Cortical and Affective Regulation of Autonomic Coordination

Abstract

Although anatomical research clearly demonstrates the ability of the sympathetic and parasympathetic branches of the autonomic nervous system to independently influence cardiac function, little research has examined whether coordinated activation is typical, or whether the extent of autonomic coordination is situationally-dependent. This study examines the extent of coordination between sympathetic (cardiac pre-ejection period: PEP) and parasympathetic (respiratory sinus arrhythmia: RSA) influences on cardiac function to determine whether coordination is a trait-like between-person characteristic, or a state-varying within-person phenomenon, and if so, whether variability in autonomic coordination is modulated by cognitive (P3b amplitude) or affective state. Kindergarten-aged children (n = 257) completed a go/no-go task administered in blocks designed to induce affective states through the delivery of reward (Blocks 1 and 3) and frustration (Block 2). Results from multilevel models that allowed for simultaneous examination of between-person and within-person associations in the repeated measures data suggested that (a) children with higher overall RSA also tended to have higher overall PEP; (b) at within-person level, RSA and PEP tended to be reciprocally coordinated; but that (c) when frustration invokes cognitive disengagement, coordination between parasympathetic and sympathetic systems demonstrate compensatory coordination. These findings highlight the extent to which coordination of autonomic systems is a dynamic state-like phenomenon rather than a trait-like individual differences characteristic.

1. Introduction

1.1. Cortical Regulation of Autonomic Coordination

Although extensive evidence confirms the ability of each branch of the autonomic nervous system to act independently (Berntson, Cacioppo, & Quigley, 1993; Berntson, Cacioppo, Quigley, & Fabro, 1994), the extent and nature of coordination between sympathetic and parasympathetic branches may have additional meaning beyond the amount of activation in either branch alone (Quas et al., 2014; Rudd & Yates, 2018). The extent of coordination between sympathetic and parasympathetic influences on cardiac function is presumably regulated by higher-order central nervous system networks capable of integrating complex information related to situational context and motivational goals to deliver efferent signals to the peripheral nervous system in support of the behavioral responses necessary to meet the current environmental demands (e.g., Critchley et al., 2003; Smith, Thayer, Khalsa, & Lane, 2017). Several theories propose that measures of autonomic response to psychological demands can be interpreted indices of central nervous system activity, most frequently with regard to the involvement of the prefrontal cortex in regulating parasympathetic activity (e.g., Thayer & Lane, 2009). However, very few studies have directly examined the association between individuals’ on-going parasympathetic and sympathetic activity, or whether coordination between branches is associated with cortical function. Furthermore, the majority of studies examining cortical and autonomic associations do so from a between-person perspective, for example, examining whether individuals who are higher than their peers on cognitive or cortical measures of prefrontal function are also higher than their peers on measures of parasympathetic function. This approach does not adequately assess the hypothesized within-person process whereby the extent to which an individual’s autonomic branches are coordinated fluctuates dynamically over time in accordance with changes in cortical activation. Thus, the present study seeks to extend the research on coordination between central and autonomic function by examining, from both between-person and within-person perspectives, how differences and changes in an electrophysiological index of attentional control (P3b) are related to the coordination of sympathetic (cardiac pre-ejection period; PEP) and parasympathetic (respiratory sinus arrhythmia; RSA) indices of cardiac function.

1.2. Cortical Control of Autonomic Function

Several theories posit that greater parasympathetic cardiac function is associated with better cognitive control (Porges, 2007; Thayer & Lane, 2009). According to the polyvagal model (Porges, 2007), higher resting heart rate variability (frequently quantified as RSA), an index of parasympathetic control, reflects a greater capacity to regulate physiological arousal dynamically. The withdrawal of parasympathetic control supports cognitive engagement by enabling increases in physiological arousal as needed to meet cognitive demands. Empirical evidence from across the life span suggests that higher resting heart rate variability is associated with better cognitive performance, including among pre-school aged children (Marcovitch et al., 2010), school aged children (Staton, El-Sheikh, & Buckhalt, 2009), adolescents (Chapman, Woltering, Lamm, & Lewis, 2010), young adults (Capuana et al., 2014), and adults (Hansen, Johnsen & Thayer, 2003). Fewer studies have directly examined cortical activity, but findings from those that have indicate that there is a functional association between prefrontal function and resting RSA. Specifically, higher resting RSA is associated with localized connectivity strength within the anterior cingulate cortex (Jennings, Sheu, Kuan, Manuck, & Gianaros, 2016), increased regional cerebral blood flow across multiple cortical regions associated with sustained attention and mental effort (Lane et al., 2009), and better performance on a brain-computer interface task driven by event related potential activity (P300) (Kaufmann, Vogele, Sutterlin, Lukito, & Kubler, 2012). Together, these findings indicate that individuals with relatively higher resting heart rate variability also demonstrate relatively better cognitive function and/or more cortical activity.

Despite the robust association reported for resting RSA, associations between cortical function and RSA reactivity to cognitive challenges are mixed. Results from some studies align with the polyvagal model, with better cognitive performance during a go/no-go task associated with higher RSA at rest, and lower RSA during the task (Chapman et al., 2010). Other studies, however, did not find associations between cognitive performance and RSA reactivity to task (Staton et al., 2009). The mix of evidence suggests that the association between RSA reactivity and cognitive performance may be complex, with the relative benefits of reactive increases or decreases in RSA varying as a function of the task (Sulik, Eisenberg, & Spinrad, 2015), or of participant characteristics (Skowron, Cipriano-Essel, Gatzke-Kopp, Teti, & Ammerman, 2014; Giuliano, Gatzke-Kopp, Roos, & Skowron, 2017; Giuliano et al., 2018b).

Although fewer studies have examined the association between sympathetic arousal and cognitive function, some evidence suggests that higher sympathetic arousal is associated with better cognitive performance. Research with children suggests that cardiac measures of both higher parasympathetic and higher sympathetic activity at rest are associated with better subsequent inhibitory control performance (Giuliano et al., 2018a). Research with adults suggests that cardiac indices of higher sympathetic activity are associated with better attentional control and executive function (Duschek, Muckenthaler, Werner, & del Paso, 2009; Duschek, Hoffmann, Reyes Del Paso, & Ettinger, 2017) and electrophysiological indices of attentional control (Giuliano et al., 2018b). In addition, neuroimaging research indicates an association between dorsal anterior cingulate activation and sympathetically mediated increases in blood pressure in the context of cognitive effort (Critchley, 2009). Although the existing literature indicates that individual differences in both sympathetic and parasympathetic function are associated with differences in cognitive control, none of these studies have examined if and how the coordination between autonomic branches is associated with cognitive function.

1.3. Autonomic Coordination

The understanding that each branch of the autonomic nervous system can act independently has prompted researchers to focus on measures that facilitate better isolation of each branch (e.g. RSA and PEP). However, research explicitly examining the extent to which autonomic branches function independently, or whether patterns of coordination have psychophysiological significance, has received less attention. The implications of autonomic coordination are best illustrated in the context of the functional surface of autonomic space first described by Berntson, Cacioppo, and Quigley (1991). This functional surface is illustrated in Figure 1, where vertical height indicates the functional state of arousal of the physiological system as a result of the contributions from each autonomic branch. Research on autonomic coordination has typically characterized profiles of coordination according to the direction of change in each branch from baseline to task conditions. Change in each branch can be classified as an increase (+) or decrease (−), and coordination profiles can then be populated as a simple cross tabulation of these two dichotomous dimensions. Illustrated in Figure 1 are commonly identified profiles of reciprocal sympathetic activation (increase in sympathetic activation with concurrent decrease in parasympathetic activation), reciprocal parasympathetic activation (decrease in sympathetic activation with concurrent increase in parasympathetic activation), co-inhibition (decrease in sympathetic activation with concurrent decrease in parasympathetic activation), and co-activation (increase in sympathetic activation with concurrent increase in parasympathetic activation). An examination of these autonomic response profiles across 3 studies encompassing children between 3 and 8 years of age indicated that, although there appeared to be some changes in the distribution of children in each of the autonomic profiles across age groups, there was no evidence of a single dominant or prototypical response style (Alkon et al., 2003). Studies examining psychological correlates of autonomic response styles have been somewhat mixed. One study of typically developing young adults found that autonomic co-activation was associated with higher levels of self-reported “grit” (Silvia, Eddington, Beaty, Nusbaum, & Kwapil, 2013). In contrast, several recent studies of adolescents exposed to inter-parental conflict found that co-activation was associated with greater increases in symptoms of psychopathology over time (McKernan & Lucas-Thompson, 2018; Philbrook, Erath, Hinnant, & Eli-Sheikh, 2018).

Figure 1.

Adapted from the model of autonomic space reported originally by Berntson, Cacioppo, & Quigley, 1991, this figure illustrates how levels of activation in the sympathetic and parasympathetic branches contribute collectively to the overall functional state of physiological arousal. States of autonomic coordination defined by relative increases and decreases in each branch are illustrated with regard to their implications for physiological arousal.

Recently our lab sought to examine autonomic coordination as a dynamic process wherein individuals continually move within the autonomic space, rather than assuming that individuals only exist within a small set of static profiles. We examined sympathetic (PEP) and parasympathetic (RSA) cardiac measures quantified in 30 sec epochs across a 12 min task as children passively viewed a series of four film clips depicting different emotional themes (fear, sad, anger, happy) (Gatzke-Kopp & Ram, 2018). Across the 31 epochs spanning film clips that lasted between 2 and 3 minutes and were interspersed with 30 sec neutral clips and 30 sec fixation blocks, there was considerable variability in the nature and intensity of the stimuli. Our analyses capitalized on this variability to examine whether and how dynamic movement across epochs in one autonomic branch was correlated, within-person, with the dynamic movement in the other branch. Dynamic movement within the autonomic space can be conceived of in 4 main patterns illustrated in Figure 2: (1) independent movement of the parasympathetic system (illustrated with the purple line), where changes in arousal are solely a function of activity in the parasympathetic system with the range of functional arousal achievable constrained by the intercept point on the sympathetic axis, (2) independent movement of the sympathetic system (illustrated with the red line), where changes in arousal are solely a function of activity in the sympathetic system, with the range of functional arousal constrained by the intercept on the parasympathetic axis, (3) reciprocal coordination (illustrated by the blue diagonal line) where increases in one system are accompanied by decreases in the other such that each branch contributes in the same way toward the change in arousal, and (4) compensatory coordination (illustrated by the green line) where increases in one branch are accompanied by increases in the other and vice versa, such that each branch contributes in opposing ways toward the change in arousal. Changes in arousal that are mediated by independent activation in one branch reflect a lack of autonomic coordination. In contrast, the compensatory and reciprocal patterns reflect presence of autonomic coordination, with reciprocal coordination indicated by a positive within-person correlation between repeated measures of PEP and RSA (because longer PEP reflects lower sympathetic activation), and compensatory coordination indicated by a negative within-person correlation between repeated measures of PEP and RSA. Results from our analyses indicated that the prototypical child’s autonomic branches were reciprocally coordinated, and there was no evidence of developmental change of this coordination over repeated assessments obtained in kindergarten, 1^st, and 2^nd grade.

Figure 2.

Dynamic changes in overall arousal (height on Y axis) can be accomplished through movement in either or both autonomic branches. Changes mediated entirely through fluctuations in sympathetic arousal are illustrated by the red line. Note that the level of arousal achievable through independent changes in sympathetic activation are constrained by the intercept on the parasympathetic axis. Conversely, changes mediated entirely through fluctuations in parasympathetic arousal are illustrated by the purple line. Reciprocal coordination, encompassing the states of reciprocal sympathetic and reciprocal parasympathetic activation illustrated in Figure 1, is illustrated by the blue line. Compensatory coordination, encompassing the states of co-activation and co-inhibition illustrated in Figure 1, is illustrated by the green line. Note that movement along the blue line results in rapid changes in functional arousal whereas movement along the green line results in a relatively stable level of arousal.

Importantly, this dynamic conception and approach to studying autonomic coordination accommodates the fact that activation in either or both the parasympathetic and sympathetic branch is likely to fluctuate; demonstrating increases and decreases over time. Unlike profile classification approaches that only consider unidirectional change (baseline to task) this analytical approach accommodates multidirectional change, both the ups and downs in RSA and PEP that occur as individuals move between baseline and task, within task, and between task and baseline. This accommodation, however, does imply that movement along the diagonal corridor (reciprocal coordination) does not differentiate between the reciprocal sympathetic and reciprocal parasympathetic profiles because in both cases, task-related increases/decreases in activation of the sympathetic or parasympathetic branches will eventually and presumably reverse when the task ends. In sum, conceptual and analytical approaches that make use of repeated measures of parasympathetic and sympathetic activation (e.g., many 30 sec epochs) provide new opportunity to study autonomic coordination as a dynamic process wherein individuals continually move within the autonomic space.

1.4. State-Moderation of Autonomic Coordination

In the original presentation of the autonomic space model, Berntson and colleagues (1991) present the argument that the adaptive value of any given point on the autonomic space grid may be determined by the situational demands. Thus, rather than being a stable, trait-like characteristic of an individual, autonomic coordination may vary as a function of contextual demands. Indeed, our prior analyses of children’s autonomic coordination during passive film viewing found that the extent of autonomic coordination was moderated by the nature of the emotion depicted in the film at any given time (Gatzke-Kopp & Ram, 2018). Reciprocal coordination was strongest during approach-oriented emotion scenes (e.g. happy, angry), and weaker during avoidance-oriented emotion scenes (fear, sad), consistent with the notion that approach-oriented emotions require greater physiological arousal to meet the behavioral demands of approach. However, this evidence does not address whether autonomic coordination is associated with cortical prefrontal activity.

Anatomical networks extending from the prefrontal cortex are believed to coordinate arousal-related systems (autonomic and neuroendocrine) in support of both emotional and cognitive behavior (McKlveen, Myers, & Herman, 2015; Van Eden & Buijs, 2000). Thus, it is possible that reciprocal coordination would be greater at times when individuals are more cognitively engaged. Alternatively, the polyvagal theory proposes an autonomic hierarchy where mild increases in arousal demands associated with simple cognitive function are first supported by withdrawal of the parasympathetic system, with the sympathetic system engaging only if the arousal needs cannot be met by the parasympathetic system alone (Porges & Furman, 2011). The proposed hierarchy suggests that cortical engagement is associated with parasympathetic activity and sympathetic activity sequentially rather than concurrently through autonomic coordination. Despite evidence that activity in both branches of the autonomic nervous system is influenced by cortical structures associated with cognition (e.g. Giuliano et al., 2018a Giuliano et al., 2018b), it remains unclear how cortical activation is related to autonomic coordination, theoretically or empirically.

1.5. The Present Study

The current analyses extend our and others’ previous findings by examining whether autonomic coordination is modulated by cortical regulation of attentional control, and/or as a function of the affective context across the experimental task. Kindergarten age children completed a go/no-go task, during which cardiac measures of sympathetic (PEP) and parasympathetic (RSA) activity were measured along with the P3b event-related potential (as an index of children’s attentional activation). Between-person analyses examine whether individual differences in central nervous system activity (P3b) are related to autonomic nervous system activity (PEP and RSA) and the extent of autonomic coordination between sympathetic and parasympathetic branches. Within-person analyses examine whether autonomic coordination changes dynamically in conjunction with within-person changes in P3b amplitude induced by a frustration manipulation previously shown to induce meaningful individual differences in P3b for a subset of children prone to disruptive behavior (Gatzke-Kopp et al., 2015). By design, this experimental manipulation induces an affective perturbation that facilitates examination of both how the affective context influences autonomic coordination, and how autonomic coordination is related to between-person differences and within-person changes in cortical function.

Using a multilevel model, we examine how autonomic coordination is associated with the following three perspectives: (1) Consistent with previous findings (e.g., Gatzke-Kopp & Ram, 2018) we expect that children’s autonomic coordination is modulated by the changes in affective context induced through experimental manipulation (task condition); (2) Following theory and empirical findings implying that cognitive engagement is associated with a reciprocal coordination between autonomic branches, we examine the association between autonomic coordination and P3b amplitude according to the following multi-level hypotheses: Between-person, individuals with relatively higher P3b average amplitude will have a relatively stronger reciprocal coordination (denoted by positive correlation between PEP and RSA); Within-person, the strength of association between RSA and PEP will be moderated by within-person variation in P3b amplitude such that a stronger positive correlation will emerge in the block when P3b amplitude is relatively higher; (3) Alternatively, if as implied by the polyvagal model, cognitive engagement is supported primarily through the parasympathetic system, we hypothesize that P3b amplitude will be correlated with between-person and within-person differences in RSA, but not with the coordination of RSA and PEP. Results inform competing theoretical models and contribute to understanding of how central and autonomic systems interact by examining the data from three perspectives.

2. Method

2.1. Participants

Data for the present analyses were drawn from the baseline physiological assessments obtained in a longitudinal clinical trial of a socioemotional intervention program for early-onset aggression (Gatzke-Kopp, Greenberg, Fortunato, & Coccia, 2012). The sample was drawn from the Harrisburg School District in central Pennsylvania through a screening process that over-selected children with elevated externalizing symptoms. The full sample represented the full range of severity on the externalizing scale (M = 23.35, SD = 12.80; observed range = 10–60; possible range = 10–60) and nearly the full range of severity on the internalizing scale (M = 9.28, SD = 4.62; observed range = 5–27; possible range = 5–30), and thus may be considered representative of a psychologically heterogenous population of American children. In total, 352 children (61.6% male) were recruited for participation in the 3-year longitudinal study (including n = 13 children who dropped out after completing the initial physiological assessment). Consistent with the community demographics, the majority of parents identified their children as African American (68%), with the remaining identifying their children as Latino (20%), Caucasian (11%), or other (1%). Although portions of these data have been examined in other ways (Gatzke-Kopp et al., 2015, Willner, Gatzke-Kopp, Bierman, Greenberg, & Segalowitz, 2015), the research questions and combination of data used here are fully unique.

The analysis sample in the present study consists of n = 257 children (65% male, M_age = 6.04 years, SD = .38) with valid physiology data from the initial (prior to intervention) physiological assessment. Data from n = 48 children were missing due to the following reasons: the child refused participation (n = 2), excess school absence during baseline assessment phase (n = 6), or researcher error/ equipment failure (n = 40). Data from an additional n = 47 participants were excluded from the current analyses because one or more of the physiological measures was corrupted by movement artifact. The 257 children with relatively complete psychophysiology data did not differ from those with missing data with regard to sex χ²(1, 352) =.45, p = .50 or likelihood being African American, χ²(1, 352) = 2.15, p = .14, but did have, on average, marginally lower scores on the externalizing screening scale, F(1, 351) = 3.95, p = .048 (d = 0.23; complete group: M = 10.14, SD = 5.13; group with missingness: M = 9.03, SD = 4.46).

2.2. Procedure

All procedures were approved by the institutional IRB. Parents were visited at their home by research staff, informed of all procedures related to the study, and provided written consent for their child to participate. Psychophysiological assessments took place during the school day in a mobile laboratory that was installed inside of a recreational vehicle. On the day of the assessment each child was summoned from class and greeted by two research assistants (RA) who explained the protocol to the child. After receiving verbal assent from the child, RAs applied 7 electrodes to measure autonomic function (heart rate, impedance, and skin conductance) as well as a 32-channel EEG cap, described below. The child was then seated in front of a computer monitor where they were told to sit quietly while they “traveled through space to another planet where they would play a game”. The computer screen showed a moving starfield video for 2 minutes while baseline physiological data were recorded. The child then completed a Go/No-go task (described below), had a short snack break, and then travelled through space again to watch a movie (emotion induction task not examined here).

2.3. Go/No-go Task

To enhance engagement, stimuli consisted of 45 distinct cartoon drawings of alien critters that the child was instructed to “zap” by pressing a button on the response box (Go trial). The child was instructed not to zap a critter twice in a row, and thus to avoid pressing the button if the critter on the screen was identical to the one that just appeared (No-go trial). The task was programmed to adjust the inter-trial interval dynamically in an effort to equalize objective performance (error rate) across participants (Lewis, Lamm, Segalowitz, Stieben & Zelazo, 2006).

2.3.1. Affective manipulation

Prior to starting the task, children were shown a collection of prize bags filled with small assorted toys and told they could win a prize only if they earned “enough” points. The necessary number of points was not specified to ensure that children remained motivated to earn points throughout the task. The task consisted of 3 blocks of trials that included 75 Go trials interspersed with 35 No-go trials. Blocks, which each lasted about 2.5 minutes, were separated by 30 second rest (baseline) periods. In the first block, rewards (points) were governed by an algorithm that provided for consistent accumulation of points by strongly rewarding correct responses while weakly punishing incorrect responses. In the second block the algorithm reversed, resulting in a continuous loss of points despite comparable performance. In the final block the original algorithm was reinstated, allowing children to regain their points and ultimately earn the prize. To ensure that the reward structure was easy to comprehend by the kindergarten-aged children performance feedback was conveyed visually through multiple cues. After incorrect trials, a large red rectangle appeared on the screen, framing the stimulus prompt. After correct trials, no rectangle appeared. Approximately every 10 trials (jittered), a “thermometer” that indicated their current cumulative point total appeared in the center of the screen along with a cartoon face that provided proximal reward feedback; with a thumbs-up smiling face or thumbs-down frowning face indicating whether the child had won or lost points since the last time the thermometer was shown.

2.3.2. Task design variables

Two variables were invoked to enable an examination of how the experimental manipulation was associated with physiological activity: Block number, coded 0, 1, 2 for analysis, was used to examine linear change across the task. Frustration, coded = 0 for blocks where correct responses were highly rewarded, and = 1 for the block where points were lost, was used to examine differences associated with the affective manipulation (e.g., coding across blocks = 0, 1, 0).

2.4. Cardiac Recording

Electrocardiograph data were obtained using 3 disposable, pre-gelled cardiac electrodes placed over the child’s distal right collar bone, lower left rib, and lower right rib. Impedance cardiograph data were obtained using an additional 4 spot electrodes that were placed on the front of the torso on the collarbone and sternum, and on the back of the torso at positions one inch outside of each of the frontal electrodes. Data were collected continuously at 500 Hz using the Biolab 2.4 acquisition system (Mindware, Westerville, OH).

2.4.1. RSA

Respiratory sinus arrhythmia (RSA) was computed using Mindware HRV (v. 3.0.20) software. ECG files were visually inspected to identify and correct missing or erroneous beats. The cleaned interbeat interval series was subjected to a fast Fourier transformation to compute power in the 0.12–1.04 Hz frequency band. Data were output in 30sec epochs, matching with epoch length used to calculate PEP (described below) and allowing for both valid estimation of power in this frequency band and sensitivity to dynamic changes across the task (Berntson et al., 1997). Individual epochs with substantial movement artifact that obscured detection of R spikes for several consecutive beats, and epochs in which peak respiration power (estimated using the impedance wave) fell outside of the respiratory frequency defined for extracting RSA, were removed.

Repeated measures of RSA for Block 1 included a 30s epoch of initial resting baseline RSA prior to the start of the task, up to 6 30s epochs of RSA obtained while the child was zapping aliens in the reward condition, and a 30s epoch of between-block resting baseline (starfield space travel). Repeated measures of RSA for Block 2 (frustration) included up to 6 30s epochs of RSA obtained while the child was zapping aliens in the frustration condition and a 30s epoch of between-block resting baseline (during which children may have still been frustrated). Repeated measures of RSA for Block 3 included up to 6 30s epochs of RSA obtained while the child was zapping aliens in the reward condition. Of note, in addition to data loss due to artifact, the number of epochs within a block could differ between children because of the dynamic algorithm that adjusted the inter-trial interval length in order to standardize the task difficulty. Altogether, children provided up to 21 epochs of RSA data (M = 17.74, SD = 3.46, min = 3, max = 21), which included up to 8 epochs in Block 1 (M = 7.06, SD = 1.12), up to 7 epochs in Block 2 (M = 5.88, SD = 0.95), and up to 6 epochs in Block 3 (M = 5.25, SD = 0.89).

2.4.2. PEP

In parallel, cardiac pre-ejection period (PEP) was computed using Mindware IMP (v. 3.0.3) software. The IMP program was used to generate an ensemble average of all heartbeats within the same 30-second epochs described above for RSA, superimposed with the ensemble average of the corresponding impedance cardiogram data. An automated algorithm identified critical points within both wave forms (R peak in the electrocardiogram and the Z peak in the impedance cardiogram) and estimated the placement of the B point (Lozano et al., 2007). Trained researchers visually inspected each epoch to identify potential artifact, and determine whether erroneous placement of the B point could be corrected manually. If correction was not possible, the epoch was removed.

In total, this analysis makes use of 4,559 epochs of cardiac recording (RSA and PEP) obtained from 257 children during completion of the Go/No-go task.

2.5. EEG Recording

EEG data were recorded from a BioSemi ActiveTwo system (BioSemi, Amsterdam, Netherlands) with 32 standard extended 10–20 scalp channels using sintered silver electrodes. Additional electrodes were placed on the left and right suborbital ridge under the pupil and 1 cm outside the left and right lateral canthi. Data were recorded using DC amplifiers with a gain of 1 using 24-bit A-D conversion, and digitized according to the BioSemi zero reference principle (the voltage at each site is quantified relative to the common mode sense and driven right leg loop) at 1024 Hz. As a substitute for impedance measures quantifying signal quality, electrode offsets were maintained below 50 μV. Data were low-pass filtered at 512 Hz online, and subsequently filtered offline with a 1– 30 Hz bandpass in Brain Vision Analyzer.

Details regarding the quantification of the P3b in the present sample have been published previously (Gatzke-Kopp et al., 2015; Willner et al., 2015). In brief, the P3b in response to go-stimuli were examined because each Go trial presented an essentially novel stimulus. Although the P3b is a stimulus-locked ERP, only trials in which the participant made a correct response between 100 and 1,000 ms after stimulus onset were segmented to ensure the trial was valid and the participant was paying attention. Trials were segmented from −200 to 1,000 ms relative to stimulus onset, and baseline-corrected to the mean amplitude across −200 ms to stimulus onset. Eye blink artifacts were corrected using the Gratton and Coles algorithm, as implemented by Brain Vision Analyzer 2.0 (Gratton, Coles, & Donchin, 1983). Artifact detection eliminated individual trials with a voltage step of more than 100 μV between sampling points, or a voltage reading outside the range of −75 μV to 75 μV. As elsewhere, P3b was quantified as the mean voltage from 500 to 700 ms post-stimulus within each of the three blocks. P3b was calculated for blocks with a minimum of 3 usable trials after artifact rejection. The number of usable trials was, in Block 1: M = 39.33 (SD = 13.82), range 3 – 69; in Block 2: M = 34.25 (SD = 13.11), range 3 – 71; and in Block 3 M = 41.73 (SD = 15.02), range 3 – 71. Number of usable trials was not significantly correlated across blocks or with P3b amplitude for that, or any other block (all ps > .29).

2.6. Data Alignment

The repeated measures of RSA and PEP (up to 21 30s epochs) and the repeated measures of P3b (up to 3 blocks) were aligned with respect to the task design variables. Specifically, block-level P3b scores were attached to all the epochs within Blocks 0, 1 and 2, respectively. Example data from one child are shown in Figure 3. Three lines indicate how the repeated measures of RSA (blue squares), PEP (red triangles), and P3b (green circles) changed across the Go/No-go task (values standardized for plotting purposes only). Illustrative of our interest in the within-person coordination of parasympathetic and sympathetic activity associations, the within-person correlation between RSA and PEP for this participant is strong (r_within = .72 and .70) in the non-frustration blocks, and weaker (r_within = .27) in the frustration block. Similar plots for other children in the sample show different patterns (i.e., between-person differences in within-person coordination).

Figure 3.

Example data from one child. Three lines indicate how standardized values for the repeated measures of RSA (blue squares), PEP (red triangles), and P3b (green circles) changed across the Go/No-go task. Within-person correlations between RSA and PEP are shown separately for Blocks 1, 2 (frustration), and 3.

2.7. Data Analysis

The nested nature of the data (epochs nested within blocks nested within persons) was accommodated analytically using a multilevel modeling framework (Snijders & Bosker, 1999). An unconditional means model (no predictors) indicated that of the total variance in the RSA scores, 80% was attributable to differences across persons, 2% was attributable to differences across blocks, and 18% was attributable to differences across epochs. Given that such a small portion of the variance in RSA was attributed to differences across blocks, we collapsed to a two-level structure, epochs as nested within persons, with block included as a time-varying (epoch-level) factor. Following usual practice, other time-varying predictors were split into time-invariant (between-person) and time-varying (within-person) components (see Bolger & Laurenceau, 2013). For example, the repeated measures of each child’s PEP and P3B across epochs were split into person-mean scores (PEPperson_i, P3Bperson_i) that were calculated as the arithmetic mean of all the repeated measures, and epoch-specific scores (PEPepoch_it, P3Bepoch_it) that were calculated as deviations from the person-means. Using these variables, between- and within-person associations of RSA and PEP, and moderation by P3B and the affective manipulation, were examined using a multilevel model of the form

$$\begin{gathered} {RSA}_{it}={\beta }_{0i}+{\beta }_{1i}({PEPepoch}_{it})+{\beta }_{2i}({P3Bepoch}_{it})+{\beta }_{3i}({Block}_{it})+ \\ {\beta }_{4i}({Frustration}_{it})+{\beta }_{5i}({PEPepoch}_{it}\ast {P3Bepoch}_{it})+ \\ {\beta }_{6i}({PEPepoch}_{it}\ast {Frustration}_{it})+{\beta }_{7i}({P3Bepoch}_{it}\ast {Frustration}_{it})+ \\ {\beta }_{8i}({PEPepoch}_{it}\ast {P3Bepoch}_{it}\ast {Frustration}_{it})+e_{it} \end{gathered}$$ (1)

where the repeated measures of RSA for individual i at epoch t, RSA_it, are modeled as a function of person-specific intercepts (β_0i), within-person associations with epoch-to-epoch changes in PEP (β_1i) and P3B (β_2i), time-related changes indicative of fatigue or habituation (β_3i), changes associated with the affective manipulation (β_4i), the relevant interactions among these time-varying variables (β_5i, β_6i, β_7i, β_8i), and residual error (e_ti) that was assumed normally distributed. Person-specific coefficients were simultaneously modeled as a function of person-level predictors,

$$\begin{gathered} {\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}({PEPperson}_i)+{\gamma }_{02}({P3Bperson}_i)+{\gamma }_{03}({PEPperson}_i\ast \\ {P3Bperson}_i)+u_{0i} \end{gathered}$$ (2)

$${\beta }_{1i}={\gamma }_{10}+{\gamma }_{11}({PEPperson}_i)+{\gamma }_{12}({P3Bperson}_i)+u_{1i}$$ (3)

$${\beta }_{2i}={\gamma }_{20}$$ (4)

$${\beta }_{3i}={\gamma }_{30}$$ (5)

$${\beta }_{4i}={\gamma }_{40}$$ (6)

$${\beta }_{5i}={\gamma }_{50}$$ (7)

$${\beta }_{6i}={\gamma }_{60}$$ (8)

$${\beta }_{7i}={\gamma }_{70}$$ (9)

$${\beta }_{8i}={\gamma }_{80}$$ (10)

where the γs are sample-level parameters that describe between-person associations and prototypical within-person associations, and us are residual unexplained between-person differences that may be correlated. Person-level predictors were sample-centered to facilitate interpretation of model parameters as representing effects for the prototypical child (as described by the average demographics above). Additional interaction terms were included, but trimmed iteratively to obtain the parsimonious model shown above. Models were fit to the data using the lme4 library in R (Bates, Maechler, Bolker & Walker, 2015) with restricted maximum likelihood estimation.

3. Results

Results from the multilevel model used to examine between-person and within-person associations among autonomic and cortical measures are shown in Table 1. The prototypical child’s level of RSA in the initial non-frustration block was estimated as γ₀₀ = 7.082 (p < .001).

Table 1.

Results From Multilevel Model Examining Between-person and Within-person Association of RSA and PEP and How It Is Moderated by P3b and Frustration.

Parameter	Estimate	95% CI
Fixed Effects
Intercept, γ00	7.082*	[6.930, 7.234]
PEPperson, γ01	0.016*	[0.004, 0.028]
P3Bperson, γ02	0.005	[−0.052, 0.063]
PEPperson x P3Bperson, γ03	−0.00004	[−0.005, 0.005]
PEPepoch, γ10	0.018*	[0.010, 0.025]
PEPepoch x PEPperson, γ11	− 0.001*	[−0.002, −0.0004]
PEPepoch x P3Bperson, γ12	−0.002	[−0.004, 0.001]
P3Bepoch, γ20	−0.018	[−0.036, 0.001]
Block, γ30	−0.077*	[−0.100, −0.055]
Frustration, γ40	−0.047*	[−0.085, −0.008]
PEPepoch x P3Bepoch, γ50	−0.003	[−0.007, 0.001]
PEPepoch x Frustration, γ60	−0.014*	[−0.025, −0.002]
P3Bepoch x Frustration, γ70	0.007	[−0.029, 0.043]
PEPepoch x P3Bepoch x Frustration γ80	0.013*	[0.005, 0.021]
Random Effects
Intercept, σ2u0i	1.493	[1.239, 1.760]
PEPepoch, σ2u1i	0.0007	[0.0003, 0.0012]
Corr Intercept, PEPepoch, ρ[u0i, u1i]	0.001	[−0.251, 0.254]
Residual, σ2eit	0.356	[0.340, 0.371]
Model Fit
AIC	9556
BIC	9672
−2LogLik	−4760

Note: N = 4559 observations nested within 257 persons; CI = confidence interval; AIC = Akaike information criterion; BIC = Bayesian information criterion; Corr = correlation

^*indicates p < .05.

3.1. Between-Person Cortical-Autonomic Associations

Consistent with the between-person notion of reciprocal coordination, there was a significant association between individual differences in PEP and initial level of RSA, γ₀₁ = 0.016 (p = .01), such that, as shown in the top panel of Figure 4, higher levels of RSA (higher parasympathetic activation) were related to longer PEP values (lower sympathetic activation). However, individual differences in P3b did not moderate this association, γ₀₃ = −0.00004 (p = .99), and were not related to individual differences in RSA γ₀₂ = 0.005 (p = .86). The lack of evidence for both the hypothesis that cortical activity is related to autonomic function and the hypothesis that cortical activity is a regulator of autonomic coordination here highlight that the interrelated function of these systems is not a between-person phenomenon, but a within-person phenomenon.

Figure 4.

Between- and within-person associations of RSA and PEP. Top panel: Solid black line shows the positive between-person association of children’s average PEP and average RSA. Bottom panel: Line plots of the within-person association between epoch-to-epoch PEP and epoch-to-epoch RSA. Solid black line shows the prototypically positive within-person association, indicating reciprocal coordination of sympathetic and parasympathetic activity. Light gray lines show the range of individual differences in the extent of coordination.

3.2. Within-Person Cortical-Autonomic Associations

Systematic within-person changes in RSA were evidenced by both the significant linear decline in RSA across the task blocks, γ₃₀ = −0.077 (p < .001), and a significant effect of the frustration manipulation on RSA. Independent of the greater physiological effort/engagement that ensued linearly as the task progressed, RSA was lower, on average, in the frustration block (Block 2) than in the reward blocks (Blocks 1 and 3), γ₄₀ = −0.047 (p = .02). This suggests that the affective manipulation influenced individuals’ parasympathetic function in the hypothesized and intended way. As well, and in line with the polyvagal hypothesis, there was some evidence suggesting that when cortical activation was relatively higher, there was a relative decrease in RSA, γ₂₀ = −0.018 (p = .06), but this association was not statistically significant.

Also in line with our hypotheses, there was evidence of within-person coordination between autonomic branches. For the prototypical child, the two branches of the autonomic system exhibited reciprocal coordination; epoch-to-epoch changes in RSA across the task were significantly and positively associated with epoch-to-epoch changes in PEP, γ₁₀ = 0.018 (p < .001). This result is illustrated in the bottom panel of Figure 4, where the solid bold line depicts the tendency for PEP to be relatively shortened (sympathetic system engaged) in epochs where RSA is relatively lower (parasympathetic system withdrawn), and vice versa; in epochs where RSA is relatively higher (parasympathetic system engaged) PEP is relatively lengthened (sympathetic system withdrawn). Notably, however, the light gray lines illustrate the extensive individual differences in both the strength and direction of this within-person association.

Of note, although the extent of within-person autonomic coordination was not moderated by individual differences in P3b, γ₁₂ = −0.002 (p = .15), it was moderated by between-person overall levels of PEP. Specifically, the strength of autonomic coordination was lower among children who had longer PEP values overall, γ₁₁ = −0.001 (p < .001), suggesting that these individuals were more likely to accomplish changes in arousal through independent movement of the parasympathetic system (i.e. the purple line in Figure 2).

Specific to our main research question, the within-person association between RSA and PEP across epochs was not moderated by within-person changes in P3b, γ₅₀ = −0.003 (p = .17), suggesting that autonomic coordination is not being regulated by changes in attentional allocation. Autonomic coordination was significantly moderated by the frustration manipulation γ₆₀ = −0.014 (p = .02), consistent with the proposition that autonomic coordination is more consistently related to affective rather than cognitive demands. However, these effects were qualified by a significant 3-way interaction indicating that the frustration manipulation moderated the association between P3b amplitude and extent of coordination between RSA and PEP, γ₈₀ = 0.013 (p = .001). This 3-way interaction (which supersedes γ₅₀ and γ₆₀) is shown in Figure 5. The left panel shows that, during non-frustration blocks, there was a positive association between RSA and PEP, and the strength of this is association did not differ as a function of individuals’ relative amplitude of P3b (within-person). In other words, when an individual’s P3b amplitude was relatively lower during a reward block (−1SD = −1.48) compared to other blocks the degree of coordination between autonomic branches was not different from when an individual’s P3b amplitude was relatively higher during a reward block (+1SD = +1.48) compared to other blocks. In contrast, as illustrated in the right panel, when an individual’s P3b amplitude was relatively lower during the frustration block (−1SD = −1.48) compared to other blocks the degree of coordination reversed from reciprocal (positive) coordination to compensatory (negative) coordination. Follow-up probes using Johnson-Neyman method (Preacher, Curran, & Bauer, 2006; implemented using interactions package in R, Long, 2019) revealed that in the no-frustration blocks, the within-person association between RSA and PEP exhibited reciprocal coordination (β_1i ≥ 0.01) when an individual’s P3b amplitude was within −6.02 (the most negative P3b deviation = −4.07 SD) and +2.22 of their average P3b (that is, there was reciprocal coordination unless P3b was relatively high, > 1.37 SD); whereas in the frustration block, the within-person association between RSA and PEP exhibited compensatory coordination (β_1i ≤ −0.01) when an individual’s P3b amplitude was more than −1.94 below their average P3b (that is, compensatory coordination emerged when P3b was relatively low, < 1.31 SD). In sum, the results suggest that when frustration invokes cognitive disengagement, parasympathetic and sympathetic systems may end up working against each other.

Figure 5.

Moderation of the within-person association between RSA and PEP by P3b and Frustration (3-way interaction) Left panel: During non-frustration blocks, differences in extent of coordination of RSA and PEP (indicated by slope of the lines) when P3B is low (−1SD = red dashed line) or high (+1SD = green solid line) are negligible. Right panel: During the frustration block, coordination of RSA and PEP is negative when P3b is low (red dashed line) but not when P3b is high (green solid line). The pattern of slopes shows that while the parasympathetic and sympathetic systems generally work together in reciprocal coordination, when frustration invokes cognitive disengagement, the systems may end up working against each other.

4. Discussion

The present study examined patterns of autonomic coordination to determine whether coordination was associated with characteristics of individuals, contextual demands, or dynamic fluctuations in cognitive processes. In order to examine how autonomic and cortical systems respond dynamically to challenge, participants were administered a cognitive inhibitory control task that was embedded within an affective manipulation. Analyses indicated that individuals with higher overall levels of RSA also had longer PEP intervals on average. In addition, dynamic within-person processes indicated significant reciprocal coordination between RSA and PEP from epoch to epoch, which is somewhat inconsistent with the notion that changes in arousal as a function of task demands are typically supported by parasympathetic withdrawal independent of sympathetic activity (e.g. Porges, 2001). However, substantial between-person variation in the extent of autonomic coordination was also evident. This study further sought to examine whether autonomic coordination was moderated by affective demands of task condition, or by dynamic fluctuations in cortical activation. Interestingly, a significant 3-way interaction emerged indicating that autonomic coordination was only associated with cortical activation in the frustration condition.

4.1. Autonomic-Cortical Associations

A confluence of evidence supports an association between individual differences in RSA and cognitive task performance (e.g. Marcovitch et al., 2010; Staton et al., 2009), with a more limited range of studies extending this association to physiological measures of frontal cortical activity including the P3b ERP (Kaufmann et al., 2012). Findings from the current study, however, failed to replicate this effect, as individual differences in average P3b amplitude were not found to correlate with initial RSA levels in the current study. Although no association was detected across individuals, a marginal effect (p = .05) indicated potential evidence for a correspondence between P3b amplitude and RSA dynamically across the task. Specifically, individuals’ RSA levels were lower during epochs in which their P3b amplitudes were higher, suggesting that withdrawal of RSA corresponded dynamically with cognitive function. This pattern is consistent with the Polyvagal Model that proposes that RSA withdrawal is an adaptive mechanism by which to initiate a regulated increase in physiological arousal in response to mild challenges (Porges, 2007).

Results of the present study further examine the tenets of the Polyvagal Model by additionally examining concurrent sympathetic activity. The Polyvagal Model specifically proposes a hierarchical relationship between sympathetic and parasympathetic activation, which suggests that in relatively low-threat contexts the parasympathetic system is sufficient to meet arousal demands (Porges & Furman, 2011). This proposition leads to the hypothesis that activity in each branch of the autonomic nervous system function relatively independently. In contrast, our results indicated correlations between sympathetic and parasympathetic function at the between-person and within-person levels. The autonomic space model demonstrates empirically the ability of each branch of the autonomic nervous system to act independently of the other branch (Berntson et al., 1994). This ability suggests that the autonomic system possesses a great degree of flexibility in responding, that covers the full spectrum of combinatorial patterns. As such, it is reasonable to hypothesize that coordination is not a trait characteristic, but rather a phenomenon that may emerge in response to situational demands, while maintaining the flexibility to diverge and converge as needed.

4.2. Affective Context

The ability to rapidly increase or decrease arousal that would be facilitated by reciprocal autonomic coordination may be particularly relevant to affective responding, and autonomic activation can be regulated by midbrain regions responsive to affective state (Ulrich-Lai & Herman, 2009). In previous analyses we found that autonomic coordination varied across an emotion induction paradigm, with the strength of coordination moderated by the motivational direction of the affective state. Specifically, reciprocal coordination was greater while children viewed approach-oriented emotional scenes (happy, angry) than avoidance-oriented scenes (fear, sad) (Gatzke-Kopp & Ram, 2018). We postulate that approach motivation is facilitated by the ability to significantly and efficiently increase arousal to support behavioral responses, where avoidance motivation may benefit from a de-coupling of autonomic branches. Avoidance motivation may still involve an increase in arousal (such as in the case of fear) but is more likely to be accompanied by behavioral inhibition while remaining vigilant, and thus the ability of the branches to move independently to support this complex set of goals may be adaptive.

The current study extends this finding by examining coordination across a different type of affect induction. Consistent with the previous findings, autonomic coordination was evident across the reward blocks, when approach-motivation should be high (Brenner, Beauchaine, & Sylvers, 2005). A moderating effect of the frustration condition was superceded by the 3-way interaction including cortical activation (P3b). Specifically, autonomic coordination differed as a function of whether individuals had relative increases or decreases in P3b in response to the frustration condition. The magnitude of P3b in this context may provide insight into how individuals respond affectively to the frustration condition. The frustration condition is defined by the removal of a previously provided incentive, and the lack of reward may suggest that approach-motivation would be dampened relative to the reward blocks. However, frustration, similar to anger, may be an approach-oriented emotion where an increase in arousal supports efforts to overcome obstacles to a goal. Research examining this hypothesis indicates that the extent to which anger is associated with approach-oriented patterns of EEG asymmetry is a function of whether the individual feels empowered or helpless to resolve their frustration (Harmon-Jones, Sigelman, Bohlig, & Harmon-Jones, 2003). In the current study, individuals who demonstrated relatively higher P3b amplitude during the frustration block continued to demonstrate reciprocal autonomic coordination, suggesting that they remained in an approach-oriented state, persevering in the face of frustration. In contrast, individuals who demonstrated a reduced P3b in response to frustration showed an inverse pattern of autonomic coordination. This may reflect a tendency for these individuals to disengage from the task.

It is unknown whether the ability to respond to the frustration condition by increasing attentional control resulted in greater reciprocal coordination through top-down cortical regulation of autonomic arousal, or whether the ability to maintain a stable pattern of reciprocal coordinated autonomic function enabled the allocation of additional attentional resources. The connectivity between central and autonomic systems includes both efferent and afferent pathways, and autonomic states have been shown to affect cognitive processing (Critchley & Garfinkel, 2015; Garfinkel et al., 2016). If cortical activation serves to regulate autonomic coordination, it might be expected that decreases in cortical activation observed during the frustration condition would be associated with a lack of coordination. In contrast, there appeared to be a reversal in the nature of coordination rather than an effect on strength of coordination. Taken together, autonomic coordination appears to be primarily modulated by affective/ motivational state. It is possible that dynamic changes in cortical activation (P3b) would be associated with dynamic changes in parasympathetic activation (RSA) in a cognitive task that was not presented within the context of an incentive manipulation.

It is important to note that although the pattern of reciprocal autonomic coordination was significant, the correlation between epoch-level sympathetic and parasympathetic activity was modest (r = −0.10). In other words, the prototypical pattern of coordination identified by the model does not necessarily characterize all individuals in the sample. Individuals varied extensively with regard to the strength of association (whether changes in one branch correlated with changes in the other) and direction (whether such changes were better characterized as reciprocal or compensatory). These findings hint at the possibility that psychophysiological coordination may be a meaningful measure of individual differences, beyond autonomic reactivity of each branch on its own. The current findings suggest that additional research is needed to establish a knowledge base of experimental paradigms that are likely to elicit certain patterns of autonomic function to guide research examining the implications of individual differences in coordination.

By focusing on within-person dynamics in autonomic movement, the analytical approach used here does not make a distinction between all the autonomic reactivity profiles depicted in Figure 1. The repeated measures analysis does captures movement along the reciprocal axis, but does not differentiate individuals who move toward reciprocal sympathetic and reciprocal parasympathetic poles. Given that the implications for arousal are markedly different between these two profiles, it may be of value to examine whether individual differences exist not just in the tendency to move along this corridor, but a tendency to move in one particular direction in response to task demands (and perhaps return through a different corridor). Research suggests that central-autonomic integration engages distinct profiles and combinations of peripheral arousal to support, and potentially define, distinct affective states (Critchley, 2009).

4.3. Limitations and Future Considerations

Several aspects of the current study should be noted when interpreting the findings. First, the examination of autonomic coordination in the present study was restricted to the cardiac system. Although we have previously reported evidence of coordination between RSA and electrodermal activity (Gatzke-Kopp & Ram, 2018), the implications of autonomic coordination in the same target organ may differ from coordination across different organs. Indeed, some evidence indicates that specific nuances of situational demands result in different patterns of autonomic activation across multiple organs (Critchley, 2009; Saper, 2002), and as such the extent to which the findings reported here replicate with other physiological indices requires empirical investigation. Likewise, the P3b represents only one possible index of cortical activation and it is possible that other cortical markers of cognitive function (i.e. inhibitory control) may demonstrate a stronger association with autonomic coordination.

Second, as with any study, several characteristics of the study sample limit the extent to which findings generalize to other populations. Participants in this study were children assessed during their kindergarten year. Although we previously failed to detect any effect of age on autonomic coordination, data were only available through Grade 2. It remains unclear whether major developmental changes, such as occurs during adolescence, would influence the processes examined here. In addition, the majority of participants in this sample were drawn from and are representative of a diverse population. Although it remains critically important to expand basic research to historically understudied populations (see Gatzke-Kopp, 2016), the demographics of this sample may limit the comparability of these findings to a literature focused on study of more affluent and/or samples of European descent. Some research suggests an impact of socioeconomic adversity on autonomic and cortical function (Giuliano et al., 2018a), and thus the current findings may reflect adaptations to experiences of stress not typical in all populations. In addition, some research has reported different patterns of association between parasympathetic function and cognitive performance in African-American compared with European-Americans (Jennings, Allen, Gianaros, Thayer, & Manuck, 2015). Furthermore, children participating in the study had been oversampled for aggressive behavior. Given that aggressive children are hypothesized to display distinctive patterns of physiological regulation (Beauchaine et al., 2013; Lorber, 2004), the use of such a sample further precludes generalization to a normative sample.

In line with current measurement norms, our results are based on up to 21 repeated measures that indicate how cortical and autonomic systems operate during temporal epochs that were 30-second (RSA and PEP) or longer (P3b). This allowed for precise examination of the coordination and interaction among these systems – at a specific time-scale that may or may not match the actual time-scale at which these systems operate and/or influence one another. Although our analysis pushes forward consideration of autonomic coordination as a within-person process, future work should locate and make use of measurement paradigms and designs that support examination of how these systems operate at faster time-scales (e.g., second-to-second).

4.4. Conclusion

Overall, our results suggest that reciprocal autonomic coordination may have adaptive implications in meeting cognitive task demands, particularly in emotionally arousing contexts. Consistent with previous results, coordination appears to support emotions consistent with behavioral activation, regardless of affective valence. Coordination between cortical and autonomic systems was evident, but only when examined as a within-person process. These findings highlight the importance of considering how autonomic activity spans the full spectrum of autonomic space dynamically in response to changes in situational demands.