Acta Psychologica 149 (2014) 148–156

Contents lists available at ScienceDirect

Acta Psychologica

j ourna l homepage: www.e lsev ie r .com/ locate /actpsy

Emotional modulation of attention affects time perception:Evidence from event-related potentials

Maria Tamm a,b,⁎, Andero Uusberg a,b, Jüri Allik a,b,c, Kairi Kreegipuu a,b

a Department of Experimental Psychology, Institute of Psychology, University of Tartu, Estoniab Doctoral School of Behavioural, Social and Health Sciences, University of Tartu, Estoniac Estonian Academy of Sciences, Estonia

⁎ Corresponding author at: Department of ExperimPsychology, University of Tartu, Estonia. Tel.: +372 73759

E-mail address: Maria.Tamm@ut.ee (M. Tamm).

http://dx.doi.org/10.1016/j.actpsy.2014.02.0080001-6918/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 18 January 2013Received in revised form 27 January 2014Accepted 19 February 2014Available online 18 March 2014

PsycINFO classification:2300 Human Experimental Psychology2360 Motivation & Emotion2346 Attention2530 Electrophysiology2560 Psychophysiology

Keywords:Time perceptionInterval productionAffectERP

Emotional effects on human timeperception are generally attributed to arousal speeding up or slowing down theinternal clock. The aim of the present study is to investigate the less frequently considered role of attention as analternativemediator of these effectswith the help of event-related potentials (ERPs). Participants produced shortintervals (0.9, 1.5, 2.7, and 3.3 s) while viewing high arousal images with pleasant and unpleasant contents incomparison to neutral images. Behavioral results revealed that durationswere overproduced for the 0.9 s intervalwhereas, for 2.7 and 3.3 s intervals, underproductionwas observed. The effect of affective valencewas present forthe shorter durations anddecreased as the target intervals became longer.More specifically, the durations for un-pleasant imageswere less overproduced in the 0.9 s intervals, and for the 1.5 s trials, durations for unpleasant im-ages were slightly underproduced, compared to pleasant images, which were overproduced. The analysis ofdifferent ERP components suggests possible attention processes related to the timing of affective images in addi-tion to changes in pacemaker speed. Early Posterior Negativity (EPN)was larger for positive than for negative im-ages, indicating valence-specific differences in activation of early attention mechanisms. Within the early P1 andthe Late Positive Potential (LPP) components, both pleasant and unpleasant stimuli exhibited equal affectivemodulation. Contingent Negative Variation (CNV) remained independent of both timing performance and affec-tivemodulation. This pattern suggests that both pleasant and unpleasant stimuli enhanced arousal and capturedattention, but the latter effect was more pronounced for pleasant stimuli. The valence-specificity of affective at-tention revealed by ERPs combinedwith behavioral timing results suggests that attention processes indeed con-tribute to emotion-induced temporal distortions, especially for longer target intervals.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Our sense of time can be altered by context, being particularly de-pendent on emotional states. The capacity to estimate time is assumedto rely on an internal clock-like mechanism (Buhusi & Meck, 2005;Gibbon, Church, & Meck, 1984; Treisman, 1963). This conceptual clockconsists of a pacemaker transmitting pulses to an accumulator througha switch. Temporal decisions observed in laboratory timing tasks furtherrequire comparing the content of the accumulator with previouslystored target durations in working memory. In principle, independentinfluences on any part of this clock-like mechanism – pacemaker,switch, accumulator, andmemory –may generate distortions in subjec-tive time. Affective temporal distortions are mostly explained by the ef-fects of emotional arousal on an internal clock. However, theoreticalarguments as well as empirical findings suggest that an interferencebetween affective attention and a switch component of an internal

ental Psychology, Institute of02; fax: +372 7376152.

clock-like mechanism may also be involved. The present study usesERP components associated with emotional arousal, affective attention,and time perception to disentangle arousal and attention contributionsto timing illusions.

One phenomenon capable of influencing time is emotion (for re-views see Droit-Volet & Gil, 2009; Droit-Volet &Meck, 2007; Wittmann& Paulus, 2008). As a general rule, durations of emotional stimuli tend tobe judged as longer than neutral stimuli. This effect has been observedwith emotional faces (Droit-Volet, Brunot, & Niedenthal, 2004; Gil &Droit-Volet, 2011; Lee, Seelam, & O'Brien, 2011) and images (Angrilli,Cherubini, Pavese, & Manfredini, 1997), as well as with emotionalsounds (Noulhiane, Mella, Samson, Ragot, & Pouthas, 2007). Moreover,highly arousing negative stimuli (i.e., angry faces, negative sounds) arejudged as significantly longer than low arousal stimuli of negative orpositive valence (e.g., fearful, sad, and happy faces; positive sounds)(Noulhiane et al., 2007; Tipples, 2008). However, emotional lengthen-ing of time is typically demonstrated for intervals shorter than 1 s andis rarely observed for durations longer than 2–3 s (e.g., Droit-Voletet al., 2004). Angrilli et al. (1997) even showed, in a temporal reproduc-tion task, that the relative overestimation of time for 2 s intervals shifted

149M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

to underestimation at longer intervals (4 and 6 s). Emotions, therefore,seem to have complex effects on time perception.

Emotional lengthening of subjective time can first of all be attributedto arousal, which is known to increase the rate at which pulses areemitted from the internal pacemaker. For instance, subjective time ismodulated when arousal is directly manipulated by body temperature(Wearden & Penton-Voak, 1995) or drug administration (Maricq,Roberts, & Church, 1981). Therefore, increases in physiological arousalinduced by emotional events (Cacioppo & Gardner, 1999) should simi-larly enhance the speed of the pacemaker. In fact, temperature effectson time may be mediated by the higher levels of emotional discomfortand arousal accompanying increased body temperature (Tammet al., 2014).

Alternatively, shifts in timingperformance can also be induced by at-tention processes. According to the internal-clock model, attention caninfluence sense of time by determining the fraction of pulses thatreach the accumulator (Pouthas & Perbal, 2004; Zakay, 1998; Zakay &Block, 1996). Affective stimuli are known to be natural targets for selec-tive attention— a phenomenon known asmotivated attention (for a re-view, see Vuilleumier, 2005). Conceivably, emotional stimuli maytherefore alter the switch component of the timing mechanism viatheir effects on attention. However, the direction of these effectsseems to depend on whether affective stimuli are distractors or targetsfor the timing task (Schirmer, 2011). In the former case, given that se-lective attention is resource-limited, emotional stimuli may draw atten-tion away from interval timing. This in turn leads to fewer temporalunits being processed and consequently results in temporal underesti-mations (Penney, 2003; Zakay, 1998; Zakay & Block, 1996).When emo-tional stimuli are to be timed, however, the additional attentioncaptured by emotional stimuli could facilitate the perceiving of the tem-poral aspects of those stimuli (e.g., carefully counting all the ticks),resulting in relative overestimation of durations (Schirmer, 2011).

Emotional effects on attention also provide some of the evidencethat is not congruent with the arousal–pacemaker explanation of emo-tional time distortion (Droit-Volet & Gil, 2009). For example, Gable andPoole (2012) propose that the attentional capture caused by high ap-proach motivation can explain why pleasant stimuli shorten timewhile equally arousing negative stimuli do not. Differences in emotionalprocessing as a function of interval length also suggest the involvementof several separate timingmechanisms (Angrilli et al., 1997; Gil & Droit-Volet, 2011). Temporal distortions in emotional stimuli within shorterdurations should reflect arousal-based effects and the activation of au-tomatic processes, whereas timing of longer durations requires morecognitive resources, such as attention. In combination, the theoreticaland empirical arguments suggest that the interplay between arousaland attention mechanisms may result in the affective modulation oftime perception.

While arousal and attention could both conceivably underlie emo-tional influences on subjective time, their contributions are often diffi-cult to disentangle at the behavior level. One potential solution is tosearch for the neural correlates of the different underlying processes,such as event-related potentials (ERP). To date, the quest to find theelectrophysiological correlates of time perception has mostly concen-trated on Contingent Negative Variation (CNV). CNV is a slow frontalERP component occurring during the anticipation stage of a required ac-tion or relevant feedback (van Rijn, Kononowicz, Meck, Ng, & Penney,2011) and it has also been related to timing performance in general(for reviews see Macar & Vidal, 2004, 2009; van Rijn et al., 2011) andemotional time modulations in particular (Eimer & Holmes, 2007;Gan, Wang, Zhang, Li, & Luo, 2009). However, the specific functionalcorrelates of CNV in the context of timing remain unclear (cf. Macar &Vidal, 2009; van Rijn et al., 2011). According to one bold proposal, agradually increasing CNV amplitude directly reflects the accumulationof pulses, probably in the Supplementary Motor Area (Macar & Vidal,2009). This account can explain the correlations between CNV ampli-tudes and time production errors (Bendixen, Grimm, & Schröger,

2005; Macar & Vidal, 2002; Macar, Vidal, & Casini, 1999). However, fail-ures to replicate such correlations as well as other findings contradictthis idea (Kononowicz & van Rijn, 2011). Alternatively, CNVmay simplyreflect peripheral aspects of timing performance, such as motor prepa-ration and decision processes (van Rijn et al., 2011). Understandingthe relationship between timing and CNV thus requires further re-search. To this end, the present study investigates the sensitivities ofCNV amplitudes to timing performance with and without emotionalinfluences.

The usefulness of ERPs in timing research need not be confined toCNV dynamics. That is, ERPs may also reveal aspects of emotionalarousal and attention experienced by participants during timing perfor-mance. In particular, emotional stimuli are known to enhance theamplitudes of several posterior components such as P1, Early PosteriorNegativity (EPN), and Late Positive Potential (LPP) (Hajcak,MacNamara, & Olvet, 2010; Olofsson, Nordin, Sequeira, & Polich, 2008;Schupp, Flaisch, Stockburger, & Junghöfer, 2006; Schupp et al., 2004).These components, particularly LPP, can first of all be used to obtain di-rect and fairly objective measures of the overall emotional arousal in-duced by certain stimuli in specific participants. This is an importantadvantage over those studies where arousal levels were only assumedfrom normative stimulus ratings or indirectly assessed from subjectiveself-ratings. Moreover, some affective ERP components can also belinked to more distinct underlying mechanisms. EPN and LPP havebeen associated with two consecutive stages of affective attention(Schupp et al., 2006; Uusberg et al., 2013). At the first and largely auto-matic stage, motivationally significant stimuli are selected for furtherprioritized processing, which occurs at the second stage. We thus as-sume that EPN can be considered amore selectivemeasure of the extentto which affective stimuli automatically capture attention, compared toLPP, which reflects more controlled processes, as well as P1, whichprobably corresponds to pre-attentive perceptual encoding. This partialdissociation will be used to differentiate the contributions of attentionand arousal phenomena on time perception. If EPN predicts behaviorbetter than LPP, distortions should be attributed, at least partially, toattention-related mechanisms. Meanwhile, if LPP is a better match toperformance data, the underlying mechanism is more likely to involveaffective arousal.

2. Method

2.1. Participants

The sample consists of 62 university students (42 females) aged 19–51 years. All participants were healthy, had normal or corrected-to-normal vision, and spoke fluent Estonian. All participants providedinformed consent and the study was approved by the Ethics ReviewCommittee of Human Research at the University of Tartu, Estonia.

2.2. Stimuli

Images from the International Affective Picture System (IAPS) wereselected based on published normative ratings (Lang, Bradley, &Cuthbert, 2005). Unpleasant and pleasant image sets were constructedwith similarly high arousal ratings, accompanied by a minimally arous-ing neutral category. Separate positive stimuli were selected for malesand females, as the latter are known to report less arousal as well as re-duced valence in response to some erotic images. As a result, negativeimages depicting mutilated human bodies (IAPS images 3051, 3060,3170; mean normative valence M = 1.9, SD = 1.5; arousal M = 6.7,SD = 2.2), neutral images of everyday objects (7000, 7004, 7009;valenceM=5.0, SD=0.8; arousalM=2.5, SD=1.8) and positive im-ages of opposite sex nudes or erotic couples (4645, 4676, 4677 for fe-males; valence M = 6.7, SD = 1.7; arousal M = 6.0, SD = 2.2; and4652, 4664, 4670 for males; valence M = 7.9, SD = 1.1; arousal M =7.4, SD = 1.7) were used in the study. All images contained a few

150 M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

large objects on a uniform background. The stimuli were presented on a19-in. CTR monitor (Mitsubishi Diamond Pro, refresh rate 120 Hz) at1 m distance (screen subtending 22.95°/arc horizontally and 17.29°/arc vertically). In addition to pictorial stimuli, a blank gray screen waspresented in one quarter of the trials. This condition was included toa) obtain a baseline CNV uncontaminated by stimuli and affectiveprocessing and b) draw explicit attention away from the distinction be-tween affective categories by creating an evenmore salient dissimilaritybetween picture and non-picture trials.

2.3. Procedure

The time production task was administered last within a battery ofthree behavioral tasks (after a stop-signal task and a delay discountingtask). Each task required about 20 min and the whole session about2 h. Sessions were scheduled to start at either 9 AM or 12 PM. Duringtask performance, participants sat alone in a quiet, dimly lit, and electri-cally shielded room. Before each task,written aswell as oral instructionswere provided together with a short practice session supervised by theexperimenter. In the time production task, participants were asked toproduce different intervals by two subsequent mouse-clicks whileviewing affective images. They were instructed not to divert gaze fromthe screen and avoid excessive eye-blinks during the productioninterval.

Four target intervals were used: 0.9, 1.5, 2.7, and 3.3 s. Each intervalwas presented with each stimulus type (negative, positive, neutral, andno stimulus) in a randomized order within one experimental block. Al-together, 4 blocks were presented, resulting in 192 trials in total. Eachexperimental trial started with an instruction screen stating the dura-tion to be produced (e.g., 1.5 s). The nextmouse-click by the participant(hereafter referred to as R1) started the production interval as well asthe presentation of the affective image. After the second button-press(R2), the image was replaced with a black screen for 1.5 s. Finally, aknowledge result (KR) was shown for a duration of between 0.8 and1.2 s, which was randomly selected. On 20% of trials within eachblock, the KR provided informative feedback stating that the completedtrial was “Correct!” in cases where the produced interval laywithin 10%of the target interval or otherwise that the interval was “Too short!” or“Too long!.” On the remaining 80% of trials, the words “No feedback”were presented. This feedback ratio was selected to encourage activeanticipation of the KR without inducing excessive learning effects.

Several self-report measures were also collected in the days beforethe experimental session, as part of an online-battery. During the ses-sion, participants were additionally screened for subjective fatiguebefore and after each task. Finally, after completing all tasks, the partic-ipants filled in a short questionnaire measuring various potentially rel-evant factors, including handedness, recent consumption of cigarettes,alcohol, and coffee. The self-report data are not analyzed in this paper.

2.4. EEG recording and pre-processing

A continuous EEG was recorded with a BioSemi ActiveTwo system(BioSemi, Netherlands) from 32 scalp electrodes positioned accordingto the 10/20 system, as well as four electrodes capturing horizontaland vertical eye-movements. The recording used the BioSemi specialvirtual reference between the CMS (Common Mode Sense) and DRL(Driven Right Leg) electrodes at 1024 Hz, with a band-pass filter withhalf-amplitude cut-offs at 0.16 and 100 Hz. The data were laterresampled to 256 Hz and an average reference was calculated. Offlineprocessing was conducted with EEGLAB (Delorme & Makeig, 2004)and Matlab R2008b (MathWorks, USA) software.

Artifacts were dealt with using Independent Component Analysis(ICA) to correct for eye-movements, followed by removal of noisy chan-nels and data segments. An Infomax ICA algorithmwas applied to train-ing data consisting of 1000ms non-baseline corrected epochs that werehigh-pass filtered (1 Hz Butterworth with 12 dB/oct) and pruned of

artifactual channels (EEGLAB rejchan, probability N 5 SD), aswell as seg-ments with excessive muscular noise (EEGLAB rejspec, 20–40 Hz). Theindependent components corresponding to eye-blinks and horizontaland vertical eye movements were visually identified and rejected foreach subject (e.g., Mognon, Jovicich, Bruzzone, & Buiatti, 2010). The re-maining components were used to reconstruct the original, unfilteredEEG-activity (as suggested by Debener, Thorne, Schneider, & Viola,2010). The ICA-pruned continuous data were then segmented intotwo different sets: stimulus-locked (from −200 before to 900 msafter R1, 200 ms pre-stimulus baseline) and response-locked (from−900 before to 200 ms after R2; 100 ms before and after the R2 base-line). All trials with responses more than 2 SD away from the samplemean were discarded. The remaining segments were screened for arti-facts using the threshold (rejthresh;+/−125 μV) and spectral (rejspec:15–30 Hz) criteria in a dynamic algorithm, so that if a single channelwas responsible for excluding more than 2% of segments, the channelwas rejected and the procedure repeated on the remaining channels.All rejected channels were spherically interpolated (EEGLAB interp) tosimplify subsequent analyses (whichwere based on larger scalp regionsrather than single electrodes). On average, 91.3% of trials remained(SD = 9%, range 67.2–100%) in the stimulus-locked data and 94.2%of trials (SD = 9.5%, range 70.8–100%) in the response-locked set.

2.5. Analysis

2.5.1. Behavioral data analysisProduced intervals that differed from the mean of each target inter-

val by more than 2 standard deviations were excluded from furtheranalysis. Interval production data are presented as relative temporalproductions following Gil and Droit-Volet (2011), that is, “(mean tem-poral estimate − target duration) / the target duration”. This transfor-mation provides information about the direction and magnitude of atemporal distortion,with positive values indicating that the stimuli pro-duced estimates longer than the target duration (overproduction) andnegative values indicating shorter productions than the target duration(underproduction). A repeatedmeasures ANOVAwith a Tukey post-hocanalysis was performed on behavioral data with affective category(pleasant, unpleasant, and neutral pictures) and duration (0.9, 1.5,2.7, and 3.3 s) as factors. For assessing the precision of temporal perfor-mance, a Coefficient of Variation (CV) was calculated (standarddeviation / mean; Wearden & Lejeune, 2008) and analyzed similarlyto produced intervals.

2.5.2. EEG data analysisStimulus- and response-locked ERP dynamics were investigated for

distinct aims. The former were assumed to reflect initial affective re-sponses to different stimuli while the latter were mined for potentialcorrelates of timing performance and affective influences on them.Three affect-sensitive components were identified in stimulus-lockedERPs: P1, Early Posterior Negativity (EPN), and Late Positive Potential(LPP). In the response-locked data, Contingent Negative Variation(CNV) was analyzed. Statistical analysis was comprised of two steps: amass-univariate ANOVA to test the central hypothesis, followed by amore detailed repeated measures ANOVA of the averaged dynamicswith Tukey post-hoc testing. False Discovery Rate correction was ap-plied to mass univariate p-values to correct for the inflated Type Ierror risk induced by multiple statistical comparisons.

Note that analyzingCNV from thepresent data is complicated, due toa potential overlap between timing and emotional processes. More spe-cifically, as Fig. 2 illustrates, affective content increased voltage positiv-ity at the fronto-central area, which exhibits a negative-trending CNV.To disentangle the latter activity from the former, we quantified CNVfrom the response-locked rather than the stimulus-locked data by cal-culating the difference between the 900 to 700 ms pre-R2 voltage andthe baseline measured from around the R2 (from −100 ms to100 ms). In addition, only longer production intervals (2.7 and 3.3 s)

151M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

were used in the CNVanalysis, assuming that the last 900msof the ERPsreflects the neural generators of CNV, rather than any systematic affec-tive effects, as LPP is known to reach a stable plateau after initial stimu-lus processing (Hajcak et al., 2010). To further confirm the validity ofthis approach, the trials from the gray screen conditionwere used to cal-culate independent CNV estimates for all durations.

3. Results

3.1. Temporal production

Fig. 1 depicts the relative time distortions observed in this study. Arepeated measures ANOVA of the depicted data revealed a significantmain effect for affective category, F(2, 122) = 5.91, p b 0.01. The maineffect for duration, F(3, 183) = 53.74, p b 0.0001, was also significant,due to the relative overproduction of shorter durations and underpro-duction of longer durations. A significant interaction occurred betweenthese two factors, F(6, 366) = 11.17, p b 0.0001, meaning that thetemporal distortion induced by different emotional stimuli wasdependent on the duration to be produced (see Fig. 1). Tukey post-hoc tests (p b 0.05) revealed that, when the target duration was 0.9 s,negative stimuli (M = 0.06) were less overproduced than neutral(M= 0.13) and positive (M= 0.12) stimuli. For the 1.5 s target inter-val, there was a significant difference between positive (M = 0.02)and neutral (M = −0.03) stimuli. The target duration of 2.7 s wasunderproduced for all stimuli, but the underproduction of positive stim-uli (M=−0.14) was significantly different from that of neutral stimuli(M = −0.09). No significant stimulus difference occurred for the un-derproduction of the target duration of 3.3 s. In summary, for shorter in-tervals, the durations were overproduced whereas, for the longerintervals, underproduction of durations occurred. The valence effectwas also more prominent for shorter intervals and completely absentfor the 3.3 s interval.

For the Coefficient of Variation (CV), a main effect of affective cate-gory was present, F(2, 122) = 8.55, p b 0.001, indicating a higher CVfor the processing of unpleasant stimuli (M= 0.22) compared to pleas-ant and neutral stimuli (M = 0.20 and M = 0.21, respectively). Amain effect for duration also occurred, F(3, 183)= 66.95, p b 0.0001, in-dicating a progressive decrease in CV with increasing target durations.In addition, the interaction between these factors was significant, F(6,366) = 3.51, p b 0.01. A Tukey post-hoc test (p b 0.05) revealed thatthe main difference occurred within the shortest duration (0.9 s),where the CV for the unpleasant stimuli (M = 0.28) was significantlylarger than for the pleasant and neutral stimuli (M = 0.25 and M =

Fig. 1. Relative temporal productions plotted against target durations for the negative,positive and neutral pictures. **p b 0.0001; * b 0.01.

0.25, respectively; see Fig. 2). No significant differences were found forthe longer stimulus durations.

3.2. Stimulus-locked affective ERP components

To identify the time-course and topography of affectivemodulationsof stimulus-locked ERPs, we averaged negative and positive waveformsfrom all intervals and calculated the difference between this aggregatedaffective ERP and the mean response to neutral images. Averagingacross valence is justified given that affective ERP modulations areknown to be sensitive to arousal rather than valence (Hajcak et al.,2010; Olofsson et al., 2008). The resulting affective difference waves ineach scalp electrode are plotted in panel A of Fig. 3. The panel alsoshows a bar reflecting the number of electrodes exhibiting a significantaffective main effect at a given point of time. As can be seen, affectivecontent significantly modulated ERPs from around 60 ms onwardswithin three expected time-windows: P1 (60–150 ms), EPN (150–320 ms), and LPP (320–500 ms). Panel B of Fig. 3 plots scalp distribu-tions of voltages averaged within these time-windows. These plotswere used to define representative scalp regions for each component:occipital (O1, O2, PO3, PO4) for P1, as well as EPN and central-posterior (Cz, CP1, CP2, Pz) for LPP. The mean waveforms from allthree picture conditions within these two regions are plotted in panelsC and D of Fig. 3.

Fig. 4 indicates clear affective effects in all three stimulus-lockedcomponents. Note that affective content increased voltage positivity inP1 and LPP and negativity within EPN. This is the reason why EPN is re-ferred to as a negativity even though the underlying ERP component islargely positive. To determine which pair-wise differences between af-fective categories reached significance, aswell as to studypossible inter-val effects, a repeated measures ANOVA with Tukey post-hoc tests wasconducted on voltages averaged within the identified time-windowsand scalp regions. For P1, the main effect for affective category was sig-nificant, F(2, 122)= 29.44, p b .0001. Post-hoc tests revealed that affec-tivemodulations induced by both negative and positive stimuli differedfrom the neutral category, but not from each other (see Fig. 4 andTable 1). For EPN, the main effect for affective category was also signif-icant, F(2, 122)= 93.55, p b .0001, togetherwith the affect-by-durationinteraction, F(6, 366)= 4.18, p b .001. Post-hoc tests indicated that, de-spite the significant interaction, all three stimulus categories differedfrom each other in all durations (see Table 1). Specifically, pleasant im-ages showed stronger affective modulations than the unpleasant ones.The main effects for affective category, F(2, 122) = 107.07, p b .0001,and for duration, F(3, 183)= 5.78, p b .001, were significant predictorsof LPP dynamics. Similarly to P1, the affective modulation for negative

Fig. 2. Coefficient of Variation (CV) of temporal productions for the negative, positive andneutral pictures. Spreads denote standard errors. *p b 0.01.

Fig. 3. Affectivemodulation of stimulus-locked ERP-s. Panel A. Neutral waveform subtracted from an average of positive and negative at all electrodes. Lower panel depicts the number ofelectrodeswith significant affective effect at a given time-point (ANOVA, FDR corrected p b .05). Panel B. Scalp distributions of the affective differencewave averagedwithin time-windowsidentified from Panel A. Panel C. Average ERP-s from an occipital region (O1, O2, PO3, PO4) showing affectivemodulation of P1 and EPN. Panel D. Average waveform from central-parietalregion (Cz, CP1, CP2, Pz) showing affectivemodulation of the LPP. Shades aroundwaveforms on panels C and D denote standard errors. Vertical lines on panels A, C and D demarcate stim-ulus onset and subsequent time-windows used for averaging.

152 M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

and positive stimuli of LPP was different from neutral stimuli, but notfrom one another. Also, the shortest (0.9 s) duration showed lower af-fective modulation compared to the 1.5 s and 3.3 s durations (seeTable 1). Given the long latency of the LPP window, the duration maineffect may simply reflect stronger interference between stimulus per-ception and response preparation processes in the shorter comparedto the longer production trials.

Fig. 4. Affective modulation of three ERP componen

3.3. CNV

Based on previous reports and scalp-plots of average response-locked activity measured in this study, CNV was averaged from Fz,FC1, and FC2 (see scalp maps in Fig. 5; note that, in the response-locked data, negative trends are reflected in positive voltages). Beforestudying affective effects in CNV, we attempted to replicate the

ts. Spreads denote standard errors. *p b 0.001.

Table 1Event-related potentials P1, EPN and LPP for different experimental stimuli across the tar-get durations.

Target duration (s) Negative Positive Neutral

P1 0,9 4.55(3.6) 5.12(3.0)** 3.75(3.3)1,5 4.98(3.5)** 4.74(2.7)** 3.57(2.8)2,7 4.66(3.8)** 4.58(3.1)** 3.43(3.0)3,3 4.70(3.6) 4.61(3.4) 3.91(3.4)

EPN 0,9 8.73(5.0)** # 7.40(4.0)** 11.04(4.8)1,5 9.46(4.7)* # 6.94(4.3)** 10.67(4.0)2,7 8.93(5.1)** # 6.83(4.2)** 11.39(4.5)3,3 8.91(4.6)** # 6.46(4.4)** 11.50(4.6)

LPP 0,9 0.01(2.5)** −0.13(2.5)** −1.85(2.5)1,5 0.69(2.4)** 0.60(2.5)** −1.65(2.1)2,7 0.73(2.5)** 0.002(2.2)** −1.44(2.2)3,3 0.55(2.8)** 1.10(2.4)** −1.30(2.2)

Notes:**p b .001; *p b .01; significant difference from the Neutral stimuli; #p b .001;significant difference from the positive stimuli. EPN — Early Posterior Negativity (μV);LPP — Late Positive Potential (μV); mean values with standard deviations.

153M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

relationship between CNV amplitude and timing accuracy. All trialswere distributed to three bins – relatively short, average, and longresponses – based on the within-subjects quintiles of produced timedistributions within each interval. This approach ensured that eachlevel of the timing–accuracy factor had a comparable number of obser-vations. The waveforms within these levels were averaged across allpicture stimuli for the two longest intervals and the gray screen trialsfor all durations. Waveforms for different intervals can be averagedgiven that CNV has been shown to be largely independent of specific in-terval durations (e.g., Gibbons & Rammsayer, 2004; Praamstra, Kourtis,Kwok, & Oostenveld, 2006). Somewhat surprisingly, as can be seen inpanel A of Fig. 5, differences in timing behavior were not reflected inCNV amplitudes. The absence of timing effects was replicated in thegray screen condition. These findings support the view that CNV maynot be a direct correlate of timing performance.

Fig. 5. Timing-performance and CNV. Panel A. Average responses to all picture stimuli from2.7 aintervals. The scalp maps showmean amplitudes from 900 to 700 ms prior to response. ERP ploand over-produced. Shaded areas denote standard errors.

Note that the failure to find any timing performance effects on CNVamplitudes undermines the validity of using this component as a directmeasure of emotional modulations on timing. We nevertheless con-ducted the initially planned comparisons of CNV amplitudes measuredfrom trials with different affective contents (again using only the twolongest intervals). Fig. 6 indicates that, in terms of CNV, the three affec-tive conditions did not differ. A repeated measures ANOVA of CNV am-plitudes (averaged within Fz, FC1, and FC2, from −900 to −700 mspre-response) in both time intervals revealed no significant main or in-teractive effects. The absence of affective effects in CNV is in line withthe behavioral data, where the two longer intervals analyzed here re-vealed only a single affective modulation. Unfortunately, CNVs fromthe shorter intervals cannot be reliably assessed from the present data.

4. Discussion

The central aim of the current studywas to investigate emotional ef-fects on interval production in relation to the affective modulation ofERPs. Behavioral results revealed valence effects for shorter intervals,as unpleasant stimuli resulted in relatively shorter productions thanpleasant stimuli. Although the valence effect decreased with longer tar-get intervals, pleasant stimuli durations were more underproduced forthe 2.7 s interval. Several stimulus-locked ERP components were usedto objectively characterize the states generated by affective images.While P1 and LPP showed valence-independent arousal-sensitivity,EPN was larger for positive than for negative images, indicating higheractivation of early attention mechanisms by the erotic compared tothe aversive content. No evidence was found for affective or timing ef-fects on CNV, adding to the growing body of evidence questioning thelink between this component and the internal clock-like mechanism.Taken together, these results confirm the central hypothesis that theconceptual link between emotional arousal and pacemaker rate is notsufficient to explain the complex pattern of emotional temporal distor-tions. Instead, our results implicate the involvement of attention mech-anisms, although further research is needed to differentiate between

nd 3.3 s production trials. Panel B. Average responses to the gray screen condition across allts show CNV (Fz, FC1, FC2) waveforms for three types of trials: under-produced, accurate

Fig. 6. Affective modulation of response-locked CNV. Shaded waveforms denote standarderrors.

154 M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

the variousways inwhich affectivemodulation of attention can contrib-ute to interval timing.

In termsof timingperformance,we observed a steady decrease in theratio between produced and required intervals from above to below 1 s.The 0.9 s interval was significantly overproduced whereas durations of2.7 and 3.3 swere underproduced for all pictorial stimuli. The 1.5 s inter-val seemed to constitute a cross-over condition between average over-and underproductions. According to the Vierordt's Law, a progressivechange fromoverestimation of short intervals to underestimation of lon-ger intervals is often observed in time reproduction tasks, with an indif-ference point occurring somewhere around the duration range wherejudgments are rather veridical (Lejeune & Wearden, 2009).

For the current purposes, the absolutemodulation of time is less im-portant than the relative differences between time distortions inducedby the various affective stimuli. In this regard, we found that unpleasantstimuli significantly shortened the produced intervals and increasedvariability compared to the neutral baseline only in the briefest, 0.9 sinterval. Appetitive images had an exclusive effect on 1.5 and 2.7 sdurations. For the 1.5 s durations, pleasant stimuli durations wereoverproduced compared to unpleasant stimuli, whereas for 2.7 s, pleas-ant stimuli durations were more underproduced than for the neutralimages. No significant affective modulations were observed for the lon-gest interval, although trends were similar to the 2.7 s condition.

This admittedly complex pattern can be summarized in three gener-alizations. First, in most significant as well as trend-level cases, affectivecontent increased rather than decreased the speed of subjective time.This finding aligns with most previous reports of subjective elongationof duration with emotional stimuli (Angrilli et al., 1997; Droit-Voletet al., 2004; Gil & Droit-Volet, 2011; Noulhiane et al., 2007). Second,the present results also replicated the general finding that emotions in-duce changes in the internal clock for relatively short intervals, in ourcase up to 2.7 s. Third, the power of affective distortions showed aduration-dependent valence-sensitivity, whereby the shortest produc-tion trials were affected by unpleasant and the longer ones by pleasantstimuli.

The interpretation of these observations can be facilitated by the ad-ditional information made available in this study by the ERP analyses.Building on the extensive literature, we used stimulus-locked ERPs toobjectively characterize the affective states that caused the behavioraltiming distortions. First of all, as a general manipulation check, allthree components of interest – P1, EPN, and LPP – confirmed that bothtypes of emotional stimuli generated clear and significant shifts awayfrom the neutral waveform in the expected direction. ERPs also provid-ed insights into the qualitative differences between the states inducedby the pleasant and unpleasant stimuli and these might, in turn, be rel-evant for the valence-specific timing distortions observed in this study.

We found a dissociation, whereby erotic images generated stronger af-fective modulation than the aversive stimuli in EPN, but not in either P1or LPP. The present findings thus suggest that pleasant images weresomewhat more captivating of early attention resources than negativeimages, while the arousal levels, reflected predominantly by LPP,remained indistinguishable. We also investigated CNV as a potential di-rect correlate of timing processes. However, several analyses confirmedthat, in the present experiment, CNV was completely unrelated to bothtiming performance, as well as its affective distortions. In this regard,the present study adds weight to the view (van Rijn et al., 2011)questioning the proposed association between CNV and timing perfor-mance (Macar & Vidal, 2004, 2009).

An issue central to the aims of this paper is whether behavioraltiming results can be explained by emotional arousal effects on an inter-nal pacemaker or should attention effects on a switch-like componentalso be incorporated. First of all, it is relatively clear that emotionalarousal alone cannot account for the intricate pattern of emotional dis-tortions we observed. A distinctive feature of the behavioral results wasthe duration-dependent valence-specificity of produced intervals. Morespecifically, productions of 0.9 s were sped up by unpleasant but notpleasant stimuli, and this pattern also extended to 1.5 s in terms of thedifference between positive and negative stimuli. The opposite wastrue for 2.7 s where the pleasant stimuli were more underproduced. Inaddition to the valence effect formean relative production, the observeddifferences in the variability of produced durations indicate changes inthe accuracy of timing performance. The Coefficient of Variation (CV)did not remain constant across durations, thus not exhibiting scalarproperties (Wearden, 1999). Larger variation togetherwith significantlyshorter productions of 0.9 s with unpleasant images suggests that thepacemaker rate is modulated by the valence of the stimuli. However,this pattern of results was not observed for the other target durations.The lack of significant differences for longdurations cannot be explainedby the use of counting. On the one hand, although no direct instructionswere provided for chronometric counting and the target intervals usedwere not integers, the use of counting cannot be ruled out. On the otherhand, according to Wearden (1991), counting generally yields verysmall variation values (0.1 or less), and the rather large CV values inthe current study (despite the systematic decline with increasing targetduration) render the use of the aforementioned strategy questionable.Noulhiane et al. (2007) explained similar results with simultaneous in-terference between activation and attention processes, reducing the dif-ference between the emotional and neutral stimuli, especially for longdurations. This explanation also fits our findings, which suggest an in-creasing role of attention over time in addition to accelerations in pace-maker rate. However, the exact source of variance is difficult to detect asthe various clock components and their respective contributions to tem-poral behavior are inconspicuous.

In addition to the behavioral data contradicting the simplified viewof arousal-induced temporal distortions, the present findings in ERPsprovide some indirect evidence for the involvement of attention mech-anisms. Specifically, we found a consistent difference between the affec-tive states induced by pleasant and unpleasant images in the EPNcomponent associated with early selective attention. This suggeststhat the appetitive erotic stimuli initially drew attention more potentlythan the aversive images. We assumed that consistent arousal differ-ences between the images would have emerged in all components,the LPP in particular. However, the P1 and LPP did not reveal anyvalence-by-duration interactions suggesting that, at least for the first900 ms of picture viewing, the overall arousal of the states induced byimageswas not a function of the target interval. This implies that the be-havioral valence-sensitivity of timing performance is somehow relatedto the valence differences in automatic affective attention.

More generally, it is plausible that a dynamic interplay betweenarousal and attention mechanisms is the most adequate explanationfor emotional timing distortions. The relative contribution of thesemechanisms may also depend on the durations to be perceived or

155M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

produced. Specifically, arousal-induced timing differences might be rel-atively more dominant in shorter durations, while the contribution ofattention becomes more evident in longer durations. Similarly to thepresentfindings, negative stimuli have previously been shown to be sig-nificantly overestimated in durations under 1 s (Gil & Droit-Volet, 2012;Tipples, 2008). These shorter temporal productions with unpleasantcompared to pleasant pictures have been associated with specificdisgust- or fear-induced arousal effects. It can thus be speculated thata negativity bias in the automatic activation of emotional arousal under-lies the present and previous findings regarding productions of veryshort trials. The production of longer durations,meanwhile,was consid-erably affected by the processing of pleasant rather than unpleasant af-fect. Conceptually, longer intervals to be timed should involve highercognitive processes relatively more so than shorter intervals, which re-main dominated by the automatic processes involved in emotion per-ception and the rapid detection of threatening stimuli (Gil & Droit-Volet, 2012).

Unfortunately, the present data alone do not reveal exactly how at-tention was involved in generating behavioral results. This is because,even while EPN demonstrated enhanced attention capture of pleasantstimuli, it remains unclear if this opened or closed the switch theorizedto operate in internal clockmodels. On the one hand, affective attentionmay enhance the functioning of the switch by drawing attention to thestimulus. Prioritized processing of emotional stimuli inherent in moti-vated attention conveys evolutionary benefits, such as faster detectionof threats or enemies. It would thus be reasonable to assume that theprioritized processing of emotional stimuli does not exclude the tempo-ral aspects of the stimuli (Schirmer, 2011). On the other hand however,emotional stimuli may also function as a source of distraction, divertingattention resources away from the timing task and thus closing theswitch more often. This is particularly likely if the intervals to be pro-duced do not in fact perceptually “bind” to the presented visual stimuli(i.e., participants “produced intervalswhile viewing images” rather than“controlled the duration of the images”). These two ways of performingthe task remain currently indistinguishable. As we only observed highlyarousing stimuli in the current study, low arousing stimuli should be in-cluded for further understanding the underlying mechanisms of emo-tional time perception.

In conclusion, the present study suggests that both arousal and at-tention processes are involved in the affective modulation of time per-ception. Although behavioral results show inconsistent valence effects,they nevertheless offer support to the central thesis of this paper thatthe arousal–pacemaker link is not sufficient to explain affective distor-tions of time. If this were the case, temporal effects should haveemerged largely independently from valence, as well as duration. In-stead, the complex pattern observed here indicates substantial contri-butions of other underlying mechanisms. The study also illustrated thefeasibility of using stimulus-related ERP components to elucidate thecauses of timing distortions. Meanwhile, the analyses of CNV suggestthis component probably does not live up to the promise of being a di-rect correlate of mental timing itself.

Acknowledgments

We thank Annika Teder, Anu Ilves, Kadi Tulver, Kerttu Mäger, andKristjan Mesipuu for their help in conducting the experiment. Thisstudy was supported by a Primus grant (#3-8.2/60) from theEuropean Social Fund to Anu Realo and grants from the EstonianMinis-try of Science and Education (SF0180029s08 and IUT02-13).

References

Angrilli, A., Cherubini, P., Pavese, A., & Manfredini, S. (1997). The influence of affective fac-tors on time perception. Perception & Psychophysics, 59, 972–982.

Bendixen, A., Grimm, S., & Schröger, E. (2005). Human auditory event-related potentialspredict duration judgments. Neuroscience Letters, 383(3), 284–288.

Buhusi, C. V., & Meck, W. H. (2005). What makes us tick? Functional and neural mecha-nisms of interval timing. Nature Reviews Neuroscience, 6(10), 755–765.

Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of Psychology, 50,191–214.

Debener, S., Thorne, J., Schneider, T. R., & Viola, F. C. (2010). Using ICA for the analysis ofmulti-channel EEG data. In M. Ullsperger, & S. Debener (Eds.), Simultaneous EEG andFMRI: Recording, analysis, and application. Oxford University Press.

Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal ofNeuroscience Methods, 134(1), 9–21.

Droit-Volet, S., Brunot, S., & Niedenthal, P.M. (2004). Perception of the duration of emo-tional events. Cognition and Emotion, 18, 849–858.

Droit-Volet, S., & Gil, S. (2009). The time–emotion paradox. Philosophical Transactions ofthe Royal Society, B: Biological Sciences, 364, 1943–1953.

Droit-Volet, S., & Meck, W. H. (2007). How emotions colour our perception of time. Trendsin Cognitive Sciences, 11(12), 504–513.

Eimer, M., & Holmes, A. (2007). Event-related brain potential correlates of emotional faceprocessing. Neuropsychologica, 45(1), 15–31.

Gable, P. A., & Poole, B.D. (2012). Time flies when you're having approach motivated fun:Effects of motivational intensity on time perception. Psychological Science, 23,879–886.

Gan, T., Wang, N., Zhang, Z., Li, H., & Luo, Y. (2009). Emotional influences on time percep-tion: Evidence from event-related potentials. NeuroReport, 20(9), 839–843.

Gibbon, J., Church, R. M., & Meck, W. H. (1984). Scalar timing in memory. In J. Gibbon, & L.G. Allan (Eds.), Timing and time perception (pp. 52–77). New York, NY: New YorkAcademy of Sciences.

Gibbons, H., & Rammsayer, T. H. (2004). Current–source density analysis of slow brain po-tentials during time estimation. Psychophysiology, 41(6), 861–874.

Gil, S., & Droit-Volet, S. (2011). “Time flies in the presence of angry faces”… depending onthe temporal task used! Acta Psychologica, 136, 354–362.

Gil, S., & Droit-Volet, S. (2012). Emotional time distortions: The fundamental role ofarousal. Cognition & Emotion, 26, 847–862.

Hajcak, G., MacNamara, A., & Olvet, D.M. (2010). Event-related potentials, emotion, andemotion regulation: An integrative review. Developmental Neuropsychology, 35(2),129–155.

Kononowicz, T. W., & van Rijn, H. (2011). Slow potentials in time estimation: The role oftemporal accumulation and habituation. Frontiers in Integrative Neuroscience, 5.

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2005). International affective picture system(IAPS): Instruction manual and affective ratings. The Center for Research in Psycho-physiology, University of Florida (Retrieved from http://www.unifesp.br/dpsicobio/adap/instructions.pdf).

Lee, K. H., Seelam, K., & O'Brien, T. (2011). The relativity of time perception produced byfacial emotion stimuli. Cognition & Emotion, 25(8), 1471–1480.

Lejeune, H., & Wearden, J. H. (2009). Vierordt's the experimental study of the time sense(1868) and its legacy. European Journal of Cognitive Psychology, 21, 941–960.

Macar, F., & Vidal, F. (2002). Time processing reflected by EEG surface Laplacians.Experimental Brain Research, 145(3), 403–406.

Macar, F., & Vidal, F. (2004). Event-related potentials as indices of time processing: A re-view. Journal of Psychophysiology, 18(2–3), 89–104.

Macar, F., & Vidal, F. (2009). Timing processes: An outline of behavioural and neural indi-ces not systematically considered in timing models. Canadian Journal of ExperimentalPsychology — Revue Canadienne, 63(3), 227–239.

Macar, F., Vidal, F., & Casini, L. (1999). The supplementary motor area in motor and sen-sory timing: evidence from slow brain potential changes. Experimental Brain Research,125(3), 271–280.

Maricq, A. V., Roberts, S., & Church, R. M. (1981). Methamphetamine and time estimation.Journal of Experimental Psychology: Animal Behavior Processes, 7, 18–30.

Mognon, A., Jovicich, J., Bruzzone, L., & Buiatti, M. (2010). ADJUST: An automatic EEG arti-fact detector based on the joint use of spatial and temporal features. Psychophysiology,48, 229–240.

Noulhiane, M., Mella, N., Samson, S., Ragot, R., & Pouthas, V. (2007). How emotional audi-tory stimuli modulate time perception. Emotion, 7, 697–704.

Olofsson, J. K., Nordin, S., Sequeira, H., & Polich, J. (2008). Affective picture processing: Anintegrative review of ERP findings. Biological Psychology, 77(3), 247–265.

Penney, T. B. (2003). Modality differences in interval timing: attention, clock speed, andmemory. In W. H. Meck (Ed.), Functional and neural mechanisms of interval timing(pp. 209–234). Boca Raton, FL: CRC Press.

Pouthas, V., & Perbal, S. (2004). Time perception depends on accurate clock mechanismsas well as unimpaired attention and memory processes. Acta NeurobiologiaeExperimentalis, 64, 367–385.

Praamstra, P., Kourtis, D., Kwok, H. F., & Oostenveld, R. (2006). Neurophysiology of implic-it timing in serial choice reaction-time performance. The Journal of Neuroscience,26(20), 5448–5455.

Schirmer, A. (2011). How emotions change time. Frontiers in Integrative Neuroscience,5(58).

Schupp, H. T., Cuthbert, B. N., Bradley, M. M., Hilman, C. H., Hamm, A. O., & Lang, P. J.(2004). Brain processes in emotional perception: Motivated attention. Cognition &Emotion, 18(5), 593–611.

Schupp, H. T., Flaisch, T., Stockburger, J., & Junghöfer, M. (2006). Emotion and attention:event-related brain potential studies. Understanding Emotions, vol. 156. (pp. 31–51).Elsevier, 31–51.

Tamm, M., Jakobson, A., Havik, M., Burk, A., Timpmann, S., Allik, J., et al. (2014). The com-pression of perceived time in a hot environment depends on physiological and psy-chological factors. The Quarterly Journal of Experimental Psychology, 67(1), 197–208.

Tipples, J. (2008). Negative emotionality influences the effects of emotion on time percep-tion. Emotion, 8, 127–131.

156 M. Tamm et al. / Acta Psychologica 149 (2014) 148–156

Treisman, M. (1963). Temporal discrimination and the indifference interval: Implicationsfor a model of the “internal clock”. Psychological Monographs, 77.

Uusberg, A., Uibo, H., Kreegipuu, K., Tamm, M., Raidvee, A., & Allik, J. (2013).Unintentionality of affective attention across visual processing stages. Frontiers inEmotion Science, 4, 969. http://dx.doi.org/10.3389/fpsyg.2013.00969.

Van Rijn, H., Kononowicz, T. W., Meck, W. H., Ng, K. K., & Penney, T. B. (2011). Contingentnegative variation and its relation to time estimation: a theoretical evaluation.Frontiers in Integrative Neuroscience, 5.

Vuilleumier, P. (2005). How brains beware: neural mechanisms of emotional attention.Trends in Cognitive Sciences, 9(12), 585–594. http://dx.doi.org/10.1016/j.tics.2005.10.011.

Wearden, J. H. (1991). Do humans possess an internal clock with scalar timing proper-ties? Learning and Motivation, 22, 59–83.

Wearden, J. H. (1999). “Beyond the fields we know…”: Exploring and developing scalartiming theory. Behavioural Processes, 45, 3–21.

Wearden, J. H., & Lejeune, H. (2008). Scalar properties in human timing: Conformity andviolations. The Quarterly Journal of Experimental Psychology, 61, 569–587.

Wearden, J. H., & Penton-Voak, I. S. (1995). Feeling the heat: Body temperature and therate of subjective time, revisited. The Quarterly Journal of Experimental Psychology,48(2), 129–141.

Wittmann, M., & Paulus, M. P. (2008). Decision making, impulsivity and time perception.Trends in Cognitive Science, 12, 7–12.

Zakay, D. (1998). Attention allocation policy influences prospective timing. PsychonomicBulletin & Review, 5(1), 114–118.

Zakay, D., & Block, R. A. (1996). The role of attention in time estimation processes.Advances in Psychology, 115, 143–164.