Evidence for bivariate systems: An empirical test of appetition and aversion across domains

Research in diverse domains of psychology has independently identified two behavioral systems, one concerned with obtaining positive outcomes, the other concerned with avoiding negative outcomes. This basic distinction, described in different domains of inquiry with varying terminology, may be integrated within a single appetitive–aversive systems model. The present research was designed to examine the viability of the appetitive–aversive distinction as an organizational construct underlying various particular measures and concepts. In four studies, individual difference measures from different domains were examined with exploratory (Study 1) and confirmatory (Studies 2–4) factor analyses. We expected and found that measures tapping sensitivity to rewards or positive outcomes would load on a common appetitive latent factor, whereas, measures tapping individual differences in sensitivity to punishment or negative outcomes would load a common aversive latent factor. Results strongly supported the hypothesized two-factor structure over alternative models and indicated that the latent appetitive and aversive variables accounted for about half the variance in the observed variables. 2003 Elsevier Science (USA). All rights reserved. 1. Appetition and aversion across domains: An empirical test Over the past few decades, researchers from seemingly diverse areas of psychology have recognized, either explicitly or implicitly, the existence of two distinct systems in human behavior, one appetitive and the other aversive (e.g., Cacioppo & Gardner, * Corresponding author. Fax: 1-310-206-5895. E-mail address: [email protected] (S.L. Gable). 0092-6566/$ see front matter 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S0092-6566(02)00580-9 350 S.L. Gable et al. / Journal of Research in Personality 37 (2003) 349–372 1999; Carver, 1996; Fowles, 1994; Gray, 1990; Higgins, 1998; Lang, 1990; Miller, 1959; Schneirla, 1959). The appetitive system regulates responses to potentially rewarding stimuli and thereby serves as an approach-oriented system, whereas, the aversive system regulates responses to potentially punishing stimuli and serves as an avoidance-oriented system. Although these two behavioral systems have been given different names depending on the research programs in which they have emerged (e.g., approach/avoidance, positive/negative, discrepancy-reducing/discrepancy-enlarging, activation/inhibition), the basic appetitive/aversive distinction can be found in many research domains, such as motivation, attitude evaluation, personality, emotion, mood, and coping (e.g., Cacioppo, Gardner, & Berntson, 1997; Elliot & Church, 1997; Eysenck, 1967/1981; Moos, 1997; Watson, Clark, & Tellegen, 1988). The emergence of a seemingly comparable appetitive–aversive distinction across discrete processes has not gone unnoticed, as evidenced by recent attempts to integrate theories and findings from independent research programs (e.g., Carver, 1996; Carver, Sutton, & Scheier, 2000). For example, Carver et al. s (2000) broad review of several literatures concluded that approach behaviors and positive affect are managed by one regulatory system, whereas, avoidance behaviors and negative affect are managed by a separate regulatory system. Other authors, citing the rich history of the approach–avoidance distinction not only in humans but also in other species, have asserted that the distinction between appetition and aversion may be fundamental and innate (e.g., Elliot & Covington, 2001), a conclusion also supported by neurophysiological studies (e.g., Harmon-Jones & Allen, 1997; Sutton & Davidson, 1997). The present research is an empirical attempt to structurally integrate findings from several different domains of psychological inquiry within an appetitive–aversive systems model. That is, we sought to determine whether the similar dimensions that have emerged in different research domains might represent a meaningful commonality, suggesting the possibility of an underlying structural association, or whether it is simply an interesting but circumstantial parallelism. We have adopted a confirmatory factor analysis approach to examining the latent structure of various measures that, although tapping discrete constructs, may reflect a common distinction. Thus, the present work was intended to serve as a quantitative complement to more conceptually based integrations of the evidence concerning appetition and aversion. 1.1. The appetitive–aversive distinction in various research domains Although interest has waxed and waned over the years, appetition and aversion have a long history in psychology (see Elliot, 1999; Higgins, 1998, for reviews). James (1890) wrote about pleasure as a reinforcer and pain as an inhibitor of behavior, and Pavlov (1927) described two types of reflexes, one orienting toward the stimulus and the other turning away from the stimulus, a description literally adopted as a manipulation by later researchers (e.g., Cacioppo, Priester, & Berntson, 1993). Schneirla (1959) documented the communality across diverse species of ‘‘towardness’’ (approach) and ‘‘awayness’’ (withdrawal), and suggested that these processes may have had evolutionary roots. Similarly, Miller s (1959, 1961) classic research on separate approach (appetitive) and withdrawal (aversive) learning processes proS.L. Gable et al. / Journal of Research in Personality 37 (2003) 349–372 351 duced graphical representations of their interplay in approach and avoidance conflict gradients. More recent research has seen the appearance of the appetitive–aversive distinction in domain-specific conceptualizations, examples of which we chronicle below. We note here that we based domain classifications (e.g., personality, motivation) on their most typical representation in the literature, and were employed for parsimony and are not intended as rigid classifications. 1.1.1. Affect Studies of mood and affect have garnered considerable evidence that activated positive affect (PA) is conceptually and statistically distinct from activated negative affect (NA; Watson et al., 1988; Watson & Tellegen, 1985). High PA is characterized by elation, excitement, and enthusiasm, whereas, low PA is characterized by feeling dull and sluggish. Adjectives such as distressed, fearful, and hostile describe high NA, whereas, calm, placid, and relaxed represent low NA. Although the exact magnitude of the correlation between positive and negative affect has been debated (Feldman Barrett & Russell, 1998; Green, Goldman, & Salovey, 1993; Watson, Wiese, Vaidya, & Tellegen, 1999), the typical modest correlation between PA and NA (e.g., r 1⁄4 :28; Diener, Smith, & Fujita, 1995) tends to be somewhat higher when measurement error is accounted for (e.g., r 1⁄4 :44; Diener et al., 1995). Thus, although positive and negative affect are not entirely orthogonal, their covariation excludes the largest portion of each variable s variance, leaving considerable room for independence. Consistent with this idea, Watson and his colleagues (1999) have argued that PA is the subjective feeling experienced when a biologically based appetitive or approach system is activated, whereas, NA is the subjective feeling experienced when a biologically based aversive or avoidance system is activated. 1.1.2. Personality In Eysenck s (1967/1981) model of personality structure, extraversion and neuroticism, two independent dimensions of personality, may be interpreted to represent appetitive and aversive functions, respectively. That is, individual differences in extraversion represent differing sensitivities of an appetitive system, whereas, individual differences in neuroticism reflect differing sensitivities of an aversive system. Although extraversion is often described in terms of sociability, recent research strongly suggests that reward sensitivity is at the core of this trait (Lucas, Diener, Grob, Suh, & Shao, 2000). Similarly, neuroticism may reflect the strength of an individual s emotional response to negative stimuli (Watson & Pennebaker, 1989); for example, people high on neuroticism react more strongly to a laboratory induction of negative mood than do people low on neuroticism (Larsen & Ketelaar, 1991). Similar interpretations may be applied to the independent neuroticism and extraversion dimensions of the ‘‘Big Five’’ theory of personality (McCrae & Costa, 1987). 1.1.3. Cognitive evaluation Evidence collected by Cacioppo and colleagues (e.g., Cacioppo & Gardner, 1993; Cacioppo et al., 1997) indicates that the positive and negative aspects of an attitude object are assessed via independent mechanisms. Attitude measurement traditionally 352 S.L. Gable et al. / Journal of Research in Personality 37 (2003) 349–372 uses a single bipolar dimension (i.e., negative to positive). Cacioppo et al. (1997) suggest that a bivariate approach, in which the positive attributes of an object are evaluated by a mechanism functionally and stochastically separate from the mechanism that evaluates its negative attributes, may be more accurate. The output of these distinct systems is then integrated to yield an overall attitude. Among other advantages, this model helps differentiate attitudinal indifference from ambivalence. 1.1.4. Motivation Several theories of motivation and behavioral self-regulation converge on the appetitive/aversive distinction. For example, in Carver and Scheier s (1990) model of self-regulation, some feedback processes attempt to reduce the discrepancy between informational input from the environment and the individual s internal reference (called discrepancy-reducing), whereas, other feedback processes attempt to enlarge this discrepancy (called discrepancy-enlarging). Carver (1996) has equated these two systems with approach and avoidance processes, respectively. Elliot has distinguished between approach and avoidance goals, examining these forms of self-regulation in terms of domain-general personal strivings (Elliot & Sheldon, 1998) and domain-specific achievement goals (Elliot & Church, 1997). Approach and avoidance goals have been linked to distinct antecedents and consequences (Elliot, 1999). Higgins s (1998) theory of regulatory focus also distinguishes between two independent forms of self-regulation, one focused on the promotion (attainment) of positive end-states, the other focused on the prevention of negative end-states. Gray s (1987) neurobiologically based theory of motivation (see also Fowles, 1994) posits distinct appetitive and aversive motivation systems, referred to as the Behavioral Activation System (BAS) and the Behavioral Inhibition System (BIS), respectively. Gray s model, which is based on extensive animal research, supports the independence of these systems at the level of neurobiological mechanisms. (For example, Gray (1990) and LeDoux (1995) have suggested that the amygdala is involved in fear and withdrawal, whereas, the nucleus accumbens is active in approach behaviors.) BAS (the appetitive system) activates behavior in response to signals of reward and nonpunishment, whereas, BIS (the aversive system) inhibits behavior in response to signals of punishment, nonreward, and novelty. Gray s (1990, 1994) theory also links behavioral self-regulation to emotion: BAS activation is associated with hope and approach behaviors, whereas, BIS activation is linked to anxiety and avoidance behaviors. 1.1.5. Neurophysiology Building on the work of Gray and others, several researchers have suggested that separate appetitive and aversive dimensions may have emerged in many areas of research because they express two underlying neurobiological systems in the brain (e.g., Davidson, 1992). Carver (1996) noted that ‘‘Although these two tendencies [approach and avoidance] are often layered across each other in the topography of behavior, they are conceptually distinct from each other. Being distinct, they may be managed by different structures in the nervous system.’’ (p. 320). S.L. Gable et al. / Journal of Research in Personality 37 (2003) 349–372 353 Recent neurophysiological research has supported this proposition. For example, Sutton and Davidson (1997) found that Gray s BIS and BAS constructs were associated with different components of resting prefrontal asymmetry as measured with electroencephalographic (EEG) technology (see also Harmon-Jones & Allen, 1997). Subjects with higher BAS showed greater relative left prefrontal activation, whereas, those with higher BIS scores showed greater relative right prefrontal activation. Similarly, laboratory-induced motivational experiences may correspond to neural activity changes. In one study, reward contingencies were associated with left frontal activation, whereas, punishment contingencies were associated with right frontal activation (Sobotka, Davidson, & Senulis, 1992). The existence of two neurobiologically based functional systems may help explain why conceptually similar appetitive and aversive distinctions are represented in diverse phenomena. 1.2. Putting it together: Basic, functionally independent systems In each of the research areas reviewed above, different starting points and diverse domains of inquiry support the same general distinction between two processes, one concerned with appetition and the other concerned with aversion. Of course, these observed two-factor structures differ among themselves in important ways and their seeming likeness may be nothing more than an interesting coincidence. However, it is also possible that their resemblance reflects a common underlying structure, one whose behavioral expression includes both commonalities and unique properties. The goal of the present research was to determine whether a common structure might underlie individual differences across several conceptual domains. Support for a two-factor appetitive–aversive structure would suggest that these two systems may represent a fundamental organization that influences more specific cognitive, affective, and behavioral processes. If such support is found, these more specific processes would be seen to possess important, heretofore not well-understood underlying commonalities with each other. Moreover, a common structure of separate appetitive and aversive systems would imply that, across diverse specific manifestations: (1) appetitive and aversive systems tend to be activated by different environmental stimuli; (2) appetitive and aversive processes may operate by different processes which are not merely mirror images of each other; and (3) appetitive and aversive processes are likely to be associated with different outcomes. We recently published evidence in support of these three implications (Gable, Reis, & Elliot, 2000). In three studies of motivational predispositions and reactions to daily events, we found that the occurrence of daily negative events was strongly associated with increased NA, whereas, the occurrence of daily positive events was associated with increased PA. Furthermore, high BIS sensitivity (representing the aversive system) was associated with higher average negative affect, whereas, high BAS sensitivity (representing the appetitive system) was predictive of increased positive affect. Most importantly, the relationship between BIS– BAS and affect was characterized by different processes. The BAS–PA relationship 354 S.L. Gable et al. / Journal of Research in Personality 37 (2003) 349–372 was described by differential exposure—high BAS individuals were more likely to seek out and initiate positive-affect producing events. On the other hand, the BIS– NA relationship was better explained by differential sensitivity—the higher an individual s BIS score, the greater their affective response to negative events. Prior research has to some extent supported the proposed organization of various affective, personality, motivational, and attitudinal measures along the lines of appetite and aversion. For example, Lucas and Fujita (2000) found consistent connections between extraversion and positive affect, regardless of the method by which either was assessed. Jorm et al. (1999) obtained support for a two-factor distinction that differentiated BIS-negative affectivity from BAS-positive affectivity; Wilson and Gullon (1999) showed correlations between extraversion and PA, and neuroticism and NA accros the life span; and Carver and White (1994) found relationships between BIS and negative temperament and BAS and positive temperament. The present research goes beyond prior work by extending the range of variables to be examined, both in terms of the phenomena to be considered and the methods by which they are assessed. Such extensions are essential to make a compelling case for appetition and aversion as fundamental systems underlying a broad and diverse set of psychological domains. We refer to these systems as functionally independent to reflect their independent activation and operation, although it is likely that their behavioral manifestations are inversely related to some degree. As shown by Cacioppo et al. (1997), the outward displays of these two systems may depend on physical and environmental constraints that limit expression of two underlying systems to a single bipolar response. For example, organisms can either approach or withdraw from a stimulus at any given moment, regardless of the underlying positivity and negativity of their attitudes toward the stimulus. Furthermore, because feedback mechanisms between these two systems are likely to exist, the reciprocal relationship between them may also vary according to environmental conditions. For example, Zautra, Potter, and Reich (1998) have argued that under high stress, the separate affect systems may merge functionally (i.e., high negative affect dampens the positive affect system) in order to reduce uncertainty and conserve resources. Thus, the appearance of bipolarity may at times mask the functional separateness of the appetitive and aversive systems. 1.3. Hypotheses for the present research In the present set of four studies, we sought to determine whether a common structure underlies the various manifestations of the appetitive–aversive distinction reviewed above. Participants completed an array of individual difference measures representing distinct domains of psychological research and methodological variations. These data were examined with confirmatory factor analysis (CFA), a method well-suited to the identification of higher-order structures underlying covariances among a series of manifest variables. With CFA, it is possible to determine whether the imposition of a higher-order structure, represented in latent variables, can account for the observed pattern of covariation, and if it does, S.L. Gable et al. / Journal of Research in Personality 37 (2003) 349–372 355 which of several possible models best fits the observed data. We expected that one set of measures would reflect a latent factor representing appetitive processes, whereas, a second set of measures would tap a latent factor representing aversive processes. The specific measures were chosen to represent personality, affect, and motivation, classified a priori as ‘‘appetitive’’ or ‘‘aversive’’ on theoretical grounds. Measures assessing an individual s orientation to the presence (or absence) of positive environmental features were considered appetitive, whereas, measures assessing an individual s orientation to the presence (or absence) of negative environmental features were considered aversive. The heterogeneity of these measures not withstanding, we hypothesized that a common two-factor structure would emerge; that this structure would distinguish appetitive variables from aversive variables; and that this structure would fit the data better than other plausible structures (e.g., single-variable models; groupings based on research domain). We also incorporated several methodological variations in order to demonstrate that the appetitive–aversive distinction would emerge even when common method variance works against it. Finally, we again note that measures were classified into domains (e.g., personality, motivation) based on their most typical representation in the literature. Domain classifications were used for parsimony and are not intended as rigid classifications. Data collection procedures for Studies 1–4 were similar. Undergraduate participants enrolled in introductory personality courses in four different semesters, completed individual difference measures during multiple sessions held at different times during the semester. Each study at minimum included measures from the domains of personality, motivation, and affect. We used exploratory principal components analysis in Study 1 to examine the viability of our model. Studies 2–4 were analyzed with confirmatory factor analysis.