Today´s work experience: Precursors of both how I feel and how I think about my job?

An intriguing finding is that relations between trait affectivity, state affect, and JS seem to depend on the JS questionnaire that was used, especially if the questionnaire focused more on the cognitive or affective component of JS ( Kaplan et al., 2009 ). Associations were stronger with the use of Kunin Faces Scale (KFS) of JS (1955)–a mono-item measure of overall satisfaction that focuses primarily on the affective component of JS–compared to other scales. One explanation for this moderating effect might be the relation and weight of the affective and cognitive components of JS in questionnaires. Thus, comparing the affective and cognitive JS components of different scales is interesting ( Fisher, 2000; Moorman, 1993; Organ & Near, 1985 ). Brief and Roberson (1989) studied the relation of three different scales of JS with affective experiences. In contrast to the Job Descriptive Index (JDI) and the Minnesota Satisfaction Questionnaire (MSQ) ( Weiss, Dawis, England, & Lofquist, 1967 ), KFS (Kunin, 1955 ) was correlated with cognitions and retrospectively measured affect at work (state affect during the past week; Job Affect Scale, JAS) ( Brief, Burke, George, Robinson, & Webster, 1988 ). JDI and MSQ were related to cognitions about the job only. Therefore, Brief and Roberson concluded that in terms of affect and cognition, the KFS is “the most balanced of the JS scales” (p. 723). Niklas and Dormann (2005) also showed that state affect had comparably the largest impact on JS measured with KFS. Niklas and Dormann (2005) suggested that state affect influences KFS at the time when it is measured. We expect, therefore, the association between state affectivity and current JS to depend on the type of JS measure used. Thus, the first study hypothesis postulates that the association between state affectivity and current JS is stronger when the latter is measured by KFS compared to measurement with a more cognitive scale, such as the one by Wegge and Neuhaus (2002) that asks for evaluation facets of JS (hypothesis 1). In addition, the association between positive and negative work experiences and current JS should be stronger when the latter is measured by KFS compared to measurement with a more cognitive scale (hypothesis 2).

For a long time, research on JS focussed on the cognitive evaluation of various job conditions, like satisfaction with salary, supervisors, colleagues, and work conditions. It was a central benefit from the Affective Events Theory (AET) ( Weiss & Cropanzano, 1996 ) to shed light on the affective component of JS by demonstrating that affective experiences in the workplace have an impact on global JS and its consequences. Therefore, during the last decade, attention increased on the effects of emotions at work ( Weiss, 2002 ). Job conditions underlying cognitive evaluation and affective experiences triggering affective parts of JS lead to the idea of partly independent processing pathways for affective and cognitive JS. There should be some overlap, because for instance emotional experiences at work correspond to events that elicit emotions, and these are more likely to appear in the background of unfavourable job characteristics, e.g., low autonomy ( Weiss & Cropanzano, 1996 ). According to AET, controlling for the cognitive component of current JS increases the associations of the affective component of JS with state affect and trait affectivity, while controlling for current affective JS component should certainly reduce these associations. This is the second primary focus of this study. The second hypothesis of the study therefore postulates that control of the cognitive component of JS by controlling ( Wegge & Neuhaus, 2002 ) job facets scale in predicting KFS will increase associations between KFS and previous experience, while control of KFS in predicting satisfaction with job facets components will decrease association between satisfaction with job facets and previous experience (hypothesis 3).

Table 1 shows correlations at day-level between study variables. The correlation between both JS scales was .48. Correlations between positive trait affect and negative trait affect were the highest, followed by correlations between positive trait affect and positive state affect. At day-level, the correlation coefficient between positive and negative state affect was also high. Correlations between negative trait affectivity and state affect were moderate to high. Correlations between measures of current JS and positive trait and state affect were high while correlations between measures of current JS and negative trait and state affect were moderate. Correlations coefficients between JS and valence of events were moderate for positive valence while coefficients were small to moderate for negative valence. Both the KFS and the facets scale of JS showed strong correlations with trait and state affectivity and valence, but coefficients were consistently stronger for KFS than for the facets scale of JS. The first hypothesis postulated the correlation between state affect and job satisfaction measured with KFS to be larger than with facets scale. A test of difference in dependent correlation ( Steiger, 1980 ) showed significant differences ( z = 1.80, p = .036, one-tailed, for positive state affect; z = 1.95, p = .026, one-tailed, for negative state affect). A test of the second hypothesis did not confirm differences between scales ( z = 0.19, p = .424, one-tailed, for positive work experience; z = -0.43, p = .335, one-tailed, for negative work experience).

The measurement of fluctuating JS is crucial in research on occupational health, work motivation, turnover, and work safety ( Ilies & Judge, 2004; Klumb, Elfering, & Herre, 2009 ). To our knowledge this is the first time-based experience sampling study that compared two measures of current JS with respect to previous work experience. Based on AET, the study hypothesised that current state affect and valence of work experiences would be closely associated with KFS, a measure of JS measure that is more focused on the affective component of JS than with the job facets scale, that is more cognitive in nature ( Elfering & Grebner, 2010, 2011 ). The hypothesis was partly confirmed for state affect and by ICC that showed more daily variability in KFS than in the facets measure. The lower ICC value for the KFS measure indicates greater reactivity of the affective component of job satisfaction to work events. However, this interpretation is only valid when the reliability of KFS is not lower than the reliability of the facets measure. Meanwhile, the reported estimate of .69 for the KFS reported by Wanous et al. (1997) is not different from the reliability of the facets measure found in the current study. Thus, the replication seems necessary to a restest of KSL because the reliability estimate that was reported by Wanous et al. (1997) was based on the correlation of Kunin's (1955) scale with other JS scales and therefore is an estimate of construct validity and not of reliability. Items that were compared were not parallel (i.e., they did not have the same standard error, the same random errors, etc.). In order to estimate the reliability of the KFS, the appropriate reliability coefficient is a test-retest one. Retest coefficient captures random, specific, and transient errors ( Salgado, 2015; Schmidt, Le, & Ilies, 2003 ).

The finding is important both theoretically and practically. First, it shows that the affective component of JS is more reactive to emotional states at work than the cognitive component of JS. However, rejection of the second hypothesis is not in line with AET that affective JS component could be supposed more closely bound in time with work events while the cognitive component of JS is less closely bound to the time when work events occurred. In this study, the valence of work experiences was related equally with both job satisfaction scales. However, in the multilevel regression analysis–beyond state affect–trait affectivity and positive valence of preceding work experiences had an incremental value only in the prediction of KFS, not in the prediction of the facets scale values. A preliminary interpretation of this finding is that the unique influence of specific work-related experience beyond emotional state, e.g., experience of competence, is unlikely to influence directly the cognitive component of JS but the affective component first. In summary, research on AET using experience sampling methods should consider the use of both scales. Whenever the evidence for daily work experience on change in JS is reviewed, the kind of JS measures should be considered as a moderator. If the use of both scales is impossible, KFS is recommended not only for its one item stand-alone validity ( Wanous et al., 1997 ), but also for its sensitivity to within-person changes of daily job satisfaction.

Work events’ influence on current JS seems to differ depending on positive and negative valences ( Fisher, 2002; Maybery, Jones-Ellis, Neale, & Arentz, 2006 ). In the prediction of KFS, negative work experience had no incremental value above negative state affect and negative trait affectivity. This finding is in line with AET that supposed negative state affect mediated the effects of negative work experience on JS ( Weiss & Cropanzano, 1996 ). Positive work experience, however, seems to have an influence on JS that is independent from current positive state affect and positive trait affectivity. We consider specific success-related emotions like feelings of competence to be involved. Positive work experiences that are perceived as individual success, including goal-attainment, pro-social success (i.e., preventing others from failure and harm), and positive feedback are linked to feelings of competence and JS ( Grebner, Elfering, & Semmer, 2008, 2010 ). One might point to the empirical redundancy of positive and negative trait affectivity and positive and negative mood that are highly related and ask for the real contribution of these scales to the prediction of JS ( Le, Schmidt, Harter, & Lauver, 2010; Schmidt & Harter, 2008 ). Indeed, Judge, Hulin, and Dalal (2012) make the point that it is difficult to comprehend a person who exhibits high scores on both PA and NA. However, there is also evidence that PA and NA are the affective manifestations of two relatively independent bio-behavioural systems (i.e., an approach system and an avoidance/withdrawal system; Watson, Wiese, Vaidya, & Tellegen, 1999 ). In addition, the associations of positive and negative affectivity with job satisfaction do not seem to be redundant. In the meta-analysis of Connolly and Viswesvaran (2000) , correlations between job satisfaction and positive affectivity, corrected for coefficient alpha in the two measures, were .49 for positive affectivity ( N = 3,326, k = 15) and r = .33 for negative affectivity (N = 6,233, k = 27). The pattern is also observed in daily event sampling studies. For instance, Judge and Ilies (2004) reported average state job satisfaction to correlate r = .35 with positive trait affectivity and r = -.16 with negative trait affectivity as rated by significant others. In the current study, the pattern is rather comparable with higher correlations between job satisfaction and positive affectivity and lower correlations between job satisfaction and negative affectivity. Thus, we suggest a real but small unique association of positive and negative affectivity that is related with the affective component of job satisfaction measure.

The second hypothesis suggested different effects of mutual control of affective and cognitive JS component in multilevel regression analyses. Controlling for the cognitive component of current JS should increase the proposed associations of the affective component of JS with state affect and trait affectivity, while controlling the affective JS component should certainly reduce these associations to accentuate the differences. Results did not confirm the second hypothesis, as all associations remained nearly unchanged. The preliminary interpretation is that without mutual control the unique affective versus cognitive characteristics of both scales are already linked to state affect.

The diary assessment may have forced participants to reflect on their work experiences, leading thereby to unusual scores for JS. Thus, the reactivity of the sampling method–self-observation–might have changed attitudes and behaviour towards the job ( Klumb et al., 2009 ). Moreover, we exclusively used self-reports to measure variables. This can lead to inflated stressor-strain associations through correlated measurement errors (common method variance; e.g., Semmer, Grebner, & Elfering, 2004 ). Replication in a larger sample is necessary. Given the intensity of the data collection efforts, with multiple observations per participant being d daily, it is difficult to collect such data with large numbers of participants. In comparison to other experience sampling studies on the topic, the current sample size ( N = 40) is small but not at the lower end ( N = 27 in Ilies & Judge, 2002; N = 24 in Weiss, Nicholas, & Daus, 1999 ). The small sample lowers the generalizability of results and we can not exclude variance restriction in JS measurement. Moreover, the Kunin's (1955) scale measured global job satisfaction and the scale of Wegge and Neuhaus (2002) addressed different facets of job satisfaction. A replication should rely on a Kunin's scale that addresses different facets of job satisfaction, too. Finally, we did not test retest reliability of KFS.

One of the advantages of this study, however, is its high participation rate and the combination of data from different JS measures. In the meanwhile, it is important to keep in mind that the separation of cognition and affect in content and measurement is not so easy as separation in theory. Moreover, even if we could clearly separate measures of cognition and affect, given their relationship, causality in both directions have to be supposed.

Despite much research on JS, construct validation is increasingly necessary. This study has shown the relative weight of affective and cognitive components within JS measure to contribute to the strength of associations with daily work experience and state affect. Ambulatory assessment in occupational research should rely on both attitude components of JS when AET is in focus.

The authors of this article declare no conflict of interest.

The Swiss National Science Foundation supported part of this research (SNF 100011-105538 and SNF 100014_144057).