Autonomic specificity of discrete emotion and dimensions of affective space: a multivariate approach

https://doi.org/10.1016/j.ijpsycho.2003.08.002Get rights and content

Abstract

The present study addressed autonomic nervous system (ANS) patterning during experimentally manipulated emotion. Film clips previously shown to induce amusement, anger, contentment, disgust, fear and sadness, in addition to a neutral control film, were presented to 34 college-aged subjects while skin conductance, blood pressure and the electrocardiogram (ECG) were recorded, as was self-reported affect. Both mean of and mean successive difference of heart period were derived from the ECG. Pattern classification analyses revealed emotion-specific autonomic patterning for all emotion conditions except disgust; all emotion conditions exhibited significant patterning using self-report. Discriminant function analysis was used to describe the location of discrete emotions within dimensional affective space using both self-report and ANS variables. Findings suggest that the dimensions of valence and activation portray the structure of self-reported emotion, but that valence is more accurately described as approach–withdrawal when applied to autonomic responses during discrete emotions. The findings provide further support for the existence of emotion-specific ANS activity, and are consistent with a hybrid discrete–dimensional model of affective space.

Introduction

The issue of emotion-specific autonomic nervous system (ANS) activity is arguably one of the most enduring research topics in psychology. This essential tenet of James’ (1884) venerable model of emotion processing holds that basic human emotions have distinct ANS patterns. Although substantial empirical support has accrued for the autonomic discriminability of at least some emotions (e.g. Ekman et al., 1983, Levenson et al., 1990, Levenson et al., 1992, Sinha et al., 1992, Stemmler, 1989, Witvliet and Vrana, 1995; see Levenson, 1992, for review), there remains a lack of consensus on the convergence of these results in support of emotion-specific ANS patterns (cf. Cacioppo et al., 1993).

The present study was conducted to address a number of prominent theoretical and methodological issues in this considerable body of literature. First, studies of ANS specificity are either implicitly or explicitly based on a particular structural model of affect; this issue is explored in detail in the present investigation. Second, the nature of ANS patterns may vary as a function of the emotion-induction technique and dependent autonomic measures selected; the present study employs a particular method (films) to manipulate affect in conjunction with a broad array of ANS measures. Finally, most research on ANS specificity has been based on univariate analyses; multivariate techniques were used in the current study to detect autonomic patterns. These issues are discussed below, followed by a description of the present study.

The selection of induced emotions is a crucial design feature of ANS specificity studies, and should be guided by a clearly specified theoretical model of affective space. These models generally fall into one of two categories: discrete or dimensional. Discrete models focus on universal set of ‘primary’ emotions (e.g. fear, anger or disgust), which are typically seen in terms of their evolutionary adaptive value (Plutchik, 1980). This functional view is consistent with James’ (1884) characterization of ‘standard’ emotions, and portrays emotions in terms of their ability to coordinate multiple behavioral and physiological responses to produce an appropriate response to environmental demand (Levenson, 1988). By contrast, dimensional models describe affective space with a limited number of underlying dimensions.

A prevalent dimensional model is the circumplex, comprising a 2D circular array of affect descriptors (Larsen and Diener, 1992, Russell, 1980) Two dimensions of the affective circumplex have consistently emerged: (1) valence or hedonic tone (positivity or negativity) and (2) arousal or activation (energy level). An alternative dimensional model posits orthogonal axes labeled as positive affect and negative affect, rotated 45° relative to the valence and activation axes (Watson and Tellegen, 1985). In contrast to the circumplex, this model views activation and valence as inseparable, and depicts valence as independent positive and negative factors rather than as anchors of a single bipolar dimension. Both models have accrued considerable empirical support, and their relative merits have been actively debated in the literature (e.g. Feldman Barrett and Russell, 1998, Russell and Carroll, 1999, Watson et al., 1999).

Although discrete and dimensional models are often presented as mutually exclusive, the most fruitful approach for ANS specificity research may be a hybrid of the two (Levenson, 1988). This approach was employed in a study that illustrated a hierarchical relationship between lower-order discrete emotions and higher-order emotion dimensions (Nyklicek et al., 1997). In this view, emotions of the discrete model represent unique points in dimensional affective space.

The choice of affect induction method is also critical in ANS specificity research. Diverse forms include: (a) ‘real-life’ inductions (e.g. Stemmler, 1989); (b) reading affective statements (e.g. Velten, 1968) or scenarios (e.g. Witvliet and Vrana, 1995); (c) directed facial expressions (e.g. Ekman et al., 1983); (d) imagery (e.g. Lang, 1979); (e) music (e.g. Nyklicek et al., 1997); (f) slides (Lang et al., 1988); and (g) films (Gross and Levenson, 1995). All of these methods have their relative merits, and the types of induction used will vary with theoretical and practical concerns. One such factor is the match of the manipulation with the underlying structural model of affect upon which it is based. For example, the facial action task is intended to induce discrete facial expressions of emotions such as anger or fear (Ekman et al., 1983). By comparison, the targets of dimensionally based affect manipulations (e.g. Velten, 1968) are more diffuse affective states (i.e. quadrants of the circumplex). Thus, the resolution of dimensionally based affect manipulations is theoretically coarser than those based upon discrete models, and it is necessary when addressing some aspect of discrete emotions, as is the case in the present study, to employ a manipulation that is congruent with the latter type. For this reason, films selected on discrete emotion criteria were used in this study.

Furthermore, films were chosen due to their distinct practical advantages: (a) ease of standardization, (b) lack of need for deception and (c) ecological validity (Gross and Levenson, 1995). Finally, because different stimulus situations may evoke dissimilar patterns of somatic responding (Lacey, 1967), it can be difficult to distinguish the physiological effects of emotion from those due to the experimental context of the induction (Stemmler, 1992b, Stemmler et al., 2001). Hence, a relatively neutral ‘context without emotion’ condition was included as a control for these potentially independent sources of variability.

Another relevant issue is the collection of a representative montage of ANS variables. Breadth can be obtained with measures that reflect varying autonomic inputs (Stemmler, 1992a). To achieve this aim, the following measures were employed in the present study: (a) blood pressure (BP), which reflects vascular α-adrenergic (sympathetic) activity as well as cardiac vagal and β-adrenergic input (Smith and Kampine, 1984); (b) skin conductance (SC), which provides a relatively pure sympathetic index that varies with affective arousal (Dawson et al., 2000); (c) heart period (HP), reflecting mixed vagal and β-adrenergic input (Smith and Kampine, 1984) and (d) mean successive differences (MSD) of HP variability, a widely used measure of cardiac vagal control (Friedman et al., 2002).

Finally, to detect coordination of multiple response systems associated with ANS patterning, analytic techniques sensitive to variable configurations must be applied. Pattern classification analysis is ideal for this research question because it affords simultaneous consideration of multiple response variables (Fridlund et al., 1984). In this technique, classification functions are generated for each of several output classes (e.g. emotions). Cases are then ‘blindly’ assigned to classes based on a vector of input elements (e.g. ANS responses). The success of the classification functions, and thus the discriminability of the classes based upon respective input elements, is evaluated by testing the percentage of correct classifications against chance level using a standardized normal test statistic (Huberty, 1994).

A secondary aim of this study was to place discrete emotions, with respect to both ANS and self-report variables, in dimensional affective space. To this end, discriminant analysis, which yields a smaller number of discriminant rather than classification functions, was used to generate dimensions (i.e. discriminant functions) that maximally separate groups in multivariate space. Once these functions are identified, group centroids (multivariate group means) can be plotted, thus describing the location of the discrete emotions in dimensional space and facilitating interpretation (i.e. labeling) of the discriminant functions.

Section snippets

Goals of the present study

A recent study based upon the hybrid discrete–dimensional model of affect addressed a number of the above theoretical and methodological issues (Nyklicek et al., 1997). Key elements of the study include a dimensionally based affect manipulation (music), a broad set of autonomic indices, and multivariate pattern classification analysis. Results yielded ANS patterns that reliably distinguished among four affect conditions. Moreover, both self-report and ANS activity produced a 2D structure

Subjects

Thirty-four subjects (16 males and 18 females) were recruited via fliers and received their choice of either extra credit or $10 for participation in an experimental session lasting approximately 1.5 h. Subjects were screened on the basis of age (18–26 years old; M: 18.7 years; S.D.: 0.88) and health status as determined by questionnaire. Those indicating a history of neurological deficits or cardiovascular disorder, or taking medication for hypertension, depression or anxiety were excluded.

In

Discussion

The presence of unique ANS patterns during emotional states, the primary hypothesis of the study, was supported by statistically significant classification using ANS variables. That disgust failed to show significance was likely due to one of two reasons. First, though a diverse montage of ANS variables was used, none specifically indexed gastric activity. Thus, autonomic changes unique to disgust were likely not captured. Second, data suggest that physiological responses to potent disgust film

Acknowledgements

This research was supported by a grant to the second author from the College of Arts and Sciences at Virginia Tech. Portions of these data were presented at the annual meetings of the Society for Psychophysiological Research (October, 2002) and the Society for Behavioral Medicine (April, 2002). The authors would like to thank James B. Weaver III for his technical support, and Lauren Eadie for her assistance in data collection.

References (41)

  • R.J. Davidson

    Anterior cerebral asymmetry and the nature of emotion

    Brain Cognition

    (1992)
  • R. Plutchik

    A general psychoevolutionary theory of emotion

  • E. Velten

    A laboratory task for induction of mood states

    Behav. Res. Ther.

    (1968)
  • R.M. Bagby et al.

    The twenty-item Toronto Alexithymia Scale—I: item selection and cross-validation of the factor structure

    J. Psychosom. Res.

    (1992)
  • R.M. Bagby et al.

    The twenty-item Toronto Alexithymia Scale—II: convergent, discriminant, and concurrent validity

    J. Psychosom. Res.

    (1992)
  • A.T. Beck et al.

    Manual for the Beck Depression Inventory-II

    (1996)
  • K.G. Belani et al.

    Accuracy of radial artery blood pressure determination with the Vasotrac

    Can. J. Anaesth.

    (1999)
  • J.T. Cacioppo et al.

    The psychophysiology of emotion

  • M.E. Dawson et al.

    The electrodermal system

  • P. Ekman et al.

    Autonomic nervous system activity distinguishes among emotions

    Science

    (1983)
  • L. Feldman Barrett et al.

    Independence and bipolarity in the structure of current affect

    J. Pers. Soc. Psychol.

    (1998)
  • A.J. Fridlund et al.

    Pattern recognition of self-reported emotional state from multiple-site facial EMG activity during affective imagery

    Psychophysiology

    (1984)
  • B.H. Friedman et al.

    Quantification issues in time- and frequency-domain measures of heart rate variability

    IEEE Eng. Med. Biol.

    (2002)
  • Friedman, B.H., Christie, I.C., Sargent, S.L, Weaver, J., Self-reported sensitivity to continuous non-invasive blood...
  • J.J. Gross et al.

    Emotion elicitation using films

    Cognition Emotion

    (1995)
  • E. Harmon-Jones et al.

    Anger and frontal brain activity: EEG asymmetry consistent with approach motivation despite negative affective valence

    J. Pers. Soc. Psychol.

    (1998)
  • C.J. Huberty

    Applied Discriminant Analysis

    (1994)
  • W. James

    What is an emotion?

    Mind

    (1884)
  • R.A. Johnson et al.

    Applied Multivariate Statistical Analysis

    (1992)
  • Kallenberg, J., Pennings, M., Vingerhoets, A.J., 2001. Blood pressure and emotional reactions of medical and...
  • Cited by (287)

    • Assessing the existence of a semantic space to express action tendencies

      2022, International Journal of Industrial Ergonomics
    • A CONCEPTUALIZATION OF A JOYFUL LEADERSHIP STYLE

      2024, Research on Emotion in Organizations
    View all citing articles on Scopus
    View full text