Conflict precedes affective priming
August 2014
Student: E.A. Kovacs
Student number: 10671544
Master Thesis Psychology
Specialization: Brain and Cognition
Faculty of Social and Behavioral Sciences
University of Amsterdam
Supervisor Dr. R.H. Phaf
Second assessor Prof. dr. J.G.W. Raaijmakers
Wordcount: 7174
Table of contents Table of contents ... 2 Abstract ... 3 Introduction ... 4 Method Participants ... 11 Design ... 11
Material & Apparatus ... 13
Procedure ... 15
Results ... 16
Discussion ... 24
References ... 31
Abstract
It is currently unknown what the critical factor determining whether affective priming can be obtained or not is. Phaf and Rotteveel (2012) have recently proposed the affective monitoring hypothesis, which argues that conflict (i.e., neural competition) is required to elicit affect from neutral stimuli, but, as is argued here, perhaps also from affective stimuli. The present experiment contrasted conditions with and without conflict in the affective priming task. The conflict was manipulated by a) filtering emotion faces to BSF and LSF primes and b) parametrically varying the presentation durations of those primes. The main expectations for the study were that affective priming would be larger for the LSF than the BSF condition, because LSF elicits the conflict required for affective priming. We also expected some priming to result from the semantic congruence between prime and target. Results revealed that with affective priming there is no need to differentiate between conscious and unconscious processing, but merely between optimal and suboptimal presentation. Consciousness is thus not, as was presumed for a long time, the critical factor. The critical factor is conflict, which can be elicited by many sources that can strongly interfere with one another, and conflict resolution. We can state that there is a clear dissociation between suboptimal, LSF, global focus, non-analytic, open-mindset priming and optimal, BSF, local focus, analytic, closed-mindset priming.
Introduction
Emotion is traditionally viewed as an inner subjective experience, or feeling, of for example happiness, fear, anger or sadness, which is evoked by a particular situation or can occur spontaneously. Contrary to this popular view, many modern researchers have argued that nonconscious processes form the core of emotions (LeDoux, 1996; Phaf & Rotteveel, 2012; Winkielman & Berridge, 2004; Zajonc, 1980). LeDoux (1996) even observed that “The nonconscious processes are the fundamental facts of an emotion, and the conscious feelings are the frills that add icing to the emotional cake” (p. 302).
A paradigmatic example of research supporting the primacy of nonconscious affective processes are the priming experiments by Murphy and Zajonc (1993). For affective priming an emotional face, a neutral polygon or an empty screen, was presented as a prime before a meaningless (i.e., for the American participants) Chinese ideograph, either for 4 ms (suboptimal presentation) or 1000 ms (optimal presentation). Only suboptimal exposures of emotional faces produced shifts in subjects’ ratings of liking and good-bad judgments. With non-affective (‘cognitive’) priming this pattern of results was reversed with larger priming in optimal conditions (Murphy & Zajonc, 1993). The main conclusion was that when affect is elicited outside of conscious awareness, with suboptimal exposures, it is diffuse and nonspecific and can be transferred to meaningless stimuli. These findings seem to contradict the notion that emotions are primarily conscious experiences.
The findings closely fit within the dual-route model of LeDoux (1996). LeDoux proposed a model in which affective information is processed by the brain in parallel through a direct ‘quick and dirty’ route to the amygdala and an indirect route via the neocortex, which performs a slower but more detailed analysis of emotional stimuli. Simple features of the suboptimally presented emotion faces can already be processed by the direct
route, whereas optimally presented faces are processed more fully by both direct and the indirect route. In other words, when showing a stimuli suboptimally (e.g., for 4ms, as in the Murphy and Zajonc experiments) the quick direct route can alter subjects’ ratings of liking and good-bad judgments. This is not the case when showing a stimuli optimally (e.g., for 1000ms) because the indirect route ‘kicks in’ and inhibits the primary reaction. This can be compared to the startle reaction people often have to a fluff (direct route), before realizing it is not a spider (indirect route).
An interesting suggestion has been made by Vuilleumier, Armony, Driver, and Dolan (2003), that the low spatial frequencies of visual stimuli are primarily able to address the direct route. All visual stimuli are made up of high and low spatial frequencies and the full spectrum creates a clear image for the human eye. Technology has made it possible to exclude either high (or low) spatial frequencies from a picture, so that only low (or high) spatial frequencies remain. Only low spatial frequencies will produce a blurry picture, whereas only high spatial frequencies will produce a picture build up of detailed contours. Using event-related functional magnetic resonance imaging (fMRI) in humans, Vuilleumier, Armony, Driver, and Dolan (2003), found that neural responses in the fusiform cortex were greater with high-spatial-frequency faces, whereas amygdala responses were greater for low-spatial-frequency faces. More interestingly, Vuilleumier, Armony, Driver, and Dolan (2003), found that the low-spatial-frequency fear expression was more difficult to recognize consciously than the high-spatial-frequency or broad-spatial-frequency (full spectrum) fear expression. This strongly suggests that low-spatial-frequency components of emotional expressions, as opposed to high-spatial frequency or broad-spatial frequency components, are primarily processed nonconsciously.
The elicitation of nonconscious emotion with affective priming, however, has not been uniformly replicated. Andrews, Lipp, Mallan, and Konig (2011), for instance, found no evidence for suboptimal affective priming with emotional facial expression primes. Neither did Hermans (1996), whilst investigating automatic stimulus evaluation. It seems plausible that there is an additional, hitherto unknown, mechanism at work that accounts for these differences in findings with those of Murphy and Zajonc (1993). It is currently unknown what the critical factor is determining whether affective priming can be obtained or not.
A suggestion for such a critical factor can be derived from Sansom-Daly and Forgas (2010). They demonstrated that perceptual disfluency can influence processing quality and magnify priming effects. Priming effects were larger for perceptually disfluent (blurred) faces, consistent with disfluent images triggering more elaborate, constructive, and longer processing (Sansom-Daly & Forgas, 2010). A more recent paper (Alexopoulos, Fiedler & Freytag, 2012) hypothesized that disfluent primes, which resist an easily completed encoding process, should induce an open mindset and thereby result in stronger congruity effects than fluent primes inducing closed mindsets. Another study that implies a role for (dis)fluency of processing was performed by Barbot and Kouider (2012). They found in a continuous flash suppression task that while brief periods of invisible (i.e., suboptimal) prime presentation resulted in classical congruent priming on neutral face recognition, long periods of similarly invisible prime presentation led to a reversal of priming. In contrast, when the prime was visible, longer exposure resulted in the classical facilitation effects, revealing qualitative differences between conscious and nonconscious processes, which constitutes stronger evidence for nonconscious processing than the classical indirect-without-direct effects pattern of results (Merikle, 1992; Schmidt & Vorberg, 2006).
Perceptual disfluency may serve as the critical factor in the occurrence of nonconscious affective priming. This seems to fit well within the Affective Monitoring hypothesis of Phaf and Rotteveel (2012).Affective monitoring links affect to attention within modular competitive neural networks by proposing that affect is elicited by the temporal gradient of a local match-mismatch (i.e., a competitive) process of node activations. Negative affect is raised by mismatches, incongruency, disfluency, novelty, or dissonance (i.e., sustained competition), whereas positive affect follows from matches, congruency, fluency, familiarity, and resonance, at least when an initial mismatch is solved quickly (i.e., resolved competition). Both types of affect thus require initial competition (i.e., conflict) that can either be solved quickly (i.e., positive affect) or not (i.e., negative affect). It is hypothesized here that such initial conflict is also a prerequisite for affective priming. The present research investigated whether conflict is crucial for affective priming and whether differences in conflict resolution can account for stronger suboptimal than optimal affective priming. The hypothesis was that stronger suboptimal than optimal affective priming effects will ensue when conflict precedes or co-occurs with the priming. The present experiment contrasted conditions with and without conflict in the affective priming task. Conflict was raised by filtering the emotion (i.e., happy and angry) faces, so that blurred faces with only low spatial frequencies (LSF) remain. In addition, conflict resolution was manipulated by parametrically varying presentation times of the BSF (i.e., unfiltered) and LSF emotion faces that serve as primes.
In the present experiment affective priming on the speed of evaluating positive and negative words was investigated. With positive priming the evaluation speed of positive words increases and of negative words decreases, and vice versa with negative priming (cf. Fazio, Sanbonmatsu, Powell, & Kardes, 1986). Only right-handers participated in the
experiment, because there are indications of lateralization differences in emotional processing between right- and left-handers (Casasanto, 2009), and for interactions between lateralization and spatial-frequency processing (Peyrin, Mermillod, Chokron, & Marendaz, 2006).
Based on pilot research with affective priming a number of recommendations have been followed up. These include omitting forward and backward masks around the primes, because they can act as an independent source of conflict that may obscure the other priming effects.
Furthermore, research has shown that priming is stronger when participants are in an overall positive mood (Rotteveel & Phaf, 2007). Positive moods probably enhance priming due to non-analytic processing strategy that is adopted by the participants (Whittlesea & Price, 2001; see also, Rotteveel de Groot, Geutskens, & Phaf, 2001). Whittlesea and Price indeed found that the induction of an analytic (i.e., detailed and local) processing mode made both suboptimal mere exposure and implicit memory priming effects disappear that were present in non-analytic (i.e., broad and global) processing modes. It is therefore important to stimulate a non-analytic strategy with the participants. For this reason, active attempts were made to make the participants feel comfortable and happy. This was done by prolonging and personalizing the introduction, chitchatting in the breaks between experimental blocks, offering the participants a candy and eating several together with the experimenter. Hence, a highly relaxed atmosphere was created for the participant.
A non-analytic strategy will also be induced by familiarizing the participants with the target words in advance. Familiarization was done by presenting the target-words beforehand to the participants and letting the participants classify them as positive or negative prior to the actual experiment. In that way the words become more familiar and
they will therefore be more easily classified as positive or negative. Semantic evaluations are made beforehand and are thus less demanding in the experiment itself whereby a more non-analytic strategy can be adopted.
Whether the induction of the non-analytic strategy is successful was tested in a global-local focus test. Non-analytic modes, and presumably more positive moods (cf. Gasper & Clore, 2002) promote a global focus (i.e., on large-scale features of the stimulus) whereas analytic modes foster a local focus (i.e., on fine-scale features). A paper-and-pen global-local focus test, similar to the one used by Gasper and Clore, therefore served to measure the processing mode adopted by the participants.
Murphy and Zajonc (1993) gave different sets of instructions in optimal and suboptimal conditions which may alternatively account for the dissociation between conscious and nonconscious affective processing that they obtained. With optimal presentation, participants were instructed to ignore the primes, whereas with suboptimal presentation they were not. This might explain why the participants only incorporated these primes in their ratings in the latter conditions (Rotteveel et al., 2001). Therefore, in this experiment all participants are instructed to ignore the primes and to focus on the target word to be able to judge it as quickly as possible.
For the expectations of the study it is important to keep in mind that reactions to positive stimuli are generally faster than to negative stimuli (Taylor, 1991). As reaction times will be converted to affect indices; a subtraction of reactions to positive and negative words, therefore the affect index for a single prime face will be mostly negative, even in negative affective states. Negative values of the congruency-index, in which the two affect indices for happy and angry primes are subtracted from one another, represent congruent affective priming, whereas positive values represent incongruent priming. Congruency-indices
represent affective priming in a way that small negative indices represent large affective priming effects and near-zero indices represent little affective priming.
The expectations for the study were that nonconscious affective priming would be larger for the LSF than the BSF condition, because LSF elicits the conflict required for affective priming. We expected some conscious priming to result from the semantic congruence between prime and target. This type of priming was thought to be larger for BSF than for LSF primes and also to occur at longer SOAs. The nonconscious affective priming was expected to decrease as presentation duration increases due to conflict resolution. Some priming however was expected to remain, because still no fluent processing of the blurred LSF faces with longer presentation durations was expected. If the short presentation durations and LSF stimuli caused the most nonconscious affective priming, this would have confirmed that conflict is required for affective priming and that differences in conflict resolution account for the stronger suboptimal than optimal affective priming condition.
An alternative interpretation would be in terms of LeDoux’ (1996) dual-route model, namely that there will be no difference in the priming by BSF and LSF emotion faces. In this case somewhat more priming by BSF than LSF faces was expected due to the additional, presumably conscious, priming by the high spatial frequency (HSF) components. Inhibitory influences from the high road in LeDoux’ model were expected to reduce affective priming in both conditions (LSF/BSF) with longer presentation durations, because as presentation duration increases, information will be processed more consciously according to LeDoux (1996). According to LeDoux’ (1996) dual-route model priming ought to decrease as information is processed more consciously, but less for the LSF than the BSF faces.
Expectations for a perfomed face detectibality task were that for BSF faces accuracy of detectability will increase somewhat with longer presentation durations, but will be above
chance level even for the shortest durations due to the absence of masking. For LSF faces, face detectability will be more difficult (cf. Vuilleumier et al., 2003) even with longer presentation durations.
Expectations for the global-local focus test were that participants who show to have a more local preference, and presumably more analytic processing, on the task will display less nonconscious affective priming than participants who show to have a global preference, and more non-analytic processing (cf. Whittlesea & Price, 2001). For the conscious type of presumably more semantic priming the reverse may be true.
Method
Participants
Forty-six students (average age 21.2 year; SD = 1.9 years, 17 male; 29 female) from different Dutch universities participated in the experiment, after signing informed consent, for course credit or financial compensation. All participants were native Dutch speakers, had normal or corrected-to-normal vision and were strongly right-handed. Handedness was scored by the Dutch Van Strien handedness questionnaire (Van Strien, 1992). Only strictly right-handed students (i.e., students with the highest possible score (10) on right-handedness) were allowed to participate in the study.
Design
The experiment consisted of three experimental blocks and one block manipulation check, all comprising 160 trials per block. The experimental blocks, in which participants had to judge a target word as positive or negative after seeing an emotional face, had a 2 (positive or negative word) x 2 (LSF or BSF face) x 5 (presentation time) within-subjects design. The
dependent variable consisted of the reaction times of correct responses to positive and negative words (e.g., judging a positive word as positive). The differences in these reaction times represent affect-indices per condition. Subsequently, a calculation of the differences in affect-indices gave priming congruency-indices for happy and angry prime faces.
The fourth block, in which participants had to detect male or female faces, had a 2 (LSF or BSF) x 5 (presentation time) within-subjects design. The block was included as a manipulation check to check for level of consciousness in each condition. Detection accuracy (e.g., judging a male face as male) served as dependent variable.
Table 1: Calculation of affect-indices and congruency-indices
As the experimental task of target word evaluation is quite easy, participants with less than 90% accuracy were considered not sufficiently motivated and excluded from the analysis. Outliers, reaction times differing more than 1.5 IQR (interquartile range) per condition, were also excluded from the analysis with the Boxplot method in SPSS. To enhance the accuracy of the measurements, a speed-accuracy correction was performed on the raw data (e.g., see Smilek, Enns, Eastwood, & Merikle, 2006). For every participant, reaction times per condition were corrected for the proportion correct responses in the
Prime Target RT Affect-indices Congruency-indices
Smile Xa Xa – Xb for Affect-index for
Angry Xb happy faces happy faces
Smile Xc Xc - Xd for Affect-index for
condition by the following formula: reaction time / (1 – (number of mistakes / number of trials)) = new reaction time. Due to this correction, longer reaction times with few errors remain the same, whereas shorter reaction times with more errors become longer.
To determine the adopted strategy of participants, a global-local focus test, as used by Gasper and Clore (2002, see also Kimchi &Palmer, 1982), was performed after the manipulation check. Participants saw an overall shape (e.g., a triangle) made up of smaller geometric figures (e.g., squares). Their task was to indicate which of two alternative figures (e.g., a square made of triangles or a triangle made of squares) was more similar to the previous one (see Appendix C). A median split was performed on the local-global results to determine more locally and more globally focussed groups.
Overall ANOVAs were performed on the average reaction times according to the above design. All reaction times were converted to affect-indices and congruency-indices, as discussed previously. The means were used to conduct one sample t-tests to evaluate whether their means are significantly different from 0. Means of detection accuracy were used to conduct one sample t-test to evaluate whether their means are significantly different from 50, chance level.
Material & Apparatus
Face stimuli (60 angry BSF, 60 angry LSF, 60 happy BSF, 60 happy LSF) for the experimental trials were selected from the Karolinska Directed Emotional Faces set (Lundqvist, Flykt, & Öhman, 1998). All faces were either filtered to contain only low spatial frequencies (LSF) with a low-pass Gaussian filter (SD = 0.015; less than 17 cycles per image width) or contained all spatial frequencies (BSF). Target words consisted of 120 happy and 120 angry related words selected from the EmoClar database by Phaf, Van der Leij, Stienen and Bierman
(2006). The positive and negative words (see Table 2 and Appendix A and B) were matched on length and word frequency (Keuleers & Brysbaert, 2010). The faces as well as the words were randomly selected by the computer, so that they could be repeated within an experimental session.
Table 2: Length and frequency of target words
All participants were seated on a fixed distance of 30 cm from the computer screen by use of a chinrest. The face stimuli had a size of 562 (width) x 762 (height) in pixels and 15 (width) x 20 (height) in cm. All faces occupied the same surface on a rectangle that subtended a visual angle of 15.2° (width) x 19.9° (height). Response buttons were fixed over participants, left for negative (in the priming task) and male (in the manipulation check), and right for positive (in the priming task) and female (in the manipulation check).
A prime and a target word were subsequently presented on a gray background. Participants were to evaluate the target word by pressing the congruent response button. The target word remained visible on the screen until response. Participants were instructed to respond as quickly as possible. The inter-trial interval was jittered between 1000ms and 2000ms for all trials, to prevent participants from adopting a fixed response pace.
Positive Negative
Length (in characters) Range 4 – 7
Mean 5.9 5.7
Frequency per million words Range 0.2 – 477.2
Items (16) for the global-local focus task, as used by Gasper and Clore (2002), were selected from Kimchi and Palmer (1982). The items were printed on paper and presented on the desk between experimenter and participant at a comfortable reading distance for the participant. All participants adopted approximately the same distance of 50 cm from the items, as they were instructed to point at their chosen alternative.
Procedure
The experiment was introduced as an investigation into the speed of evaluating words, which would be preceded by sometimes barely visible faces. Upon arrival, participants read the information brochure and signed informed consent. Hereafter, participants were seated behind the computer and given specific verbal information and instructions. Participants were told that the experiment was a replica of standard priming experiments, because results had been contradictory with one another. Participants were openly told the expectations of the study and that it is currently unknown what the critical factor determining whether affective priming occurs or not is. The participants were told that this experiment could provide the answer, most importantly to demonstrate the absence of deception but also to generate a feeling of importance and thereby motivate them. Instructions included the explicit remark that participants were to ignore the facial primes and focus solely on the target word to be able to judge it as quickly as possible.
After familiarization with all target words (i.e., reading and categorizing them as positive or negative beforehand), ten practice trials were performed. In case of any mistakes, the practice trials would be repeated. Participants were specifically instructed to respond as quickly as possible to target words, and that eventual mistakes caused by the aspiration to be quick would be logical. However, when participants made too many mistakes (e.g.,
eventually crossing the 10% inaccuracy limit), participants were instructed to be slightly more accurate. Three affective priming blocks of 160 trials then followed. Blocks were separated by short breaks, in which the experimenter tried to seek contact with the participants, and motivate them. A manipulation check of level of consciousness was then performed in the fourth block also consisting of 160 trials. In the exit interview participants were asked for their opinion, whether they had used any strategies and whether they noticed something extraordinary in the experiment. All abnormalities were registered in a logbook. Their answers were taken into account whilst analyzing the data.
Results
In total 46 out of 48 participants were analyzed for the results, two participants were excluded from the analysis because of less than 90% accuracy on target word evaluation. In the remaining participants accuracy varied between 93% and 99%. Means and standard deviations for all conditions are displayed in Table 3.
Table 3: Overview table of the average reaction times (SD) in ms
BSF LSF
Happy Angry Happy Angry
Positive Negative Positive Negative Positive Negative Positive Negative 10 640 (118) 648 (112) 648 (101) 647 (98) 641 (90) 643 (88) 656 (107) 637 (95) 30 639 (99) 642 (97) 652 (110) 623 (101) 616 (97) 645 (93) 672 (172) 630 (106) 90 627 (105) 637 (94) 669 (95) 622 (111) 633 (95) 641 (117) 653 (104) 620 (80) 270 628 (90) 628 (86) 634 (87) 615 (88) 641 (101) 633 (115) 633 (97) 630 (92) 1000 632 (104) 632 (94) 646 (94) 619 (77) 607 (86) 611 (94) 618 (103) 622 (107)
As expected, overall mean reaction times decreased as presentation duration of the primes increased and shorter reaction times ensued when face emotion and target word were congruent than when they were incongruent with respect to affective value. In the 2 x 2 X 5 ANOVA on the average reaction times presentation duration (i.e., 10ms, 30ms, 90ms, 270ms, 1000ms) indeed significantly reduced reaction times (F(4, 180) = 4.596, p <0.05). Long presentation durations resulted in short reaction times and short presentation durations resulted in longer reaction times (10ms mean=645ms; SD=101ms, 30ms mean=640ms; SD=112ms, 90ms mean=638ms; SD=101ms, 270ms mean=630ms; SD=94ms, 1000ms mean=623ms; SD=95ms).
Furthermore, the interaction effect of facial emotion (i.e., happy or angry) and target word (F(1, 45) = 12.172, p <0.01) showed that the both faces and words were affectively valenced and that affective priming took place in the experiment. Reaction times to happy faces with positive words (mean=630ms; SD=98ms), as well as to angry faces with negative words (mean=627ms; SD=96ms), were faster than reaction times to happy faces with negative words (mean=636ms; SD=99ms) and to angry faces with positive words (mean=648ms; SD=109ms). Unexpectedly, a marginally significant effect for target word (i.e., positive or negative) showed up (F(1, 45) = 3.575, p = 0.065), which was opposite to the standard valence effect (Taylor, 1991). Negative words (mean=631ms; SD=98ms) were judged faster than positive words (mean= 639ms; SD=104ms), presumably because negative words were evaluated by clicking the left button with the index finger and positive words by clicking the right button with the less frequently used middle finger.
The interaction-effect between facial emotion, target word and presentation duration proved marginally significant (F(4, 180) = 2.135, p = 0.0783). Priming appeared to decrease as presentation duration increased, which corresponds to the classical
stronger-suboptimal-than-optimal affective priming effect by Murphy and Zajonc (1993; see also Rotteveel et al., 2001). A more detailed insight in this effect can be derived from the affective priming congruency-indices that will be discussed later in the results section.
A median split was performed on the local-global results to determine group with more local and more global focussed, which is assumed to correspond to analytic and non-analytic processing strategies (cf. Whittlesea & Price, 2001). When adding this variable to the ANOVA two interactions with local-global group emerged. A marginally significant interaction effect for prime filtering (i.e., BSF or LSF) and global or local group was found (F(1, 44) = 3.675, p = 0.0617), which was further qualified by an interaction effect between prime filtering, facial emotion, target word and global or local focus (F(1, 44) = 12.856, p <0.01). As can be seen from Table 4, this interaction seems to indicate that LSF priming was larger with global than local groups, but the reverse was true for BSF priming. This effect too can be more easily interpreted from the congruency-indices.
Table 4: Means and standard deviations in ms for the interaction between prime filtering, facial emotion, target word and global or local focus
BSF LSF Happy Positive Negative Angry Positive Negative Happy Positive Negative Angry Positive Negative Local 619 (92) 632 (85) 655 (95) 615 (72) 625 (89) 612 (68) 630 (99) 623 (74) Global 645 (110) 642 (105) 645 (100) 634 (112) 630 (98) 654 (121) 660 (135) 632 (111)
All reaction times were converted to affect indices, and subsequently to congruency-indices. Average congruency-indices and their standard deviations for all conditions, without the local-global group split, are displayed in Table 5. Negative values of the congruency-index reflect congruent priming, whereas positive values would reveal incongruent priming. The present results, however, only showed negative values for short durations and near-zero values for longer durations.
Table 5: Means and standard deviations for congruency-indices in ms Mean BSF (SD) Mean LSF (SD) 10ms -9 (125) -20 (143) 30ms -32 (138) -71 (188) 90ms -58 (163) -41 (161) 270ms 1000ms -18 (124) -27 (130) 4 (156) 1 (117)
Figure 1: Congruency-indices as a function of presentation duration of the prime and prime filtering -80 -70 -60 -50 -40 -30 -20 -10 0 10 10 30 90 270 1000 Pr im in g in d ex (m s) Presentation duration (ms) BSF LSF
The 2 x 5 ANOVA on the affect-indices again showed a marginally significant decrease with presentation duration (F(4, 180) = 2.135, p = 0.0783), but no significant main effect for prime filtering or an interaction effect for prime filtering and presentation duration (Fs < 1). Indications for congruent affective priming could only be found with 30ms and 90ms durations. In one sample t-tests (hypothesized mean = 0) |t(45)| exceeded the value of one only in five conditions. For BSF faces this was the case for 30 ms (t(45) = -1.554, NS, Cohen’s d = 0.23), 90ms (t(45) = -2.403, p < 0.05, d = 0.36), 1000 ms (t(45) = -1.391, NS, d = 0.21), and for LSF faces 30ms (t(45) = -2.557, p < 0.05, d = 0.38), 90 ms (t(45) = -1.735, p = 0.895, d = 0.25). Paired t-tests of the differences between conditions corroborated the stronger-suboptimal-than-optimal affective priming pattern, but only for LSF faces. All differences within the BSF conditions proved to be unreliable. For LSF faces (marginally) significant differences were found between the 10ms and 30ms condition (t(45) = 1.967, p = 0.0553, d = 0.28), the 30ms and 270ms condition (t(45) = -2.283, p < 0.05, d = 0.34), and the 30ms and 1000ms condition (t(45) = -2.156, p < 0.05, d = 0.32 ).
When the global–local group split was performed (median=9, local group n=21; global group n=25) the same significant interaction effect ensued for prime filtering and global-local group (F(1, 44) = 12.856, p < 0.01), as in the ANOVA on the reaction times. Table 6 shows that priming of LSF faces is stronger for participants with a global focus than for participants with a local focus, whereas priming of BSF faces was stronger for participants with local focus than for participants with a global. In absolute terms there were even indications that the Murphy and Zajonc priming pattern was specific to LSF primes and globally focused participants, but the interaction between filtering duration and group did not reach significance (F(4, 176) = 1.745, p = 0.14). From it can even be seen that BSF priming
was only present for the local, presumably analytic, group, and that it extended to longer durations than the LSF priming for the global group.
Table 6: Means and standard deviations for congruency-indices in ms, after performing the global-local group split
BSF LSF
Local (SD) Global (SD) Local (SD) Global (SD)
10 -24 (106) 4 (139) -7 (122) -31(160)
30 -73 (133) 3 (134) -1 (98) -129 (224)
90 -94 (114) -28 (193) 5 (95) -80 (194)
270 -33 (113) -6 (134) 39 (121) -26(177) 1000 -40 (146) -16 (117) -5 (120) 5 (116)
Figure 2: Congruency-indices for BSF and LSF primes as a function of presentation duration, for the local and global group
-140 -120 -100 -80 -60 -40 -20 0 20 40 60 10 30 90 270 1000 Pr im in g in d ex (m s) Presentation duration (ms) BSF global LSF local BSF local LSF global
-60 -50 -40 -30 -20 -10 0 10 Local Global Pr im in g in d ex (m s) Presentation duration (ms) BSF LSF
Table 7: Means and standard deviations for congruency-indices in ms N BSF (SD) LSF (SD)
Local 21 -53 (124) 6 (111) Global 25 -8 (144) -52 (182)
Figure 3: Congruency-indices as a function of prime filtering and global-local focus.
One sample t-tests (hypothesized mean = 0) confirmed that BSF priming was only present in the local group and LSF priming only in the global group. For the global group the LSF 30 ms condition differed significantly from zero (M = 129 ms, SD = 224 ms, t(24) = -2.880, p <0.05, d = 0.58) and the LSF 90ms condition marginally significantly (M = -80 ms, SD = 195, t(24) = -2.055, p = 0.0509, d = 0.41). For the local group both the BSF 30ms condition (M = -73 ms, SD = 133 ms, t(20) = -2.513, p <0.05, d = 0.55) and BSF 90ms condition (M = -94 ms, SD = 114 ms, t(20) = -3.783, p <0.05, d = 0.82) proved to be significant.
Paired sample t-tests only revealed (marginally) significant differences between condition for LSF faces in the global group and for BSF faces in the local group, but not in for LSF faces in the local group en for BSF faces in the global group. For the BSF local group a
marginally significant difference was found between the 10ms and 90ms condition (t(20) = 2.061, p = 0.0526, d = 0.45), and for the LSF global group (marginally significant differences were found for the 10ms and 30ms condition (t(24) = 2.939, p <0.05, d =0.59), the 30ms and 270ms condition (t(24) = -1.920, p = 0.0668), the 30ms and 1000ms condition (t(24) = -2.556, p <0.05, d = 0.39), and the 90ms and 1000ms condition (t(24) = -2.142, p <0.05, d = 043). Hence, with the global-local split more priming is found with shorter durations by LSF faces in the global group, whereas more priming by BSF faces is found in the local group.
The manipulation check indicated remaining awareness for all durations and with the two types of filtering, but the level of awareness was affected by the manipulations. Table 8 shows the mean accuracies and standard deviations of the gender decision task. Particularly with the shorter durations, gender detection of LSF faces was lower than of BSF faces. Overall, angry faces were less readily detected than happy faces. Analyses of the manipulation check revealed significant effects for presentation duration (F(4, 180) = 80.053, p <0.0001), prime filtering (F(1, 45) = 54.403, p <0.0001), facial emotion (F(1, 45) = 41.685, p <0.0001), and presentation duration x prime filtering (F(4, 180) = 7.865, p <0.0001).
Table 8: Accuracy of the gender decision task (SD) in percentage
BSF LSF
Happy(SD) Angry(SD) Happy(SD) Angry(SD)
10 84 (16) 78 (16) 69 (18) 64 (19)
30 89 (14) 83 (14) 80 (15) 76 (13)
90 92 (10) 90 (12) 88 (10) 82 (13)
270 95 (8) 90 (11) 96 (6) 87 (14)
50 60 70 80 90 100 10 30 90 270 1000 A cc u ra cy (%) Presentation duration (ms) BSF LSF
One sample t-tests (hypothesized mean = 50) were conducted on the accuracies of the gender decision task to evaluate whether their means are significantly different from 50. All conditions were significantly different from 50 (P < 0.000), implying conscious recognition of the primes in all conditions. Presentation duration, however, does influence to what extend a prime is consciously recognized.
Figure 4: Measured face detectability as a function of presentation time.
Discussion
The current experiment contrasted conditions with and without conflict in an affective priming task, to seek evidence for the hypothesis that conflict (i.e., neural competition) is required to elicit affect from affective stimuli. Conflict elicited by low-spatial frequency filtering enhanced affective priming (cf. Sansom-Daly & Forgas, 2010), particularly for the group of participants that focussed on large-scale, global and low spatial frequency features of the stimuli. BSF faces of course also contain the low spatial frequencies, which may explain the remaining affective priming with these faces. Here, however, the global group
showed almost no affective priming and the priming seemed confined to the local group. Consistent with the affective monitoring hypothesis, the unresolved conflict with shorter presentation durations led to more priming than the resolved conflict with longer durations, particularly for the LSF primes that evoke the most conflict. This pattern of results also conceptually replicates the affective priming results of Murphy and Zajonc (1993), but now with a different dependent measure (i.e., speed of evaluating positive and negative stimuli; cf. Fazio et al., 1986). In contrast to Murphy and Zajonc, however, the shortest duration of 10 ms revealed little priming, and there was conscious identification of the primes even in the 30 ms and 90 ms conditions where the priming was largest. Such results can be seen as further support for the use of the term “suboptimal” instead of “subliminal”. This also suggests that consciousness is not the critical factor in the stronger-suboptimal-than-optimal priming effect, but that some other process is involved. We argue here that this factor is conflict and the degree to which it is resolved.
As mentioned before, affective priming was greater for primes with low-spatial frequency filtering than primes that contained the full spectrum of spatial frequencies. This finding is in accordance with the hypothesis that conflict precedes affective priming; LSF primes elicit conflict required for affective priming and hereby create stronger priming effects. BSF primes, however, also induced some priming. While this could be due to the low-spatial frequencies of a full spectrum prime, a more complete interpretation can be made after splitting the group into global or local focus preference. The global-local focus split, quite unexpectedly, was of great influence to the results. It appeared that priming for LSF faces was significantly stronger for people who had adopted a more global focus (i.e., a non-analytic strategy) whereas priming for BSF faces was significantly stronger for people who had a adopted a local focus (i.e., an analytic strategy). It also appeared that for people
with a global focus more nonconscious (or rather suboptimal) affective priming developed, whereas for people with a local focus more conscious (or rather optimal) priming developed. This difference can also be linked to the article of Alexopoulos, Fiedler and Freytag (2012), who hypothesized that disfluent primes, which resist an easily completed encoding process, should induce an open mindset (i.e., a global focus) and thereby result in stronger congruity effects than fluent primes inducing closed mindsets (i.e., a local focus).
It seemed that there was dissociation between 1) priming induced by BSF primes, local focus, an analytic strategy, long presentations durations and closed mindsets and 2) priming induced by LSF primes, global focus, a non-analytic strategy, short presentation durations and open mindsets. Furthermore it seemed that the combination of positive mood, non-analytic strategy and global focus did not simply enhance priming effects. Rather it seemed that two different processes were at work. It appeared that for people with a local focus there was weak suboptimal priming with short presentation durations but strong optimal priming with long presentation durations, whereas for people with global focus there was strong suboptimal priming with short presentation durations but weak optimal priming with long presentation durations.
This pattern corresponds with the findings of Barbot and Kouider (2013), who found priming for both short and invisible primes and long and visible primes. Possibly the short and invisible priming corresponds with LSF priming, whereas long and visible priming corresponds with BSF priming, and that global-local focus, or rather (non)analytic strategy decides which of two priming occurs. Short and invisible priming could correlate with LSF priming because both elicit conflict required for affective priming. Perhaps with longer presentation durations, where no conflict is elicited, a form of cognitive priming occurs. In fact, the results of Barbot and Kouider (2013) highly correspond with the results of the
current experiment when replacing visible and invisible priming respectively with no conflict and conflict conditions. In other words, both experiments call for affective conflict-dependent suboptimal priming versus cognitive conflict-inconflict-dependent optimal priming. Further research is required to confirm this hypothesis.
When looking at the results for the manipulation check, faces that are happy, BSF and have optimal presentation durations were recognized most accurately. However face recognition was above chance level for every condition, meaning that the experiment does not differentiate between conscious and unconscious processes, but merely the amount of conscious processing allowed for by presentation time. The terms optimal and suboptimal are therefore preferred over terms such as luminal and subliminal, and conscious and subconscious. This also calls for a reinterpretation of results of Murphy and Zajonc (1993). Apparently consciousness is not the critical factor in the stronger-suboptimal-than-optimal priming effect, but conflict is. The stronger-suboptimal-than-optimal priming effect is only true for LSF primes, for BSF primes the opposite is true. The 4ms condition, as used by Murphy and Zajonc (1993), apparently elicited conflict required for affective priming, whereas the 1000ms condition, did not elicit conflict, and therefore resulted in cognitive priming. In this way, the Murphy and Zajonc findings (1993) do not fit within the dual route model (LeDoux, 1996), which only distinguishes between a quick/dirty route and a slow/high route but does not incorporate conflict.
Although the results contained some strong suggestions, and evidence, for the hypothesis that conflict precedes affective priming, an optimal pattern of results as would be expected has not been found. A factor that plays a great role herein, are the standard deviations. Quite strong and convincing mean differences have been found, however the standard deviations were so big that statistical analyses failed to prove significant. The
standard null-hypothesis significance testing (NHST) in itself is already not recommendable, as statistical significance says little about power of measure, while insignificances can still be very meaningful. In a recent article, Cumming (2013) even observed that “We need to shift from reliance on NHST to estimation and other preferred techniques. The new statistics refers to recommended practices, including estimation based on effect sizes, confidence intervals, and meta-analysis” (p. 7). Either way, measures ought to be more accurate and standard deviations ought to be smaller. The previously referred to pilot research contained forward and backward masks around the primes and had standard deviations of around 60ms, while this experiment had standard deviations of around 100ms. The variance in reaction times can firstly be reduced by experimenting on a homogenous group of participants. However the great variance is presumably caused by uncertainty of participants. After seeing an emotional face, the interval between face and target word was randomly chosen by the computer. The uncertainty of when the target word would appear, gives variance amongst reaction times. With the pilot study there was a mask between the face and the target word, eliminating the uncertainty.
For follow-up studies a couple of recommendations should be considered. A first remark concerns the 10ms condition. In all analyses the 10ms condition was too weak to show any priming results. In spite of the hypothesis that conflict is a requirement for affective priming, too much conflict and the inability to resolve it could weaken the affective priming. The complete hypothesis, namely, is that easily resolved conflict, in the form of fluent processing after conflict facilitates priming, whereas too much conflict, in the form of continuous disfluent processing does not. It is possible that with short presentation durations conflict was too large, so only disfluent processing resulted. When comparing this to the findings of Murphy and Zajonc (1993), who did find affective priming in the 4ms
condition, which is more likely to be 10ms as a result of old equipment, a possibility is that the short presentation duration elicited enough conflict to create affective priming, whereas the combination of short presentation duration and LSF-conflict elicited too much conflict so that only disfluent processing remained, and no affective priming occurred.
A second recommendation is the usage of global-local group as manipulation. In the present experiment all participants were positively stimulated to adopt a non-analytic strategy, but still half of participants had a local focus. In a follow-up study it would be beneficial if one group is stimulated to adapt a non-analytic strategy whereas the other group is stimulated to adopt an analytic strategy. Both groups will have to perform a global-local focus test at the end of the experiment. In the current experiment it was assumed that a local or global focus indicates the adoption of an analytic or non-analytic strategy, however it is even possible that when manipulating analytic strategy, a dissociation ensues between global-local focus and (non)analytic strategy.
Another remark concerns our own expectations. The idea is that longer presentation durations correspond to more resolution of the conflict, as proposed by the affective monitoring hypothesis. However the same hypothesis also states that negative affect is elicited by persistent conflict, whereas positive affect arises if it is swiftly resolved (Phaf & Rotteveel, 2012). This would mean that longer exposure durations lead to more positive judgments, after both happy and angry faces. As participants had to judge target words on their positive or negative value, interference of this positive fluency is expected. With negative target words this interference was in contradiction with the task and could thereby possibly produce more conflict, both leading to longer reaction times. The facial expressions probably play a role in the resolution of conflict. Happy faces trigger more resolution than
angry faces. However, when primes are presented too briefly, the conflict cannot be resolved at all anymore.
Although this study does not provide complete unambiguous answers to specific research questions, nevertheless it does give strong conjectures towards solving the critical factor determining whether affective priming occurs or not is. At the least, we can state that with affective priming there is no need to differentiate between conscious and unconscious processing, but merely between optimal and suboptimal presentation. Consciousness is thus not, as was presumed for a long time, the critical factor. The critical factor is conflict, which can be elicited by many sources that can strongly interfere with one another, and conflict resolution. We can also state that there is a clear dissociation between suboptimal, LSF, global focus, non-analytic, open-mindset priming and optimal, BSF, local focus, analytic, closed-mindset priming. Lastly, there is one thing we can say with absolute certainty: emotions are not simply the inner subjective experience, or feelings, of for example happiness, fear, anger or sadness. Emotion in its purest form, or rather affect, is elicited by the temporal gradient of a local match-mismatch (i.e., a competitive) process of node activations (Phaf & Rotteveel, 2012), and conscious feelings are merely “the frills that add icing to the emotional cake” (LeDoux, 1996, p.302).
References
Alexopoulos, T., Fiedler, K., & Freytag, P. (2012). The impact of open and closed mindsets on evaluative priming. Cognition and Emotion, 28, 978-994.
Andrews, V., Lipp, O. V., Mallan, K. M., & Konig, S. (2011). No evidence for subliminal affective priming with emotional facial expression primes. Motivation and Emotion, 35, 33 – 43.
Barbot, A. & Kouider, S. (2012). Longer is not better: Nonconscious overstimulation
reverses priming influences under interocular suppression. Attention perception and psychophysics, 74, 174–184.
Casasanto, D. (2009). Embodiment of abstract concepts: Good and bad in right- and left- handers. Journal of Experimental Psychology: General, 138, 351-367.
Cumming, G. (2014). The new statistics: why and how. Psychological Science, 25, 7-29. Fazio, R.H., Sanbonmatsu, D.M., Powell, M.C., & Kardes, F.R. (1986). On the automatic
activation of attitudes. Journal of Personality and Social Psychology, 50, 229–238. Gasper, K. & Clore, G.L. (2002). Attending to the big picture: Mood and global versus local
processing of visual information. Psychological Science, 13, 34-40.
Hermans, D. (1996). Automatic stimulus evaluation: An experimental analysis of the
preconditions for evaluative stimulus discrimination using an affective priming paradigm. Unpublished doctoral dissertation, University of Leuven, Belgium.
Keuleers, E. & Brysbaert, M. (2010). SUBTLEX-NL: A new measure for Dutch word frequency based on film subtitles. Behavior Research Methods, 42, 643-650.
Kimchi, R., & Palmer, S.E. (1982). Form and texture in hierarchically constructed patterns. Journal of Experimental Psychology: Human Perception and Performance, 8, 521–535.
LeDoux, J.E. (1996). The emotional brain: The mysterious underpinnings of emotional life. New York: Simon & Schuster.
Lundqvist, D., Flykt, A., & Öhman, A. (1998). The Karolinska Directed Emotional Faces.
Psychology section, Department of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden.
Merikle, P.M. (1992) Perception without awareness, American Psychologist, 47, 792-795. Murphy, S.T., & Zajonc, R.B. (1993). Affect, cognition and awareness: affective priming
with optimal and suboptimal stimulus exposures. Journal of personality and social psychology, 64, 723–739.
Peyrin, C., Mermillod, M., Chokron, S., & Marendaz, C. (2006). Effect of temporal
constraints on hemispheric asymmetries during spatial frequency processing. Brain and Cognition, 62, 214-220.
Phaf, R.H., & Rotteveel, M. (2012). Affective Monitoring: a generic mechanism for affect elicitation. Frontiers in Psychology, 3, 47.
Phaf, R.H., Van der Leij, A.R., Stienen, B.M. & Bierman, D. (2006). Positieve, neutrale en negatieve woorden bij minimale aanbieding: Een ordening door perceptuele clarificatie [Positive, neutral, and negative words at minimal presentation levels: Ordering by perceptual clarification], Amsterdam, The Netherlands: Technical Report, Universiteit van Amsterdam
Rotteveel, M., de Groot, P., Geutskens, A., & Phaf, R.H. (2001). Stronger Suboptimal Than Optimal Affective Priming? Emotion, 1, 348–364.
Rotteveel, M., & Phaf, R.H. (2007). Mere exposure in reverse: Mood and motion modulate memory bias. Cognition and Emotion, 21, 1323-1346.
Sansom-Daly, U.M., & Forgas, J.P. (2010). Do Blurred Faces Magnify Priming Effects? The Interactive Effects of Perceptual Fluency and Priming on Impression Formation. Social Cognition, 28, 630–640.
Schmidt, T., & Vorberg, D. (2006). Criteria for unconscious cognition: Beyond the zero-awareness criterion. Perception & Psychophysics, 68, 489-504.
Smilek, D., Enns, J.T., Eastwood, J.D., and Merikle, P.M. (2006) Relax! Cognitive strategy influences visual search. Visual Cognition, 14, 543-564.
Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: the mobilization- minimization hypothesis. Psychological bulletin, 110, 67-85
Van Strien, J.W. (1992). Classificatie van links- en rechtshandige proefpersonen. Nederlands Tijdschrijft voor de Psychologie, 47, 88-92.
Vuilleumier, P., Armony, J.L., Driver, J., & Dolan, R.J. (2003). Distinct spatial frequency sensitivities for processing faces and emotional expressions. Nature Neuroscience, 6, 624-631.
Whittlesea, B.W.A., & Price, J.R. (2001). Implicit/explicit memory versus
analytic/non-analytic processing: Rethinking the mere exposure effect. Memory & Cognition, 29, 234-246.
Winkielman, P., & Berridge, K.C. (2004). Unconscious Emotion. Current Directions in Psychological Science, 13, 120-123.
Zajonc, R.B. (1980). Feeling and thinking: Preference need no inferences. American Psychologist, 35, 151-175.
Appendix A
120 positive target words (word – length – frequency)
Gein 4 1,49 Knus 4 2,36 Pret 4 9,03 Zoen 4 9,99 Vers 4 17,86 Wijs 4 36,25 Sexy 4 36,34 Lach 4 46,72 Wens 4 59,39 Spel 4 95,11 Seks 4 100,14 Life 4 113,95 Baby 4 151,8 Fijn 4 198,06 Vrij 4 232,91 Blij 4 277,3 Aaien 5 2,1 Party 5 4,96 Fraai 5 6,4 Hoera 5 7,62 Zalig 5 9,15 Humor 5 19,6 Bruid 5 21,13 Winst 5 21,63 Zacht 5 22,85 Jeugd 5 27,88 Kunst 5 37,09 Zomer 5 42,9 Vrede 5 65,01 Sterk 5 95,22 Feest 5 98,26 Droom 5 109,42 Trots 5 111,46 Nieuw 5 119,85 Geluk 5 166,02 Prima 5 276,79 Samen 5 370,64 Vital 6 1,07 Extase 6 1,42 Geinig 6 1,62 Lollig 6 3,52 Komiek 6 4,07 Intiem 6 6,04 Vredig 6 6,86 Jarige 6 7,68 Warmte 6 10,5 Zuiver 6 10,63 Glorie 6 11,21 Gloria 6 11,34 Nuttig 6 11,96 Handig 6 13,81 Begrip 6 13,97 Zoenen 6 14,73 Passie 6 14,77 Mazzel 6 16,08 Moedig 6 18,84 Klasse 6 19,87 Vrijen 6 22,89 Natuur 6 26,96 Wensen 6 28,84 Cadeau 6 29,29 Gezond 6 31,83 Geduld 6 34,28 Wonder 6 44,91 Schoon 6 49,28 Gelukt 6 65,43 Zingen 6 65,52 Ruimte 6 77,25 Lichen 6 88,89 Dansen 6 102,91 Success 6 103,64 Muziek 6 107,46 Winnen 6 116,56 Voelen 6 117,86 Gevoel 6 128,27 Aardig 6 191,95 Liefde 6 208,9 Slapen 6 209,22 Lekker 6 276,15 Rustig 6 354,52 Kleurig 7 0,18 Gelukje 7 0,23 Jeugdig 7 0,5 Snoezig 7 0,91 Begeren 7 0,91 Vleiend 7 1,3 Profijt 7 1,53 Hoopvol 7 1,74 Beeldig 7 1,92 Swingen 7 1,97 Verrukt 7 2,01 Strelen 7 2,65 Bekwaam 7 2,84 Melodie 7 3,04 Amusant 7 3,59 Fluiten 7 4,67 Gunstig 7 4,71 Geestig 7 5,03 Elegant 7 5,28 Stralen 7 5,28 Grapjas 7 5,53 Welzijn 7 5,83 Ambitie 7 5,85 Komedie 7 6,47 Social 7 7,94 Rijkdom 7 9,97 Vermaak 7 11,18 Oprecht 7 11,34 Knuffel 7 11,94 Vreugde 7 13,56 Gedicht 7 14,73 Applaus 7 17,38 Beleefd 7 17,49 Winnaar 7 23,6 Prettig 7 27,37 Vrolijk 7 31,51 Energie 7 38,81 Prinses 7 42,47 Bloemen 7 46,19 Zwanger 7 62,77
Appendix B
120 negative target words (word – length – frequency)
Grof 4 11,89 Wond 4 14,02 Coma 4 14,09 Vies 4 19,8 Vuil 4 24,06 Vals 4 26,6 Heks 4 26,76 Saai 4 31,47 Graf 4 34,53 Zwak 4 35,4 Boos 4 105,79 Ziek 4 129,2 Haat 4 151,32 Fout 4 165,27 Wees 4 233,69 Pijn 4 266,16 Bang 4 477,21 Zweep 5 4,87 Tumor 5 5,79 Plaag 5 8,42 Satan 5 8,8 Boete 5 8,83 Woest 5 9,03 Schok 5 11,66 Demon 5 15 Wreed 5 15,18 Vloek 5 18,77 Woede 5 24,95 Virus 5 28,91 Vrees 5 29,43 Zoned 5 33,48 Alarm 5 34,78 Straf 5 35,45 Kogel 5 43,72 Wraak 5 44,02 Brand 5 44,39 Ruzie 5 64,19 Angst 5 69,34 Zweer 5 79,42 Slaan 5 94,51 Wapen 5 140,59 Moord 5 141,99 Kwaad 5 157,83 Doden 5 217,98 Kerker 6 4,25 Onweer 6 4,28 Zeiken 6 4,3 Fatal 6 4,64 Rouwen 6 4,78 Giftig 6 4,99 Wurgen 6 5,05 Akelig 6 5,1 Debiel 6 5,15 Zweten 6 5,6 Leegte 6 5,9 Rotten 6 6,47 Maniak 6 6,88 Somber 6 6,88 Boeven 6 7,45 Engerd 6 7,66 Lullig 6 7,82 Afgaan 6 8,51 Zinken 6 9,6 Arrest 6 11,55 Crisis 6 11,89 Stress 6 13,86 Herrie 6 13,97 Gebrek 6 14,02 Hitler 6 15,5 Triest 6 17,52 Klagen 6 19,76 Kanker 6 21,88 Smerig 6 24,35 Zielig 6 24,86 Verzet 6 27,01 Lelijk 6 29,23 Lijden 6 35,88 Ziekte 6 37,14 Paniek 6 39,86 Geweld 6 42,17 Risico 6 44,23 Strijd 6 44,73 Duivel 6 45,12 Gemeen 6 45,37 Aanval 6 47,59 Huilen 6 54,52 Geweer 6 55,39 Gewond 6 60,62 Vijand 6 60,69 Lasting 6 67,69 Gevaar 6 91,54 Honger 6 92,45 Wapens 6 109,83 Schuld 6 178,1 Slecht 6 267,81 Bliksem 7 10,63 Brutal 7 11,07 Verward 7 11,57 Mislukt 7 11,78 Zinloos 7 13,74 Schamen 7 15,96 Aanslag 7 17,06 Dwingen 7 17,72 Verraad 7 19,07 Overval 7 21,29 Schande 7 23,42 Leugens 7 26,8 Ellende 7 27,01 Eenzaam 7 31,56 Twijfel 7 34,03 Jaloers 7 42,9 Misdaad 7 43,22 moorden 7 45,28 Gevecht 7 48,98 Verlies 7 49,07 Slechts 7 87,95 Ongeluk 7 100,96 Pistol 7 102,63 Vechten 7 126,37 Sterven 7 175,26
Appendix C
