Exciting Games
Sensation seeking and the
evaluation of novelty
February 2017
Student: M. Lilienblum
Student number: 5814162
Masterthesis Psychology
Specialisation: Brain and Cognition
Faculty of Social and Behavioural Sciences
University of Amsterdam
Supervisor: Dr. R.H. Phaf
2
Table of contents
Table of contents ... 2
Abstract... 3
Introduction...4
Method
Participants...14
Design…...14
Material & Apparatus...16
Procedure...23
Results ...24
Discussion...50
References...54
Appendix A...58
Appendix B...63
3 Abstract
The Affective Monitoring hypothesis views novelty as a conflict between stimulus and memory representations, which should automatically elicit negative affect. Conversely, sensation seeking may correspond to strong exploratory tendencies that match novel
stimulation, and thus may elicit positive affect. The research question of the present research was whether high sensation seekers deviate from the general principle that novelty has an intrinsic negative valence. The conflicting expectations in the Iowa Gambling Task (IGT) resemble the conflict elicited by novelty. On the basis of percentile scores, the unselected participant group was divided into high, medium, and low sensation seekers. Implicit affect elicited by the card choice in the IGT was measured by an affective priming task.
Interestingly, only high sensation seekers developed a lower choice score for risky decks than for safe decks during the IGT. Implicit affective responses did not correspond to this
development, which suggests that choice behavior in the IGT is not affectively marked. If, however, affect plays a role in decision behaviour, these results indicate that for high sensation seekers, novelty is intrinsically negative.
4
Exciting Games
Sensation seeking and the evaluation of novelty
Imagine dining at a restaurant, free to choose between your usual dish and a new one – would you go for the novel experience or stick to what has already been tried and tested? In former times, a choice for novel food might have determined your existence when the food happened to be poisonous. From an evolutionary viewpoint, a preference of familiar and avoidance of novel experiences might have been beneficial for survival (Bornstein, 1989). This is
consistent with the Affective Monitoring hypothesis (Phaf & Rotteveel, 2012), arguing that familiarity initially elicits positive affect. The Affective Monitoring hypothesis explains the elicitation of affect in terms of the dynamics of conflict resolution. Neural competition may be a fundamental process in many different functions (e.g., attention, see Desimone & Duncan, 1995). Novelty, for instance, can be envisaged as the competitive conflict between a stimulus representation and available memory representations (Phaf & Rotteveel, 2012).
According to affective monitoring, whether the resolution of initial conflict happens quickly or not is instrumental in the elicitation of positive or negative affect, respectively. With regard to the causation of positive affect by familiarity, the Affective Monitoring
hypothesis (Phaf & Rotteveel, 2012) suggests that a familiar stimulus is already represented in memory, which leads to the swift resolution of conflict and thus makes processing fluent. In contrast, a new stimulus cannot be processed easily. The resulting mismatch evokes
unresolved or slowly resolved conflict, eliciting negative affect (Phaf & Rotteveel, 2012). Therefore, novelty is generally expected to elicit intrinsically negative affect. Noordewier, Topolinski and Van Dijk (2016) described a similar mechanism in a dual-process view for surprise, which may also be characterized by initial conflict. According to this mechanism, surprise is intrinsically negative due to unexpectedness and interruption (Noordewier et al., 2016). In a subsequent step, the unexpected event can be interpreted (i.e., cognitively
5 mastered) and, depending on the stimulus, positively evaluated (Noordewier et al., 2016). Another, somewhat similar causation of negative affect (e.g., anxiety) by novelty has previously been postulated by Gray (1978; 1990). He argued that a novel stimulus would activate the behavioural inhibition system (BIS), which corresponds to an individual’s state of anxiety. In the activated BIS, novelty would be signalled as a potential threat by an approach-avoidance conflict, leading to behavioural inhibition, approach-avoidance reactions and negative affect (Gray & McNaughton, 2000). Individuals with BIS sensitivity (i.e., high BIS) would be especially prone to negative affect elicited by novelty (Gray 1978; 1990; Gray & McNaughton, 2000).
From a different perspective, novelty might actually be associated with positive affect. As was already proposed by Stang (1975), the processing of novel stimuli can sometimes be rewarding and also lead to an approach towards these stimuli, whereas familiar (i.e., boring) stimuli would be avoided. Familiar stimuli are, however, clearly preferred above novel stimuli after a single or a few presentations, as can be deduced from the classical mere exposure effect (Bornstein, 1989). To explain this apparent inconsistency, Bornstein, Kale, and Cornell (1990) suggested that enduring familiarity would lead to negative affect due to boredom, whereas novel stimuli could break the routine. Individual differences could modulate the relation between novelty and affect. Zheng Xu, Jin, Sheng, Ma, Zhang, and Shen (2010) observed that the appraisal of novelty was positive and more pronounced in high sensation seekers than in low sensation seekers. Sensation seeking is defined by Zuckerman (1994) as ‘‘the seeking of varied, novel, complex, and intense sensations and experiences, and the willingness to take physical, social, legal, and financial risks for the sake of such experience’’ (Zuckerman, 1994, p. 27). A vivid example of a sensation seeker is Flaviu Cernescu, riding a unicycle on the top of a 256m chimney without safety equipment. When asked if he is afraid to die, he admits his fear, but explains that he wants to fulfill his dreams and experience life, which makes him perform novel, dangerous stunts (Cernescu, 2016; Steve, 2016).
6 An explanation for the positive-negative discrepancy of novelty could be based on affective monitoring and an explorative behavioural SEEKING system (Panksepp, 1998), which are regarded as basic systems operating in parallel (Phaf & Rotteveel, 2012). In the case of individuals with a strongly active SEEKING system, their tendency to explore the environment (see Zuckerman, 1994) could conflict with the repeated confrontation by familiar stimuli, which should elicit negative affect (Phaf & Rotteveel, 2012). This conflict may, however, be quickly resolved by the presentation of novel stimuli, matching the explorative tendency (Phaf & Rotteveel, 2012). Nieuwenhuizen (2015) linked this mechanism to affect elicitation in sensation seekers. Although novelty remains intrinsically negative, also for high sensation seekers, the prolonged experience of familiarity elicits boredom and conflicts with prevailing action tendencies, so that eventually novelty solves this conflict, resulting in positive affect (Nieuwenhuizen, 2015). Conversely, novelty might conflict with low sensation seekers’ risk avoidance (Zuckerman, 1994), resulting in negative affect.
The positive-negative paradox in the elicitation of affect by novelty leads to the research question whether high sensation seekers deviate from the general principle that novelty has an intrinsic negative valence, and if so, to what degree. In addition, does BIS sensitivity strengthen the elicitation of negative affect by novelty?
To investigate this issue, the Iowa Gambling Task (IGT: Bechara, Damasio, Damasio, & Anderson, 1994) was combined with affective priming. The IGT has not before been
combined with affective priming and may in this combination provide a suitable measurement of the affect elicited by different conditions of risk and novelty.
The IGT was developed by Bechara et al. (1994) for investigating the influence of non-conscious and conscious emotions on decision-making under ambiguity and risk (e.g., conflicting expectations). Cards with gains and losses are picked from four decks of which two are risky and two are safe. Risky decks offer high gains but also high losses and are clearly disadvantageous in the long run. Safe decks contain low gains but also low losses and
7 are advantageous in the long run. Participants do not know these rules beforehand but
discover bit by bit the properties of each deck while playing, developing preferences for decks.
According to affective monitoring, the level of conflict in the IGT (i.e.,ambiguity / risk) resembles the conflict elicited by novelty (e.g., conflicting expectations), and should thus be suited for investigating whether novelty primarily elicits negative or positive affect. Safe and risky decks provided different conditions of novelty. At the start, both decks would elicit conflicting expectations of gains and losses. Initial immediate bigger gain in the risky decks could solve this conflict more strongly than small gains in the safe decks, so that the former would raise more positive affect than the latter. Because the risky decks contained also big losses throughout the task, participants would be confronted with highly variable outcomes over time. Due to the high level of variation (i.e., big gains and big losses), conflicting
expectations would remain for risky decks. This would eventually lead to unresolved conflict, eliciting negative affect. Less risky and less variable outcomes in the safe decks would lead to more conflict resolution in the safe decks, reversing the implicit preference. Eventual
avoidance would indicate a negative evaluation of risky decks and therefore a negative
evaluation of the conflict elicited by conflicting expectations, which is assumed to correspond with novelty.
Even before participants could explicitly formulate the different properties of the decks, a shift in card selection from risky to safe decks was observed during the IGT
performance (Bechara et al., 1994). Experiments with the IGT (Bechara, Tranel, Damasio, & Damasio, 1996) indicated a development of implicit expectations for risky and safe decks, measured by anticipatory Galvanic Skin Response (GSR; see Venables & Christie, 1975). The GSR was more pronounced for risky decks even if participants had not yet explicit knowledge of the risk of the deck and did not yet change their choice behaviour (Bechara et al., 1997). The underlying mechanism was explained by Bechara et al. (1997) referring to somatic
8 markers (Damasio, 1994), which were non-consciously signalling risky options that had led to bad outcomes in the past, guiding behaviour towards safe options. It is however unclear if this underlying mechanism was related to implicit negative affect because GSR is primarily sensitive for arousal, which may be both negatively and positively valenced (Bradley,
Codispoti, Cuthbert, Sabatinelli & Lang, 2001a, b). In order to differentiate between positive and negative affect, in our experiment, the GSR measurement in the IGT was replaced by more valence dependent affective priming. Affective priming happens when a preceding prime has the same valence as the target, which facilitates the processing of the latter (see Fazio, Sanbonmatsu, Powell, & Kardes, 1986). During the IGT, an implicit learning task, participants would build up expectations for the card decks during gambling. Subsequently, during an affective priming task, participants would react as quickly as possible to facial stimuli that appeared on the decks before the card value was shown. The preceding
experiences would lead to expectations which would influence the reaction time for safe and risky decks. Relatively faster responses to positive stimuli and slower responses to negative stimuli reflect a positive valence of affect, whereas the reverse pattern corresponds to negative affect (Fazio et al., 1986). Because affective priming does not depend on subjective reports, it is an implicit measurement of affect. Implicit affect is less likely to be subject to conscious constructions and therefore presumably more reliable than choice behaviour or explicit preference. Under the assumption that implicit affect is steering choice behaviour in the IGT (Bechara et al., 1997), a similar shift in implicit affect as in choice behaviour would imply that novelty is intrinsically negative.
The IGT provided a paradigm with more and less exploratory, novel, conditions and was expected to match high sensation seekers’ highly active exploratory behaviour system. Following their explorative tendency, they would feel more attracted to explore and choose more cards from risky than from safe decks. Presumably more sensitive to conflict due to boredom, the high level of novelty in the risky decks would solve the conflict in high
9 sensation seekers more strongly, eliciting positive affect. Because risky decks would fit
sensation seekers’ tendency to explore more than safe decks, they should remain positive longer for the high sensation seekers than for the low sensation seekers. During the game, increasing familiarity with the properties of all decks and a decrease in novelty over time would eventually lead to boredom in high sensation seekers. Once their exploratory behaviour system would be no longer satisfied by the game, the conflict with their exploratory
tendencies would remain unresolved. This would result in negative affect also for the risky decks. In the end, the higher level of conflicting expectations for the risky than for safe decks would remain, resulting in negative affect.
Inversely, continuous gambling (i.e. non-familiarity, ambiguity) and confrontation with novelty may not match the behavioural system in high BIS individuals. In these
individuals, more positive affect would be elicited by safe decks that provide less conflicting expectations (i.e., novelty). In contrast, unresolved conflict would be elicited by the higher levels of novelty in risky decks, resulting in negative affect. If the task, however, would be experienced too threatening altogether, no clear differentiation of the affective measures between safe and risky decks may arise in high BIS individuals (Miu, Heilman & Houser, 2008).
Expectations
With respect to the choice scores (see Figure 1), for the whole group a preference should arise for safe decks over time, as was observed in Bechara et al. (1994). Participants would learn during the game that it is advantageous to choose cards from safe decks.
10
Figure 1. Expected proportion of chosen cards over time for all participants.
Low sensation seekers/high BIS would pick proportionally more cards from safe decks over time. In the second and third block, this development should happen at a fast pace (Figure 2). High sensation seekers/low BIS would show an initial preference (e.g. in the first block) for the risky decks, indicated by choosing proportionally more cards from these decks. Because novelty should wear off over time, a preference in card choice would develop for the safe decks during the second and third block, but at a slower pace than in low sensation seekers/high BIS participants (Figure 3).
11
Figure 2. Expected proportion of chosen cards over time for low sensation seekers / high BIS.
Figure 3. Expected proportion of chosen cards over time for high sensation seekers / low BIS.
With regard to affect, for all participants a development towards more positive affect for safe decks and more negative affect for the risky decks is expected when performing the IGT over time (Figure 4).
12
Figure 4. Expected affect index for safe and risky decks over time for all participants.
Low sensation seekers / high BIS were expected to develop increasing positive affect for safe decks. Simultaneously, they would develop increasing negative affect for risky decks. The increase in negative affect for the risky decks was expected to develop fast, because low sensation seekers / high BIS might be more sensitive to the negative properties of the
processing of novelty (Figure 5). High sensation seekers / low BIS would show initially more positive affect towards risky decks due to a higher level of novelty in these decks. However, novelty would wear off over time, resulting in less positive affect for the risky decks over time. Simultaneously, a slow decrease of negative affect for safe decks was expected over time (Figure 6).
13
Figure 5. Expected affect over time for low sensation seekers/high BIS.
Figure 6. Expected affect over time for high sensation seekers/low BIS.
With regard to gain, it was expected that choosing comparatively more risky decks would in total lead to loss while choosing more safe decks would result in gain. Because it was
14 it was expected that these groups would experience more loss than low sensation seekers/high BIS.
The explicit indication of preference for the decks was expected to be more positive for high sensation seekers than for low sensation seekers towards risky decks. However, because it was expected that the group would have learned to associate risky decks with loss, that the evaluation overall would be more negative for risky decks than for safe decks.
Method Participants
Sixty-two unselected participants took part in the study (aged 18-36 with average age 22; SD = 4, 39 female). All participants had normal or corrected-to-normal vision. At recruitment, participants were tested on their handedness by the Edinburgh Handedness Inventory (Oldfield, 1971). Participants with a score lower than 9 out of 10 on right-handedness were excluded from participation in order to avoid potential differences in emotional processing between left- and right-handers (Casasanto, 2009).
Participants signed informed consent. For their participation, students received one course credit. Furthermore, each participant received 10€ as their gain from playing the IGT. Participants with less than 90% accuracy in correctly evaluating happy and angry faces during the whole experiment were considered not sufficiently motivated and excluded from the analyses. The groups of low, average, and high sensation seeking / BIS were assigned with a percentile split of 33%.
Design
The experiment consisted of 6 blocks, with 3 IGT blocks and 3 affective-priming blocks. The affective-priming blocks consisted of 40 trials. Each affective-priming block was preceded by an IGT block, containing two subblocks of each 16 trials (Figure 7). At the beginning, a
15 practice trial was conducted for evaluating the facial stimuli in order to make sure that
participants understood how to correctly react to happy and angry faces. A baseline measure of reaction time to the stimuli followed. During the practice block and the evaluation block, in a random order, all 40 stimuli were presented. Afterwards, a short practice block of the
affective priming task followed. In this block, the number of positive and negative stimuli was balanced. Randomly selected stimuli appeared in a random order. Positive and negative card values were randomly selected but were not balanced. The position of risky and safe decks for the practice trials varied for each participant but stayed the same during the experiment.
Exercise blocks and baseline
Experimental blocks
IGT Block 1 AP Block 1 IGT Block 2 AP Block 2 IGT Block 3 AP Block 3
Figure 7. Schedule of exercise, baseline, practice trials and IGT (sub)blocks and
affective-priming (AP) blocks with their approximate duration.
The design for the IGT was a 4 (sensation seeking high/low, BIS high/low) x 2 (high vs. low risk decks) x 6 (subblocks over time) mixed factorial design. The dependent variable for the IGT was the proportion choice of the decks. The design for the affective-priming task was a 4 (sensation seeking high/low, BIS high/low) x 2 (high vs. low risk decks) x 2 (happy or angry face) x 3 (blocks over time) mixed factorial design. The dependent variable was the
16 reaction time on correct responses to positive and negative target stimuli (see Fazio et al., 1986). The reaction times were transformed into an affect index per block for each of the four decks. In the affect index the reaction times to positive target stimuli were subtracted from the reaction times to negative target stimuli (AI = RTneg – RTpos). Relatively more positive values in one condition than in another condition indicate that affect was more positive in the former condition than in the latter. More negative values reflect relatively more negative affect. To correct for individual response bias to the stimuli, the reaction times of the baseline measurement to each type of stimulus were subtracted from those during the affective priming task for each participant separately. Trials with reaction times shorter than 300 ms and longer than 900 ms were considered outliers and were excluded from further analyses. Finally, an explicit preference for the decks was determined at the end of the experiment.
In this study, null-hypothesis significance testing (NHST) was not applied. Absence of significance does not mean that the effect is indeed absent (Dienes, 2012) and in the case of significance, this does not mean that the effect is really present (and may even be false in majority of cases; Ioannidis, 2005). Therefore, the “New Statistics” (Cumming, 2014) was applied in this study, only relying on effect size and confidence intervals (Lakens, 2013), and attempts were made to arrive at meaningful interpretations of the data.
Material and Apparatus
Forty facial target stimuli (20 angry, 20 happy; equal number male/female) for the experimental trials were selected from the Karolinska Directed Emotional Faces set (Lundqvist, Flykt, & Öhman, 1998). The target stimuli were presented only once within a block, but repeated across blocks. Presentation order of decks, facial target stimuli and card value were determined randomly (card value and target stimuli without replacement). Participants evaluated the facial target stimulus by pressing the left button of a fast gaming mouse for positive (happy) and right button for negative (angry) faces with their right index
17 finger or middle finger of their right hand, respectively. Response buttons positions were fixed over participants. The light in the room was kept constant at high levels. Participants sat in front of the computer with a distance to the screen varying between 40cm and 60cm per participant. The screen had a size of 61cm, each card deck had a size of 3.5 (width) x 5.5 (height) in cm.
The experimental tasks in this experiment were adjusted versions of the IGT from Bechara et al. (1994), which were administered on the computer. Participants picked a card from one of four decks. Each type of deck contained cards with different values in different frequencies according to a schedule (see Table 1). Risky decks had big gains but also big losses and a high frequency of losses, leading to loss in the long run. Safe decks had small losses and small gains, but led to gain in the long run. Participants did not know beforehand which decks were advantageous or disadvantageous, and had to figure this out while playing. The earned points of each trial in each IGT block and each affective priming block added up throughout the game. The total number of earned points during each affective priming block was zero (see Table 1), but this was not mentioned to participants.
18 Table 1. Gain-Loss values for risky and safe decks per 10 experimental trials
Deck Card Values Risky Risky Safe Safe
1 -50 -50 15 15 2 100 100 -15 -15 3 -100 -100 20 20 4 150 150 20 20 5 150 150 -20 -20 6 -150 -150 25 25 7 200 200 25 25 8 -200 -200 25 25 9 300 300 -25 -25 10 -500 -500 30 30 Total -100 -100 +100 +100
Four decks of cards were presented. A counter bar showed the current value (starting at zero at the beginning of the experiment). After picking a card by clicking on the deck, the value of the card was shown immediately, accompanied by a sound for either gain or loss. The gain or loss, directly added to or subtracted from the current value on the counter bar, was visible for 1750 ms and then the card turned over. The cursor remained invisible until the next trial started (Figure 8).
19
Figure 8. IGT trial. Both positive and negative values were possible. The value was subtracted
of the total displayed on the counter bar.
In the three affective priming blocks an arrow presented for 500 ms first indicated the deck of the to-be-turned-over card. Then a fixation cross was presented on the deck for 500 ms. Subsequently, an angry or happy facial stimulus appeared on the position of the card on the deck. Participants evaluated as fast as possible the face, which disappeared after 500 ms. Next, a blank card appeared on the deck for 1000 ms. Next, the value of the card was shown for 1750 ms, accompanied by a gain or a loss sound and was added/subtracted to the counter bar. The counter bar disappeared during the evaluation phase (before the fixation cross), and returned when the value on the card was shown. The gains and losses associated with the card followed the same scheme as in the IGT trials. After an ITI between 1000 ms – 1500 ms, the next trial started (Figure 9).
InterTrialInterval 1000ms-1500ms
Until a card is chosen
Appearance of card value and sound 1750ms
20
Figure 9. Schematic depiction of an affective-priming trial. An arrow pointed to the deck
where a face will appear. Participants evaluated the face and after a short delay, the value of the card was shown accompanied by a sound and added on the counter bar.
ITI InterTrialInterval jittered 1000ms-1500ms
Value on counterbar depends on scores of earlier trials
Appearance of arrow below deck (random) 500ms
Fixation cross 500ms
Counterbar invisible during fixation cross, stimulus, blank card
Facial stimulus (happy/angry; random) 500ms
Participants react with button for left (happy), right (angry)
Blank card appears after participants’ response 1000ms
Appearance of card value and sound 1750ms
Counterbar shows current (pos. or neg.) gain
21 After the last block of the affective priming task (the end of the experimental tasks), participants were asked to indicate their liking for the four decks, by choosing a value (1 dislike – 5 like) on a 5-point Likert scale under each of the decks, under the heading “How much did you like to pick cards from each of the decks?” (Figure 10).
Figure 10. Indication of explicit preference.
Sensation Seeking Scale-V (SSS-V)
The participants’ level of sensation seeking was assessed with the English version of
Zuckerman’s SSS-V (1994), which was administered on the computer. The SSS-V consists of four 10-item subscales (see Appendix A): Thrill/Adventure Seeking (TAS), Experience Seeking (ES), Disinhibition (Dis) and Boredom Susceptibility (BS). The reliability of the subscales is good with Cronbach’s α = 0.80 for the TAS subscale, ES subscale α = 0.75, Dis subscale α = 0.80 and BS subscale α = 0.76 (Roberti, Storch, & Bravata, 2003). Participants made two-alternative forced choices on these 40 items. Each item contains one option representing the presence (scored 1) or the absence (scored 0) of sensation seeking. The total score indicates the overall level of sensation seeking.
Two items were reworded with respect to the original version. These items
simultaneously serve as examples of the sub-scales. The item from the ES sub-scale “I would How much did you like to pick cards from each of the decks?
22 like to meet some persons who are homosexual (men or women)/I stay away from anyone I suspect of being ‘gay or lesbian’” was worded “I would like to meet some gay or lesbian people/I am not interested in meeting gay or lesbian people”. From the Dis sub-scale, the item “A person should have considerable sexual experience before marriage/It’s better if two married persons begin their sexual experience with each other” was transformed into “A person should gain extensive sexual experience with several different partners/People should share their sexuality with partners who are ‘the right one’”. An example of the TAS sub-scale is “I would like to try sky diving/I would never want to try jumping out of a plane with or without a parachute” and an example of the BS subscale is “I enjoy looking at personal videos or travel photos/Looking at someone’s personal videos or travel photos bores me
tremendously”. The order of items and response options was randomized. Participants were instructed to answer all items and to not skip any item, to choose only one option per item and to be as honest as possible.
Behavioral Inhibition Scale (BIS)
Participants filled in the BIS scale (Carver & White, 1994) on the computer to estimate the level of behavioral inhibition (see Appendix B). The reliability varies between α = 0.59 - 0.79 (Van Den Berg, Franken, & Muris, 2010). The questionnaire consists of 7 items (e.g.
“Scolding or criticism hurts me quite a bit”) on a 4-point scale from 1 (very false for me) to 4 (very true for me). A low score reflects a low level and a high score a high level of behavioral inhibition. Participants were instructed to answer all items and to not skip any item, to choose only one option per item and to be as honest as possible.
For both the SSS-V and the BIS, the experimenter covered the second screen where participants’ choices on the questionnaires were displayed, so that the answers were only visible on the screen of the participant. It was stressed that anonymity of their answers was guaranteed.
23 Procedure
The experimenter explained that the aim of the experiment was to determine whether playing a game could influence reactions to a face. To motivate participants, it was stressed that they could win real money with a maximum amount of 10€, calculated from the points they gained during the game. After reading the information brochure, informed consent was signed. The experimenter asked participants to switch off their mobile devices. Then the game was explained: ”You can see four decks; each card of these decks can bring a win or a loss. You
can choose from whichever deck you like. The decks might differ from each other, but their position remains the same during the game. Both gains and losses can be on the cards, and both the gains and the losses can be quite large. You start with a virtual budget of 0€.”
Participants were further instructed that the game consisted of free-choice blocks and
evaluation blocks, but remained unaware of the total of cards to be picked. It was stressed that in the evaluation blocks, an arrow would choose the cards and they only had to evaluate the faces as fast and as accurate as possible. It was explicitly stressed that in these blocks, gain or loss had no feedback function for evaluative responses and that the following gain or loss should be regarded only as a property of the deck. Also it was mentioned that in every block, each value on the cards would add to the gain or loss and that they should try to gain as much as possible. Furthermore, it was explicitly mentioned that the decks remained on the same position during the whole experiment. When the experiment started, the experimenter wore ear protection and told the participants that their choices and the accompanying sound would not be monitored by the experimenter.
The baseline block and the exercise blocks took about 4 min. altogether. The IGT took about 3 min. per block, the affective-priming task about 6 min. per block. After the second and after the fourth block, there was a short break of 2.5 min. The experimenter informally enquired whether everything was fine, stressed the importance of fast reactions and accuracy and also reassured that in the end, the gain in points would be transformed into monetary gain.
24 After the sixth block, participants indicated their explicit preference for the decks on a scale. Including the indication of explicit preference, the task took about 30 min. The SSS-V questionnaire followed, administered on the computer after the experimental session in order to avoid any influence of the scale on participants’ affective state. To fill in the SSS-V took about 15 min, the BIS questionnaire took about 2 min. In the exit interview, the experimenter asked participants how they liked the tasks, whether they had developed a specific strategy, if they had realized differences between the decks and if yes, when. Also, they were asked for their opinion about the questionnaires. Demographic variables as gender and age and particularities during the experiment were registered in a logbook. The exit interview took about 5 min. In total, the experiment took approximately 55 min.
Results
Sixty-two unselected participants took part in the study. Five of them were excluded from further analyses. One did not follow instructions (only chose cards from the middle decks which were closest to the place where the cursor appeared, because did not want to move the computer mouse too much), one had extremely slow reaction times (>3SD from the overall mean) and three exceeded the limit of 10% incorrect responses to happy and negative faces during the affective priming task. The results of 57 participants were analyzed. They (37 female) were aged 18-36 yr. with average age 22 yr. (SD = 4).
The SSS-V had a reliability of α = 0.74. The subscale Thrill Seeking had α = 0.77, Experience Seeking α = 0.45, Disinhibition α = 0.62, Boredom Susceptibility α = 0.52. The BIS scale had a reliability of α = 0.75. For the whole group, the mean score on sensation seeking was M = 22, ranging from 8-30 (SD = 5) and for BIS M = 21, ranging from 12-28 (SD = 3.5). A negative correlation, based on 57 data points, could be observed between sensation seeking and BIS, r = -0.36, which means that participants scoring high on sensation seeking had lower scores on BIS and low sensation seekers had higher scores on BIS. Because a high
25 BIS score suggests more behavioral inhibition, low sensation seekers seemed to be more behaviorally inhibited and high sensation seekers seemed to be less behaviorally inhibited.
In the exit interview, 20 participants (13 female, 7 low sensation seekers, 7 high sensation seekers) could not specify the differences in gains and losses between the decks. In the following, this group is addressed as unaware group. Thirty-seven participants reported to have noted the differences. This group is addressed as aware group. Only two of them could specify when they realized the differences, reporting the beginning of the second IGT block.
Whole group analyses IGT
For all participants, a classical switch in choice behaviour from risky to safe decks, as observed by Bechara et al. (1994), appeared. Because the exact moment of the switch was unknown, the differences in proportion of deck choice between the first and the last subblock were compared for safe and risky decks. These differences were, however, small (for safe decks: d = -0.26, CI 95% [-0.78, 0.27]; for risky decks: d = 0.26, CI 95% [-0.27, 0.78]) and suggest that participants might not have learned to develop the classical choice behaviour (Table 2, Figure 11).
26 Table 2. Mean proportion of deck choice (SD) over time, per subblock for all participants
(N=57) Safe Risky 95% CI MSafe-Risky Cohen’s d Subblock 1 0.47 (0.16) 0.53 (0.16) [-0.12, 0.0] -0.38 Subblock 2 0.49 (0.21) 0.51 (0.21) [-0.10, 0.06] -0.10 Subblock 3 0.49 (0.23) 0.51 (0.23) [-0.11, 0.07] -0.09 Subblock 4 0.53 (0.24) 0.47 (0.24) [-0.03, 0.15] 0.25 Subblock 5 0.50 (0.28) 0.50 (0.28) [-0.11, 0.11] 0.00 Subblock 6 0.53 (0.29) 0.47 (0.29) [-0.05, 0.17] 0.21
27 In the unaware group, a larger switch appeared in card choice between the first and the last subblock (for safe decks: d = -0.34, CI 95% [-1.23, 0.54]; for risky decks: d = 0.34, CI 95% [-0.54, 1.23]). The switch suggests that these participants, lacking explicit knowledge, implicitly learned to distinguish between safe and risky decks and developed classical choice behaviour (Table 3, Figure 12). The aware group (Table 4, Figure 13) showed a smaller switch in card choice between the first and the last subblock (for safe decks: d = -0.23, CI 95% [-0.87, 0.42]; for risky decks: d = 0.23, CI 95% [-0.42, 0.87]). However, this probably does not represent a meaningful difference with the unaware group. In absolute value, the differences were larger for the aware than the unaware group. The larger variance in the former than in the latter group must account for these negligible effect size differences.
Table 3. Mean proportion of deck choice (SD) per subblock for unaware participants (N=20)
Safe Risky 95% CI MSafe-Risky Cohen’s d Subblock 1 0.49 (0.10) 0.51 (0.10) [-0.09, 0.05] -0.2 Subblock 2 0.50 (0.12) 0.50 (0.12) [-0.08, 0.08] 0.0 Subblock 3 0.49 (0.18) 0.51 (0.18) [-0.14, 0.10] -0.11 Subblock 4 0.46 (0.18) 0.54 (0.18) [-0.20, 0.04] -0.44 Subblock 5 0.50 (0.20) 0.50 (0.20) [-0.13, 0.13] 0.0 Subblock 6 0.54 (0.18) 0.46 (0.18) [-0.04, 0.20] 0.44
28 Table 4. Mean proportion of deck choice (SD), per subblock for aware participants (N=37)
Safe Risky 95% CI MSafe-Risky Cohen’s d Subblock 1 0.47 (0.18) 0.53 (0.18) [-0.14, 0.02] -0.33 Subblock 2 0.49 (0.25) 0.51 (0.25) [-0.14, 0.10] -0.08 Subblock 3 0.49 (0.26) 0.51 (0.26) [-0.14, 0.10] -0.08 Subblock 4 0.57 (0.26) 0.43 (0.26) [0.02, 0.26] 0.54 Subblock 5 0.50 (0.32) 0.50 (0.32) [-0.15, 0.15] 0.0 Subblock 6 0.53 (0.33) 0.47 (0.33) [-0.10, 0.22] 0.18
29
Figure 13. Mean proportion of deck choice over time, per subblock for aware participants.
Affective Priming
Over the whole group, affective priming did not appear to be a sensitive dependent variable in the IGT. Safe decks did not seem to elicit more positive affect than risky decks, either - in general (d = 0.06, 95% CI [-0.46, 0.58]), or per block (Table 5). When the first block was compared to the last, affect remained the same for safe (d = 0.07, 95% CI [-0.45, 0.59]) and risky decks (d = -0.09, 95% CI [-0.61, 0.43]). With regard to card choice, only a negligible relationship with affect could be observed for safe (all |r| < 0.15) and risky blocks (all |r| < 0.19) over participants (i.e., based on 57 data points). Calculated over the overall means of the affective priming- and IGT blocks (with each couple of IGT subblocks averaged), based on six data points for the six conditions, the correlation between affect and choice was r = -0.26, indicating a weak negative relationship. Surprisingly, when participants chose more cards from a deck, the affect index was more negative and vice versa.
30 Table 5. Means (SD) of affect index in ms over time for all participants (N=57)
Safe Risky 95% CI MSafe-Risky Cohen’s d Block 1 6 (60) -2 (54) [-13, 29] 0.14 Block 2 14 (56) 11 (58) [-18, 24] 0.05 Block 3 2 (58) 3 (57) [-23, 21] -0.02
For unaware participants (Table 6), safe decks in general elicited slightly more positive affect than risky decks (d = 0.07, 95% CI [-0.80, 0.95]). When the first block was compared to the last, affect slightly increased for safe (d = -0.13, 95% CI [-1.0, 0.75]) and decreased for risky decks (d = 0.16, 95% CI [-0.72, 1.04]). Over participants, based on 20 data points, affect was correlated with card choice (i.e., each couple of IGT subblocks
averaged). The correlation for safe decks in the first block was r = -0.36, in the second block r = -0.17, and in the third block r = -0.3, indicating that for safe decks the more positive the evaluation of a deck, the less often a card from the desk was chosen. For risky decks, the correlation, based on 20 data points, was r = 0.26 in the first, r = 0.23 in the second and r = 0.36 in the third block. More importantly, calculated over conditions, based on six data points, the correlation between choice and affect was r = 0.56, indicating a moderate positive
relationship between card choice and affect. Affective priming results, over conditions, appeared to be more clearly related to choice behaviour in the unaware group than in the whole group. These results suggest that in unaware participants, lacking explicit knowledge about the decks, implicit affect might have contributed to choice behaviour.
In aware participants, overall, safe and risky decks did not differ in affect elicitation (d = 0.06, 95% CI [-0.59, 0.7]). When the first block was compared to the last (Table 7), affect slightly decreased for safe (d = 0.16, 95% CI [-0.48, 0.81]) and increased for risky decks (d = -0.21, 95% CI [-0.86, 0.44]). When affect of each affective priming block was correlated with
31 card choice of each IGT block (i.e., each couple of subblocks averaged), based on 37 data points, the correlations for safe decks were weak (all |r| < 0.2). For risky decks, the
correlation, based on 37 data points, was r = 0.26 in the first, r = 0.19 in the second and r = -0.31 in the third block. Affect was not clearly related to choice behaviour. Calculated over conditions, based on six data points, the correlation between choice and affect was r = -0.26, indicating a small, negative relationship. When fewer cards were chosen from risky decks, more positive affect was elicited. In aware participants, affective priming did not correspond to card choice and therefore did not seem to be involved in decision making. It may be conjectured that only when participants did not have other (conscious) clues to base their decisions on, the affective markings of the decks played a role in their choices.
Table 6. Means (SD) of affect index in ms over time for unaware participants (N=20) Safe Risky
95% CI MSafe-Risky Cohen’s d Block 1 0 (63) 1 (40) [-35, 34] -0.01 Block 2 7 (37) 10 (49) [-32, 26] -0.07 Block 3 7 (45) -7(60) [-21, 49] 0.26
Table 7. Means (SD) of affect index in ms over time for aware participants (N=37) Safe Risky
95% CI MSafe-Risky Cohen’s d Block 1 10 (59) -3 (60) [-15, 41] 0.22 Block 2 17 (64) 11 (64) [-24, 36] 0.09
32 Gain
The total gain from the IGT averaged over all participants was -151 (SD = 851). The setup of the decks was clearly successful, because the safe decks (M = 475, SD = 195) overall resulted in a profit and the risky decks (M = -625, SD = 806) in a loss (d = 1.9, 95% CI [1.3, 2.5]). During the middle block, the gain in both safe and risky decks was the highest (Table 8), which corresponds with the finding of the highest positive affect for both safe and risky decks in the second block (Table 5).
Table 8. Average gains (SD) from the IGT per safe and risky deck per block for all
participants (N=57) Safe Risky Block1 Block2 Block3 148 (79) 167 (82) 160 (98) -230 (484) -193 (445) -203 (450)
The correlation over all participants (i.e., 57 data points) between affect and gain was weak for each block in safe decks (all |r| < 0.15). For risky decks, the correlations between gain and affect over participants were also weak (all |r| < 0.26). When calculated over conditions, based on six data points, the correlation between affect and gain was r = 0.37, indicating that affect possibly reflected actual gains and losses instead of conflicting expectations of the decks.
The total gain from the IGT averaged over unaware participants was -468 (SD = 832). The gain in safe decks remained the same, whereas loss in risky decks increased throughout the task (Table 9). Affect did not seem to be strongly related to this development. The
correlation over unaware participants (i.e., 20 data points) between affect and gain was weak for each block in safe decks (all |r| < 0.19). For risky decks, the correlations between gain and
33 affect over participants were in the first block, r = 0.15, in the second block, r = 0.29, and in the third block, r = -0.56. The latter correlation, however, should be interpreted with caution because two high gain values for risky decks in the third block influenced the strength and direction of the correlation. When calculated over conditions, based on six data points, the correlation between affect and gain was r = 0.31, indicating a weaker relationship than between affect and card choice.
The total gain from the IGT averaged over aware participants was 21 (SD = 822), which differed moderately from unaware participants’ gain (d = -0.59, 95% CI [-1.15, 0.04]). In this experiment at least, explicit decision making seemed more effective than intuitive, affectively based choices. For aware participants, gain in safe decks slightly increased, whereas loss in risky decks decreased over time (Table 10). The correlation over aware participants (i.e., 37 data points) between affect and gain was weak in safe decks (all |r| < 0.17). For risky decks, the correlations between affect and gain over participants were r = 0.32 in the first block, r = 0.16 in the second block and r = -0.04 in the third block. More
importantly, when calculated over conditions, based on six data points, the correlation between affect and gain was r = 0.39, indicating that affect was related more strongly to gain than to card choice.
Table 9. Average gains (SD) from the IGT per safe and risky deck per block for unaware
participants (N=20) Safe Risky Block1 Block2 Block3 160 (53) 146 (64) 160 (71) -235 (526) -347 (446) -352 (492)
34 Table 10. Average gains (SD) from the IGT per safe and risky deck per block for aware
participants (N=37) Safe Risky Block1 Block2 Block3 140 (89) 179 (89) 160 (110) -227 (468) -109 (427) -122 (410) Explicit preference
The explicit preference of all participants (Table 11) was somewhat higher for safe than for risky decks (d = 0.33, 95% CI [-0.19, 0.86]). The correlation over participants, based on 57 data points, between affect in the last block for safe decks and explicit preference, was weak to moderate (r = 0.31). The more positive affect for safe decks was elicited during the task, the higher the explicit preference at the end and vice versa. For risky decks, explicit
preference and affect in the last block were negatively correlated (r = -0.26). When implicit affect in the last blockwas positive, participants explicitly evaluated risky decks negatively in the end. Possibly, realising that they lost with risky decks might have influenced explicit preference. This may reveal a dissociation between implicit and explicit affect.
Table 11. Mean (SD) of explicit preference for safe and risky decks for all participants (N=
57)
All participants
Safe Risky 95% CI MSafe-Risky Cohen’s d 3.3 (0.9) 3.0 (0.9) [-0.04, 0.64] 0.33
35 For unaware participants, however, explicit preference (Table 12) was clearly higher for safe than for risky decks (d = 0.79, 95% CI [0.12, 1.7]). This may reflect that unaware participants, in the absence of other clues, relied on affective markings for their choices. The correlation over participants (i.e., 20 data points), between affect during the last block and explicit preference, was however weak for both safe (r = -0.08) and risky decks (r = -0.17).
Aware participants showed an explicit preference for safe decks (Table 13), but the difference was small (d = 0.2, 95% CI [-0.44, 0.85]). The correlation over participants (i.e., 37 data points) between affect during the last block and explicit preference, was weak to
moderate for both safe (r = 0.41) and risky decks (r = -0.31). More positive affect
corresponded to more positive ratings for safe decks and vice versa, but more positive affect for risky decks corresponded with more negative explicit ratings. Although it is unknown when exactly participants had gained explicit knowledge, it can be assumed that the majority had explicit knowledge at least during the last block. An absent consequent correspondence between implicit affect and explicit ratings suggests that explicit knowledge might have influenced at least explicit preference.
Table 12. Mean (SD) of explicit preference for safe and risky decks for unaware participants
(N=20)
Unaware participants
Safe Risky 95% CI MSafe-Risky Cohen’s d 3.6 (0.72) 3.0 (0.8) [0.10, 1.1] 0.79
36 Table 13. Mean (SD) of explicit preference for safe and risky decks for aware participants
(N=37)
Aware participants
Safe Risky 95% CI MSafe-Risky Cohen’s d 3.1 (0.96) 2.9 (1) [-0.26, 0.66] 0.20
Over all participants, typical IGT behaviour showed up in the classical measures of choice and gain, but similar patterns did not emerge in affective priming. The finding that patterns of card choice seemed only weakly related to implicit affect might question if affect was involved in decision behaviour at all. Interestingly, when analysed over conditions, unaware participants’ affect seemed to correspond to choice behaviour, whereas this was not the case for aware participants. Therefore, it can be assumed that affect still might have played a role in the IGT when participants did not rely on explicit knowledge in their choice behaviour.
It was investigated whether the differences in affect elicitation could be magnified, in different directions, by group distinctions in sensation seeking and BIS. Because reliable norm data were not available, the classification of the subgroups was based on the percentile scores of this sample. A 33% percentile split was performed to assign participants to a high, moderate or low level group of both sensation seeking and BIS. Groups were created for low (8-20, N = 19, 13 female), moderate (21-24, N = 20, 13 female) and high (25-30, N = 18, 11 female) levels of sensation seeking. Three groups were also created for low (12-19, N = 18, 10 female), moderate (20-22, N = 17, 12 female) and high BIS (23-28, N = 22, 15 female). In the subsequent analyses, only the extreme scoring groups were considered.
37 Subgroup analyses
IGT
For low sensation seekers (Table 14, 15, Figure 14), the differences in choice behaviour between the first and last subblock were negligible (for safe decks: d = 0.04, 95% CI [-0.86, 0.94]; for risky decks: d = -0.04, 95% CI [-0.94, 0.86]). Only the high sensation seekers seemed to be able to learn in the IGT (Table 14, 15, Figure 15). Their choice preference clearly reversed from risky decks in the first subblock to safe decks in the last subblock (for safe decks: d = -0.39, 95% CI [-1.32, 0.54]; for risky decks: d = 0.39, 95% CI [-0.54, 1.32]). The change in gambling behavior observed by Bechara et al. (1994) was thus observed only for the high sensation seeking group (Figures 14, 15).
Table 14. Mean proportion of safe deck choice (SD) over time, for LSS (N=19) and HSS
(N=18) Safe LSS HSS 95%CI MLSS-HSS Cohen’s d Subblock1 0.49 (0.20) 0.47 (0.10) [-0.08, 0.12] 0.13 Subblock2 0.53 (0.20) 0.49 (0.17) [-0.08, 0.16] 0.22 Subblock3 0.52 (0.23) 0.47 (0.19) [-0.09, 0.19] 0.24 Subblock4 0.53 (0.24) 0.48 (0.24) [-0.11, 0.21] 0.21 Subblock5 0.50 (0.24) 0.52 (0.30) [-0.20, 0.16] -0.07 Subblock6 0.48 (0.27) 0.56 (0.31) [-0.27, 0.11] -0.28
38 Table 15. Mean proportion of risky deck choice (SD) over time, for LSS (N=19) and HSS
(N=18) Risky LSS HSS 95% CI MLSS-HSS Cohen’s d Subblock1 0.51 (0.20) 0.53 (0.10) [-0.12, 0.08] -0.13 Subblock2 0.47 (0.20) 0.51 (0.17) [-0.16, 0.08] -0.22 Subblock3 0.48 (0.23) 0.53 (0.19) [-0.19, 0.09] -0.24 Subblock4 0.47 (0.24) 0.52 (0.24) [-0.21, 0.11] -0.21 Subblock5 0.50 (0.24) 0.48 (0.30) [-0.16, 0.20] 0.07 Subblock6 0.52 (0.27) 0.44 (0.31) [-0.11, 0.27] 0.28
39
Figure 15. Proportion of deck choice for high sensation seekers (HSS).
The differences in choice behaviour were less pronounced for the BIS groups than for the sensation seeking groups. Between the first and the last subblock, the differences were similar for both low BIS (for safe decks: d = -0.25, 95% CI [-1.18, 0.68]; for risky decks: d = 0.25, 95% CI [-0.68, 1.18]) and high BIS (for safe decks: d = -0.26, 95% CI [-1.10, 0.58]; for risky decks: d = 0.26, 95% CI [-0.58, 1.10]). Both the low and high BIS groups showed a similar development in gambling behavior as in the classical IGT (Table 16, 17, Figures 16, 17), but this development was not as pronounced as in the high sensation seeking group.
40 Table 16. Mean proportion of safe deck choice (SD) over time, for low (N=18) and high BIS
(N=22)
Safe
Low BIS High BIS
95%CI ML-H Cohen’s d Subblock1 0.46 (0.15) 0.49 (0.14) [-0.12, 0.06] -0.21 Subblock2 0.49 (0.19) 0.50 (0.17) [-0.13, 0.11] -0.06 Subblock3 0.47 (0.28) 0.52 (0.24) [-0.22, 0.12] -0.19 Subblock4 0.47 (0.23) 0.55 (0.25) [-0.23, 0.07] -0.33 Subblock5 0.52 (0.29) 0.51 (0.26) [-0.17, 0.19] 0.04 Subblock6 0.52 (0.30) 0.55 (0.29) [-0.22, 0.16] -0.10
Table 17. Mean proportion of risky deck choice (SD) over time, for low (N= 18) and high BIS
(N= 22)
Risky
Low BIS High BIS
95%CI ML-H Cohen’s d Subblock1 0.54 (0.15) 0.51 (0.14) [-0.06, 0.12] 0.21 Subblock2 0.51 (0.19) 0.50 (0.17) [-0.11, 0.13] 0.06 Subblock3 0.53 (0.28) 0.48 (0.24) [-0.12, 0.22] 0.19 Subblock4 0.53 (0.23) 0.45 (0.25) [-0.07, 0.23] 0.33 Subblock5 0.48 (0.29) 0.49 (0.26) [-0.19, 0.17] -0.04 Subblock6 0.48 (0.30) 0.45 (0.29) [-0.16, 0.22] 0.10
41
Figure 16. Proportion of deck choice over time for low BIS.
42 Affective priming
Aggregated over all decks in all blocks, more negative affect was elicited in low (M = -3 ms,
SD = 43 ms) than in high sensation seekers, who displayed overall more positive affect (M = 7
ms, SD = 44 ms). This suggests more enjoyment of the whole task for high sensation seekers than for low sensation seekers (d = -0.23, 95% CI [-0.88, 0.42]). Overall, high sensation seekers displayed the same level of positive affect (d = 0.00, 95% CI [-0.92, 0.92]) for both risky (M = 7 ms, SD = 46 ms) and safe decks (M = 7 ms, SD = 50 ms). Low sensation seekers displayed overall only slightly more negative affect towards risky decks (M = -8 ms, SD = 48 ms) than towards safe decks (M = 2 ms, SD = 43 ms), (d = 0.22, 95% CI [-0.68, 1.12]). Risky decks revealed the biggest difference between low and high sensation seekers, eliciting more negative affect in low sensation seekers than in high sensation seekers (d = -0.32, 95% CI [-0.97, 0.33]). For safe decks, the difference in affect between low and high sensation seekers was small (d = -0.10, 95% CI [-0.75, 0.54]).
When analysed over time (Table 18, 19, Figure 18), in the first block, there was no difference between groups for safe decks and risky decks. In the second block, while the difference remained small for safe decks, risky decks elicited moderately more positive affect in high sensation seekers than in low sensation seekers. In the third block, both safe and risky decks elicited more negative affect in low sensation seekers than in high sensation seekers.
Low sensation seekers showed a clear development over time to more negative affect towards both types of decks (Table 18, 19, Figure 18), which shows that negative affect was elicited by both decks in low sensation seekers. They developed more negative affect from the first to the last block for safe (d = 0.34, 95% CI [-0.56, 1.25]) and for risky decks (d = 0.26, 95% CI [-0.65, 1.16]).Their appreciation of the whole task seemed to decline. For the high sensation seekers affectremained at similar, slightly positive, levels over time for safe (d = 0.09, 95% CI [-0.84, 1.01]) and risky decks (d = -0.07, 95% CI [-1.0, 0.85]). They showed continued but limited enjoyment during the whole task (Table 18, 19, Figure 18).
43 Interestingly, this development in the affective priming task between low and high sensation seekers contrasts with choice behaviour in the IGT, where only high sensation seekers showed a development and low sensation seekers did not. This further suggests that choice behaviour and affective priming did not reflect the same underlying process in this experiment.
When low sensation seekers’ affect was correlated with card choice over participants (i.e., 19 data points), the correlation in the first block was r = -0.06, in the second block r = 0.3, and in the third block r = 0.3. For risky decks, the correlation was r = -0.53 in the first, r = -0.08 in the second and r = -0.35 in the third block. Correlated over conditions, based on 6 data points, r = -0.3 indicates that low sensation seekers’ affect did not correspond to card choice. For high sensation seekers, the correlation over participants (i.e., 18 data points) for safe decks in the first block was r = -0.49, in the second block r = 0.26 and in the third block r = 0.14. For risky decks, the correlation in the first block was r = 0.47, in the second block r = 0.27 and in the third block r = 0.31. Correlated over conditions, based on 6 data points, r = -0.05 suggests that high sensation seekers’ positive affect did not correspond to choice
behaviour.
Table 18. Means (SD) of affect index in ms over time for safe decks for LSS (N=19) and HSS
(N=18) Safe LSS HSS 95% CI MLSS-HSS Cohen’s d Block 1 6 (54) 7 (62) [-39.9, 37.9] -0.02 Block 2 14 (46) 12 (60) [-33.8, 37.8] 0.04 Block 3 -13 (57) 2 (49) [-50.4, 20.4] -0.28
44 Table 19. Means (SD) of affect index in ms over time for risky decks for LSS (N=19) and HSS
(N=18) Risky LSS HSS 95% CI MLSS-HSS Cohen’s d Block 1 -3 (54) 0 (52) [-38.4, 32.4] -0.06 Block 2 -4 (57) 17(54) [-58.0, 16.0] -0.38 Block 3 -17 (56) 4 (61) [-60.1, 18.1] -0.36
Figure 18. Affect Index over time for safe and risky decks for low and high sensation seekers
(SS).
Low BIS displayed in general more positive affect for all decks (M = 3 ms, SD = 39 ms) than high BIS (M = -2 ms, SD = 40 ms), corresponding to only a small difference between groups (d = 0.13, 95% CI [-0.50, 0.75]). Overall, low BIS displayed the same level
45 of affect for safe decks (M = 4 ms, SD = 45 ms) as for risky decks (M = 2 ms, SD = 41 ms; d = 0.05, 95% CI [-0.88, 0.97]). High BIS showed slightly more positive affect towards safe decks (M = 2 ms, SD = 45 ms) than towards risky decks (M = -6 ms, SD = 43 ms; d = 0.18, 95% CI [-0.72, 1.08]). Compared to one another, high BIS displayed overall somewhat more negative affect towards risky decks than low BIS (d = 0.19, 95% CI [-0.43, 0.81]). For safe decks, a difference in affect elicitation between low BIS and high BIS was nearly absent (d = 0.04, 95% CI [-0.58, 0.67]).
Analysed over time and per block, the results were inconsistent (Table 20, 21). Low BIS showed a development of positive affect for risky decks from the first to the last block (d = -0.45, 95% CI [-1.38, 0.49]) but remained slightly negative for safe decks (d = -0.02, 95% CI 0.94, 0.90]). High BIS remained slightly negative for risky decks (d = 0.08, 95% CI [-0.76, 0.91]) but showed a development of negative affect from the first to the last block for safe decks (d = 0.48, 95% CI [-0.36, 1.33]). The differences found for low and high BIS with regard to affect for safe and risky decks are in general and over time inconsistent and not as clear as they are for low and high sensation seekers. Therefore, further analyses with regard to the relationship between affect and choice behaviour were not conducted.
Table 20. Means (SD) of affect index in ms over time for safe decks for low BIS (N=18) and
high BIS (N=22)
Safe
Low BIS High BIS 95% CI ML-H Cohen’s d Block 1 -5 (51) 12 (49) [-49.2, 15.2] -0.34 Block 2 20 (59) 9 (50) [-24.5, 46.5] 0.20 Block 3 -4 (49) -14 (58) [-24.3, 44,3] 0.18
46 Table 21. Means (SD) of affect index in ms over time for risky decks for low (N=18) and high
BIS (N=22)
Risky
Low BIS High BIS 95% CI ML-H Cohen’s d Block 1 -14 (39) -6 (59) [-39.5, 23.5] -0.16 Block 2 11 (52) -1 (50) [-20.9, 44.9] 0.24 Block 3 8 (58) -10 (45) [-15.8, 51.8] 0.35
Gain
The profit/loss from the IGT showed a small difference between groups (d = 0.17, 95% CI [-0.48, 0.81]), as low sensation seekers had less loss and more gain with a minimum of -1375 points and a maximum of 1710 points (M = -86, SD = 915) than high sensation seekers with a minimum of -1630 and a maximum of 935 points (M = -230, SD = 768).
For low sensation seekers, the correlation over participants, based on 19 data points, between gain per IGT block and affect for safe decks was r = -0.05 for the first, r = 0.26 for the second and r = 0.32 for the third block. For risky decks, the correlations for the first block was r = -0.02, for the second block r = 0.38 and r = -0.3 for the third block. This possibly reflects an influence of gain (see Table 22, 23). For high sensation seekers, the correlation over participants (i.e., 18 data points) between affect and gain was weak for safe decks (all |r| < 0.23). For risky decks, in the first block the correlation was weak to moderate, r = 0.35, but was weak in the second and third block (all |r| < 0.09). Calculated over conditions, based on six data points, the correlation between affect and gain was r = 0.62 for low sensation seekers and r = 0.03 for high sensation seekers. It seemed as if for high sensation seekers, gain was mainly unrelated to affect. Conversely, in low sensation seekers, gain seemed to be related to affect throughout the task.
47 Table 22. Mean gain (SD) from the IGT per block for safe decks for LSS (N=19) and HSS
(N=18) Safe LSS HSS Block1 Block2 Block3 169 (88) 169 (83) 141 (90) 146 (50) 153 (79) 185 (116)
Table 23. Mean gain (SD) from the IGT per block for risky decks for LSS (N=19) and HSS
(N=18) Risky LSS HSS Block1 Block2 Block3 -211 (492) -34 (446) -321 (480) -344 (436) -239 (500) -131 (386)
With regard to BIS (Table 24, 25), participants low in BIS won a minimum of -1665 and a maximum of 870 points (M = -319, SD = 737). For high BIS, the minimum was -1005 and the maximum 2050 (M = 0.05, SD = 769), showing a moderate difference between gain in low and high BIS (d = -0.42, 95% CI [-1.05, 0.21]). However, when gain was correlated with affect over participants for low BIS (i.e., 18 data points), the relation was weak for safe decks (all |r| < 0.17). For risky decks, the correlation was r = 0.17 in the first, r = 0.27 in the second and r = -0.39 in the third block. For high BIS, the correlation over participants (i.e., 22 data points) was weak for both safe (all |r| < 0.25) and risky decks (all |r| < 0.16). Analysed over
48 conditions, based on six data points, the correlation was weak for low BIS with r = 0.24 and slightly moderate for high BIS with r = 0.42.
Table 24. Mean gain (SD) per block and per safe deck from the IGT for low and high BIS Safe
Low BIS High BIS
Block1 Block2 Block3 139 (69) 144 (81) 149 (104) 146 (64) 176 (83) 160 (96)
Table 25. Mean gain (SD) per block and per risky deck from the IGT for low and high BIS Risky
Low BIS High BIS
Block1 Block2 Block3 -358 (452) -258 (410) -136 (409) -77 (522) -168 (382) -236 (449) Explicit preference
Both low and high sensation seekers showed a preference for safe above risky decks (Table 26), but the differences between both groups are so small that they are probably absent (all |d| < 0.10).
When affect in the last block was correlated with explicit preference of low sensation seekers, based on 19 data points, the correlation was r = 0.59 for safe decks and r = -0.66 for risky decks. For high sensation seekers, the correlation based on 18 data points was r = 0.29 for safe decks and r = -0.30 for risky decks. Overall, the correlations reflect some
49 inconsistency between explicit and implicit measurements of affect for risky decks for both low and high sensation seekers.
Table 26. Mean scores (SD) of explicit preference of low (N=19) and high sensation seekers
(N=18) for safe and risky decks
Safe Risky
LSS HSS 95% CI MLSS-HSS LSS HSS 95% CI MLSS-HSS 3.3 (0.9) 3.3 (1.0) [-0.64, 0.64] 3.0 (0.8) 2.9 (1.1) [-0.54, 0.74]
Explicit preference for safe and risky decks did not differ much in both BIS groups (all |d| < 0.21) and in view of the large variance (Table 27), these should probably be attributed to noise. When explicit preference was correlated with affect in the last block for low BIS, based on eighteen data points, the relationship was r = 0.04 for safe decks and r = -0.21 for risky decks. For high BIS, based on 22 data points, the correlation was r = 0.18 for safe decks and r = -0.24 for risky decks. These correlations are small and again illustrate an inconsistency between implicit and explicit measurement of affect.
Table 27. Mean scores (SD) of explicit preference of low (N=18) and high BIS (N=22) for
safe and risky decks
Safe Risky
Low BIS High BIS 95% CI ML-H Low BIS High BIS 95% CI ML-H 3.3 (1.0) 3.1 (0.9) [-0.42, 0.82] 3.0 (1.1) 3.2 (0.9) [-0.85, 0.45]
50 Discussion
The current experiment aimed to investigate a basic premise of the Affective Monitoring hypothesis (Phaf & Rotteveel, 2012), which argues that novelty corresponds to conflict, resulting in intrinsically negative affect. The IGT provided different conditions of conflicting expectations (i.e., a similar conflict as with novelty) in the form of reward chances of safe and risky decks. Over time, more conflict resolution in the safe decks than in the risky decks should have led to a higher preference for the safe decks and avoidance of risky decks.
The classical IGT choice behaviour could only be weakly replicated over the whole group, showing that after an initial preference for risky decks, preference shifted to safe decks at the end. Bigger gains in risky than in safe decks might have led to risky card choice in the beginning. Later on, being confronted with more frequent and larger losses in risky than in safe decks, participants reversed their choice behaviour. The avoidance of risky decks
presumably corresponded to a negative evaluation of these decks. Therefore novelty, which is assumed to be similar to the conflict elicited by risky decks, seemed negatively evaluated. In the affective priming results, however, the classical IGT preference development of the whole group could not be observed. Even a slightly negative correlation with the choice scores emerged. Affect did not seem to play a role in the shift of choice behaviour. The divergence between choice behaviour and affective priming suggests that both tasks did not reflect the same underlying process in this experiment. Affect seemed to result from actual gains and losses rather than from expectations for the characteristics of the decks. Furthermore,
explicitly reported preference gave no information on preferences for decks as the correlations were inconsistent. Explicit preference therefore seemed not to provide a good measurement of affect, at least less informative than implicit affective priming.
The finding that implicit affect was not clearly related to choice behaviour could even raise some doubt about whether affect is at all involved in IGT decision making, as was argued by Bechara et al. (1997), based on the somatic marker hypothesis (Damasio, 1994).
