination, F\l, 90) = 29.86, p = 12.96,p< .001, and task combination, F(3, 30) = 13.12,p
< .001; and the Age x Task Combination interaction, F(3, < .001. The interaction between age and task combination
90) = 4.49,p<.01. was not significant, however, F(3, 90) = 1.07. There is some
It is possible that the absolute DA cost measure gives a indication from Table 4 that the easiest task combination
misleading picture of age differences, however, because the (position/voice) was associated with a smaller age difference
in relative DA costs than the other combinations, but because

Table 3
Mean Absolute Divided Attention Costs (in Milliseconds) as Table 4
a Function of Age and Task Combination in Experiment 1 Mean Divided Attention Costs as a Function of Age and
Target Task Task Combination in Experiment 1
Visual Auditory Position Character
DA costs and
Easy Difficult Easy Difficult group Voice Category Voice Category
Shared task (position) (character) (voice) (category)
Absolute
Young Young 93 73 174 275
Easy 26 77 160 105 Old 214 289 478 646
Difficult 41 239 271 310 Relative
Old Young .26 .13 .33 .39
Easy 150 354 277 332 Old .43 .42 .65 .70
Difficult 246 613 601 679 Note. DA = divided attention.
AGING, TASK DIFFICULTY, AND DIVIDED ATTENTION 273

the interaction was not significant, we should conclude that


the lower costs for the younger group were relatively constant
across all conditions.
To summarize, the results of all analyses are clear in show-
ing that the increased difficulty (as indexed by RTs) associated
with performing perceptual-motor tasks under divided atten-
tion as compared with focused attention conditions is affected
by the nature of the tasks themselves. Tasks that necessitate
deeper cognitive processing are more negatively affected, even
when ratio measures are taken (relative DA costs, Table 4).
The data are also clear in showing that older people are
relatively more penalized by division of attention—again, this
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

is true for both absolute and relative measures. There is some


This document is copyrighted by the American Psychological Association or one of its allied publishers.

ambiguity, however, surrounding the question of whether task


difficulty further exacerbates this age-related decrement;
analysis of the absolute DA cost data suggests that it does,
whereas the relative DA cost data suggest that it does not.

Divided Attention and Complexity

Salthouse (1982) made the interesting suggestion that divi-


sion of attention simply acts to increase the overall complexity
of the situation, thereby necessitating the involvement of more
mental operations. With respect to the effects of aging, it may
be the case that the observed age decrement in performance
would therefore be as great in a single complex task as in two
shared tasks. The alternative possibility is that there is some-
thing about the divided attention situation itself that is espe- 200 400 600 800 1000 1200
cially difficult for the older person to handle. RT-YOUNG
One way in which the present data may be examined in
Figure 3. Mean reaction times for young and old subjects on
order to address this issue is to plot mean young RTs against
auditory and visual tasks in Experiment 1. (E = easy version of task;
mean older RTs for the various tasks (Salthouse, 1978). In D = difficult version of task; A = task alone; thus EA = easy auditory
the case of tasks performed singly, we would expect that as task alone, DE = difficult visual task combined with easy auditory
more complex tasks are considered, RTs go up for both age task, etc.)
groups. Furthermore, it is suggested in previous work (Cerella
et al., 1980) that the complexity function is reasonably linear,
with a slope greater than 1.0. That is, for a given increase in and the percentage of variance accounted for by the linear
RT for a group of young subjects, as tasks increase in com- regression is 81%. These data, therefore, give good support to
plexity, there will be a correspondingly greater increase in Salthouse, Rogan and Prill's (1984) suggestion—division of
RTs for the older group. The question of interest is what attention apparently acts in a way that is very similar to an
happens when tasks performed under divided attention con- increase in single-task complexity.
ditions are added to this plot. If divided attention acts simply
to increase complexity, then the further points should lie on Discussion
the same function; however, if divided attention has some
especially disruptive effect on older people, these further A full discussion of the results and their implications will
points would be expected to form a different function with a be given after Experiment 2 is reported. For the moment it
higher slope or higher intercept than the first. may be emphasized that in contrast to the claims of Somberg
The data from the present study are shown plotted in this and Salthouse (1982), Experiment 1 showed strong age effects
way in Figure 3. Separate regression equations were calculated in the costs associated with division of attention. Older adults
for single- and dual-task data. For single task, the slope and were affected more by division of attention in the number of
intercept were 1.31 and 10.2, respectively; for dual task, the errors made, in absolute DA costs, and in relative DA costs.
slope and intercept were 1.46 and 111.6, respectively. Com- In line with the suggestions of Salthouse et al. (1984) and the
parisons of slopes and intercepts across task conditions indi- results of Craik (1973), this marked difference in conclusions
cated that neither of the sets of parameters differed from one may be attributable to the different tasks used by Somberg
another significantly (all ts < 1.0). One linear function de- and Salthouse on the one hand and the present investigators
scribes the total data set quite well; the product-moment on the other. The present tasks all clearly involved controlled
correlation between RT-older and RT-young is r( 10) = +.90, processing in the sense of Shiffrin and Schneider (1977),
274 JOAN M. McDOWD AND FERGUS I. M. CRAIK

whereas the detection and repetition tasks used by Somberg On the basis of the results from Experiment 1, we predicted
and Salthouse appear to be more automatic in nature. Also, that there would be an age decrement in divided attention
the present tasks were arguably more complex and involved performance even on the position task and that this decrement
deeper levels of processing, although neither variable was would be amplified by increases in complexity. On the basis
defined unambiguously. Experiment 2 represents an attempt of Experiment 1, it was also expected that the age-related
to manipulate complexity and degree of choice in a more decrement would be greater in the deeper faces task and that
objective manner. this decrement would again be exacerbated by complexity.
One major purpose of Experiment 1 was to examine the Finally, the data may be used to explore further the similarity
suggestion that age differences in divided attention are ampli- of division of attention to other forms of task complexity.
fied by the use of more complex or more difficult tasks. The
hypothesis received good support from graphic analysis (Fig-
ures 1 and 2) and from an analysis of absolute RT costs but Method
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

was not well supported by the results of relative RT costs.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Two further points that received confirmation were, first, that


Subjects
deeper (or more complex) tasks are associated with greater
DA costs than are shallow tasks under dual-task conditions The young subjects were Erindale College undergraduates who
and, second, that the effects of divided attention are very participated either for course credit or for $5. The older adults were
similar to those associated with an increase in single-task healthy, community-living volunteers, recruited over the course of
complexity. Experiment 2 examined the same issues but several years to serve as subjects in experiments on aging. They were
attempted to place task complexity on a more objective not paid for their participation but were reimbursed for travel ex-
footing by using choice RT tasks involving two, four, or eight penses to and from the laboratory.
In the position task, the mean age of younger subjects was 21.0
choices. Thus complexity was defined in terms of the amount
years (SD = 2.7), and the mean age of the older group was 71.9 years
of information associated with each decision. Again, the
(SD = 5.3). Mean Mill Hill vocabulary scores were 14.2 and 16.8 for
choice RT tasks were performed by young and older subjects
the young and older groups, respectively; this difference was signifi-
under conditions of focused and divided attention. Two visual cant by / test, ((34) = 3.58, p < .01. For the faces task, the mean ages
choice RT tasks were used; the first required only perceptual- were 21.0 years (SD = 4.4) and 67.3 years (SD = 5.2) for the young
motor analysis, and the second required substantial cognitive and old groups, respectively; the corresponding vocabulary scores
analysis. were 13.4 and 16.2, respectively. The age-related increment in the
vocabulary test was again significant, £(34) = 3.81,/>< .01.

Experiment 2

In overview, Experiment 2 explored age differences in


Tasks
divided attention performance as a function of task complex-
Position task. The general paradigm for both visual tasks was that
ity—defined quantitatively in terms of number of choices— of continuous choice reaction time. Three levels of single-task diffi-
and task difficulty—manipulated qualitatively as described culty were included: two-choice, four-choice, and eight-choice reac-
below. In Experiment 2, performance was measured on the tion time. For the position task, the visual display was generated on
visual tasks only: The visual task in a given condition was the video display terminal of a PDP-11 computer. It consisted of a
either performed alone or in combination with a standard rectangle with its long side on the horizontal axis. The rectangle was
auditory task. Thus the function of the auditory task was divided into two, four, or eight separate lanes, depending on the task
simply to provide the second task in dual-task conditions; condition. In front of the subject was a row of eight keys, each of
performance changes were measured on the visual task. Two which was assigned to one of the lanes in the visual display. The
target was a set of two asterisks, which on a given trial would appear
visual tasks were used; the first was the position task from
in one of the possible lanes of the rectangle. The subject's task was to
Experiment 1—subjects pressed keys corresponding to lights
respond as quickly as possible to the appearance of the target by
in two, four, or eight positions. As in the previous experiment, pressing the appropriate response key. Response times were recorded
this position task was judged to involve minimal cognitive by the PDP-11 computer.
processing. The second visual task involved cartoon faces The visual task was subject paced; as soon as a response was made,
presented on a computer screen. The faces were either male the next target would appear. The location of the target asterisks on
or female, were shown with a hat or without a hat, and were a given trial was randomly determined by the computer program.
smiling or frowning. In the two-choice condition, subjects Reaction times associated with incorrect responses were not included
responded on the basis of sex (regardless of other attributes); in any data analysis. A block of trials consisted of as many responses
in the four-choice condition, both sex and attire (hat/no hat) as could be made in 45 s.
Faces task. In this task, the visual stimuli were schematic faces
were relevant; and in the eight-choice condition, all three
created by using the Sprite graphic capabilities of the Commodore 64
dimensions were relevant. Relative to the positions task, the
computer. The faces varied on three dimensions, and each dimension
relation between stimulus and response in the faces task was had two levels, making a total set of eight different faces. The
of low compatibility. On the basis of the work of Greenwald dimensions were sex, male or female; attire, hat or no hat; and
(1972) on ideomotor compatibility, the faces task was consid- expression, smile or frown. The basic face was identical for both men
ered to require appreciably more cognitive processing than and women, but the hats were different for the two sexes, and the
did the position task. man wore a bow tie, whereas the lady wore a necklace.
AGING, TASK DIFFICULTY, AND DIVIDED ATTENTION 275

In front of the subject was a row of eight response keys that were dual-task conditions were collected under conditions in which the
assigned to the various faces, depending on the choice condition. The concurrent auditory task was performed with at least 75% accuracy.
subject's task was to classify the face presented on a given trial on the
basis of one, two, or three of the varied dimensions. The row of eight
keys was separated by a space into two groups of four keys. One Procedure
group was assigned to the female faces, and one group was assigned
to the male faces. The assignment of attribute to response key was At the beginning of the experimental session, the subject was seated
arbitrary but constant, and the assignments within each sex were in front of a video screen and a row of response keys, and the nature
mirror images of each other. Both hands were used in each choice of the two tasks was explained by the experimenter. Several practice
condition: one finger from each hand in the two-choice condition, trials on the auditory task were allowed for the subject to become
two fingers from each hand in the four-choice condition, and four familiar with the task. Then a total of 7 single- and 7 dual-task trials
fingers of each hand in the eight-choice condition. Subjects rested were carried out at each level of choice in the visual reaction time
their fingers on the relevant response keys while waiting to respond; task. The 14 trials were carried out in the following order: 3 single-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

responses were registered by depressing any key. task trials, 4 dual-task trials, 2 single, 3 dual, and 2 single. The first
This document is copyrighted by the American Psychological Association or one of its allied publishers.

For each condition, a card was prepared for the subject to use as a single-task and the first dual-task trials were considered to be practice
guide indicating the relevant attribute-response key assignments. This and were not included in any further analyses. The alternating of
card was placed immediately below the computer screen on which single- and dual-task trials was intended to control for practice effects
the faces appeared, and the card was present throughout the relevant across trials and to ensure, as far as possible, the validity of the single-
trials. Thus subjects could refer to the card when necessary, and any to dual-task trial comparison. The subjects were informed that they
load on memory was minimized. were to try to detect all the targets on the auditory task and that they
All eight faces appeared in each of the three choice conditions, but would have to repeat trials in which their accuracy fell below 75%.
the number of features critical to classification was varied. In the two- Other than the initial practice, the auditory task was performed only
choice condition, responses were made on the basis of sex. Subjects under dual-task conditions. The experimenter initiated each trial with
simply pressed one of two response keys to indicate whether the a verbal ready-begin signal and operated the tape recorder and the
character on the screen was male or female. In the four-choice computer that controlled the visual display. Each trial was terminated
condition, the relevant attributes were sex and attire; subjects pressed at the end of 45 s by the experimenter's verbal stop signal.
one of four keys to indicate the presence or absence of a hat within The order in which the two-, four-, and eight-choice tasks were
the particular sex. In the eight-choice condition, all three attributes performed was counterbalanced across subjects to control for possible
were relevant: sex, attire, and expression. In this case, subjects pressed practice or fatigue effects. At the end of the experiment, any dual-
one of eight response keys to indicate the particular combination of task trials in which a 75% accuracy level had not been reached on
attributes present in the stimulus face. Response times were recorded the auditory task were repeated. Thus each subject produced six
using the millisecond timer in the Commodore 64 computer. single-task and six valid dual-task trials at each of the three levels of
The task was subject paced; a stimulus face stayed on the screen choice in the visual task. Experimental sessions lasted approximately
until a response was made by the subject. As soon as a response was 1 hr.
made, the next stimulus face appeared. The order of appearance of
the stimulus faces was randomly determined. Reaction times associ-
ated with incorrect responses were not included in any data analyses. Results
A block of trials consisted of as many responses as could be made in
45s.
In all, 72 subjects participated in the experiment; 18 young subjects Errors
and 18 older subjects performed each task. Thus, task was a between-
subjects variable, as, of course, was age; the remaining two variables Statistical analyses were carried out only on RTs that were
(choice and single/dual) were manipulated within subjects. associated with correct responses. However, both older and
Auditory task. Subjects monitored a continuous series of spoken younger subjects maintained a high degree of accuracy across
digits for the occurrence of targets, defined as any run of three levels of complexity, so that few observations were lost by
consecutive odd digits (e.g., 1, 7, 5, or 3, 9, 1). The digits were discounting error RTs. The high levels of accuracy can be
prerecorded and played back on a tape recorder at a rate of 1 digit seen in Table 5, which shows mean proportions of errors on
every 1.5 s. Response to a target was vocal. When subjects heard the
the visual tasks as a function of age, choice, and single/dual-
third digit in a target run, they spoke the third digit out loud, and the
task condition.
experimenter recorded its accuracy. A trial consisted of 30 digits
Separate ANOVAS were carried out on arcsin transformations
filling 45 s. Within each trial, there were four target runs of three
consecutive odd digits. of the error proportions in the position and faces tasks. For
The purpose of the auditory task was simply to ensure concurrent the position task, neither age nor single/dual affected the
processing of two tasks in the divided attention condition; no detailed proportion of errors reliably. There was a significant effect of
measures of performance on this task were taken. Instead, subjects choice, however, F(2,68) = 20.9, p <. 001, and the interaction
were required to maintain a constant level of performance on the of choice and single/dual was also reliable, F(2, 68) = 3.19, p
auditory task, and the focus of interest was the level of performance < .05. For the faces task, the error rates were somewhat
they achieved on the concurrent visual reaction-time task. Detection higher. In this case, choice was again the only significant main
accuracy in the auditory task had to equal or exceed 75% (i.e., subjects
effect, F(2, 68) = 89.4, p <. 01; the only reliable interaction
were required to detect three of the four targets on a given trial in
was the three-way interaction among age, choice, and single/
order for the trial to count). If detection accuracy fell below 75% on
dual, F(2, 68) = 6.57, p < .01. This interaction reflects the
a dual-task trial, that trial was discounted and more trials were
performed at the end of the experiment to fill in the missing obser- fact that older and younger subjects performed comparably
vations. Thus all the data included for analysis from the choice RT in the two-choice and four-choice tasks but that the older
276 JOAN M. McDOWD AND FERGUS [. M. CRAIK

Table 5 involving age—that among Age x Choice X Task, F(2, 136)


Mean Error Proportions in Experiment 2 as a Function of = 5.53, p < .01. This effect reflects the fact that age differences
Age, Task, and Experimental Condition in performance were more strongly affected by increasing
Task
complexity in the faces task than in the position task. Cru-
cially, neither the Age x Single/Dual x Choice interaction
Two choice Four choice Eight choice
Task and nor the Age x Single/Dual x Task interaction approached
group Single Dual Single Dual Single Dual significance. Thus, contrary to prediction, the age decrement
Position observed under dual-task (or divided attention) conditions
Young .01 .01 .02 .01 .03 .03 was not amplified reliably by an increase in task complexity
Old .01 .02 .02 .02 .03 .04 (degree of choice) nor by an increase in task difficulty (from
Faces position to faces tasks).
Young .02 .03 .02 .02 .06 .07
Old .02 .02 .03 .02 .08 .13
In line with the results of Experiment 1, the present data
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

did show that the difference between single- and dual-task


This document is copyrighted by the American Psychological Association or one of its allied publishers.

performance was amplified reliably both by task complexity


subjects made relatively more errors in the eight-choice con- and task type. That is, for Single/Dual x Choice, F(2, 136) =
dition, especially under dual-task conditions. 27.0, p< .01, and for Single/Dual x Task, F( 1, 68) = 49.4, p
The main point to emerge from the error analysis is that, < .01. These effects are qualified by a highly reliable three-
with the exception of the eight-choice faces task, errors were way interaction, however: For Single/Dual x Choice x Task,
quite rare. Such errors as there were increased somewhat with F(2, 136) = 35.8, p < .02. This three-way interaction reflects
task difficulty; therefore, as in Experiment 1, reaction times the fact (see Table 7, absolute DA costs) that divided attention
for the more difficult conditions are likely to be somewhat costs drop with increasing complexity in the position task,
low. There are no indications of trade-offs between time and but rise substantially in the faces task.
errors across the conditions of the experiment. Divided attention costs. The absolute DA costs were calcu-
lated by subtracting single-task RTs from dual-task RTs in all
Reaction Times conditions. These values are shown in the top half of Table
7. Further analyses were not performed on these derived data
Table 6 shows the mean RTs for both age groups in all because the effects were already shown as interactions with
conditions.1 There are clear effects of age, task, choice, and single/dual in the ANOVA just reported. However, the absolute
single/dual conditions. These effects were confirmed by a DA costs shown in Table 7 make it clearer that DA costs
four-factor ANOVA on the data in Table 6; the main effects of decline with increasing complexity in the position task but
age, F(\, 68) = 27.8; task, F(l, 68) = 214.9; choice, F(2, 136) rise with complexity in the faces task. Also the age difference
= 260.1; and single/dual, F(l, 68) = 142.1, were all reliable in DA costs tends to decline with complexity in the position
a t p < . 0 1 or better. task but increases quite substantially in the faces task; this
Of the interactions involving age, the Age x Task interac- latter trend in the data is reflected by a tendency for the four-
tion was marginally significant, F(l, 68) = 3.69, p < .06, way interaction among Age X Single/Dual X Complexity x
showing that the difference in RTs between the position and Task to approach significance, F(2, 136) = 2.56, .05 < p <
faces task was somewhat greater for the older group. Both the .10.
Age x Choice interaction, F(2, 136) = 10.66, and the Age x Tables 6 and 7 thus show that whereas absolute RTs
Single/Dual interaction, f(1,68) = 18.12, were highly reliable, increased with the degree of choice in the position task, the
showing that older subjects' performance was more detrimen- cost of performing the task under divided attention conditions
tally affected than that of their young counterparts, both by actually decreased with degree of choice for both age groups.
the increase in complexity associated with a greater degree of Speculative explanations for this unpredicted effect are sug-
choice and by division of attention. These two-way interac- gested in the Discussion section.
tions were qualified by one reliable three-way interaction The bottom half of Table 7 shows relative DA costs for
both tasks—that is, the absolute DA costs divided by the time
to perform the task under single-task conditions. For these
Table 6
tasks, relative DA costs declined with an increase in task
Mean Reaction Times (in Milliseconds) as a Function of
Age, Task, and Experimental Condition in Experiment 2
Experimental condition 1
Note that these response times appear unusually short relative to
values typically reported in the literature for similar tasks. It is possible
Two choice Four choice Eight choice
Task and that these short response times are an artifact of the nature of the
group Single Dual Single Dual Single Dual visual display. That is, the stimuli were displayed in a double-height
Position mode: The first line was built on the screen, followed by the second
Old 182 383 394 541 533 680 line, after which the response timer was started. Although the building
Young 143 190 279 318 393 413 of the display is rapid, it is not instantaneous. Thus it is possible that
Faces subjects could initiate their responses on the basis of the visual
Old 600 964 765 1177 1665 2520 information in the first line, before the response time clock was
Young 472 684 649 840 1194 1634 started. This, in turn, would produce artificially short response times.
AGING, TASK DIFFICULTY, AND DIVIDED ATTENTION 277

Table 7
Mean Divided Attention Costs as a Function of Age, Task, D POSITION TASK
and Choice Condition in Experiment 2 A FACES TASK
Experimental condition

Task and group Two choice Four choice Eight choice

Absolute DA costs
Position
Old 201 147 147
Young 47 39 20
Faces
Old 364 412 855
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Young 212 191 440


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Relative DA costs
Position
Old 1.14 0.37 0.27
Young 0.35 0.15 0.08
Faces
Old 0.60 0.52 0.43
Young 0.44 0.29 0.36
Note. DA = divided attention. 200 «30 600 800 1000 1200 1400 1600 1800

RT-YOUNG ( m s e c )
complexity, partly reflecting the higher baselines in the four-
Figure 4. Mean reaction times for young and old subjects on the
and eight-choice conditions and also (in the position task) visual tasks in Experiment 2. (Numbers 2, 4, 8 refer to choice
reflecting the fact that absolute DA costs decline with increas- condition; S = single task condition; D = dual task condition.)
ing choice. Separate ANOVAS conducted on the data from each
task revealed significant effects of age in each case, F(l, 34)
DA costs were greater on the deeper faces task than on the
= 34.1, p < .01, for the positions task, and F(l, 34) = 4.46, p
<. 05, for the faces task. Degree of choice was associated with position task (the F value for the Single/Dual x Task inter-
reliable decreases in relative DA costs for both tasks, and action was F[ 1, 68] = 49.4, p< .001). It should be noted that
finally, the Age x Choice interaction was reliable in the case whereas we have referred to the difference between the two
of the position but not for the faces task. tasks as one of different depths of processing or amounts of
cognitive involvement, there is also a clear difference in S-R
compatibility between the tasks. The position task involves
Divided Attention and Complexity very compatible S-R mapping, whereas the faces task is much
less compatible. It is known that highly S-R compatible tasks
As in Experiment 1, the mean RTs for the young and older
are performed faster and are less vulnerable to interference
groups on each task condition were plotted together (Figure
from other tasks (Fitts, 1964).
4). Again, separate regression equations were calculated for
Divided attention costs were also amplified by task com-
single- and dual-task data. For single tasks, the slope and
plexity, but the picture is clouded by the fact that whereas the
intercept were 1.38 and negative 31.9, respectively; for dual
degree of choice is associated with increases in absolute DA
tasks, slope and intercept were 1.48 and 41.7, respectively.
costs on the faces task, there is a slight trend in the opposite
Again, within each parameter across task type, these values
direction on the position task. Two speculative reasons for
did not differ from one another significantly (again, all ts, <
the unexpected pattern of results on the position task are,
1.0). Figure 4 shows that the points are well fitted by one
first, that the overall decision time is taken up by mental
linear function. The product-moment correlation between
operations that vary considerably in the resources they re-
young and older group means across the 12 conditions was r
quire. If the proportion of relatively undemanding operations
= +.94, and the percentage of variance accounted for by the
increases with degree of choice, then the effects of divided
linear regression was 88. It may therefore be concluded (in
attention would be absorbed more easily in the four- and
line with the conclusion from Experiment 1) that the effects
eight-choice conditions. For example, the times taken to
of division of attention are equivalent to those caused by
identify the location of the stimulus and to select and execute
increases in task complexity associated with degree of choice
the correct response may increase (as well as central decision
and with qualitative variations in processing requirements.
times) with greater degrees of choice, and these relatively
undemanding operations may occupy time that can also be
Discussion used to perform the secondary task. A second possibility
follows from Kahneman's (1973) suggestion that subjects
Before discussing the results relating to the effects of aging, simply cannot harness as much effort in easy tasks as they
it is of some interest to note that divided attention costs were can in difficult tasks. It may be that in the simple two-choice
strongly affected by the task performed. As in Experiment 1, condition of the position task, no more effort could be directed
278 JOAN M. McDOWD AND FERGUS I. M. CRAIK

to the task to prevent it from suffering some decrement under divided attention costs would be amplified by increases in
dual-task conditions. In the case of the more difficult eight- task difficulty.
choice task, however, increased effort is garnered and per- With respect to the first issue, Experiment 1 showed that
formance benefits, thus preventing any substantial decrement absolute DA costs rose significantly with task difficulty—both
due to divided attention. Clearly, both suggestions are specu- the difficulty of the task being measured and the difficulty of
lative and post hoc; it may be sufficient to note that increases the task with which it was combined. When absolute costs
in complexity associated with increasing degrees of choice do were pooled for particular task combinations, it was again
not necessarily lead to concomitant increases in DA costs. found that DA costs rose as the component task increased in
With respect to aging, the major predictions were that an difficulty; moreover, this result held both for absolute and
age decrement in divided attention performance would again relative DA costs. In the first experiment, difficulty was ma-
be found and that this age-related effect would be amplified nipulated by varying the depth of processing required for the
by differences in task type and by increases in task complexity. tasks. In Experiment 2, both depth (or S-R compatibility)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Some of the predictions were borne out by the data, but some and degree of choice were manipulated. The overall ANOVA
This document is copyrighted by the American Psychological Association or one of its allied publishers.

received no support. First, the overall age decrement in per- showed a strong interaction between task and divided atten-
formance was exacerbated by complexity in both tasks, and tion costs, F(l, 68) = 49.4, p < .001, reflecting the fact that
this two-way interaction was qualified by a significant three- DA costs were substantially higher in the faces task. The
way interaction among age, complexity, and task; that is, the effects of choice were less clear-cut, however, inasmuch as the
Age x Complexity interaction was larger in the faces task, position task yielded anomalous results; DA costs fell, if
which presumably involved greater amounts of central proc- anything, with increasing complexity—possibly because the
essing. This finding echoes similar patterns of results involving costs were absorbed in longer peripheral processing times
age differences, degree of choice, and decreasing stimulus- associated with the more complex conditions or possibly
response compatibility (Welford, 1958). because the more complex conditions allowed for more effort
Strong age differences in divided attention performance to be mobilized and deployed. However, the faces task did
were found in both tasks, and this was true both for absolute show a substantial interaction between single/dual and com-
and for relative DA costs. However, contrary to prediction, plexity. On balance, therefore, the experiments provided rea-
the Age x Single/Dual interaction was not affected by either sonably strong evidence that at least absolute DA costs in-
degree of choice or by the qualitative nature of the task. That crease with the difficulty of the component tasks, in which
is, neither the Age x Single/Dual X Complexity interaction, difficulty is manipulated by increasing the degree of cognitive
nor the Age x Single/Dual x Task interaction approached involvement necessitated by the task and (in certain cases, at
significance (F< 1.5, in both cases). Table 7 shows that age least) by increasing the degree of choice involved.
differences in absolute DA costs actually decline somewhat The second issue concerned the reality of age-related dec-
with increasing complexity on the position task, and this may rements in performance under dual-task conditions. Here,
reflect features of this particular task, as speculated above. the results of the present studies were quite unambiguous;
For the faces task, the pattern of RT differences is in accord- highly reliable age decrements in divided attention perform-
ance with prediction, but the Age x Single/Dual x Complex- ance were found in all conditions of both experiments, and
ity interaction does not reach significance. The failure to find this result held for both absolute and relative DA costs. Taking
effects of either task complexity or task type on age-related the present results along with the similar findings from many
DA costs thus fail to support the suggested reason for the previous studies (see Craik, 1977, and Salthouse, 1982, for
difference in results reported by Somberg and Salthouse reviews), it can be reasonably claimed that the results of
(1982) and Salthouse et al. (1984). We return to this problem Somberg and Salthouse (1982) are the anomalous ones. These
in the General Discussion section. investigators found no age differences in divided attention
Finally, Figure 4 lent strong support to the notion that costs, conceivably because the tasks used were simple, repet-
division of attention acts to increase task complexity in a itive, and involved relatively automatic processing.
fashion very similar to the increases associated with increasing The third, and major, point addressed by the present ex-
choice and with increased task difficulty. There is apparently periments was therefore to examine the possibility that age
no reason to believe that the difficulty associated with division differences in DA costs are negligible when the component
of attention presents a special problem to older people that is tasks are simple and relatively automatic but that an age
different from that faced on any complex task. decrement appears with complex tasks and is amplified by
further increases in complexity. The present experiments
provide at least some support for this hypothesis. The only
General Discussion contrary evidence comes from the results of the position task
in Experiment 2.
The two experiments reported in this article were carried In any event, the crucial Age x Single/Dual x Task Diffi-
out with three main issues in mind. The first was to examine culty interaction was found in Experiment 1, although it was
the effects of different tasks on performance under divided statistically reliable for absolute but not for relative DA costs.
attention conditions; the second was to gain further insight In the faces task of Experiment 2, the RT data show the
into the circumstances in which age differences in divided hypothesized pattern of results but did not reach statistical
attention performance do and do not occur; and the third was significance; the interaction was reliable for the error data,
to examine the specific hypothesis that age differences in however. There is, therefore, some limited support for the
AGING, TASK DIFFICULTY, AND DIVIDED ATTENTION 279

notion that the age decrement in DA costs is amplified by and theoretical approach. Human Factors, 26, 71-82.
increases in task complexity. A further possibility is that the Allport, D. A. (1980). Attention and performance. In G. L. Claxton
presence or absence of an age difference in DA costs depends (Ed.), New directions in cognitive psychology (pp. 112-153). Lon-
don: Routledge & Kegan Paul.
on other task parameters such as whether the tasks require
Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. H.
automatic or controlled processing (Hasher & Zacks, 1979;
Bower (Ed.), The psychology of learning and motivation (Vol. 8,
Shiffrin & Schneider, 1977).
pp. 47-90). New York: Academic Press.
What are the implications of the present results for theories Birren, J. E. (1965). Age changes in speed of behavior: Its central
of aging? One finding that seems reasonably clear is that tasks nature and physiological correlates. In A. T. Welford & J. E. Birren
performed under divided attention conditions lie on the same (Eds.), Behavior, aging, and the nervous system (pp. 191-216),
regression function as tasks performed under conditions of Springfield, IL: Charles C Thomas.
focused attention (Figures 3 and 4). The slopes of the regres- Broadbent, D. E., & Gregory, M. (1965). Some confirmatory results
sion lines relating young to older RTs were 1.56 and 1.47 for on age differences for simultaneous stimulation. British Journal of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Experiment 1 and 2, respectively—values in good agreement Psychology, 56, 77-80.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Broadbent, D. E., & Heron, A. (1962). Effects of a subsidiary task on


with the mean slope of 1.62 reported by Cerella et al. (1980)
performance involving immediate memory in younger and older
for subjects over 60 years of age compared with subjects in
men. British Journal of Psychology, 53, 189-198.
their 20s. The implication of the present findings is therefore
Cerella, 3. (1985). Information processing rates in the elderly. Psycho-
that the dual-task situation represents one of several possible, logical Bulletin. 98, 67-83.
and apparently equivalent, ways of increasing overall task Cerella, J., Poon, L., & Williams, D. (1980). Age and the complexity
complexity (Salthouse, 1982) that differentially penalizes hypothesis. In L. Poon (Ed.), Aging in the 1980s (pp. 332-340).
older adults. Washington, DC: American Psychological Association.
Why are older people penalized more by increasing com- Craik, F. I. M. (1973, August). Signal detection analyses of age
plexity? One line of explanation is that suggested by Birren differences in divided attention. Paper presented to the Annual
(1965) and developed by Salthouse (1982, 1985) to the effect Convention of the American Psychological Association, Montreal.
Craik, F. I. M. (1977). Age differences in human memory. In J. E.
that mental operations take longer to perform with increasing
Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging
age and that this behavioral slowing is amplified as the task
(pp. 384-420). New York: Van Nostrand Reinhold.
involves a greater number of operations. It has also been Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A
suggested (Cerella, 1985; Cerella et al., 1980) that central or framework for memory research. Journal of Verbal Learning and
higher order processes show greater effects of slowing than do Verbal Behavior, 11, 671-684.
peripheral, sensory-motor processes, and this suggestion fits Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the
well with the observation from the present experiments that retention of words in episodic memory. Journal of Experimental
the effects of aging on decision times were greater in tasks Psychology, 104, 268-294.
ostensibly requiring deeper processing. Dark, V. J., Johnston, W. A., Myles-Worsley, M., & Farah, M. J.
A related factor that may be relevant here is the subject's (1985). Levels of selection and capacity limits. Journal of Experi-
mental Psychology: General, 114, 472-497.
ability to deal with the overall task in a coordinated, holistic
Fitts, P. M. (1964). Perceptual-motor skill learning. In A. M. Melton
fashion. One effect of aging may be a progressive inability to
(Ed.), Categories of human learning (pp. 244-285). New York:
deal with a complex task in a holistic way and instead having Academic Press.
to deal with the task as a collection of components. The Greenwald, A. G. (1972). On doing two things at once: Time sharing
resources required to coordinate the component operations as a function of ideomotor compatibility. Journal of Experimental
(the concurrence costs, in Navon & Gopher's, 1979, phrase) Psychology, S, 52-57.
are likely to be greater in the second case. As a final specula- Hasher, L., & Zacks, R. T. (1979). Automatic and effortful processes
tion, dissimilar tasks are easier to perform concurrently, pos- in memory. Journal of Experimental Psychology: General, 108,
sibly because they draw on different resource pools (Allport, 350-388.
Heron, A., & Chown, S. (1967). Age and function. Boston: Little
1980; Navon & Gopher, 1979), and it is possible that two
Brown.
concurrent tasks may be more easily kept functionally distinct
Hyde, T. S., & Jenkins, J. J. (1969). Differential effects of incidental
by the younger information-processing system. Cerebral atro-
tasks on the organization of recall of a list of highly associated
phy in the aging brain may make it more difficult to maintain words. Journal of Experimental Psychology, 82, 472-481.
distinct and insulated resource pools (Kinsbourne, 1980), and Inglis, J., & Caird, W. K. (1963). Age differences in successive
such decreased functional distance would act to make each responses to simultaneous stimulation. Canadian Journal of Psy-
task interfere with the smooth execution of the others to a chology, 17, 98-105.
greater extent. Johnston, W. A., & Heinz, S. P. (1978). Flexibility and capacity
Sensible choices among these and other forms of explana- demands of attention. Journal of Experimental Psychology: Gen-
tion can be made only on the basis of experimental observa- eral, 107, 420-435.
Kahneman, D. (1973). Attention and effort. New York: Prentice Hall.
tions. The present experiments represent a contribution to
Kausler, D. H. (1982). Experimental psychology and human aging.
that data base.
New York: Wiley.
Kinsbourne, M. (1980). Attentional dysfunction in the elderly: The-
References oretical models and research perspectives. In L. W. Poon, J. L.
Ackerman, P., Schneider, W., & Wickens, C. D. (1984). Deciding the Fozard, L. S. Cermak, D. Arenberg, & L. W. Thompson (Eds.),
existence of a time-sharing ability: A combined methodological New directions in memory and aging: Proceedings of the George A.
280 JOAN M. McDOWD AND FERGUS I. M. CRAIK

Talland Memorial Conference (pp. 113-129). Hillsdale, NJ: Erl- human information processing: 2. Perceptual learning, automatic
baum. attending, and a general theory. Psychological Review, 84, 127-
Kirchner, W. K. (1958). Age differences in short-term retention of 190.
rapidly changing information. Journal of Experimental Psychology, Somberg, B. L., & Salthouse, T. A. (1982). Divided attention abilities
55, 352-358. in young and old adults. Journal of Experimental Psychology:
Navon, D., & Gopher, D. (1979). On the economy of the human Human Perception and Performance, 8, 651-663.
processing system. Psychological Review, 86, 214-255. Treisman, A. M. (1964). Verbal cues, language and meaning in
Salthouse, T. A. (1978). Age and speed: The nature of the relationship. selective attention. American Journal of Psychology, 77, 206-219.
Unpublished manuscript, Washington University, St. Louis, De- Treisman, A. M. (1969). Strategies and models of selective attention.
partment of Psychology. Psychological Review, 76, 282-299.
Salthouse, T. A. (1982). Adult cognition: An experimental psychology Welford, A. T. (1958). Aging and human skill. London: Methuen
of human aging. New York: Springer-Verlag. Press.
Salthouse, T. A. (1985). Speed of behavior and its implications for Welford, A. T. (1977). Motor performance. In J. E. Birren & K.. W.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

cognition. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the Schaie (Eds.), Handbook of the psychology of aging (pp. 450-496).
This document is copyrighted by the American Psychological Association or one of its allied publishers.

psychology of aging (pp. 400-426). New York: Van Nostrand New York: Van Nostrand Reinhold.
Reinhold.
Salthouse, T. A., Rogan, J. D., & Prill, K. (1984). Division of
attention: Age differences on a visually presented memory task. Received February 9, 1987
Memory & Cognition. 12, 613-620. Revision received July 13, 1987
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic Accepted July 14, 1987

APA Buys Clinician's Research Digest

APA has acquired the Clinician's Research Digest and will take over format publication of
the digest as of July 1, 1988. Presently published by the California-based Relational Dy-
namics Institute, CRD offers practitioners brief summaries of clinically relevant research
findings and other clinical information.
Clinton W. McLemore, PhD, president of Relational Dynamics, founded CRD in 1983.
McLemore will continue to serve as CRD editor through June 30, 1988. A new editor, to be
selected, will take over as of July 1, 1988.
The CRD acquisition was proposed by the ad hoc Committee on Practitioner Publications
(PPC), chaired by Charles D. Spielberger. From 1984 to 1987, the PPC, established by the
P&C Board at the behest of the BOD's Subcommittee on the Future of Professional Educa-
tion in Psychology, made several recommendations for practice-oriented publications tai-
lored for health service providers, school/educational psychologists, and I/O psycholo-
gists—including the development of monograph series for each group.
A continuing education program, which has also been acquired by APA, is offered in
conjunction with the CRD. RDI will continue to operate the program under license from
APA.
For the present, information on subscriptions to CRD and the CRD CE Program can be
obtained from Clinical Information Services, P.O. Box 61025, Pasadena, California
91106-9990. CRD will be issued monthly beginning in January 1988. 1988 subscription
rates: individuals, $48; institutions, $62.

You might also like