Four-Quadrant Investigation of Job-related Affects and Behaviours

Emphasizing differences in activation as well as valence, six studies across a range of situations examined relations between types of job-related core affect and 13 self-reported work behaviours. A theory-based measure of affect was developed, and its four-quadrant structure was found to be supported across studies. Also consistent with hypotheses, highactivation pleasant affect was more strongly correlated with positive behaviours than were low-activation pleasant feelings, and those associations tended to be greatest for discretionary behaviours in contrast to routine task proficiency. Additionally as predicted, unpleasant jobrelated affects that had low rather than high activation were more strongly linked to the negative work behaviours examined. Theory and practice would benefit from greater differentiation between affects and between behaviours. Affects are “primitive, universal, and simple, irreducible on the mental plane” (Russell, 2003, p. 148), “single feeling[s] at a slice in time” (Yik, Russell, & Steiger, 2011, p.705), and are central to many psychological constructs such as mood, emotion, well-being, happiness, satisfaction, and strain. Early research into workers’ affect focussed on associations with potential environmental antecedents such as personal discretion, workload, social support and so on (reviewed, for example, by Warr, 2007), but links with work behaviours have increasingly received attention. This paper identifies limitations in knowledge about affectbehaviour links, and extends research through theory-based differentiation between types of each. Four types of affect will be compared in tests of hypotheses about their associations with two categories of job behaviour. The paper goes beyond earlier affect-and-behaviour publications in its comprehensive examination of multiple forms of affect and in its comparison between those forms in relation to theoretically-distinct kinds of behaviour. This extension is important, for example, to identify affect-behaviour relationships which may have been obscured in previous less-differentiated research. Our focus is on different forms of what Russell (e.g., 1980, 2003) refers to as “core affect” – “that neurophysiological state consciously accessible as the simplest raw (nonreflective) feelings” (2003, p. 148). Core affect can exist without being given a label or attributed to any cause, and is considered by Russell and others to represent an integral blend of two primary attributes – pleasure and arousal. Those are shown as horizontal and vertical dimensions respectively within the circumplex in Figure 1 (e.g., Remington, Fabrigar, & Visser, 2000; Russell, 1980, 2003; Seo, Feldman Barrett, & Bartunek, 2004; Yik, Russell, & Feldman Barrett, 1999; Yik et al., 2011). The horizontal dimension, ranging from unpleasant to pleasant, covers affective valence; and the vertical arousal dimension, from low to high mental activation, concerns a person’s “state Affects and Behaviours 2 of readiness for action or energy expenditure” (Russell, 2003, p.156). A related notion has been viewed by laboratory researchers as “motivational intensity” or “the impetus to act” (Gable & Harmon-Jones, 2010, p. 323), and in Thayer’s (1989) biopsychological model, the upper two quadrants are viewed as “tense arousal” and “energetic arousal”. Illustrative feelings in relation to the two core affect dimensions are located around the outside of Figure 1, and summary labels are indicated in each quadrant: High-activation unpleasant affect (HAUA) such as anxiety; high-activation pleasant affect (HAPA) for example enthusiasm; low-activation unpleasant affect (LAUA) illustrated by depression; and low-activation pleasant affect (LAPA) such as feeling comfortable. Core affect is present in specific emotions in combination with particular cognitions. For example, anger, fear, distress, and jealousy all contain unpleasant, high-activation feelings. However, although they are similar in containing core affect those emotions differ from each other and from core affect in containing additional elements peculiar to themselves. Particular emotions involve their own kinds of cognitions and behavioural tendencies over and above their constituent core feelings (e.g., Lerner & Keltner, 2000; Yik et al., 2011). For example, when viewing pride as feeling good about oneself, the feelings involved represent instances of core affect whereas the self-focus is an additional cognitive component (Russell, 2003, p. 148). These partial content differences mean that different emotions are associated with behaviours in different ways. For example, the emotion of hostility may be linked to counterproductive work behaviour (Judge, Scott, & Ilies, 2006), but that particular association is not expected for every other emotion. The study of a particular emotion (e.g., hostility) is different from examination both of other emotions (e.g., guilt) and of the core feelings that are affect alone. In studying core affect it is thus necessary to use instruments which do not additionally assess the content of specific emotions. Measures have varied in that respect. For instance, scales by Burke, Brief, George, Roberson, and Webster (1989), Daniels (2000), Van Katwyk, Fox, Spector, and Kelloway (2000) and Watson, Clark, and Tellegen (1988) contain core-affect items but also a range of emotion-linked terms such as aggressive, annoyed, confused, disgusted and proud, whereas instruments by Warr (1990) and others are restricted to feelings that vary only in their valence and activation. The present studies take the latter approach, in order to examine only “primitive, universal and simple” feelings – the ones that can be common to all specific emotions – rather than also the additional elements that can vary between those emotions. Another important conceptual issue in this area is the overlap between a “mood” and an “emotion”. Researchers have disagreed in their use of these terms, and the distinction between them is generally agreed to be “blurry” (George, 2011, p. 148). The two constructs can be broadly distinguished in respect of intensity (emotions are generally more intense), duration (emotions tend to be briefer) and specificity (emotions are more targeted on particular features or people) (e.g., Cropanzano, Weiss, Hale, & Reb, 2003). For example, one might experience “anger” either as an emotion in response to a particular event (e.g., insulting behaviour by a customer) or as a generalized “bad mood” over a longer period. Russell (2003) defined mood as “prolonged core affect with no object or with a quasi-object” (p. 147). In many cases of affect it is arguable which is the more appropriate label (a mood or an emotion), and moods themselves contain emotions. The emphasis here is upon the nature of the constituent feelings within both of those. Affects and Behaviours 3 ASSESSING CORE AFFECT AT WORK Drawing on the circumplex framework, how should core affect in work settings be measured? Researchers (not themselves studying links with behaviour) have explored alternative factorial structures and assessment dimensions (e.g., Cropanzano et al., 2003; Remington et al., 2000; Russell, 1979, 1980, 2003; Watson & Tellegen, 1985, 1999; Yik et al., 2011), typically envisaging measurement axes across sections of Figure 1 – either diagonal (e.g., bottom-left to top-right) or horizontal (i.e., all unpleasant versus all pleasant). However, the few investigators focussing on the four separate sections (Burke et al., 1989; Mäkikangas, Feldt, & Kinnunen, 2007; Van Katwyk et al., 2000) have all reported that between-affect correlations particularly support a four-quadrant interpretation. For example, Mäkikangas et al. (2007) concluded from a longitudinal study of managers that the “fourfactor solution showed a better approximation with the data than the alternative models at both time points” (p. 213). Burke et al. (1989) described how the “confirmed four factors” (p. 1097) are paralleled by both neurological processes and clinical syndromes. A four-quadrant perspective offers greater specificity of analysis and precision of prediction, and will be adopted here. We provide the first examination of diverse work behaviours as a function of all four kinds of core affect. Several previous studies of links with behaviour have been directed at “positive affect” or (less often) “negative affect”. These might be expected to cover all feelings on the right-hand or left-hand side of the figure, involving all levels of activation. However, that has rarely been the case, in part because many measures have been based on the Positive and Negative Affect Schedule (PANAS) created by Watson et al. (1988). Those investigators (see also Watson, 1988; Watson & Tellegen, 1985) viewed positive affect (PA) and negative affect (NA) in terms of diagonal axes from LAUA to HAPA and HAUA to LAPA respectively. However, their PANAS scales were restricted to high-PA and high-NA quadrants only, containing activated positive terms such as enthusiastic, excited and inspired and activated negative terms such as distressed, jittery and nervous. Thus, although the two PANAS scales are often referred to as measuring “positive affect” (PA) and “negative affect” (NA), they in fact tap only certain kinds of those affects – feelings in the two upper segments of Figure 1 (e.g., Remington et al., 2000). This restricted coverage was later recognized by the scales’ originators. For example, they indicated that “to avoid terminological ambiguity, we have renamed the two factors Positive Activation and Negative Activation respectively, and use the abbreviations PA and NA in reference to these new labels only” (Tellegen, Watson, & Clark, 1999, p. 298). However, in practice these more appropriate labels have rarely been used in that way. Research described as linking positive affect to behaviour has thus frequently been restricted to the upper right-hand quadrant in Figure 1, whereas pleasantness extends through both sections on that side of the figure. Likewise, investigations of unpleasant affect using only PANAS items do not allow for the possibility that low-activation negative feelings differ in causes or consequences from the activated forms which alone are covered by that instrument. For example, Wright and Staw (1999), Dalal, Lam, Weiss, Welch, and Hulin (2009, Study 2), Fritz and Sonnentag (2009) and Foo, Uy, and Baron (2009) used PANAS (i.e., activated) items for positive and negative affect, and Tsai, Chen, and Liu (2007) supplemented the positive PANAS items with other activated terms – cheerful, delighted, happy and joyful. George and Zhou (2007) used all 20 PANAS items, and Den Hartog and Affects and Behaviours 4 Belschak (2007) drew items mainly from that set. In order to assess a wider range of activation new, more comprehensive scales of affect will be applied here. Based on these theoretical and practical arguments and on initial empirical support, we hypothesise the following: Hypothesis 1: A four-quadrant model of affect which differentiates high-activation unpleasant affect, high-activation pleasant affect, low-activation unpleasant affect, and lowactivation pleasant affect provides a better fit to empirical data than do alternative models of affect. Assuming the above hypothesis is correct, the quadrants’ relationships with other variables require investigation. In the next section, we view the four quadrants in relation to each other, and additionally focus upon possible variations between types of work behaviour. Overarching theoretical perspectives are lacking about between-behaviour differences, and a framework will be introduced in terms of both valence and discretionary content. PREDICTIONS ABOUT CORE AFFECTS AND BEHAVIOURS The valence of behaviours can be examined as their desirability versus undesirability to an individual or organization, recognizing that in certain cases those two assessments do not fully coincide. Thus positive behaviours include task performance, proactivity, initiativetaking, meeting targets, and citizenship contributions of many kinds. On the other hand, negative work behaviours include effort withdrawal or disengagement, theft, sabotage, workplace violence, bullying, incivility, and other “counterproductive” activities. Those harm the organization, either directly by affecting its functioning or property or indirectly by reducing the effectiveness of other employees (Fox, Spector, & Miles, 2001). In addition to differences between behaviours in terms of valence, it is important to distinguish between those which are required by a role and the more spontaneous, discretionary activities that exceed core obligations (Katz, 1964). An example of the latter is citizenship behaviour, such as helping others or expressing loyalty to the organization (Smith, Organ, & Near, 1983). In a similar vein, scholars have distinguished “extra-role” from in-role behaviour, with the former referring to positive, discretionary activities that are not specified in advance nor clearly recognized by formal reward and punishment systems (e.g., Van Dyne & LePine, 1998). Personal discretion is also fundamental to proactivity – self-initiated behaviour aimed at bringing about a change in the environment (e.g., Parker, Bindl, & Strauss, 2010), such as when taking charge of a situation (Morrison & Phelps, 1999) or using one’s personal initiative (Frese & Fay, 2001). As expanded later, we expect that affect-types will relate differently to discretion-based behaviours compared to routine in-role performance (Borman & Motowidlo, 1993) and task proficiency (Griffin, Neal, & Parker, 2007), which focus on carrying out pre-specified requirements. Hypothesis 2 proposes that links between pleasant affect and positive behaviours will be stronger for pleasant feelings that are activated rather than less activated. Those feelings (HAPA in Figure 1) have energizing potential that builds upon personal resources and can encourage approach behaviour. Energy within this kind of affect is likely to be important to self-start and to sustain challenging change activity, perhaps in the face of resistance and setbacks (Bindl & Parker, 2010). On the other hand, pleasant feelings with low activation (LAPA) are instead expected to be less related to positive behaviours. Feelings of that kind contain no impetus for action, generating reflection more than activity (Frijda, 1986) and in Affects and Behaviours 5 some circumstances broadening cognition rather than targeting attention on personal goalattainment (Gable & Harmon-Jones, 2010). Activated pleasant feelings are central to several models of self-regulation addressed primarily to behaviour outside organizations. For instance, Carver and Scheier’s (1998) framework embodies an affective feedback loop through which successful behaviour towards a target gives rise to feelings in terms specifically of activated elation; and Higgins’s (1997) regulatory focus theory argues that successful approach behaviours are linked to pleasant feelings that are especially characterized by cheerfulness and similar activated states. In respect of positive activity, both self-regulation models thus emphasize the top-right (HAPA) quadrant of Figure 1. Hypothesis 2 has received partial support from some previous research. Pleasant feelings that are activated (but described in original reports as “positive affect” as a whole) were found to be significantly associated with self-reported personal initiative by Den Hartog and Belschak (2007) and Fritz and Sonnentag (2009), and with manager-rated proactive behaviour by Tsai et al. (2007) and Parker, Collins, and Grant (2008). Significant associations with high-activation pleasant affect have also been reported for self-reported entrepreneurial effort (Foo et al., 2009), proactive goal regulation (Bindl, Parker, Totterdell, & Hagger-Johnson, 2012), and citizenship behaviours (Dalal et al., 2009; Tsai et al., 2007). However, these studies did not examine the full range of pleasant affect, and research that contrasts the two quadrants in their association with a wide range of positive work behaviours is still required. That empirical comparison will be made here. Hypothesis 2: The relationships of pleasant affect with positive behaviours are stronger for pleasant affect that is activated (HAPA) than for pleasant affect with low activation (LAPA). Hypothesis 3 differentiates between behaviours. Links between activated pleasant feelings and positive behaviours (above) are expected to be strongest for behaviours that are more discretionary – with self-set goals that require personal initiative, effort and perhaps risk-taking, as in proactive suggestion-making and undertaking citizenship activities beyond prescribed requirements. This theme was central to George and Brief’s (1992) model of positive mood at work – viewed as having “the most effect on behaviors that are performed of one’s own free will” (p. 324). Mischel (e.g., 1977) argued that the impact of within-person variables depends on “situational strength”, such that personal states were more influential in “weak” situations – when more mental and behavioural options were available. Similarly, Spector and Fox (2002) emphasized that behaviours that are voluntary leave “far more room for the action of personal intentions than more constrained and routinized task-related job performance” (p. 270). In empirical terms, affect-related comparisons between highand low-discretion behaviours are lacking, and we extend previous research by comparing observed patterns for the two kinds of behaviour. Hypothesis 3: The relationships of high-activation pleasant affect with positive behaviours are stronger for positive behaviours that are more discretionary than for positive behaviours that are less discretionary. Hypothesis 4 concerns core affect that is low in pleasure. It proposes that associations between unpleasant affect and negative behaviours tend to be stronger for negative feelings Affects and Behaviours 6 that are of low rather than high activation. In terms of Figure 1, many negative job behaviours are thus predicted to be more correlated with core job-related affect in the bottom-left quadrant (low-activated and unpleasant) than in the top-left quadrant (high activation and unpleasant). This reversal for negative behaviours of the Hypothesis-2 emphasis on more activated affect for positive activities may initially appear counterintuitive, but it is in fact central to several established perspectives and reflects the fact that low-activation unpleasant affect (LAUA) is the diagonal opposite of pleasant affect that is activated (HAPA). Lazarus (1991) emphasized that unpleasant affect with low activation is primarily linked to the loss or absence of something desired which requires action for its replacement, and laboratory studies have identified contrasting associations with unpleasant feelings of lowversus highactivation. For instance, Raghunathan and Pham (1999) found that high-activation unpleasant affect such as anxiety primed behaviours likely to reduce uncertainty, whereas low-activation unpleasant affect (sadness) tended towards the remediation of perceived deficits. In addition, at the heart of general models by Carver and Scheier (1998) and Higgins (1997) is the notion that a deficiency of desirable features and poor progress towards a goal generates unpleasant feelings that have low rather than high activation. Affect with low activation as well as low pleasure is viewed as being associated particularly with the absence of something that is wanted and which requires action for its attainment. Although depressed (LAUA) feelings linked to non-attainment of goals can in extreme cases lead to behavioural disengagement from a situation, total withdrawal (apart from quitting a job) is not possible for workers constrained by role demands and task targets. Depressed feelings at work are thus likely to be associated with those forms of personal withdrawal that are possible, such as distancing oneself from others (social withdrawal) or reducing one’s involvement in a situation (effort withdrawal or silence). Hypothesis 4: The relationships of unpleasant affect with negative behaviours are stronger for unpleasant affect that is of low activation (LAUA) than for highactivation unpleasant affect (HAUA). Note that this hypothesized importance for negative behaviours of low-activation unpleasant core affect does not extend to all kinds of emotion. As described earlier, the present perspective is restricted to core feelings alone whereas specific emotions also have content that is additional to the core and differs between emotions. For example, the activated emotion of hostility can predict counterproductive work behaviours (e.g., Judge et al., 2006), but the content of hostility is dissimilar from other emotions such as, for instance, shame or guilt, for which different predictions may be appropriate, and also goes beyond the core affect under investigation here. Different hypotheses and investigations are needed for different emotions in relation to different behaviours. Yik et al. (2011, p.710) identify a general agreement that “no single measure of fit for structural models should be relied on exclusively”, and previous studies have used indicators comprising either two or four sectors of Figure 1. In addition to testing hypotheses about the four quadrants individually (above), those multi-quadrant indicators also require examination to learn about findings’ descriptive comparability with other research. Relationships with behaviours will therefore also be summarized for two kinds of dual-quadrant measure. First, measures of all pleasant affect and of all unpleasant affect have been computed, ranging across both activation quadrants for a single direction of valence. We thus combined HAUA Affects and Behaviours 7 and LAUA scores into a comprehensive index of unpleasant affect, and HAPA and LAPA into pleasant affect as a whole. Second, we examined the two-quadrant axes identified by Watson and colleagues (above), combining diagonally-opposite lowand high-activation affects – HAUA with LAPA and LAUA with HAPA. Those axes are central to much affect theorising (e.g., Watson et al., 1988) and also to the self-regulation models of Higgins (1997) and Carver and Scheier (1998) and the overall happiness framework of Warr (2007). For example, Higgins’s (1997) regulatory focus theory treats self-regulation as either promotion-focused or preventionfocused. When a person is promotion-focused, he or she is oriented towards approach behaviour and “emotional experience varies along a cheerful-dejected dimension” (Brockner & Higgins, 2001, p. 39) – described here as running from HAPA to LAUA. However when a person is prevention-focused, tending towards avoidance of a situation, the pattern is different: “emotional reactions vary along a quiescence-agitation dimension” (p. 39) – described here as from LAPA to HAUA (see also Carver, 2003). After reversing the direction of negative feelings, diagonal two-quadrant compound scores were computed to represent Anxiety-Comfort (reversed-HAUA and LAPA) and Depression-Enthusiasm (reversed-LAUA and HAPA). Finally, in order to provide descriptive information about overall affect, responses from all quadrants were brought together into a global feeling-good score (again reversescoring the negative items). Van Katwyk et al. (2000) suggested that this four-quadrant index “offers the most comprehensive assessment” (p. 224), and the “single, integral blend” of both pleasure and activation is central to the model of Seo et al. (2004; see p. 426). We expect that combining together responses from pairs of quadrants or from all quadrants will reduce measurement sensitivity in comparison with the four separate scores.