R

Dp

AI

h

•••

ARRAA

KDPRPDR

1

lidpcttsti

oB

0h

Behavioural Brain Research 250 (2013) 304– 307

Contents lists available at SciVerse ScienceDirect

Behavioural Brain Research

j ourna l h o mepa ge: www.elsev ier .com/ locate /bbr

esearch report

ual-task performance is differentially modulated by rewards andunishments

li Yildiz, Witold Chmielewski, Christian Beste ∗

nstitute for Cognitive Neuroscience, Department of Biopsychology, Ruhr-University Bochum, Germany

i g h l i g h t s

Reward anticipation decreases dual-task performance.Punishment anticipation increases dual-task performance.Dual-task performance is modulated by dopaminergic mechanisms.

a r t i c l e i n f o

rticle history:eceived 2 April 2013eceived in revised form 30 April 2013ccepted 6 May 2013vailable online 13 May 2013

eywords:ual-task

a b s t r a c t

Dual-task situations play a pivotal role in daily life and are subject to a research in cognitive psychol-ogy and neuroscience. From a neuroscience perspective, the response selection bottleneck may be partlyconstituted by the dopaminergic system. The dopaminergic system plays a pivotal role in reward andpunishment effects. In the current study we therefore investigated the effects of rewards and punish-ments as a potential modulator of dual-tasking processes. We examined dual-task performance in thepsychological refractory period (PRP) paradigm, where the task order was either predictable, or unpre-dictable. Three groups were tested; a punishment group (N = 14), a reward group (N = 18) and a control

sychological refractory period (PRP)ewardunishmentopamineesponse selection

group (N = 16). The results show that in the punishment condition, dual-task performance is increasedrelative to controls (i.e., faster RTs). In the reward condition performance decreased relative to con-trols. The effects observed were of moderate to high effect sizes. However, the effects were only evidentwhen task performance was unpredictable. These divergent effects of rewards and punishments on dual-tasking may be explained by the differential involvement of different dopamine receptors in rewards andpunishments, and their effects on the amount and flexibility of task goals in working memory.

. Introduction

“Multitasking” is challenge in daily life and requires a highevel of cognitive control. The underlying mechanisms have beennvestigated for a long time using dual-task paradigms. A classicalual-task paradigm is the “psychological refractory period” (PRP)aradigm [1]. In this paradigm, responses are required on two suc-essive tasks. The typical finding is that responses on the secondask are slower when this task was presented shortly after the firstask (=PRP effect) [2,3]. With increasing time between the tasks (i.e.,

timulus onset asynchrony, SOA), this effect vanishes [e.g. 4,5]. Evenhough the precise nature of the slowing of the response to task 2s still a matter of debate [for review 4], all theoretical accounts

∗ Corresponding author at: Institute for Cognitive Neuroscience, Departmentf Biopsychology, Ruhr-Universität Bochum, Universitätsstrasse 150, D-44780ochum, Germany. Tel.: +49 234 322 4630; fax: +49 234 321 4377.

E-mail addresses: christian.beste@rub.de, christian.beste@cityweb.de (C. Beste).

166-4328/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.bbr.2013.05.010

© 2013 Elsevier B.V. All rights reserved.

assume that the PRP occurs because response selection capacity islimited [2,3,6,7]. Factors that modulate processing capacity shouldhence also modulate dual-tasking abilities.

Several results suggest that processing capacity, as well asprocessing chrematistics of a network and response selection pro-cesses are modulated by the dopaminergic system. The dual-statetheory of the dopamine systems states that network characteris-tics change depending on whether a dopamine D1, or D2-receptorneural transmission dominates [8,9]. A highly active dopamine D2system has been shown to allow the establishment of multiple rep-resentations in prefrontal cortical networks and working memory,i.e., processing capacity of the network is increased [8,9]. This stateis more responsive and flexible, but also more interference-pronecompared to a network state dominated by high dopamine D1receptor turnover. In the dopamine D1 dominated network state

processing capacities are more restricted [8]. Modulations of thedopamine D1 and dopamine D2 receptor system may hence affectperformance in dual-tasking through their effects on processingcharacteristics in the neuronal network.

ain Re

obAvaoacpi‘b

Doltoias[crac[rfd

2

2

(T(htccU

2

(sbsbtltdstoatstIg2lFdea

A. Yildiz et al. / Behavioural Br

Interestingly, the dopamine D1 and D2 receptor system can notnly be differentially modulated by pharmacological interventions,ut also by factors related to rewards and punishments [10,11].

number of studies suggest that reward processing is mediatedia the dopamine D1-system, whereas the effects of punishmentsre likely mediated via the D2-system [10–14]. The expectancyf punishments in case of low dual-task performance may pro-ctively shift the mode of the network towards a ‘D2-state’. As aonsequence processing capacity may become larger and dual-taskerformance increases. Opposed to this, the expectancy of rewards

n case of good dual-task performance may pro-actively induce aD1-network state’. As a consequence, dual-task performance maye lower.

However, besides the considerations of the dopamine D1 and2 system in terms of processing capacity, these systems are alsof importance when considering that the PRP reflects a processingimitation at the response selection stage [2,3]. Response selec-ion processes have frequently been conceptualized as a propertyf basal ganglia-prefrontal networks [15–19]. Response selections thought to be constituted by the interaction of two pathways:

selection pathway dominated by dopamine D1 neurotransmis-ion and a control pathway dominated by the D2 receptor system20,21]. This underlines that besides modulations of processingapacity, dual-task performance may also be modulated becauseesponse selection processes are affected by modulating the D1nd D2 receptor systems. However, for the response selectionomponent the flexibility to switch between the tasks is critical22–24]. It is therefore possible that the degree as to whetherewards and punishments are able to modulate dual-task per-ormance further depends on predictability of required responsesuring dual-tasking.

. Materials and methods

.1. Participants

In this experiment N = 48 (29 women), healthy right-handed students23.63 ± 3.79 (SD) years of age) were recruited at the Ruhr-University of Bochum.hey were randomly assigned to three different conditions: the punishment groupN = 14), the reward group (N = 18) and the control group (N = 16). All participantsad normal or corrected-to-normal vision and normal hearing performance. All par-icipants enrolled into the study had a comparable socio-economic background andomparable monthly income. For the participation the students received financialompensation. The study was approved by the Ethics committee of the Ruhr-niversity of Bochum.

.2. Principle outline of the experimental paradigm

A PRP paradigm was used in this study. The two tasks used were an auditorytone task) and a visual task (letter task). For each task one of two task-specifictimuli was presented, which had to be identified by pressing a stimulus-specificutton: Two sine wave tones (a low 500 Hz tone and a high 1300 Hz tone) were pre-ented in the tone task. The participants were instructed to respond to the low toney pressing a stimulus-specific button with their left index finger and to the highone with their right index finger. In the letter task we used the white-colouredetters ‘H’ and ‘O’ on a dark blue screen (visual angle: 1.8◦ × 2.3◦). In this task forhe stimulus H the subjects have to press the according button with their left mid-le finger and for O the corresponding button with their right middle finger. Everytimulus applied a stimulus presentation duration amount of 200 ms. One trial con-ained both, the tone and the letter task, which were presented successively withne of the four different stimulus onset asynchrony (SOA; 16 ms, 133 ms, 500 msnd 100 ms). The trials started with the presentation of a central fixation cross athe screen. The first stimulus S1 (tone) was presented 1 s later. Then follows the pre-entation of the second stimulus S2 (letter) in a predefined SOA Participants wereold to respond in minimum time and with a maximum of accuracy to each stimulus.n addition there were instructed to place equal emphasis on both tasks and not toroup responses [25]. The time window for responding to a stimulus was limited to000 ms. Otherwise the trial was considered as a miss. After a missed trial the fol-

owing trial started within 1500 ms (randomly jittered between 500 and 2500 ms).or the RT data analysis across SOAs the data was screened for trials in which theifference in RT between task 1 and task 2 was 100 ms or less, to account for possibleffects of ‘response grouping’. The trials following a regular responded trial startedfter response–stimulus interval (RSI) with a jitter between 1000 and 4000 ms. For

search 250 (2013) 304– 307 305

the data collection and stimulus presentation we used the software ‘Presentation’(Neurobehavioral System Inc.).

2.3. Administration of the PRP task

The PRP task was administered in two different forms: form A and form B. Inform A the stimuli were presented in a fixed order (S1 = tone; S2 = letter). In formB, the order of the presentation for the first and the second task (T1 and T2) wasrandomized (S1 could be tone or letter). Subjects were always requested to respondfirst to the first stimulus appearing (irrespective of the task). This means that in formB the RT1 and RT2 comprise response to tones and letters. In form A RT1 refers toresponses on tones and RT2 refers to responses on letters. Each form (fixed, random)was administered two times in a counterbalanced order (either ABAB or BABA tocontrol block-sequence effects). ABAB and BABA blocks occurred equally frequentwithin each experimental group. One block of forms A and B consisted of 320 trials.Thus, the whole experiment consisted of 1280 trials.

2.4. Experimental groups

Reward and punishment manipulations were done using a between-subjectdesign. In a within-subject design carry-over effects may affect the results; e.g. it ispossible that rewards have a different effect, when reward manipulation is done atthe beginning of the experiment, or after a session where the subjects received pun-ishments. Counterbalancing may overcome these problems, but it is easier to run abetween-subject experimental design. The control group received a fixed amount of30 Euros, which was paid at the end of the experiment for participating in the studyindependent from their achievement (incorrect trials: wrong reaction or exceed-ing the time limit per reaction >600 ms). The participants in the punishment groupreceived an amount of 30 Euros at the beginning of the experiment. However theywere informed to be punished by 3 Euro cents for each incorrect trial and that theyreceive the amount of 30 Euros minus the sum of the value of each incorrect trialconducted throughout the experiment. In the reward group the participants wereinformed to receive a reward to the amount of 3 Euro cents for every correct trialwhich were also paid after the experiment as a sum (also summing up to ∼30 Euros).Hence, there was no amount of money allotted to the participants before the experi-ment. This group was told that their compensation for taking part in the experimenttotally depends on their performance.

2.5. Statistical analysis

Statistical analysis was performed using SPSS 20 (SPSS, Inc., 2009, Chicago, IL).Data analysis was conducted using mixed effects and univariate ANOVAs. To testthe normal distribution of the variables we calculated Kolmogorov–Smirnov test.All post hoc tests were Bonferroni-corrected. Greenhouse–Geisser correction wasused, if necessary. For the statistical analysis of the data, data were not groupedwith respect to the modality of the stimulus, but for its occurrence (i.e., first orsecond stimulus). Hence, data were pooled across the different modalities in therandom block. In the statistical analysis the modality of the T1 and T2 stimulus istherefore discarded in the random block.

3. Results

The net amount of money in the punishment group was 24.53(1.64). This equals 182.22 (55.23) punished error trials. In thereward group the net amount of money was 25.13 (1.55) and thusnot different from the punishment group (p > .4). The amount ofmoney lost did not differ across blocks (i.e., from A and form B)(p > .4). The net amount of money in the reward and punishmentgroup differed from the control group receiving a fixed amount of30 Euros (p < .001).

For task 1 (T1) the mixed effect ANOVA reveal the followingresults. There was a main effect block (F(1, 45) = 106.1; p < .001;�2 = .702), with reaction times (RTs) being shorter in the predictablecondition (529 ± 11) (form A), compared to the unpredictable con-dition (601 ± 11) (form B). Moreover, there was a main effectgroup (F(2, 45) = 3.32; p = .045; �2 = .129). Bonferroni corrected pairwais comparisons revealed only a difference between punishment(530 ± 18.6) and the reward group (594 ± 16.4) (p = 0.41), but nodifference between the control group (571 ± 17.4) and the punish-ment and the reward group (p > .4). All other main or interaction

effects were not significant (all F < 1.01; p > .3).

Considering the task 2 (T2) the mixed effects ANOVA showsa main effect of SOA (F(3, 45) = 397.9; p < .001; �2 = .965) withRTs for the SOA16 (667.4.9 ± 11.5); SOA133 (585 ± 11.2); SOA500

306 A. Yildiz et al. / Behavioural Brain Research 250 (2013) 304– 307

F r all SOa

((tRptp

waitttfg1gSt(ct(b

(g(mbf

ig. 1. Average reaction times (RTs ± SEM) in ms on task 1 (T1) and task 2 (T2) ovend the unpredictable (right).

462.8 ± 8.7) and SOA1000 (418.9 ± 7) all differing from each otherp < .001). Also, a main effect “form (predictable vs. unpredictableask order)” (F(1, 45) = 185.6; p < .001; �2 = .805) was evident, withTs for the predictable task order (498.2 ± 9) being shorter, com-ared to the unpredictable task order (568.9 ± 10). Importantly,here was an interaction “SOA × form × group” (F(6, 135) = 2.58;

= .021; �2 = .103) (refer Fig. 1).To explore this interaction further separated mixed ANOVAs

ere run for the predictable and unpredictable task form, sep-rately. For the predictable condition (form A), there was nonteraction “SOA × group” (F(6, 135) = 0.57; p > .5; �2 = .034), but forhe unpredictable block (block B) there was a significant interac-ion “SOA × group” (F(6, 135) = 7.91; p < .001; �2 = .26). To examinehe interaction between SOA and group for the unpredictable taskorm, univariate ANOVAs were conducted where the effects ofroup were examined for each SOA length separately. For the SOA000 condition there were no significant differences between theroups (F(2, 45) = 1.38; p = .261;�2 = .058). The same was true for theOA 500 condition (F(2, 45) = 2.56; p = .089; �2 = .102). Analysinghe SOA 133 condition, a significant group effect was evidentF(2, 45) = 7.016; p = .002; �2 = .238). The corresponding Bonferroni-orrected pairwise comparisons showed that RTs were shorter inhe punishment group (561 ± 23), compared to the control group643 ± 21) (p = .039; Cohen’s d = 1.06). No difference was evidentetween the control and the reward group (p = .873).

In the SOA 16 condition, there was a significant effect “group”F(2, 45) = 10.93; p < .001; �2 = .327). Compared to the controlroup (713.6 ± 57.8), RTs were slower in the reward condition

786 ± 101.7) (p = 0.34; Cohen’s d = 0.67) and faster in the punish-

ent group (646 ± 87) (p = .016; Cohen’s d = 0.93). However, it maye argued that effects of rewards and punishments are not specificor the RTs on task 2, since there was also a modulations of RTs in

As for the punishment, reward and the control conditions in the predictable (left)

task 1 responses (refer Fig. 1) suggesting for a general “speedingeffect” [26]. In order to account for a possible modulation of RT2responses by a general “speeding effect” we subjected the meanreaction time on the T1 stimulus as a covariate in the mixed effectsANOVA. This did not change the model and the interaction was stillsignificant (F(6, 135) = 2.22; p = .030; �2 = .101).

4. Discussion

In the current study we examined the effects of rewards andpunishments on dual-task performance. The results show thatrewards and punishments differentially modulate the dual-taskperformance in case of unpredictable task order. In case of pre-dictable task order no effects specific to rewards and punishmentswere evident. Effects of rewards and punishments were evidentfor the T1 task and the T2 task. However, the analyses show thatchanges observed in task 2 cannot be attributed to reflect generalspeeding effects induced by reward and punishment modulations.There were generally no group dependent modulations in errorrates on task 1 and task 2. This shows that the reward and pun-ishment manipulation affects the speed of performance with noevidence for a speed-accuracy trade-off.

Differences between the reward, the punishment and the con-trol group were only evident in short SOAs when the task order wasunpredictable. In the SOA 16 ms condition, the RT on T2 is shorter incase of punishment, compared to the control group. In the rewardcondition, RTs on T2 were longer compared to controls.

In the introduction we noted that two factors may be importantfor the modulation of dual-task performance by rewards and pun-ishments: (i) processing capacity and (ii) limitations in responseselection.

ain Re

bi[tfisaimpstTmvpdd

brDWnarPmis

tit[masLotmwo

apdmts

spstrdi

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

303–7.[28] Kahneman D, Tversky A. Prospect theory: an analysis of decisions under risk.

Econometrica 1979;47:313–27.[29] Tversky A, Kahneman D. Advances in prospect theory: cumulative representa-

tion of uncertainty. Journal of Risk and Uncertainty 1992;5:297–323.

A. Yildiz et al. / Behavioural Br

Numerous studies have shown that unpredictable switchingetween responses affects PRP performance, which stress the

mportance of a response selection limitation in dual-tasking22–24]. In particular it has been shown that processes related to T1ask disengagement, or suppression of T1 task sets are importantor response selection in the PRP, which are especially demand-ng in short SOAs [23]. In this regard, the results suggest that taskets inhibition processes are more efficient in case of punishments,nd less efficient in case of rewards. In this regard it is interest-ng that in the SOA 133 condition, punishments, but not rewards

odulated task 2 RTs, compared to controls. It may therefore beossible that punishments are a more effective modulator of tasket suppression or disengagement processes. This may be due tohe fact that punishments and rewards were of similar magnitude.he ‘prospect theory’ [28,29] proposes a higher sensitivity againstoney loss, compared to a similar amount of money gain. These

ariations in sensitivity may drive differences between rewards andunishments in the modulation of response selection processes atifferent SOA length and hence in conditions imposing differentemands on response selection mechanisms.

One can only infer the possible neuronal mechanisms inducedy rewards and punishments. Previously, it has been shown thatewards and punishments differ regarding their effect on dopamine1 (rewards) and dopamine D2 receptors (punishments) [11–14].ith respect to models of response selection in fronto-striatal

etworks it has been suggested that selection and control mech-nisms are implemented via the D1 and D2 receptor system,espectively [20]. The more prominent effects of punishments onRP task performance may hence suggest that control processes areore important than selection processes. In this way the limitation

n dual-tasking more reflects a response control, than a responseelection deficit.

An alternative explanation of the observed effects refers tohe effects of rewards and punishment on processing capac-ty. It has been suggested that in a dopamine ‘D2-state’ moreask goal representations are accessible in working memory8,9]. The ‘D2-state’ is more responsive and flexible, but also

ore interference-prone than the ‘D1-state’. However, mech-nisms enabling flexible behavioural adaption have long beenuggested to promote performance in dual-task situations [22–24].ikely, modulatory differences between rewards and punishmentsn the response selection bottleneck are due to more flexibleask goal representations that are promoted in case of punish-

ents, compared to rewards. This would also explain why effectsere restricted to the block presenting the tasks in random

rder.Clearly, it is only possible to infer the possible neuronal mech-

nisms of the effects of reward and punishments on dual taskerformance. Studies in neuropsychiatric diseases affecting theopamine system and the effects anti-psychotic treatment byeans of dopamine D2-receptor antagonists shall be conducted

o directly examine the importance of the dopamine D2-receptorystem for dual task performance.

In summary, the study shows that the dual-task responseelection bottleneck is differentially modulated by rewards andunishments. Punishments lead to a weakening of the responseelection bottleneck, while rewards entail a more restrictive bot-leneck. The differential effects of rewards and punishments onesponse selection may be due to the involvement of differentopamine receptor systems, which have been shown to play an

mportant role in response selection.

search 250 (2013) 304– 307 307

Acknowledgements

This work was supported by a Grant from the DeutscheForschungsgemeinschaft (DFG) BE4045/10-1 to C.B. We thank allparticipants.

References

[1] Pashler H. Dual-task interference in simple tasks: data and theory. Psycholog-ical Bulletin 1994;116:220–44.

[2] Welford AT. The psychological refractory period and the timing of high-speed performance: a review and a theory. British Journal of Psychology1952;43:2–19.

[3] Pashler H. Processing stages in overlapping tasks: evidence for a central bottle-neck. Journal of Experimental Psychology: Human Perception and Performance1984;10:358–77.

[4] Wu C, Liu Y. Queuing network modeling of the psychological refractory period(PRP). Psychological Review 2008;115:913–54.

[5] Meyer DE, Kieras DE. A computational theory of executive cognitive pro-cesses and multiple-task performance: Part 1. Basic mechanisms. PsychologicalReview 1997;104:3–65.

[6] Tombu M, Jolicoeur P. A central capacity sharing model of dual-taskperformance. Journal of Experimental Psychology: Human Perception and Per-formance 2003;29:3–18.

[7] Navon D, Miller JK. Queuing or sharing? A critical evaluation of the single-bottleneck notion. Cognitive Psychology 2002;44:193–251.

[8] Durstewitz D, Seamans JK. The dual-state theory of prefrontal cortex dopaminefunction with relevance to catechol-o-methyltransferase genotypes andschizophrenia. Biological Psychiatry 2008;64:739–49.

[9] Seamans JK, Yang CR. The principal features and mechanisms of dopaminemodulation in the prefrontal cortex. Progress in Neurobiology 2004;74:1–58.

10] Schultz W, Tremblay L, Hollerman JR. Reward processing in pri-mate orbitofrontal cortex and basal ganglia. Cerebral Cortex 2000;10:272–84.

11] Kravitz AV, Tye LD, Kreitzer AC. Distinct roles for direct and indirect pathwaystriatal neurons in reinforcement. Nature 2012;15:816–8.

12] Bromberg-Martin ES, Matsumoto M, Hikosaka O. Dopamine in motivationalcontrol: rewarding, aversive, and alerting. Neuron 2010;68:815–34.

13] Ferguson SM, Eskenazi D, Ishikawa M, Wanat MJ, Phillips PE, Dong Y, Roth BL,Neumaier JF. Transient neuronal inhibition reveals opposing roles of indirectand direct pathways in sensitization. Nature Neuroscience 2011;14:22–4.

14] Hikida T, Kimura K, Wada N, Funabiki K, Nakanishi S. Distinct roles of synaptictransmission in direct and indirect striatal pathways to reward and aversivebehavior. Neuron 2010;66:896–907.

15] Wichmann T, DeLong MR. Functional neuroanatomy of the basal ganglia inParkinson’s disease. Advances in Neurology 2003;91:9–18.

16] Gerfen CR, Surmeier DJ. Modulation of striatal projection systems by dopamine.Annual Review of Neuroscience 2011;34:441–66.

17] Cisek P, Kalaska JF. Neural mechanisms for interacting with a world full of actionchoices. Annual Review of Neuroscience 2010;33:269–98.

18] Chakravarthy VS, Joseph D, Bapi RS. What do the basal ganglia do? A modelingperspective. Biological Cybernetics 2010;103:237–53.

19] Redgrave P, Prescott TJ, Gurney K. The basal ganglia: a vertebrate solution tothe selection problem. Neuroscience 1999;89:1009–23.

20] Humphries MD, Stewart RD, Gurney KN. A physiologically plausible model ofaction selection and oscillatory activity in the basal ganglia. Journal of Neuro-science 2006;26:12921–21242.

21] Frank MJ. Dynamic dopamine modulation in the basal ganglia: a neuro-computational account of cognitive deficits in medicated and nonmedicatedParkinsonism. Journal of Cognitive Neuroscience 2005;17:51–72.

22] Jentzsch I, Leuthold H, Ulrich R. Decomposing sources of response slowing inthe PRP paradigm. Journal of Experimental Psychology: Human Perception andPerformance 2007;33:610–26.

23] Sigman M, Dehaene S. Dynamics of the central bottleneck: dual-task and taskuncertainty. PLoS Biology 2006;4:e220.

24] Lien MC, Proctor RW. Stimulus–response compatibility and psychologicalrefractory period effects: implications for response selection. PsychonomicBulletin and Reviews 2002;9:212–38.

25] Borger R. The refractory period and serial choice-reactions. Quarterly Journalof Experimental Psychology 1963;15:1–12.

26] Fischer R, Hommel B. Deep thinking increases task-set shielding andreduces shifting flexibility in dual-task performance. Cognition 2012;123: