Cardiac signals have previously been associated with a wide range of psychopathologies, including substance abuse (Levin et al., 1992; Ingjaldsson et al., 2003), panic disorder and generalized anxiety disorders (Yeragani et al., 1993; Friedman and Thayer, 1998), obsessive compulsive disorder (Pittig et al., 2013), depression (Kemp et al., 2010), schizophrenia (Clamor et al., 2016) and post-traumatic stress disorder (Cohen et al., 1997). The link between cardiac signals and psychopathology is usually explained by the fact that these psychopathologies affect the function of the autonomic nervous system (ANS), which, in turn, directly impacts cardiac activity.

More recently, cardiac signals such as heart rate variability (HRV) have also been related to cognitive function in healthy individuals. For example, higher resting HRV has been associated with more adaptive attention to emotional stimuli (Park and Thayer, 2014), smaller startle responses (Ruiz Padial et al., 2003), better deliberate suppression of unwanted memories (Gillie et al., 2014) and thoughts (Gillie et al., 2015), more accurate and faster working memory retrieval (Hansen et al., 2003), and better self-control over dietary choices (Segerstrom and Nes, 2007). Furthermore, HRV has been shown to increase transiently during emotion regulation (Butler et al., 2006), and during difficult memory retrieval (Gianaros et al., 2004). Further to this, a direct causal relationship between ANS function and cognition has also been demonstrated via vagus nerve stimulation (for working memory processes; Clark et al., 1999).

Cardiac measurements have also been associated with performance in higher-level decision-making tasks. For example, individuals with low resting heart rates have been found to be more likely to make risky decisions (Schmidt et al., 2013). Transient heart rate, however, has been shown to slow directly after a loss in a gambling task, and this decrease in heart rate begins earlier (often prior to the outcome) in individuals who have better general performance in the task (Crone et al., 2004). It has also been shown that individuals with a higher power in the low frequency component of resting HRV generally perform better on gambling tasks (Drucaroff et al., 2011), and there is some evidence that direct vagus stimulation (which increases HRV) can enhance performance in these tasks (Martin et al., 2004). Furthermore, HRV (and its high frequency component) decreases during “stressful” unfair economic offers (such as those in the ultimatum game; Armin et al., 2011; Dulleck et al., 2014). These pressure-induced decreases in HRV have been shown to be more pronounced in individuals who cope less efficiently under high pressure situations (Laborde et al., 2014). While the reasons for these associations are not yet fully understood, these relationships demonstrate that in addition to their association with clinical disorders, cardiac signals are also associated with aspects of decision behaviors in healthy individuals.

One popular account, referred to as the neurovisceral integration model, posits that cardiac signals, and in particular HRV, index the state of a self-regulation network that spans across both the ANS and the central nervous system (Thayer and Lane, 2000, 2009). According to this model, the self-regulation network facilitates physiological, cognitive, and behavioral adaptability to environmental change. It predicts that a lack of such adaptability is associated with low HRV, whereas high HRV reflects a healthy, adaptive system.

The neurovisceral integration model places specific emphasis on the inhibitory role of the parasympathetic nervous system (PNS), the dysfunction of which results in prolonged and inappropriate physiological, emotional and behavioral responses (Thayer and Lane, 2000). As it has a relatively rapid influence on cardiac activity, one common method of estimating PNS function is via the high-frequency spectral component of HRV (HF-HRV; Berntson et al., 1997). Other studies have interpreted the low-frequency component of HRV (LF-HRV) as a reflection of sympathetic cardiac influence (Drucaroff et al., 2011; Dulleck et al., 2014), although this interpretation has been critisized (Reyes del Paso et al., 2013; see Discussion). The neurovisceral integration model also delineates neuroanatomical evidence for the relationship between cardiac signals and cognitive function. Specifically, this relies on the central autonomic network, which is responsible for central cardiac control, and comprises of the ventromedial prefrontal cortex (vmPFC), the central nucleus of the amygdala, anterior cingulate cortex, the insula, as well as several hypothalamic nuclei (Benarroch, 1993). Many of these areas have significant structural overlap with those that support the types of cognitive functions typically associated with cardiac signals (Thayer and Lane, 2000). In particular, cerebral blood flow in vmPFC has been correlated with changes in HRV induced by both emotional images (Lane et al., 2009) and working memory tasks (Gianaros et al., 2004). It has been suggested that vmPFC is the main locus of interaction between cognitive function and cardiac control (Thayer and Lane, 2000).

One fundamental aspect of self-regulation that has not previously been investigated from the perspective of cardiac physiology is regulatory behavior in the temporal domain, principally, impulsivity. While impulsivity is a multifaceted construct, two key aspects are of interest here: decision impulsivity, related to trading off immediate and delayed rewards, and impulsivity as it relates to time perception, which is required for appropriately timing actions (Evenden, 1999).

The main behavioral model of decision impulsivity is temporal discounting (TD; Bickel, 2015). It is typically characterized using a TD task, which assesses individuals’ choices between different magnitudes of reward available at different delays (Frederick et al., 2002). The rate of discounting (i.e., the rate of devaluation of reward per unit of time) has been shown to be relatively stable and heritable (Anokhin et al., 2015; Bickel, 2015), and has recently been proposed as an endophenotype (Bickel, 2015). Higher discount rates are apparent in multiple psychopathologies such as substance abuse, problem gambling, attention deficit hyperactive disorder, schizophrenia, depression and obesity (Bickel et al., 2012).

In the time perception domain, impulsivity is based on the notion that impulsive individuals overestimate, and hence under-reproduce time intervals (Wittmann and Paulus, 2008; Rubia et al., 2009; Moreira et al., 2016). For example, by using a duration reproduction (DR) task (in which participants reproduce a sample interval with a manual response), it has been shown that individuals with a impulsive personality traits tend to terminate their reproductions earlier than those without impulsive personality traits, as if they perceive the passage of time to be faster than it objectively is (van den Broek et al., 1992). Performance deficits in these types of time estimation tasks have been related to a range of psychiatric conditions: Parkinson’s disease, depression, bipolar disorder, schizophrenia, attention deficit hyperactivity disorder, autism, as well as anxiety disorders (Allman and Meck, 2012; Teixeira et al., 2013). Notably, there is considerable overlap in psychiatric disorders associated with abnormal cardiac signals, time perception and TD (Kemp et al., 2010; Allman and Meck, 2012; Bickel et al., 2012; Teixeira et al., 2013; Clamor et al., 2016).

Unlike TD, time perception has previously been investigated with respect to cardiac signals. For instance, one such study found that individuals with higher resting HRV were more accurate in a DR task (Pollatos et al., 2014). Another study found that individuals were more accurate in this task when they had a higher rate of heart rate slowing during the encoding of intervals (Meissner and Wittmann, 2011). Thus, there is some existing evidence that time perception accuracy and autonomic function are associated.

As the vmPFC is one of the primary regions involved in both central cardiac control and impulsivity (Kim and Lee, 2011), our first hypothesis was that specific aspects of cardiac signals are associated with behavioral performance during impulsivity-related tasks. Specifically, we hypothesized that HF-HRV (as an index of PNS function) would be associated with lower discount rates, and longer reproductions of time. Given that a common mechanism may underlie both decision impulsivity and time perception (Wittmann and Paulus, 2008; Moreira et al., 2016), and that damage to the vmPFC often results in both abnormal time perception and steeper TD (Berlin et al., 2004; Moreira et al., 2016), another hypothesis was that these two behavioral measures are related. To test these hypotheses, we measured cardiac activity while participants were completing a TD task and a DR task.

In order to investigate the relationship between time perception, decision impulsivity and cardiac signals, we assessed cardiac activity of individuals while they performed DR and TD tasks. We found a negative correlation between the mean heart rate and discount rate, suggesting that individuals with higher heart rates were more patient. We also found positive correlations between LF-HRV and the parameters of the psychophysical function for DR, suggesting that those with higher LF-HRV power perceived time differently during the encoding and reproduction of the interval, and had poorer sensitivity to longer intervals. Additionally, we found a positive correlation between HF-HRV and the exponent parameter of the psychophysical function, suggesting that those with high HF-HRV power increasingly underestimated longer intervals.

TD has previously been related to other physiological measures, such as pupil dilation (Lempert et al., 2016), but to our knowledge, our study is the first to relate it to cardiac signals. The neurovisceral integration model (Thayer and Lane, 2000, 2009) is built on the notion that greater regulatory control is associated with greater inhibitory parasympathetic functionality. Thus, it would predict that discount rate was negatively associated with HRV. Here, we observed that healthy individuals with higher heart rates had lower discount rates, indicating lower impulsivity, which does not support this prediction.

However, recent work has shown that during TD tasks, choices toward delayed rewards are more likely when sympathetic responses (measured via pupil dilation) to these options are greater (Lempert et al., 2016). As heart rate is primarily mediated by sympathetic activity, it is possible that the observed association between heart rate and discount rate constitutes a similar phenomenon: individuals with lower discount rates were also the ones with higher sympathetic responses, which reflected their tendency to choose delayed rewards. In support of this interpretation, previous studies have shown that higher levels of impulsivity (measured using personality questionnaires) were associated with lower resting heart rates (Mathias and Stanford, 2003) and with lower heart rates during the preparation of an independent task (Allen et al., 2009). Thus our findings provide further evidence that sympathetic nervous system responses may be associated with decision impulsivity.

We also observed a correlation between the parameters of the psychophysical function for DR and both LF- and HF-HRV. It is generally accepted that HF-HRV reflects PNS cardiac influence (Berntson et al., 1997), although it has also been reported that this can be confounded by respiratory patterns (Grossman and Kollai, 1993). While we did not directly measure respiratory activity in this study, an analysis controlling for this possible influence (Grossman and Kollai, 1993) did not alter our results. It has also been noted that interindividual associations between HF-HRV and parasympathetic cardiac influence are modest (Grossman and Taylor, 2007). Thus, while these findings appear to indicate that individuals with higher parasympathetic cardiac influence had poorer sensitivity to longer intervals, some caution should be taken in interpreting our results solely along these lines.

The interpretation of LF-HRV is more contentious than that of HF-HRV. In the psychological literature, increased LF-HRV has previously been related to fatigue and attentional deficits (Egelund, 1982; Mascord and Heath, 1992; Fairclough and Graham, 1999). This may provide an explanation of our findings, as lowered attentional capacity may lead to shorter encoded durations (Zakay and Block, 1995), or stronger temporal context effects and a regression to the mean of the sample durations (Jazayeri and Shadlen, 2010).

The physiological source of LF-HRV is highly contentious. It has been argued that differences in LF-HRV are due to either gross sympathetic activation (Malliani et al., 1991), a combination of parasympathetic and sympathetic influences (Reyes del Paso et al., 2013), baroreflex function (changes in HR due to changes in blood pressure; Goldstein et al., 2011), or the output of a central, LF oscillator (Cooley et al., 1998). On one hand, there is existing support for the possibility that the relationship between LF-HRV and time perception is sympathetically mediated. Time perception has been reported to “accelerate” relative to objective time during life-threatening situations (Stetson et al., 2007), under conditions of high body temperature (Wearden and Penton-Voak, 1995), as well as during general emotional arousal (Gil and Droit-Volet, 2012), which corresponds to sympathetic physiological changes (Thompson, 2013). Likewise, previous studies have shown that another measure of high sympathetic activity (cardiac pre-ejection period) is associated with lower temporal sensitivity (Cellini et al., 2015). On the other hand, the positive correlations among LF-HRV, HF-HRV and rMSSD (as well as the negative correlation between HF-HRV and the exponent parameter of temporal reproduction), suggest that our measurement of LF-HRV may represent predominantly parasympathetic influence. However, if this is the case, then our results directly contradict previous findings showing that parasympathetic function and vagal control are associated with increased time perception accuracy, rather than the decrease in accuracy we observed in those with high LF-HRV (Cellini et al., 2015). One possible resolution of this discrepancy is that this previous study used different time scales (around 1 s) and different timing mechanisms may be recruited for different durations (Ivry and Spencer, 2004). This interpretation is also inconsistent with the relationships between HRV, working memory and time perception, as higher vagal control has been associated with better working memory (Hansen et al., 2003), which is, in turn, associated with increased time perception accuracy (Broadway and Engle, 2011). Further to this, we did not observe any relationships between time perception and HRV in the time domain (rMSSD). Thus, it is difficult to comment on the physiological source of the observed result. Future studies could clarify this by employing more interpretable measures of SNS activity, such as skin conductance level (Mella et al., 2011), or levels of neurotransmitters such as noradrenaline (Zygmunt and Stanczyk, 2010). However, our results are consistent with the psychological attention-based interpretation of LF-HRV, and more broadly the associations between HRV frequency components and time perception measures support previous suggestions that periodic internal signals may constitute a time-keeping mechanism (Craig, 2009; Wittmann, 2015).

It is important to note that some of the observed correlations (e.g., between discount rates and mean heart rate) were only significant before correction for multiple comparisons. The uncorrected results are nevertheless interesting, given the absence of existing literature on this topic, and the exploratory nature of our study. We emphasize that these findings require further investigation.

Some limitations of the current study could be addressed in future work. For example, in addition to using directly interpretable measures of SNS activity, future studies could also employ other non-invasive physiological measures, such as eyeblink rate or pupillometry. On the other hand, an assessment of action, as opposed to choice impulsivity (such as a Go-Nogo task), or personality traits, may also be desirable (Glicksohn et al., 2006). It would also be of interest to investigate whether the observed correlations between cardiac measures and time perception would extend to other time estimation paradigms, such as a temporal bisection task, or shorter or longer intervals. Finally, this line of inquiry may be of potential utility in the diagnosis and treatment of psychopathologies that involve impulsivity, such as substance abuse, problem gambling, attention deficit hyperactive disorder, schizophrenia, depression and obesity (Bickel et al., 2012).

We note that as our physiological measures were not isolated to the presentation of stimuli, we were not able to determine whether the observed correlations were driven by task-related responses or trait-like autonomic function. Further studies with baseline measures, longer recordings, and more isolated task phases may be able to address this question, as well as whether the observed relationships were mediated by other personality traits.

While we found that cardiac signals were independently correlated with both TD and DR, we did not find any direct relationship between these two components of impulsivity (Wittmann and Paulus, 2008; Moreira et al., 2016). One possible reason for this is the difference in temporal scale (months in TD vs. seconds in DR), which may recruit different cognitive mechanisms. Future research could address this by using a discounting task with time delays more similar to those used in time perception tasks.

In conclusion, our study shows that differences in ANS function may help to explain inter-individual heterogeneity in both TD and time perception. The association between TD and heart rate supports the notion that low arousal might be related to higher impulsivity, similar to previous perspectives on trait impulsivity (Eysenck, 1993) and previous research using alternate measures (Mathias and Stanford, 2003; Allen et al., 2009; Lempert et al., 2016). Given the conflicting interpretations of cardiac indices, our results concerning DR are difficult to interpret from a physiological perspective. Psychologically, however, our results appear to reflect an association mediated by differences in attention. We found no evidence for a relationship between TD and time perception, which reinforces the idea that these measure different aspects of impulsivity, which appears to be a rather complex construct (Evenden, 1999). Our findings further show that ANS function could provide distinct indices for such aspects of impulsivity, opening up new avenues for future research to decompose impulsivity beyond TD and time perception (e.g., response inhibition; Krypotos et al., 2011).

BJF and DLC contributed to the design of the study, collection and analysis of the data, and drafting of the manuscript. SB and CM contributed to the design of the study, statistical analysis, and drafting of the manuscript. All authors gave final approval for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.