Abstract

Neuroimaging studies of internally generated behaviors have shown seemingly paradoxical results regarding the dorsolateral prefrontal cortex (DLPFC), which has been found to activate, not activate or even deactivate relative to control conditions. On the one hand, the DLPFC has been argued to exert top–down control over generative thought by inhibiting habitual responses; on the other hand, a deactivation and concomitant decrease in monitoring and focused attention has been suggested to facilitate spontaneous associations and novel insights. Here, we demonstrate that prefrontal engagement in creative cognition depends dramatically on experimental conditions, that is, the goal of the task. We instructed professional pianists to perform improvisations on a piano keyboard during fMRI and play, either with a certain emotional content (happy/fearful), or using certain keys (tonal/atonal pitch-sets). We found lower activity in primarily the right DLPFC, dorsal premotor cortex and inferior parietal cortex during emotional conditions compared with pitch-set conditions. Furthermore, the DLPFC was functionally connected to the default mode network during emotional conditions and to the premotor network during pitch-set conditions. The results thus support the notion of two broad cognitive strategies for creative problem solving, relying on extrospective and introspective neural circuits, respectively.

Introduction

Neuroimaging studies on internally generated behaviors and creativity have identified brain regions involved in the generation of novel and meaningful content, for example, using musical improvisation or verbal fluency tasks as model behaviors. These regions include the dorsolateral prefrontal and inferior frontal cortices (DLPFC and IFG), parietal association area, anterior cingulate cortex, dorsal premotor area (PMD), and the presupplementary motor area (preSMA) (Bengtsson et al. 2007; Berkowitz and Ansari 2008; de Manzano et al. 2012a, 2012b; Limb and Braun 2008; Liu et al. 2012; Donnay et al. 2014). Specific outcomes, however, vary between studies. For the DLPFC, findings even appear to be mutually contradictory, as this region has been found to activate, not activate, and even deactivate during creative performance. On the one hand, the DLPFC (particularly the mid-DLPFC or BA46) has been argued to exert active top–down control over generative thought by inhibiting habitual responses (de Manzano et al. 2012b); on the other hand, a deactivation has been suggested to facilitate more spontaneous associations and novel insights (Limb and Braun 2008; Liu et al. 2012). While such puzzling inconsistencies have been noted in reviews (Dietrich and Kanso 2010; Beaty 2015), no attempts have been made to explain experimentally how this peculiar range of outcomes is possible. Here, we present a theoretical perspective and experimental results, which indicate that creative cognition can involve different neural networks under different sets of circumstances.

First, we introduce the notion that creativity (defined as the generation of original and useful ideas) can be achieved through a variety of cognitive processes. Nijstad et al. (2010), for example, suggest 2 pathways which are qualitatively different, that is, the flexibility pathway and the persistence pathway. “Pathway,” in this psychological literature, refers to strategy rather than neural structures and “persistence” should be understood as focused task-directed cognitive effort, not perseverance. According to this model, creative achievement can thus be the outcome of either free associative thinking—characterized by fluent and flexible switching between cognitive categories—or more systematic, effortful, and focused exploration of fewer categories; or a combination of the 2.

Second, the consideration of different strategies during creative performance has been entertained in artistic fields for some time. In the context of musical improvisation, Clarke (1988) described 3 principles: 1) Current behavior may be part of a hierarchical structure, to some extent worked out in advance, and to some extent constructed online; 2) current behavior may be part of an associative chain of actions; 3) current behavior may be selected from a number of actions contained within the performer's repertoire. Clarke also asserts that in practice, these principles are generally intermingled throughout performance. The parallels between these principles and the core elements of the dual pathway model are obvious.

Third, it is reasonable to assume that different strategies for creative thinking are also implemented differently at the neural level. If we apply the notion of dual strategies within a creative task, we should be able to contrast one strategy that involves top–down cognitive control, attentional monitoring, and explicit organization of behavior, with another that draws more on implicit, spontaneous recombinations of expertise, established representations and routines. Thus, the former should load more on prefrontal and frontoparietal networks involved in action planning and working memory. For the latter strategy, such explicit processing might even hinder fluent extemporaneous behavior in much the same way as attending to the components of a well-learned skill can impair concurrent performance (Beilock et al. 2002; Gray 2004).

Fourth, we recently found that pianists more trained in improvisation display an overall lower activity in frontoparietal association areas during improvisation, as well as greater effective functional connectivity among prefrontal, premotor, and motor regions (Pinho et al. 2014). In other words, skilled improvisation may be characterized by both low demands on executive control and a more efficient interaction within the network of involved brain areas. In terms of the creative strategies discussed above, experienced improvisers may be able to rely on less attention demanding, yet flexible skills, while novices with, for example, a less extensive storage of musical patterns (Pressing 1988), may have to resort to more explicit organization of behavior.

Consequently, we may formulate the hypothesis that the DLPFC is recruited during creative performance to allow for explicit top–down control and action selection, primarily in situations where knowledge structures and memory processes are not defined or refined enough to guarantee novel and useful output by means of spontaneous free association. However, when well-rehearsed skills and knowledge are on par with challenge, explicit control processes may instead need to be inhibited not to undermine fluent and flexible performance. Thus, we wanted to conduct an experiment where alternating between different task constraints during free generation would bias cognition towards either a more explicit or implicit strategy, to test for constraint-specific processing in the DLPFC. To that end, we asked professional piano players to perform piano improvisations on a piano keyboard during fMRI and either play on certain keys (tonal/atonal sets) or, express a certain emotion (happy/fearful). We hypothesized that the pitch-set conditions (playing certain keys) would require maintenance of the response set in working memory and integration of sensory information and consequently greater activity in the DLPFC and regions involved in visuospatial working memory processes, that is, the PMD and superior parietal cortex, as melodic/ordinal structures are mainly represented and processed as spatial information (Bengtsson et al. 2004, 2006). During emotional conditions, where a more spontaneous and flexible cognitive strategy would be favored, we instead expected activity in these regions to be reduced, particularly since explicit attention will interfere with selective emotion processing (Schupp et al. 2007). For emotional conditions, we additionally predicted an increased activity in the insula, the amygdala, and the dorsomedial and ventromedial prefrontal cortices, that is, in regions linked to music and emotion (reviewed in Koelsch 2010).

We have previously suggested that the premotor network may change its interactions with other regions during creative performance depending on task specific demands. This suggestion was based on finding higher effective functional connectivity between the preSMA and cerebellum during rhythmic improvisation compared with melodic improvisation (de Manzano et al. 2012a). In combination with our previous finding that expertise in improvisation is related to higher effective functional connectivity between prefrontal and premotor regions during improvisation (Pinho et al. 2014), we expected to see a similar effect here within-person, that is, effective functional connectivity would be higher between the DLPFC and premotor regions for pitch-set than for emotional conditions.

Materials and Methods

Participants

Thirty-nine right-handed pianists (24 males, 15 females; age mean/SD = 32 ± 11 years). Handedness was determined with the Swedish version of the Edinburgh Handedness Inventory (Oldfield 1971). All participants were healthy, with no history of psychotropic medication or neurological disease. All pianists were active performers with wide experience in classical or jazz piano playing. Apart from one subject (a self-taught professional jazz piano player), all pianists carried either a university degree in piano performance or were students at the Royal College of Music in Stockholm, the Guildhall School of Music and Drama in London or the Nordic Masters in Jazz programme (Nomazz)—Norwegian University of Science and Technology in Trondheim. Given the number of participants and their level of expertise, the sample can be considered representative of the population of professional piano players in the Stockholm region. The sample has been used in a previous study investigating how between-participant differences in improvisation practice relates to brain activity during improvisation (Pinho et al. 2014). The participants displayed a wide range of both improvisational and classical piano training/experience (improvisation hours: mean 9100, SD ± 13 014; classical playing hours: mean 14685, SD ± 13 062). The effects of improvisational training were found to be consistent across the improvisation conditions. The research question investigated here, concerning the within-subject effect of different task instructions on brain activity during improvisation, is therefore studied as independent from the previous investigation. The previous study also reported an age effect. The present imaging analyses were thus adjusted for both improvisation experience and classical training as well as age. The experiments were carried out with the understanding and formal consent of each participant, in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki), and ethically approved by the Regional Ethical Review Board in Stockholm (Dnr 2011/637-32 and Dnr 2011/1682-32). Participants were reimbursed with 600 SEK.

Piano Keyboard, Auditory Feedback, and Musical Recordings

The participants improvised on a custom designed MR-compatible fiber optic piano keyboard (LUMItouch, Inc.) of one octave (12 piano keys, ranging from F to E) during scanning. Two different MIDI outputs were generated. The first output was relayed to a sound module (Roland SD-50) which synthesized the piano sound (General MIDI Level 2, European Pf) which was given as audio feedback (keys F2–E3) to the participants. The second output was relayed to a PC which recorded the behavioral output using a music production software (Cubase 5; Steinberg).

Visual Stimuli

Visual instructions were given at the start of each experimental trial. The display was controlled by a custom-made E-Prime script (E-Prime 2.0 Professional; Psychological Software Tools, Inc.). The instructions were presented via MR-compatible OLED display goggles (Nordic Neuro Lab's visual system) mounted on the head coil. There were 2 main types of instructions (see Supplementary Fig. S1), introducing the pitch-set and emotional constraints/conditions, respectively. The pitch-set slides displayed a musical staff with a set of 6 allowed notes, corresponding to either the tonal or atonal conditions. There were 24 trials of each condition in the experiment. Pitch-set slides were randomized, with no repeats, to each trial for each participant from a pool of 32 slides/condition. The emotional slides displayed a clip art face expressing the emotions/conditions happy or fearful. These were also displayed in a randomized order. The reason for using 2 conditions per constraint was to reduce the otherwise confounding effect of a particular emotion or structure on the main contrast of interest (pitch-set vs. emotional).

Training Session and Preparations

Participants were given detailed instructions on how to perform the specific experimental conditions in a training session right before scanning. Thus, participants could get familiarized with the piano keyboard and with the different experimental conditions. The participants were told to do free improvisations and play throughout the duration of the trial. They were instructed to play only with the right hand and not look at the piano keys/right hand while playing, in order to match conditions inside the scanner. They then completed one training session (similar in procedure, visual stimuli and audio feedback to the real experiment). Visual stimuli were presented on a laptop; pitch-set slides did not overlap with those later used during scanning. The training sessions lasted about 30 min. All participants testified that they had understood the instructions before going into the scanner. No major difficulties executing the paradigm were observed by the experimenters nor reported by the participants. The participants were scanned in supine position with the piano keyboard placed on their lap. The right arm was supported by a sponge pad in order to avoid fatigue and to minimize arm movements. Ear plugs and headphones were used to reduce scanner noise and to allow auditory feedback from the piano and verbal communication with the experimenters located in the control room. The volume of the audio feedback and dioptric settings in the goggles were adjusted for each participant.

Experimental Procedure

The experiment (illustrated in Supplementary Fig. S2) was carried out in 6 sessions. Each session consisted of 16 trials, that is, every condition was presented 4 times in each session, which gives a total of 96 trials per participant. The trials were composed of 4 consecutive blocks: instruction (3.5 s), performance (15 s), a distractor task (11 s), and rest (6 s). Thus, one trial lasted for 35.5 s. Consequently, we collected 6 min of improvisation in each condition for each participant. The sessions were separated by short breaks (1–2 min), during which the recording software and scanner were prepared for the following session.

During instruction, the visual stimuli would introduce either a pitch-set (tonal or atonal) constraint or emotional (happy or fearful) constraint. In the pitch-set conditions, the participants were to improvise using only the 6 displayed notes. In tonal, the notes/pitches were a subset taken from a specific western musical scale (major, minor). The pitches thus naturally suggested a tonality for the improvisation. Pitch-sets for atonal, in contrast, were randomly generated to fulfill 2 different criteria: 1) that the pitches were not all part of the same major or minor scale, and 2) that there would be at least one interval equal to or larger than a minor third (in order to avoid chromatic sequences). Thus, this pitch-set did not suggest any particular tonality. The specific pitch-sets were unique for each trial of tonal and atonal. For the emotional conditions, the participants were to express or convey a particular emotion (happy or fearful). The order of trials were randomized with the limitations that a certain condition could only appear maximum 2 times in a row and that no more than 3 consecutive trials would consist of conditions with the same constraint (pitch-set or emotional). The visual stimuli were only present during the instruction.

The performance phase was interrupted by the distractor task. This task was included to interrupt musical cognitive processes from the previous trial and to minimize planning of the next improvisation. The distractor task was constructed as a visual aesthetical judgment task. The participant would view a fixation cross for 3 s, then watch a computer generated image for 5 s and then, during the final 3 s of the task, provide a rating according to own aesthetical judgment of the image by pressing on the keys on the piano (higher note meant a higher rating). Results from analyses of this task will be presented in a separate manuscript.

Data Acquisition

All behavioral (musical) data were recorded in MIDI format and analyzed using a custom-made script in MATLAB 7 (The MathWorks, Inc.). The fMRI data were acquired using a 3-T scanner (Discovery MR750w 3.0T, GE Healthcare) with a 32-channel coil (MR Instruments, Inc.) at the MR Research Center of the Karolinska Hospital. Imaging was performed using a gradient echo pulse, echo-planar (EPI) T2*-weighted sequence with blood oxygenation level-dependent (BOLD) contrasts, using the following parameters: repetition time = 2.5 s; echo time = 30 ms; flip angle = 90°; field of view (FOV) = 28.8 cm; slice spacing = 0 mm; voxel size = 3 × 3 × 3 mm3; data acquisition matrix = 96 × 96, interpolated during reconstruction to 128 × 128; slice order = interleaved; number of slices = 48. Two hundred twenty-eight functional images volumes were acquired per session, giving a total of 1368 image volumes per participant. At the beginning of each session, 10 “dummy” image volumes were scanned, but not saved, to allow for T1-equilibration effects. Subsequently, a 3D fast spoiled gradient echo T1-weighted anatomical image volume covering the whole brain was acquired: voxel size = 1 × 1 × 1 mm3; axial slice orientation; flip angle = 12°; inversion time = 450 ms; FOV = 24 cm.

Analysis of Behavioral Data

A set of criteria were employed in order to distinguish valid performances: First, trials in which the participant started playing during the instruction phase were excluded. Second, trials in which the participant had not initiated performance within 6.25 s from the start of the performance phase were excluded. Third, performances from the pitch-set conditions were evaluated to make sure instructions were followed to a reasonable degree. A valid pitch-set performance would correspond to a trial where 1) the participant used at maximum one wrong pitch and 2) at least 5 different valid pitches were used. Using this filtering, we arguably removed performances during which the participant held an incorrect representation of the instructions, but at the same time, allowed for small involuntary slips. In total, ∼11% of tonal and ∼22% of atonal trials were removed. Lastly, sessions in which there were no correct trials for at least one condition were excluded in the analysis. Consequently, 2 participants had 4 sessions and 6 participants had 5 sessions. For the remaining 31 participants, all 6 sessions were used. Another 5 sessions (2% of the data) had to be removed due to various technical problems, for example, with MIDI recordings. Supplementary Table S1 describes how many trials were discarded in each condition because of technical issues or due to violations of the performance criteria.

A Lempel-Ziv complexity measure (Lz) was used in order to characterize the complexity of the musical samples. This index is a measure of the number of unique patterns present in a sequence (Doğanaksoy and Göloğlu 2006). For the present analyses, the melodic structure of each improvisation was represented as a sequence of numbers. Single notes were represented as integers (0–11) corresponding to their pitch. If the interonset interval between 2 consecutive notes was <75 ms, those notes were considered to be part of the same single event (i.e., a musical chord). These musical chords were represented as larger integers that uniquely identified the component pitches of the chord. In addition to Lz, we also measured the number of keys played (Nkeys) in each condition. These measures were primarily created in order to control for motor output in the imaging analysis. Differences in Lz and Nkeys between pitch-set and emotional constraints were analyzed using dependent two-sample t-tests. Further differences between tonal, atonal, happy, and fearful were analyzed using repeated-measures ANOVAs.

Image Processing and Statistical Analysis

The MRI data were processed and analyzed using the SPM8 software package (Wellcome Department of Imaging Neuroscience, London, UK) in MATLAB R2010a.

For each participant, all fMRI image volumes were realigned to the first image of the first session (Friston et al. 1995) and unwarped (Andersson et al. 2001). The T1-weighted (anatomical) image was then coregistered onto the fMRI unwarped mean image (Ashburner and Friston 1997). The brain extraction tool was used in order to remove nonbrain tissue from the coregistered anatomical image (Smith 2002). This image was then segmented in order to estimate the deformation field for the normalization of all functional and anatomical images (Ashburner and Friston 2005).

Comparing Experimental Conditions

The fMRI data were modeled using a general linear model (GLM) and the standard hemodynamic response function (HRF). The model included 10 regressors: 1) instruction, 2) tonal, 3) atonal, 4) happy, 5) fearful, 6) bad performances, that is, excluded performances, 7) distractor presentation (fixation cross and image presentation), 8) distractor rating. To control for motor output, 2 nuisance regressors were included: 9) Nkeys and 10) Lz. These regressors were first centered to the mean and then Lz was orthogonalized with the respect to Nkeys. Thus, also the shared variance between these regressors was captured by the model. Both nuisance regressors were convolved with the HRF and then entered into the model. Rest conditions were part of the implicit baseline. The high-pass filter was set to 497 s (i.e., the Nyquist rate, or 2 × the maximum period between the experimental condition and its repeat). The design matrix weighted each preprocessed image according to its overall variability to reduce the impact of movement artifacts (Diedrichsen and Shadmehr 2005). The contrasts of interest were pitch-set−emotional and emotional−pitch-set. The contrasts for the nuisance covariates Nkeys−rest and Lz−rest were also explored. Contrasts were weighted by the number of included sessions. The contrast images resulting from this first-level analysis were smoothed using a Gaussian kernel with a full-width-at-half-maximum (FWHM) of 10 mm.

A second-level/group analysis was performed on the contrast images using one sample t-tests. Given that we previously found a between-participants effect of improvisation practice and participant age on brain activity in the right DLPFC (Pinho et al. 2014), we included these variables and the reported amount of classical training as nuisance covariates. A participant gray matter mask was created and used as an explicit mask. Regions where significant effects were found were manually labeled using visual inspection and the WFU PickAtlas (Maldjian et al. 2003, 2004). Results are shown if they survive multiple comparisons correction using a voxel-level family-wise error rate (FWE) at P = 0.05 and an extent threshold of 3 voxels.

Connectivity Analysis

A psychophysiological interaction (PPI) analysis (Friston et al. 1997) was performed to analyze differences in effective functional connectivity between the DLPFC and other brain regions when switching between pitch-set and emotional.

Despite the experimental and clinical focus on the DLPFC in structural and functional imaging, the variability of the location of this area, differences in opinion on exactly what constitutes DLPFC (BA46, or BA46 and lateral BA9, or BA46 and lateral BA9 and BA10), and inherent difficulties in segmenting this highly convoluted cortical region have contributed to a lack of widely used standards for how to define DLPFC regions of interest (ROI) for imaging analyses. Therefore, it could be argued that a task-related functional definition constitutes a viable option. In our previous studies, we have found improvisation-related activity within the middle frontal gyrus in what would correspond to the mid-DLPFC or BA46 (Bengtsson et al. 2007; de Manzano et al. 2012a, 2012b). As discussed in the introduction section, we expected to replicate this finding using the pitch-set conditions. However, we have also previously shown that the activity level in this region is dependent on improvisation training (Pinho et al. 2014). Thus, we defined the DLPFC ROIs based on clusters of activity that increased reliably above baseline across participants during pitch-set conditions, that is, using the contrast pitch-set−rest (FWE-corrected), within a spatial range roughly corresponding to BA46 (Supplementary Fig. S3). This meant that ROIs in the left and right hemisphere would not be identical, but as laterality differences for BA46 are likely (as discussed in, e.g., Rajkowska and Goldman-Rakic 1995), one could argue that the derived ROIs would correspond to the most relevant voxels based on functional criteria. The derived right DLPFC ROI had a center of mass at MNI coordinates [x = 40, y = 42, z = 29] and a volume of 861 mm3. The left DLPFC ROI had a center of mass at [x = −41, y = 36, z = 26] and a volume of 2099 mm3. These ROIs are localized in what would correspond to the (mid-)DLPFC (Mylius et al. 2013) or roughly BA46 (Brodmann 1909; Rajkowska and Goldman-Rakic 1995).

The BOLD time series of each seed region was deconvolved from the HRF in order to obtain the underlying neural time course of activity. The PPI was then estimated by multiplying the neural response with a block regressor representing the conditions pitch-set and emotional, and then reconvolving it with the HRF. Afterwards, a GLM analysis was performed with 3 regressors: the PPI, the regional BOLD response, and the experimental conditions. The high-pass filter was set to 497 s. Each preprocessed image was weighted with its overall variability (more variable images receiving a lower weighting) to reduce the impact of movement artifacts (Diedrichsen and Shadmehr 2005). The resulting contrast images were smoothed with an isotropic Gaussian kernel of 10-mm FWHM. Lastly, a second-level analysis was performed using a one sample t-test on the contrast images from the first-level analysis to get the contrasts for pitch-set−emotional and emotional−pitch-set. The same covariates (improvisation practice, age, and classical training) and gray matter mask was used here as in the previous analysis.

Results

Analysis of Behavioral Data

Measures of motor output (Nkeys) and complexity (Lz) were compared between conditions. The mean and standard deviation for Nkeys during pitch-set and emotional conditions were 30.17 ± 1.42 and 36.90 ± 1.77, respectively. Nkeys was significantly higher during emotional than pitch-set conditions [t(2,38) = 5.04, P = 0.000]. Similarly, Lz was higher [t(2,38) = 7.00, P = 0.000] for emotional (8.52 ± 0.33) than for pitch-set (6.54 ± 0.14) conditions. Because of these differences, Nkeys and Lz were introduced as nuisance covariates in the analysis of the imaging data. Further comparisons between happy, fearful, tonal, and atonal are found in the Supplementary material—Analysis of Behavioral Data.

Analysis of fMRI Data

Comparing Experimental Conditions

Table 1 and Figure 1A illustrate the results of the contrast pitch-set−emotional. In line with the main hypothesis, the pitch-set conditions induced a comparably greater activation of the bilateral DLPFC, which in the right hemisphere extended throughout the middle frontal gyrus into the PMD. In addition, there was greater activity in the bilateral parietal lobes. Also in line with predictions, the reverse contrast (emotional−pitch-set; Table 2, Fig. 1B) revealed comparably greater activation of the left dorsomedial prefrontal cortex in the superior medial gyrus, the left medial orbital gyrus and bilateral insulae, extending into the amygdala. Additional results from comparisons between happy−fearful and tonal−atonal are shown in the Supplementary material—Analysis of fMRI Data, and Supplementary Table S2.

Table 1

Imaging results for the contrast pitch-set−emotional

  Coordinatesa
 
Voxel levelb
 
 
Regionc Sided x y z t P Sizee 
MFG 38 59 15 9.96 0.000 9860 
MFG −41 54 15 5.46 0.018 165 
SMedG 29 40 5.11 0.042 
ITG −51 −31 −18 5.39 0.021 17 
ITG 56 −37 −20 6.87 0.000 1091 
IPL 42 −51 41 11.82 0.000 14161 
IPL −36 −54 42 8.03 0.000 3060 
CCrusI −32 −66 −35 5.15 0.039 
IOG −57 −67 −18 5.63 0.012 119 
  Coordinatesa
 
Voxel levelb
 
 
Regionc Sided x y z t P Sizee 
MFG 38 59 15 9.96 0.000 9860 
MFG −41 54 15 5.46 0.018 165 
SMedG 29 40 5.11 0.042 
ITG −51 −31 −18 5.39 0.021 17 
ITG 56 −37 −20 6.87 0.000 1091 
IPL 42 −51 41 11.82 0.000 14161 
IPL −36 −54 42 8.03 0.000 3060 
CCrusI −32 −66 −35 5.15 0.039 
IOG −57 −67 −18 5.63 0.012 119 

Note: CCrusI, cerebellum crus I; IOG, inferior occipital gyrus; IPL, inferior parietal lobe; ITG, inferior temporal gyrus; MFG, middle frontal gyrus; SMedG, superior medial gyrus.

aMNI coordinates.

bVoxel-level FWE-corrected statistics.

cAnatomical location of peak voxels. Regions are sorted anterior to posterior.

dLeft/right hemisphere.

eNumber of significant voxels in the same cluster.

Table 2

Imaging results for the contrast emotional−pitch-set

  Coordinatesa
 
Voxel levelb
 
 
Regionc Sided x y z t P Sizee 
SMedG −8 65 30 7.28 0.000 1496 
MedOrbG −8 50 −11 5.54 0.014 58 
IFG (triangularis) 48 33 5.83 0.007 88 
IFG (orbitalis) −35 33 −14 5.58 0.013 119 
IFG (triangularis) −48 30 −2 5.23 0.032 15 
STG −42 14 −21 5.89 0.006 79 
Insula −39 18 5.21 0.033 19 
PreCG 51 −9 55 5.41 0.020 21 
Insula −41 −10 −11 5.57 0.013 95 
STG 44 −15 −5 5.34 0.024 29 
MCC −11 −21 43 6.43 0.001 426 
CS −29 −31 55 6.21 0.003 331 
RO 41 −31 18 5.28 0.028 33 
IOG 30 −88 −11 8.12 0.000 4440 
SOG −15 −93 27 8.48 0.000 3532 
  Coordinatesa
 
Voxel levelb
 
 
Regionc Sided x y z t P Sizee 
SMedG −8 65 30 7.28 0.000 1496 
MedOrbG −8 50 −11 5.54 0.014 58 
IFG (triangularis) 48 33 5.83 0.007 88 
IFG (orbitalis) −35 33 −14 5.58 0.013 119 
IFG (triangularis) −48 30 −2 5.23 0.032 15 
STG −42 14 −21 5.89 0.006 79 
Insula −39 18 5.21 0.033 19 
PreCG 51 −9 55 5.41 0.020 21 
Insula −41 −10 −11 5.57 0.013 95 
STG 44 −15 −5 5.34 0.024 29 
MCC −11 −21 43 6.43 0.001 426 
CS −29 −31 55 6.21 0.003 331 
RO 41 −31 18 5.28 0.028 33 
IOG 30 −88 −11 8.12 0.000 4440 
SOG −15 −93 27 8.48 0.000 3532 

Note: CS, central sulcus; IFG, inferior frontal gyrus; IOG, inferior occipital gyrus; MCC, middle cingulate cortex; MedOrdbG, medial orbital gyrus; preCG, precentral gyrus; RO, rolandic operculum; SMedG, superior medial gyrus; SOG, superior occipital gyrus; STG, superior temporal gyrus.

aMNI coordinates.

bVoxel-level FWE-corrected statistics.

cAnatomical location of peak voxels. Regions are sorted anterior to posterior.

dLeft/right hemisphere.

eNumber of significant voxels in the same cluster.

Figure 1.

Contrasting brain activity between pitch-set and emotional conditions. All colored areas represent significant differences at a FWE-corrected statistical threshold of P < 0.05. (A) Brain regions showing higher activity during pitch-set than emotional conditions. (B) Brain regions showing higher activity during emotional than pitch-set conditions. Cutouts reveal activity in the insula (left panel) on a sagittal slice at MNI coordinate x = −39 and in the CS/MCC (middle panel) on an axial slice at z = 57.

Figure 1.

Contrasting brain activity between pitch-set and emotional conditions. All colored areas represent significant differences at a FWE-corrected statistical threshold of P < 0.05. (A) Brain regions showing higher activity during pitch-set than emotional conditions. (B) Brain regions showing higher activity during emotional than pitch-set conditions. Cutouts reveal activity in the insula (left panel) on a sagittal slice at MNI coordinate x = −39 and in the CS/MCC (middle panel) on an axial slice at z = 57.

We also plotted how activity in the right and left DLPFC during performance related to baseline (i.e., pitch-set−rest and emotional−rest), in each individual (see Supplementary Fig. S4). There were individuals who showed activity either above or below baseline in both regions for both contrasts; however, proportions differed between instances. For the right DLPFC, the ratios of participants with activity above:below baseline were 33:6 and 20:19 during pitch-set and emotional conditions, respectively. For the left DLPFC, the corresponding ratios were 35:4 and 30:9. For the right DLPFC, activity dropped from above to below baseline in 8 participants when comparing pitch-set with emotional conditions; for the left DLPFC, 5 participants showed the same pattern; 2 participants showed this pattern for both regions.

The contrasts for the nuisance covariates (Nkeys−rest and Lz−rest) were also explored. Nkeys correlated with activity in the bilateral putamen, the primary sensorimotor cortex, the supplementary motor area (SMA), and right cerebellum, as well as with the right superior temporal gyrus (STG), the left thalamus, and the bilateral middle occipital gyri (see Supplementary Table S3). The contrast Lz−rest did not reveal any further findings.

Connectivity Analysis

The whole-brain PPI analysis showed that the left and right DLPFC (the seed regions) had constraint-dependent patterns of connectivity to other brain regions. During pitch-set conditions, there was greater effective functional connectivity between the right DLPFC and the bilateral PMD and STG as well as the left ventral premotor cortex, SMA, primary sensorimotor cortices, left parietal lobe, and right cerebellum (Table 3, Fig. 2A). During emotional conditions, there were instead increases in correlations between the right DLPFC and several regions of the default mode network (Buckner et al. 2008), including medial prefrontal and medial parietal regions (Table 3, Fig. 2B). For the left DLPFC, there were no significant findings for pitch-set−emotional, but the reverse contrast showed significant increases in effective functional connectivity (Table 4) which largely resembled the pattern for the right DLPFC.

Table 3

Imaging results from the PPI analysis based on the right DLPFC as seed region and the contrasts pitch-set−emotional and emotional−pitch-set

  Coordinatesa
 
Voxel levelb
 
 
Contrast/regionc Sided x y z t P Sizee 
Pitch-set−emotional 
 SFG 27 −7 58 5.28 0.032 32 
 STG 56 −16 6.41 0.002 1541 
 PostCG −57 −27 52 7.28 0.000 13026 
 STG 68 −30 16 5.27 0.033 37 
 SPL −17 −67 66 5.45 0.021 
Emotional−pitch-set 
 SFG −21 66 19 8.64 0.000 29466 
 ITG −44 −44 6.15 0.003 1698 
 Amygdala 27 −1 −20 5.77 0.009 368 
 MTG 56 −4 −23 5.31 0.030 44 
 Insula 35 −13 21 5.91 0.006 227 
 Caudate −20 −24 27 5.63 0.013 31 
 CalG −23 −55 10 7.54 0.000 70346 
 Cvermis −58 −44 5.75 0.010 245 
 CcrusII −30 −79 −36 5.60 0.014 219 
  Coordinatesa
 
Voxel levelb
 
 
Contrast/regionc Sided x y z t P Sizee 
Pitch-set−emotional 
 SFG 27 −7 58 5.28 0.032 32 
 STG 56 −16 6.41 0.002 1541 
 PostCG −57 −27 52 7.28 0.000 13026 
 STG 68 −30 16 5.27 0.033 37 
 SPL −17 −67 66 5.45 0.021 
Emotional−pitch-set 
 SFG −21 66 19 8.64 0.000 29466 
 ITG −44 −44 6.15 0.003 1698 
 Amygdala 27 −1 −20 5.77 0.009 368 
 MTG 56 −4 −23 5.31 0.030 44 
 Insula 35 −13 21 5.91 0.006 227 
 Caudate −20 −24 27 5.63 0.013 31 
 CalG −23 −55 10 7.54 0.000 70346 
 Cvermis −58 −44 5.75 0.010 245 
 CcrusII −30 −79 −36 5.60 0.014 219 

Note: CalG, calcarine gyrus; CcrusII, cerebellum crus II; Cvermis, cerebellar vermis; ITG, inferior temporal gyrus; MTG, middle temporal gyrus; postCG, postcentral gyrus; SFG, superior frontal gyrus; SPL, superior parietal lobe; STG, superior temporal gyrus.

aMNI coordinates.

bVoxel-level FWE-corrected statistics.

cTitle of contrasts in italics/anatomical locations of peak voxels. Regions are sorted anterior to posterior.

dLeft/right hemisphere.

eNumber of significant voxels in the same cluster.

Table 4

Imaging results from the PPI analysis based on the left DLPFC as seed region and the contrast Emotional−pitch-set

  Coordinatesa
 
Voxel levelb
 
 
Regionc Sided x y z t P Sizee 
MedOrbG 54 −8 5.25 0.034 35 
MedOrbG −5 53 −9 5.39 0.024 56 
IFG (orbitalis) −35 35 −11 5.17 0.042 
SMedG −6 30 61 5.69 0.011 100 
IFG (triangularis) 57 27 22 6.50 0.001 209 
MTG 59 −6 −21 5.16 0.043 
CCrusII 36 −76 −42 5.50 0.018 95 
CCrusII −30 −82 −39 6.16 0.003 752 
CCrusI 15 −82 −32 5.49 0.019 83 
  Coordinatesa
 
Voxel levelb
 
 
Regionc Sided x y z t P Sizee 
MedOrbG 54 −8 5.25 0.034 35 
MedOrbG −5 53 −9 5.39 0.024 56 
IFG (orbitalis) −35 35 −11 5.17 0.042 
SMedG −6 30 61 5.69 0.011 100 
IFG (triangularis) 57 27 22 6.50 0.001 209 
MTG 59 −6 −21 5.16 0.043 
CCrusII 36 −76 −42 5.50 0.018 95 
CCrusII −30 −82 −39 6.16 0.003 752 
CCrusI 15 −82 −32 5.49 0.019 83 

Note: CCrusI, cerebellum crus I; CCrusII, cerebellum crus II; IFG, inferior frontal gyrus; MedOrbG, medial orbital gyrus; MTG, middle temporal gyrus; SMedG, superior medial gyrus.

aMNI coordinates.

bVoxel-level FWE-corrected statistics.

cAnatomical location of peak voxels. Regions are sorted anterior to posterior.

dLeft/right hemisphere.

eNumber of significant voxels in the same cluster.

Figure 2.

Brain regions to which the right DLPFC ROI (green) show increased constraint-dependent (effective) functional connectivity. All colored areas represent significant increases at a FWE-corrected statistical threshold of P < 0.05. (A) Contrasting pitch-set—emotional conditions (yellow). (B) Contrasting emotional—pitch-set conditions (blue).

Figure 2.

Brain regions to which the right DLPFC ROI (green) show increased constraint-dependent (effective) functional connectivity. All colored areas represent significant increases at a FWE-corrected statistical threshold of P < 0.05. (A) Contrasting pitch-set—emotional conditions (yellow). (B) Contrasting emotional—pitch-set conditions (blue).

Supplementary Analysis

Since we, with the same dataset, were previously able to show a negative relation between improvisation experience and brain activity during improvisation in the right DLPFC and right angular gyrus (Pinho et al. 2014), we performed additional analyses to investigate if there was any interaction between conditions and improvisation experience (in primarily the right DLPFC). Figure 3 illustrates overlapping activity between 1) the contrast pitch-set−emotional, 2) regions where brain activity during improvisation is negatively related to improvisation experience, and (iii) the DLPFC ROI in the right hemisphere. Supplementary Figure S5 illustrates the relation between improvisation experience and brain activity during the 4 conditions with regard to the right DLPFC (i.e., the prefrontal cluster reported in the previous study). Lastly, we present results from 2 additional second-level analyses (Supplementary Table S4). These were based on the first-level contrasts pitch-set–rest and emotional–rest and covariates representing age, classical training, and improvisation training. The 2 analyses differed in that one of them also included the interaction term between brain activity and improvisation experience. Based on the similarity between outcomes, we argue that the interaction term does not influence the results in a way that motivates further action.

Figure 3.

Activity related to (i) the contrast pitch-set−emotional (FWE corrected) in red; (ii) regions where brain activity during improvisation is negatively related to improvisation practice (Pinho et al. 2014), in cyan; (iii) the right DLPFC ROI used here for the PPI analysis, in green (part of (i)); overlap between (i) and (ii), in yellow; overlap between (i), (ii), and (iii), in magenta.

Figure 3.

Activity related to (i) the contrast pitch-set−emotional (FWE corrected) in red; (ii) regions where brain activity during improvisation is negatively related to improvisation practice (Pinho et al. 2014), in cyan; (iii) the right DLPFC ROI used here for the PPI analysis, in green (part of (i)); overlap between (i) and (ii), in yellow; overlap between (i), (ii), and (iii), in magenta.

In order to assess if the results were biased by the removal of trials with pitch-set violations in the pitch-set conditions, we performed an additional first-level analysis without removing any trials due to pitch-set violations and subsequently compared the results of the contrasts pitch-set−emotional and emotional–pitch-set, to those where pitch-set violations were excluded, using a repeated-measures second-level GLM. Improvisation training, classical training, and age were entered as covariates. No significant differences were found in the whole-brain analyses, nor when doing small volume corrections based on the left and right DLPFC ROI used in the PPI analyses. A similar analysis was performed for the contrasts tonal–atonal and atonal–tonal, with similar results.

Discussion

In the present study, we sought to address a paradox in current creativity research, that is, why creative cognition has been linked to both activation and deactivation of prefrontal regions. As predicted, we found that the activity and connectivity of the DLPFC was highly dependent on condition, that is, improvisational constraint/task goal. In conditions that arguably favored an explicit approach to creative thinking, the DLPFC showed comparably higher activity and higher effective functional connectivity with premotor and parietal areas. In the emotional conditions, which may have allowed for the use of a more implicit strategy, the DLPFC showed less activity and greater effective functional connectivity with the default mode network (Fransson 2005). Based on these results, we argue that the DLPFC is not necessary for creative cognition per se, but rather that both the level and the regional distribution of prefrontal activity during creative performance is a function of the required level of explicit response selection during extemporization.

Pitch-Set Improvisations

During the pitch-set conditions, compared with the emotional conditions, there was an increase of activity in primarily the PMD and frontoparietal networks with a concomitant increase in effective functional connectivity between the DLPFC and the PMD, SMA, primary sensorimotor cortex, and the cerebellum. In other words, the visuospatial working memory system showed higher activity and stronger interactions with the motor system. Upon closer consideration and in view of all outcomes, it appears unlikely that this activity was “primarily” related to mental effort in maintenance of visual instructions (Zatorre et al. 1994; Volle et al. 2005) or purely related to perception–action transformations (Yamagata et al. 2012). Then we would presumably have seen a constraint-by-expertise interaction (since the emotional stimuli were arguably less complex), which we do not. In a previous study (using professional piano players), we investigated the parametric modulation of pitch-set size on brain activity during improvisation and pseudo-random sequence generation and found no difference between pitch-sets composed of either 2, 6, or 12 notes. Expertise and the ability to chunk items can alleviate working memory processes and allow pitch-sets to be perceived as complex stimuli that prime distributed representations across both visuospatial and auditory modalities, which are further associated with motor responses (Yang et al. 2014). Moreover, the pitch-set was in the present case not only maintained, but used actively throughout performance. Participants presumably also placed their fingers on several (up to 5 of 6) of the corresponding keys, which would greatly reduce the need for explicit maintenance in working memory. The sets were also presented in a musically “well-organized” way (ascending scales), which would make them easier to remember (Halpern and Bower 1982). Lastly, we found no indication that pitch-set violations would be related to brain activity in any region. We therefore speculate that pitch-set violations might have occurred mostly due to misperception or failure to encode all items during the 3.5-s instruction time. Such difficulties might increase particularly in atonal trials in which pitch-sets do not follow musically syntactic (tonal) rules.

Instead, we argue that DLPFC activity during pitch-set conditions was predominantly related to integrating goal-oriented information, that is, internal (musical) and external (response set) constraints, for attentional selection, that is, cognitive control of action sequencing and motor execution. Growing evidence supports that explicit response selection is based on competition between multiple actions that are prepared and represented in parallel in the PMD and posterior parietal cortex (Cisek 2007; Klaes et al. 2011). This competition can be biased by, for example, subjective strategies (Dorris and Glimcher 2004). The DLPFC is strongly implicated in response value computations (Camus et al. 2009; Sokol-Hessner et al. 2012) based on the conjunction of relevant sensory and cognitive information that accumulates over time (Domenech and Dreher 2010). The resulting frame of reference may bias competition and increase activity in cells representing a relatively better matching action alternative while suppressing others. In line with this model, the DLPFC and the PMD have been shown to coactivate specifically when attentional selection and sequence generation are to be unified for serial information processing (Abe et al. 2007). Thus, the DLPFC is suggested to inform selection and sequencing in the PMD of competing musical structures or phrases that are retrieved and represented in parallel in the dorsal visuomotor system. The DLPFC, PMD, and parietal cortex in combination, can therefore also be said to constitute an “intentional framework” for sensorimotor behavior (Shadlen and Kiani 2013), which essentially correspond to the extrospective task-positive network that routinely activates during and goal-directed motor task performance (Fox et al. 2005; Fransson 2005). Consistent with this view are several studies associating activity in these regions with subjective consciousness, the use of explicit strategies, intention, and sense of agency (e.g., Lau and Passingham 2006; Desmurget et al. 2009; Manenti et al. 2010). Our observed increase in effective functional connectivity between the right DLPFC on one hand and the PMV, SMA, and M1 on the other, could indicate that top–down cognitive control extends even to the level of motor execution. In support of this notion, Hasan et al. (2013) showed a task specific (free choice vs. externally specified action) timing and muscle specific influence of the DLPFC (BA 46) on the primary motor cortex.

Emotional Improvisations

When improvisation was constrained to express a certain emotion, activity in the DLPFC, PMD, and parietal regions decreased compared with pitch-set conditions—in several participants from above to below baseline—and the DLPFC BOLD signal showed increased correlations with default mode regions. Many of these typically introspective or task-negative regions, for example, the medial orbital and dorsomedial prefrontal cortices actually increased their activity when comparing emotional with pitch-set conditions. This finding is consistent with what Hutcherson et al. (2012) described as a shift from a lateral to a medial “value system” for cognitive regulation during decision making. The medial prefrontal cortex (MPFC) has been associated with the representation and elaboration of affective meaning of stimuli and perceiving one's affective state (Kober et al. 2008; Waugh et al. 2014). Importantly, MPFC regions have also been shown to integrate sensory information and provide predictions about specific outcomes associated with stimuli, choices, and actions—especially their moment-to-moment value based on current internal states (reviewed in Rudebeck and Murray 2014). Their (functional) interconnections with cortical, striatal, and limbic regions, such as the insula and amygdala, where activity also increased during emotional conditions, allow convergence of sensorimotor integration and visceromotor control in the processing of emotionally salient information and regulation of behavior. Thus, these regions interface between percepts, cognitive context, and core affect, also in relation to music (Koelsch 2010). Janata et al. (2002) was able to show that there is a topographic representation of tonality structure in the MPFC and—in a later study—that the same regions may enable associative processes between music, emotions, and memories (Janata 2009). The difference observed in the contrast tonal−atonal, that is, a higher activity in the ventral MPFC, could therefore indicate processing of tonality structure, but also that tonal sequences were processed as being more familiar or appreciated.

The relative increases of activity in the MPFC and insula were paralleled by increases in the IFG, rolandic operculum, STG, MCC, primary motor cortex, and occipital cortex. The IFG and STG are known to be involved in processing of musical syntax and semantics (Koelsch 2005). On a general level, the IFG has been implicated in implicit memory processes and controlled selection, retrieval and sequencing of semantic associates in competition, complex auditory-motor transformations, and serial production (Thompson-Schill et al. 1997; Uddén and Bahlmann 2012; Cogan et al. 2014). IFG activity has also been shown to correlate with both within- and between-person differences in both verbal and nonverbal divergent thinking (Gonen-Yaacovi et al. 2013) and Benedek et al. (2014) recently argued that IFG might have a particular role in overcoming dominant but uncreative responses. We suggest that while the value reference was explicitly elaborated in the MPFC to guide response selection during emotional conditions, it was chiefly controlled by the IFG (rather than the PMD), based on retrieval and sequencing processes that to a greater extent utilized internalized musical syntactic rules and semantic associations. Furthermore, the MPFC, insula and IFG in combination with the rolandic operculum, precentral gyrus, and middle cingulate cortex constitute most of the neural network identified for upper-limb motor imagery (Hetu et al. 2013). The occipital cortex is also implicated in music imagery (Meister et al. 2004) and melodic (vs. rhythmic) processing during piano performance (Bengtsson and Ullén 2006). In conjunction, our observations thus implicate differences beyond value or goal representation to what Cocchi et al. (2013) identified and described as a shift between 2 large-scale “meta-systems” for complex cognitive control.

In brief, these mechanisms would correspond to a shift from an executive meta-system where the DLPFC drives integration of sensory, autonomic, and goal-related information to implement adaptive control, to an integrative meta-system constituted primarily by the default mode network, where largely automated processes in specialized brain systems are organized under the influence of the MPFC for the flexible integration of exogenous and endogenous information. The present findings therefore also highlight that the default mode network is not task-negative per se, but can be actively engaged in goal-directed cognition and behavior depending on the nature of the task, that is, when drawing on interoceptive activities such as visual and motor imagery. A growing literature indicates that the combined functionality of the default mode network and IFG is particularly important for enabling divergent thinking, that is, for the combination of generative and selective processes (respectively) in line with theories of blind variation and selective retention (Beaty et al. 2014).

Functional relationships between the 2 meta-systems are typically reciprocal in nature (Fox et al. 2005; Fransson 2005). As a consequence, decreased or disrupted activity in one system may facilitate processing in the other (Galea et al. 2010; Manenti et al. 2010; Lee et al. 2013). Thus, the differences in activity between the lateral and medial prefrontal cortices during pitch-set and emotional conditions could have been additionally augmented by reciprocal interactions between the 2 described control systems. In fact, it would be difficult to explain why activity in the right DLPFC would drop below baseline levels in several participants, if the reduction was only related to a release from response set related processing. One could therefore also speculate on a suppression of systems involved in explicit response selection in favor of associative memory retrieval, sequence processing/fluency of performance, and the representation of affective information. A certain disinhibition of generative processes is also conceivable. The increased effective functional connectivity between the right DLPFC and active regions during emotional conditions, does however suggest some remaining role. We speculate this to be monitoring of progress toward goals over time (Benn et al. 2014) and preparation for action-driven conflict resolution (Schulte et al. 2009).

In line with the introduction and Clarke's description of generative principles being intermingled throughout performance (Clarke 1988), we could expect a dynamic use of strategies also in the present experiment. Furthermore, introducing a pitch-set constraint does not exclude emotional content; nor does introducing an emotional constraint exclude that the participant improvises using structural constraints. This is why we choose to phrase the influence of each type of constraint as a “bias” on cognition. It should be pointed out that contaminations such as those discussed (i.e., emotional processing in pitch-set conditions, and vice versa), will tend to reduce differences between experimental conditions. That we nonetheless find clear differences in brain activity and effective functional connectivity patterns when contrasting pitch-set and emotional conditions makes a strong case for having to acknowledge that creative cognition and associated neural systems may operate in a more complex, dynamic, and adaptive fashion than traditionally assumed.

We found higher behavioral motor output and complexity in emotional compared with pitch-set conditions. We controlled for motor output in the imaging analysis, but it is also important to note the direction of effects, emotional conditions, associated with higher motor output, brought about a deactivation of the DLPFC. If anything, the behavioral data thus support that activity was reduced in order to facilitate and/or not interfere with behavioral fluency in more structurally complex improvisations. Conversely, more cognitively demanding improvisations and increased activity in the DLPFC were related to lower Nkeys and reduced complexity. Importantly, the nuisance covariates appeared to be more related to basic aspects of perception and motor execution than higher order cognition. There was, for example, no effect of these regressors on brain activity in prefrontal or parietal regions.

Dual Strategies for Creative Ideation and Problem Solving

In summary, the results appear to largely support to our hypotheses initially derived from the dual pathway model. One could imagine creative thinking to range on a continuum between extremes of implicit and explicit cognition; and that along this dimension there is a point beyond which, prefrontal regions are recruited to ensure novel and original content, that is, when tasks introduce prerequisites that go beyond existing knowledge, skills, and abilities. Here during pitch-set conditions, cognition was pushed beyond that point when new information had to be perceived and used as a reference during subsequent improvisation. Expertise is likely another important factor. We have previously found pseudo-random generation by professional musicians to be accompanied by higher frontoparietal activity than improvisation (de Manzano et al. 2012b). In line with this, we recently also found—using the same dataset as here—that more improvisation training is linked to less activity in the frontoparietal working memory network and greater effective functional connectivity between prefrontal, premotor, and motor regions (there were no interactions between tasks and expertise) (Pinho et al. 2014). Possibly, once a sufficient level of domain expertise has been reached, a deactivation of the DLPFC and/or a disinhibition of generative processes becomes a viable and cognitively efficient approach, because long-term memory structures and retrieval strategies enable spreading activation and free association to yield spontaneous yet relevant responses at high fluency. If less experienced improvisers tried something similar, they would presumably produce less adequate responses that are either too simplistic or aesthetically inappropriate. Only with deliberate top–down control supported by prefrontal functionality might they reach similar aesthetic criteria. In some support of that argument, our less expert performers, who on average displayed higher activity in frontoparietal working memory systems, did not produce less complex improvisations. Thus, there might even in neural terms be dual strategies for creative performance.

The present results could also help to explain the apparent inconsistencies between earlier studies. We observe that studies showing a deactivation of the DLPFC used jazz pianists and studies showing an activation of the DLPFC used classical pianists. Improvisation is a much more central feature of jazz than classical music and jazz pianists are generally more experienced with improvisation. Second, returning to Clarke (1988): “[…] free jazz is principally characterized by associative structure, since it eschews the constraints of a pre-planned structure, and attempts to avoid the use of recognizable ‘riffs’. More traditional improvisation tends towards the hierarchical principle, in its adherence to a fairly strict harmonic outline.” Thus, when jazz and classical musicians are asked to perform spontaneous improvisations, they might tend toward different styles of improvisation and thus different creative strategies. Third, one could also point to specific methodological differences affecting cognitive load., for example, free improvisation on a well-learned scale or chord structure, as in Limb and Braun (Limb and Braun 2008), is conceivably a less intricate task than ornamentation of a melody presented “a prima vista,” as in Bengtsson et al. (2007). Lastly, choice of control task could have played a role. One could, for example, expect less explicit control and less deactivation of the default mode network while playing a scale or reproducing a composition from memory, as in Limb and Brown (2008), than when performing non-trivial a prima vista sight readings as in (de Manzano et al. 2012a, 2012b). This would have affected the likelihood of seeing an activation or deactivation in associated regions during improvisation.

Conclusion

We demonstrate that creative performance can be associated with both an increase and reduction of activity in the DLPFC (and other working memory regions) within-person, depending on task prerequisites. The findings clearly indicate that creativity is the result of a dynamic interplay between several brain regions, networks, and systems, and that the patterns of brain activity during creative problem solving depend strongly on the employed problem solving strategies. Specifically, we suggest that creative cognition can be biased toward either of 2 general cognitive strategies or control modes that to some extent differentiate with regard to involved neural systems, as indicated by the observer condition-based differences in the underlying distribution of regional brain activity and effective functional connectivity between the DLPFC and other brain areas. Thus, creative thinking is far more multifaceted in its nature and neural underpinnings than what has been commonly assumed, in particular with regard to goal representation and response selection. We suggest that a theoretical framework, which builds on the idea that some of the observed dynamics of creative thinking can be discussed in terms of an interaction between explicit and implicit cognition may be fruitful, and propose a number of factors which may influence cognitive strategies and neural processing.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

This research was supported by the Bank of Sweden Tercentenary Foundation (M11-0451:1), the Foundation for Science and Technology (FCT, Portugal; SFRH/BD/33895/2009), the Swedish Research Council (521-2010-3195) and the Sven and Dagmar Salén Foundation.

Notes

We thank Dr Jonathan Berrebi for technical support and Dr Rita Almeida for consultations on statistics, as well as Dr László Harmat and Diana Muessgens for assistance during experiments. Conflict of Interest: None declared.

References

Abe
M
,
Hanakawa
T
,
Takayama
Y
,
Kuroki
C
,
Ogawa
S
,
Fukuyama
H
.
2007
.
Functional coupling of human prefrontal and premotor areas during cognitive manipulation
.
J Neurosci
 .
27
:
3429
3438
.
Andersson
JL
,
Hutton
C
,
Ashburner
J
,
Turner
R
,
Friston
K
.
2001
.
Modeling geometric deformations in EPI time series
.
Neuroimage
 .
13
:
903
919
.
Ashburner
J
,
Friston
K
.
1997
.
Multimodal image coregistration and partitioning—a unified framework
.
Neuroimage
 .
6
:
209
217
.
Ashburner
J
,
Friston
KJ
.
2005
.
Unified segmentation
.
Neuroimage
 .
26
:
839
851
.
Beaty
RE
.
2015
.
The neuroscience of musical improvisation
.
Neurosci Biobehav Rev
 .
51
:
108
117
.
Beaty
RE
,
Benedek
M
,
Wilkins
RW
,
Jauk
E
,
Fink
A
,
Silvia
PJ
,
Hodges
DA
,
Koschutnig
K
,
Neubauer
AC
.
2014
.
Creativity and the default network: a functional connectivity analysis of the creative brain at rest
.
Neuropsychologia
 .
64C
:
92
98
.
Beilock
SL
,
Carr
TH
,
MacMahon
C
,
Starkes
JL
.
2002
.
When paying attention becomes counterproductive: impact of divided versus skill-focused attention on novice and experienced performance of sensorimotor skills
.
J Exp Psychol Appl
 .
8
:
6
16
.
Benedek
M
,
Jauk
E
,
Fink
A
,
Koschutnig
K
,
Reishofer
G
,
Ebner
F
,
Neubauer
AC
.
2014
.
To create or to recall? Neural mechanisms underlying the generation of creative new ideas
.
Neuroimage
 .
88
:
125
133
.
Bengtsson
SL
,
Csíkszentmihályi
M
,
UIlén
F
.
2007
.
Cortical regions involved in the generation of musical structures during improvisation in pianists
.
J Cogn Neurosci
 .
19
:
1
13
.
Bengtsson
SL
,
Ehrsson
HH
,
Forssberg
H
,
Ullén
F
.
2004
.
Dissociating brain regions controlling the temporal and ordinal structure of learned movement sequences
.
Eur J Neurosci
 .
19
:
2591
2602
.
Bengtsson
SL
,
Ullén
F
.
2006
.
Different neural correlates for melody and rhythm processing during piano performance from musical scores
.
Neuroimage
 .
30
:
272
284
.
Benn
Y
,
Webb
TL
,
Chang
BP
,
Sun
YH
,
Wilkinson
ID
,
Farrow
TF
.
2014
.
The neural basis of monitoring goal progress
.
Front Hum Neurosci
 .
8
:
688
.
Berkowitz
AL
,
Ansari
D
.
2008
.
Generation of novel motor sequences: the neural correlates of musical improvisation
.
Neuroimage
 .
41
:
535
543
.
Brodmann
K
.
1909
.
Vergleichende Lokalisationslehre der Grosshirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues
 .
Leipzig
:
Barth
.
Buckner
RL
,
Andrews-Hanna
JR
,
Schacter
DL
.
2008
.
The brain's default network: anatomy, function, and relevance to disease
.
Ann N Y Acad Sci
 .
1124
:
1
38
.
Camus
M
,
Halelamien
N
,
Plassmann
H
,
Shimojo
S
,
O'Doherty
J
,
Camerer
C
,
Rangel
A
.
2009
.
Repetitive transcranial magnetic stimulation over the right dorsolateral prefrontal cortex decreases valuations during food choices
.
Eur J Neurosci
 .
30
:
1980
1988
.
Cisek
P
.
2007
.
Cortical mechanisms of action selection: the affordance competition hypothesis
.
Philos Trans R Soc Lond B Biol Sci
 .
362
:
1585
1599
.
Clarke
EF
.
1988
.
Generative principles in music performance
. In:
Sloboda
J
, editor.
Generative processes in music
 .
Oxford
:
The Clarendon Press
. p.
1
26
.
Cocchi
L
,
Zalesky
A
,
Fornito
A
,
Mattingley
JB
.
2013
.
Dynamic cooperation and competition between brain systems during cognitive control
.
Trends Cogn Sci
 .
17
:
493
501
.
Cogan
GB
,
Thesen
T
,
Carlson
C
,
Doyle
W
,
Devinsky
O
,
Pesaran
B
.
2014
.
Sensory-motor transformations for speech occur bilaterally
.
Nature
 .
507
:
94
98
.
de Manzano
Ö
,
Ullén
F
.
2012a
.
Activation and connectivity patterns of the presupplementary and dorsal premotor areas during free improvisation of melodies and rhythms
.
Neuroimage
 .
63
:
272
280
.
de Manzano
Ö
,
Ullén
F
.
2012b
.
Goal-independent mechanisms for free response generation: creative and pseudo-random performance share neural substrate
.
Neuroimage
 .
59
:
772
780
.
Desmurget
M
,
Reilly
KT
,
Richard
N
,
Szathmari
A
,
Mottolese
C
,
Sirigu
A
.
2009
.
Movement intention after parietal cortex stimulation in humans
.
Science
 .
324
:
811
813
.
Diedrichsen
J
,
Shadmehr
R
.
2005
.
Detecting and adjusting for artifacts in fMRI time series data
.
Neuroimage
 .
27
:
624
634
.
Dietrich
A
,
Kanso
R
.
2010
.
A review of EEG, ERP, and neuroimaging studies of creativity and insight
.
Psychol Bull
 .
136
:
822
848
.
Doğanaksoy
A
,
Göloğlu
F
.
2006
.
On Lempel-Ziv complexity of sequences
. In:
Gong
G
,
Helleseth
T
,
Song
H-Y
,
Yang
K
, editors.
Sequences and their applications–SETA 2006
 .
Berlin
,
Heidelberg
:
Springer
. p.
180
189
.
Domenech
P
,
Dreher
JC
.
2010
.
Decision threshold modulation in the human brain
.
J Neurosci
 .
30
:
14305
14317
.
Donnay
GF
,
Rankin
SK
,
Lopez-Gonzalez
M
,
Jiradejvong
P
,
Limb
CJ
.
2014
.
Neural substrates of interactive musical improvisation: an FMRI study of “trading fours” in jazz
.
PLoS One
 .
9
:
e88665
.
Dorris
MC
,
Glimcher
PW
.
2004
.
Activity in posterior parietal cortex is correlated with the relative subjective desirability of action
.
Neuron
 .
44
:
365
378
.
Fox
MD
,
Snyder
AZ
,
Vincent
JL
,
Corbetta
M
,
Van Essen
DC
,
Raichle
ME
.
2005
.
The human brain is intrinsically organized into dynamic, anticorrelated functional networks
.
Proc Natl Acad Sci USA
 .
102
:
9673
9678
.
Fransson
P
.
2005
.
Spontaneous low-frequency BOLD signal fluctuations: an fMRI investigation of the resting-state default mode of brain function hypothesis
.
Hum Brain Mapp
 .
26
:
15
29
.
Friston
KJ
,
Buechel
C
,
Fink
GR
,
Morris
J
,
Rolls
E
,
Dolan
RJ
.
1997
.
Psychophysiological and modulatory interactions in neuroimaging
.
Neuroimage
 .
6
:
218
229
.
Friston
KJ
,
Frith
CD
,
Frackowiak
RS
,
Turner
R
.
1995
.
Characterizing dynamic brain responses with fMRI: a multivariate approach
.
Neuroimage
 .
2
:
166
172
.
Galea
JM
,
Albert
NB
,
Ditye
T
,
Miall
RC
.
2010
.
Disruption of the dorsolateral prefrontal cortex facilitates the consolidation of procedural skills
.
J Cogn Neurosci
 .
22
:
1158
1164
.
Gonen-Yaacovi
G
,
de Souza
LC
,
Levy
R
,
Urbanski
M
,
Josse
G
,
Volle
E
.
2013
.
Rostral and caudal prefrontal contribution to creativity: a meta-analysis of functional imaging data
.
Front Hum Neurosci
 .
7
:
465
.
Gray
R
.
2004
.
Attending to the execution of a complex sensorimotor skill: expertise differences, choking, and slumps
.
J Exp Psychol Appl
 .
10
:
42
54
.
Halpern
AR
,
Bower
GH
.
1982
.
Musical expertise and melodic structure in memory for musical notation
.
Am J Psychol
 .
95
:
31
50
.
Hasan
A
,
Galea
JM
,
Casula
EP
,
Falkai
P
,
Bestmann
S
,
Rothwell
JC
.
2013
.
Muscle and timing-specific functional connectivity between the dorsolateral prefrontal cortex and the primary motor cortex
.
J Cogn Neurosci
 .
25
:
558
570
.
Hetu
S
,
Gregoire
M
,
Saimpont
A
,
Coll
MP
,
Eugene
F
,
Michon
PE
,
Jackson
PL
.
2013
.
The neural network of motor imagery: an ALE meta-analysis
.
Neurosci Biobehav Rev
 .
37
:
930
949
.
Hutcherson
CA
,
Plassmann
H
,
Gross
JJ
,
Rangel
A
.
2012
.
Cognitive regulation during decision making shifts behavioral control between ventromedial and dorsolateral prefrontal value systems
.
J Neurosci
 .
32
:
13543
13554
.
Janata
P
.
2009
.
The neural architecture of music-evoked autobiographical memories
.
Cereb Cortex
 .
19
:
2579
2594
.
Janata
P
,
Birk
JL
,
Van Horn
JD
,
Leman
M
,
Tillmann
B
,
Bharucha
JJ
.
2002
.
The cortical topography of tonal structures underlying Western music
.
Science
 .
298
:
2167
2170
.
Klaes
C
,
Westendorff
S
,
Chakrabarti
S
,
Gail
A
.
2011
.
Choosing goals, not rules: deciding among rule-based action plans
.
Neuron
 .
70
:
536
548
.
Kober
H
,
Barrett
LF
,
Joseph
J
,
Bliss-Moreau
E
,
Lindquist
K
,
Wager
TD
.
2008
.
Functional grouping and cortical-subcortical interactions in emotion: a meta-analysis of neuroimaging studies
.
Neuroimage
 .
42
:
998
1031
.
Koelsch
S
.
2005
.
Neural substrates of processing syntax and semantics in music
.
Curr Opin Neurobiol
 .
15
:
207
212
.
Koelsch
S
.
2010
.
Towards a neural basis of music-evoked emotions
.
Trend Cogn Sci
 .
14
:
131
137
.
Lau
HC
,
Passingham
RE
.
2006
.
Relative blindsight in normal observers and the neural correlate of visual consciousness
.
Proc Natl Acad Sci USA
 .
103
:
18763
18768
.
Lee
TG
,
Blumenfeld
RS
,
D'Esposito
M
.
2013
.
Disruption of dorsolateral but not ventrolateral prefrontal cortex improves unconscious perceptual memories
.
J Neurosci
 .
33
:
13233
13237
.
Limb
CJ
,
Braun
AR
.
2008
.
Neural substrates of spontaneous musical performance: an fMRI study of jazz improvisation
.
PLoS One
 .
3
:
e1679
.
Liu
S
,
Chow
HM
,
Xu
Y
,
Erkkinen
MG
,
Swett
KE
,
Eagle
MW
,
Rizik-Baer
DA
,
Braun
AR
.
2012
.
Neural correlates of lyrical improvisation: an fMRI study of freestyle rap
.
Sci Rep
 .
2
:
834
.
Maldjian
JA
,
Laurienti
PJ
,
Burdette
JH
.
2004
.
Precentral gyrus discrepancy in electronic versions of the Talairach atlas
.
Neuroimage
 .
21
:
450
455
.
Maldjian
JA
,
Laurienti
PJ
,
Kraft
RA
,
Burdette
JH
.
2003
.
An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets
.
Neuroimage
 .
19
:
1233
1239
.
Manenti
R
,
Cotelli
M
,
Calabria
M
,
Maioli
C
,
Miniussi
C
.
2010
.
The role of the dorsolateral prefrontal cortex in retrieval from long-term memory depends on strategies: a repetitive transcranial magnetic stimulation study
.
Neuroscience
 .
166
:
501
507
.
Meister
IG
,
Krings
T
,
Foltys
H
,
Boroojerdi
B
,
Muller
M
,
Topper
R
,
Thron
A
.
2004
.
Playing piano in the mind—an fMRI study on music imagery and performance in pianists
.
Brain Res Cogn Brain Res
 .
19
:
219
228
.
Mylius
V
,
Ayache
SS
,
Ahdab
R
,
Farhat
WH
,
Zouari
HG
,
Belke
M
,
Brugieres
P
,
Wehrmann
E
,
Krakow
K
,
Timmesfeld
N
et al
.
2013
.
Definition of DLPFC and M1 according to anatomical landmarks for navigated brain stimulation: inter-rater reliability, accuracy, and influence of gender and age
.
Neuroimage
 .
78
:
224
232
.
Nijstad
BA
,
De Dreu
CKW
,
Rietzschel
EF
,
Baas
M
.
2010
.
The dual pathway to creativity model: creative ideation as a function of flexibility and persistence
.
Eur Rev Soc Psychol
 .
21
:
34
77
.
Oldfield
RC
.
1971
.
The assessment and analysis of handedness: the Edinburgh inventory
.
Neuropsychologia
 .
9
:
97
113
.
Pinho
AL
,
de Manzano
O
,
Fransson
P
,
Eriksson
H
,
Ullen
F
.
2014
.
Connecting to create: expertise in musical improvisation is associated with increased functional connectivity between premotor and prefrontal areas
.
J Neurosci
 .
34
:
6156
6163
.
Pressing
J
.
1988
.
Improvisation: methods and models
. In:
Sloboda
JA
, editor.
Generative processes in music
 .
New York
:
Oxford University Press
. p.
129
178
.
Rajkowska
G
,
Goldman-Rakic
PS
.
1995
.
Cytoarchitectonic definition of prefrontal areas in the normal human cortex: II. variability in locations of areas 9 and 46 and relationship to the Talairach Coordinate System
.
Cereb Cortex
 .
5
:
323
337
.
Rudebeck
PH
,
Murray
EA
.
2014
.
The orbitofrontal oracle: cortical mechanisms for the prediction and evaluation of specific behavioral outcomes
.
Neuron
 .
84
:
1143
1156
.
Schulte
T
,
Muller-Oehring
EM
,
Vinco
S
,
Hoeft
F
,
Pfefferbaum
A
,
Sullivan
EV
.
2009
.
Double dissociation between action-driven and perception-driven conflict resolution invoking anterior versus posterior brain systems
.
Neuroimage
 .
48
:
381
390
.
Schupp
HT
,
Stockburger
J
,
Bublatzky
F
,
Junghofer
M
,
Weike
AI
,
Hamm
AO
.
2007
.
Explicit attention interferes with selective emotion processing in human extrastriate cortex
.
BMC Neurosci
 .
8
:
16
.
Shadlen
MN
,
Kiani
R
.
2013
.
Decision making as a window on cognition
.
Neuron
 .
80
:
791
806
.
Smith
SM
.
2002
.
Fast robust automated brain extraction
.
Hum Brain Mapp
 .
17
:
143
155
.
Sokol-Hessner
P
,
Hutcherson
C
,
Hare
T
,
Rangel
A
.
2012
.
Decision value computation in DLPFC and VMPFC adjusts to the available decision time
.
Eur J Neurosci
 .
35
:
1065
1074
.
Thompson-Schill
SL
,
D'Esposito
M
,
Aguirre
GK
,
Farah
MJ
.
1997
.
Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation
.
Proc Natl Acad Sci USA
 .
94
:
14792
14797
.
Uddén
J
,
Bahlmann
J
.
2012
.
A rostro-caudal gradient of structured sequence processing in the left inferior frontal gyrus
.
Philos Trans R Soc Lond B Biol Sci
 .
367
:
2023
2032
.
Waugh
CE
,
Lemus
MG
,
Gotlib
IH
.
2014
.
The role of the medial frontal cortex in the maintenance of emotional states
.
Soc Cogn Affect Neurosci
 .
9
:
2001
2009
.
Volle
E
,
Pochon
JB
,
Lehericy
S
,
Pillon
B
,
Dubois
B
,
Levy
R
.
2005
.
Specific cerebral networks for maintenance and response organization within working memory as evidenced by the “double delay/double response” paradigm
.
Cereb Cortex
 .
15
:
1064
1074
.
Yamagata
T
,
Nakayama
Y
,
Tanji
J
,
Hoshi
E
.
2012
.
Distinct information representation and processing for goal-directed behavior in the dorsolateral and ventrolateral prefrontal cortex and the dorsal premotor cortex
.
J Neurosci
 .
32
:
12934
12949
.
Yang
T
,
Gathercole
SE
,
Allen
RJ
.
2014
.
Benefit of enactment over oral repetition of verbal instruction does not require additional working memory during encoding
.
Psychon Bull Rev
 .
21
:
186
192
.
Zatorre
RJ
,
Evans
AC
,
Meyer
E
.
1994
.
Neural mechanisms underlying melodic perception and memory for pitch
.
J Neurosci
 .
14
:
1908
1919
.