approach and avoidance motivation
A report about a research project completed at the Faculty of Social and Behavioural Sciences
January 13th until July 13th
Research group: Work and Organizational Psychology
University of Amsterdam
Number of credits: 36
Elisabeth M¨uhlfeld (10664106) Programme: Neurobiology
Track: Cognitive Neurobiology and Clinical Neurophysiology
Supervisor: Nathalie Boot, MSc Examiner: Matthijs Baas, PhD Co-assessor: Conrado Bosman Vittini, PhD
EEG patterns underlying creative cognition under
approach and avoidance motivation
Elisabeth M ¨uhlfeld (10664106)
AbstractCreative ideas are thought to be achievable through divergent as well as convergent thinking processes. However, despite the fact of being subject to several studies, the underlying basis of both is poorly understood. This electroencephalography (EEG) study sought to find oscillation patterns of creative cognition under approach and avoidance motivation during the pasta task. The focus was thereby set on convergent and divergent thinking processes. In the pasta task, subjects got presented three examples with the same word ending for e.g. pasta names. They were then asked to generate new names for that category. Names that have the same endings as the example words were categorized as items that emerged from convergent thinking processes whereas names with a different word ending were characterized as divergent items. Approach and avoidance motivation was additionally incorporated in this experiment since it is known that these manipulations might influence creative cognition. Furthermore, we measured eye-blink rates (EBRs) because of their assumed relationship with flexible thinking which was also one of our hypotheses. Moreover, we expected approach motivation to be associated with higher levels of flexibility as compared to avoidance motivation.
EEG findings uncover significant differences in power values of the lower theta band between convergent and divergent thinking periods across all regions separately as well as together. High theta power is commonly linked to working memory performance and general memory retrieval which fits the outcome of higher theta power during convergent ideation when primes were recalled to generate analogous responses regarding the word ending. Differences in EEG patterns between tasks under approach and avoidance motivation were not found. At the behavioral level, results reveal no significant influences from approach or avoidance motivation on neither convergent or divergent item generation nor on any of the other measures (flexibility, fluency, switches, repetitions). Also there was no significant relation between EBRs and any of the factors. Interaction analyses between Motivation, Thinking and Switching disclose a slightly more pronounced flexible processing style in form of more switches and divergent thinking under avoidance motivation which might be explained by our small subject number as well as by the restrictiveness of the pasta task as compared to more ‘free-association’ tasks.
Keywords
EEG — Convergent Thinking — Divergent Thinking — Approach — Avoidance — Theta
Contents
Introduction 1
1 Methods 6
1.1 Participants . . . 6
1.2 Experimental task and conditions. . . 6
1.3 Procedure . . . 7
1.4 Data acquisition and analysis . . . 7
1.5 EBR measurement . . . 9
2 Results 9 2.1 Behavioral results . . . 9
2.2 EEG results. . . 10
2.3 EBR analysis results . . . 11
3 Discussion 12 3.1 Limitations of the present study . . . 15
3.2 Future avenues . . . 16
4 Conclusion 17
References 17
Appendix 21
Introduction
‘There is no doubt that creativity is the most important human resource of all. Without creativity, there would be no progress,
and we would be forever repeating the same patterns.’ Edward de Bono
-As emphasized by Edward de Bono, a Maltese inventor, author and physician, creativity is important in many fields like edu-cation, arts, science, economy and diverse others. Although it is subject to studies in various areas such as psychology, sociology, pedagogy, cognitive science and neuroscience and despite its importance in day-to-day life, there is unfortunately up to date no accord of how creativity might be manifested in the brain. This question is approached in an exploratory manner in the present study, focusing thereby on underlying oscillatory patterns.
People generally agree that ideas need to be novel and task-appropriate in order to be creative. Yet, the task-appropriateness of ideas (like especially works of art) might change with time
and corresponding alterations in culture and taste. This might lead to the more specific definition of creative ideas of be-ing not only task-relevant, but also appropriate for experts (Nijstad, De Dreu, Rietzschel, & Baas,2010).
In addition, a differentiation between ‘big C’ and ‘little C’ is usually made. ‘Big C’ refers to creative genius like for example Mozart, Einstein and Da Vinci. ‘Little c’ is a term for day-to-day creativity, thus for the small creative problems we solve in every day life (Gardner,1993). Generally, the latter notion of creativity is subject to research, since it is easier to simulate in experiments and might comprise the same me-chanisms as ‘big C’ creativity (Ward, Smith, & Finke,1999). Another common distinction is made between convergent and divergent thinking styles, where latter ones are most of the time associated or even equated with creativity (Cropley,
2006). Divergent items are usually the result of open-ended tasks where multiple solutions are expected (Mumford,2001;
Cropley,2006). Examples of divergent thinking measures include the Torrance Test of Creative Thinking and the Alter-native Uses Test (AUT) (Guilford,1967). In the AUT, subjects are asked to generate as many alternative uses for a common object as they can think of. Results are then evaluated regard-ing elaboration, i.e. the amount of detail given in the answers, and with help of the measures fluency, flexibility and original-ity. These criterions are traditionally used to assess creativity scores in divergent thinking tasks (Guilford,1967;Mumford,
2001). Thereby, fluency refers to the number of produced answers, originality captures the ‘novelty’ and ‘uniqueness’ of generated solutions and flexibility scores usually comprise the number of switches between categories and the uniqueness of solutions.
Contrary to divergent thinking, convergent thinking is in gen-eral compared to analytical cognition like arithmetical or log-ical operations that normally result in a single solution and involve manipulations of knowledge (Cropley,2006). Often, Mednick’s Remote Associates Test (RAT) (Mednick & Med-nick,1967) is used to assess convergence levels. In this test, people are asked to generate one word that associates three given ones. Since this results in a single solution and also requires ‘logical linking’, the RAT is regularly employed as a convergent thinking test. Yet, it can also be used to evaluate divergent thinking as for example in a study about the influ-ence of oxytocin on creativity byDe Dreu et al.(2013). The author motivates this latter usage by explaining that divergent thinking is needed to come up with several potential solutions before a single solution has to be elected with help of conver-gent thinking processes. Thus, the RAT can theoretically be used to assess both kinds of thinking.
This latter fact already suggests that divergent thinking is not the only important process in creative cognition. Instead, it is proposed that both, convergent and divergent processes to-gether, are needed to accomplish creative results that meet novelty as well as appropriateness criteria (Cropley,2006). In addition, convergent processes might also denote thinking
styles that follow particular rules and can result in more than one solution. This latter definition is also used in the present study when referring to convergent item generation.
Further support for the belief that convergence plays a role in creativity comes fromFischer and Hommel(2012) refer-ring to an unpublished paper by their group which suggests that divergent thinking is associated with flexibility, conver-gent thinking with persistence and that the usage of either of the pathways can result in creative outcomes (Hommel, Ak-bari Chermahini, van den Wildenberg, & Colzato,submitted). This potential involvement of two different pathways in cre-ative production and problem solving which also might ex-plain the importance or rather the possibility of convergence in these processes is discussed below in the sectionthe dual pathway to creativity model, stating that both, persistence and flexibility pathway, may lead to equally high levels of creativ-ity (De Dreu, Baas, & Nijstad,2008;Nijstad et al.,2010).
The dual pathway to creativity model
The dual pathway to creativity model was first proposed by
De Dreu et al.(2008) and further elaborated and discussed by
Nijstad et al.(2010). In this model potential mechanisms of ‘little c’ creativity are captured. The main argument of the the-ory is that creativity can be reached via either or both of two pathways: the flexibility pathway and the persistence pathway. The flexibility pathway
The flexibility pathway accounts for creativity reached through flexible processing and holistic thinking, switching between categories and connecting distant ideas (Nijstad et al.,2010) and it is used under activating moods with a positive tone (De Dreu et al.,2008). On a psychological level, potential rea-sons for increased cognitive flexibility in people in a positive mood state might be their elevated tendency for exploratory behavior and the readiness to assume risk which might be trig-gered by positive mood related feelings of safeness (De Dreu et al.,2008).
On a neurobiological level heightened flexibility might be explained by reduced levels of latent inhibition (Nijstad et al.,2010) which is in general associated with higher levels of creativity (Carson, Peterson, & Higgins,2003;Fink, Slamar-Halbedl, Unterrainer, & Weiss,2012). Decreased latent in-hibition is assumed to result in a larger pool of accessible stimuli and thus in the generation of more original ideas. This might in turn only be of help for creative ideation if people also exhibit high mental abilities which facilitate filtering out of inappropriate ideas (Nijstad et al.,2010).
A speculative mechanism, as proposed byNijstad et al.(2010), includes the apparent relation between latent inhibition and dopamine (Eysenck,1993). Accordingly, positive mood leads to a release of dopamine (Ashby, Isen, et al.,1999) which in turn attenuates latent inhibition (Gray et al.,1995) and sim-ilarly heightens cognitive flexibility (Cools,2008). This fits
well with the dopaminergic theory of positive affect, stating that increased dopamine levels correlate with positive affect leading to a more flexible processing style and thus also to a facilitation of creative problem solving (Ashby et al.,1999). Additional evidence for the influence of dopamine on cog-nitive flexibility comes from eye-blink rate (EBR) studies. Correspondingly, it was shown that spontaneous EBRs act as a clinical marker of dopamine levels (Karson,1983;Taylor et al.,1999). This idea arose in the first place from dopamine-attributed dysfunctions (Colzato, van den Wildenberg, van Wouwe, Pannebakker, & Hommel,2009). Accordingly, re-duced EBRs have been found in recreational cocaine users (Colzato, van den Wildenberg, & Hommel,2008) and Parkin-son patients (Deuschl & Goddemeier,1998) while elevated EBRs were observed in schizophrenics (Freed et al.,1980). Dopamine functioning might predict flexibility in creative pro-cesses (Chermahini & Hommel,2010;Ashby et al.,1999) and its relationship to creative performance is thought to be non-monotonic, but rather to have a ‘U-shape’ (Goldman-Rakic, Muly III, & Williams,2000), exhibiting highest degrees of flexibility for a medium EBR. In addition,Chermahini and Hommel(2010) found a negative correlation between EBRs and convergent thinking, i.e. the higher the dopamine level, the more impaired was convergent thinking.
The persistence pathway
People using the persistence pathway reach creative ideas through systematic search and the exploration of categories in depth, not rendering flexible thinking and switching between categories necessary. While the flexibility pathway is possi-bly running under activating moods with positive tone, the persistence pathway might be utilized under activating moods with negative tone (De Dreu et al.,2008). Negative mood states that convey feelings of danger and discomfort may then result in these systematic and incremental search processes (Nijstad et al.,2010). On a ‘deeper’ and neurobiological level, dopamine and its equilibrium in frontal areas might regulate which of the two pathways is used (Braver & Cohen,2000;
Cools,2008).
Approach and avoidance motivation
More influential on the ‘selection’ of a pathway than dopamine levels, activation and hedonic tone, might be regulatory focus. In that context approach motivation denotes the directing of be-havior towards pleasure or a positive stimulus, and avoidance motivation turns the behavior away from a negative stimulus or pain (Elliot,2006).
Indications that higher cognitive flexibility might be related to approach motivation were pointed out byFriedman and F¨orster(2001,2002). They tested the influence of promotion and prevention cues as well as arm flexor and extensor con-tractions as approach and avoidance motor actions on creative
performance.
Additionally, Friedman and F¨orster(2001, 2002) deduced that avoidance-related motivation elicits a more perseverant processing style that leads to less creative outcomes. Neverthe-less, this latter conclusion of less creative outcomes might not generally be true as discussed byBaas, De Dreu, and Nijstad
(2011). The authors stressed there thatFriedman and F¨orster
(2001,2002) only focused on approach and avoidance motiva-tion under regulatory closure i.e. under successful attainment of either their approached goal or its successful avoidance. Yet, unfulfilled avoidance motivation that is usually accompa-nied by activating rather than deactivating moods might elicit similar creative results in terms of fluency and originality by utilizing a more persistent processing style. Hence the drawn conclusion ofFriedman and F¨orster(2001,2002) that perse-verant processing styles lead to less creative outcomes may possibly not generally be applicable.
In terms of the dual pathway to creativity model, potentially two distinct pathways, i.e. the flexibility in case of approach and the persistence pathway in case of avoidance motivation, get activated (Nijstad et al.,2010). Thus, under avoidance motivation, participants might counteract their less flexible thinking style by being more persistent, effortful and con-trolled (Roskes, De Dreu, & Nijstad,2012). This could lead to equally creative outcomes in terms of fluency and original-ity as results achieved with the flexibiloriginal-ity pathway (Baas et al.,2011;Nijstad et al.,2010). Following, one can say that motivation in form of approach or avoidance has a substantial influence on activation levels and mood in general and thus also on the selection of the pathway and creative congition. Besides,De Dreu et al.(2008) notice that deactivating moods, irrespective of whether under regulatory closure or not, do not seem to affect creativity at all and that activating positive moods are associated with higher flexibility (see alsoAshby et al.,1999) whereas activating negative ones are linked to higher levels of persistence.
These points already give a hint on the fact that the relation between creativity and mood is not simple: not only hedonic tone, but also activation levels, regulatory focus and closure as well as a combination thereof play an important role in creativity (Baas, De Dreu, & Nijstad,2008) and make it hard to holistically evaluate in experiments.
EEG and oscillations
Some of above discussed aspects were incorporated in EEG creativity studies seeking to capture underlying oscillatory patterns (e.g. Fink & Neubauer, 2006; Fink et al., 2009;
Bekhtereva, Dan’ko, Starchenko, Pakhomov, & Medvedev,
2001;Benedek, Bergner, K¨onen, Fink, & Neubauer,2011). EEG measurements record ‘brain activity’ by determining dif-ferences in the electrical potential on various positions on the scalp. Source of electric field potentials are for the biggest part excitatory and inhibitory postsynaptic potentials (Srinivasan,
of those electrical fields results in a decrease of the signal with increasing distance from the source, which leads to farther propagation for stronger electrical fields. However, electrical fields are only measurable if they are perpendicular oriented to the scalp while parallel oriented fibers produce weaker potentials that are much harder to detect. Thus, EEG measure-ments only capture potentials that are already quite strong and synchronized (Roach & Mathalon,2008).
Hans Berger, who invented the EEG in 1924, at first detected alpha waves which range between 8 and 12 Hz (Berger,1929). In creativity research, a distinction between lower (7.5 - 9 Hz) and upper (9.5 - 12.5 Hz) alpha band is frequently made (Petsche, Kaplan, Von Stein, & Filz,1997). This is done due to the fact that those two bands seem to be function-ally different: the upper alpha band was shown to be related to semantic memory (Klimesch, Schimke, & Pfurtscheller,
1993; Klimesch, Doppelmayr, Pachinger, & Ripper,1997) whereas the lower one is associated with attention and episodic memory and can further be separated into two ‘sub-bands’ (Klimesch, Doppelmayr, Schimke, & Ripper,1997;Klimesch,
1999). Other frequency bands are delta (0-4 Hz), theta (4-7 Hz), beta (12.5-30 Hz) and gamma (25-100 Hz, typically around 40 Hz). These oscillations are thought to be function-ally relevant and to emerge from behavior-dependent networks consisting of cortical neurons (Buzs´aki & Draguhn,2004). In addition, a very important aspect is the assumption that oscillations might account for the synchronization of both, local and distant networks. This might be the case due to low energy costs that emerge from coupling rhythms to tempo-rally coordinate neurons. Much higher costs come along with synchrony induced by e.g. strong common inputs (Buzs´aki & Draguhn,2004).
Not only the fact that oscillations are considered to convey interesting information about the formation and communi-cation of neural assemblies and that creativity is thought to rely at least partly on broad collaboration of neuronal groups (Jauˇsovec & Jauˇsovec,2000;Buzs´aki & Draguhn,2004; Tha-gard & Stewart,2011) but also the circumstance that espe-cially alpha oscillations were widely found to exhibit a con-nection to creativity in the broad sense motivated us to make use of time-frequency analyses of EEG data from a creativity task to take a closer look at potential underlying oscillatory mechanisms.
Why time-frequency analysis as opposed to ‘simple’ event-related potential analysis
At this point it might be appropriate to shortly stress why we made use of time-frequency analysis as opposed to a more traditional and computationally simpler event-related potential (ERP) analysis. Averaging across EEG epochs that are time-locked to a specific stimulus or event yields ERPs. These ERPs are then visible as positive or negative voltage deflections. Seemingly ‘random’ fluctuations in the EEG get cancelled out by averaging across a huge number of them
approaching zero with increasing number of epochs. The positive or negative fluctuations visible in the ERPs are often interpreted as being indicative for specific cognitive and sen-sory processes (Roach & Mathalon,2008). However, ERPs are often not capable of completely ‘grasping’ the electrophy-siological responses which has several reasons. First, evoked responses are not completely independent of the EEG, second, they are not stable and third, the EEG itself is influenced by task-related events (Makeig,1993). This results in the fact that ERPs are difficult to interpret and not easily linkable to physiological mechanisms which is also due to the relatively small understanding of underlying neurophysiological mecha-nisms of ERPs (Cohen,2014).
Cohen(2014) gives three major reasons for the favored use of time-frequency analyses over ERP analyses: First, the results yielded with time-frequency analyses as opposed to especially ERPs can most often be interpreted on the neurophysiological basis of oscillations which are quite well understood. Addi-tionally, oscillations are studied with a variety of techniques in neurosciences. Hence findings from those different mea-surements can easily be linked. And finally, time-frequency analyses simply uncover more of the brain dynamics measu-rable with EEG than ERPs do.
For creativity research both types of analyses are used whereas latter one more frequently. In what follows some reviews and studies that include amongs others those two kinds of analyses are presented.
In a recent review of EEG, ERP and neuroimaging studies
Dietrich and Kanso(2010, p.1) point out that ‘creative think-ing does not appear to critically depend on any sthink-ingle mental process or brain region [...]’ and that at least with current neuroimaging techniques it could not be localized.
Dietrich and Kanso(2010) reviewed a total of 63 articles categorized into divergent thinking, artistic performance and insight studies. Summarized findings of divergent thinking studies revealed ‘diffuse prefrontal activation’ that could not further be specified. Also regarding lateralization, which is one hypothesized characteristic in that category,Dietrich and Kanso(2010) found no consistent results. A further topic were changes to the alpha band. Findings there reached from in-creases to dein-creases in synchrony and power and additionally to conclusions implying that divergent thinking is not associ-ated with alpha but instead with delta, beta or theta (Dietrich & Kanso,2010). Besides evoking diffuse activation patterns as divergent thinking studies did, artistic performances were found to, quite unsurprisingly, mostly elicit activations in temporoparietal and motor regions that might be related to task specific movements. The latter category, insight studies, hinted at differences in activation patterns of the anterior cin-gulate cortex as well as the prefrontal cortex. However, these conclusions were rather vague and finally lead together with contradictory findings from the first two groups of studies to the above cited statement ofDietrich and Kanso(2010, p. 1). Another review of EEG, fMRI, PET and SPECT creativity
Figure 1. The design of one trial. A trial, which always corresponded to a category, started with the presentation of three
example words. These words all exhibited the same word ending, like e.g. lunia, fridia and ezilia in the flower category. Pressing ENTER lead to the thinking period marked by a white cross on a black background. Participants were asked to press SPACE when they had an idea to start typing the word. Another ENTER press marked the beginning of the next thinking period. All thinking periods of one trial together summed up to 30 seconds. Before the next trial started, participants were asked to rate their motivation in and the difficulty of the foregoing trial.
studies from the same year deduced that there is ‘little clear evidence of overlap in [...] [the] results’ (Arden, Chavez, Grazioplene, & Jung,2010, p. 1). Regarding EEG research,
Arden et al.(2010) conclude that one might associate higher alpha synchronization with creative cognition in general but that this synchronization might rather be attributed to working memory, episodic short term memory demands and ‘suppres-sion of unattended positions during visual spatial orienting’ (Arden et al.,2010, p. 5).
The view that higher apha synchronization is not merely linkable to creative cognition but that differences therein might also result from other factors is shared byBenedek et al.(2011). Accordingly,Benedek et al.(2011) state that it is essential to differentiate between low and high internal processing demands since they produce different degrees of synchronization of oscillations. Thus one can not simply as-sume that divergent thinking tasks result in higher frontal alpha synchronization than convergent ones which is often done (Fink, Grabner, Benedek, & Neubauer,2006;Fink et al.,
2009;Benedek et al.,2011;Fink & Benedek,2012). Instead, differences in alpha activity for convergent and divergent item generation might arise because of the discrepancy between in-ternal processing demands of divergent and convergent tasks, which are frequently high for former and lower for latter ones. Correspondingly,Benedek et al.(2011) found that un-der high internal attention demands, alpha synchronization at frontal sites was present for convergent as well as for di-vergent thinking tasks with latter ones additionally exhibiting alpha synchronization in right posterior parietal regions. A further investigation of alpha activity by the same group
(Benedek, Schickel, Jauk, Fink, & Neubauer,2014) focused only on divergent thinking tasks and differentiated two kinds of them by means of ‘task-immanent differences in bottom-up processing demands’ (Benedek et al.,2014, p.1): whether a continuous relying on external information, thus resulting in constant bottom-up processing, was needed to complete the task or not. The first kind of task was termed sensory-intake and the second one sensory-independence task. Additionally two more conditions in form of low (stimuli were visible on screen throughout the trial) and high (masking the stimulus after 500ms) internal processing demands as applied in the study from 2011 (Benedek et al.,2011) were incorporated. Results revealed that alpha power increases were generally higher for the sensory-independence task and that internal attention demands only increased alpha power in the sensory-intake task and not in the sensory-independence task (Benedek et al.,2014).
The general inconsistency in results, as also became apparent in above paragraphs, and the sparseness of creativity research in general might be traced back to two big problems: 1. There is a lack of a clear conceptual definition of creativity. 2. There are several methodological problems in existing stu-dies affecting reliability and validity of creativity measures. Methodological problems include the huge variety of applied tasks that are thought to capture creative thinking processes. Further sources of inconsistent findings might lie in manip-ulations of the tasks: if they are carried out under approach or avoidance motivation (Baas et al.,2011,2008;De Dreu et al.,2008;Spielberg, Stewart, Levin, Miller, & Heller,2008;
cope with low or high memory load (Jonides et al., 1997;
Rypma, Berger, & D’esposito,2002) and processing demands (Benedek et al.,2011) and if the tasks were of the sensory-intake or sensory-independence type (Benedek et al.,2014). Additionally, it is important to deliberately determine if the resting state is carried out under similar with eyes-closed or eyes-open. Correspondingly,Mo, Liu, Huang, and Ding
(2012) andYan et al.(2009) found that those two manipula-tions during the resting state result in different activity patterns and thus affect study outcomes. These are at least some of the points that might influence measurement outcomes and thus lead to disperse outcomes when comparing studies.
In this study, divergent and convergent thinking, two aspects of creative cognition, are assessed by the pasta task, which is explained in more detail in themethodssection. Generated words can be assigned to either of the two groups, convergent and divergent, and hence be attributed to those specific types of thinking. Since dopamine is thought to play a role in cre-ative processes, we additionally measured EBRs, which are assumed to reflect dopamine levels in the striatum (Karson,
1983;Taylor et al.,1999). It is then investigated whether we can use EBRs to predict flexibility levels of responses. Addi-tionally, we added a further manipulation in form of approach and avoidance motivation which are believed to influence cre-ative cognition (Baas et al.,2011;Roskes et al.,2012). As mentioned above, the present study uses the term ‘con-vergent’ for produced items that are similar to the presented examples in terms of word endings. Thus, in this study, con-vergent processes do not result in a single solution but in multiple ones. Opposed to that, convergent thinking is most often referred to as arithmetic or logical reasoning leading to a single either correct or wrong solution. Exceptions to this general understanding and usage of convergence are studies byDijksterhuis and Meurs(2006) andDe Dreu et al.(2013). They also use the pasta task to discern between divergent and convergent ideation in the same manner as in the present ex-periment which employs an extended version of the task used in their studies. However, neitherDijksterhuis and Meurs
(2006) norDe Dreu et al.(2013) made use of EEG to measure these processes. Due to the fact that this study is, at least to my knowledge, the first one to investigate divergent and convergent thinking as defined here by means of EEG, it is difficult to propose strong and already supported hypotheses. Hence, the following ones are of highly speculative nature.
Hypotheses
a) Medium EBRs predict highest level of cognitive flexi-bility
As discussed in the section aboutthe dual pathway to cre-ativity model, EBRs seem to reflect dopamine levels (Karson,
1983;Taylor et al.,1999) and thus might indicate flexibility scores of creative cognition (Chermahini & Hommel,2010,
2012;Ashby et al.,1999). Accordingly, it is hypothesized that also in the present study a positive correlation between
cognitive flexibility and EBRs will be found. Since this re-lationship is supposed to have a U-shape (Goldman-Rakic et al.,2000;Chermahini & Hommel,2012), we expect the highest flexibility for medium EBRs and, consistent with the findings ofChermahini and Hommel(2010), low EBRs to predict higher levels of convergent thinking.
b) Higher flexibility in the approach as compared to the avoidance motivation condition
A second hypothesis is that the approach as compared to the avoidance motivation condition will result in higher cognitive flexibility. This assumption relies on conclusions made by
Friedman and F¨orster(2001,2002) who looked for the in-fluence of approach and avoidance motor actions on creative performance. What they found was that approach motivation seems to be related to higher flexibility than avoidance moti-vation which is explained in more detail inthe dual pathway to creativity model.
1. Methods
1.1 Participants
Thirty-seven (11 male and 26 female) participants took part in that study which called for participants on a website for experiments conducted at the University of Amsterdam. Their mean age was 21.46 (SD 2.41). They participated on a volun-tary basis and received either a financial reward of 20 euros or course credit. The reward could be increased up to 26 euros depending on their task performance and also people who chose to receive course credit could gain the bonus. All parti-cipants had normal or corrected-to-normal vision. Partiparti-cipants gave written informed consent and the study was approved by the ethical committee of the Psychology Department of the University of Amsterdam.
One female participant was excluded from the analysis since the amount of ‘clean’, artifact-free data was insufficient for an analysis.
1.2 Experimental task and conditions
Participants had to solve the pasta task while being recorded with EEG. In the pasta task (Marsh, Ward, & Landau,1999;
Dijksterhuis & Meurs,2006;De Dreu et al.,2013) subjects got presented three examples of non-existent words within a category that all had the same word ending (e.g. fussilini, krapiand falucci in the pasta category) and were then asked to generate as much new names as possible within 30 seconds for that category. All in all, participants had to create names for 30 different categories (seeAppendix) which have previously been tested and selected from a bigger sample in a behavioral pilot study.
A trial was composed of different ‘phases’: example pre-sentation, thinking period, typing the word and ratings of motivation and difficulty. As illustrated inFigure 1, to se-parate the ‘thinking periods’ from the ‘reading and writing
epochs’ button presses were used: after reading the examples, participants were instructed to press ENTER to start the thin-king period during which a white cross on a black background appeared. To indicate that they have an idea and start writing, they had to press SPACE. The end of the typing period and simultaneous start of the next thinking period was marked by a further ENTER press. Participants had 30 seconds of ‘pure thinking time’ for every category hence the time they needed for reading examples and typing words was not sub-tracted from these 30 seconds. We separated these periods to ensure higher reliability of EEG measurements, and, to guarantee a long enough recording time, participants could not proceed to the next period by pressing a button within the first 1500ms. At the end of each trial, participants were asked to rate their motivation in the foregoing trial as well as its subjective difficulty on a scale from one to five with one denoting not motivated and not difficult at all and five really motivated and really difficult.
The whole experiment was divided into six blocks (Figure 2), each containing five trials that corresponded to one category. The manipulations in form of approach and avoidance moti-vation were applied in a counterbalanced order: Before each block, a two-digit number was presented. In the approach motivation condition participants could gain one euro if they successfully remembered the number after the block. In the avoidance motivation condition they would loose one euro if they were not able to recall the number at the end of the block. To provide participants with a minimum earning of 20 euros also in case they could not remember any of the numbers, they got an initial bonus of three euros. This results in a potential additional gain of six euros. Participants did not receive feedback regarding their ‘number recalling’ until the end of the whole experiment in order to avoid influencing their performance.
1.3 Procedure
Participants were seated in a dimmed EEG room. Before the start of the first trial, a resting period of five minutes where participants were asked to fixate on a white cross in the center of a black screen was carried out with eyes open.
To familiarize subjects with the task, two practice trials were carried out after the resting condition. They started with the task instructions on the screen followed by two runs with test categories (seeAppendix). After the practice trials, partici-pants had the possibility to ask questions about the instructions and the task if not clear. Optional breaks were scheduled be-tween the blocks and after the third block one experimenter entered the EEG room for a ‘compulsory break’ to examine if the participant was still feeling fine. All in all each participant spent about scarcely one hour with the experiment.
1.4 Data acquisition and analysis
The EEG recordings from the scalp were made using a BioSemi ActiveTwo 64-channel active EEG system (BioSemi Inc.,
Am-sterdam, The Netherlands). Electrodes were located at 64 positions: at frontal polar (Fpz) left (Fp1), anterio-frontal
(AFz) left (AF3, AF7), frontal (Fz) left (F1, F3, F5, F7),
fronto-central (FCz) left (FC1, FC3, FC5) and fronto-temporal left
(FT7) sites. Further electrode positions were at central (Cz)
left (C1, C3, C5), temporal left (T7), centro-parietal (CPz) left
(CP1, CP3, CP5), temporo-parietal left (TP7), parietal (Pz) left
(P1, P3, P5, P7, P9), parieto-occipital (POz) left (PO3, PO7)
and occipital (Oz) left (O1) sites. Electrode positions were
analogous for the right side. Additional electrodes were Iz, and CMS and DRL as grounding electrodes located on the back of the head.
Data was sampled at 1024 Hz and the reference electrodes were located on the earlobes. Four supplementary external electrodes were applied: two vertical and two horizontal elec-trodes around the eyes to detect eye-blinks and lateral eye movements. This results in a total of six external electrodes. The signal was filtered with a high-pass filter, cutting out signal below 0.16 Hz and a low-pass filter, cutting out fre-quencies above 100 Hz.
Behavioral data analysis
The log-file generated by the ‘presentation’ software (Neu-robehavioral Systems, Inc.) during the experiment was ana-lyzed with matlab (The MathWorks, Inc.).
For that goal, categories along with their corresponding an-swers were extracted from the log-file. Furthermore informa-tion regarding approach or avoidance motivainforma-tion trials was added. If the ending of a generated response, which, de-pending on the category, consisted of one or two letters, was similar to that of the previous one, the response was marked as convergent. If this was not the case, then the answer was categorized as a divergent response. Additionally, we looked for repetitions: successive answers that had both the same ending were identified as repetitions.
Two further criteria we were interested in were flexibility and switches. Flexibility scores were calculated by taking the unique responses with respect to word endings from one cate-gory. The number of switches was defined as the number of successive words with non-repeated endings within a category. For all outcome variables we subtracted duplicates and ‘real’, already existing names.
In addition to these aspects, we also calculated averages: the mean of total answers per participant and category, of diver-gent responses, of flexibility scores and of switches. Further-more, deviations from those average answers were determined. All this information was stored in a big matrix that was after-wards used for the preprocessing as well as the time-frequency analysis.
Focusing on the motivational manipulation of the study, we further took a closer look at the influence of approach and avoidance motivation on behavioral outcomes. This resulted in scores for total, convergent and divergent responses under approach as well as under avoidance motivation along with repetition, switch and flexibility scores for both
manipula-Figure 2. The design of one block. Each block starts with the presentation of a two-digit number. In the approach motivation
condition the successful recall at the end of the block resulted in an additional gain of one euro. In the avoidance motivation condition the failing to recall the number lead to a loss of one euro. The number presentation was followed by five trials and the recall of the at the beginning presented number. In total, the experiment consisted of six blocks, separated by small optional breaks and one compulsory short break after the third block. Approach and avoidance motivation conditions were applied in a counter-balanced order resulting in three approach and three avoidance motivation blocks.
tions. These values were then averaged across all categories and stored in a separate matrix.
Behavioral data statistics
Behavioral data was analyzed with SPSS for Windows (ver-sion 22, SPSS Inc.). Relationships between approach and avoidance motivation on the measures convergence, diver-gence, flexibility, fluency, repetitions and switches as well as relationships between those factors were computed by means of correlation analyses and a repeated measures ANOVA was used to assess interaction effects between motivation (ap-proach vs. avoidance), thinking mode (convergent vs. diver-gent) and switching behavior (repetitions vs. switches). Originality will not be part of the rating since the pasta task does not account for outcomes that can reasonably be assessed regarding this feature.
Preprocessing
It has to be noted that although this study comprises a total of 37 subjects, the preprocessing as well as the time frequency analysis in this report was carried out with only the first 20 participants. EBR and behavioral analyses included all the subjects.
The entire preprocessing was carried out with the EEGLAB toolbox for matlab (seeDelorme & Makeig,2004). Epoch times were set from 2.5 seconds before until 1.5 seconds after a response was given, yielding an epoch length of 4 seconds. With a resampling rate of 512 Hz this results in an epoch length of 2048 points.
A high-pass filter at 0.5 Hz was applied to remove slow drifts and re-referencing was performed to the two earlobes. To remove ‘bad trials’, characterized by e.g. muscle artifacts, we visually inspected all epochs and then manually discarded bad ones. Afterwards, we ran an independent component analysis (ICA) to take out further artifacts that were mainly caused by eye-blinks and other clearly distinguishable non-brain related artifacts. Additionally, ICA decreases volume
conduction and helps to find single activation patterns of par-ticular sources (Makeig, Debener, Onton, & Delorme,2004). The ICA was performed by the EEGlab function pop runica() where after we again selected ‘bad components’ by visual inspection in order to subsequently remove them.
Time-frequency analysis
First, single-trial data was concatenated into one long time series to increase time-frequency decomposition performance and to produce cleaner code, avoiding an additional loop over trials. Then, the fast Fourier transform (FFT) of this time se-ries was taken. Subsequently, we performed a complex Morlet wavelet convolution where after the long time series was cut into single-trial data again. We used 30 frequencies increasing logarihtmically from 1 to 40 Hz. The number of wavelet cy-cles was selected to range from 3 to 10 changing as a function of frequency. Inverse FFT was applied to perform convolu-tion. Following, the squared magnitude of the outcome of the convolution was taken to derive frequency band-specific power. Due to the nature of the recorded trials, i.e. the fact that they did not contain a reliable pretrial baseline period, we had to perform a different baseline correction: we looped over all conditions and divided the time-frequency data pointwise by the average across the conditions for the time period in which we were interested (-1500ms - 0ms) with 0 being the point at which subjects pressed the SPACE button to type their answer. Finally, time-frequency data was separately assessed for frontal (Fpz, Fp1, Fp2, AFz, AF3, AF4, AF7, AF8, Fz, F1,
F2, F3, F4, F5, F6, F7, F8, FCz, FC1, FC2, FC3, FC4, FC5,
FC6), central (Cz, C1, C2, C3, C4, C5, C6, T7, T8, Pz, CPz,
CP1, CP2, CP3, CP4, CP5, CP6) and parietal (TP7, TP8, P1,
P2, P3, P4, P5, P6, P7, P8, P9, P10, POz, PO3, PO4, PO7, PO8,
Oz, O1, O2) regions as well as for the alpha (8.69 - 12.73 Hz)
and the theta (4.05 - 8.69 Hz) frequency band.
As opposed to simple Fourier transforms, convolution with complex Morlet wavelets comprises the temporal aspect of time-frequency representations of EEG data. This
charac-teristic emerges as a result of windowing sine waves with a Gaussian (= Morlet wavelets) that are then temporally loca-lized while ‘simple’ sine waves as used in Fourier transforms fluctuate over the entire time series and thus do not convey any temporal information (Cohen,2014). Moreover, wavelet convolution only assumes stationarity for the signal during the time period where the wavelet resembles a sine wave and not across the whole time series as is the case for standard Fourier transforms (Cohen,2014).
Time-frequency power statistics
Matlab (The MathWorks, Inc.) was used to perform statistical analysis and visualizations.
To carry out time-frequency power statistics we chose nonpara-metric permutation testing on a group-level. Advantages of nonparametric tests include the fact that they are (a) extremely general, (b) ensuring that the validity is not determined by the probability distribution of the analyzed data, and (c) capable of easily incorporating corrections for the multiple compar-isons problem (Maris & Oostenveld,2007;Cohen,2014). The multiple comparisons problem arises when evaluating effects with various tests across for example electrodes, fre-quency bands and/or time points and is particularly present for exploratory studies like the present one. Corrections for that can be performed with e.g. Bonferroni correction which is, however, really conservative, resulting in a very small p-value derived by dividing the ‘original, standard’ p-value by the number of comparisons. An additional problem with Bonfer-roni correction is that it assumes independency of the applied tests which is usually not the case for time-frequency analyses (Cohen, 2014). Nonparametric permutation testing on the other hand can quite easily integrate pixel- or cluster-based statistics that correct for multiple comparisons.
For our analysis, the p-value for the pixel-based statistics was set to 0.05, running a total of 1000 permutations. In each iter-ation subjects were randomly shuffled across conditions and t-maps of the corresponding null hypothesis were generated. The Z maps were then calculated by taking the ‘real’ t-values and subtracting the permuted t-values.
1.5 EBR measurement
To detect eye-blinks and horizontal eye movements facial electrodes have been used. Two vertical Ag-AgCl electrodes were applied above and below the left eye to detect eye blinks by measuring voltage differences between the two electrodes. Gazes to left and right were recorded with electrodes placed lateral to the external canthi.
EBRs during the five minutes resting condition were used to test the firsthypothesisthat assumes an ‘inverted U-shape relation’ between EBRs and flexibility scores and low EBRs to predict high levels of convergence.
As it was shown that EBR is stable during daytime but in-creases in the evening (Barbato et al.,2000), no recordings have been carried out after 6 pm.
2. Results
2.1 Behavioral results
Convergent items. Words that had the same ending as the example words in terms of either the last or the last two letters were categorized as convergent items. The average amount of convergent items for a subject per category was 4.406 (SD 2.610).
Divergent items. Words with a different word ending than the primes were classified as divergent items. On average, participants generated 1.999 divergent items per category (SD 1.306). The proportion of divergent answers per category is visualized infigure 3 A.
Fluency. Fluency was calculated by taking the number of generated names of one trial. The average fluency score per category and participant was 6.406 (SD 2.691).
Flexibility. Flexibility was defined as the number of pro-duced words with different endings. As an illustration, the flexibility score for a category with the answers samus, riada, segus, lamido, subridaand halegi would be 4. The average flexibility score per participant and category was 1.987 (SD 0.714) and is illustrated category-wise infigure 3 B.
Switches. The number of times participants alternated between different word endings was defined as switches. On average, subjects switched 2.110 times in one category (SD 1.201).
Repetitions. Correspondingly, category repetitions were labeled as the number of times that subjects generated names with the same ending in a row. On average, subjects exhibited 4.649 repetitions in one category (SD 2.309).
For all these measures we carried out a correlation analy-sis whose results are illustrated intable 1.
Positive significant correlations between flexibility and diver-gence scores (r = 0.95, p = 0.000) as well as between flexi-bility and switches (r = 0.96, p = 0.000) were found. These outcomes were not surprising because of following reasons: Divergent items were defined as words with endings that dif-fer from those of the primes. The probability that a subject generates a divergent answer and all successive ones within that category have the same ending is really low. Regarding the second significant correlation the explanation is even more simple: High flexibility is only possible if there are many switches since flexibility scores are calculated by taking the number of generated words with different endings.
Likewise the positive significant correlation between diver-gence and switches (r = 0.91, p = 0.000) as well as between convergence and repetitions (r = 0.88, p = 0.000) can be ex-plained. Also the significant negative correlation between
Figure 3. A. The proportion of divergent answers per category, averaged across subjects. B. The number of different word
endings per category, averaged across subjects, which refers to flexibility scores.
flexibility and convergence (r = -0.41, p = 0.14) and flexibility and repetitions (r = -0.41, p = 0.015), as well as between convergence and switches (r = -0.40, p = 0.019) and between switches and repetitions (r = -0.37, p = 0 .028) are obvious. Finally, we detected significant positive correlations between convergence and fluency (r = 0.88, p = 0.000) and fluency and repetitions (r = 0.76, p = 0.000) hinting on the fact that the number of generated words was generally high if they were convergent ones.
Approach and avoidance motivation
Furthermore we examined interaction effects in a repeated measures analysis of variance (ANOVA) with factors Motiva-tion (Approach vs. Avoidance), Thinking (Convergent vs. Di-vergent Thinking) and Switching (Repetitions vs. Switches). There were no main effects on Motivation, F(1, 35) = 2.737, p = 0.107, η2p= 0.073, and significant effects on Thinking, F(1,
35) = 15.590, p = 0.000, ηp2= 0.308, and Switching, F(1, 35)
= 11.958, p = 0.001, η2p= 0.255, indicating that there was a significantly higher proportion of convergent than divergent items as well as of repetitions as compared to switches. Motivation x Thinking. The Motivation x Thinking in-teraction, F(1, 35) = 4.363, p = 0.044, η2p= 0.111, revealed a
significant interaction effect. In both Motivation modes there were about two times as much convergent items than divergent ones. However, under approach motivation participants gener-ated significantly more convergent items than under avoidance motivation. Likewise, under avoidance motivation they pro-duced essentially more divergent items as compared to under approach motivation.
Motivation x Switching. Also there was a significant
ef-fect between Motivation and Switching, F(1, 35) = 19.112, p = 0.000, ηp2= 0.353, yielding more repetitions under
ap-proach and more switches under avoidance motivation. As in the Motivation x Thinking interaction, also here there were generally more repetitions under both thinking conditions. Thinking x Switching. Further, convergent thinking is signi-ficantly linked to more repetitions and divergent thinking to more switches, F(1, 35) = 103.846, p = 0.000, η2
p= 0.748.
Motivation x Thinking x Switching. The overall interaction effect between Motivation, Thinking and Switching, F(1, 35) = 6.935, p = 0.012, ηp2= 0.165, indicated that under approach motivation during convergent thinking there are more repe-titions produced than in the divergent thinking mode, while under avoidance motivation the proportion of switches in the divergent thinking mode is higher than it is under approach motivation.
Additionally, we carried out a correlation analysis to test for relations between approach and flexibility scores, r = 0.059, p = 0.734, and avoidance and flexibility scores, r = 0.086, p = 0.618, neither of them yielding significant results.
2.2 EEG results
Oscillation patterns related to thinking mode
Looking at the differences in convergent and divergent condi-tions averaged across all subjects and all electrodes gives the impression of several dissimilarities in power values at lower frequencies between -1500ms and -500ms.
cor-Figure 4. Time-frequency plots of power averaged across all electrodes illustrating differences between convergent and
divergent thinking. Panel A shows the unthresholded Z map and panel B the uncorrected thresholded Z map (p < 0.05 uncorrected; suprathreshold pixel clusters are idicated by contour lines). In panel C the pixel-corrected Z map can be seen, showing significant (p < 0.05 corrected) dissimilarities around -1000ms until -750ms and -1200ms until -1100ms as well as about 150ms later at very low frequencies (1-2 Hz) and quite high ones (36-40 Hz), respectively.
rections (p < 0.05), there are indeed two small ‘clusters’ that differ significantly between the two conditions which can be seen in figure 4. Higher power values for the lower delta range (around 1-2 Hz) can be observed between -1000ms and -750ms in the convergent thinking condition while dis-crepancies in higher frequencies (36-40 Hz) are visible from -1200ms until -1100ms as well as shortly afterwards.
Oscillation patterns related to motivation, different regions, repetitions and switches and the alpha band
Comparing oscillation patterns of approach and avoidance motivation did not reveal any significant differences in power. Also contrasting power values of different regions (frontal, central and parietal) as well as of hemispheres of the two conditions did not yield any meaningful differences.
Likewise there were no significant differences between switches and repetitions and in the alpha band between conditions, re-gions or hemispheres.
Theta band
However, in the theta band significant dissimilarities between convergent and divergent thinking in all regions separately as well as averaged across all regions could be detected.Figure 5provides plots from unthresholded Z map (panel A), uncor-rected thresholded Z map (p < 0.05 uncoruncor-rected) (panel B) and pixel-based corrected Z map (p < 0.05 corrected) (panel C) of theta band power in convergent vs. divergent ideation across all conditions and subjects and uncovers higher power in the lower theta band for convergent ideation.
Taking a closer look at theta band power and hemispheric differences reveals a temporally slighter larger difference in power for the left hemisphere. Also left-sided as compared to right-sided theta power captures a bigger range of the fre-quency band: For the left side, the difference in power spreads from around 4.05Hz to 6Hz as compared to 4.05Hz to 5.2Hz for the right hemisphere (Figure 7).
Taking a closer look at detected differences in hemispheric theta, we observed that the bigger range of theta power re-garding frequencies comes from central regions: In central left regions, theta power differences between convergent and divergent thinking processes were prominent from 4.05 - 6 Hz whereas in frontal and parietal left regions those differences were only visible between 4.05 and 5.2 Hz.
2.3 EBR analysis results
EBR analysis was performed for 35 subjects from the origi-nally 37. Additioorigi-nally to one subject that was excluded for the whole study due to an insufficient amount of artifact-free data to analyse, one participant had to be excluded from the EBR analysis since this person had, despite the contrary instruction, his eyes closed during the measurement.
We defined an eye-blink as a voltage change of 100 µV in the vertical electro-oculogram within a time interval of 500ms. The total amount of measured eye-blinks was then divided by the duration of the resting period which was 5 minutes long. The mean EBR per minute across subjects during the resting period was 15.26 (SD 12.02) ranging from a minimum EBR of 2.29 to a maximum EBR of 47.49 per minute. In view of those two values it can be seen that there are huge interindi-vidual differences in EBRs which is in line with previous EBR studies (Bentivoglio et al.,1997;Karson,1987).
Figure 6shows a plot of EBRs per minute in relation to fle-xibility. SPSS curve fitting procedure was used to generate linear (R2= 0.011) and quadratic (R2= 0.011) best fit for EBRs vs. flexibility scores
Correlation coefficients and significance levels of the 7 fac-tors flexibility, convergence, divergence, switches, fluency, repetitions and EBR are depicted intable 1. No significant correlations between EBR and any of the factors was found.
Figure 5. Time-frequency plots of power averaged across all electrodes illustrating differences between convergent and
divergent thinking in the theta band. Panel A shows the unthresholded Z map and panel B the uncorrected thresholded Z map (p < 0.05 uncorrected; suprathreshold pixel clusters are indicated by contour lines). In panel C the pixel-corrected Z map can be seen, showing significant (p < 0.05 corrected) dissimilarities around 1000ms until 780ms in the lower theta band (4.05 -5.2 Hz).
FLEX CON DIV SWI FLU REP
FLU .76** SWI .05 -.37* DIV .91** .29 -.21 CON -.20 -.40* .88** .88** FLEX -.41* .95** .96** .06 -.41* EBR R .10 .04 .11 .14 .10 -.05 L .01 .002 .01 .02 .09 .002 Q .01 .01 .02 .02 .1 .05
Table 1. Correlation coefficients (R) and significance levels
(** p < .01, * p < .05) between spontaneous EBR and scores of flexibility (FLEX), convergence (CON), divergence (DIV), switches (SWI), fluency (FLU) and repetitions (REP). Also R2values of linear (L) and quadratic (Q) fits of EBR as a predictor of FLEX, CON, DIV, SWI, FLU and REP are given.
3. Discussion
This study sought to reveal EEG patterns during convergent and divergent thinking by using the pasta task. In this task sub-jects were asked to generate new names in various categories after they got presented three example words. Generated answers were then categorized as convergent or divergent de-pending on word endings. In addition, we measured EBRs and included approach and avoidance motivation as an extra manipulation.
Analysis of EEG patterns uncover some significant differ-ences between convergent and divergent thinking: Generally around 1000ms before the response period (typing a word) there are essential discrepancies in very low (1 - 2 Hz) and quite high (36 - 40Hz) frequency bands between the two con-ditions.
Further significant differences were found in the lower theta
Figure 6. Flexibility scores as a function of spontaneous
EBR per minute. Regression lines for linear (black) and quadratic fit (red) are given.
range (4.05 - 5.2 Hz) at around -1000ms until -780ms between the two conditions for frontal, central and parietal regions sep-arately as for all regions together. Moreover, we found that these theta changes were in a bigger frequency range promi-nent for the left hemisphere (4.05 - 6 Hz). Going into further detail with theta activity revealed that these more ample power differences arise from left central regions. Only there, and not in frontal and parietal regions, theta power differences between the two conditions ranged from 4.05 to 6 Hz and not only between 4.05 and 5.2 Hz.
of produced convergent items than divergent ones indepen-dently of manipulation of motivation. Repeated measures ANOVA to test for interaction effects between motivation, thinking mode and switching behavior show a slightly higher rate of divergent items and switches generated under avoi-dance as compared to divergent items and switches produced under approach motivation. However, we failed to find any sig-nificant correlation between motivation and flexibility scores. Analyzing EBRs in relation to the factors flexibility, conver-gence, diverconver-gence, switches, fluency and repetition yielded some positive and negative significant correlations. On the whole, these reliable relationships confirmed the validity of our definitions of measures. The positive correlation between flexibility and switches for example arises due to the fact that flexibility scores are calculated by taking the number of gene-rated words with different word endings which evidently also results in a corresponding number of switches. The positive relationship between convergence and fluency indicates that the number of produced words was generally high if they were convergent.
No relationship between EBR and motivation and EBR and other measures
Correlation and analyses for best fit (linear and quadratic) between EBR and the factors flexibility, convergence, diver-gence, switches, fluency and repetitions did not yield any significant results. The fact that we did not find a quadratic relationship between EBR and flexibility, as was however hypothesized, was quite surprising since there are studies (Chermahini & Hommel,2010,2012) that reported a signifi-cant inverted U-shape relation between those two. One aspect that could possibly explain the fact of our failing to find a quadratic relationship between EBR and flexibility scores might lie in the nature of the applied task. WhileChermahini and Hommel(2010) employed the alternative uses task (AUT) to assess divergent thinking and flexibility levels we utilized the pasta task as one single task to differentiate between con-vergent and dicon-vergent ideation. In the AUT open-ended ques-tions are posed that require multiple answers, more speci-fically subjects are asked to list as many alternate uses for an object as they can think of within a specific time window (Guilford,1967). The number of different categories used for every item then yields the flexibility scores. As stated above in theresults section, we calculated flexibility scores for the pasta task by taking the number of produced words with different word endings.
A further big difference between the AUT and the pasta task is the ‘degree’ of restrictiveness. Compared to a more ‘free-association’ task like the AUT, the pasta task is more limited in terms of connecting distant ideas. Moreover, due to the fact that we provided the subjects with three primes before starting their thinking period, the participants might in the first place be quite biased to produce more convergent items, following the example words, and thus exhibit generally lower flexibility scores that might just be not prominent enough to establish
any relationship with EBRs.
No pronounced influence of motivation on behavioral task performance
Following this latter remark of having showed primes, it is not further surprising that significantly more convergent than divergent items were produced across all categories. What was surprising is that under avoidance motivation there was a slightly higher proportion of produced divergent items and switches than under approach motivation. Together with the interaction effect between motivation, thinking mode and switching, yielding similar results, i.e. a higher amount of switches in the divergent thinking mode under avoidance as compared to under approach motivation, we can follow that in this study avoidance motivation can be slightly more linked to more divergent, flexible items and switching than approach motivation.
Additionally, we were not able to find a significant relation-ship between flexibility and approach motivation but contrary even a slightly tighter, but far from significant, relationship between avoidance motivation and flexibility. This latter out-come actually opposed one of the twohypotheses, i.e. higher flexibility scores under approach as compared to avoidance motivation which was based on findings byFriedman and F¨orster(2001,2002). Taken together, these latter inferences are quite unexpected and require some explanation.
Again, one potential explanation might be the relatively small number of participants we had for our study, though this rea-son might not be a panacea to explain away all unexpected findings.Friedman and F¨orster(2001), on whose conclusions the hypothesis of a positive correlation between approach mo-tivation and flexibility scores was based, used a total of 90 subjects for their study whileFriedman and F¨orster(2002) had 30 participants for a sub-study on creative insight problem-solving where they evaluated task performance and not flex-ibility scores per se. In a further sub-study, reported in the same paper (Friedman & F¨orster,2002), they had 26 subjects to perform the AUT task and found a positive relationship between promotion motor actions and task performance, im-portantly not regarding flexibility scores but regarding the level of creativity of the given answers rated by independent scorers. Thus,Friedman and F¨orster(2001) andFriedman and F¨orster(2002) deducted from findings about ‘better’ creative task performance that promotion cues lead to a higher risk behavior as compared to a more perseverant processing style which does not necessarily imply higher flexibility scores as were used in this study. Referring to the potential explanation given above about EBRs and flexibility, the issue is exactly the same withFriedman and F¨orster(2001)’s andFriedman and F¨orster(2002)’s study: the nature of the applied task. Hence, the general failing to find links between EBRs or motivation manipulations and any of the creativity factors in this study might purely lie in the character of the pasta task, particularly in its above described restrictiveness.
Figure 7. Pixel-corrected Z maps of theta power difference between convergent and divergent thinking periods for left (panel
A) and right (panel B) hemispheres. Left-hemispheric theta is slightly more spacious in the temporal domain and shows a significant power difference from convergent as compared to divergent ideation for a larger frequency range (4.05 - 6Hz for the left side as compared to 4.05 - 5.2Hz for the right side.
show a positive relationship between approach motivation and flexibility scores is the omission of controlling for emotion since it is known that especially hedonic tone influences cre-ative processing (De Dreu et al.,2008;Baas et al.,2008,2011) which was not examined in the present study and thus might have influenced results.
A final potential reason might be that our applied manipula-tion was not strong enough and would possibly have worked better if e.g. a higher amount of money could have been lost or won. An analysis of the relationship between manipulation of motivation and motivation ratings that have been given after each block could shed some light on that matter.
Additional correlation analyses between the six factors flexibili-ty, convergence, divergence, switches, fluency and repetition did not produce any surprising or interesting results that need some explanation or require to be discussed in more detail (seeresults).
Higher theta power over frontal and all other regions du-ring convergent as compared to divergent task
Our finding of significantly higher theta in convergent as compared to divergent ideation over all regions fits reports from previous studies investigating those two types of think-ing. Correspondingly,M¨olle et al.(1996) found higher theta power over frontal regions during convergent as compared to divergent thinking and mental relaxation (see alsoM¨olle, Marshall, Wolf, Fehm, & Born,1999) and furthermore signifi-cantly higher theta power over central regions in convergent as compared to divergent ideation. Notably, they used very different tasks to assess convergent and divergent processes, an arithmetical thinking task and a ‘stories’ task proposed by
Guilford(1977), respectively.
This higher theta power in convergent as compared to di-vergent thinking periods might be explained by its apparent
relation to working memory which is also supported by stud-ies byJensen and Tesche(2002) and a review about working memory and EEG theta oscillations bySauseng, Griesmayr, Freunberger, and Klimesch(2010). Accordingly,Klimesch, Doppelmayr, Pachinger, and Ripper(1997),Klimesch, Dop-pelmayr, Schimke, and Ripper(1997) andKlimesch(1999) detected larger theta power for good working memory perfor-mance and related that finding to working memory demands, memory retrieval (Klimesch et al.,2001) and the encoding of new information. In their studies, subjects had to view words and recall them. The more words they remembered the higher theta power was during encoding those words and, importantly especially for the comparison to our findings, also during recalling them. Relating those results to ours, one could argue that during convergent thinking, where we found higher theta power, participants had higher working memory demands since they recalled primes and subsequently gene-rated words with similar endings.
The finding of more spacious theta power differences between convergent and divergent thinking periods in terms of time (which though was really small) and frequency range at left-hemispheric (4.05 - 6Hz) as compared to right-left-hemispheric (4.05 - 5.2Hz) sites might be explained by the verbal nature of the task. Accordingly, a PET study bySmith, Jonides, and Koeppe(1996) showed that a verbal memory task as com-pared to a spatial one elicited more pronounced activity in the left hemisphere which is generally linked to language (Frost et al.,1999). Additionally, we found that the larger frequency range of theta power differences in the left hemisphere arise specifically in left central regions. Left frontal and left pari-etal regions exhibited same power differences regarding the frequency span as were detected in right-hemispheric regions.
