Arousal: Reports of Its Demise May Be Premature

It was with great interest that we read the paper entitled “Arousal may not be anything to get excited about”. The authors are to be commended for addressing this important issue and have provided a cogent critique of non-specific or general arousal. We have written extensively over the years on this topic and share many of the same concerns and recommendations but are more sanguine about the issue (Friedman & Thayer, 1998; Friedman, 2010; Friedman & Thayer, 2024). Based on its continuing utility as we will illustrate here, we think that the reports of the demise of arousal as a useful construct in the study of human behavior may be premature.

Early discussions of the arousal concept are more nuanced than represented in this paper and in the usage by more modern researchers (Hebb, 1955; Osgood, 1962; Osgood & Suci, 1955; Schlosberg, 1954). This lack of nuance has contributed to the controversy about the term and its utility. However, the lack of appreciation of finer distinctions, including that arousal likely involves subsystems and not just one homogeneous system, and to differ both quantitatively and qualitatively (Hebb, 1955), is due to the problematic usage of the concept and not due to the concept itself.

In studies over several decades, we have found arousal to be informative in our understanding of emotion, and in particular its physiological concomitants. In one study of the cardiovascular concomitants of emotions in real-life settings (ecological momentary assessed emotion) in two samples of community-dwelling adults, we found that both negative (anxious/annoyed) and positive (elated/happy) emotions, derived from a circular emotion rating scale (circumplex), were associated with similar increases in blood pressure and heart rate (Jacob et al., 1999). Moreover, support for an arousal dimension, based on the combination of self-report and physiological responses, was provided by the use of circular-linear correlations (correlation between the orientation of the linear dimension that provided the best fit with the circular self-report data and the cardiovascular data), which revealed that the orientation of axes in the affective circumplex was consistent with an arousal, but not a positive affect/negative affect dimension. This raised questions about the consistently reported deleterious consequences of negative emotions. That is, if positive emotions were associated with blood pressure increases of similar magnitude as negative emotions, why were only negative emotions associated with poor health outcomes. To further investigate this discrepancy between the similar cardiovascular responses of positive versus negative emotions and differential health outcomes, a subsequent study examined the momentary and delayed heart rate responses during positive and negative emotional episodes (Brosschot & Thayer, 2003). We found that the initial increase in heart rate was similar for both positive and negative emotions and thus related to the arousal dimension, but the delayed responses were associated with valence, with negative emotions being associated with prolonged heart rate increases. Without conceptualizing these studies in the context of an arousal dimension separate from valence, it would have been difficult to explain our findings. In another study that examined heart rate and skin conductance responses to male and female pictures of emotion derived from the Ekman faces in a sample of men and women, we found that the affective space for women was more complex and included both valence and arousal dimensions whereas men showed only a valence dimension in their self-report and physiological responses (Johnsen et al. 1995; Thayer & Johnsen, 2000). Importantly, we found that small differences in arousal as indexed by skin conductance were significant for women in their affective responses to the faces (Johnsen et al., 1995). This study allowed us to reconcile the discrete emotion theories with dimensional emotion theories that were hotly contested at the time. Other studies from our group on autonomic nervous system specificity for basic emotions further support this perspective (Christie & Friedman, 2004; Stephens et al., 2010). These results also illuminate that arousal effects may manifest themselves differently in different groups of people, as suggested by the early proponents of arousal theory (Hebb, 1955; Schlosberg, 1954)

Recent conceptualizations of “arousal”, broadly defined, have placed it in the context of nonlinear dynamical systems models of behavior and emotion. Calderon et al. (2016) provide a detailed exposition of the concept of general arousal. They propose that general arousal (GA) is a “primitive and elementary non-specific neuronal ‘force’ that activates a set of ascending and descending systems” and may be defined by a differential equation with weighted parameters that represent different types of arousal, both named and unnamed, that are sensitive to both internal and external demands of the organism (Calderon et al. 2016).

We too, and others, have proposed a dynamical systems approach to emotions (Faith & Thayer, 2001; Friedman & Thayer, 2024; Globus & Arpaia, 1994; Thayer & Lane, 2000). In this conceptualization, arousal is a control parameter in a dynamical system, and along with valence, defines the state-space of emotion. For example, using P-technique factor analysis of self-reported affect to a wide range of affective stimuli, we showed that the dimensions of change in 20 out of 23 individuals involved a valence dimension and an arousal dimension (Faith & Thayer, 2001). In a prior study of physiological responses to musically-induced emotions, we found that ratings of arousal correlated with respiration frequency such that greater arousal was associated with higher respiration frequency (Nykliček et al., 1997). This is noteworthy, given a recent study reporting that fluctuations in arousal as indexed by respiratory variation and heart rate variability were causally related to the synchronization of global brain waves (Raut et al., 2021). Specifically, these researchers note that “traveling waves spatiotemporally pattern brain-wide excitability in relation to arousal” (Raut et al., 2021, p. 1). That is, afferent arousal signals from the periphery drive patterns of brain activity. Relatedly, we have shown that “ascending modulations from vagal activity precede neural dynamics and correlate with the reported level of arousal” during viewing of emotional films (Candia-Rivera et al., 2022, p. 1). We noted that “perceived arousal level modulates the amplitude of ascending heart-to-brain neural information flow” (Candia-Rivera et al. 2022). As such we provided evidence for William James's concept of the role of arousal in emotion (James, 1884). That is, physiological and behavioral responses (arousal) precede subjective emotional experience (Friedman, 2010). Another recent study showed that during up- and down-regulation of emotion, an overlapping set of brain regions including the insula, the dorsal anterior cingulate cortex, and the frontal pole were associated with the subjective reports of emotional intensity (Min et al., 2022). These latter two studies are consistent with the “valence general affective workspace hypothesis”, which states that valence is not mapped in the brain, but that arousal signals provide input for the brain that may trigger valenced emotional processes based on individual and environmental factors (Lindquist et al., 2016). The similarity of this more modern neuroscience-based theorizing to that of early arousal theories such as Schlosberg (1954) and Hebb (1955) is striking.

Based on its continuing utility as we have illustrated here, we think that the reports of the demise of arousal as a useful construct in the study of human behavior may be premature. Nevertheless, continued refinement and careful operationalization as suggested by Smith, Woodard, and Pollak will aid researchers as they apply the concept of arousal to emotion and related areas.