1
Affective priming: An unmasked effect?
5814162
Marianne Lilienblum
Bachelorproject Affective Priming Hans Phaf & Alexander Krepel May 2015
2 Abstract
Affective priming remains an effect with several secrets to be explored. Factors that contribute to a successful effect might be related to conflict, as is suggested by Phaf and Rotteveel (2012) in their Affective Monitoring framework. It states that conflict is necessary to elicit affect, which is crucial for eliciting an effect by affective priming. Since in previous studies by Kovacs (2014a ; 2014b), the masking of stimuli might function as a source of conflict which obscured the real priming effect, this study consisted of conditions with and without masked affective stimuli. To create the necessary conflict for an effective affective priming effect, the conditions consisted of low- and broad spatial frequency primes and short and long presentation duration of the stimuli. It was expected that low spatial frequency and short presentation of stimuli would elicit more priming than less difficult to perceive stimuli, e.g. broad spatial filtered and long lasting. However, the main expectation was that with unmasked stimuli, the priming effect would be greater than with masked stimuli. Individual differences in focus and personality also were taken into account, expecting individuals to differ in priming in terms of local and global focus and neuroticism and extraversion. Results showed mainly priming effects under rather optimal conditions, with the least priming in the suboptimal condition. With regard to masked stimuli, priming effects differed not much between masked and nonmasked conditions. The findings with regard to participant’s foci and to personality are mixed. However, there was still support for conflict as a possible crucial factor for affective priming.
3 Affective priming: An unmasked effect?
The affective state of an individual plays an important role in therapeutic interventions - could we use knowledge about it to improve interventions? As supposed by Evans (2010), affective priming could be very useful in this respect. A strong positive emotional prime could work as a facilitation tool to alter a client’s attitude or willingness towards a treatment in a positive way. But does affective priming really work that easily? There has been quite a bit of discussion on the nature of affective priming. Indeed, as has been found out by Murphy and Zajonc (1993), affective priming can affect an individual’s choice with regard to the
evaluation of stimuli following an affective prime. In their study, participants’ judgment on Chinese ideographs was only influenced in the direction of an emotional facial prime, when the prime was administered suboptimally. This implies that affective priming includes
processes differently from non-affective primes. Also, it adds evidence to the assumption that emotional processing can take place without conscious awareness, which is stated in LeDoux’ Two Route Model (1996): In contrast to non-affective stimuli, affective stimuli are processed via a direct route which leads to automatic processing of the affective information. The second route is addressed by stimuli that are presented optimally, which makes conscious access possible and which can alter the response to these stimuli.
Liddell et al. (2004) add some support for the finding that affective stimuli are
processed differently in subliminal and supraliminal conditions. In contrast to neutral stimuli, fear stimuli elicited an automatic response for face processing in the subliminal condition, while an inhibitory component was observed in the supraliminal condition. So far, the
processing itself did not play a major role in priming, yet it could be a key to the answer why the effect of subliminal affective priming is so difficult to replicate, as is reported by Andrews et al. (2011). They propose that masking might have led to no affective priming, with too stringent masking preventing any subtle priming effects. Without masks, the priming effect was visible. Interestingly, when using masked facial stimuli, also Kovacs (2014a) did not find a priming effect for suboptimal conditions, whereas without masks, the effect could be obtained (Kovacs, 2014b). The only difference between Kovacs’ studies is the use of masking. That the effect of affective priming did appear in her experiments with non-masked primes, leads to the
assumption that the masks might have led to an impact on the effect of the prime in the earlier study (Kovacs, 2014b). Kovacs placed this finding within the Affective Monitoring Framework (Phaf & Rotteveel, 2012), in which it is assumed that influency of processing resulting in conflict and therefore in negative affect, could be the underlying mechanism. Phaf and Rotteveel (2012)
4 suggest that affect occurs as a result of conflicts or (mis)matches. Conflict can be caused by a disruption of fluency in processing. It is assumed that the resolving of conflict elicits positive or negative affect, with quickly resolved conflict resulting in positive affect and unresolved conflict in negative affect, while no conflict leads to no affect at all.
Supposed by Kovacs (2014b), processing of the prime would be distorted because of the mask, leading to negative affect. Since in every condition a mask was applied which led to an elicited negative affective effect that did not differ between conditions, the masking obscured the priming effect. In line with expectations, Kovacs found that low spatial filtered (LSF) primes elicited overall a larger priming effect than broad spatial filtered (BSF) primes, a finding consistent with Sansom-Daly & Forgas (2010). Due to the LSF primes, which are more difficult to perceive, conflict might be elicited, which is not the case for BSF primes. Since conflict leads to affect, which is an important requirement for affective priming to occur, the interpretation of the results fits into the Affective Monitoring Framework. Additionally, Kovacs (2014b) found that there was a difference between globally and locally focused individuals; globally focused individuals were more sensitive to LSF facial stimuli, whereas individuals with a local focus showed a significant stronger priming effect for BSF facial stimuli. Based on the Affective monitoring framework (Phaf & Rotteveel, 2012), she suggested that the underlying mechanism could be a difference in elicited conflict.
This difference between individuals in elicited conflict while processing negative affective stimuli, leads to an interesting factor that applies to affective priming: individual differences in affective processing. Dannolowski et al. (2006) discovered that patients, suffering from a clinical depression, showed greater sensitivity for negative expressions and an automatic negative processing bias for facial affect. In a second priming study
(Dannolowski et al., 2007), primed with positive and negative stimuli, depressive patient’s processing of negative stimuli was more intense than that of positive stimuli. These findings suggest that a negative affective state might lead to an automatic way of processing affective stimuli, which is different from people without an (e.g. negative) affective condition.
Individual differences in processing of affective stimuli are also found in non-patient studies. Not only the intensity but a direct relation between neuroticism and priming effects for negative stimuli was discovered by Robinson et al. (2007). They tested whether neuroticism was related to affective priming and found that it was predictive for priming effects for negative, but not for positive targets, which is in line with the previous mentioned findings. That a personality trait might influence affective processing was also suggested by Augustine et al. (2013). They found that differences in scores on neuroticism and extraversion could
5 better explain differences in priming effects than the priming itself. Participants scoring
higher on neuroticism scales were more sensitive to negative facial stimuli and more
influenced by negative affect and not by positive affect, while participants with high scores on extraversion showed stronger priming effects for positive stimuli.
Frühholz et al. (2010) found that a high score on neuroticism was related to a high sensitivity to emotional conflicts during processing of negative and positive facial
expressions. Also Haas et al. (2007) found out that neuroticism was related to high emotional conflict. An explanation for the link between neuroticism and emotional conflict could be that these individuals have more trouble resolving conflict. Therefore, the affect elicited through difficulties in emotional processing could lead to a stronger negative affect which again could lead to participants with a high score on neuroticism experiencing a stronger affective priming effect.
The current study sought to find evidence for the assumption that conflict functions as a part of the effect of affective priming. Since there is not much evidence about the factors contributing to conflict, we will estimate the impact of these factors. First, we explored an external source of conflict in the visual disruption of processing due to masked stimuli, with masking as the obscuring factor of the priming effect. This was examined under conditions with LSF and BSF facial primes and also with a short and long duration of prime presentation. We also took into account the local or global focus of participants as a possibly distinguishing factor for greater priming by LSF or BSF facial primes. We also included neuroticism as an internal factor that might contribute to elicit conflict in individuals high in this trait. Getting to know more about factors that determine processing of conflict, we might get more insight into how affective priming works.
We expected to find a stronger affective priming effect in conditions in which conflict is elicited. This is the case for short presentation duration and LSF stimuli since the prime is difficult to see. We expect that in masked conditions, the priming effect will be the same along the different conditions so that there will be less priming effect compared to non-masked conditions. For individuals with a local focus, we expected more priming by BSF primes and for globally focused individuals, we expected more priming by LSF primes.
With regard to individual differences in personality, we expected individuals scoring high on neuroticism to show greater priming effects, due to greater levels of conflict, than individuals scoring high on extraversion.
6 The answers to above mentioned questions will be explored with a priming task that contains facial emotional expressions. Because affective priming is different for low spatial primes (Veuillemier et al., 2003) and for suboptimal presentation (Winkielman & Berridge, 2004), we use low-spatial filtered emotional faces and broad spatial filtered faces and a suboptimal and an optimal presentation time. The primes will consist of happy and angry faces which are followed by targetwords that are negative or positive. Participants’ reaction time to the targetwords will indicate the effect of the primes in the above described
conditions. Calculating affect indexes and congruency indexes based on the reaction times will give insight in underlying processes. A manipulation check will control for the level of conscious perception of the stimuli in the different conditions. A global-local test will be administered aswell as a personality test.
Method
Participants
Thirty-nine participants (aged 18-42 with average age 23 year; SD = 4,6 years, of which 27 women and 12 men) took part in the experiment. Participating students received 1,5 course credits. All participants were were strongly right-handed, which was scored by the Dutch Van Strien handedness questionnaire (Van Strien, 1992). Participants with a score lower than 9 were excluded from participation. Furthermore, all participants had corrected-to-normal vision and were native Dutch speakers.
Design
The experiment contained four experimental blocks and one manipulation block, each block consisted of 160 trials. Before the start of the experimental blocks, participants had to classify twenty targetwords to estimate the accuracy of the participant’s judgment. In the experimental blocks, participants had to classify a target word as positive or negative after seeing an emotional facial expression. The experimental blocks had a 2 (mask/nonmask) x2 (BSF/LSF) x2 (facial expression happy/angry) x2 (900msec/30msec) x2 (targetword
positive/negative) within-subjects design. The reaction time of correct responses to the targetwords was the dependent variable.
To control for individual differences in speed, a speed-accuracy correction was performed on the original reaction times (see Smilek, Enns, Eastwood & Merikle, 2006). For every condition, the individual reaction times were adjusted for the proportion correct
7 responses by the formula: RT/(1 - (number of incorrect responses/number of trials))= adjusted reaction time. As a result, shorter reaction times with more incorrect responses become longer while longer reaction times with no incorrect responses will not be adjusted. So, incorrect responses and reaction time are combined in the adjusted reaction times as one variable. On this single variable, the calculation of affect- and congruency indexes is based.
Calculation
The differences in reaction times were transformed into affect indexes per condition. Since people tend to react faster on positive than on negative words (Taylor, 1991), the reaction times of positive words were subtracted from the reaction times to negative words. Following the same facial prime in the same masking, duration and spatial frequency, the reaction times to the positive targetwords were subtracted from the reaction times to the negative targetwords. In other words, RT’s to the positive, congruent pair of stimuli were subtracted from the incongruent pair, resulting in a positive affect index value (=the effect of happy faces on RT). The value represents the resting value after the –as assumed quicker- RT’s to the congruent positive set are subtracted from those of the incongruent set.
The negative affective index value (the effect of the negative face prime) is the RTs of the congruent negative pair of stimuli minus the RTs of the incongruent pair of stimuli (a positive word following a negative prime, see Table 1).
Resulting positive values represent that the overall reaction for that condition was faster to positive words, whereas a negative value represents a faster reaction to a negative word.
The affect index consists of values that show the difference of congruent-incongruent positive RTs resp. the difference of congruent-incongruent negative RTs. Subtracting the negative affect values from the positive affect values, the difference serves as congruency index. The values of the congruency index show the strength of the affective priming effect. When the reaction time was faster to both congruent sets (positive-positive;
negative-negative), the congruency index gets higher. When the reaction time was higher for only one congruent set; it gets lower. When the reaction time was faster to both incongruent sets, the value gets (more) negative.
So, a higher score means a great effect of affective priming, whereas a smaller score means a small effect of affective priming. A negative score represents the strength of the opposite effect, which means a faster reaction to the targetword when the preceding faceprime had shown the opposite emotion.
8 It is important to know that there is no fixed strength of priming connected to a certain value; the values show the strength of effect with regard to the values in the other conditions which then might be higher or lower than other values.
Table 1
Calculation of reaction time into affect indexes and congruency indexes
Materials
Stimuli & Apparatus
Affective facial expression stimuli (60 angry BSF, 60 angry LSF, 60 happy BSF, 60 happy LSF) were selected from the Karolinska Directed Emotional Faces set (Lundqvist, Flykt, & Öhman, 1998). The facial stimuli were filtered and therefore contained only low spatial frequencies (LSF) with a low-pass Gaussian filter (SD = 0.015; less than 17 cycles per image width) or they remained unfiltered (BSF). Each pair of facial stimuli consisted of the same greyshaded face portraying a happy or a sad expression, both in low spatial frequency (LSF) or unfiltered frequency (normal).
The target words were selected from the EmoClar database (Phaf, Van der Leij, Stienen and Bierman, 2006). In total 80 positive and 80 negative words were selected. The targetwords (see Table 2 and Appendix 1a,1b) were matched on length (Keuleers & Brysbaert, 2010). All stimuli and targetwords were selected at random by the computer. In each block, all targetwords appeared. The facial stimuli were repeated randomly during the blocks.
Face prime Targetword RT (ms) Affect indexes (ms) Congruency indexes (ms)
Negative Positive
RTa Positive: RTa (negative word) - RTb (positive word) Affect-index for
RTb happy prime minus Negative RTc Negative: RTc (negative word) - RTd (positive word) Affect-index for
9
Priming task
The size of the facial stimuli was 562 (width) x 762 (height) in pixels and 15 (width) x 20 (height) in cm. Participants sat in front of the computer, with the use of a chinrest, on a fixed distance of 30 cm from the screen. The left button was for responding to the negative targetwords in the experimental task and the right button for positive words.
Following a preceding prime of either 30ms or 900ms, a targetword was presented on a gray background. Participants were asked to press the corresponding button to evaluate the words. Until response, the targetword stayed on the screen. In the instruction, participants were asked to respond immediately to the target. To prevent a fixed response pattern, inter-trial interval was jittered between 1000ms and 2000ms for all trials. Due to the low level of difficulty of the experimental task, participants with less than 90% accuracy were excluded from further analysis. Reaction times differing more than 2 SD (standard deviations) from the overall mean were considered as outliers and were iteratively removed.
In the masking condition, a scrambled-face mask will appear before and after the prime. The duration of the mask is 100 msec in both masking conditions. Dependent on the prime´s duration, a grey slide will appear to keep SOA constant.
Alternatively to a mask, there will be a blank grey sheet appearing before and after the prime, functioning as a mask to avoid an after image of the prime (see figures 1-4). With a prime lasting 900ms, in the mask condition a forward and a backward mask appeared for 500ms resp. 100ms. In the unmasked condition, instead of the masks, grey slides appeared for the same duration as did the masks. Following the backward mask or the grey slide, after 100ms the targetword appears and lasted until the participant responded to the targetword. With the prime lasting 30ms, in the mask condition a forward and a backward mask appeared for 500ms resp. 100ms. In the unmasked condition, instead of the masks, grey slides appeared for the same duration as did the masks. Following the backward mask or the grey slide, after 100ms the targetword appears and lasted until the participant responded to the targetword. To keep the duration between stimulus onset and target the same for both long and short prime duration, in the short-prime masked condition a grey slide lasting 870ms was used as a filler between backward mask and targetword. In the short-prime unmasked condition, a grey slide, lasting 970 ms, appeared between prime and targetword. The total duration from stimulus onset to targetword 1500 ms for all conditions (see figure 1-4).
10 Visual timeline
Figure 1. Prime 900 ms, with masks Figure 2. Prime 900 ms, no masks
Figure 3. Prime 30 ms, with masks Figure 4. Prime 30 ms, no masks
Table 2
Length of target words
Positive Negative Length (in characters) Range 4-7 4-7
11
Manipulation check
To check for the level of conscious perception, a manipulation check was included to check whether it was possible to recognize the faces in the different conditions of spatial frequency and presentation time and masking, or in other words, the level of conscious perception was assessed. The manipulation block did not differ from the experimental blocks, except for the targetwords, which were absent. Instead, participants were instructed to
evaluate the gender of the faces by pressing the left button for male and the right button for female faces. The dependent variable was detection accuracy of gender. It was checked by one-sample t-test if detection accuracy, thus the means of the conditions, differed significantly from chance level (50%).
Global-Local task
Because of differences in strategy like Kovacs (2014b) reported, a global-local focus test (Gasper & Clore 2002; Kimchi & Palmer, 1982) was administered after the manipulation check. Sixteen items were selected for the global-local task, as used by Gasper and Clore (2002). They were administered on the computer screen, with use of the chinrest at a distance of 30 cm from the screen. Participants had to make a decision which of two geometric figures (a triangle consisting of squares or a square built out of triangles) at the bottom was the most similar to the figure on the top. Participants had to respond
with either the left or the right button to choose the corresponding figure on the left resp. right side. The figure on the top
resembled both in a global shape and also in the entities it consisted of (see Appendix 2). A median split was performed to assign participants to a global respectively local focus.
Figure 5.
BIS/BAS
Since as recently illustrated by Smits & Boeck (2006), there is a link between neuroticism and extraversion with a high score on the scale of Behavioral inhibition system sensitivity (BIS) resp. behavioral activation system sensitivity (BAS). Thus, participant’s levels of neuroticism and extraversion were assessed using Carver and White’s (1994) BIS/BAS scales after the computer session. A translated version of the BIS/BAS
questionnaire was administered on paper. Participants were asked to indicate on a 4 point likert scale how strongly they agreed or disagreed with the statements. The questionnaire
12 consisted of 7 BIS and 13 BAS items (with fillers a total of 24 items). An example for an item of the BIS dimension is “I worry about making mistakes.”, an example for the BAS
dimension is “I’m always willing to try something new if I think it will be fun”. Participants were instructed to answer all items and to not skip any item, to give only one answer per item, being as honest as possible and to answer each item separately, without thinking about the items before.
The BIS/BAS scale was administered after the experimental session in order to avoid any influence of the scale on the participant’s affective and attentional state.
A median split was performed to assign participants to a more neurotic or more extravert personality.
Procedure
The participant’s handedness was administered with a handedness questionnaire (Van Strien, 1992). The experimenter explained that the effect of priming was not yet fully
established and that the aim of the experiment was to determine whether a facial expression could influence the reaction to a word. To motivate participants, they were also told that this experiment could lead to an answer to this question. After having read the information brochure, participants and experimenter signed the informed consent.
Then, participants were familiarized with the target words. The experimenter pointed to each word on a list while the participant categorized them as being positive or negative.
The experimenter asked participants to switch off mobile phones and made sure that the participant was seated comfortably. If necessary, the height of the computer screen was adjusted aswell as the height of the chinrest. Further, participants were instructed that there might appear a flash or a face but that there would always be a word and that they had to react as fast as possible as they saw the word. The experimenter emphasized that the other stimuli might help to keep focused at the middle of the screen but that it was all about the word.
Then the practice block started. When this practice block was completed with less than 10% false responses, the experimental task started. If not, the participant had to repeat the practice block. When the participant successfully accomplished the practice block, the experimental priming task started.
In the breaks between the experimental blocks, the experimenter informally talked to the participant, thereby creating motivation for the participant to put effort in the task.
13 After the experimental priming task, the manipulation check followed. The
experimenter instructed participants that speed didn’t matter but accuracy; if they didn’t see a face, they should use their intuition.
As the last task, the global-local task was administered. Participants stayed seated in the same position in front of the computer with their chin on the chinrest. The experimenter gave the instruction that speed didn’t matter but their first impression.
Afterwards, participants filled in the BIS/BAS questionnaire, which was administered on paper. The experimenter instructed participants to answer the questions as honest as possible, following their intuition.
In the exit interview, the experimenter asked participants how they perceived the task, if they used any strategies and if they recognized something with regard to the experiment. Demographic variables as gender and age and abnormalities during the experiment were registered in a logbook. In total, the experiment took about 90 minutes.
Results
Although all 46 participants didn’t exceed the limit of 10% false responses and scored on handedness not lower than 9, however seven participants were excluded from further analysis because their overall reaction times were outliers. The data of thirty-nine participants served for final analysis. The power for finding an effect of masks was 0.36. Due to the low level of power, the chance of a significant p-value is very low, even if there was an effect. A nonsignificant result is therefore not meaningful and can not be treated as part of the analysis of possible effects found in the experiment. Therefore, statistical test results will only be mentioned for the sake of completeness for certain parts. These statistical test results will not appear in other parts since it is not justifiable to mention test results when the test was taken under conditions that lead to outcomes which might not show effects even if they were there. Instead, effect sizes will be used to illustrate the possible presence of an effect.
14
Manipulation check
The manipulation check served as control for the level of awareness in the different conditions. The accuracy and standard deviations of the gender decision task are displayed in table 3.
Table 3
Mean accuracy and SD in gender decision task in percentages LSF BSF
Happy Angry Happy Angry
Mask 30ms 52 (18) 44** (19) 50 (15) 50 (17) Mask 900ms 95* (7) 85* (16) 96* (7) 94* (8)
No Mask 30ms 86* (11) 78* (13) 91* (9) 86* (13) No Mask 900ms 95*(7) 88* (11) 96* (6) 94* (14)
*. Significant at the .01 level (2-tailed). **. Significant at the .05 level (2-tailed).
It can be seen that with short, masked primes, the detection accuracy was lower than in longer or unmasked trials. There was not much difference in recognition with regard to the spatial frequency for happy faces. However, the detection was less accurate for angry filtered faces than for angry unfiltered faces. Overall, the accuracy differs from 50% in all optimal
conditions significantly except for conditions with masking, short prime and LSF. Interestingly, the combination of these conditions with an angry prime also significantly differed from 50%, t(38) = -2.16, p < .05. This means that detection accuracy in this condition was below chance; the masked short angry LSF face was difficult to recognize.
15
Experimental task
Data – Reaction times (RT)
Of 46 participants, seven were excluded of the analysis due to a mean reaction time that was considered as an outlier. With regard to accuracy, the average percentage of correct responses was 97.4%.
The means and standard deviations of each experimental condition are given below in Table 4.
Table 4
Overview of means and SD of reaction time of each condition in ms LSF BSF
Happy Angry Happy Angry
Positive Negative Positive Negative Positive Negative Positive Negative Mask 30ms 643 (100) 614 (93) 643 (98) 624 (85) 647 (97) 620 (90) 636 (94) 613 (73) Mask 900ms 682 (77) 657 (91) 692 (97) 643 (67) 690 (99) 642 (72) 674 (84) 659 (100)
No Mask 30ms 662 (102) 638 (95) 657 (87) 629 (91) 649 (87) 613 (95) 663 (87) 623 (84) No Mask 900ms 669 (101) 641 (89) 697 (89) 643 (99) 668 (95) 634 (87) 667 (94) 640 (92)
Contrary to the standard valence effect (Taylor, 1991), participants reacted significantly faster to negative targetwords (M = 633ms, SE = 11.5) than to positive
targetwords (M = 665ms, SE = 12.3), r = .64, [t(38) = -5.08, p < .005]. The effect size shows that this effect is large and so represents a substantive finding. This might be explained by the finding in the exit interviews that participants used a button apparatus for left-handed
individuals, which means that the button for the middle finger was actually at the distance for the index finger, and the button for the index finger was at the distance for the middle finger. As a result, many participants had to cant the button box. Still, some participants felt that clicking the button for a positive answer went less quickly and that clicking the left button was more comfortable. Also, participants reported that they were very used to often use the index finger (e.g. computer mouse) for confirmation and not for negative actions. This can have led to quicker answers for a negative word and slower answers for a positive word.
16
Data – congruency indexes
As displayed in table 5, the indexes for masked conditions are predominantly negative, whereas the indexes for the nonmasked conditions are predominantly positive. It is striking that the highest value is displayed for a LSF, long lasting prime for both masked and
unmasked conditions. A big negative value is displayed for the long BSF masked condition, whereas the value is nearly zero for the long BSF unmasked condition. This leads to the assumption that the combination of optimal priming and a mask leads to incongruent priming.
Table 5
Means and standard deviations for congruency indexes for presentation duration Mask NoMask
Short LSF -10 (128) 4 (90) Short BSF -4 (112) 4 (77) Long LSF 24 (92) 26 (119) Long BSF -33 (128) -7 (113)
The graphical illustration (figure 6) shows clearly the direction of the effects, independent of numerical scores. As expected, there was in general a smaller effect of affective priming for masked stimuli than for nonmasked stimuli, which is in line with the expectation. However, the differences are small.
Figure 6. Congruency indexes for presentation duration with and without masked prime, subdivided in LSF and BSF.
17 LSF: The effect of priming is greater for 900ms than for 30ms. This is the case for both conditions Mask and NoMask. In the short masked condition, which is the suboptimal condition, the effect is smaller for masked stimuli than for nonmasked stimuli. So for LSF, optimal presentation leads to priming for both masked and nonmasked conditions.
BSF: The effect is greater for 30ms than for 900ms for both Mask and NoMask conditions. For long presentation duration, which is the most optimal condition, the effect is smaller for the Mask condition. So, for BSF, the suboptimal presentation leads to priming, whereas the optimal condition leads to incongruent priming; people reacted faster to the targetword when primed with a prime of the opposite emotion. This incongruent effect in the BSF conditions is enhanced with masked stimuli.
Furthermore, short presentation duration, a masked stimulus and filtered stimuli are suboptimal conditions for perception, which can consequently lead to suboptimal priming. Optimal conditions (long presentation, no mask, unfiltered stimuli) can lead to an optimal effect of priming. The latter was the case for the LSF condition. In the BSF condition, there was a suboptimal effect of priming, but this might rather appear because of the great
incongruent effect than that there was actually suboptimal priming.
For the suboptimal condition (LSF, short prime, masked) and the most optimal condition (BSF, long prime, unmasked), the least affective priming took place. Masks enhanced the incongruent priming effect in these conditions.
When participants were assigned to either a more global or a more local focus, globally and locally focused individuals show no clear differences with regard to spatial frequency (figure 7). For both, the priming effect is stronger for long LSF primes. This finding is not in line with the expectation that only globally focused individuals would show more priming in LSF conditions.
18 Figure 7. Congruency indexes for presentation duration for Mask and NoMask conditions, subdivided in global and local focus for LSF and BSF.
Globally focused participants showed more affective priming in LSF conditions than in BSF conditions, which is in line with the expectation. In LSF, an affective priming effect occurred in all conditions except for the suboptimal condition (masked, short, LSF). The BSF condition led to less affective priming; only for the short, unmasked BSF condition, there was affective priming. In the conditions in which the BSF prime was long lasting, an incongruent priming effect occurred; this means that when the BSF prime was long time visible, globally focused participants reacted faster to a targetword that had the opposite valence of the BSF prime. For both LSF and BSF, the suboptimal condition led to no priming.
Locally focused individuals showed in the LSF condition an affective priming effect for long presentation of the stimuli but not for short presentation times, showing an optimal priming effect. Only for masked BSF primes, the pattern was reversed – masked primes led to an incongruent priming effect when long presented, whereas long presented unmasked stimuli led to congruent affective priming.
An interaction effect could be observed for the nonmasked BSF condition, where locally focused individuals showed congruent affective priming for long duration and globally focused individuals showed congruent affective priming for short duration.
19
Personality
For the analysis of possible influences on affective priming due to personality, participants were assigned to a more neurotic or a more extravert personality with a median split.
The reliability for the scale was α = .69 for the neuroticism (BIS) scale and α = .76 for the extraversion (BAS) scale. These reliabilities are not optimal, but suggest sufficient
reliability to consider the scores in the analysis.
As is shown in table 6, participants scoring high on neuroticism showed more priming in unmasked and optimal conditions, whereas participants scoring high on extraversion only showed affective priming in the suboptimal condition. Both findings are contrary to our expectation that participants scoring high on neuroticism would show more affective priming than participants scoring high on extraversion.
Table 6
Means and standard deviations congruency index for participants high in neuroticism (N=15) and extraversion (N=19)
Mask NoMask 30ms 900ms 30ms 900ms Neuroticism -6.5 (102) -16.5 (134) 8.5 (108) 20.5 (141) Extraversion 17.5 (124) -19.5 (108) -3 (154) -2.5 (98)
As can be seen in figure 8, participants high scoring on neuroticism showed affective priming in unmasked conditions, with the most affective priming in the long presentation duration, showing an optimal priming effect in nonmasked conditions. Participants scoring high on extraversion showed the opposite; the most affective priming occurred in the masked, short condition. Also, high-scorers on extraversion show incongruent priming for the long masked condition thereby showing a suboptimal effect of affective priming in masked conditions.
20 Figure 8. Congruency index for participants high in neuroticism (BIS) and extraversion (BAS).
When split up into the two different spatial frequencies, LSF and BSF, there was no clear pattern. However, for participants scoring high on neuroticism, LSF primes elicited more affective priming than BSF. The effect was even greater in non-masked and longer lasting conditions whereas BSF primes lead to incongruent priming. So people scoring high on neuroticism showed the most priming when the prime was LSF, but in optimal conditions.
Participants scoring high on extraversion showed a suboptimal affective priming effect for BSF stimuli and an optimal affective priming effect for LSF stimuli. The suboptimal effect for BSF stimuli was enhanced in the masked condition whereas the optimal effect for LSF stimuli was enhanced in the nonmasked condition. In other words, long presentation times of BSF resp. LSF stimuli led to a more incongruent resp. a more congruent priming effect.
Table 7
Means and standard deviations congruency index for participants low in neuroticism (N=24)
Means and standard deviations congruency index for participants high in neuroticism (N=15)
Mask NoMask
Low High Low High
Short LSF -25 (141) 14 (104) -16 (61) 36 (119)
Short BSF 10 (116) -27 (103) 19 (60) -19 (96)
Long LSF 20 (79) 30 (114) 7 (79) 57 (162)
21 Table 8
Congruency index for participants low in extraversion (N=20)
Congruency index for participants high in extraversion (N=19)
Mask NoMask
Low High Low High
Short LSF -22(145) 3 (110) 29 (87) -21 (88)
Short BSF -38(66) 32 (138) -6(87) 15 (66)
Long LSF 33(113) 13 (66) 30(135) 22 (102)
Long BSF -16(104) -52 (150) 12(129) -27 (94)
For both groups, as for the total congruency index, long duration of a LSF prime had the biggest affective priming effect. As it is not much difference for both masked and unmasked conditions in the general index, a difference becomes visible when the group is split up into people scoring high or low on either neuroticism or extraversion. For high scorers on neuroticism, affective priming is greater and enhanced in unmasked conditions. For high scorers on extraversion, no distinctive pattern was visible.
Figure 9. High scorers on neuroticism (BIS) and extraversion (BAS) on spatial frequency, duration and masking.
22
Data – Affect indexes
The experimental reaction times were transformed into affect indexes (table 9). In these indexes, the amount of positive and negative affect is shown. The valence of the prime does not matter since it is just about conflict. The expectation was that there would be more negative affect in masked conditions than in nonmasked conditions which would lead to an overall negative affect, overshadowing the effect of affective priming. Short duration and LSF primes were expected to elicit more conflict.
Table 9
Overview of general means (SD) of affect indexes for each condition in ms LSF BSF
Mask 30ms -24 (89) -25 (85) Mask 900ms -37 (76) -32 (85)
No Mask 30ms -27 (76) -38 (80) No Mask 900ms -41 (76) -31 (82)
Figure 10. Affect indexes for LSF and BSF between the conditions Mask and NoMask
In table 9 and figure 10, it is noticeable that each of the affect indexes is negative. This means that participants reacted faster to negative targetwords that to positive targetwords, which is against the expectation that the reaction times to positive stimuli are faster. An explanation is the button box, which is already mentioned in the analysis of the exit
23 interviews. There are nearly no differences in affect between the Mask and NoMask
condition. The least negative affect is shown in the short conditions, with an exception for the unmasked BSF condition. This is contrary to the expectation that in suboptimal conditions, more negative affect would be elicited due to the greater difficulty of processing.
Data – affect indexes and personality
Furthermore, the affect indexes were examined to check whether participants scoring high on neuroticism showed more negative affect, since it was expected that negative affect in high-scorers on neuroticism would lead to more affective priming.
Figure 11. BIS affect Figure 12. BAS affect
In contrast to extraverted participants, those scoring high on neuroticism show an inverse pattern with regard to the LSF condition. Masked long duration elicits positive affect, whereas unmasked long duration elicits negative affect. This is contrary to the expectation of high scorers on neuroticism to show stronger negative conflict compared to extraverts. In the BSF condition, no positive affect is shown.
24 Table 10
High BIS (N=15) Overview of means (SD) of affect indexes for each condition in ms LSF BSF Mask 30ms -26 (85) -23 (92) Mask 900ms 6 (73) -28 (103) No Mask 30ms -20 (80) -29 (94) No Mask 900ms -43 (92) -22 (76) Table 11
High BAS (N=19) Overview of means (SD) of affect indexes for each condition in ms LSF BSF
Mask 30ms -43 (77) -28 (88) Mask 900ms -53 (62) -32 (84)
No Mask 30ms -55 (74) -58 (63) No Mask 900ms -49 (74) -40 (77)
Participants scoring high in extraversion showed overall negative affect. Expecting that participants scoring high on neuroticism would show more negative affect and that participants scoring high on extraversion should show an opposite pattern, this finding is contrary to the expectation. In the masked condition, for LSF faces, the overall affect is more negative than for BSF faces.
Discussion
The current study sought to find evidence for the assumption that conflict functions as a part of the effect of affective priming. First, we explored an external source of conflict in the visual disruption of processing due to masked stimuli, with masking as the obscuring factor of the priming effect. We also took into account the individual focus of participants as a possibly distinguishing factor for LSF or BSF facial expressions. Second, we also included
25 exploratively neuroticism as an internal factor that might contribute to elicit conflict in
individuals high in this trait. The results were mixed with generally more optimal than suboptimal priming, masked and nonmasked conditions revealing small differences and LSF eliciting more priming that BSF, but unexpectedly not clearly related to focus. Splitting participants into high scorers on neuroticism or extraversion, some patterns emerged for neuroticism. Also, the Affective Monitoring hypothesis could not be proved wrong and therefore conflict as a contributing factor can not be rejected.
Compared to nonmasked conditions, we expected that in masked conditions the priming effect would be less. This was true for masked stimuli eliciting slightly less priming than nonmasked stimuli. However, this difference was small. Furthermore, we expected to find a stronger affective priming effect in conditions in which conflict is elicited. This would be the case for masked, short presentation duration and LSF stimuli since the prime is difficult to see. If the in the suboptimal condition priming was found, it would have been clear that conflict is necessary for affective priming. We did not find priming effects for suboptimal conditions but for optimal conditions; in the suboptimal condition, no affective priming was found. Also in the most optimal condition, no affective priming took place which is in line with our expectations that conflict previously must be elicited in order to lead to affect. It seemed that when consciously perceived, participants tended to react against the by the prime intended effect. This is in line with LeDoux (1996), who argues that with conscious
processing, the initial reaction to the prime can be stopped and altered. However, in our experiment, reaction times served as dependent variable and it is not clear, in which way the reaction of participants might have been altered. A possible explanation are the facial stimuli, which participants experienced as exaggerated; the exit interviews revealed that several participants laughed because of the faces. It is salient that the BSF condition led to less priming than the LSF condition. Therefore, a possibility for further research is the use of pictures that show situations, like fire or physical attack, to elicit an emotional response.
It were the intermediate conditions leading to priming; when a prime was optimally presented without masks, we did find more affective priming for LSF than for BSF.
This leads to the suggestion that conflict, due to low spatial frequency, was necessary to elicit affect and therefore leading to affective priming, which is also in line with the Affective Monitoring hypothesis.
Considering the affect indexes, in the suboptimal condition and in most masked conditions, less negative affect was found, contrary to our expectation. This could be
26 explained by the fact that the suboptimal condition was so difficult to perceive that even non-conscious processing was not possible. It seems that the invisibility of the prime did not allow for processing, which therefore could not be distorted through conflict the therefore no affect was elicited. Interestingly, Andrews (2011) also found no effect in masked conditions and argued that in studies where the effect was found, masking was methodologically not stringent applied. The manipulation check, which served as a measurement of conscious perception, might have not given a good indication of the differences between conditions. The instructions, being the opposite from the instructions of the experimental task, might have led to different processing. Also, the evaluation of gender might be different than the evaluation of emotions, since gender is more neutral and might not depend that much on participant’s mood. Yet the processing is essential for the effect. Indeed, Deruelle and Fagot (2004) suggest independent processing routes for functionally different types of information such as emotion and gender (Deruelle & Fagot, 2004). Therefore, the manipulation check might not have given a reliable estimate of the amount of perception of the primes in the different conditions. This is interesting for future research since it was salient that the suboptimal prime did not lead to affective priming in most conditions.
Contrary to our expectations, for both globally and locally focused participants, there was not a clear distinction in priming effects. However, again LSF primes led to more affective priming than BSF primes. However, as the median split shows, nearly half of the participants were assigned to either a more local focus, which contradicts with Kovacs
(2014b), in which the majority of participants had a global focus. The difference is that in this study, participants were forced to sit close (30cm) to the screen while in Kocavs’ experiment, the test was administered on paper. It remains unclear if the effect found by Kovacs, that locally and globally focused individuals differ in terms of spatial frequency preferences, is valid, since the distance between participants and the paper was not fixed. But also in the current study, the near distance to the screen could have influenced the actual perception and therefore the focus of the participants. Huntsinger et al. (2010) found out that a positive mood led to an enhancement of a previously primed focus. Assumed that the close distance to the screen worked as a prime, leading to a more local view, a positive mood then would have increased a local focus, whereas a negative mood would have weakened a local focus. Therefore, participants in a positive mood would have experienced a greater local focus and participants in a negative mood would have experienced a less local focus. Since it was the end of a long experiment, it is not clear in which mood the participants did the global-local
27 task. In future research, this methodological implication should be taken into account to get a more reliable measure of foci.
With regard to individual differences in personality, we expected individuals scoring high on neuroticism to show greater priming effects, because they would experience greater levels of conflict than individuals scoring high on extraversion. The explorative assessment of personality and priming effects showed mixed results. For both, participants scoring high on extraversion and participants scoring high on neuroticism, long duration of a LSF prime had the biggest affective priming effect. Interestingly, for high scorers on neuroticism, affective priming was greater and enhanced in unmasked LSF conditions. For individuals scoring high on extraversion, affective priming only occurred in the suboptimal condition. This could be explained by the fact that that high scorers on neuroticism, feeling more conflict due to the masks, would show less priming in those conditions. Individuals scoring high on extraversion, feeling less overall conflict, would show more priming in masked conditions, which is not the case though. Taking the affect indexes into account, it is visible that high scorers on
neuroticism scored in the opposite way in LSF conditions than extraverts, with masking in the long LSF condition leading to more positive affect for high scorers on neuroticism and
negative affect for high scorers on extraversion, contrary to the expectation. Taken together, high scorers on neuroticism experience the most priming with long LSF primes, with masking a little less. They also experience positive and negative affect for masked resp. nonmasked stimuli in these conditions. This finding suggests that indeed affect is necessary for affective priming, with masks playing a role in low spatial frequencies. Since there was not a clear pattern for BSF or for high scorers on extraversion, the link between personality and affect should be further explored.
This experiment had several methodological shortcomings. These can explain why some effects did not come out as clearly as expected and can deliver important knowledge on how to improve the experiment. Below, some suggestions will be discussed for future
research.
First of all, the button box for left handed individuals, while strongly right handed individuals participated, was a heavy distortion of reaction times and led to errors due to automatic use of the index finger. Also, the button box in itself, due to the unnatural position of the fingers on the left-handed box, could have led to negative affect during the whole experiment. Also the scrambled faces could have led to conflict as being not perceived as a
28 whole face, but still some pieces. In the exit interviews, participants reported in the exit
interviews seeing parts of faces, which suggests that the content of the mask, rather than the appearance of a mask, might have led to a distortion in processing.
Also, the experiment took so long that participants reported becoming easily tired. Since fatigue and an exhausting task easily lead to conflict, in future studies, the experimental task could be distributed on subsequent days to keep the level of attention and therefore the processing of stimuli on a stable level during the experiment.
In the exit interviews, some participants reported that targetwords, such as “lightning”, were ambiguous. Since before starting the experiment a selection was made, this selection might need to be stricter. Also, many words were not contemporary anymore. For this reason, the frequency of the words was also not taken into account. Using a contemporary list of words could help to use words matched on frequency and being less ambiguous.
The standard deviations were quite big in this study, which implies imprecise
measurements. The affect elicited in some participants by the exaggerated faceprimes might have contributed to these differing reaction times. But also the button box might have led to a distortion. Also the reliability of the scales for measuring neuroticism and extraversion could have been higher. For future research, it can be recommended to ask participants to fill in the questionnaire in the beginning, because the exhausting experiment might have contributed that participants were less motivated to read the questions properly.
In this experiment, the research questions could not be completely answered. Instead, it becomes even clearer that the effect of affective priming is more complex and needs further investigation. The Affective Monitoring hypothesis can not be rejected, however. Since there was no priming effect for the most optimal conditions, it still can be assumed that conflict is necessary to elicit affective priming. This is especially illustrated by the fact that the LSF primes led to more affective priming than BSF primes. Interestingly, personality seems to influence affect, resulting in different strengths of priming for high scorers on extraversion and for high scorers on neuroticism, which again was salient in the LSF condition. This, together with the above mentioned implications, can serve as an important basis for future research, being another important step into the direction of finding the factors contributing to affective priming.
29 Literature:
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(1), 33–43.
Augustine, A. A., Larsen, R.J., & Elliot, A.J. (2013). Affect is greater than, not equal to, condition: Condition and person effects in affective priming paradigms. Journal of Personality 81(4), 355-364.
Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and affective responses to impending reward and punishment: The BIS/BAS Scales. Journal of Personality and Social Psychology 67(2), 318–333.
Dannlowski, U., Kersting, A., Donges, U.–S., Lalee–Mentzel, J., Arolt, V. & Suslow, T. (2006). Masked facial affect priming is associated with therapy response in clinical depression. European Archives of Psychiatry and Clinical Neuroscience 256(4), 215-221.
Dannlowski, U., Ohrmann, P., Bauer, J., Kugel, H., Arolt, V., Heindel, W., & Suslow, T. (2007). Amygdala reactivity predicts automatic negative evaluations for facial emotions. Psychiatry research 154(1), 13-20.
Deruelle, C. & Fagot, J. (2005). Categorizing facial identities, emotions, and genders:
Attention to high- and low-spatial frequencies by children and adults. Journal of Experimental Child Psychology 90(2), 172-184.
30 Evans, I. M. (2010). Positive affective priming: A behavioral technique to facilitate
therapeutic engagement by families, caregivers, and teachers. Child & Family Behavior Therapy 32(4), 257-271.
Frühholz S., Prinz, M., & Herrmann, M. (2010). Affect-related personality traits and contextual interference processing during perception of facial affect. Neuroscience Letters 469(2), 260–264.
Gasper, K. & Clore, G.L. (2002). Attending to the big picture: Mood and global versus local processing of visual information. Psychological Science 13(1), 34-40.
Haas, B.W., Omura K, Constable, R.T., & Canli T. (2007). Emotional conflict and neuroticism: Personality-dependent activation in the amygdala and subgenual anterior cingulate. Behavioral Neuroscience 121(2), 249–256.
Huntsinger, J. R., Clore, G. L.& Bar-Anan, Y. (2010). Mood and global–local focus: Priming a local focus reverses the link between mood and global–local processing. Emotion 10(5), 722-726.
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(4), 521–535.
31 Kovacs, E. (2014a). Conflict precedes affective priming. Unpublished masterthesis. Brain & Cognition, University of Amsterdam.
Kovacs, E. (2014b). Conflict precedes affective priming. Unpublished Internship report. Brain & Cognition, University of Amsterdam.
LeDoux, J.E. (1996). The emotional brain: The mysterious underpinnings of emotional life. New York: Simon & Schuster.
Liddell, J.B., Williams, L.M., Rathjen, J., Shevrin, H., & Gordon, E. (2004). A Temporal Dissociation of Subliminal versus Supraliminal Fear Perception: An Event-related Potential Study. Journal of Cognitive Neuroscience 16(3), 479–486.
Lundqvist, D., Flykt, A., & Öhman, A. (1998). The Karolinska Directed Emotional Faces. Psychology section, Department of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden.
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(5), 723-739.
Phaf, R.H. & Rotteveel, M. (2012). Affective monitoring: A generic mechanism for affect elicitation. Frontiers in Psychology 3(47), 1-17.
32 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.
Robinson, M.D., Ode, S., Moeller, S.K., & Goetz, P.W. (2007). Neuroticism and affective priming: Evidence for a neuroticism-linked negative schema. Personality and Individual Differences 42(7), 1221-1231.
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(5), 630–640.
Smilek, D., Enns, J.T., Eastwood, J.D., and Merikle, P.M. (2006) Relax! Cognitive strategy influences visual search. Visual Cognition 14, 543-564.
Smits, D.J.M., & Boeck, P.D. (2006). From BIS/BAS to the Big Five. European Journal of Personality 20(4), 255–270.
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(2), 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,
33 624-631.
Winkielman, P., & Berridge, K.C. (2004). Unconscious emotion. Current Directions in Psychological Science 13(3), 120-123.
34
Appendix 1a
Positive targetwords matched on length
4 Zacht Zuiver
Knus Jeugd Glorie
Pret Kunst Nuttig
Zoen Zomer Handig 7
Vers Vrede Zoenen Kleurig
Wijs Sterk Passie Jeugdig
Lach Feest Mazzel Hoopvol
Wens Droom Moedig Verrukt
Baby Trots Natuur Melodie
Fijn Nieuw Cadeau Gunstig
Vrij Geluk Gezond Geestig
5 Prima Geduld Elegant
Aaien Samen Wonder Stralen
Party 6 Schoon Welzijn
Fraai Geinig Gelukt Ambitie
Hoera Lollig Zingen Rijkdom
Zalig Komiek Dansen Oprecht
Humor Vredig Succes Freugde
Bruid Jarige Muziek Winnaar
Winst Warmte Winnen Prettig
Aardig Vrolijk
Liefde Energie
35
Appendix 1b
Negative targetwords matched on length
4 Virus Somber
Wond Vrees Engerd
Coma Alarm Crisis
Vies Straf Stress 7
Vals Wraak Triest Bliksem
Zwak Brand Klagen Mislukt
Boos Ruzie Kanker Schamen
Ziek Angst Smerig Aanslag
Haat Slaan Zielig Dwingen
Fout Wapen Lelijk Verraad
Pijn Moord Lijden Overval
5 Kwaad Ziekte Schande
Tumor Doden Paniek Leugens
Plaag 6 Geweld Ellende
Satan Zeiken Gemeen Eenzaam
Boete Rouwen Aanval Twijfel
Schok Giftig Huilen Jaloers
Wreed Wurgen Geweer Misdaad
Vloek Akelig Gewond Moorden
Woede Rotten Vijand Gevecht
Gevaar Verlies
Schuld Sterven
36
Appendix 2
