Title: The domain generality-specificity dilemma re-assessed: a systematic review on the neural foundations of creativity Date of submission: 22-12-2019
Name student and ID: Fabienne Q. van Rossenberg1, 10530843 Name supervisor: dr. Claire E. Stevenson2
Name co-assessor: dr. Matthijs Baas3
1 University of Amsterdam, Msc. in Brain and Cognitive Sciences
2 University of Amsterdam, Department of Psychology, PO Box 15906, 1001 NK Amsterdam 3 University of Amsterdam, Department of Psychology, PO Box 15919, 1001 NK Amsterdam
Abstract
Many neuroscientific reviews of creativity have tried in vain to answer the question of whether creativity is domain-general or domain-specific. Very little evidence of overlapping brain regions across creativity domains was observed by them, causing the neural mechanisms underlying creativity to remain poorly understood. Almost all these reviews base their results on studies that used divergent thinking or insight as measures of creativity. However, both divergent thinking and insight tests have been considered domain-general tests, which makes them a biased source when studying the
generality-specificity dilemma. Therefore, the current review aimed to provide a systematic review of neuroimaging and neuroelectric studies assessing creativity from multiple domains, such as creative writing, visual arts and musical creativity, whilst deliberately excluding studies on divergent thinking and insight. A total of 41 studies were included in this review belonging to a variety of creativity domains, as derived from the creative achievement questionnaire (CAQ).
The results of this review portrayed a highly variegated landscape of reported brain regions, both within and across creativity domains. Among the most consistent findings for neuroimaging studies was the involvement of brain regions belonging to the executive control network and the default-mode network for creative compared to control tasks, though lateralization differed between studies. Neuroelectric studies most consistently reported an association between frontal alpha-band activity in creative compared to control tasks, though the direction of activity (increased or decreased alpha activity) varied between studies. These findings are in line with results from studies assessing divergent thinking, insight problem solving and intelligence. Although methodological artifacts within and across creativity domains compromise drawing reliable conclusions, the outcome of this review suggests that a specific creativity network does not exist, but rather that creativity is dependent on general cognitive processes.
1. Introduction
The question of whether creativity, the ability to produce novel and useful ideas or products (Stein, 1953; Martindale, 1999; Runco & Jaeger, 2012), is domain-general or domain-specific is a hot topic of debate (Baer, 2010; 2012; 2019; Baer & Kaufman, 2005; Barbot, Besançon, & Lubart, 2016; Lubart & Guignard, 2004; Qian, Plucker & Yang, 2019). The domain-general perspective argues that creativity resembles a general ability, like intelligence, which suggests that creative performance in one domain (e.g., visual) would predict that in other domains (e.g., verbal, musical). This is because according to these theories, creative outcomes in different domains rely on the same general creative processes, such as the ability to shift perspectives and think divergently (Beaty, Silvia, Nusbaum, Jauk, & Benedek, 2014). In contrast, the domain-specific perspective argues that people can be highly creative in one domain, but not necessarily in other domains. According to these theories, creative outcomes in a domain rely on expertise and processes that are highly specific for that domain (Baer, 2012; 2015). Knowing whether creativity is domain-general or domain-specific has important implications for creativity theory, measurement, and training.
There are many inconsistent findings that either favor the domain general or the domain specific perspective. For example, behavioral evidence supporting domain-generality of creativity has come from studies which showed relatively high inter-correlations between individual creativity performance indicators in different domains (e.g., Chen, Himsel, Kasof, Greenberger, & Dmitrieva, 2006; Runco, 1987). A study showing that performance on verbal and figural divergent thinking as measured with the Torrance test of creative thinking (TTCT; Torrance, 1966) predicted creative achievements in other domains, provide further support (Plucker, 1999). Furthermore, research has shown that creativity training in one domain can enhance creative performance in other domains (Fink, Graif & Neubauer, 2009; Scott, Leritz, & Mumford, 2004; Sowden, Clements, Redlich & Lewis, 2015), although the transfer to real-world creativity remains small (Zeng, Proctor, & Salvendy, 2011).For example, Sowden et al. (2015) found that children who took part in improvisational dance or improvisational verbal and acting games performed better on two divergent thinking tasks and a creative ‘toy’ design task. In addition, professional dancers produced a larger number of alternative uses for objects, as measured with the Alternative Uses Test (AUT; Guilford, 1967), compared to novice dancers (Fink, Graif, & Neubauer, 2009).
Meanwhile, a growing body of research has reported results that favor domain-specificity. For instance, multiple behavioral studies show little to no relationship of individual creative performance indicators across domains (Baer, 1991, 1993; Kaufman & Baer, 2004; Han & Marvin, 2002; Runco, 1987) or even within domains (e.g., poetry vs. story writing, Baer, 1994a; painting vs. drawing, Conti, Coon, & Amabile, 1996). Han (2003) reported that children’s divergent thinking ability, measured with the Wallach-Kogan Creativity Test (Wallach & Kogan, 1965) and the Real World Divergent Thinking Test (Okuda, Runco, & Berger, 1991), was not related to creative performance in other
domains, including language, art and mathematics. In addition, studies by Baer (1994b; 1996) showed that creative writing training for poetry did not enhance creative story writing, suggesting that creative ability does not even transfer within the verbal domain.
Besides behavioral research on the generality-specificity dilemma, scientists put much effort into unraveling the neural correlates of creativity. Previous studies have examined whether
neuroimaging and neuroelectric techniques could find a consistent neural basis for creativity across domains, as this would suggest creativity is a domain-general construct. Beaty, Benedek, Silvia and Schacter (2016) reviewed brain activity and connectivity studies in relation to creative tasks across domains, including divergent thinking, poetry composition and musical improvisation. They concluded that an interaction between two brain networks, the default mode network and executive control network, was present for creativity tasks compared to control tasks in different domains. Further research showed that functional coupling of the two large-scale networks predicted individual differences in performance on the AUT, as well as creative achievement (Beaty et al., 2015; 2018). Observed brain areas included the medial prefrontal cortex (mPFC) and posterior cingulate cortex (PCC) for the default mode network, and the dorsolateral prefrontal cortex (DLPFC) and the anterior cingulate cortex (ACC) for the executive control network. The involvement of the PFC during creativity tasks from multiple domains has received much support from other reviews on
neuroimaging studies (Arden, Chavez, Grazioplene, & Jung, 2010; Boccia, Piccardi, Palermo, Nori, & Palmiero, 2015; Dietrich, 2004; Dietrich & Kanso, 2010; Gonen-Yaacovi et al., 2013; Shen et al., 2016; 2018). These results suggest that the PFC plays a key role in creative behavior. However, the precise PFC regions have been found to differ according to domains and task-specific factors (Arden et al., 2010; Boccia et al., 2015; Dietrich & Kanso, 2010; Gonen-Yaacovi et al., 2013; Shen et al., 2016; 2018). For example, a review from Boccia et al. (2015) reported consistent activation of right DLPFC for the musical and visuo-spatial domains and left DLPFC for the verbal domain, whereas the right inferior frontal gyrus (IFG) was found to be consistently activated during verbal and visuo-spatial domains, but not the musical domain. Similar mixed findings have been reported on the involvement of parietal and temporal regions which are connected to frontal areas, such as the inferior parietal lobule (IPL) and the superior temporal gyrus (STG; Boccia et al., 2015; Dietrich & Kanso, 2010; Gonen-Yaacovi et al., 2013).
Another brain measure which has been associated with creativity is alpha power (or alpha synchronization). Alpha-band activity, measured by electroencephalography (EEG), represents a frequency range between 8-13 Hz and is generally thought to reflect a state of minimal arousal during which a person is awake, but relaxed. Fink and Benedek’s (2014) review of EEG measures of creative thinking reported increased alpha power as one of the most consistent findings in neuroscientific research on creativity. In stark contrast, Dietrich and Kanso (2010) reported that findings of increased alpha power were heavily fragmented among studies on divergent thinking, artistic creativity and studies on insight. Similarly, Boot, Baas, Mühlveld, de Dreu and van Gaal (2017) found decreased
delta power, rather than increased alpha power, during divergent thinking compared to a convergent thinking control task.
Altogether, both brain and behavioral studies regarding the domain generality-specificity dilemma have shown incredibly mixed results. One of the main difficulties in studying this dilemma stems from creativity’s multifaceted nature. Although creativity could be simply defined as the ability to create novel and useful outcomes (Runco & Jaeger, 2012; Stein, 1953), it remains a complex construct which can manifest itself in many different ways. As a result, researchers have tried to decompose creativity into subtypes in order to make it tractable. The single most studied creativity subtype is divergent thinking, defined by Guilford (1950; 1967) as the ability to generate multiple solutions to an open-ended problem. Another frequently studied subtype of creativity is insight problem solving. Insight is a sudden comprehension of the solution of a problem, also referred to as the “Aha! moment”, that can be the result of approaching the problem from different and unexpected angles (Sternberg & Davidson, 1995). However, the strong focus on divergent thinking and insight in creativity research provides a biased information base to resolve the generality-specificity dilemma. Dietrich (2007a; 2015; 2019) argues that divergent thinking represents a false category formation, because the exact opposite of divergent thinking, convergent thinking, can also lead to creative products (Runco, 2004; Simonton, 2015). As Dietrich (2019, page 2) reported: “If both divergent and convergent thinking can lead to both creative and non-creative thinking, divergent thinking is
incapable of identifying the processes that turn normal thinking into creative thinking. The treatment and the control condition cannot contain the same variable.” Furthermore, divergent thinking involves many different mental processes such as (working) memory, cognitive control, attentional flexibility and spontaneous thought (Beaty, Benedek, Kaufman, & Silvia, 2015; Nijstad & Stroebe, 2006; Sowden, Pringle, & Gabora, 2015), which makes it a compound construct, just like creativity. Similarly, it has been suggested that creative insight combines elements of divergent and convergent thinking and that it involves of heterogeneous set of cognitive processes (Shen et al., 2016; de Dreu, Nijstad, & Baas, 2012). In line with this, some researchers refer to divergent thinking and creative insight as being domain-general tasks (Beaty et al., 2016). Consequently, almost all reviews and meta-analyses aiming to find a neural basis of creativity based their results on studies that used divergent thinking tasks (Arden et al., 2010; Beaty et al., 2015; Boccia et al., 2015; Dietrich, 2004; Dietrich & Kanso, 2010; Fink & Benedek, 2014; Gonen-Yaacovi et al., 2013; Mihov et al., 2010) or insight tasks (Dietrich & Kanso, 2010; Kounios & Beeman, 2014; Shen et al., 2016; 2018) as measures of
creativity. As a result, their arguments are one-sided and conclusions perhaps biased. In order to assess whether the neural correlates of creativity support the domain-generality or domain-specificity theory, new reviews are needed in which divergent thinking and insight studies are not taken into account. Therefore, the aim of this study was to provide a systematic review of neuroimaging and neuroelectric studies assessing creativity, such as creative writing, visual arts and musical creativity, but not
For this review, we considered the following neuroimaging techniques. Functional magnetic resonance imaging (fMRI), structural magnetic resonance imaging (sMRI), diffusion tensor imaging (DTI), positron emission tomography (PET) and functional near-infrared spectroscopy (fNIRS).fMRI provides a measure of brain activity over time using a magnetic field to measure the ratio of
oxygenated blood to deoxygenated blood, generally referred to as the blood oxygen level dependent (BOLD) signal. This technique assumes that when neuronal activity increases in a certain brain area, blood flow rich with oxygen towards that area increases more rapidly than neurons are able to use the oxygen. This results in more oxygenated blood compared to deoxygenated blood, thus showing an increase in BOLD signal. However, peak BOLD levels occur several seconds after the increase of neuronal activity, causing fMRI to provide poor temporal resolution, but high spatial resolution. In contrast to fMRI, sMRI provides a static anatomical image of the brain. Morphological information can be used to relate behavior to brain anatomy or assess anatomical differences between different groups of participants. DTI provides estimates of location, orientation and anisotropy of white matter tracts in the brain, thereby elucidating neuronal networks across the brain. Similar to fMRI, both PET and fNIRS assume that activated neurons require more oxygen and therefore excite greater blood flow toward those regions. PET uses a radioactive tracer in the bloodstream in order to measure regional cerebral blood flow (rCBF), for which more radiation equals more blood flow. In contrast, fNIRS uses the different light absorption spectra within the near-infrared range (700-1000 nm) of oxygenated and deoxygenated blood to monitor changes in blood flow. A drawback of fNIRS is that its measurement is mostly limited to the prefrontal cortex.Furthermore, we included four neuroelectrictechniques, which were electroencephalography (EEG), magnetoencephalography (MEG), event-related potentials (ERP) and transcranial direct-current stimulation (tDCS). EEG and MEG measure cortical
electromagnetic fields generated by an increase or decrease in neural activity. When a large number of parallel aligned neurons receive signals in the form of electrical impulses at the same time, an
electromagnetic field is created which can be detected by sensors placed on the scalp. However, the more distant neurons are located from the scalp, the weaker the electrical signal. This makes EEG and MEG mostly sensitive to cortical regions, but not subcortical regions. The difference between EEG and MEG is that EEG can detect neural activity in cortical sulci (the grooves in the surface of the cerebral cortex) as well as the top of cortical gyri (the ridges of the folds in the surface of the cerebral cortex), whereas MEG is most sensitive to activity in sulci only. From EEG and MEG recordings it is possible to isolate very specific electrophysiological brain responses to a single stimulus or event, called event related potentials (ERPs). This type of analysis allows for the extraction of highly specific sensory, cognitive and motor events compared to general EEG analysis, which focuses more on larger brain processes. In contrast to MRI techniques, EEG, MEG and ERP techniques provide excellent temporal resolution, but rather poor spatial resolution. Last but not least, tDCS is used to modulate neural activity by delivering a low positive (anodal) or negative (cathodal) electric current to the scalp. The delivered current is not strong enough to trigger action potentials in neurons, but rather works by
strengthening or weakening synaptic transmission between neurons in specified brain areas. It becomes possible to identify whether facilitation or disruption of activity in the target area influences behavioral outcomes, thereby causally showing its involvement in cognitive processes, such as creativity.
The names of the neuroscientific methods combined with the keyword “creativity” were used to identify published reports by conducting an online database search. From the output, studies were selected based on inclusion and exclusion criteria in order to find evidence in favor of generality, specificity or both. We considered various results to be supportive of domain-generality. For instance, when brain areas or networks are consistently reported across creativity domains, but these regions are not related to other general cognitive processes (e.g., executive control, fluid intelligence); when similar brain areas or networks are reported across studies irrespective of task-related differences and when creativity ratings based on subject’s performance (e.g., CAT; Amabile, 1982) positively or negatively correlate with similar brain areas or networks across studies. Consequently, we consider opposing results for these three examples to reflect evidence in favor of domain-specificity. That is, if brain areas or networks fail to show similarities across domains, tasks and creativity ratings. In addition, previous research has shown a clear link between domain-specific creativity and expertise (Baer, 2015). If studies report differences in brain areas or networks across domains for experts compared to non-experts, we regard this as supporting domain-specificity. Furthermore, we consider the possibility that results support both generality and domain-specificity, which may suggest that creativity emerges from an interplay of domain-general and domain-specific brain areas or networks, as has been previously suggested by Barbot, Besançon and Lubart (2016).
2. Method
We conducted and report our systematic review of the literature on neuroimaging and neuroelectric studies on creativity following the PRISMA guidelines (Liberati et al., 2009). The literature search was carried out using four online bibliographical databases with the Ovid interface: PsycINFO, Medline, Embase and ERIC (no time restrictions, all databases). Search queries were constructed by combining keywords and subject headings (tailored towards each database) related to creativity and brain measures, including neuroimaging techniques such as fMRI, sMRI, DTI, PET, NIRS, and neuroelectric techniques such as EEG, MEG, ERP and tDCS. The Boolean operator AND was used to combine the search queries. For instance, for the PsycINFO database, the search query consisted of the following keywords and Boolean connectors: TS = (creativ*.ti,ab,id AND functional magnetic resonance imaging.ti,ab.id). For a complete overview of the search query that was used per database, see Appendix A.
After deduplication, the search yielded a set of 2,244 articles. These articles were then screened for inclusion or exclusion based on the following a priori inclusion criteria: (1) the paper reported an empirical study; (2) participants were not from a clinical population; (3) the study included a neuroscientific method; (4) neural activation was correlated with creative performance on a
behavioral task that did not solely measure divergent thinking, such as the AUT (Guilford, 1967), the TTCT (Torrance, 1966) or the Wallach-Kogan test battery (Wallach & Kogan, 1965); (5) neural activation was correlated with creative performance on a behavioral task that did not solely measure creative insight, such as the Remote Associates Test (Mednick, 1968) or match stick arithmetic task; (6) the title, abstract or keywords contained the words ‘creativity’ or ‘creative’ and the name of a neuroscientific method or subsequent brain measure; (7) the paper was written in English. With the use of these criteria 130 articles were selected.
After the screening phase, full texts of the remaining reports were read in order to further select papers for the systematic review. We included tasks that were related to the domains assessed in the Creative Achievement Questionnaire (CAQ; Carson et al., 2005). The CAQ provides a measure of individual creativity by assessing the quantity and quality of creative products generated by the individual within the domains of visual arts, music, dance, architectural design, creative writing, humor, inventions, scientific discovery, theatre/film and culinary arts. Therefore, the inclusion criteria for the selection phase, in addition to the inclusion criteria of the screening phase, were as follows: (8) the task that was used to measure creativity must entail one of the domains from the CAQ. ; (9) the brain measure must be acquired whilst participants performed the creative task. This was done in order to obtain more direct neural measures of creative behavior and to avoid studies that assess creativity with correlational analyses only. However, exceptions were made for structural imaging studies since this technique does not allow for direct brain measures of creativity; (10) statistics on the relation between a creativity measure and brain measure must be available; (11) the study must include at least
eight or more participants; (12) if the paper reports a study in which the experimental condition aimed to measure creativity, a relevant control condition must be used. For example, if an fMRI study aimed to measure creative melodic improvisation during the experimental condition as opposed to staring at a fixation cross as the only control condition, no conclusions can be derived regarding the creativity element. Instead, the control condition should, for instance, contain the performance of pre-learned melodies. This way, brain activity related to motor execution and audio/syntactic processes is controlled for. Altogether, the literature search resulted in a final selection of 41 articles. See Appendix B for an overview of the process according to the PRISMA 2009 flowchart diagram for systematic reviews (Moher et al., 2009).
3. Results
The final selection consisted of 41 studies using different neuroscientific methods. The majority of studies (20 out of 41) used fMRI to obtain brain measures related to creativity, followed by ten studies using EEG, six studies using sMRI and two studies using tDCS. In addition, PET, fNRIS and MEG were each represented by one study. See Table 1 for a complete overview of the included studies. Data was extracted in order to find out whether the results provided support for domain generality, domain specificity or both.
For this review, articles were sorted per domain. Analyses showed that sixteen studies represented the musical domain, followed by nine studies in the visual arts domain and seven studies in the creative writing domain. The scientific discovery domain and dance domain each included two studies and the humor and inventions domains each included one study. No studies were present for the remaining three domains, which are architecture, theatre/film and culinary arts. In addition, three structural studies which used the CAQ as a measure of creativity were included. Per domain, results from neuroimaging studies and neuroelectric studies shall be discussed in light of the domain-generality/specificity dilemma.
3.1. Musical domain
All studies within the musical domain used musical improvisation in order to study creativity. However, musical improvisation tasks differed between studies. Participants were instructed to improvise various kinds of music (e.g., classical, jazz, freestyle rap) by playing on a simple scanner-proof piano keyboard for neuroimaging studies or via mental imagery for neuroelectric studies. Other differences between tasks consisted of type of improvisation (e.g. rhythmic improvisation, melody improvisation, or both), type of control conditions (e.g. sight-reading or playing a pre-learned melody), and the amount of freedom regarding improvisations (e.g. completely free, according to a certain note-sequence, or improvisations based on a previously heard stimulus). In total, ten studies on musical creativity used fMRI, three studies used EEG, two studies used tDCS and one study used MEG.See Table 1 for a complete overview of the studies included in the musical domain.
3.1.1.Creative versus control tasks
Across musical improvisation studies, the involvement of the dorsolateral prefrontal cortex (DLPFC) was frequently reported. The DLPFC has been functionally implicated in many different cognitive abilities, including working memory (Curtis & D’Esposito, 2003), cognitive flexibility, inhibition, and motor planning. Eight out of ten studies using fMRI reported changes in DLPFC activity during improvisation compared to control conditions (Bengtson et al., 2007; De Manzano & Ullén, 2012a; 2012b; Dhakal et al., 2019; Donnay et al., 2014; Liu et al., 2012; McPherson et al., 2016; Villareal et al., 2013). From the remaining two fMRI studies that did not mention DLPFC
activity, one study solely reported neural differences of the creative versus control task between expert and non-expert pianists (Berkowitz & Ansari, 2010), which may explain the absence of DLPFC activity. Although the DLPFC seems consistently involved during musical creativity, the direction of activity (increased versus decreased activity) and lateralization differed between studies. For example, one study reported increased activity in the right DLPFC during improvisation (Bengtsson et al., 2007), whereas another study reported decreased activity in the right DLPFC (Liu et al., 2012). Moreover, one study reported decreased bilateral DLPFC activity during improvisation (McPherson et al., 2016), whereas three studies reported increased activity in bilateral DLPFC (De Manzano & Ullén, 2012a; Villareal et al., 2013; Donnay et al., 2014) and two studies reported increased activity in the left DLPFC (de Manzano & Ullén, 2012b; Dhakal et al., 2019). Although lateralization and the direction of activity did not complement one another across studies, one tDCS study provided causal evidence for the involvement of the right DLPFC during creative improvisation (Rosen et al., 2016). The study reported that the causal link between the right DLPFC and creative performance was moderated by the amount of musical experience. They showed that anodal tDCS of the right DLPFC improved the quality of improvisations for the least experienced musicians relative to sham, whereas both anodal and cathodal stimulation of the right DLPFC hindered performance for the
most-experienced musicians (professional jazz pianists).
Besides DLPFC activity, all studies using fMRI reported the involvement of one or more components of the motor cortex during improvisation compared to control conditions, apart from one study (Berkowitz & Ansari, 2010). Motor cortex areas included the primary motor cortex (M1), premotor cortex (PMC) and supplementary motor area (SMA). All fMRI studies consistently reported increased activity in motor areas as opposed to decreased activity, though differences in lateralization were still present. For example, musical improvisations were related to activations in the right pre-SMA only (Bengtsson et al., 2007), left pre-pre-SMA only (Dhakal et al., 2019; Liu et al., 2013) or bilateral SMA (de Manzano et al., 2012a; 2012b; Donnay et al., 2014; McPherson et al., 2016;
Villareal et al., 2013). In addition, causal evidence for the importance of M1 for musical creativity was provided by a tDCS study (Anic, Olsen, & Thompson., 2018). The results showed that the group of jazz pianists who received excitatory tDCS on bilateral M1 generated significantly more creative improvisations compared to jazz pianists who received inhibitory tDCS.
Furthermore, seven studies using fMRI reported increased activity during improvisation in the inferior frontal gyrus (IFG), an area which has been associated with musical perception (for review see Koelsch & Siebel, 2005) and response inhibition (Hampshire, Chamberlain, Monti, Duncan & Owen, 2010). These activations were either centered in the left IFG (Berkowitz et al., 2008; Dhakal et al., 2019; Liu et al., 2012; McPherson et al., 2016) or bilaterally (Donnay et al., 2014; de Manzano & Ullén, 2012a; 2012b). In contrast, three studies did not report involvement of the IFG during
improvisation (Bengtsson et al., 2007; Berkowitz & Ansari, 2010; Villareal et al., 2013). De Manzano and Ullén (2012b) discussed in their report why their results showed presence of IFG activity during
improvisation, whereas the results of previous work from Bengtsson et al. (2007) did not. They hypothesized that this finding may be due to differences in the degree of musical freedom between the tasks. For instance, Bengtsson et al. (2007) instructed participants to improvise based on a previously presented musical template, whereas de Manzano and Ullén (2012b) allowed participants to have total freedom when improvising a melodic line of 24 pitches. Interestingly, the study of Villareal et al. (2013) was the only other study in the musical domain, besides Bengtsson et al. (2007), which instructed its participants to create a musical improvisation based on a previously heard rhythm. They also failed to show IFG activity, suggesting that the IFG may be important for musical freedom. These findings highlight the notion that task-related factors influence the activation of different creativity-related brain areas and networks. However, other studies revealed activation in similar brain networks regardless of task differences, which may be indicative of domain-generality. For example, de
Manzano and Ullén (2012a) reported that the same brain regions, mainly the pre-SMA, left dorsal premotor cortex (dPMC), DLPFC and IFG, were activated for both melodic improvisation and rhythmic improvisation, although increased functional connectivity between the pre-SMA and cerebellum was only present for rhythmic improvisation, but not melodic improvisation. Similarly, two more studies reported activation of completely overlapping neural networks for melodic and rhythmic improvisation (Berkowitz & Ansari, 2008) or scale and jazz improvisation (Donnay et al., 2014).
Besides the DLPFC, motor cortex and IFG, studies reported various other brain regions in frontal, temporal, parietal and occipital lobes during musical improvisation, such as the anterior cingulate cortex (ACC; Berkowitz & Ansari, 2008; de Manzano & Ullén, 2012b), medial prefrontal cortex (mPFC; Berkowitz & Ansari, 2008; Liu et al., 2012; McPherson et al., 2016), superior frontal gyrus (SFG; Adhikari et al., 2016; Berkowitz & Ansari, 2008), the inferior parietal lobule (IPL), which includes the angular gyrus (AG) and supramarginal gyrus (SMG; Adhikari et al., 2016; Berkowitz & Ansari, 2008; Donnay et al., 2014; McPherson et al., 2016), the posterior cingulate cortex (PCC; Berkowitz & Ansari, 2008; Boasen et al., 2018; Liu et al., 2012) and superior and middle temporal gyrus (STG; MTG; Adhikari et al., 2016; Bengtsson et al., 2007; Boasen et al., 2018; Liu et al., 2012; Dikaya & Skirtach, 2015; Donnay et al., 2014). Again, direction of activation and lateralization differed greatly (see Table 1 for an extensive overview), strengthening the notion that neural correlates related to musical creativity show little overlap.
However, one important finding must receive serious attention. The fMRI study of de
Manzano and Ullén (2012b) incorporated an extra control condition of significant importance. Besides sight-reading, a common control condition, participants were instructed to generate random
keypresses. The instructions entailed the description that every keypress could be seen as the result of a rolling dice. By adding this control condition, they made sure that differences in the neural correlates of improvisation compared to random generation were solely due to musical creativity, but not the ‘randomness’ which could be involved in creating something novel. However, the improvisation >
random generation contrast did not yield significant activations. In addition, when they contrasted both conditions to sight-reading, the exact same brain areas became activated. These brain areas included the left DLPFC, pre-SMA, IFG and ACC, all of which have been implicated in musical creativity. On the one hand, this finding raises the question whether the current musical improvisation paradigms can capture the neural correlates of musical creativity and not just general cognitive skills necessary for creativity. On the other hand, this finding may strengthen the view that (musical) creativity does not have its own neural correlates, but rather depends on the cooperation of various general cognitive skills.
3.1.2.Creative performance
Similar to the findings of de Manzano and Ullén (2012b), one EEG study reported that playing a pre-learned melody (PP) and playing an improvised melody (PI) elicited exactly the same coherence pattern within a network consisting of the right DLPFC, SMA, left SFG, left IPL and right STG (Adhikari et al., 2016). However, although the networks of the PP and PI conditions were highly similar, differences did emerge when specifically assessing the level of creative performance (e.g., the amount of originality or novelty) in relation to brain regions, instead of contrasting creative behavior against non-creative behavior. For instance, higher originality of improvisational performance was related to increased EEG coherence between the SFG and SMA, SFG and IPL, and SMA and IPL. Higher originality was also related to negative causal interactions from the SFG to SMA, IPL to SFG and IPL to STG (Adhikari et al., 2016). The importance of frontal regions for creative performance was further strengthened by an EEG study which showed that right frontal alpha synchronization positively correlated with the quality of improvised melodies (Lopata et al., 2017). Furthermore, one fMRI study showed that more innovative performance of freestyle rap improvisations was
significantly associated with increased activity in the left mPFC, PCC, MTG and superior temporal sulcus (STS) compared to less innovative performance (Liu et al., 2012). Another fMRI study divided their participants, all music college students, in two groups based on their level of creative
performance and subsequently investigated differences in brain activity during improvisation (Villareal et al., 2013). Their results showed that improvisation was related to increased left DLPFC and right insula activity in the high creative group compared to the low creative group.
3.1.3.Expertise
Furthermore, three studies investigated neural differences of improvisation in expert musicians compared to non-expert musicians. One fMRI study reported that expert pianists showed significantly decreased activity in the right temporoparietal junction (TPJ) compared to non-experts during
improvisation, even though the groups were behaviorally matched regarding their creative
performance (Berkowitz & Ansari, 2010). In addition, two EEG studies reported that expert musicians showed increased alpha activity in prefrontal regions during musical improvisation compared to
non-experts, although one study reported left lateralized frontal alpha activity (Dikaya & Skirtach, 2015), whereas the other reported right lateralized frontal alpha activity (Lopata et al., 2017). In addition, expert musicians showed lower alpha power in bilateral temporal areas, whereas amateurs showed higher alpha power in temporal areas (Dikaya & Skirtach, 2015). Like Berkowitz & Ansari (2010), differences in behavioral performance were not observed between experts and non-experts (Lopata et al., 2017).
3.1.4.Interim discussion on musical creativity
Altogether, studies portrayed highly variable outcomes regarding the neural correlates of musical creativity. The involvement of the DLPFC, motor areas and IFG during musical improvisation were the most consistent findings across studies. Still, lateralization and direction of activity of these brain regions, as well as less frequently reported brain regions, differed extensively. Moreover, the findings indicate that frontal activity may be important for quality of creative performance. Although all studies that measured creative performance associated improved quality of musical improvisations with increased frontal brain activity, the exact anatomical locations within frontal areas lacked consensus, ranging from the DLPFC, the mPFC to the SFG. Only three studies examined effects of expertise on the neural correlates of musical creativity. They agreed that more experienced musicians showed increased task-related brain activity in prefrontal regions as well as decreased brain activity in temporal regions compared to less experienced musicians, though lateralization varied between studies. Despite differences in brain activity between experts and non-experts, performance on musical improvisations were behaviorally matched, suggesting that expertise may influence a reorganization of brain function necessary for carrying out the task, irrespective of actual performance. In addition, the results point toward the idea that musical creativity may arise from cooperation between brain regions within a network of general cognitive skills, a notion that is specifically strengthened by the finding that creative musical improvisation and random generation of key presses elicited activity in the exact same brain regions (de Manzano and Ullén, 2012b). Still, it remains difficult to determine whether musical creativity is related to a distinct set of brain areas due to the great variety in results. Even within this one domain many differences regarding the neural correlates of creativity exist, which will inevitably make it harder to draw reliable conclusions across domains. One of the explanations for the varying results probably lies with the large number of methodological artifacts. Although all studies used musical improvisation as a measure of creativity, task-designs still differed regarding type of comparison, instructions, sample size and participants, thereby compromising reliable comparison.
3.2. Visual arts domain
Nine studies belonged to the visual arts domain, of which five studies used fMRI, two studies used EEG, one study used sMRI and one study used fNIRS. Unlike the studies incorporated in the musical domain, which only used musical improvisation as a measure of creativity, studies from the
visual arts domain displayed a great variety of tasks aiming to assess creativity. For instance, task designs consisted of mentally composing or planning an artwork, observational drawing (drawing based on an example, e.g., participants viewed a picture of a human hand and were instructed to draw the hand as accurate as possible), designing book covers, free drawing or doodling, and word-drawing (drawing based on the meaning of words, like the game Pictionary).
3.2.1.Creative versus control tasks
Despite the variety of tasks, four fMRI studies reported activity in brain regions associated with the executive control network (ECN) and the default-mode network (DMN; De Pisapia et al., 2016; Ellamil et al., 2012; Saggar et al., 2015; 2016). These brain regions included the DLPFC, ACC, IFG and superior parietal lobes for the ECN and the mPFC, PCC, precuneus, IPL, TPJ and
(para-)hippocampus for the DMN. Two studies reported that word-drawing compared to zigzag-drawing showed increased activity in ECN and DMN regions, whereas the reverse contrast showed increased activity in regions related to the DMN only (Saggar et al., 2015; 2016). The latter finding may be expected since zigzag-drawing does not require much attention and therefore may allow for processes like mind-wandering. However, increasing creativity ratings of drawn words were
associated with increased recruitment of bilateral cerebellum, lingual gyrus, brain stem, FFG and right ITG (Saggar et al., 2015). These regions are not related to either the ECN or DMN but rather with recognition (lingual gyrus, FFG, ITG) and motor activity (cerebellum), although the cerebellum has previously been linked to creativity as well (Vandervert, Schimpf, & Liu, 2007). Furthermore, functional connectivity between ECN and DMN nodes was assessed during mentally planning of an artwork and mentally visualizing plain letters of alphabet (De Pisapia et al., 2016). Results showed that the creative task was associated with increased connectivity between ECN nodes and DMN nodes, specifically between the right DLPFC, right IFG, PCC and precuneus. In addition, the creative task was associated with decreased functional connectivity within ECN nodes, specifically between bilateral DLPFC and right IFG (de Pisapia et al., 2016). Importantly, one fMRI study assessed the neural correlates of visual creativity by comparing two different aspects of the creative process, which are generation and evaluation of a creative product (Ellamil et al., 2012). In order to design book covers based on a summary text, subjects were first instructed to draw or write ideas for book covers during the generation condition and subsequently evaluate their ideas during the evaluation condition. Both conditions were contrasted with a control condition of tracing lines. Interestingly, the generation > evaluation contrast showed activity in different brain regions compared to the evaluation >
generation contrast. The generation phase was related to increased activity in hippocampal areas, IPL, PMC and FFG, whereas the evaluation phase was related to increased activity in ACC, DLPFC, mPFC, PCC, Precuneus, TPJ, RLPFC and SMA. Note that the brain regions of the evaluation phase are all, apart from the SMA, associated with the ECN or DMN. In fact, a functional connectivity analysis revealed increased coupling of the ECN and DMN during idea evaluation. Besides
differences, similar brain regions were involved in both phases as well, such as the IFG, MTG, SPL and cerebellum. This study highlights the need for careful decomposition of the creative process.
3.2.2.Expertise
Most studies of the visual arts domain compared artists with non-artists. Only one study using fNIRS did not find significant differences between groups (Kaimal et al., 2017), although it must be noted that they focused on left mPFC activity alone. In contrast, it was shown that artists exhibited higher functional connectivity between the precuneus and other regions, such as the left DLPFC, right PCC, left PMC and right FFG, compared to non-artists (de Pisapia et al., 2016). The association between expertise and the precuneus is supported by a structural MRI study which showed larger grey matter volume (GMV) in the right precuneus for artists compared to non-artists (Chamberlain et al., 2014). Furthermore, two EEG studies reported decreased alpha band activity in frontal regions in artists as opposed to non-artists (Bhattacharya & Petsche, 2005; Kottlow et al., 2011). However, both studies showed contrasting results as well. For instance, Kottlow et al. (2011) reported decreased alpha power in occipitoparietal areas as well as the right ITG and right MTG during drawing for artists compared to non-artists. Instead, Bhattacharya and Petsche (2005) reported increased delta band synchrony in frontal and posterior occipitotemporal areas during mentally composing a drawing in artists compared to non-artists. Previous research has provided evidence that delta waves may relate to attention focused on an internal representation (Harmony, 2013), suggesting that artistic expertise may come with more vivid imagery. A possible explanation for the discrepancy in results between the two EEG studies could be the difference in task designs. For example, Bhattacharya and Petsche (2005) instructed participants to mentally compose a drawing of their own choosing, leaving space for creative thought, whereas Kottlow et al. (2011) instructed participants to draw copies of pre-made drawings. Perhaps the latter task, a form of observational drawing, should be regarded as a measure of technical artistic ability rather than creative artistic ability. This notion is strengthened by their finding that control conditions of the observational drawing task (imagining the pre-made drawing and imagining the motor action of drawing) showed decreased alpha power in the same brain areas (right ITG, right MTG and right PCC) as the actual drawing condition itself, suggesting that these areas are not creativity-related but can rather be attributed to general cognitive skills necessary for carrying out the experiment. Two other studies used observational drawing tasks to measure creativity and both reported the involvement of motor areas. First, Chamberlain et al. (2014) reported a positive correlation between drawing ability and white matter volume in the posterior cerebellum as well as grey matter volume in the left anterior cerebellum and right SMA. Second, Schlegel et al. (2015) found that participants who received art training showed increasing activation in the right anterior cerebellum over time during a gesture-drawing task (quickly drawing form and movement of human figures) compared to participants who received chemistry training. Subsequently, art students improved in their ability to draw human figures from observation compared to chemistry students,
suggesting that expertise influences brain function. Another study examined the effect of creativity training on the brain as well (Saggar et al., 2016). Participants were divided into two groups. The experimental group received five weeks of a creative capacity building program (CCBP), whereas the control group received five weeks of a language capacity building program (LCBP). Spontaneous creative improvisation was measured during pre- and post-intervention tests with the use of a word-drawing task. Results indicated that the CCBP group showed a trend, though not significant, towards enhanced creative post-test performance on the word-drawing task compared to pre-test performance, whereas the LCBP group did not. However, the CCBP group did show significantly reduced post-test activity in the SMA, ACC, right DLPFC, bilateral parietal lobules, left temporal-occipital fusiform, lingual gyrus and cerebellum, whereas the LCBP group showed increased post-test activity in these areas. Furthermore, the CCBP group showed increased task-related post-test cerebellar-cerebral connectivity compared to the LCBP group, including the right ACC, right IFG, bilateral DLPFC, frontal pole, SFG and paracingulate gyrus. According to Saggar et al. (2016), their results indicate that enhanced improvisation-based creative capacity requires less engagement of executive functioning regions and more involvement of spontaneous implicit processing regions.
3.2.3.Interim discussion on visual arts creativity
The findings of the selected studies for the visual arts domain suggest that creativity involves a dynamic interaction process between the ECN and DMN, two brain networks linked to executive functioning (ECN) and spontaneous thought (DMN). Amongst the brain regions that were associated with these networks as well as frequently reported by visual arts studies were the DLPFC and IFG, two regions that were similarly observed in musical creativity tasks, and the mPFC. Other frequently observed regions belonging to the ECN and DMN networks were the ACC, PCC, Precuneus, IPL, SPL and OFC. Notably, however, visual art studies that failed to observe activity in frontal areas such as the DLPFC and mPFC generally used observational drawing as a measure of creativity. Although observational drawing studies were selected based on the inclusion criteria, the question arises whether these types of tasks reflect creative behavior. Rather, it seems as if observational drawing assesses technical drawing ability and miss the creative component of producing spontaneous and/or original art. However, one can argue that drawing in itself counts as creative behavior regardless of what is produced, which makes assessing creativity a matter of defining creativity. Observational drawing tasks mostly reported motor related brain areas such as the SMA and cerebellum. In fact, the
cerebellum is another brain region that has been reported relatively often across visual art studies (five out of nine studies), but less so by musical creativity studies. An extensive body of research has shown the importance of cerebellar activity for fine motor skills (Ito, 2008), which makes sense when
thinking of the precise movements that come into play whilst drawing. Besides its relation to motor skills however, prior studies have linked cerebellar activity to the production of creativity and innovation (Vandervert, Schimpf, & Liu, 2007). The finding that creativity ratings of drawn words
were associated with increased cerebellar activity (Saggar et al., 2015) strengthens the notion that the cerebellum may play a key role in visual arts creativity. Besides cerebellar involvement, visual art studies also reported more temporal brain regions related to object recognition compared to musical creativity studies.
When neural correlates of artists were compared with non-artists, differences were observed. Though based on few studies, a general tendency occurred that artists showed decreased activity in ECN and DMN related brain regions during creativity tasks compared to non-artists, as well as increased long-range functional connectivity between frontal and more posterior regions. Exact brain areas however differed per study. Due to the small number of studies reporting on expertise-related effects in both the visual arts domain and the music domain, it was hard to compare results. Still, a discrepancy in results of EEG studies was found, showing that visual arts studies observed decreased alpha band activity in frontal regions for experts compared to non-experts during creative tasks, whereas musical creativity studies reported increased alpha band activity for experts.
Besides the small number of visual arts studies, methodological artifacts may also explain much variance. Whereas musical creativity studies all used the same concept of measuring creativity, which was musical improvisation, visual arts studies differed greatly in task-designs. For instance, studies used actual drawing and mentally preparing a drawing, as well as free-drawing and
observational drawing. Another possibility for the large amount of mixed results may come from the lack of decomposition of the creative process. Ellamil et al. (2012) showed that generation of ideas for book covers was associated with widespread activity of DMN regions, whereas idea evaluation was associated with increased activity in both DMN and ECN regions. These findings suggest that each phase of the creative process requires the involvement of different brain constructs. Without careful decomposition of the creative process it becomes impossible to differentiate between creative phases. As a result, creative tasks provide a measure of brain activity which involves all the creative phases together. However, some task designs may tap more into aspects of idea generation for artistic
performance whereas other task designs require more evaluative processes. Cross comparison of these studies therefore becomes more difficult and could explain the large amount of mixed results. This problem is probably not limited to visual arts creativity but may also apply for musical creativity, which evidently makes it harder to determine whether creativity is a domain general or domain specific construct.
3.3. Creative writing domain
Seven studies were incorporated in the creative writing domain. Creative task conditions were relatively consistent across studies since all studies used a story generation task to measure creative performance, except for one sMRI study that compared GMV of expert writers to non-expert writers. Tasks commonly required participants to generate stories based on a cue that was presented to them
prior to story generation. Cues differed between studies as they entailed literary texts, art images or semantically related and unrelated words. In contrast to the relative consistency of creative task conditions, control conditions differed extensively as they included generating uncreative stories (versus creative stories), recitation of text, generation of memorized and non-memorized facts, memorizing word lists, random typing movement and listing details of art images.
3.3.1.Creative versus control tasks
Four out of seven studies reported the involvement of the DLPFC during creative writing (Bechtereva et al., 2004; Erhard et al., 2014; Liu et al., 2015; Shah et al., 2013), although lateralization and the direction of activity differed across studies. For example, two studies reported increased left DLPFC activity during story generation (Bechtereva et al., 2004; Shah et al., 2013), whereas another study reported decreased bilateral DLPFC activity during story generation (Liu et al., 2015). The presence of DLPFC activity also varied when taking different creative writing phases into account, such as brainstorming, writing and revision. It was shown that left DLPFC activity increased during brainstorming about potential stories, but not during actual creative writing (Shah et al., 2013). In addition, Liu et al. (2015) showed a decrease in DLPFC activity during generation of poems, but an increase in DLPFC activity during revision of the generated poems.
Furthermore, five out of seven studies reported mPFC activity during creative task conditions (Bechtereva et al., 2004; Erhard et al., 2014; Howard-Jones et al., 2005; Liu et al., 2015; Shah et al., 2013). Similar to DLPFC involvement, the presence of mPFC activity differed between creative writing phases. One study reported mPFC activity during brainstorming on creative stories, but not during creative writing (Shah et al., 2013), whereas another study reported mPFC activity during creative writing but not during brainstorming (Erhard et al., 2014). Moreover, one study reported mPFC activity during the generation of new poems, but not during revision of the generated poems (Liu et al., 2015). Still, a more conclusive pattern emerges regarding the direction of activity since each study reported increased mPFC activity as opposed to decreased activity. Liu et al. (2015) further explored differences between generation of new poetry and revision of self-generated poem using independent component analysis (ICA). They identified five clusters of brain regions that were associated with poetry composition. One cluster included DMN regions (e.g., mPFC and PCC) and another included ECN regions (e.g., DLPFC and IPS). Idea generation resulted in negative
correlations between DMN and ECN clusters. During idea revision, however, the correlation between the networks increased, suggesting that relative increased cooperation between the DMN and ECN plays an important role for the evaluation of self-generated poems. Furthermore, the same study found that brain activity in the mPFC during generation of poems only slightly differed from activity during generation of non-memorized facts, whereas mPFC activity was not found for generation of
memorized facts. Generation of poems and non-memorized facts both require participants to come up with spontaneous ideas, though poems are considered creative whereas non-memorized facts are not.
For this reason, Liu et al. (2015) argued that the presence of mPFC activity relates to spontaneous cognitive ability, which is irrespective of a creativity component. However, this statement is
contradicted by the results of Howard-Jones et al. (2005). In their study they placed specific emphasis on the generation of creative stories based on semantically unrelated words versus the generation of non-creative stories. They reported increased activity in bilateral MFG (part of mPFC) for the creative condition compared to the non-creative condition, suggesting that the mPFC may be specifically important for creativity after all. Besides DLPFC and mPFC activity, many other brain regions were reported in relation to story generation, though less consistent. These included the left ACC (Howard-Jones et al., 2005; Shah et al., 2013), striatum (Erhard et al., 2014; Liu et al., 2015), MTG (Bechtereva et al., 2004; Shah et al., 2013), cerebellum (Neumann et al., 2018; Shah et al., 2013) and more (see Table 1).
3.3.2.Expertise
The importance of frontal areas for creative writing is further strengthened by studies that assessed effects of expertise. For instance, a structural MRI study compared whole-brain GMV of expert creative writers to non-expert writers and reported increased GMV in bilateral MFG and right SFG for experts (Neumann et al., 2018). In addition, one EEG study investigated alpha power in relation to story generation in actors versus non-actors (Rodionov et al., 2013). Although both groups showed globally enhanced alpha power during story generation based on an art image compared to listing details of the same art image, actors showed increased alpha power in frontal, temporal and central areas during story generation compared to non-actors. Two fMRI studies further assessed differences between expert and novice creative writers. First, Liu et al. (2015) reported that experts showed significantly greater decreased activity in the DLPFC and the intraparietal sulcus compared to novice writers during poem generation, whereas novice writers showed greater increased activity in the left DLPFC and lingual gyrus compared to experts during poem revision. Furthermore, high craft poems (e.g., more creative) were associated with increased coupling of the mPFC with language-related regions, decreased coupling of the mPFC with posterior parietal areas and decreased coupling of the DLPFC and sensorimotor areas for expert compared to novice writers. This finding suggests that with expertise comes a reorganization of involved brain areas in order to produce creative products. Second, Erhard et al. (2014) reported that experts showed increased activity in right lateralized motor areas, insula and putamen during brainstorming minus the reading condition compared to non-experts. Experts showed increased activity in bilateral DLPFC, left mPFC and left caudate during the creative writing phase compared to non-experts. However, activity in these areas disappeared when the control condition copying was subtracted from the creative writing condition, only showing increased right middle cingulate activity for experts compared to non-experts.
3.3.3.Interim discussion on creative writing
Among the most consistent findings within the creative writing domain was the involvement of the DLPFC and mPFC. Both regions complement findings of the musical and visual arts domain. Especially the DLPFC has been similarly observed in both the musical and the visual domain, whereas the mPFC has been mostly observed in the visual arts domain, but not so much the musical domain. In line with results from the musical domain, creative writing studies also reported differences in
lateralization and increased versus decreased activity as well as lateralization.
Three studies in the creative writing domain assessed brain activity after separating the creative process into multiple phases. Two studies compared brainstorming about a new story with writing of the self-generated story using the same experimental paradigm (Erhard et al., 2014; Shah et al., 2013). Both studies failed to report overlapping brain regions associated with brainstorming or creative writing, although the lack of overlap could be explained by the difference in subject groups. One study focused on the difference between brainstorming and creative writing within one group of ‘regular’, non-experienced subjects (Shah et al., 2013), whereas the other study focused on the difference between brainstorming and creative writing in expert writers versus non-expert writers (Erhard et al., 2014). Furthermore, one study assessed differences in brain activity during generation of new poems versus evaluation of the self-generated poems. Their results showed decreased coupling of ECN and DMN regions during idea generation, but relatively increased coupling between these networks during idea evaluation. Interestingly, these findings compliment a visual arts study, which showed that idea generation of book covers was associated with widespread activity of DMN regions, whereas idea evaluation was associated with increased functional coupling between ECN and DMN regions (Ellamil et al., 2012). These studies suggest that self-generated ideas in poetry and visual art may benefit from involvement of the DMN, but that such ideas may need top-down modulation during evaluation as reflected by DMN-ECN coupling. Together these studies stress the need to implement the decomposition of the creative process in task-designs.
Furthermore, the results from creative writing studies on expertise partly overlap with the other creativity domains. For example, creative writing is associated with decreased activity in the DLPFC and IPS during the creative task for experts compared to non-experts irrespective of creative writing phase. This result complements findings of the visual art domain, which showed decreased activity in ECN regions (including the DLPFC) and DMN regions for artists compared to non-artists. In line with this thought, previous research showed that experts may use general cognitive brain functions more efficiently, which is reflected by decreased activity in these regions (Debarnot, Sperduti, Rienzo & Guillot, 2014). This trend, however, is not supported by musical improvisation studies as they reported increased activity in prefrontal regions including the DLPFC, rather than decreased activity. Moreover, the production of high craft poems in expert writers versus non-expert writers was associated with increased mPFC coupling with language areas and posterior parietal
regions, and weaker DLPFC coupling with superior parietal areas. In contrast, visual arts studies reported increased DLPFC coupling with a wide range of posterior areas, although these findings are not based on the level of creative performance. In addition, musical creativity studies reported
decreased activity in temporal regions for experts, compared to non-experts, which was not supported by visual arts or creative writing studies. Besides fMRI studies, EEG studies also portray a partly complementary pattern regarding expertise. One creative writing study reported increased alpha activity in frontal, temporal and central regions during creative writing for experts compared to non-experts. In contrast, visual arts studies observed decreased alpha activity in frontal regions for artists compared to non-artists. Musical studies reported increased alpha activity in prefrontal regions, which is similar to the creative writing study. However, they also report decreased alpha activity in temporal regions, which does not support the creative writing study. It should be kept in mind, however, that the reported findings on expertise are based on few studies per creativity domain.
3.4. Other domains
Two EEG studies represented the dance domain. Both studies used dance improvisation as opposed to dancing a choreography in order to differentiate between creative and non-creative behavior. A main difference in task design between the two studies was that the first instructed participants to mentally perform an improvisation dance (Fink et al., 2009), whereas the second instructed participants to physically touch index fingers with the experimenter and perform a contact improvisation with the right arm alone (Goldman et al., 2019). The results of these studies exhibited little overlap. Fink et al. (2009) showed greater alpha synchronization in frontal, frontocentral and centrotemporal brain regions in dance improvisation compared to the control condition, whereas Goldman et al. (2019) reported greater alpha power in posterior regions reflecting the visual cortex. In addition, according to Fink et al. (2009) professional dancers exhibited more right-hemispheric alpha synchronization in parietotemporal and parietooccipital regions during mentally performing
improvisation dance compared to novice dancers, whereas Goldman et al. (2019) did not observe any effect of expertise.
The scientific discovery domain was represented by two studies as well. Both studies used a hypothesis generation task as a measure of creativity, however, their results barely complement each other. Jin et al. (2006) showed increased functional coupling, as measured with EEG, between left posterior and right anterior brain regions, as well as left posterior and right posterior regions during hypothesis generation compared to rest. In contrast, Lee et al. (2012) reported increased functional connectivity, as measured with fMRI, between left-lateralized brain regions, including the STG, MTG, MFG, putamen and para-hippocampus. In addition, they studied the difference in functional
connectivity between groups with different levels of expertise regarding hypothesis-generation. These groups consisted of professional biologists, science high school students and general high school students. Their results showed that functional connectivity between pairs of left-lateralized brain
regions was largest for biologists and weakest for general high school students, suggesting an effect of expertise. However, it remains unclear whether their results may be confounded by age as biologists were much older compared to high school students and age was not corrected for. Moreover,
participants in the study of Jin et al. (2006) consisted of 5th grade students, which may explain the lack of consistency between both studies.
Furthermore, the humor domain was represented by one study. Creativity was assessed with a humor generation task during which participants were instructed to make up captions based on a neutral drawing that were either funny or mundane (Amir & Biederman, 2016). Their results showed greater activation in the mPFC, bilateral striatum, bilateral temporo-occipital junction (TOJ) and the primary visual cortex for the generation of humorous captions compared to mundane captions. Activity in the mPFC and striatum decreased with occupational experience, whereas the reverse was true for the right TOJ. In addition, better creative performance, based on subjective and expert-based funniness ratings, was associated with increased activity in the striatum and bilateral TOJ, but not the mPFC for professional comedians compared to controls.
Like the humor domain, the inventions domain consisted of one study only. For this study, a design task was used to measure creativity (Kowatari et al., 2009). Art students and non-art students viewed pictures of pens whilst lying in the MRI scanner. They were instructed to invent new designs for pens during the experimental condition and to count the pens on the pictures during the control condition. Contrasts between inventing and counting pens did not show any significant differences, suggesting that the same brain regions were activated during both conditions. These brain regions included the right IFG, right inferior parietal cortex (associative visual cortex), bilateral FFG, MTG, STG and hippocampus. Differences did emerge, however, when art experts were compared with novices during the design condition. Experts showed only right-lateralized activity in the PFC and inferior parietal cortex, whereas novices showed bilateral activity in both regions during invention. In addition, ACC activity during invention was present for novices, but not for experts, which may reflect higher task-difficulty for novice subjects (Fincham & Anderson, 2006). After the scan session,
participants drew their inventions on paper. Independent design-experts provided creativity ratings based on the originality of the drawings. Results showed that the expert group obtained higher
creativity ratings compared to novices. Moreover, creativity ratings were positively correlated with left and right PFC connectivity for experts, whereas creativity ratings were negatively correlated with activity in the left or right parietal cortex for novices.
Finally, three selected studies examined the correlation between CAQ scores and GMV. Chen et al. (2014) reported positive correlations between total CAQ scores and GMV in the left SFG, right vmPFC, right DLPFC, left precuneus, bilateral ITG and bilateral thalamus. In addition, they reported negative correlations between total CAQ score and GMV in the bilateral ACC, extending to SMA.
Instead of total CAQ score, Shi et al. (2017) used the two-factor solution by Carson et al. (2005) to divide total CAQ scores into artistic creativity (including theatre, humor, music, visual arts and creative writing domains) and scientific creativity (including invention, scientific discovery and culinary arts domains). Partly overlapping with the findings of Chen et al. (2014), they also reported negative correlations between CAQ score and GMV in the ACC and SMA, though this was only the case for artistic creativity and not scientific creativity. In contrast, scientific creativity scores were positively correlated with GMV in the left MFG and left inferior occipital gyrus. Furthermore, Zhu et al. (2016) reported that both total CAQ score and artistic creativity scores were positively correlated with GMV in the right PMC.
Interim discussion on dance, scientific discovery, humor and inventions domains
The creative domains included in this section entail so little studies that conclusions based on study comparison within and across domains must be considered with great caution. Within the dance domain, both EEG studies reported increased alpha activity during dance improvisation, though the specific brain regions differed. Few studies from other domains reported on differences in EEG signals between creative conditions versus control conditions. Only one EEG study from the musical domain also showed increased alpha activity during musical improvisation in brain regions that partly overlap with brain regions reported by dance studies. In addition, only one dance study found differences in brain activity between dance experts and non-experts, showing increased alpha activity in
parietotemporal and occipital regions. Increased alpha activity in experts versus non experts was similarly observed during musical improvisation and creative writing, whereas drawing was associated with decreased alpha activity in artists versus non-artists. Brain regions in which alpha change was detected show some similarities across domains. For instance, the musical, visual arts and creative writing domains all reported alpha activity change in frontal areas in experts, whereas the dance domain and visual arts domain reported parietal, temporal and occipital brain regions (though the visual arts domain showed decreased alpha activity instead of increased alpha activity).
Findings from the scientific discovery domain show very little similarities within and across domains. One of the few similarities with other domains entailed the finding of increased functional connectivity in experts compared to non-experts during hypothesis generation (Lee, 2012). Similarly, studies from the visual arts and inventions domains also reported increased functional connectivity in experts versus non-experts. However, the brain areas that were functionally more connected showed little overlap across the three domains.
Increased activity in the mPFC associated with generation of humorous captions compared to generation of mundane captions for the humor domain complements findings of visual arts and creative writing domains. Activity in the striatum and TOJ have not been mentioned by other studies included in this review, except for one creative writing study that found increased functional coupling between the mPFC and the caudate nucleus (a sub-part of the striatum) during poem generation. The
authors suggested that the involvement of striatal activity may reflect activity of the reward-system, suggesting that generation of funny captions provides greater feelings of reward. This idea was further supported by the finding of increased activity in the striatum for greater funniness ratings in experts. Similar to visual arts and creative writing studies, the humor study also reported decreased activity in the mPFC for experts, based on occupational experience.
The results of the pen design task from the invention study complement the findings of the book cover design task from a visual arts study (Ellamil et al., 2012), since both studies reported activation in the IFG, FFG, MTG and hippocampus during creative conditions. Furthermore, activity in the IFG and temporal regions was also observed in musical improvisation studies. However, the invention study also reported a lack of significant differences in brain activity between the
experimental and control condition, suggesting that activity in these regions may represent active use of general cognitive skills rather than creativity. Altogether, the dance, scientific discovery, humor and invention domains show extremely little overlapping brain regions or networks when compared to each other. Although one could argue that this may suggest the non-existence of a general creativity network, this is almost impossible to conclude due to the little number of studies as well as the enormous variety in task-designs per domain.
Finally, the structural studies assessing the association between creative achievement and brain volume mostly listed regions that were frequently observed in musical, visual arts and creative writing studies, such as the DLPFC, precuneus, ACC and SMA. These studies therefore strengthen the possibility that these regions are important for creative behavior.
