Report

Perceived time expands and contracts within each

heartbeat
Graphical abstract Authors
Irena Arslanova, Vassilis Kotsaris,
Manos Tsakiris

Correspondence
irena.arslanova@rhul.ac.uk

In brief
Arslanova et al. show that cardiac signals
arising from the heart distort experienced
time. Durations of non-arousing stimuli
are contracted during systole but
expanded during diastole. This balance is
disrupted when arousal increases. The
work provides causal evidence for the
involvement of interoceptive processes in
human time perception.

Highlights
d Time-locking identical stimuli to distinct cardiac phases
distorts perceived duration

d The cardiac-led time distortion is modulated by experienced

arousal

d At low arousal, systolic time contraction is counteracted by

diastolic time expansion

d At high arousal, diastolic time expansion shifts toward time

contraction

Arslanova et al., 2023, Current Biology 33, 1–7

April 10, 2023 ª 2023 The Author(s). Published by Elsevier Inc.
https://doi.org/10.1016/j.cub.2023.02.034 ll
 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
OPEN ACCESS

Report
Perceived time expands and contracts
within each heartbeat
Irena Arslanova,1,4,5,* Vassilis Kotsaris,2 and Manos Tsakiris1,3
1Department of Psychology, Royal Holloway, University of London, Egham TW20 0EY, UK
2School of Psychology, University of Kent, Canterbury CT2 7NP, UK
3Centre for the Politics of Feeling, School of Advanced Study, University of London, London WC1E 7HU, UK
4Twitter: @irena_arslanova
5Lead contact

*Correspondence: irena.arslanova@rhul.ac.uk
https://doi.org/10.1016/j.cub.2023.02.034

SUMMARY

Perception of passing time can be distorted.1 Emotional experiences, particularly arousal, can contract or
expand experienced duration via their interactions with attentional and sensory processing mechanisms.2,3
Current models suggest that perceived duration can be encoded from accumulation processes4,5 and from
temporally evolving neural dynamics.6,7 Yet all neural dynamics and information processing ensue at the
backdrop of continuous interoceptive signals originating from within the body. Indeed, phasic fluctuations
within the cardiac cycle impact neural and information processing.8–15 Here, we show that these momen-
tary cardiac fluctuations distort experienced time and that their effect interacts with subjectively experi-
enced arousal. In a temporal bisection task, durations (200–400 ms) of an emotionally neutral visual shape
or auditory tone (experiment 1) or of an image displaying happy or fearful facial expressions (experiment 2)
were categorized as short or long.16 Across both experiments, stimulus presentation was time-locked to
systole, when the heart contracts and baroreceptors fire signals to the brain, and to diastole, when the
heart relaxes, and baroreceptors are quiescent. When participants judged the duration of emotionally neu-
ral stimuli (experiment 1), systole led to temporal contraction, whereas diastole led to temporal expansion.
Such cardiac-led distortions were further modulated by the arousal ratings of the perceived facial expres-
sions (experiment 2). At low arousal, systole contracted while diastole expanded time, but as arousal
increased, this cardiac-led time distortion disappeared, shifting duration perception toward contraction.
Thus, experienced time contracts and expands within each heartbeat—a balance that is disrupted under
heightened arousal.

RESULTS Cardiac systole contracts while diastole expands the

experienced duration of emotionally neutral auditory
Across two experiments, participants first learned to discrimi- and visual stimuli
nate a short (200 ms) from a long (400 ms) reference duration Participants’ performance on the duration bisection task was
and were then asked to judge whether intermediate test dura- modeled with psychometric functions, where the point of subjec-
tions (200, 250, 300, 350, and 400 ms) were more like the short tive equality (PSE) reflects the stimulus duration at which the
or the long reference16 (Figure 1A; STAR Methods). Across both participant is equally likely to respond ‘‘short’’ or ‘‘long.’’ Thus,
experiments, stimulus presentation was time-locked to either shifts in PSE indicate relative under- or overestimation of stim-
the systolic (R + 100 ms) or the diastolic (R + 500 ms) cardiac ulus durations (Figures 2A and 2B). In experiment 1, repeated-
phase (Figure 1B). In experiment 1, participants (n = 28) per- measures ANOVA with factors modality (auditory, visual) and
formed the task for visual and auditory stimuli in separate cardiac phase (diastole, systole) on PSE values yielded a signif-
blocks, judging the duration of emotionally neutral visual im- icant main effect of cardiac phase (F(1,27) = 8.1, p = 0.01, h2p =
ages or auditory tones. In experiment 2, a new group of partic- 0.23), with stimuli presented at diastole judged, on average, 7 ms
ipants (n = 39) judged the duration of images depicting happy longer (M = 290, SD = 23) than those presented at systole (M =
or fearful facial expressions presented in a random order. In 297, SD = 26). The main effect of modality was also significant
experiment 2, after the temporal bisection task, participants (F(1,27) = 5.7, p = 0.02, h2p = 0.17) with tones judged, on
were presented with each face again for 300 ms and were average, 13 ms longer (M = 287, SD = 15) than visual stimuli
asked to rate how aroused it made them feel on an adapted (M = 300, SD = 30). The interaction between modality and car-
5-point self-assessment mannequin (SAM) scale from calm (1) diac phase was not statistically significant (F(1,27) = 1.9, p =
to aroused (5). The averaged arousal ratings are shown in 0.18, h2p = 0.06), suggesting that the same temporal distortions
Figure 1A. occurred in both visual and auditory modalities. In addition to

Current Biology 33, 1–7, April 10, 2023 ª 2023 The Author(s). Published by Elsevier Inc. 1
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
OPEN ACCESS Report

Figure 1. Task design

(A) Schematic trial structure. A stimulus was presented on the screen, and participants judged whether the presented stimulus was long or short. In experiment 1,
we manipulated the stimulus modality (auditory or visual), whereas in experiment 2, the emotional valence of the stimulus (happy or fearful face). At the end of
experiment 2, participants rated the level of arousal they experienced in response to each presented face on a scale from 1 to 5. The plot shows individual and
group-level ratings (mean and SEM).
(B) Schematic task design. Stimulus onset was time-locked to distinct cardiac phases: systole (red) and diastole (blue). The figure shows how the most
ambiguous duration (300 ms) did not overlap across the phases. The black bars represent stimulus durations. Participants first learned to discriminate between
the short (200 ms) and the long (400 ms) reference, and during bisection, they were presented with additional intermediate test durations.

PSE, just noticeable difference (JND) reflects temporal sensi- statistically significant effect (beta = 24.1, SE = 2.4, c2 = 48.7,
tivity, where smaller values indicate better discriminability of p < 0.001). Thus, the effect of the cardiac phase on duration
changes in stimulus durations (Figure 2C). In experiment 1, the perception was independent from heart rate changes.
cardiac phase did not have a significant effect (F(1,27) = 1.5,
p = 0.24, h2p = 0.05), but JND values were significantly affected Cardiac distortion of experienced duration extends to
by modality (F(1,27) = 100.0, p < 0.001, h2p = 0.79) with higher low arousal stimuli but disappears when subjective
sensitivity for tone durations (M = 22, SD = 8) as compared arousal increases
with visual durations (M = 46, SD = 17). That auditory stimuli First, experiment 2 (Figure 2D) was analyzed like experiment 1,
are perceived to last longer and result in more precise duration with a repeated-measures ANOVA coding for the cardiac phase
representation than visual stimuli is a common finding.17 The (diastole, systole) and the emotional valence of the presented
interaction between cardiac phase and modality was not statis- face (happy, fearful). There was a significant main effect of car-
tically significant (F(1,27) = 0.001, p = 0.94, h2p < 0.001). diac phase on PSE values (F(1,38) = 20.9, p < 0.001, h2p =
Because we observed heart rate changes over the course of the 0.35), with stimuli presented at diastole judged, on average,
trial, in particular, heart deceleration before and during the stim- 9 ms longer (M = 305, SD = 25) than those at systole (M = 314,
ulus presentation, we tested whether the cardiac phase effect SD = 26). The effect of valence was not statistically significant
was independent from the effects arising from the heart rate (F(1,38) = 1.3, p = 0.27, h2p = 0.03) and neither was the cardiac
changes. For that reason, we ran a mixed linear model (MLL) on phase by valence interaction (F(1,38) = 1.3, p = 0.26, h2p = 0.03;
PSE and JND values with predictors coding for cardiac phase Figure 2E). In contrast to experiment 1, the cardiac phase
and modality while controlling for heart deceleration just before extended a small but statistically significant influence on JND
and during the stimulus presentation (STAR Methods). The magni- values (F(1,38) = 4.5, p = 0.04, h2p = 0.14) with sensitivity being
tude of heart deceleration did not affect PSE values (beta = 0.1, higher during the diastolic (M = 42, SD = 17) than during the sys-
SE = 2.2, c2 = 0.0, p = 0.96). Cardiac phase retained its effect tolic phase (M = 45, SD = 17). There was no interaction
when heart deceleration was added into the model (beta = 6.8, (F(1,38) < 0.001, p = 1.0, h2p < 0.001) and no main effect of
SE = 2.4, c2 = 7.7, p = 0.006) and so did the effect of modality valence (F(1,38) = 0.58, p = 0.45, h2p = 0.001; Figure 2F).
(beta = 12.7, SE = 5.3, c2 = 5.3, p = 0.02). The interaction between Second, we considered the arousal ratings given to each pre-
cardiac phase and modality remained statistically not significant sented face (Figure 1A) because such ratings can capture the vari-
(beta = 6.6, SE = 4.8, c2 = 1.9, p = 0.17). Heart deceleration did ance in people’s affective response to the stimuli and conse-
not affect JND values either (beta = 0.6, SE = 1.5, c2 = 0.2, p = quently in the arousal-led effects on time perception.2,3 For that
0.66). When heart deceleration was added into the model, cardiac reason, we included arousal ratings for each stimulus, indepen-
phase effect remained statistically not significant (beta = 1.9, SE = dently of its valence, into the LMM model and controlled for the
1.9, c2 = 1.0, p = 0.31) as well as its interaction with modality heart deceleration like in experiment 1 (STAR Methods). PSE
(beta = 0.3, SE = 3.8, c2 = 0.0, p = 0.93). Modality retained its values were significantly affected by an interaction between

2 Current Biology 33, 1–7, April 10, 2023

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
Report OPEN ACCESS

Figure 2. PSE and JND values estimated from the psychometric functions as a function of cardiac phase, modality (experiment 1), or emotion
(experiment 2)
(A) Fitted individual (gray) and group-level (black) cumulative Gaussian functions showing the proportion of long responses as a function of test durations across
modality (auditory, visual) by cardiac (systole, diastole) conditions in experiment 1. The vertical lines show the average PSE.
(B) PSE values in experiment 1 from systolic (red) and diastolic (blue) conditions, across modality conditions. Labels in gray indicate the main effects and in-
teractions of the repeated-measures ANOVA, n = 28. The cardiac phase influenced PSEs without interaction by modality. Systolic durations were under-
estimated, whereas diastolic durations overestimation. The dots represent individual data while the boxplots reprsent group-level data (median and quartiles).
Distribution plots were created with ‘‘raincloud’’ R package.18,19
(C) Same for JND values. Only modality influenced JNDs.
(D) Fitted individual (gray) and group-level (black) cumulative Gaussian functions showing the proportion of long responses as a function of test durations across
emotion (happy, fearful) by cardiac (systole, diastole) conditions in experiment 2.
(E) PSE values in experiment 2 from systolic (red) and diastolic (blue) conditions, across emotion conditions. Labels in gray indicate the main effects and in-
teractions of the repeated-measures ANOVA, n = 39. The cardiac phase influenced the PSEs in the same way as in experiment 1 without an interaction by emotion.
(F) Same for the JND values. The cardiac phase influenced JNDs without interaction by emotion. Participants were more sensitive to duration differences at
diastole compared with systole. n.s., non-significant; **p < 0.001; *p < 0.05.

cardiac phase and arousal ratings (beta = 5.2, SE = 2.3, c2 = 5.3, experienced duration, while diastole expanded it. Experiment 2
p = 0.02). The main effect of arousal ratings was not significant replicated the observed cardiac-led temporal distortion and
(beta = 0.9, SE = 1.7, c2 = 0.3, p = 0.59), but the cardiac phase re- showed that it was further modulated by the subjectively experi-
tained its statistical significance (beta = 9.2, SE = 2.3, c2 = 16.1, enced arousal in response to the presented facial expressions.
p < 0.001). A significant cardiac phase by arousal interaction was As the experienced arousal increased, the relative cardiac-led
followed up with simple slopes analysis (Figure 3), which showed temporal contraction-expansion was disrupted, biasing duration
that the opposing effects of systole and diastole on time percep- perception toward contraction.
tion were present for low and average arousal ratings (beta = There has been a growing interest in the idea of embodied
14.4, SE = 3.2, t = 4.5, p < 0.001 and beta = 9.2, SE = 2.3, t = time, according to which experienced time is influenced by phys-
4.1, p < 0.001, respectively) but disappeared when arousal ratings iological, interoceptive, signals from the body.20,21 For example,
increased (beta = 4.1, SE = 3.2, t = 1.3, p = 0.21). For JND values, past studies have examined interoceptive attention effects,22
none of the predictors and their interactions made a statistically the cortical processing of cardiac signals,23 and heart rate
significant contribution. The main effect of the cardiac phase changes.24–27 Specifically, asking participants to focus on bodily
was no longer significant (beta = 2.8, SE = 2.1, c2 = 1.8, p = 0.18). sensations exaggerates the emotional distortions of time so that
negative experiences seem to last even longer and positive ones
DISCUSSION seem to pass even quicker compared with when participants
focus externally.22 In addition, the strength of cortical processing
Experiment 1 showed that when timing emotionally neutral audi- of afferent cardiac signals indexed by heart-evoked potential
tory and visual stimuli, cardiac systole contracted the (HEP) is modulated by the over- and underestimation of elapsed

Current Biology 33, 1–7, April 10, 2023 3

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
OPEN ACCESS Report
Figure 3. Simple slopes analysis breaking
down the cardiac phase by subjective arousal
interaction on PSE values in experiment 2
A linear mixed model was run on the PSE values
modeling the effects of the cardiac phase (systole,
diastole), subjective arousal, and changes in heart
rate. The cardiac effect on PSEs was concentrated
at low and mean levels of arousal but disappeared
at high arousal levels. Arousal ratings were
mean-centered. The lines represent the slopes and
the shading represents the 95% confidence band.
n.s., non-significant; **p < 0.001; *p < 0.05.

similar to how time dilates post-saccade to

compensate for the contraction caused by
the saccade.31
However, beyond this sensory modula-
tion account, it is also possible that cardiac
signals directly affect the temporal accu-
mulation processes5 via the physiological
arousal they convey. Classical pace-
maker-accumulation models conceptu-
alize timing as an accumulation of internally
time so that lower amplitude in HEPs accompany duration over- produced ticks.32,33 To highlight the embodied nature of such
estimation.23 Furthermore, when performing daily activities par- temporal accumulation, the insular cortex, considered the main
ticipants tend to report quicker passage of time during periods of interoceptive hub in the brain,34 has been found to be consis-
heightened heart rate25 (but not all studies have found significant tently engaged during duration perception tasks35,36 and is
influence of heart rate on temporal judgments26,27). argued to underpin the accumulation of ‘‘how do I feel’’ moments
Although these findings correlate interoceptive processes in across time, essentially defining the subjective passage of
time perception, these studies demonstrated in a more mecha- time.20,21 Importantly, the rate of such accumulation can be influ-
nistic way how common time distortions1–3—the contraction enced by arousal.2 It is often assumed that the systolic phase of
and the expansion of time—arise from the phasic modulations the cardiac cycle represents a state of simulated heightened
within each heartbeat. Ascending cardiac signals from the baro- arousal, in contrast to the diastole.37 On that account, the neural
receptors provide the brain with continuous information about encoding of cardiac signals could directly affect temporal accu-
the heart rate and blood pressure changes. However, barore- mulation processes, as the encoding of arousing physiological
ceptor activity is maximal during the systolic period of the car- states at the systole would lead to temporal contraction, fol-
diac cycle, and although some baroceptor activity may still be lowed by a temporal elongation during the diastolic phase.
present during the diastolic period, it is reduced or ceased rela- This is consistent with the finding that a reduction in sympathetic
tive to the systolic period.28 Thus, the method of time-locking tone induced by clonidine injection resulted in a subjective slow-
stimulus presentation to specific cardiac phases, while not ing of time.38 Combining the present paradigm with electroen-
necessarily reflecting real life perception, can be used to approx- cephalography (EEG) methods could distinguish between these
imate the causal role of ascending cardiac signals in shaping two accounts. Specifically, sensory modulation accounts would
temporal perception, as the only factor that varies across the predict that cardiac signals attenuate the amplitude of sensory
two conditions is the timing of identical stimuli relative to the event-related potentials (ERPs),8–10 whereas temporal accumu-
ECG R-wave. lation accounts would predict that cardiac signals either atten-
A fundamental question pertains to how temporal distortions uate the amplitude of the contingent negative variation (CNV)39
can arise from the ascending cardiac signals. One potential or enhance the ERP response at the offset of the stimulus.5 While
explanation stems from the observation that non-affective sen- the two accounts are not mutually exclusive, meaning that car-
sory processing is generally attenuated during the systole due diac signals likely affect a range of components, it would be
to the neural noise produced by the baroreceptor output.8–10,29 important to examine which of these modulations (sensory
According to the coding efficiency accounts of temporal pro- ERP, CNV, or offset ERP) is ultimately predictive of the subjective
cessing30 more efficient or enhanced sensory processing pre- contraction of the stimulus duration.39
dicts temporal dilation, whereas suppressed sensory processing Interestingly, the opposing temporal distortions—contraction
predicts temporal contraction. On this view, experienced time is and expansion—within each heartbeat imply that over multiple
encoded from the evolving neural networks during perception.6,7 heartbeats, such distortions would average out to produce a
Accordingly, temporal contraction during the systole may have duration estimate that is close to the veridical duration of the
resulted from a periodic sensory attenuation. The relative tempo- stimulus. We show that the opposing effect of systole and dias-
ral expansion during the diastole may then act as a compen- tole on duration perception breaks down, however, as the sub-
sating mechanism that counteracts the systolic time contraction, jectively experienced arousal increased, skewing the duration

4 Current Biology 33, 1–7, April 10, 2023

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
Report OPEN ACCESS

representation toward contraction. Although arousal has been AUTHOR CONTRIBUTIONS

associated with temporal dilation,27,40–43 some studies find
Conceptualization, M.T., V.K., and I.A.; methodology, M.T. and V.K.; formal
that if arousal arises due to anxiety or uncertainty it can lead to
analysis, I.A.; investigation, I.A.; data curation, I.A.; writing – original draft,
temporal contraction.44,45 Notably, we did not observe a main ef- I.A.; writing – review & editing, I.A. and M.T.; visualization, I.A.; supervision,
fect of arousal or valence on perceived stimulus durations in M.T.; project administration, I.A.; funding acquisition, M.T.
experiment 2. Rather, we show that subjectively felt arousal
modulated how the cardiac effects shaped perceived stimulus DECLARATION OF INTERESTS
durations, biasing temporal representation at the diastolic phase
toward contraction. Static facial stimuli do not always induce The authors declare no competing interests.
temporal distortions,46 thus future studies could use more
arousal-inducing paradigms involving, for example, unexpected Received: November 22, 2022
painful stimulation.44 Revised: January 6, 2023
Accepted: February 10, 2023
This study demonstrated how fluctuations within the cardiac
Published: March 10, 2023
phases together with the subjectively experienced arousal affect
perceptual temporal performance. Yet unexpected arousal can
REFERENCES
also reduce the metacognitive confidence in perceptual deci-
sions.47 A recent computational model of phasic cardiac influ- 1. Eagleman, D.M. (2008). Human time perception and its illusions. Curr.
ences on perception suggests that systole reduces the confi- Opin. Neurobiol. 18, 131–136. https://doi.org/10.1016/j.conb.2008.
dence in exteroceptive sensory channels.48 However, direct 06.002.
evidence as to whether the ascending cardiac signals directly 2. Lake, J.I., LaBar, K.S., and Meck, W.H. (2016). Emotional modulation of in-
impact metacognitive processing is still lacking. Incorporating terval timing and time perception. Neurosci. Biobehav. Rev. 64, 403–420.
confidence judgements within the current paradigm would thus https://doi.org/10.1016/j.neubiorev.2016.03.003.
be a fruitful avenue for future studies to help disentangle the car- 3. Droit-Volet, S., and Meck, W.H. (2007). How emotions colour our percep-
diac-driven temporal perceptual distortions from distortions in tion of time. Trends Cogn. Sci. 11, 504–513. https://doi.org/10.1016/j.tics.
2007.09.008.
metacognition.
Overall, our findings are in line with recent research showing 4. Roseboom, W., Fountas, Z., Nikiforou, K., Bhowmik, D., Shanahan, M.,
and Seth, A.K. (2019). Activity in perceptual classification networks as a
that interoceptive signals, particularly cardiac signals from the
basis for human subjective time perception. Nat. Commun. 10, 267.
heart, shape neural activity and perception8–15 and that the intri- https://doi.org/10.1038/s41467-018-08194-7.
cate interplay between the brain and the heart maintains our
5. Ofir, N., and Landau, A.N. (2022). Neural signatures of evidence accumu-
emotional experiences,49 which in turn modulate how time is lation in temporal decisions. Curr. Biol. 32, 4093–4100.e6. https://doi.org/
experienced.2,3 Thus, the current results strengthen the view 10.1016/j.cub.2022.08.006.
that time processing is an embodied process that arises from 6. Zhou, S., Masmanidis, S.C., and Buonomano, D.V. (2020). Neural se-
the continuous integration between exteroceptive and intero- quences as an optimal dynamical regime for the readout of time. Neuron
ceptive processes. 108, 651–658.e5. https://doi.org/10.1016/j.neuron.2020.08.020.
7. Tsao, A., Yousefzadeh, S.A., Meck, W.H., Moser, M.B., and Moser, E.I.
(2022). The neural bases for timing of durations. Nat. Rev. Neurosci. 23,
STAR+METHODS 646–665. https://doi.org/10.1038/s41583-022-00623-3.
8. Schulz, A., Vögele, C., Bertsch, K., Bernard, S., Münch, E.E., Hansen, G.,
Detailed methods are provided in the online version of this paper Naumann, E., and Scha €chinger, H. (2020). Cardiac cycle phases affect
auditory-evoked potentials, startle eye blink and pre-motor reaction times
and include the following:
in response to acoustic startle stimuli. Int. J. Psychophysiol. 157, 70–81.
d KEY RESOURCES TABLE https://doi.org/10.1016/j.ijpsycho.2020.08.005.

d RESOURCE AVAILABILITY 9. Al, E., Iliopoulos, F., Forschack, N., Nierhaus, T., Grund, M., Motyka, P.,
Gaebler, M., Nikulin, V.V., and Villringer, A. (2020). Heart–brain interactions
B Lead contact
shape somatosensory perception and evoked potentials. Proc. Natl.
B Materials availability
Acad. Sci. USA 117, 10575–10584. https://doi.org/10.1073/pnas.
B Data and code availability 1915629117.
d EXPERIMENTAL MODEL AND SUBJECT DETAILS 10. Al, E., Iliopoulos, F., Nikulin, V.V., and Villringer, A. (2021). Heartbeat and
d METHOD DETAILS somatosensory perception. Neuroimage 238, 118247. https://doi.org/10.
B Design and Procedure 1016/j.neuroimage.2021.118247.
B Stimuli 11. Gray, M.A., Beacher, F.D., Minati, L., Nagai, Y., Kemp, A.H., Harrison,
d QUANTIFICATION AND STATISTICAL ANALYSIS N.A., and Critchley, H.D. (2012). Emotional appraisal is influenced by car-
B ECG processing diac afferent information. Emotion 12, 180. https://doi.org/10.1037/
B ANOVA analyses a0025083.
B Linear mixed model (LMM) analyses 12. Garfinkel, S.N., Minati, L., Gray, M.A., Seth, A.K., Dolan, R.J., and
Critchley, H.D. (2014). Fear from the heart: sensitivity to fear stimuli de-
pends on individual heartbeats. J. Neurosci. 34, 6573–6582. https://doi.
ACKNOWLEDGMENTS org/10.1523/JNEUROSCI.3507-13.2014.
13. Salomon, R., Ronchi, R., Dönz, J., Bello-Ruiz, J., Herbelin, B., Martet, R.,
M.T. is supported by the European Research Council Consolidator Grant Faivre, N., Schaller, K., and Blanke, O. (2016). The insula mediates access
(ERC-2016-CoG-724537) under the FP7 for the INtheSELF project. to awareness of visual stimuli presented synchronously to the heartbeat.

Current Biology 33, 1–7, April 10, 2023 5

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
OPEN ACCESS Report
J. Neurosci. 36, 5115–5127. https://doi.org/10.1523/JNEUROSCI.4262- 32. Treisman, M. (1963). Temporal discrimination and the indifference interval.
15.2016. Implications for a model of the ‘‘internal clock’’. Psychol. Monogr. 77,
14. Ren, Q., Marshall, A.C., Kaiser, J., and Schütz-Bosbach, S. (2022). 1–31. https://doi.org/10.1037/h0093864.
Multisensory integration of anticipated cardiac signals with visual targets 33. Ivry, R.B., and Schlerf, J.E. (2008). Dedicated and intrinsic models of time
affects their detection among multiple visual stimuli. NeuroImage 262, perception. Trends Cogn. Sci. 12, 273–280. https://doi.org/10.1016/j.tics.
119549. https://doi.org/10.1016/j.neuroimage.2022.119549. 2008.04.002.
15. Marshall, A.C., Gentsch-Ebrahimzadeh, A., and Schütz-Bosbach, S. 34. Craig, A.D. (2009). How do you feel—now? The anterior insula and human
(2022). From the inside out: interoceptive feedback facilitates the integra- awareness. Nat. Rev. Neurosci. 10, 59–70. https://doi.org/10.1038/
tion of visceral signals for efficient sensory processing. NeuroImage 251, nrn2555.
119011. https://doi.org/10.1016/j.neuroimage.2022.119011. 35. Nani, A., Manuello, J., Liloia, D., Duca, S., Costa, T., and Cauda, F. (2019).
16. Kopec, C.D., and Brody, C.D. (2010). Human performance on the temporal The neural correlates of time: a meta-analysis of neuroimaging studies.
bisection task. Brain Cogn. 74, 262–272. https://doi.org/10.1016/j.bandc. J. Cogn. Neurosci. 31, 1796–1826. https://doi.org/10.1162/jocn_a_01459.
2010.08.006. 36. Wittmann, M., Simmons, A.N., Aron, J.L., and Paulus, M.P. (2010).
17. Grondin, S. (2010). Timing and time perception: a review of recent behav- Accumulation of neural activity in the posterior insula encodes the pas-
ioral and neuroscience findings and theoretical directions. Atten. Percept. sage of time. Neuropsychologia 48, 3110–3120. https://doi.org/10.1016/
Psychophys. 72, 561–582. https://doi.org/10.3758/APP.72.3.561. j.neuropsychologia.2010.06.023.

18. Allen, M., Poggiali, D., Whitaker, K., Marshall, T.R., and Kievit, R.A. (2021). 37. Critchley, H.D., and Garfinkel, S.N. (2017). Interoception and emotion.
Raincloud plots: a multi-platform tool for robust data visualization. Curr. Opin. Psychol. 17, 7–14. https://doi.org/10.1016/j.copsyc.2017.
Wellcome Open Res. 4, 63. https://doi.org/10.12688/wellcomeopenres. 04.020.
15191.2. 38. Coull, J.T. (2004). fMRI studies of temporal attention: allocating attention
within, or towards, time. Brain Res. Cogn. Brain Res. 21, 216–226.
19. Allen, M., Poggiali, D., Whitaker, K., Marshall, T.R., and Kievit, R. (2018).
https://doi.org/10.1016/j.cogbrainres.2004.02.011.
RainCloudPlots Tutorials and Codebase, Version V1.1. Zenodo. https://
doi.org/10.5281/zenodo.3368186. 39. Damsma, A., Schlichting, N., and van Rijn, H. (2021). Temporal context
actively shapes EEG signatures of time perception. J. Neurosci. 41,
20. Wittmann, M. (2013). The inner sense of time: how the brain creates a rep-
4514–4523. https://doi.org/10.1523/JNEUROSCI.0628-20.2021.
resentation of duration. Nat. Rev. Neurosci. 14, 217–223. https://doi.org/
10.1038/nrn3452. 40. Ogden, R.S., Henderson, J., McGlone, F., and Richter, M. (2019). Time
distortion under threat: sympathetic arousal predicts time distortion only
21. Craig, A.D. (2009). Emotional moments across time: a possible neural ba-
in the context of negative, highly arousing stimuli. PLoS ONE 14,
sis for time perception in the anterior insula. Philos. Trans. R. Soc. Lond. B
e0216704. https://doi.org/10.1371/journal.pone.0216704.
Biol. Sci. 364, 1933–1942. https://doi.org/10.1098/rstb.2009.0008.
41. Appelqvist-Dalton, M., Wilmott, J.P., He, M., and Simmons, A.M. (2022).
22. Pollatos, O., Laubrock, J., and Wittmann, M. (2014). Interoceptive focus
Time perception in film is modulated by sensory modality and arousal.
shapes the experience of time. PLoS ONE 9, e86934. https://doi.org/10.
Atten. Percept. Psychophys. 84, 926–942. https://doi.org/10.3758/
1371/journal.pone.0086934.
s13414-022-02464-9.
23. Richter, F., and Ibáñez, A. (2021). Time is body: multimodal evidence of 42. Mella, N., Conty, L., and Pouthas, V. (2011). The role of physiological
crosstalk between interoception and time estimation. Biol. Psychol. 159, arousal in time perception: psychophysiological evidence from an emotion
108017. https://doi.org/10.1016/j.biopsycho.2021.108017. regulation paradigm. Brain Cogn. 75, 182–187. https://doi.org/10.1016/j.
24. Meissner, K., and Wittmann, M. (2011). Body signals, cardiac awareness, bandc.2010.11.012.
and the perception of time. Biol. Psychol. 86, 289–297. https://doi.org/10. 43. Harjunen, V.J., Spape, M., and Ravaja, N. (2022). Anticipation of aversive
1016/j.biopsycho.2011.01.001. visual stimuli lengthens perceived temporal duration. Psychol. Res. 86,
25. Ogden, R.S., Dobbins, C., Slade, K., McIntyre, J., and Fairclough, S. 1230–1238. https://doi.org/10.1007/s00426-021-01559-6.
(2022). The psychophysiological mechanisms of real-world time experi- 44. Sarigiannidis, I., Grillon, C., Ernst, M., Roiser, J.P., and Robinson, O.J.
ence. Sci. Rep. 12, 12890. https://doi.org/10.1038/s41598-022-16198-z. (2020). Anxiety makes time pass quicker while fear has no effect.
26. Suárez-Pinilla, M., Nikiforou, K., Fountas, Z., Seth, A.K., and Roseboom, Cognition 197, 104116. https://doi.org/10.1016/j.cognition.2019.104116.
W. (2019). Perceptual content, not physiological signals, determines 45. Cui, Q., Liu, M., Liu, C.H., Long, Z., Zhao, K., and Fu, X. (2021).
perceived duration when viewing dynamic, natural scenes. Collabra Unpredictable fearful stimuli disrupt timing activities: evidence from
Psychol. 5, 55. https://doi.org/10.1525/collabra.234. event-related potentials. Neuropsychologia 163, 108057. https://doi.org/
27. Schwarz, M.A., Winkler, I., and Sedlmeier, P. (2013). The heart beat does 10.1016/j.neuropsychologia.2021.108057.
not make us tick: the impacts of heart rate and arousal on time perception. 46. Fayolle, S.L., and Droit-Volet, S. (2014). Time perception and dynamics of
Atten. Percept. Psychophys. 75, 182–193. https://doi.org/10.3758/ facial expressions of emotions. PLoS One 9, e97944. https://doi.org/10.
s13414-012-0387-8. 1371/journal.pone.0097944.
28. Bronk, D.W., and Stella, G. (1932). Afferent impulses in the carotid sinus 47. Allen, M., Frank, D., Schwarzkopf, D.S., Fardo, F., Winston, J.S., Hauser,
nerve I. The relation of the discharge from single end organs to arterial T.U., and Rees, G. (2016). Unexpected arousal modulates the influence of
blood pressure. J. Cell. Comp. Physiol. 1, 113–130. https://doi.org/10. sensory noise on confidence. eLife 5, e18103. https://doi.org/10.7554/
1002/jcp.1030010110. eLife.18103.
29. Skora, L.I., Livermore, J.J.A., and Roelofs, K. (2022). The functional role of 48. Allen, M., Levy, A., Parr, T., and Friston, K.J. (2022). In the body’s eye: the
cardiac activity in perception and action. Neurosci. Biobehav. Rev. 137, computational anatomy of interoceptive inference. PLoS Comput. Biol. 18,
104655. https://doi.org/10.1016/j.neubiorev.2022.104655. e1010490. https://doi.org/10.1371/journal.pcbi.1010490.
30. Kruijne, W., Olivers, C.N.L., and van Rijn, H. (2021). Neural repetition sup- 49. Candia-Rivera, D., Catrambone, V., Thayer, J.F., Gentili, C., and Valenza,
pression modulates time perception: evidence from electrophysiology and G. (2022). Cardiac sympathetic-vagal activity initiates a functional brain–
pupillometry. J. Cogn. Neurosci. 33, 1230–1252. https://doi.org/10.1162/ body response to emotional arousal. Proc. Natl. Acad. Sci. USA 119.
jocn_a_01705. e2119599119. https://doi.org/10.1073/pnas.2119599119.
31. Yarrow, K., Haggard, P., Heal, R., Brown, P., and Rothwell, J.C. (2001). 50. Brainard, D.H. (1997). The psychophysics toolbox. Spat. Vis. 10, 433–436.
Illusory perceptions of space and time preserve cross-saccadic percep- 51. Ebner, N.C., Riediger, M., and Lindenberger, U. (2010). FACES—A data-
tual continuity. Nature 414, 302–305. https://doi.org/10.1038/35104551. base of facial expressions in young, middle-aged, and older women and

6 Current Biology 33, 1–7, April 10, 2023

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
Report OPEN ACCESS

men: development and validation. Behav. Res. Methods 42, 351–362. 60. Glass, G.V., Peckham, P.D., and Sanders, J.R. (1972). Consequences of
https://doi.org/10.3758/BRM.42.1.351. failure to meet assumptions underlying the fixed effects analyses of vari-
52. Penney, T.B., and Cheng, X. (2018). Duration bisection: a user’s guide. In ance and covariance. Rev. Educ. Res. 42, 237–288. https://doi.org/10.
Timing and Time Perception: Procedures, Measures, & Applications (Brill), 3102/00346543042003237.
pp. 98–127.
61. Bates, D., Maechler, M., Bolker, B., and Walker, S. (2015). Fitting linear
53. DeBruine, L. (2021). faux: simulation for Factorial Designs R package mixed-effects models using lme4. J. Stat. Softw. 67, 1–48. https://doi.
version 1.1.0. Zenodo. https://doi.org/10.5281/zenodo.2669586. org/10.18637/jss.v067.i01.
54. Azevedo, R.T., Garfinkel, S.N., Critchley, H.D., and Tsakiris, M. (2017).
Cardiac afferent activity modulates the expression of racial stereotypes. 62. Kuznetsova, A., Brockhoff, P.B., and Christensen, R.H.B. (2017). lmerTest
Nat. Commun. 8, 13854. https://doi.org/10.1038/ncomms13854. package: tests in linear mixed effects models. J. Stat. Softw. 82, 1–26.
55. Bradley, M.M., and Lang, P.J. (1994). Measuring emotion: the self-assess- https://doi.org/10.18637/jss.v082.i13.
ment manikin and the semantic differential. J. Behav. Ther. Exp.
63. Singmann, H., Bolker, B., Westfall, J., Aust, F., and Ben-Shachar, M.
Psychiatry 25, 49–59. https://doi.org/10.1016/0005-7916(94)90063-9.
(2022). afex: analysis of factorial experiments. R package version 1.1-1.
56. Linares, D., and López-Moliner, J. (2016). quickpsy: an R package to fit https://CRAN.R-project.org/package=afex.
psychometric functions for multiple groups. R J. 8, 122–131. https://doi.
org/10.32614/RJ-2016-008. 64. Lüdecke, D., Ben-Shachar, M., Patil, I., Waggoner, P., and Makowski, D.
57. Park, H.D., Correia, S., Ducorps, A., and Tallon-Baudry, C. (2014). (2021). performance: an R package for assessment, comparison and
Spontaneous fluctuations in neural responses to heartbeats predict visual testing of statistical models. J. Open Source Softw. 6, 3139. https://doi.
detection. Nat. Neurosci. 17, 612–618. https://doi.org/10.1038/nn.3671. org/10.21105/joss.03139.
58. Motyka, P., Grund, M., Forschack, N., Al, E., Villringer, A., and Gaebler, M.
65. Bauer, D.J., and Curran, P.J. (2005). Probing interactions in fixed and
(2019). Interactions between cardiac activity and conscious somatosen-
multilevel regression: inferential and graphical techniques. Multivariate
sory perception. Psychophysiology 56, e13424. https://doi.org/10.1111/
Behav. Res. 40, 373–400. https://doi.org/10.1207/s15327906mbr4003_5.
psyp.13424.
59. Lawrence, M.A. (2016). ez: easy analysis and visualization of factorial ex- 66. Long, J.A. (2019). interactions: comprehensive, user-friendly toolkit for
periments. R package, version 4.4-0. https://CRAN.R-project.org/ probing interactions. R package version 1.1.0. https://cran.r-project.org/
package=ez. package=interactions.

Current Biology 33, 1–7, April 10, 2023 7

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
OPEN ACCESS Report

STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Deposited data
Raw behavioral data This paper https://osf.io/4dn6q/
Participant-level visualizations This paper https://osf.io/4dn6q/
Software and algorithms
MATLAB 2019B The MathWorks http://www.mathworks.com/products/matlab/; RRID:
SCR_001622
Psychtoolbox Breinard50 http://psychtoolbox.org/; RRID: SCR_002881
Labchart 8 Pro AD Instruments https://www.adinstruments.com/products/labchart;
RRID: SCR_017551
Experiment code This paper https://osf.io/4dn6q/
R version 4.2.1 (2022 - 06 - 26) R project http://www.r-project.org/; RRID: SCR_001905
R Studio version 2022.07.2 Build 576 RStudio http://www.rstudio.com/; RRID: SCR_000432
Analysis code This paper https://osf.io/4dn6q/
Other
FACES database Ebner et al.51 https://faces.mpdl.mpg.de/imeji/
Chosen and processed stimuli from the FACES Ebner et al.51 https://osf.io/4dn6q/
database
Auditory stimuli Penney & Cheng52 https://osf.io/4dn6q/

RESOURCE AVAILABILITY

Lead contact
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Irena Arslanova (irena.
arslanova@rhul.ac.uk).

Materials availability
This study did not generate new materials.

Data and code availability

d De-identified behavioral data have been deposited at OSF. They are publicly available as of the date of publication. DOIs are
listed in the key resources table.
d All original code for running the experiments as well as the analysis reported in this paper has been deposited at OSF and is
publicly available as of the date of publication. DOIs are listed in the key resources table.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

For Experiment 1, 36 participants were recruited. Data from 8 participants was excluded from the final analysis, resulting in n = 28 (5
men, aged between 18 to 40, average age 22).
For Experiment 2, 45 participants were recruited. Data from 6 participants was excluded from the final analysis, resulting in n = 39 (4
men, aged between 18 to 40, average age 23). Psychometric fits and timing of the stimuli relative to cardiac phase for each participant
(including those that were excluded from the analysis) can be found at the DOI listed in the key resources table. The criterion for
participant exclusion was as follows:

(1) More than 20% of trials were flagged in one cardiac phase condition relative to the other. The trial was flagged if: (a) Inter-beat
interval (IBI) was less than 500 ms or more than 1200 ms, which indicated impossibly low or high heart rate. (b) The actual start
of the stimulus was considerably displaced relative to the systolic or diastolic phase. (c) A wrong key was pressed, or no

e1 Current Biology 33, 1–7.e1–e4, April 10, 2023

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
Report OPEN ACCESS

response was given within the 2 second time limit. (d) The stimuli presented at the diastole exceeded the IBI by at least 200 ms,
which made it overlap with the systolic phase of the next R wave. This happened if the IBI was too short.
(2) The psychometric function fit to behavioral data did not reach a above chance level (50%) performance in any one of the con-
ditions

However, a re-analysis of the data with the excluded participants did not change the main inferences of the paper.
The sample size for Experiment 1 was based on previous studies examining heart-related effects on time perception24 and cardiac
phase effects on perception.12
The sample size for Experiment 2 was based on the effect sizes obtained in the visual task of the Experiment 1. The prediction was
that the valence of the visual stimulus (happy vs. fearful face) will modulate the cardiac phase effect. Specifically, that the fearful face
will reverse the cardiac effect relative to the happy face. We simulated data within a 2 (cardiac phase: systole, diastole) by 2 (emotion:
happy, fear) within-subject ANOVA using the ‘‘faux’’ package53 in R. We assumed that the cardiac effect under happy condition will
be equivalent to the one in the Experiment 1 (PSE = 290 ms at diastole, PSE = 300 ms at systole). Thus, we estimated the sample size
required to find an interaction, whereby PSE under fearful condition will be smaller by 10 ms at systole (PSE = 280 ms at systole)
compared to the diastole (PSE = 290 ms at diastole). SDs were kept at 30 ms and correlations at 0.5. Simulations (1000 iterations)
were run with sample sizes of 30, 35, 40, 45, 50, 55, and 60 until the power of 80% was achieved. At n = 40, we reached a power of.83
(h2p =.20).
Participants were recruited from the Royal Holloway’s community and were compensated for their time with either money (£10 per
hour) or class credit. The study was approved by Royal Holloway’s Ethics Committee in accordance with the 1964 Declaration of
Helsinki, and signed informed consents were obtained from all participants.

METHOD DETAILS

Design and Procedure

The task was identical across the two experiments, but the presented stimuli differed (see Stimuli). The task was a temporal
bisection task.16 Participants first learned to distinguish between the short (200 ms) and the long (400 ms) stimulus duration
with feedback provided after each trial (20 trials). After completing the learning phase with less than 6 errors (in case error
rate was higher than 6, they repeated 10 learning trials), participants were presented with 5 test durations (200, 250, 300,
350, 400 ms) and, on each trial, were asked to judge whether the presented test duration was more like the short or the
long reference presented during the learning. They responded by pressing A or K keyboard keys with the left and right index
fingers, respectively. The correspondence between the A/K keys and SHORT/LONG response was counterbalanced across
participants. Participants had 2 seconds to respond. In each condition, test durations were repeated 8, 16, 24, 16, and 8 times,
respectively, so that a third of trials was at the bisection point (300 ms). Stimulus durations were controlled with Psychophysics
Toolbox MATLAB extension.50
The onset of test durations was time-locked to either the systolic or the diastolic cardiac phases. Specifically, an electrocar-
diogram (ECG) was recorded throughout the experiment with three disposable ECG electrodes placed in a modified lead I chest
configuration: two electrodes were positioned underneath the left and right collarbone and another on the participant’s lower
back on the left side. The ECG signal was recorded with a Powerlab 8/35 (Powerlab, ADInstruments, http://www.
adinstruments.com/) using Labchart 8 Pro software. The sampling rate was 1,000 Hz and a hardware band-pass filter (Bio
Amp 132) between 0.3 and 1,000 Hz was applied. Heartbeats were detected online with a hardware-based function (fast output
response), which identifies the ECG R-wave with a delay smaller than 1 ms by detecting when the amplitude exceeds an indi-
vidually defined threshold. The stimuli were time-locked in a way that at least 50% of the longest stimulus (400 ms) was con-
tained within the respective cardiac phase. The systolic phase is typically defined from approximately 200 to 400 ms post
R-wave while the diastolic phase from approximately 500 to 800 ms post R-wave.54 Thus, our stimuli were presented at
100 ms post R-wave in systolic condition, and at 500 ms post R-wave in diastolic condition. To account for the fact that
wait times in the diastolic condition were always longer than systolic condition, 400 ms was removed from the inter-trial-interval
(ITI) preceding the diastolic condition. The total ITI ranged randomly between 2 to 3 seconds.
In Experiment 1, participants performed the task in two modalities – visual and auditory – in a counterbalanced order, so that half of
the participants performed the visual task first and the other half performed the auditory task first. Both tasks started with a separate
learning phase. Both tasks were divided into 4 successive blocks with breaks in-between. The reference short and long durations
were presented again (3 times each) at the beginning of each block to reduce memory demands. The cardiac phase condition
was randomized within each block.
In Experiment 2, instead of the modality manipulation, the valence of the visual stimuli was manipulated, so that participants judged
the duration of faces showing happy or fearful expressions. In the learning phase, a scrambled face served as a timed stimulus. Both
valence and cardiac phase conditions were randomized across 6 blocks with reference short and long durations presented again (for
3 times each) at the beginning of each block to reduce memory demands. After temporal bisection task, participants were presented
with each of 12 faces from the experiment for 300 ms. For each face they rated how aroused it made them feel on an adapted 5-point
Self-Assessment Mannequin (SAM)55 from calm (1) to aroused (5).

Current Biology 33, 1–7.e1–e4, April 10, 2023 e2

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
OPEN ACCESS Report

Stimuli
In Experiment 1, the auditory stimuli comprised of 300 Hz (4410 sampling Hz) neutral tones with rising and falling phases of 10 ms,
played at a 70% volume via headphones. The tones were created with audiowrite function in MATLAB according to the script pro-
vided by Penney and Cheng.52 The resulting tones can be found at the DOI listed in the key resources table. The visual stimulus was a
black star shape with a white outline presented at the center of the screen. The stimuli can be found at the DOI listed in the key re-
sources table.
In Experiment 2, the stimuli were grayscaled and cropped images of 6 people (3 men, 3 women) chosen from validated FACES
database.51 Each person was presented with either a happy or a fearful expression. The stimuli can be found at the DOI listed in
the key resources table.

QUANTIFICATION AND STATISTICAL ANALYSIS

For each participant, the proportion of ‘‘long’’ responses as a function of test durations was fitted with Gaussian cumulative psycho-
metric functions using the ‘‘quickpsy’’ R package,56 separately for each condition (in Experiment 1: modality & cardiac phase; in
Experiment 2: emotion & cardiac phase). The goodness of fits was evaluated with deviance. By re-fitting the psychometric functions
with 1000 bootstrap samples, we used the distribution of bootstrapped deviances to calculate the p-value of obtaining a deviance
greater than that of the original data. In Experiment 1, none of the participants in any of the conditions showed a statistically significant
deviance (mean p =.83, SD =.23). In Experiment 2, three participants (P24, P27, P39) showed a significant deviance in one of the
conditions, but none of the other participants showed significant deviance (mean p =.76, SD =.26). However, the results remain
the same if those participants were excluded from the analyses. We decided to retain these three participants in the analysis because
they reached a sufficient performance (i.e., above a chance level) and thus were capable of distinguishing between the stimulus
durations.
Once fitted, the point of subjective equality (PSE) and the just noticeable difference (JND) were extracted. The PSE is the duration
value at which the participant is equally likely to classify the duration as short or long. Thus, shifts in PSE indicate temporal under- or
overestimation. For example, lower PSE values suggest temporal overestimation, because participants would be predicted to begin
responding long during physically short test durations. The JND reflects the temporal sensitivity, whereby smaller values indicate
better discriminability of changes in stimulus durations.

ECG processing
The timepoints of R-wave peaks were recorded online with a hardware-based function (fast output response), which identifies the
ECG R-wave with a delay smaller than 1 ms by detecting when the amplitude exceeds an individually defined threshold. These time-
points were plotted onto the raw ECG signal for visual checks. The R-peak segments were divided into their respective trials. For each
trial, the correspondence between the stimulus onset and the R peak onset was checked, so that trials where the stimulus occurred
outside the predefined cardiac phase time-window were flagged. In addition, a trial was flagged if the inter-beat-interval (IBI) between
any of the R-peaks was less than 500 ms or more than 1200 ms, which indicated impossibly low or high heart rate. These flagged trials
were excluded from the analysis. Further, we flagged additional trials where the duration of the stimulus exceeded the average IBI on
that trial by more than 200 ms, which indicated a considerable overlap with the systolic phase of the next R-wave. For most partic-
ipants, this affected only the longest stimuli (400 ms), and thus were retained for the analysis, because removing these trials distorted
the psychometric fitting. If anything, the fact that some diastolic trials may have not been purely diastolic would result in underesti-
mation of the true cardiac effect.
To ensure that the cardiac phase influence on duration perception was independent of any heart rate changes over the course of a
trial, IBI lengths at five consecutive IBIs – two prior the stimulus presentation (during ITI: S-2 & S-1), during the stimulus presentation
(S), and two after the stimulus (during the response: S+1, S+2) – were calculated.57,58 For both experiments, across all conditions, the
heart decelerated prior and during the stimulus presentation (from S-2 to S+1), followed by a short acceleration during the response
(from S+1 to S+2). For each trial, we took the IBI length difference between S-2 and S+1, which indexed the magnitude of the heart’s
deceleration during the stimulus presentation. We averaged heart’s deceleration for each duration and response category and added
it as a control variable in the linear mixed models (see Linear mixed model (LMM) analyses below).

ANOVA analyses
First, PSE and JND values were analyzed with a type-3 repeated-measures ANOVA using the ezANOVA() function from ‘‘ez’’ R pack-
age.59 In Experiment 1, the repeated factors were: modality (auditory, visual) and cardiac phase (diastole, systole). In Experiment 2,
the repeated factors were: emotion (happy, fear) and cardiac phase (diastole, systole). The residual plots of the two ANOVA models
were visually inspected to detect violations of normality. However, given that ANOVA with a balanced design is considered robust
even with normality violations60 and that alternative non-parametric Friedman test cannot be done for factorial designs, parametric
ANOVA was used in both cases.

Linear mixed model (LMM) analyses

Second, PSE and JND values were modelled with linear mixed models (LMM), using ‘‘lme4’’61 and ‘‘lmerTest’’62 R packages.
In Experiment 1, the model structure was as follows

e3 Current Biology 33, 1–7.e1–e4, April 10, 2023

 Please cite this article in press as: Arslanova et al., Perceived time expands and contracts within each heartbeat, Current Biology (2023), https://
doi.org/10.1016/j.cub.2023.02.034

ll
Report OPEN ACCESS

outcome variable  cardiac phase * modality + heart deceleration + (1 | participant) + (1 | cardiac phase : participant) + (1 | modality :
participant)
Heart deceleration (see ECG processing above) was added as a control variable because we were not interested in its interaction
with our main predictors. Categorical variables (cardiac phase and modality) were sum-coded, while continuous variables (heart
deceleration) were mean-centered and standardized. Separate intercept was added for each participant to account for repeated
measures. In addition, we specified that random effect of ‘‘participant’’ was fully crossed within both cardiac phase and modality
conditions – each participant appears at each level of cardiac phase and modality. Statistical significance was evaluated with likeli-
hood ratio tests using the mixed() function from the ‘‘afex’’ package.63
Initial run of both PSE and JND models produced a singular fit warning, suggesting that the random effect structure did not explain
sufficient variance to warrant its inclusion. We checked the proportion of variance explained by each of the random effects with
rePCA() function from the ‘‘lme4’’ package,61 which showed that (1 | cardiac phase : participant) random effect explained 0 variance.
Thus, in the final models that random effect was removed. However, the removal of that random effect led to equivalent statistical
significance of the fixed effects as for the models with the full random effect structure. Model assumptions were checked by plotting
the model residuals with check_model() function from ‘‘performance’’ R package.64 Both PSE and JND models did not reveal any
clear violations.
In Experiment 2, the LMM model coded for subjective arousal ratings given to the happy and fearful images, which were added
irrespective of the stimulus valence (i.e., each participant had 2 arousal ratings per cardiac phase). Due to a coding error, ratings
from three participants were missing, thus the models were run on the remaining participants (n = 36). The model structure was
as follows:
outcome variable  cardiac phase * arousal rating + heart deceleration + (1 | participant) + (1 | cardiac phase : participant)
Categorical variables (cardiac phase) were sum-coded, while continuous variables (arousal rating & heart deceleration) were
mean-centered and standardized. Initial run of both PSE and JND models produced a singular fit warning, suggesting that the
random effect structure did not explain sufficient variance to warrant its inclusion. We checked the proportion of variance explained
by each of the random effects with rePCA() function from the ‘‘lme4’’ package,61 which showed that (1 | cardiac phase : participant)
random effect explained 0 variance. Thus, in the final models that random effect was removed. However, the removal of that random
effect led to equivalent statistical significance of the fixed effects as for the models with the full random effect structure. Diagnostic
plots did not show any clear assumption violations.
Statistically significant interactions were followed up by a simple slopes analysis,65 whereby the effect of one continuous variable
was examined when the interacting continuous variable was either low (-1 SD from the mean), at its mean value, or high (+1 SD from
the mean). We used sim_slopes() function from the ‘‘interactions’’ package66 in R.

Current Biology 33, 1–7.e1–e4, April 10, 2023 e4