Skip to main content
NCBI home page
Social Cognitive and Affective Neuroscience logoLink to Social Cognitive and Affective Neuroscience
. 2020 Jan 28;15(1):1–11. doi: 10.1093/scan/nsaa012

Comparing three models of arousal in the human brain

Hadeel Haj-Ali 1, Adam K Anderson 2,3, Assaf Kron 1,✉
PMCID: PMC7171372  PMID: 31993651

Abstract

The bipolar valence–arousal model is assumed by many to be an underlying structure of conscious experience of core affect and emotion. In this work, we compare three versions of the bipolar valence–arousal model at the neural domain, using functional magnetic resonance imaging (fMRI). Specifically, we systematically contrast three models of arousal: model 1—‘arousal as a separate quale from valence’, model 2—‘arousal as intensity of bipolar valence’ and model 3—‘arousal as a linear combination of unipolar pleasant and unpleasant’. Using parametric modulation analysis, we estimated the ability of each model to predict activation in arousal-related brain regions, in response to affective stimuli. The results suggest that arousal is not separable from valence in its ability to predict arousal-related neural activity. The relevance of the results to the theory of conscious affect is discussed.

Keywords: arousal, valence, fMRI, parametric modulation

Introduction

Models of conscious experience of affect (Barrett and Russell, 1999; Yik et al., 1999; Larsen et al., 2001; Kron et al., 2013; Mattek et al., 2017) are a family of models that shape the metastructure of affect theory and experimentation (Kron, 2019). On the one hand, they provide scientists with the language to ask questions about affect and, on the other hand, they provide constraints on its specific structure, thus preventing scientists from asking questions about dimensions that are not included in the model. It is not surprising then that the debates about models of affect and emotions are long and intense (e.g. Barrett and Russell, 1999; Russell and Barrett, 1999).

Traditionally, models of affect assume that conscious experience of affect is composed of arousal, which refers to a sense of energy or excitation, and valence, which refers to the experience of pleasure and displeasure (e.g. Russell, 2003). However, theoretically, valence and arousal can relate to each other in different ways—for example, they can be thought of as two independent qualia or, alternatively, arousal can be a property of valence (e.g. its intensity). The exact relationship between valence and arousal depends on the specific model of affect that one holds. In this work, we compare three main versions of the valence–arousal model. A key test to the comparison between the three models, and the focus of this study, is whether arousal is a distinct quale from valence or simply the intensity of valence. The first model we examine (henceforth model 1) assumes that arousal is a separate quale from bipolar valence (e.g. Yik et al., 1999) while models 2 and 3 assume that arousal is the intensity of valence. The second model (henceforth, model 2) interprets arousal as the intensity of bipolar valence (Bradley et al., 2001), and model 3 suggests that arousal is a linear combination of two separate unipolar dimensions of pleasant and unpleasant (e.g. Kron et al., 2013). In the current study, we compare the three models of arousal in the neural domain. If arousal is a distinct quale from valence, as suggested by model 1, we expect model 1 to predict arousal-related neural activation better than models 2 and 3.

Model 1: arousal as a separate quale from valence

The first version of the bipolar valence–arousal model posits that the conscious experience of affect is composed of two distinct qualia: arousal, which ranges from feeling calm to excited, and bipolar valence, which ranges from pleasure to displeasure (e.g. Barrett and Russell, 1999). Critically, this version of the model assumes that arousal has its own intensity and does not merely reflect the intensity of valence (Kron et al., 2013; Kron et al., 2015). Accordingly, when using self-reports that rely on this version of the model, participants are asked to report their feelings using two continuous scales: one for valence and one for arousal (Bradley and Lang, 1994).

Since this model assumes that arousal is not merely the intensity of valence, one continuous scale for valence is not enough to capture the intensity of both valence and arousal. Critically, the model assumes that participants can report two separate qualia (valence and arousal) and their intensities. Empirical support for this model was accumulated from three sources: first, from numerous studies that show the two-dimensional latent structure of words describing the current affect (e.g. Yik et al., 1999); second, some affective modulations of peripheral responses are differentially correlated with valence and arousal (e.g. Lang et al., 1993), e.g. electrodermal activation is more correlated with self-reported arousal than with bipolar valence, while heart rate attenuation after stimulus onset is more correlated with the valence dimension than with arousal; third, some brain activations were also found to be differentially correlated with the two dimensions—e.g. amygdala activation is correlated more with arousal than with bipolar valence, while orbitofrontal activation is correlated more with the valence dimension than with arousal (e.g. Anderson et al., 2003).

One limitation of model 1 is that it fails to predict the scattering of stimuli in the bipolar valence–arousal space; when rated using two scales of arousal and valence, different types of stimuli, including words (Bradley and Lang, 1999), audio recordings (Bradley and Lang, 2007) and pictures (Lang et al., 1997; Kuppens et al., 2013) consistently show what has been termed in the literature as the ‘boomerang shape’ function, which implies a strong interdependence between valence and arousal (see Figure 1a,b). If the intensity of the bipolar valence and arousal dimensions were separate, then we would expect a degree of independence between the two dimensions and the stimuli, which cover the entire bipolar valence–arousal space in a random manner (see Figure 1d).

Fig. 1.

Fig. 1

Open in a new tab

(a) The bipolar valence–arousal space of pictures from the international affective pictorial system (IAPS—Lang et al., 1997). (b) The bipolar valence–arousal space of words from the Affective Norm for English Words (ANEW—Bradley and Lang, 1999). (c) The bipolar valence and its absolute value space. (d) Scatter of stimuli on hypothetical bipolar valence–arousal space when valence and arousal are independent.

Model 2: arousal as intensity of valence

The boomerang shape scattering implies rather a different model in which arousal is the intensity of bipolar valence. The boomerang shape function is similar to the function of bipolar scores plotted against their absolute values (see Figure 1c). Conceptually, the absolute value of bipolar valence is equivalent to its intensity. That is, if the boomerang shape reflects an underlying V-shape function plus measurement errors, this would suggest that arousal is not a separate quale from valence, but rather its intensity. Bradley, Lang and colleagues, who originally adopted model 1 (i.e. ‘arousal as a separate quale from valence’—e.g. Lang et al., 1993) and used it to map stimuli in the stimuli pools (e.g. Lang et al., 1997), interpreted the boomerang shape as reflecting an absolute value function and moved to model 2 in which arousal is the intensity of bipolar valence (Bradley et al., 2001; see also Cunningham et al., 2004). The main advantage of model 2 (arousal as intensity of bipolar valence) is its ability to explain the scattering of stimuli in the bipolar valence–arousal space.

Nonetheless, model 2 has its limitations. Assuming that arousal is the intensity of valence suggests that every phenomenon that is explained by arousal should be explained, to the same degree, by the absolute value function (or quadratic function) of bipolar valence. Interestingly, and inconsistent with model 2, self-reports of bipolar valence explain only around 20% of arousal variance (as computed by Ito et al., 1998; Lang et al., 1999). In summary, both of the aforementioned versions of the bipolar valence–arousal model are inconsistent with the accumulating evidence: model 1—‘arousal as a separate quale’—fails to explain the boomerang shape of the stimuli pools and model 2—‘arousal as intensity of valence’—fails to explain cases of independence between arousal and absolute value of valence (or quadratic function).

Model 3: arousal as the summation of intensities of unipolar pleasant and unpleasant scales

In an attempt to explain existing data, comparing the adequacy of versions 1 and 2 of the model yields no conclusive results, suggesting that arousal is not a separate quale from bipolar valence on the one hand, nor the intensity of valence on the other. Recently, we suggested a model that reconciles these inconsistencies (Kron et al., 2013; Kron et al., 2015). We started with the bivariate valence model (e.g. Cacioppo and Berntson, 1994), which suggests that valence consists of two separate dimensions: one is unipolar pleasant (PL) and the other is unipolar unpleasant (UN) (see Figure 2a). The bivariate valence model posits that pleasure and displeasure systems can operate independently. There is a strong relationship between self-reports of PL\UN of the bivariate model and self-reports of the bipolar valence scale (Kron et al., 2013; Kron et al., 2015). Specifically, subtracting unipolar unpleasant scores from unipolar pleasant scores results in a vector that is almost identical to bipolar valence. The subtraction of UN from PL (PL-UN) explains around 92% of the variance of bipolar valence scores, meaning they are the same construct. The crux of the argument is that if the valence structure in conscious experience is bivariate, yet ratings are measured using a bipolar scale, reports provided by participants might then reflect the subtraction of unpleasant from pleasant. In the subtraction, information about pleasant and unpleasant feelings is inevitably lost. Consequently, an additional axis of arousal is needed to recover this lost information. That is, when valence is not measured properly, i.e. when using a bipolar valence scale, the need for arousal becomes crucial.

Fig. 2.

Fig. 2

Open in a new tab

The conversion of pleasant and unpleasant scales into bipolar valence and arousal scales. When mapped on the dual unipolar valence space (see plot a), the erotic picture and the picture of a coffee mug differ in the values of both pleasant and unpleasant valences. However, when the same pictures are mapped onto the bipolar valence–arousal space, they differ only in their arousal scores (see plot b) but not in their valence values. According to model 3, when PL + UN is plotted against PL−UN, a similar pattern to plot b emerges (see plot c), i.e. the erotic picture and the coffee mug do not differ in their PL−UN score, but they do differ in their PL + UN score.

To illustrate the loss of information in this subtraction, consider the following example: in plot 2a, an erotic picture and a coffee mug are rated using two unipolar scales and are different in both pleasant and unpleasant scales. In the subtraction of UN from PL, which is equivalent to reporting on the bipolar valence scale, the two pictures have the same valence value, that is, information about differences between pleasure and displeasure is lost (see Figure 2b). Consequently, it could be that the arousal axis is required in order to ‘recover’ the information that was artificially lost when converting bivariate valence into the bipolar structure. In other words, arousal is equal to the linear combination of two unipolar dimensions: pleasant and unpleasant (see Figure 2c). This explains why arousal is not a separate quale (it reflects information about PL and UN), and at the same time is not identical to the absolute value of bipolar valence (it contains information about PL and UN that is lost in the subtraction of UN from PL). The third model of arousal (arousal as a linear combination of PL and UN) provides clear predictions: everything that can be predicted by self-reported arousal could be predicted, to the same degree, by the linear combination of unipolar axes of pleasant and unpleasant. We supported this assumption by showing that the linear combination of unipolar pleasant and unpleasant predicts electrodermal activation to the same degree as self-reported arousal does (Kron et al., 2013; Kron et al., 2015).

In sum, we presented three versions of the valence–arousal model. Model 1 assumes that arousal and valence are separate qualia. Models 2 and 3 assume no separation between arousal and valence. Specifically, model 2 assumes that arousal is the intensity of bipolar valence, and model 3 assumes that arousal is the sum of pleasant and unpleasant in the bivariate model. A key factor in deciding between the models is the question whether arousal is or is not a distinct quale from valence.

Overview of the current study

In the current experiment, we compared the abilities of three models to predict arousal-related activation in the human brain: model 1—‘arousal as a separate quale’; model 2—‘arousal as intensity of bipolar valence’ and model 3—‘arousal as a linear combination of unipolar pleasant and unpleasant’. This specific design has three advantages over previous studies: (1) it is, to the best of our knowledge, the first to directly compare the three models; (2) while previous studies did compare models 1 and 3 (Kron et al., 2013; Kron et al., 2015), this was to predict electrodermal activation that is part of the orienting response, which is only one specific component of the affective response; (3) comparing the models in the neural domain makes it possible to examine more flexible patterns where bipolar valence and arousal have different structures according to the level of processing and neuroaxis (Cacioppo et al., 1999).

In the current study, we examined correlations between self-reports and neural activations to compare three models of feelings. Arguably, the straightforward way to compare models of feelings is in the self-reports domain, which is the best proxy to qualia (Barrett et al., 2007). Yet, when self-reports are not decisive, correlations of self-reports with neural (or physiological) data can be used as a means to validate models of emotions (e.g. Kron et al., 2013). The starting point of our argument is that participants do not judge valence and arousal to be distinct. One support for this claim is the consistent, high correlation (‘boomerang shape’) scatter of stimuli we find across all known pools of emotional stimuli (see ‘Model 1’ and ‘Model 2’ sections above). Consequently, if we were to find that self-reports of arousal explain neural activation better than self-reports of bivariate valence, then that would provide at least some support for the distinction between arousal and valence in conscious experience.

The experiment consisted of two independent parts: (a) main experiment and (b) localizer task. The main experiment was presented in an event-related design, during which participants viewed emotional visual stimuli in the MRI scanner while providing self-reports about their emotional response. Model 1—‘arousal as a separate quale’—was estimated by the traditional arousal self-reports, model 2—‘arousal as intensity of bipolar valence’—was estimated by the absolute values of the bipolar valence scale and model 3—‘arousal as a linear combination of pleasant and unpleasant’—was estimated by a summation of unipolar pleasant and unpleasant scales. Using parametric modulation analysis, we compared the ability of the three models to explain neural activation within arousal-related regions of interest. The localizer task was used to define arousal-related regions of interest (ROIs). To give model 1 (arousal as a separate quale from valence) the best chance, ROIs were defined according to model 1—that is, stimuli that were used to define the ROIs were selected according to bipolar and arousal self-reports (for elaboration on ROIs definition, see ‘Materials and methods’ section: Regions of interest (ROIs) analysis). We avoided circularity of ROIs definition (Kriegeskorte et al., 2009) to enhance the locating of neural representations of arousal that are independent of valence.

The logic of the current examination is as follows: If arousal is a separate quale from valence, then, in the self-report process, participants introspect and report separately about feelings of valence and their intensity on the valence scale, and feelings of arousal and their intensity on the arousal scale. In this case, we predict that arousal is a unique component that has a special link to arousal-related activation and, consequently, self-reports about arousal (model 1) will show an advantage over intensity of valence (models 2 and 3) in predicting arousal-related activation. However, if arousal is merely the intensity of valence, this means that when reporting about arousal, participants do not introspect and report about a different quale with a separate intensity from valence. In this case, we predict no advantage of self-reports of arousal over self-reports that reflect the intensity of valence in predicting arousal-related activations.

Finally, the comparison between the relationship of arousal and valence in the three models should come with a caveat; claiming that arousal is inseparable from valence (i.e. models 2 and 3) does not necessarily mean that arousal is not a distinct subsystem of emotion or affect. In the emotion literature, the term arousal is used in at least four different contexts (see also Satpute et al., 2018; Mather, in press). First, it is used to describe systems that relate to autonomic activation, such as the reticular activating system and the autonomic nervous system (e.g. Lang et al., 1993). Second, a system that controls wakefulness (Rainnie et al., 1994; Sakurai, 2007) and maintains consciousness and cognitive functions (Mather, in press). Third, as a name given to a factor in latent variable models (e.g. Yik et al., 1999). In the latent variable analysis, participants are not asked to report about the level of arousal; rather, arousal is a title given by the researcher to a group of items that are believed to reflect the commonality of arousal. Fourth, the term arousal refers to the observed (not latent) self-report scale (e.g. Lang et al., 1999) in which participants are asked to report about feelings named arousal. In this work, we do not argue against the specificity of physiological arousal (autonomic or wakefulness), nor do we claim that there is no such experience as arousal that is independent of valence (such as a feeling of arousal when a person walks up the stairs). Our claim concerns the meaning of arousal in the context of reporting about affective experiences.

Materials and methods

Participants

Thirty participants (17 women, mean age = 28.3 ± 3.5 years, right-handed people assessed by self-disclosure) from the University of Toronto were recruited for the study in return for payment ($80). Four participants were excluded from the study due to missing values, scanning problems and zero variability in self-report values. All participants gave their consent before participating in the experiment and reported no history of neurological and psychiatric disorders.

Stimuli

A total of 72 pictures were sampled from the International Affective Picture System (IAPS, Lang et al., 1999). In order to overcome the inherent independence of bipolar valence and arousal in the IAPS pool (the ‘boomerang shape’ of the stimuli on the space—e.g. Bradley et al., 2001), we used an in-house algorithm to select the stimuli (see Kron et al., 2013; Kron et al., 2015; Hamzani et al., 2019). The algorithm selected pictures that maximally covered all combinations of valence and arousal of the IAPS space, making sure that all the images were spread across this shape in a uniform manner (see Figure 3). This picture sampling procedure permitted optimal independence between the bipolar valence and arousal dimensions (see Kron et al., 2013, Kron et al., 2015, Hamzani et al., 2019).

Fig. 3.

Fig. 3

Open in a new tab

Scatter plot on the bipolar valence and arousal axes of the selected stimuli according to IAPS norms.

Self-reports

Self-reports were measured using two pairs of scales: (a) bipolar valence–arousal scales consisting of a bipolar valence scale ranging from −4 (very unpleasant) to 4 (very pleasant), and an arousal scale ranging from 0 (calm) to 8 (aroused); and (b) two unipolar scales for pleasant and unpleasant feelings ranging from 0 to 8. Within each subject, half of the stimuli were measured using bipolar valence and arousal scales, and the other half were measured using two unipolar scales for pleasant and unpleasant feelings.

Design

The experiment consisted of two independent parts: main experiment and localizer task. The main experiment was presented in an event-related design and the localizer task was presented in a block design. The experiment was programmed with E-Prime 2 professional software (Schneider et al., 2002).

Main experiment

The main experiment consisted of 72 trials with randomly assigned stimuli presented in an event-related design. Each trial included one visual stimulus preceded by a fixation point, for a duration of 8000 ms interstimulus interval (on average) and followed by a 5000 ms blank screen, which was followed by either a bipolar valence–arousal scale or two unipolar scales (see Figure 4). Each stimulus was presented for 6000 ms. Participants had 5000 ms per scale to provide the self-report, using a response button box to move right or left on the scale. In order to prevent any repeated exposure effects, half of the stimuli within each subject were followed by bipolar valence and arousal scales, and the other half were followed by two unipolar scales for pleasant and unpleasant feelings.

Fig. 4.

Fig. 4

Open in a new tab

Schematic trial example. Each IAPS stimulus was followed either by bipolar valence and arousal scales, or by two unipolar scales for pleasure and displeasure.

Localizer

The main experiment within each subject was followed by a localizer run, which aimed to functionally specify regions of interest, i.e. regions that are arousal specific. During the localizer run, participants passively viewed IAPS stimuli in a block design manner, which were aligned to six conditions: high arousing positive stimuli, low arousing positive stimuli, high arousing negative stimuli, low arousing positive stimuli, positive stimuli and negative stimuli. Each block consisted of five stimuli and lasted for 20 s. Figure 3 presents the scatter plot of the selected stimuli according to IAPS norms.

Procedure

The study was approved by the office of research ethics, University of Toronto. After the participants had given their consent and prior to the main experiment, they practiced reporting on both valence and arousal scales, and on unipolar pleasant and unpleasant scales outside the scanner. The experiment was then started in the scanner with the main experiment first, followed by the localizer task.

fMRI data analyses

fMRI data acquisition and preprocessing is available online as supplemental material (see SOM1).

Regions of interest (ROIs) analyses

Arousal-related ROIs were specified by the following three-step procedure; in the first step, we conducted whole-brain analysis on the localizer run (see ‘Materials and methods’ section: Localizer) based on arousal contrast [(high positive pictures + high negative pictures)—(low positive pictures + low negative pictures)]. We extracted all functional regions that survived the threshold of P = 0.0001 (uncorrected), cluster size = 10, using MarsBaR toolbox 0.44 (Brett et al., 2002) for SPM (Friston et al., 1995; https://www.fil.ion.ucl.ac.uk/spm/). In the second step, independent from the first step, we generated anatomical masks for a priori arousal-related regions based on previous literature utilizing WFU_PickAtlas (Tzourio-Mazoyer et al., 2002; Maldjian et al., 2003). The a priori regions were as follows: right and left extrastriate visual cortex (Brodmann areas 18 and 19; Lang et al., 1998; Lane et al., 1999; Todd et al., 2012), left and right amygdala (Cahill et al., 1995; Furmark et al., 1997; Anderson et al., 2003; Reyes et al., 2011), left and right thalamus (Paus et al., 1997; Van der Werf et al., 2002; Schiff, 2008), midbrain (Safron et al., 2007; Critchley, 2009), left and right insula (Nagai et al., 2004), left and right cerebellum (Critchley et al., 2000), pons (Critchley et al., 2002), putamen (Williams et al., 2001) and anterior cingulate cortex (ACC; Zhang et al., 2014). In the third step, the a priori anatomical masks were combined with the functional activations extracted from the whole-brain analysis. Final ROIs were defined as the regions of activations extracted from the whole-brain analysis that overlapped with the a priori anatomical mask. A priori regions that showed no overlap with the whole-brain analysis and, hence, were not defined as ROIs, are the right amygdala, pons, putamen and ACC.

Parametric modulation analysis

We performed a parametric modulation analysis to compare the three models of arousal. Data were analyzed using the General Linear Model (GLM), as implemented in SPM12, in Matlab (2014) and were modeled with a parametric modulation design. As a reminder, self-reports were collected from participants in the main experiment, in which participants were presented with stimuli in an event-related manner. After each stimulus, participants were asked to provide self-reports using either two unipolar pleasant and unpleasant scales, or bipolar valence–arousal scales. The parameter values that were encoded in the design matrices are the averaged standardized self-reports of all participants. For each participant, we built four design matrices; in the first, we encoded the averaged self-reported arousal (model 1) as a regressor. The second matrix included the averaged self-reported PL + UN (model 3); the third design matrix included the averaged self-reported absolute values of bipolar valence (model 2); and the fourth design matrix included the averaged self-reported bipolar valence. Next, we conducted a parametric modulation analysis with respect to the data of the main experiment i.e. ‘event-related’. In the first level of analysis, data were high-passed filtered at 128 Hz. Contrast images from the first level of analysis were then submitted to a second level of analysis, P = 0.001 (uncorrected). Finally, using MarsBaR (Brett et al., 2002), we extracted the parameter estimates of the parametric modulation analysis for the defined ROIs only.

Analytical strategy

The logic of the current examination is that if arousal is separable from valence (separate quale), we predict that self-reports about arousal (model 1) will show an advantage over intensity of valence (models 2 and 3) in explaining arousal-related activation. However, if arousal is inseparable from valence, we predict no advantage of self-reports of arousal in explaining arousal-related activations, i.e. no difference between the three models will be found. To overcome the potential problems that come with accepting the null hypothesis, we support the traditional inference with Bayesian analysis. We used parametric estimates (see ‘Parametric modulation analysis’ section above) to compare the three models of arousal. Specifically, we used parameter estimates for four types of self-reports: arousal (model 1), absolute value of bipolar valence (ABS) (model 2) and pleasant plus unpleasant (PL + UN) (model 3). To make sure activation is specific to arousal, we compared each arousal measure to self-reports of bipolar valence.

Comparisons were conducted in two steps; first, we compared the three parameter estimates within each ROI, then we compared them across all ROIs—henceforth ‘overall analysis’. For the within-region comparisons, we examined the parameter estimates of each arousal measure (arousal, ABS, PL + UN) and bipolar valence, in each ROI. In the overall analysis, we calculated the weighted averages for the four parameter estimates (three arousal measures and bipolar valence) across all ROIs. The weight for each parameter estimate was the ratio between number of voxels in the ROI and total number of voxels in all ROIs. This comparison provided a ‘bird’s-eye view’ of all the voxels across all ROIs. An additional analysis for functional activations extracted from whole brain analysis is available online as supplemental material (see SOM2).

Bayesian analyses were conducted with the JASP software (0.8.6.0 version 2018). For each contrast, we computed Bayes Factor (BF) that provides information about the ratio between the strength of evidence of the alternative hypothesis and the strength of evidence of the null hypothesis (Dienes, 2014). BFs are notated as BF10 and BF01. BF10 indicates support of the alternative hypothesis over the null hypothesis. The greater the BF10 is than 1, the stronger is the support for the alternative hypothesis over the null hypothesis, given the data. BF01 indicates support for the null hypothesis over the alternative hypothesis; the greater the BF01 is than 1, the stronger is the support for the null hypothesis over the alternative hypothesis, given the data.

Results

Reliability

Split-half reliabilities were calculated for the following self-reports: bipolar valence, arousal, pleasure, displeasure, absolute values of bipolar valence (ABS), PL + UN and PL−UN. For each self-report, participants were divided into two halves, and mean self-reports of all stimuli within each half were calculated. Next, Pearson’s correlations between the two halves were calculated. Reliability measures (Pearson’s correlations between the two halves) were as follows: 0.76 for arousal, 0.88 for bipolar valence, 0.87 for displeasure, 0.88 for pleasure, 0.75 for PL + UN, 0.76 for ABS and 0.9 for PL−UN.

Self-reports

First, we examined the intercorrelations between the three measures of arousal (arousal, PL + UN and absolute values of bipolar valence—ABS). Table 1 shows the Pearson’s correlation, P-values, Bayes factor and scatter plot of the 10 analyses. Consistent with previous results (Kron et al., 2013), strong correlations were found between arousal reports, PL + UN and absolute values of bipolar valence (see Table 1 cell nos. 2, 3 and 8 for scatter plots and cell nos. 6, 11 and 12 for inferential statistics). Specifically, PL + UN and ABS explained 53% and 37% of the variance of arousal ratings, respectively. Given that the reliability between the two sets of arousal scores (split-half correlation) was 0.76 (see ‘Reliability’ section above), the ceiling of variance explained by arousal scores was 57%. Thus, PL + UN scores explained the great majority (92%—i.e. 54% out of 57%) of possible variance in arousal, and ABS explained 64% of it. Second, we examined the correlation between each of the three measures and bipolar valence scores. Replicating previous results (Kron et al., 2013), the subtraction of unpleasant from pleasant (PL−UN) was highly associated with bipolar valence, explaining 88% of its variance (see Table 1 cells no. 20 and 24). Typical for the sample of IAPS space, low negative linear correlations were found between arousal measures (arousal, PL + UN and ABS) and bipolar valence measures (bipolar valence and PL−UN), suggesting that the negative pictorial stimuli in the sample were somewhat stronger than the positive stimuli. Note that the quadratic association between bipolar valence measures (see cell nos. 14, 15, 18 and 23) and arousal measures is estimated in this analysis by the absolute values of bipolar valence.

Table 1.

Graphics and inferential statistics for correlations between arousal measures (arousal, PL + UN and ABS), and bipolar valence measures (bipolar valence and PL−UN). Statistics include Pearson’s correlation coefficients, P-values and Bayes factors

graphic file with name nsaa012fx1.jpg
Open in a new tab

Note. BF10 indicates support for the alternative hypothesis (significant correlation between each two measures) over the null hypothesis (no correlation between each two measures).

fMRI results

Analysis within each ROI

We first examined if each measure is significantly associated with arousal-related activity (i.e. if parameter estimates of each measure are significantly different from zero). All arousal measures (arousal, ABS and PL + UN) were not significantly different from zero in L and R insula, L and R cerebellum, and hence, those regions are not presented here, and they were excluded from the overall analysis. As predicted, each of the three arousal measures (arousal—model 1, ABS—model 2 and PL + UN—model 3) showed a significant difference in most arousal ROIs (but see arousal in left and right thalamus; ABS in L and R extrastriate and PL + UN in R extrastriate) (for detailed results, see Figure 5 and Table 2). Bipolar valence consistently showed no association with activity in arousal-related ROIs (i.e. all parameter estimates of bipolar valence were not significantly different from zero—see Table 3 for Bayes factors in favor of the null hypothesis, i.e. no association between activity in arousal-related ROIs and bipolar valence).

Fig. 5.

Fig. 5

Open in a new tab

Mean averages of contrast estimates that were extracted from each region of interest, for arousal (model 1), ABS (model 2), PL + UN (model 3) and bipolar valence. ***P < 0.001, **P < 0.01, *P < 0.05.

Table 2.

Association between each arousal measure (arousal—model 1, PL + UN—model 3 and ABS—model 2) and activation in arousal ROIs (parameter estimates)

Region XYZ K Arousal BF10 PL + UN BF10 ABS BF10
t-value t-value t-value
R. extrastriate (36–69 10) 245 2.873 11.127 1.442 0.947 1.253 0.735
L. extrastriate (−34–756) 387 3.137 19.156 1.918 1.939 1.585 1.161
L. amygdala (−231–17) 32 2.073 2.505 2.592 6.401 2.738 8.505
R. thalamus (12–176) 521 0.938 0.501 2.107 2.655 2.333 3.949
L. thalamus (−12–185) 356 0.833 0.445 1.797 1.598 2.049 2.405
Midbrain (1–23 -8) 671 2.24 3.348 3.846 90.231 3.919 106.39
Open in a new tab

Note. XYZ are the central MNI coordinates, K = cluster size in voxels, R = right, L = left. BF10 is presented; the greater the BF10 is than 1, the stronger is the support for the alternative hypothesis (i.e. association between self-reports and activity in arousal-related ROIs) over the null hypothesis given the data (no association between self-reports and activity in arousal-related ROIs).

Table 3.

Association between bipolar valence and activation in arousal ROIs (parameter estimates)

Region Bipolar valence BF01
t-value
R extrastriate −0.59 4.116
L extrastriate −0.891 3.367
L amygdala 1.605 1.56
R thalamus 1.522 1.739
L thalamus 1.893 1.03
Midbrain −0.472 4.357
Open in a new tab

Note. The greater the BF01 is than 1, the stronger is the support for the null hypothesis (i.e. no association between bipolar valence and activity in arousal-related ROIs), over the alternative hypothesis (i.e. association between bipolar valence and activity in arousal-related ROIs).

Next, we directly compared each arousal measure with bipolar valence. Paired sample t-tests and Bayesian analyses revealed that arousal (i.e. model 1) parameter estimates were significantly higher than bipolar valence within the L extrastriate visual cortex, t(25) = 2.693, P < 0.01, BF10 = 7.785; R extrastriate visual cortex, t(25) = 2.226, P < 0.05, BF10 = 3.263 and midbrain, t(25) = 1.801, P < 0.05, BF10 = 1.608. There were no significant differences between arousal and bipolar valence within L amygdala, t(25) = 0.319, ns, BF01 = 3.721; L thalamus, t(25) = −0.352, ns, BF01 = 6.174 and R thalamus, t(25) = −0.194, ns, BF01 = 5.55. ABS (i.e. model 2) was significantly higher than bipolar valence within midbrain, t(25) = 3.219, P < 0.01, BF10 = 22.76. There were no significant differences between ABS and bipolar valence within L extrastriate, t(25) = 1.562, ns, BF01 = 0.891; R extrastriate, t(25) = 1.140, ns, BF01 = 571; L thalamus, t(25) = 0.331, ns, BF01 = 3.67; R thalamus, t(25) = 0.640, ns, BF01 = 2.751 and amygdala, t(25) = 0.667, ns, BF01 = 2.678. PL + UN (i.e. model 3) was significantly higher than bipolar valence, within L extrastriate visual cortex, t(25) = 1.882, P < 0.05, BF10 = 1.82 and midbrain, t(25) =3.040, P < 0.01, BF10 = 15.67. There were no significant differences between PL + UN and bipolar valence within L amygdala, t(25) =0.968, ns, BF01 = 1.929; L thalamus, t(25) = 0.335, ns, BF01 = 3.67; R extrastriate, t(25) =1.319, ns, BF01 = 1.248 and R thalamus, t(25) =0.670, ns, BF01 = 2.668.

Finally, we compared the ability of the three measures of arousal (arousal—model 1, ABS—model 2 and PL + UN—model 3) to predict arousal-related activations. Repeated-measures ANOVA revealed that all measures of arousal (arousal, ABS and PL + UN) were not significantly different from each other in their ability to predict neural activity in all arousal-related regions. Bayes factors that support the null hypothesis over the alternative hypothesis given data showed moderate support in favor of the null hypothesis within L amygdala and L thalamus. Within R and L extrastriate, R thalamus and midbrain, the evidence in favor of the null hypothesis is inconclusive (see Table 4 for Bayes factors in favor of the null hypothesis within each ROI).

Table 4.

Comparisons between parameter estimates of the three arousal measures in their ability to predict activation in arousal-related regions

Region BF01 F-value
R extrastriate 0.971 3.007
L extrastriate 1.159 2.738
L amygdala 3.101 1.401
R thalamus 1.746 2.186
L thalamus 2.495 1.634
Midbrain 0.246 4.93
Open in a new tab

Note. BF01 indicates support of the null hypotheses, i.e. no difference between the three arousal measures (Arousal—model 1, ABS—model 2 and PL + UN—model 3) in predicting arousal-related activity, over alternative hypothesis, i.e. significant difference between the three arousal measures.

Overall analyses across all ROIs

We first examined the average association of each measure with arousal-related activity (i.e. if the weighted averages of parameter estimates for each measure are different from zero). All three measures show significant associations; arousal, (mean = 0.011 ± 0.02), t(25) = 2.15, P < 0.05, BF10 = 2.86; absolute values of valence, (mean = 0.014 ± 0.02), t(25) = 2.85, P < 0.01, BF10 = 10.76 and PL + UN (mean = 0.015 ± 0.03), t(25) = 2.59, P < 0.01, BF10 = 6.43. Weighted average of parameter estimates of bipolar valence was not significant, (mean = 0.002 ± 0.01), t(25) = 0.41, ns, BF01 = 4.45.

Next, we compared the prediction ability of each of the arousal measures to the prediction ability of bipolar valence. Paired sample t-tests revealed no significant difference between the prediction ability of arousal and of bipolar valence. ABS and PL + UN prediction abilities were each significantly higher than bipolar valence (see Table 5 for t- and P-values, and Bayes factors). Finally, we compared the prediction abilities of the three arousal measures to each other (i.e. comparison of the weighted averages of parameter estimates of arousal, ABS and PL + UN). Repeated-measures ANOVA revealed no significant differences between the three measures of arousal (arousal—model 1, ABS—model 2 and PL + UN—model 3), in predicting arousal activation in arousal ROIs, F(2,50) = 0.965, ns. Further Bayesian analysis revealed strong support in favor of the null hypothesis over the alternative hypothesis, i.e. no difference between the three arousal measures in their prediction ability, given the data, BF01 = 4.274.

Table 5.

Comparison between the prediction ability of each arousal measure (Arousal—model 1, ABS—model 2 and PL + UN—model 3) and bipolar valence across all ROIs

Mean difference t-value P-value Cohen’s d BF10 BF01
Arousal—BV 0.009 1.26 0.11 0.247 — 1.349
ABS—BV 0.012 1.823 0.04 0.358 1.66 —
(PL + UN)—BV 0.014 1.775 0.044 0.348 1.54 —
Open in a new tab

Note. The comparisons presented here are between the weighted averages of parameter estimates of each arousal measure and the weighted averages of parameter estimates of bipolar valence. For significant differences, the Bayes factor in favor of the alternative hypothesis (i.e. significant difference between bipolar valence and each arousal measure) is presented as (BF10). For non-significant results, the Bayes factor in favor of the null hypothesis (i.e. no difference between bipolar valence and each arousal measure) is presented as (BF01).

Discussion

The current research systematically compared three different models of arousal: model 1—‘arousal as a separate quale from valence’, model 2—‘arousal as intensity of bipolar valence’ and model 3—‘arousal as a linear combination of unipolar pleasant and unpleasant’. While model 1 assumes that arousal is a distinct quale from valence, models 2 and 3 assume that arousal is the intensity of valence. If arousal is a separate quale from valence, then self-reports about arousal (model 1) are expected to show an advantage over the intensity of valence (models 2 and 3) in predicting arousal-related activations. The results show no advantage for self-reports of arousal (model 1) over the intensity of valence (models 2 and 3) in predicting activity in arousal-related regions. A further Bayesian analysis showed moderate to strong support in favor of the null hypothesis, i.e. no difference between arousal and intensity of valence in predicting arousal-related activations. The results suggest that, at least in the context of explaining variance in neural activity, arousal is not separable from valence. Specifically, both measures of valence, bivariate valence (two unipolar pleasant and unpleasant scales; PL + UN) and absolute value of bipolar valence (ABS), did not differ from self-reported arousal in their ability to predict neural activity in arousal-related regions.

These results are consistent with previous data that challenge the idea that self-reported arousal is separable from that of valence. The first evidence is the ‘boomerang’ shape function between bipolar valence and arousal that is repeatedly and consistently found in data pools (Bradley and Lang, 1999; Lang et al., 1999; Bradley and Lang, 2007; Dan-Glauser and Scherer, 2011; Kurdi et al., 2017). The second evidence, which is similar in its logic to the current study, shows that self-reports of arousal demonstrate an advantage over bipolar valence but not over PL + UN in predicting electrodermal activation (Kron et al., 2013). The third evidence is that self-reported PL + UN explains the majority of the variance of self-reported arousal (Kron et al., 2013). The current results replicate previous findings using self-reports and show no advantage for arousal over PL + UN in predicting neural activity in arousal-related brain areas. Yet, as mentioned in the introduction, we do not argue against the importance of physiological arousal (i.e. autonomic or experienced wakefulness). We also do not claim that there is no such experience as arousal that is independent of valence, such as a feeling of arousal when a person walks up the stairs when that is neither an enjoyable or aversive experience. Our findings are relevant to models of self-reported arousal in the context of affective response. Arousal and valence are components of the affective response, but our ability to reflect and report on them as distinct in the moment of affective experience appears to be limited; although we can hear a cello when it is played alone, this does not mean that we will recognize its sound when the whole orchestra is playing.

Both models 2 and 3 represent arousal in terms of intensity of valence; according to model 2, in the valence scale, participants report about an experience that represents the subtraction of PL−UN, and about the absolute value of PL−UN, in the arousal scale. According to model 3, participants experience PL and UN separately and transform it into PL−UN in the bipolar valence scale, and into PL + UN in the arousal scale. The current research found no difference between the two models. A critical test to decide between models 2 and 3 is the case of mixed emotions (Larsen et al., 2001; Itkes et al., 2019). Model 2, in which valence is bipolar, hinders the simultaneous representation of pleasure and displeasure. Accordingly, evidence for mixed feelings will support model 3 and evidence against mixed emotions will support model 2.

In this experiment, IAPS stimuli were chosen as emotion elicitation manipulation. There are several limitations to using these types of emotional pictures. First, the IAPS pool includes only a limited number of content categories that elicit only a limited number of discrete emotions. Second, there is a limit to the degree to which pictures can elicit a strong emotional response (e.g. attending a funeral vs viewing a picture of it). Although the other two emotion stimuli pools (IADS—for audio recording; ANEW—for English words) show the same relationship between bipolar valence and arousal, it is important to support these findings and compare the three arousal measures in other contents of categories and modalities.

The topic discussed in the current study has important implications on emotion research. Deciding between different models of emotions will potentially shape the research question, structure of the design, stimuli selection, data analysis, interpretation of the results and the accumulation of knowledge about emotions. The question about the separability of arousal and valence is critical in deciding between the models. The results of this study suggest no separability (i.e. they support models 2 and 3). Nonetheless, we believe this research question is still an open one. Specifically, in this study, we compared models of conscious experience of affect by contrasting their ability to predict arousal-related neural activity. However, the extent to which the neural framework can account for conscious experience remains an ongoing question (Edelman, 2003). It might be that arousal is separable from valence in explaining complex neural activities that recruit several brain circuits that we could not visualize here, or that similar patterns of neural activation are associated with different qualia. Consequently, more evidence is needed to reach decisive conclusions about the structure of experience of affect from neural data. Until this is accomplished, we recommend the use of two separate unipolar scales: one for pleasant feelings, and one for unpleasant feelings. As shown in this work and previous research (Kron et al., 2013; Kron et al., 2015), it is possible to compute arousal and bipolar valence from unipolar pleasant and unpleasant scores, but not vice versa.

Coda

In this paper, we compared three models of arousal (arousal as a separate quale from valence, arousal as intensity of bipolar valence and arousal as a linear combination of unipolar pleasant and unpleasant valence). Previous models assume dissociability of arousal and valence in conscious experience (e.g. Barrett and Russell, 1999). Evidence accumulated suggests that this assumption might be erroneous or a least unneeded to explain autonomic and neural activity. This research question is fundamental to emotion research—the way a researcher parses affective space, which influences all stages of experimentations from design, to data analysis, and conclusions (Kron, 2019). Yet, we believe this research question is still open and does not necessarily decide upon one model over the others. It is likely that different stages of processing rely on different valence structures (Cacioppo et al., 1999; also see Mattek et al., 2017, for a similar argument) and that people may differ in their underlying valence structure (see Barrett, 1998 and Barrett, 2004, for related example). More data are required to examine these possibilities.

Supplementary Material

scan-19-046-File002_nsaa012
Click here for additional data file. (23.5KB, doc)
scan-19-046-File003_nsaa012
Click here for additional data file. (1.2MB, doc)

Acknowledgements

We would like to thank Dr. Roee Admon and Dr. Avi Mendelsohn for their help with the fMRI analysis, and Ms. Leigh hall for proofreading the manuscript.

Funding

The current work was funded by the United States Israel Binational Science Foundation, Grant Number 2015039.

Conflict of interest

The authors declare no conflict of interests concerning the publication or the authorship of this article.

References

  1. Anderson A.K., Christoff K., Stappen I., et al. (2003). Dissociated neural representations of intensity and valence in human olfaction. Nature Neuroscience, 6(2), 196–202. [DOI] [PubMed] [Google Scholar]
  2. Barrett L.F. (1998). Discrete emotions or dimensions? The role of valence focus and arousal focus. Cognition & Emotion, 12(4), 579–99. [Google Scholar]
  3. Barrett L.F. (2004). Feelings or words? Understanding the content in self-report ratings of experienced emotion. Journal of Personality and Social Psychology, 87(2), 266–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Barrett L.F., Russell J.A. (1999). The structure of current affect: controversies and emerging consensus. Current Directions in Psychological Science, 8(1), 10–4. doi: 10.1111/1467-8721.00003. [DOI] [Google Scholar]
  5. Barrett L.F., Mesquita B., Ochsner K.N., Gross J.J. (2007). The experience of emotion. Annual Review of Psychology, 58, 373–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bradley M.M., Lang P.J. (1994). Measuring emotion: the self-assessment manikin and the semantic differential. Journal of Behavior Therapy and Experimental Psychiatry, 25(1), 49–59. doi: 10.1016/0005-7916(94)90063-9. [DOI] [PubMed] [Google Scholar]
  7. Bradley M.M., Lang P.J. (1999). Affective Norms for English Words (ANEW): Instruction Manual and Affective Ratings . Tech. Rep. No. C–1, Gainesville: Center for Research in Psychophysiology, University of Florida. [Google Scholar]
  8. Bradley M.M., Lang P.J. (2007). The International Affective Digitized Sounds (2nd Edition, IADS–2): Affective Ratings of Sounds and Instruction Manual . Tech. Rep. No. B–3, Gainesville: Center for Research in Psychophysiology, University of Florida. [Google Scholar]
  9. Bradley M.M., Codispoti M., Cuthbert B.N., Lang P.J. (2001). Emotion and motivation I: defensive and appetitive reactions in picture processing. Emotion, 1(3), 276–98. [PubMed] [Google Scholar]
  10. Brett M., Anton J.L., Valabregue R., Poline J.B. (2002). Region of interest analysis using an SPM toolbox [abstract] presented at the 8th international conference on functional mapping of the human brain. NeuroImage, 16, 1140–1. [Google Scholar]
  11. Cacioppo J.T., Berntson G.G. (1994). Relationship between attitudes and evaluative space: a critical review, with emphasis on the separability of positive and negative substrates. Psychological Bulletin, 115, 401–23. [Google Scholar]
  12. Cacioppo J.T., Gardner W.L., Berntson G.G. (1999). The affect system has parallel and integrative processing components: form follows function. Journal of Personality and Social Psychology, 76(5), 839–55. doi: 10.1037/0022-3514.76.5.839. [DOI] [Google Scholar]
  13. Cahill L., Babinsky R., Markowitsch H.J., McGaugh J.L. (1995). The amygdala and emotional memory. Nature, 377(6547), 295–6. doi: 10.1038/377295a0. [DOI] [PubMed] [Google Scholar]
  14. Critchley H.D. (2009). Psychophysiology of neural, cognitive and affective integration: fMRI and autonomic indicants. International Journal of Psychophysiology, 73(2), 88–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Critchley H.D., Corfield D.R., Chandler M.P., Mathias C.J., Dolan R.J. (2000). Cerebral correlates of autonomic cardiovascular arousal: a functional neuroimaging investigation in humans. The Journal of Physiology, 523(Pt 1), 259–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Critchley H.D., Melmed R.N., Featherstone E., Mathias C.J., Dolan R.J. (2002). Volitional control of autonomic arousal: a functional magnetic resonance study. NeuroImage, 16, 909–19. [DOI] [PubMed] [Google Scholar]
  17. Cunningham W.A., Raye C.L., Johnson M.K. (2004). Implicit and explicit evaluation: fMRI correlates of valence, emotional intensity, and control in the processing of attitudes. Journal of Cognitive Neuroscience, 16(10), 1717–29. [DOI] [PubMed] [Google Scholar]
  18. Dan-Glauser E.S., Scherer K.R. (2011). The Geneva affective picture database (GAPED): a new 730-picture database focusing on valence and normative significance. Behavior Research Methods, 43, 468–77. [DOI] [PubMed] [Google Scholar]
  19. Dienes Z. (2014). Using Bayes to get the most out of non-significant results. Frontiers in Psychology, 5, 781. doi: 10.3389/fpsyg.2014.00781. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Edelman G.M. (2003). Naturalizing consciousness: a theoretical framework. Proceedings of the National Academy of Sciences of the United States of America, 100(9), 5520–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Friston K.J., Holmes A.P., Worsley K.J., Poline J.-P., Frith C.D., Frackowiak R.S.J. (1995). Statistical parametric maps in functional imaging: a general linear approach. Human Brain Mapping, 2(4), 189–210. doi: 10.1002/hbm.460020402. [DOI] [Google Scholar]
  22. Furmark T., Fischer H., Wik G., Larsson M., Fredrikson M. (1997). The amygdala and individual differences in human fear conditioning. Neuroreport: An International Journal for the Rapid Communication of Research in Neuroscience, 8(18), 3957–60. doi: 10.1097/00001756-199712220-00021. [DOI] [PubMed] [Google Scholar]
  23. Hamzani O., Mazar T., Petranker R., Itkes O., Kron A. (2019). Semantic and affective representations of valence: prediction of autonomic and facial responses from feelings- focused and knowledge-focused self-reports. Emotion, in press. [DOI] [PubMed] [Google Scholar]
  24. Itkes O., Eviatar Z., Kron A. (2019). Semantic and affective manifestations of ambi(valence). Cognition and Emotion, 1–14. [DOI] [PubMed] [Google Scholar]
  25. Ito T.A., Cacioppo J.T., Lang P.J. (1998). Eliciting affect using the international affective picture system: trajectories through evaluative space. Personality and Social Psychology Bulletin, 24, 855–79. [Google Scholar]
  26. JASP Team (2018). JASP (Version 0.8.6.0) [Computer software]. Available: https://jasp-stats.org/.
  27. Kriegeskorte N., Simmons W.K., Bellgowan P.S., Baker C.I. (2009). Circular analysis in systems neuroscience: the dangers of double dipping. Nature Neuroscience, 12, 535–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kron A. (2019). Rethinking the principles of emotion taxonomy. Emotion Review, 1–8. [Google Scholar]
  29. Kron A., Goldstein A., Lee D.H.J., Gardhouse K., Anderson A.K. (2013). How are you feeling? Revisiting the quantification of emotional qualia. Psychological Science, 24(8), 1503–11. [DOI] [PubMed] [Google Scholar]
  30. Kron A., Pilkiw M., Banaei J., Goldstein A., Anderson A.K. (2015). Are valence and arousal separable in emotional experience? Emotion, 15(1), 35–44. [DOI] [PubMed] [Google Scholar]
  31. Kuppens P., Tuerlinckx F., Russell J.A., Barrett L.F. (2013). The relation between valence and arousal in subjective experience. Psychological Bulletin, 139(4), 917–40. doi: 10.1037/a0030811. [DOI] [PubMed] [Google Scholar]
  32. Kurdi B., Lozano S., Banaji M.R. (2017). Introducing the open affective standardized image set (OASIS). Behavior Research Methods, 49(2), 457–70. [DOI] [PubMed] [Google Scholar]
  33. Lane R.D., Chua P., Dolan R.J. (1999). Common effects of emotional valence, arousal and attention on neural activation during visual processing of pictures. Neuropsychologia, 37, 989–97. [DOI] [PubMed] [Google Scholar]
  34. Lang P.J., Greenwald M.K., Bradley M.M., Hamm A.O. (1993). Looking at pictures: affective, facial, visceral and behavioral reactions. Psychophysiology, 30(3), 261–73. doi: 10.1111/j.1469-8986.1993.tb03352.x. [DOI] [PubMed] [Google Scholar]
  35. Lang P.J., Bradley M.M., Cuthbert B.N. (1997). International Affective Picture System, Gainesville: Center for the Study of Emotion and Attention, University of Florida. [Google Scholar]
  36. Lang P.J., Bradley M.M., Fitzsimmons J.R., et al. (1998). Emotional arousal and activation of the visual cortex: an fMRI analysis. Psychophysiology, 35(2), 199–210. [PubMed] [Google Scholar]
  37. Lang P.J., Bradley M.M., Cuthbert B.N. (1999). International Affective Picture System (IAPS): Technical Manual and Affective Ratings, Gainesville: Center for the Study of Emotion and Attention, University of Florida. [Google Scholar]
  38. Larsen J.T., McGraw A.P., Cacioppo J.T. (2001). Can people feel happy and sad at the same time? Journal of Personality and Social Psychology, 81(4), 684–96. [PubMed] [Google Scholar]
  39. Maldjian J.A., Laurienti P.J., Kraft R.A., Burdette J.H. (2003). An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. NeuroImage, 19(3), 1233–9. doi: 10.1016/S1053-8119(03)00169-1. [DOI] [PubMed] [Google Scholar]
  40. Mather M. (in press). How arousal-related neurotransmitter systems compensate for age-related decline In: Gutchess A., Thomas A., editors. The Cambridge Handbook of Cognitive Aging: A Life Course Perspective, Cambridge: University Press. [Google Scholar]
  41. MATLAB (2014). The MathWorks, Natick, Vol. 2014. [Google Scholar]
  42. Mattek A., Wolford G.L., Whalen P.J. (2017). A mathematical model captures the structure of subjective affect. Perspectives on Psychological Science, 12(3), 508–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Nagai Y., Critchley H.D., Featherstone E., Trimble M.R., Dolan R.J. (2004). Activity in ventromedial prefrontal cortex covaries with sympathetic skin conductance level: a physiological account of a “default mode” of brain function. NeuroImage, 22, 243–51. [DOI] [PubMed] [Google Scholar]
  44. Paus T., et al. (1997). Time-related changes in neural systems underlying attention and arousal during the performance of an auditory vigilance task. Journal of Cognitive Neuroscience, 9(3), 392–408. [DOI] [PubMed] [Google Scholar]
  45. Rainnie D.G., Grunze H.C., McCarley R.W., Greene R.W. (1994). Adenosine inhibition of mesopontine cholinergic neurons: implications for EEG arousal. Science (New York, N.Y.), 263(5147), 689–92. doi: 10.1126/science.8303279. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Reyes B.A., Carvalho A.F., Vakharia K., Bockstaele E.J. (2011). Amygdalar peptidergic circuits regulating noradrenergic locus coeruleus neurons: linking limbic and arousal centers. Experimental Neurology, 230(1), 96–105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Russell J.A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110, 145–72. [DOI] [PubMed] [Google Scholar]
  48. Russell J.A., Barrett L.F. (1999). Core affect, prototypical emotional episodes, and other things called emotion: dissecting the elephant. Journal of Personality and Social Psychology, 76(5), 805. [DOI] [PubMed] [Google Scholar]
  49. Safron A., Barch B., Bailey J.M., Gitelman D.R., Parrish T.B., Reber P.J. (2007). Neural correlates of sexual arousal in homosexual and heterosexual men. Behavioral Neuroscience, 121, 237–48. [DOI] [PubMed] [Google Scholar]
  50. Sakurai T. (2007). The neural circuit of orexin (hypocretin): maintaining sleep and wakefulness. Nature Reviews Neuroscience, 8, 171–81. [DOI] [PubMed] [Google Scholar]
  51. Satpute A.B., Kragel P.A., Barrett L.F., Wager T.D., Bianciardi M. (2018). Deconstructing arousal into wakeful, autonomic and affective varieties. Neuroscience Letters, 693, 19–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Schiff N.D. (2008). Central thalamic contributions to arousal regulation and neurological disorders of consciousness. Annals of the New York Academy of Sciences, 1129, 105–18. [DOI] [PubMed] [Google Scholar]
  53. Schneider W., Eschman A., Zuccolotto A. (2002). E-Prime Reference Guide, Pittsburgh, PA: Psychology Software Tools. [Google Scholar]
  54. Todd R.M., Talmi D., Schmitz T.W., Susskind J., Anderson A.K. (2012). Psychophysical and neural evidence for emotion-enhanced perceptual vividness. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience, 32(33), 11201–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Tzourio-Mazoyer N., Landeau B., Papathanassiou D., Crivello F., Joliot M. (2002). Automated anatomical Labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage, 15, 273–89. [DOI] [PubMed] [Google Scholar]
  56. Van der Werf Y.D., Witter M.P., Groenewegen H.J. (2002). The intralaminar and midline nuclei of the thalamus. Anatomical and functional evidence for participation in processes of arousal and awareness. Brain Research Reviews, 39(2–3), 107–40. [DOI] [PubMed] [Google Scholar]
  57. Williams L.M., Phillips M.L., Brammer M.J., Skerrett D., Gordon E. (2001). Arousal dissociates amygdala and hippocampal fear responses: evidence from simultaneous fMRI and skin conductance recording. NeuroImage, 14, 1070–9. [DOI] [PubMed] [Google Scholar]
  58. Yik M.S.M., Russell J.A., Barrett L.F. (1999). Structure of self-reported current affect: integration and beyond. Journal of Personality and Social Psychology, 77(3), 600–19. doi: 10.1037/0022-3514.77.3.600. [DOI] [Google Scholar]
  59. Zhang S., Hu S., Chao H.H., et al. (2014). Ventromedial prefrontal cortex and the regulation of physiological arousal. Social Cognitive and Affective Neuroscience, 9(7), 900–8. doi: 10.1093/scan/nst064 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

scan-19-046-File002_nsaa012
Click here for additional data file. (23.5KB, doc)
scan-19-046-File003_nsaa012
Click here for additional data file. (1.2MB, doc)

Articles from Social Cognitive and Affective Neuroscience are provided here courtesy of Oxford University Press

RESOURCES