• Keine Ergebnisse gefunden

Order affects help-content’s effects only regarding mixed feelings

3.6 Impact of language and dimension order on mean ratings and help-related con-

3.6.2 Order affects help-content’s effects only regarding mixed feelings

As the use of printed rating booklets and oral instructions in large groups makes it necessary to pre-define the order of stimuli and types of ratings given, special care was taken to distribute

3 RESULTS

Figure 13. Magnitude of social context effects on arousal and bipolar valence ratings according to language and gender. Effect sizes (Cohen’sd with Hedge’s correction) are shown for arousal (left) as well as bipolar valence (right). Effect sizes were calculated separately for the main analyses (stars), for men (white symbols) and women (black symbols), and for English (squares) and German (circles). Error bars represent 95% CIs for Cohen’s d.

different booklet versions randomly across participants in this study. Hence, the results reported so far should be fairly independent of the specific order of rating dimensions. Nonetheless, balanced distribution cannot rule out the possibility that effects were primarily present in one of the predefined orders. Potential influences of dimension order on mean ratings as well as on the effects of picture content were therefore assessed by re-calculating mean ratings per picture separately for participants either rating arousal or pleasantness on top of each page and such rating valence or unpleasantness on top of each page. As random distribution of different booklet versions in each testing session resulted in an equal distribution of language and gender for all scale and dimension orders (see Table 1), order effects were considered across men and

Main analyses

Figure 14. Magnitude of social context effects on pleasantness, unpleasantness and mixed feelings according to language and gender. Effect sizes (Cohen’s dwith Hedge’s correction) are shown for pleasantness (left), unpleasantness (middle) as well as mixed feelings (right). Effect sizes were calculated separately for the main analyses (stars), for men (white symbols) and women (black symbols), and for English (squares) and German (circles) study language. Error bars represent 95% CIs for Cohen’s d.

women as well as across German and English language samples.

Mean ratings on the four dimensions were unaffected by dimension order in both picture subsets, all|d| ≤.24 with CIs well overlapping zero. Accordingly, need of help contents’ effects on mean ratings were virtually the same for either scale dimension order and also compared to the main analyses (see Figure 15).

Mixed feelings, in contrast, were higher for those participants who rated arousal and pleas-antness first on each new page in the “need of help”, d = .64, 95% CI [0.24,1.03], as well as the “social context” subset, d = 0.92, [0.45,1.39]. Nonetheless, need of help content’s effects were not different according to dimension order; considering each order separately would have

3 RESULTS

Figure 15. Magnitude of need of help content effects according to dimension order. Effect sizes (Cohen’s dwith Hedge’s correction) are shown for the main analyses (stars) and separately for those participants rating arousal or pleasantness on top of each page (black circles) and those rating valence or unpleasantness on top of each page (white circles). Error bars represent 95%

CIs for Cohen’s d.

lead to the same interpretation that mixed feelings are unaffected by need of help content (see bottom most part of Figure 15).

Effect sizes regarding social context were virtually the same for both dimension orders (see Figure 16 and Figure 17). Divergences in interpretation would have only resulted from an enlargement of CIs due to reduction of sample size. Despite the overall higher mixed feelings reported by those participants first rating arousal or pleasantness on each page, social contextual effects were fairly constant, too (see right panel of Figure 17). Again, all effect sizes’

CIs overlapped to a great extent. Different interpretations compared to the main analyses could have only resulted from considering the sample rating valence or unpleasantness first.

For this case, data could have been interpreted as reflecting lower mixed feelings for “child alone” compared to “adult alone” as well as“social helping” pictures.

Both of these slight divergences cannot be explained by the distribution of picture categories on odd and even trial numbers (and hence either order of rating dimensions), as at least the number of “adult alone” and “child alone” pictures was similar for odd (8 and 4, respectively) as well as even trials (12 and 16 respectively).

In sum, these additional analyses revealed that mean ratings were not influenced by the

Main analyses first valence / unpleasantness first arousal / pleasantness

Social passive - social helping Social passive - child alone Social passive - adult alone

Social helping - child alone Social helping - adult alone Child alone - adult alone

Cohen‘s d Cohen‘s d

Arousal Valence

-3 -2 -1 0 1 2 3 4 5 6 -2 -1 0 1 2 3

Figure 16. Magnitude of social context effects on arousal and bipolar valence ratings according to dimension order. Effect sizes (Cohen’s d with Hedge’s correction) are shown for the main analyses (stars) and separately for those participants rating arousal or pleasantness on top of each page (black circles) and those rating valence or unpleasantness on top of each page (white circles). Error bars represent 95% CIs for Cohen’sd.

order of dimensions on each page. The same was true for help-related content effects on mean ratings. The mean intensity of mixed feelings, however, was influenced by the order in which different ratings had to be given, even though this order was reversed for the second picture on each page. Changes in interpretation of help-related content’s effects on mixed feelings were nonetheless hardly observable.

3 RESULTS

Main analyses first valence / unpleasantness

first arousal / pleasantness Social passive

- social helping Social passive - child alone Social passive - adult alone

Social helping - child alone Social helping - adult alone Child alone - adult alone

Pleasantness Unpleasantness Mixed feelings

-2

-3 -1 0 1 2 3 4 5 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3

Cohen‘s d Cohen‘s d Cohen‘s d

Figure 17. Magnitude of social context effects on pleasantness and unpleasantness ratings and mixed feelings according to dimension order. Effect sizes (Cohen’s d with Hedge’s correction) are shown for the main analyses (stars) and separately for those participants rating arousal or pleasantness on top of each page (black circles) and those rating valence or unpleasantness on top of each page (white circles). Error bars represent 95% CIs for Cohen’s d.

4 Discussion

With this thesis I aimed to contribute to our knowledge about self-reported emotions in the domain of social perception in two ways. First, I took a methodological perspective, broadly asking how the properties of the rating scales used to measure self-reported emotional experi-ences influence these ratings. The principal question was: How do bipolar and unipolar ratings of emotional experience relate to each other? Second, I wanted to shed light on the emotional experiences accompanying help-related social situations by asking: Which emotional experi-ences are elicited by the perception of need of help and helping behavior? The contributions of this thesis can be summarized in three major findings: 1) Ratings of unipolar pleasantness and unpleasantness are closely related to bipolar valence and arousal ratings. However, mixed feelings can be evaluated only on unipolar pleasantness and unpleasantness scales. And only the use of these scales allows disentanglement of changes in positive and negative affect. 2) Seeing somebody in need of help elicits reporting of stronger unpleasant feelings, less pleasant feelings, lower valence and higher arousal. 3) Situations showing an adult helping a child do not elicit discrete patterns of self-reported emotional experiences. A summary of the magnitude and di-rection of differences in emotional ratings according to picture content is provided in Figure 11.

An overview of mean ratings on each of the four dimensions as well as mixed feelings according to picture content is given in Table 10.

Table 10. Means and SDs for all picture content categories on all four rating dimensions and for mixed feelings.

M (SD)

Picture content Bipolar valence Arousal Pleasantness Unpleasantness Mixed feelings Bird need-of-help −0.69 (2.29) 3.94 (2.43) 2.56 (2.53) 3.59 (2.82) 1.06 (1.44) Bird no-need-of-help 0.88 (2.79) 2.79 (1.69) 3.81 (2.53) 1.79 (2.24) 1.04 (1.45) Child need-of-help −1.65 (4.08) 4.08 (2.41) 1.48 (1.93) 4.56 (2.54) 0.97 (1.31) Child no-need-of-help −0.01 (1.69) 2.66 (2.14) 2.94 (2.28) 2.43 (2.33) 1.22 (1.42) Child-alone −1.02 (1.85) 3.46 (2.39) 2.12 (2.23) 3.71 (2.65) 1.07 (0.11) Adult-alone 0.10 (1.68) 1.45 (1.81) 2.61 (2.25) 1.80 (2.17) 1.04 (1.45) Social-passive 0.55 (1.71) 2.32 (2.05) 3.81 (2.33) 1.75 (2.08) 1.14 (1.42) Social-helping 0.99 (1.74) 2.39 (2.11) 4.22 (2.45) 1.58 (2.00) 1.05 (1.40)

4 DISCUSSION

Figure 18. Mean ratings per picture on bipolar valence and arousal (A) and pleasantness and unpleasantness scales (B) colored according to picture category. Each data point corresponds to mean values for one picture. Orange dots represent pictures showing need of help, green ones no need of help, gray ones pictures (also) showing an adult figure. Filled dots represent pictures of humans, hollow ones of birds. Note that ratings cluster according to the a-priori designed picture content categories.

4.1 Self-reported emotions on bipolar and unipolar scales are closely related

The first analyses of this thesis considered the relation between two different types of rating scales: bipolar scales, measuring valence from happy to sad as well as arousal from calm to excited, and unipolar scales, assessing the intensity of pleasant and unpleasant feelings from absent to strong. Essentially, the same results as reported in earlier studies (Ito et al., 1998;

Kron et al., 2013) were obtained answering questions 1 a) and b) raised at the end of this thesis’ introduction. A nearly perfect linear correlation between bipolar valence ratings and the difference between pleasantness and unpleasantness ratings (see Figure 5 B) suggests that participants are able to draw the parallel between bipolar valence and unipolar pleasantness and unpleasantness. What is more, arousal ratings could also be well described as the overall intensity of unipolar pleasantness and unpleasantness ratings (see Figure 5 A).

One may oppose that ratings of the stimuli used in this study cannot be compared to other standardized stimulus sets, e.g. the IAPS, as the NeoHelp stimuli are abstract black-and-white comic drawings that show only harmless everyday situations. Nonetheless, there is evidence

that the ratings obtained for stimuli do not stand at odds with other standardized stimulus sets.

In studies that categorize pictures on the valence dimension only, mean arousal ratings found for stimuli in the “pleasant” category have in many cases been similar to the ones reported here (see Table 1 in Gr¨uhn and Scheibe, 2008). The fact that a congruence between ratings on unipolar and bipolar rating scales was similarly present for the NeoHelp as for the IAPS pictures generalizes and therefore illustrates the robustness of pleasantness’ and unpleasantness’ close mapping on bipolar valence and arousal ratings. Therefore, the potential objection that the stimuli used here were not comparable to more established stimulus sets strengthens rather than weakens the conclusion that bipolar valence and arousal ratings are generally tightly coupled with pleasantness and unpleasantness ratings.

Given that ratings on bipolar scales were closely related to aggregated ratings on unipolar scales, it might seem that both scales can be used interchangeably. However, this argument disregards the advantage of unipolar rating scales to allow calculation of the intensity of mixed feelings (Schimmack, 2001). Across all pictures participants reported having pleasant and unpleasant feelings when viewing the same picture, as has been shown for other stimuli before (Kron et al., 2013; Larsen and Green, 2013; Schimmack, 2001). This finding cannot be explained by an aversion of participants to use the rating scales’ extremes, as participants readily used them in the majority of trials. Moreover, there was a systematic inverted u-shaped relation between the intensity of mixed feelings and bipolar valence ratings (see Figure 6) that cannot be explained by a general aversion for extreme ratings. Pictures that were rated neutral in valence on a bipolar scale elicited the highest amount of mixed feelings. This finding answers question 1 c) and suggests that assessments of valence on a bipolar scale can be regarded as asking participants for a subtotal of their pleasant and unpleasant feelings – if both are present, they will cancel each other out.

In conclusion, participants self-reported emotions on unipolar and bipolar scales were closely related to another. This finding does not contradict the notion that physiological and neuronal responses are most parsimoniously divided into the core affects valence and arousal (Anders et al., 2004; Bradley and Lang, 1994, 2000; Wilson-Mendenhall et al., 2013). What they empha-size is that despite the fact that their bodies distinguish between arousal and valence, people’s self-reports of arousal might be heavily influenced by the strength of positive and negative affect. A considerable amount of variance in self-reported arousal seems to be explicable as the overall intensity of pleasant and unpleasant feelings. As Barrett and Bliss-Moreau (Barrett and Bliss-Moreau, 2009) have aptly pointed out “[. . . ] the structure of felt experience will not correspond to the brain processes that produced those experiences in a onetoone fashion.”

-4 DISCUSSION

even though ironically their article advocates an arousal-valence model of core affect.