Serial or parallel processing in dual tasks: what is more effortful?

Abstract

Recent studies indicate that dual tasks can be performed with a serial or parallel strategy and that the parallel strategy is preferred even if this implies performance costs. The present study investigates the hypothesis that parallel processing is favored because it requires less mental effort compared to serial processing. A serial or parallel processing strategy was induced in a sample of 28 healthy participants. As measures of mental effort, we used a rating as well as heart rate (HR) and electrodermal activity. Parallel processing again showed performance costs relative to serial, whereas serial processing was judged as more effortful. Also tonic HR and phasic HR deceleration were increased with a serial strategy. Thus the preference for a parallel strategy in dual tasks likely reflects a compromise between optimizing performance and minimizing the amount of mental effort. This aspect is neglected in current dual task accounts so far.

Full text

Page 1

Serial or parallel processing in dual tasks: What is more
effortful?
CAROLA LEHLE, MARCO STEINHAUSER, AND RONALD HUBNER
Universitat Konstanz, Konstanz, Germany
Abstract
Recent studies indicate that dual tasks can be perfonned with a serial or paral1el strategy and that the paral1el strategy is
preferred even if this implies perfonnance costs. The present study investigates the hypothesis that parallel processing is
favored because it requires less mental effort compared to serial processing. A serial or parallel processing strategy was
induced in a sample of 28 healthy participants. As measures of mental effort, we used a rating as well as heart rate (HR)
and electrodennal activity. Parallel processing again showed perfonnance costs relative to serial, whereas serial pro-
cessing was judged as more effortfuL Also tonic HR and phasic HR deceleration were increased with a serial strategy.
Thus the preference for a parallel strategy in dual tasks likely reflects a compromise between optimizing perfonnance
and minimizing the amount of mental effort. This aspect is neglected in current dual task accounts so far.
Descriptors: Nonnal volunteers, Cognition and motivation, Electrodennal, Heart rate
Performing dual tasks, that is, performing two tasks at the same
time or in close succession, typical1y leads to costs compared to a
single task situation (Carrier & Pashler, 1995; Logan & Gordon,
200 I; Pashler, 1994a). This indicates that the tasks have to com-
pete for a limited capacity in the human infonnation processing
system. While much effort has been spent to prove that this al-
ways leads to a strict serial processing (Pashler, 1994b; Pashler &
Johnston, 1989), recently, an increasing number of studies has
provided evidence that the processing capacity can be shared
between the tasks in a graded fashion (e.g., Hiibner & Lehle,
2007; Miller, Ulrich, & Rolke, in press; Tombu & Jolicoeur,
2005). This means that participants can allocate a certain amount
of capacity to one task while performing-with the remaining
capacity-another task in paralleL In other words, dual tasks
can be processed with either a more serial or a more paral1el
processing strategy.l If people are free to choose between differ-
ent degrees of paral1el processing, an important question is which
strategy they prefer and for what reason.
According to most dual-task theories, a serial strategy should
be preferred because it minimizes confusion and crosstalk be-
tween the tasks (e.g., Logan & Gordon, 200 I). Crosstalk in dual
tasks arises, for instance, on the level of response categories when
both tasks make use of the same responses. In this case, a con-
gruency effect is observable, i.e., if one task is associated with a
We thank Karen Grewen, Werner Sommer, and two anonymous re-
viewers for their valuable comments on an earlier version of this paper.
Address reprint requests to: Carola Lehle, U niversitiit Konstanz,
Fachbereich Psychologie, Fach D29, D-78457 Konstanz, Germany. E-
mail: carola.lehle@uni-konstanz.de
ITo decide whether parallel processing in dual tasks is possible during
all stages or whether some processing units constitute a structural bot-
tleneck is beyond the scope of the present study.
502
different response than the other, perfonnance is usually worse
than ifboth tasks are mapped to the same response. However, up
to now, it is unclear whether a serial strategy is really more op-
timal under these conditions. Some researchers assume ihat par-
ticipants still prefer a parallel processing strategy -at least at
short stimulus onset asynchronies (Miller et aL, in press).
In previous studies (Hiibner & Lehle, 2007; Lehle & Hiibner,
in press), we investigated processing strategies io,dual tasks with
the psychological refractory period paradigm (PRP; Welford,
1952), in which the stimulus ofthe second task appears before the
processing of the first task is completed. As a result, we found
that the participants showed a strong tendency to process those
tasks in paralleL At first sight, one might suppose that such a
strategy was chosen because parallel processing leads to a benefit
in perfonnance. However, the opposite was the case. Increased
parallel processing was accompanied by longer response times
and higher error rates. We thus came to the conclusion that the
participants were either mistaken to believe that parallel pro-
cessing is beneficial in dual tasks or that they preferred a parallel
strategy for a different reason. But what reason could that be?
Presumably, it is not overt perfonnance that the participants
intend to optimize by processing dual tasks in paralleL According
to Hockey (1997), analyses of task performance need to take into
account not only overt perfonnance, such as response times and
error rates, but also trade-offs among the participant's goals and
strategies, and the amount of mental effort that is needed to
achieve the goals (e.g., Steinhauser, Maier, & Hiibner, 2007).
These considerations are based on Kahnernan's thesis that effec-
tiveness and efficiency of perfonnance should be differentiated:
"Effectiveness is a measure of the quality of perfonnance, while
efficiency is the relation between the quality ofperfonnance and
the effort invested in it" (Kahneman, 1973, p. 181).
First publ. in: Psychophysiology 46 (2009), 3, pp. 502–509
Konstanzer Online-Publikations-System (KOPS)
URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-84520
URL: http://kops.ub.uni-konstanz.de/volltexte/2009/8452

Page 2

Although many recent studies (e.g., Ruthruff, Pashler, &
Hazeltine, 2003; Ruthruff, Pashler, & Klaassen, 2001; Tombu &
Jolicoeur, 2004, 2005) demonstrated that processing two tasks
concurrently means that both tasks have to access limited re-
sources (i.e., the "central capacity" or the "central bottleneck"),
and, furthermore, that parallel processing might lead to increased
costs in overt performance, none of these studies also examined
mental effort. Accordingly, it is entirely unknown so far which
role mental effort plays in the motivation to prefer either a serial
or a more parallel processing strategy in dual tasks.
A strategy of serial processing might be more effortful in dual
tasks with the PRP paradigm because it requires inhibiting the
processing of the second stimulus for a relatively short time dur-
ing the first task and then resuming it for the second task. That
processing of previously inhibited stimuli is costly has been
shown in studies of negative priming (e.g., Tipper, 1985; Tipper
& Cranston, 1985) and inhibition of return (e.g., Posner & Co-
hen, 1984; Tipper, Weaver, Jerreat, & Burak, 1994). By copro-
cessing the stimuli from both tasks right from the beginning of a
trial, participants can avoid the effortful procedure of first in-
hibiting and then resuming the processing of the other stimulus.
In this case the strategy of parallel processing would be a com-
promise between optimizing the performance and minimizing the
processing effort-thus increasing the efficiency.
Whether parallel processing is indeed more efficient in dual
tasks than serial processing was investigated in the present study.
Because previous results have shown that participants are able to
vary the degree of parallel processing in a graded fashion according
to specific instructions (Lehle & Hubner, in press; for a review see
Navon & Gopher, 1979), we also used this method in the present
study. That is, participants were instructed to process dual tasks
either in !l serial or in a more parallel mode. Overt performance
was analyzed by measuring response times and error rates. How-
ever, additional measures were needed to quantify the processing
effort in the serial compared to the parallel instruction condition.
To obtain this objective, we asked the participants to rate the
amount of mental effort they experienced during conducting the
tasks in the serial compared to the parallel mode. If a parallel
processing strategy is consciously applied as a trade-off between
mental effort on the one side and overt performance on the other,
participants should be able to estimate the degree of effort re-
quired in the different conditions (Naccache, Dehaene, Cohen,
Habert, Guichart-Gomez, Galanaud, & WilIer, 2005). If this is
not the case, then this would suggest that the compromise be-
tween effort and performance is either not consciously accessible
or that other reasons might be responsible for the tendency to
process the tasks in parallel.
In addition, we also used psychophysiological correlates
of mental effort. It has frequently been shown that processes
paralleling the experience of mental effort involve increased
autonomic activation (see Beatty, 1982; Cacioppo, Tassinary, &
Berntson, 2000; Sourkes, 2006 for reviews). Specifically, it has
been demonstrated that mental effort is associated with increased
heart rate (e.g., Boutcher & Boutcher, 2006; Brown, Szabo, &
Seraganian, 1988; Carroll & Turner, 1986; Steptoe, Moses, Ma-
thews, & Edwards, 1990) and with increased electrodermal activ-
ity (e.g., Collet, Petit, Priez, & Dittmar, 2005; Kahneman, 1973;
Lang, Bradley, & Cuthbert, 1990; Naccache et al., 2005). There-
fore, we recorded the electrocardiogram and the electrodermal
activity of the participants while they performed the dual tasks.
We predicted that, if serial processing is more effortful com-
pared to parallel processing in dual tasks, heart rate and skin
503
conductance should be higher under the serial than under the
parallel instruction condition.
Method
Participants
Twenty-eight participants (21 female, 7 male) between 19 and 38
years of age (M = 24 years) with normal or corrected-to-normal
vision participated in the study. Participants were recruited at the
Universitat Konstanz and were paid 5€ per hour. Whereas elec-
trodermal activity was measured in all participants, heart rate
was, due to a temporary failure of the recording equipment,
recorded only in a sub-sample of 18 participants.
Apparatus
The stimuli were presented on a 21"-monitor connected to a
personal computer (PC) with a resolution of 1280 x 768 pixels,
and a refresh rate of 85 Hz. The software controlling stimulus
presentation and registration of the behavioral responses was
programmed in C++ (Microsoft Inc.). Responses had to be
given by pressing either the "shift left" or the "shift right" button
on a standard PC keyboard.
Stimuli
The stimulus set consisted of eight numerals (1,2,3,4,6,7,8, and
9). The height of the stimuli subtended a visual angle of 2° at a
viewing distance of 72 cm, and their width was about 1.36°,
depending on the specific digit. The stimulus (S!) for the first task
was presented at the center of the screen, whereas the stimulus
(S2) for the second task consisted of two copies of a numeral,
which were presented left and right of SI at an eccentricity of
1.18°. Altogether three digits appeared on each trial with S! at the
center position. SI and S2 were always different.
Procedure
The task for the participants was to judge the parity (odd, even)
of the stimuli. For the first and the second task, one of two
buttons had to be pressed with the right hand. For an "even"
number, participants had to push the left button with the index
finger, for an "odd" number the right button with the middle
finger, respectively. Each trial started with the appearance of a
fixation cross for 400 ms. After the presentation of a blank screen
of 600 ms duration, both S[ and S2 appeared in white color on a
black background.
S! and S2 were congruent on half of the trials, i.e., had the
same parity, and were incongruent on the other half, i.e., had the
opposite parity. Participants always had to respond to SI first
and subsequently to S2. The stimuli remained on the screen until
the participants' responses had occurred. Six seconds after the
last response, the fixation cross for the next trial appeared. This
long intertrial-interval was necessary to reliably assign skin con-
ductance changes to individual stimuli. Trials with a wrong an-
swer in either one or both responses were categorized as errors.
Participants received specific instructions on how they had to
allocate their capacity within a block. There were two block
types: For one type, the participants were instructed to allocate
their capacity only to SI first and to ignore S2. S2 processing
should not start before the first response had been selected. For
the other block type, the participants were instructed to distribute
their capacity also to S2 from the beginning of a trial.

Page 3

504
Four successive blocks of one instruction type alternated with
four blocks of the other type, whereas the instruction order was
balanced across participants. Altogether there were 10 blocks
with 32 trials each. Accordingly, participants performed 80 trials
under each of the four experimental conditions (serial(congruent,
serial/incongruent, parallel/congruent, parallel/incongruent).
Before the specific instructions were given, there was a training
block to familiarize the participants with the basic task. The
whole experimental session took about 90 minutes.
Rating of General Procedure and Effort
Participants had to fill in a questionnaire subsequent to the
experiment, which we had constructed for the present purpose.
One part of the questionnaire consisted of general questions
about the procedure such as enjoyment, tiredness, and possible
artifacts during the experiment (e.g., changing the seating posi-
tion or speaking loudly during the measures). In the other part,
participants had to rate the level of effort they experienced during
the experiment retrospectively and separately for the serial and
the parallel condition. This part was titled by the question "How
effortful do you judge the different conditions?" The answers
were formulated as follows: "I found the serial condition ...
"-respectively-"I found the parallel condition ... "and
choices could be made on a Likertscale ranging from 1 = "very
little effortful" to 6 = "extremely effortful."
Psychophysiological Recording
The electrocardiogram was recorded at a sampling rate of200 Hz
using two Ag/ AgCL electrodes placed on the left and right ven-
tral forearm. The electrodermal activity was recorded at a sam-
pling rate of 200 Hz using 30 mm2 unpolarizable Ag/AgCl
electrodes placed on the thenar and hypothenar surfaces of the
participant's left hand. The amplifier (Biopac, GSRIOOC, Bio-
pac Systems, Inc., Goleta, CA) used a constant voltage of 0.5 V
DC. Prior to having the electrodes attached, participants were
requested to wash their hands; subsequently, electrode sites for
the measurement of the electrocardiogram and the electrodermal
activity were prepared by cleaning the skin with ethyl alcohol
(70%). The electrodes were filled with an isotonic conductive gel
(Biopac, Gel 101) to improve sensor-skin contact. The recording
took place in a quiet and dimly lit chamber. Participants were
requested to sit quietly during the experimental blocks.
Psychophysiological Data Analysis
Unless noted otherwise, data pre-processing and analyses were
computed using MatLab (The Mathworks, Inc., Natick, MA).
Electrocardiogram. R-peaks were detected off-line with an
accuracy of 5 ms using AcqKnowledge 3.7.3 (Biopac). A con-
tinuous heart rate (HR) was obtained by transforming inter-beat
intervals into a continuous signal (cf. Koers, Mulder, & van der
Veen, 1999). This served as the base for calculating tonic and
phasic HR measures. Tonic HR was defined as the average HR in
a 2000 ms time window directly preceding the stimulus onset.
In this way, tonic HR should be independent from phasic HR
changes induced by the stimulus. To analyze phasic HR changes,
we averaged across segments from 0 ms to 3000 ms following
stimulus onset. These segments were corrected by a baseline
which corresponded to the mean HR in the 1000 ms pre-stimulus
time window. For the statistical analysis, we determined the
amplitude and the latency of the minimum HR within this time
window for each trial. Trials with response errors and artifacts
were excluded from further analyses. A segment was regarded as
being contaminated by an artifact if the standard deviation of
HR exceeded a criterion that was determined separately for each
participant by means of visually inspecting the distribution of
standard deviations across trials. The proportions of trials con-
taminated by artifacts under the four conditions were 3.12%
(congruent/serial), 3.03% (incongruent/serial), 2.44% (congru-
ent/parallel), and 2.40% (incongruent/parallel). The resulting
trial numbers are shown in Table 1.
Electrodermal Activity. As a measure of electrodermal
activity, we computed event-related skin conductance responses
(SCR). A signal change of 0.2 !!Siemens occurring within 3 s after
the stimulus was classified as an event-related SCR. For each
condition, the relative frequency of trials containing an SCR as
well as the mean amplitude of these SCRs was derived. Again,
trials with errors or artifacts were excluded from further analyses.
The data of one participant had to be excluded due to an
extremely high number of artifacts. Apart from that, only few
segments were excluded based on visual inspection (less than 1 %
of trials). The resulting trial numbers are provided in Table 1.
Table 1. Mean Response Times ( RTs) in ms and Error Rates in % of First and Second Responses as well as Mean Trial Numbers for the Two
Subsamples Used in the Analysis of Electrodermal Activity (EDA) and Heart Rate (HR) Data
Response 1
Response 2
Mean number of trials
in HR/EDA analysis
RT
% error RT
% error
HR sample (n = 18)
Serial/congruent
Serial/incongruent
Parallel/congruent
Paralleljincongruent
EDA sample (n = 27)
Serialjcongruent
Serial/incongruent
Parallel/congruent
Parallel/incongruent
711
785
777
954
1.7
2.8
1.0
4.0
1085
1140
1017
1178
3.3
2.2
3.2
4.6
61.6
68.1
68.6
64.7
727
788
774
925
1.8
3.3
1.2
6.3
1083
1118
1021
1155
4.0
2.9
3.0
6.8
67.4
73.6
72.0
67.6
Note: RT = response time (in ms), HR = heart rate, EDA = electrodermal activity.

Page 4

505
First fespoose
I ~ ~ e o n
------>-IOC,
'_t:0fl.
c::Jine.
Seeond response
r--- 1
t----.
"-----
,
~
1300
A
e
1000
*
-1
0;-
g
Cl
.,
(i) 1200
,§.
. ~ 1100
900
//
- /
1/
...,.---
"-
~
aoo
----1
l'
J.
~ 1000
c
0
CL
rA
0
tt::
"""--.
~
m
8 ~
6 ~
4 ~
2 it
o~
0
0-
al
<l>
tt::
700
S :::;
S ~
4 ..
:{ .,
....
900
f'l
600
500 I ....

a
800
....
I
11) '*

serial
parallel -
serial
parallel
Figure 1. The behavioral data for the first (A) and for the second response (B). "Con." and "inc." denote the congruent and
incongruent condition, respectively.
Results
Behavioral Data
Response times as well as error rates for the first and the second
response were entered into a two-way ANOV A with repeated
measurement on the variables Instruction (serial, parallel) and
Congruency (congruent, incongruent). Response times were an-
alyzed only for those trials on which both responses were correct.
In the following, we report the results for the whole sample.
However, the results were virtually identical when we considered
only participants used for analyzing HR data and EDA data (see
Table 1).
Response times of the first response showed significant main
effects ofInstruction, F(I,27) = 9.10, 112 = .25,p<.01, and Con-
gruency,
F(I,27) = 26.6,
and Congruency also interacted significantly, F(I,27) = 6.72,
112 = .20, p < .05. Figure lA reveals that, with a parallel instruc-
tion, responses were generally slower and the effect of Congru-
ency was strongly increased. A similar pattern emerged for the
error rates of the first response. The main effect of instruction,
i.e., a trend of more errors with the parallel instruction, just failed
from being significant F(I,27) = 3.33, 112 = .11, p = .079. Fur-
thermore, there was a significant main effect of Congruency,
F(I,27) = 23.4, 112 = .47, p< .001, and an interaction between
Instruction and Congruency, F(I,27) = 9.79, 112 = .27, p< .Ol.
Again, more errors and an increased effect of Congruency were
evident with a parallel instruction.
The analysis of the response times of the second response
revealed only a significant main effect of Congruency, F(I,
27) = 16.0,112 = .37, p< .001, and an interaction of Congruency
and Instruction, F(I,27) = 6.48, 112 = .19, p< .05. Figure IB
shows that this results from an increased effect of Congruency
with a parallel instruction. The error rates for the second re-
sponse showed a significant main effect of Instruction,
F(I,27) = 5.36, 112 = .17,p< .05, and a significant interaction be-
tween Instruction and Congruency, F(1,27) = 13.0, 112 = .33,
p< .01. Again, this mainly reflects the increased effect of
Congruency with a parallel instruction.
112 = .50,
p<.OOl.
Instruction
Effort-Rating Data
The data from the effort rating in the questionnaire were ana-
lyzed by computing the mean values for the serial and the parallel
condition. On average, participants experienced more effort
under the serial (M = 3.9) than under the parallel instruction
condition (M = 2.1). A test revealed this difference as highly
significant, t(27) = 5.27, 112 = .51, p<.OO1.
Electrocardiogram
Figure 2 shows the mean HR in a time window of 3000 ms before
and after stimulus onset. As can be seen, there is a rather constant
difference in HR between the serial and the parallel condition.
Because there seems to be no phasic HR change in anticipa-
tion of the stimulus that differs between parallel and serial con-
ditions, we computed the tonic HR during an interval of 2000 ms
preceding stimulus onset. The values of tonic, HR were then
entered into a one-way ANOV A with repeated measurement on
the variable Instruction (parallel, serial). Because tonic HR was
defined in a pre-stimulus interval and thus could not be affected
by the type of stimulus, Congruency was not considered in this
analysis. As illustrated in Figure 3A, the tonic HR was higher in
blocks with a serial instruction than in blocks with a parallel
instruction, F(1,17) = 4.55,112 = .21, p< .05.
The averaged phasic HR change is illustrated in Figure 4,
showing average wave forms for each combination of the vari-
ables Instruction (serial, parallel) and Congruency (congruent,
incongruent). As evident, stimulus presentation is followed by a
typical HR deceleration. To analyze this deceleration statisti-
cally, mean amplitudes and latency of minimum HR were
7a
76
f e 14
er::
:I:
u 12.,
C
I)
I-
70
ea
-serial C<m.
-- serial me.
-p~(X)o.
**' _parallel ine.
-3000
-2.00()
-1000
(l
S
time {ms)
1000
2000
3000
Figure 2. The time course of the tonic heart rate (HR) from 3000 ms
before stimnlus onset (S) until 3000 ms after stimulus onset. "Con." and
"inc." denote the congruent and incongruent condition, respectively.

Page 5

506
1500
78 A
i_inc.
J B
-
S
:a 76
-
~
%
i
-.
" 72
f
~1~
E
-
~
C
•
,~
- ,-£"'-'
! I
i
Il
,,11)
l:l
16"0
"-
5 c ~
.lA L.H
Q
"0 1100
~
'" 1000 III
~ 74 . 'I
'Iii 1""" 1
. - ~
I
I I.
s: ::.r:
T T

• 10
m-
j;
3 "tI
serial parallel
serial parallel
Figure 3. Tonic changes of the heart rate (HR) in the pre-stimulus interval of2000 ms (A); latency and amplitude of the deceleration
after stimulus onset (B). "Con." and "inc." denote the congruent and incongruent condition, respectively; "dec." refers to the
deceleration of the HR.
entered into two-way ANOV As with repeated measurement on
the variables Instruction and Congruency. For the amplitudes,
we obtained a significant main effect of Instruction, F(l,
17) = 5.80, r/ = .25, P < .05. As shown in Figure 3B, the ampli-
tude of the HR deceleration was stronger in blocks with a serial
instruction. Moreover, there was a tendency in blocks with a
parallel instruction that the deceleration amplitude was higher on
incongruent compared to congruent trials? For the latency of
HR minima, we obtained a significant main effect of Congru-
ency, F(l,17) = 7.66,1'/2 = .31, p<.05, and a marginally signifi-
cant interaction between Congruency and Instruction, F(I,
17) = 3.30, 1'/2 = .16, p = .087. The minimum of the HR decel-
eration is influenced by Congruency more strongly in the parallel
condition (see Figure 3B).
Electrodermal Activity
The analysis for SCR frequency revealed only a significant main
effect of Congruency, F(l,26) = 4.29,1'/2 = .14, p<.05. Figure 5
shows that this frequency was increased on incongruent trials.
Although a congruency effect was present only in the parallel
condition, the interaction between Instruction and Congruency
did not reach significance [F(l, 26) = 2.45, 1'/2 = .09, p = .129].
No significant effect on the SCR amplitude was detected (Fs< I).
The mean amplitudes in the four conditions were 0.60 ~ S i e m e n s
(congruent/serial), 0.61
~ S i e m e n s (congruent/parallel), and 0.62 ~ S i e m e n s (incongruent/
parallel).
~ S i e m e n s (incongruent/serial), 0.63
DISCUSSION
The present study was conducted to investigate whether serial
processing is more effortfnl compared to parallel processing in
PRP-like dual tasks in which S2 is presented before processing of
SI is completed. Different degrees of parallel processing were
realized by instructing the participants to perform dual tasks with
either a serial or with a parallel processing strategy. Because
response times and error rates cannot indicate the amount of
effort invested in a task (cr. Hockey, 1997), our participants rated
the degree of effort they experienced in the two conditions. If
2Please note that the differences in HR minimum in our conditions are
only weakly reflected in Figure 4. Since the minimum is at a different
point in time on each trial, averaging blurs the differences in the minimum
amplitudes.
participants intentionally chose a parallel processing strategy in
dual tasks because it is accompanied by less mental effort, they
shonld be able to report on it. Moreover, psychophysiological
measures were included in the present study to further compare
the serial and the parallel instruction condition with respect to
correlates of mental effort. Here, HR and event-related SCR
were used because these measures are known to be increased
with an increased mental effort (Boutcher & Boutcher, 2006;
Naccache et aI., 2005).
First of all, the behavioral data replicated our previous results
(Hubner & Lehle, 2007; Lehle & Hubner, in press). Response
times and error rates were higher overall if they were performed
with a parallel compared to a serial processing strategy. This
confirms that parallel processing does not lead to a benefit in
performance, but to increased costs. In the present study, of
course, the participants were instructed to adopt either a serial or
a parallel processing strategy. However, in our previous studies
including dual task conditions without specific instructions
(Hubner & Lehle, 2007; Lehle & Hubner, in press), participants
clearly preferred a parallel processing strategy and by that
accepted the costs in performance. In view of these results, we
3
i' 2
0.. e. .,
tile
fa
.c
()
Ill::
:E;
j -'1
-2
1
I)
-_. serial con.
_ - serial 'ne.
-parallel co!!.
§; parallel ire.
-
-3000 -2000 -1000
()
S
time {ms}
1000 2000
3000
Figure 4. The time course of the phasic heart rate (HR) change from 3000
ms before stimulus onset (S) until 3000 ms after stimulus onset. "Con."
and "inc." denote the congruent and incongruent condition, respectively.

Page 6

34~-----------------4
32
~ 30
~
~
28
!
i"26
.::
Ill:: 24
~
22
20
serial
parallel
Figure 5. The mean frequencies of skin conductance responses (SCR)
in the different conditions. "Con." and "inc." denote the congruent and
incongruent condition, respectively.
hypothesized that the motivation to process dual tasks in parallel
emerges from the intention to minimize mental effort.
The main question of the present study therefore was whether
the perfonnance costs with a parallel strategy are compensated
by a reduced effort. Considering the effort rating data, this was
indeed the case. In accordance with our hypothesis, participants
rated the serial processing strategy as more effortful compared to
the parallel processing strategy. Furthennore, the psychophys-
iological data seem to parallel this result. The tonic HR was
considerably increased during the condition with a serial com-
pared to a parallel processing strategy. Several studies so far have
shown that in task conditions requiring more mental effort, tonic
HR is increased compared to control conditions (e.g. Boutcher &
Boutcher, 2006; Brown et aI., 1988; Hamer, Boutcher, & Boutc-
her, 2003; Sammer, 1998; Steptoe et aI., 1990). Apart from that,
there was a phasic HR deceleration after stimulus onset in each
condition. The amplitude as well as the latency of the HR de-
celeration was higher in blocks with a serial compared to blocks
with a parallel instruction. Transient cardiac deceleration typi-
cally occurs with stimulus intake and reflects the engagement
of attention to external stimuli (e.g., Jennings & van der Molen
2002; Jennings, van der Molen, Brock, & Somsen, 1991).
The question emerges whether our tonic HR measure is
affected by phasic HR changes. Indeed, phasic effects evident
with stimulus presentation might already start in the pre-stimulus
interval, thus reflecting task preparation (cf. De Jong, 1995). For
the present results, however, inspection of the time course in the
pre-stimulus interval revealed a rather constant difference be-
tween the serial and the parallel instruction condition. This in-
dicates that our tonic HR measure-which was defined as mean
HR in a pre-stimulus interval-was not substantially influenced
by phasic HR effects emerging during task preparation. More-
over, the stimulus-locked phasic effects observed after stimulus
presentation even seem to counteract the differences obtained for
the tonic HR: There was a stronger decrease in the phasic HR in
the serial compared to the parallel condition, whereas the tonic
HR was increased under the serial compared to the parallel
instruction. Accordingly, phasic effects cannot account for the
increased tonic HR in the serial condition.
The electrodermal activity appeared to be less influenced by
the processing strategy, i.e., there was no significant effect of the
507
instruction condition on the frequency or on the amplitude of the
SCRs. There was only a significant main effect of congruency on
SCR frequency, i.e., more SCRs occurred on incongruent than
on congruent trials. However, a clear trend in the data also in-
dicated that most SCRs occurred on incongruent trials with a
parallel processing strategy. Accordingly, the pattern of SCRs
seems to reflect the amount of response conflict induced by the
stimulus. This outcome corroborates previous results showing
that the electrodennal reactivity is particularly sensitive to con-
flict on stimulus level (Naccache et aI., 2005).
It has been frequently observed in the literature that HR and
electrodennal reactivity are dissociable. This finding is usually
explained by differences in autonomic response patterns (see
Lacey, 1967) or by the hypothesis that cardiovascular and elec-
trodennal adjustments are linked to different behavioral systems
(Amodio, Master, Yee, & Taylor, 2007; Fowles, 1988; Gray,
1987). Furthennore, the HR seems to be particularly sensitive
for effort mobilization, whereas the electrodennal system is more
influenced by conflict or aversive feedback (e.g., Tranel, 1983).
The data of the present study further support this interpretation.
Taken together, the results of the present study are compatible
with the hypothesis that serial dual-task processing is more
effortful compared to parallel processing. But how can the in-
creased mental effort be explained? Shielding one task from the
other to minimize crosstalk and to achieve a serial processing
strategy requires a strict focusing of attention. This means that,
during the first task, the processing of the second stimulus has to
be inhibited for a relatively short time interval and then to be
resumed for the processing of the second task. It can be assumed
that this requires a high amount of cognitive control (e.g., Posner
& Cohen, 1984; Tipper et aI., 1994), presumably accompanied by
increased mental effort.
One could object that, in the present study, only dual tasks
were used where the task set was identical for the first and the
second task. In this case, the crosstalk between the tasks and thus
also the effort in shielding one task from the other could be
particularly high. In a previous study including also dual tasks
with different task sets, overt perfonnance was less influenced by
the other task, but the overall pattern of results was very similar
(Lehle & Hubner, in press). Participants also showed a tendency
to strategic parallel processing, although it was more costly
compared to serial processing.
The exact physiological mechanisms in the brain that lead to
the characteristic autonomic reactions associated with mental
effort are fairly unknown so far. However, mesio-frontal struc-
tures, including the anterior cingulate cortex, likely play a role in
the generation of somatic signals in response to mental effort
(e.g., Critchley, Mathias, Josephs, O'Doherty, Zanini, & Dewar,
2003; Tranel & Damasio, 1994). Moreover, the feeling of effort
that the participants experienced could also be related to signals
generated by the anterior cingulate cortex. It has been reported
that a patient with a vast lesion in these structures had lost almost
entirely the ability to experience and report a feeling of mental
effort, although cognitive abilities and overt perfonnance were
preserved (Naccache et al., 2005).
To conclude, the present results indicate that not only
overt perfonnance but also the amount of mental effort should
be taken into account to judge the efficiency of processing strat-
egies. This aspect is largely neglected in current dual task
accounts. Considering only overt performance, many theories
come to the conclusion that serial processing is advantageous in
dual tasks (e.g., Logan & Gordon, 2001; Navon & Miller, 1987;

Page 7

508
Navon & Miller, 2002). In contrast, the present data indicate that
different processing strategies in dual tasks also produce different
levels of experienced effort. This is likely to play an important
role in the choice of one strategy (i.e., parallel processing) over
the other.
When participants spontaneously adopt a parallel processing
strategy (Hubner & Lehle, 2007; Lehle & Hubner, in press), they
seem to optimize their behavior not only with respect to overt
performance, but also with respect to mental effort. Only if this
aspect is also taken into account, an adequate consideration of
serial compared to parallel processing strategies in dual tasks
seems possible. Future research should further investigate the
role of mental effort in strategic considerations during cognitive
tasks. Also, additional psychophysiological data (e.g., event-re-
lated potentials) would be desirable to investigate these effects in
more detail.
REFERENCES
Amodio, D. M., Master, S. L., Yee, C. M., & Taylor, S. E. (2007).
Neurocognitive components of the behavioral inhibition and activa-
tion systems: Implications for theories of self-regnlation. Psycho-
physiology, 45, 11-19.
Beatty, I. (1982). Task-evoked pupillary responses, processing load, and the
structure of processing resources. Psychological Bulletin, 91,276-292.
Boutcher, Y. N., & Boutcher, S. H. (2006). Cardiovascular response to
Stroop: Effect of verbal response and task difficulty. Biological Psy-
chology, 73,235-241.
Brown, T. G., Szabo, A., & Seraganian, P. (1988). Physical versus psy-
chological determinants of heart rate reactivity to mental arithmetic.
Psychophysiology, 25, 532-537.
Cacioppo, I., Tassinary, L., & Berntson, G. (2000). Handbook ofpsy-
chophysiology. Cambridge: University Press.
Carrier, L. M., & Pashler, H. (1995). Attentionallimits in memory re-
trieval. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 21, 1339-1348.
Carroll, D., & Turner, I. R. (1986). Heart rate and oxygen consumption
during active psychological challenge: The effects oflevel of difficulty.
Psychophysiology, 23, 174-181.
Collet, C., Petit, c., Priez, A., & Dittmar, A. (2005). Stroop color-word
test, arousal, electrodermal activity and performance in a critical
driving situation. Biological Psychology, 69, 195-203.
Critchley, H. D., Mathias, C. I., Iosephs, 0., O'Doherty, I., Zanini, S.,&
Dewar, B. K. (2003). Human cingnlatecortex and autonomic control:
Converging neuroimaging and clinical evidence. Brain, 126, 2139-
2152.
De long, R. (1995). The role of preparation in overlapping-task
performance. Quarterly Journal of Experimental Psychology [Aj,
48,2-25.
Fowles, D. (1988). Psychophysiological and psychopathology: A moti-
vational approach. Psychophysiology, 25, 373-391.
Gray, I. A. (1987). The psychology offear and stress. Cambridge, UK:
Cambridge University Press.
Hamer, M., Boutcher, Y. H., & Boutcher, S. H. (2003). The role of
cardiopulmonary baroreceptors during the forearm vasodilatation
response to mental stress in humans. Psychophysiology, 40, 249-253.
Hockey, G. R. I. (1997). Compensatory control in the regulation of
human performance under stress and high workload: A cognitive-
energetical framework. Biological Psychology, 45, 73-93.
Hubner, R., & Lehle, C. (2007). Strategies of flanker coprocessing in
single and dual tasks. Journal of Experimental Psychology: Human
Perception and Performance, 33,103-123.
Iennings, I. R., & van der Molen, M. W. (2002). Cardiac timing and the
central regulation of action. Psychological Research, 66, 337-349.
Iennings, I. R., van der Molen, M. W., Brock, K., & Somsen, R. I. M.
(1991). Response inhibition initiates cardiac deceleration: Evidence
from a sensory-motor compatibility paradigm. Psychophysiology, 28,
72-85.
Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NI: Pren-
ticeHall.
Koers, G., Mulder, L. I. M., & van der Veen, F. M. (1999). The com-
putation of evoked heart rate and blood pressure. Journal of Psycho-
phYSiology, 13, 83-91.
Lacey, I. I. (1967). Somatic response patterning and stress: Some revi-
sions of activation theory. In M. H. Appley & R. Trumbull (Eds.),
Psychological stress: Issues in research (pp. 14-42). New York: Ap-
pleton-Century-Crofts.
Lang, P. I., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion,
attention, and the startle reflex. Psychological Review, 97, 377-395.
Lehle, c., & Hubner, R. (in press). Strategic capacity sharing between
dual tasks: Evidence from tasks with the same and with different task
sets. Psychological Research.
Logan, G. D., & Gordon, R. D. (2001). Executive control of
visual attention in dual-task situations. Psychological Review, 108,
393-434.
Miller, I., Ulrich, R., & Rolke, B. (in press). On the optimality of serial
and parallel processing in the psychological refractory period para-
digm: Effects of the distribution of stimulus onset asynchronies.
Cognitive Psychology.
Naccache, L., Dehaene, S., Cohen, L., Habert, M.-O., Guichart-Gomez,
E., Galanaud, D., & Willer, I.-C. (2005). Effortless control: Executive
attention and conscious feeling of mental effort are dissociable.
Neuropsychologia,43,1318-1328.
Navon, D., & Gopher, D. (1979). On the economy of the human-
processing system. Psychological Review, 86,214-255.
Navon, D., & Miller, I. (1987). Role of outcome conflict in·dual-task
interference. Journal of Experimental Psychology: Human Perception
and Performance, 13,435-448.
Navon, D., & Miller, I. (2002). Queuing or sharing? A critical
evaluation of the single-bottleneck notion. Cognitive Psychology,
44, 193-251.
Pashler, H. (1994a). Dual-task interference in simple tasks: Data and
theory. Psychological Bulletin, 116, 220-244.
Pashler, H. (1994b). Graded capacity-sharing in dual-task interference?
Journal of Experimental Psychology: Human Perception and Perfor-
mance, 20, 330-342.
Pashler, H., & Iohnston, I. C. (1989). Chronometric evidence for central
postponement in temporally overlapping tasks. Quarterly Journal
of Experimental Psychology: Human Experimental Psychology, 41,
19-45.
Posner, M. I., & Cohen, Y. (1984). Components of visual orienting. In H.
Bouma & D. G. Bouwhnis (Eds.), Attention and performance x:
Control of language processes (Vol. 32, pp. 531-556). London: Law-
rence Erlbaum Associates.
Ruthruff, E., Pashler, H. E., & Hazeltine, E. (2003). Dual-task interfer-
ence with equal task emphasis: Graded capacity sharing or central
postponement? Perception & Psychophysics, 65, 801-816.
Ruthruff, E., Pashler, H. E., & Klaassen, A. (2001). Processing bottle-
necks in dual-task performance: Structural limitation or strategic
postponement? Psychonomic Bulletin & Review, 8, 73-80.
Sammer, G. (1998). Heart period variability and respiratory changes
associated with physical and mental load: Non-linear analysis. Ergo-
nomics, 41, 746-755.
Sourkes, T. L. (2006). On the energy cost of mental effort. Journal of the
History of Neurosciences, 15, 31-47.
Steinhauser, M., Maier, M., & Hubner, R. (2007). Cognitive control
under stress. Psychological Science, 18, 540-545.
Steptoe, A., Moses, I., Mathews, A., & Edwards, S. (1990). Aerobic
fitness, physical activity and psychophysiological reactions to mental
tasks. Psychophysiology, 27, 264-274.
Tipper, S. P. (1985). The negative priming effect: Inhibitory priming by
ignored objects. The Quarterly Journal of Experimental Psychology A:
Human Experimental Psychology, 37A, 571-590.
Tipper, S. P., & Cranston, M. (1985). Selective attention and priming:
Inhibitory and facilitory effects of ignored primes. Quarterly Journal
of Experimental Psychology A: Human Experimental Psychology,
37A,591-611.
Tipper, S. P., Weaver, B., Ierreat, L. M., & Burak, A. L. (1994). Object-
based and environment-based inhibition of return of visual attention.

Page 8

Journal of Experimental Psychology: Human Perception and Perfor-
mance, 20, 478-499.
Tombu, M., & Jolicoeur, P. (2004). Virtually no evidence for virtually
perfect time-sharing. Journal of Experimental Psychology: Human
Perception and Performance, 30, 795-810.
Tombu, M., & Jolicoeur, P. (2005). Testing the predictions of the central
capacity sharing model. Journal of Experimental Psychology: Human,
Perception and Performance, 31,790-802.
Tranel, D. (1983). The effects on monetary incentive and frustrative
nonreward on heart rate and electrodermal activity. Psychophysiol-
ogy, 20, 652-{)57.
509
Tranel, D., & Damasio, H. (1994). Neuroanatomical correlates of elec-
trodermal skin conductance responses. Psychophysiology, 31, 427-
438.
Welford, A. T. (1952). The 'psychological refractory period' and the
timing of high-speed performance: A review and a theory. British
Journal of Psychology, 43, 2-19.