Prefrontal cortex involvement in creative problem solving in middle adolescence and adulthood

Graphical abstract

Highlights

• •We studied neural correlates of creative development in adolescents and adults.
• •Results show better creative problem solving capacities for adolescents.
• •Successful problem solving was associated with increased left latPFC activity.
• •Adolescents showed greater activations in bilateral latPFC.
• •These findings suggest better exploration capacity in adolescents.

Abstract

Creative cognition, defined as the generation of new yet appropriate ideas and solutions, serves important adaptive purposes. Here, we tested whether and how middle adolescence, characterized by transformations toward life independency and individuality, is a more profitable phase than adulthood for creative cognition. Behavioral and neural differences for creative problem solving in adolescents (15–17 years) and adults (25–30 years) were measured while performing a matchstick problem task (MPT) in the scanner and the creative ability test (CAT), a visuo-spatial divergent thinking task, outside the scanner. Overall performances were comparable, although MPT performance indicated an advantage for adolescents in creative problem solving. In addition, adolescents showed more activation in lateral prefrontal cortex (ventral and dorsal) during creative problem solving compared to adults. These areas correlated with performances on the MPT and the CAT performance. We discuss that extended prefrontal cortex activation in adolescence is important for exploration and aids in creative cognition.

1. Introduction

The human capacity for creative problem solving is of unparalleled quality. Defined as the generation of new yet appropriate ideas, insights, and solutions (Sternberg and Lubart, 1996), creative cognition has been critical throughout human evolution and serves important adaptive purposes (Runco, 2004). It is well known that adolescence is a period characterized by transformations toward life independency and individuality (Collins et al., 1997), and a crucial phase for the development of many cognitive abilities (e.g., Casey et al., 2008, Steinberg, 2005). Yet, relatively little is known about whether and how this age period is important for creative cognition. It has been argued that creative problem solving abilities are important skills facilitating the advancement toward mature adult functioning; a transformational trajectory that requires adaptive skills (e.g., Jaquish and Ripple, 1980). Hence, adolescence is expected to be an age period of enhanced creative abilities (Kleibeuker et al., 2013).

Creative problem solving typically requires divergent thinking (generating ideas by exploring many possible solutions), and flexibility in terms of restructuring and manipulating problem information. Consider, for example, the matchstick problem (Guilford, 1967) where a spatial composition including several matchsticks has to be restructured so as to form a new pre-described composition. Using these and related tasks, neuropsychological and brain imaging studies uncovered the importance of the lateral prefrontal cortex (PFC) in creative problem solving. Miller and Tippett (1996) compared healthy controls to patients with lesions in different brain regions, and found that patients with (right) frontal lesions showed impaired creative problem solving. Performances were most significantly impaired when flexibility in terms of strategy switches was required. Another study, in healthy adults, showed increased activation in bilateral ventral and dorsal prefrontal cortex when solving matchstick problems compared to verifying a given solution to a matchstick problem (Goel and Vartanian, 2005). Furthermore, activation in the right dorsolateral PFC (DLPFC) correlated with the percentage of traced solutions, indicating that this region contributes to exploratory success.

Whereas the role of the prefrontal cortex in the development of creative problem solving is largely unexplored, large scale longitudinal brain imaging studies showed profound changes during adolescence: gray matter volume in lateral PFC matures throughout adolescence, following an inverted U-shaped pattern with a peak in early adolescence (Gogtay et al., 2004). In addition, functional brain-imaging studies have reported different developmental trajectories showing that prefrontal cortex is both more activated (e.g., Adleman et al., 2002, Crone et al., 2006c), and less activated with increasing age (e.g., Durston et al., 2006, Morton et al., 2009). These age-related changes are sometimes interpreted as an increase of the ability to recruit referred brain regions, and other times as increasing efficiency of referred brain regions. Intriguingly, some studies reported a middle adolescent specific peak in activation in lateral prefrontal cortex (e.g., Crone et al., 2006a, Dumontheil et al., 2010) and thereby challenge the abovementioned relative simplistic maturational interpretations. An alternative possibility is that prefrontal cortex function during this transitional phase may be tuned specifically toward exploration and adaptive flexibility (Crone and Dahl, 2012, Dahl, 2008, Johnson and Wilbrecht, 2011), which in turn may be specifically beneficial for creative problem solving. Recent behavioral research on creative cognition, including early-, middle-, and late-adolescents and young adults revealed a peak for visuo-spatial divergent thinking for middle-adolescents (Kleibeuker et al., 2013). These results provide initial support for an alternative maturational view, indicating increased exploratory success for middle-adolescents compared to younger and older age groups.

Here we tested the alternative maturational possibility by examining (i) how creative problem solving performance develops from adolescence to adulthood, and (ii) how adolescent specific changes in PFC functioning relate to creative problem solving.

We tested middle adolescents versus adults applying a matchstick problem task (MPT) to assess visuo-spatial creative problem solving (inside the scanner), and the creative ability test (CAT; Van Dam and Van Wesel, 2006) to assess visuo-spatial divergent thinking (outside the scanner). We obtained this latter measure to reveal individual differences in brain activations during creative problem solving related to divergent thinking capacity. We focused on visuo-spatial tasks because performance is relatively independent of conceptual development and knowledge, both which differ substantially between age groups (Kavac et al., 2010).

Based on the results obtained by Goel and Vartanian (2005), we anticipated bilateral PFC activation during creative problem solving, and predicted that activation of these regions would correlate with creative problem solving performance, as measured with MPT, and with divergent thinking capacity, as measured with CAT. Based on the hypothesis that middle adolescence is a time window of enhanced neural activity in lateral prefrontal cortex, advantageous for exploration and adaptive flexibility (Luna et al., 2010, Dahl, 2011), we predicted that adolescents would show more activation in task-relevant prefrontal cortex areas than adults, and that this activation would be associated with better creative performance.

2. Methods

2.1. Participants

Forty-two participants with no history of neurological or psychiatric disorders participated in the present study, divided across two age-groups: 25 adolescents (15–17-year-olds) and 17 adults (25–30-year-olds). The final analyses involved thirty-six participants; 20 adolescents (M_age = 16.07 years, SD = .48, 11 male), and 16 adults (M_age = 27.03 years, SD = 1.81, 7 male). Two participants were excluded from the analysis due to technical failures. Four adolescents with lowest IQ-scores were excluded to avoid significant differences between age groups. Gender distributions did not differ between age-groups (χ² (1) = .44, p = .51).

Participants were recruited from local schools and through local advertisements. All participants provided informed consent. In the case of minors, consent was also obtained from primary caregivers. Participation was compensated with either money or course credits. All procedures were approved by the Medical Ethics Committee of Leiden University Medical Center (LUMC).

To obtain an estimate of intelligence we included two subscales of the Wechsler adult intelligence scale (Digit Span and Similarities; Wechsler, 1991, Wechsler, 1997; see Soveri et al., 2011). The estimated IQ scores did not differ between age groups (M_adolescents = 24.79, SD_adolescents = 2.25; M_adults = 27.00, SD_adults = 3.41; t (34) = 1.64, p = .12, corrected for unequal variances (Levene's test for equality of variances: p < .05)).

2.2. Creative problem solving tasks

2.2.1. Matchstick problem task

Participants were presented with a computerized MPT inside the scanner, consisting of a total of 48 matchstick problems (28 experimental, 20 control). Problems contained 22-match formations that consisted of eight fully formed squares (Fig. 1; Goel and Vartanian, 2005). Underneath the matchsticks, a caption instructed participants to remove a specified number of matchsticks in order to generate a specified number of fully formed squares. On experimental trials participants had to determine whether the problem was solvable (18 out of 28 problems were solvable). These problems required divergent thinking and set-breaking (shifting between representations of the problem space) as well as convergent thinking (to verify the correctness of a possible solution). On control trials, a certain number of matches were already crossed out. Participants had to determine whether the provided solution was correct (10 out of 20 control trials were correct). Fig. 1 shows a visual display of events captioned by the time line. Each trial started with a 4-second-presentation of one of two questions: “Is there a solution for the following problem” (experimental problems) or; “Is the following solution correct” (control problems). Next, an experimental or control matchstick problem was presented for 15 s. Participants could respond by pressing a button with the right index finger (for “no”) or with the right middle finger (for “yes”). A red border appeared after 12 s to indicate that there were 3 s left to respond. A fixation cross was presented in between trials with randomly varied duration (0–7.7 s, jitter). Experimental and control problems were presented in random order over three blocks (18 trials per block); there was no repetition of matchstick problems within or between experimental and control trials.

Fig. 1.

Time-line of a match problem task trial (see text for explanation).

The dependent variables were the percentages correct responses for experimental and control trials. In addition, we distinguished between trials that were solvable and trials that were not solvable.

2.2.2. Creative ability test

To assess divergent thinking in the visuo-spatial domain, we used a pencil and paper version of the creative ability test (CAT; Van Dam and Van Wesel, 2006), which was administered outside the scanner. This task consists of nine squares that include one to five open and/or filled circles. Participants were asked to search for triads of squares with corresponding properties (i.e., same number of circles, same position of circles), such that the other six squares would not correspond on this property. A valid solution would, for example, be the notation of three squares that included exactly four circles, whereas the other six included either more or less than four circles. There was a time limit of 10 min during which participants were requested to find as many triads as possible. The dependent variable was the number of correct solutions. One adolescent was not given the CAT because of logistical reasons.

2.3. Procedure

Outside the scanner, participants received instructions and completed a four-trial practice session of the MPT. Then they were acclimated to the MRI environment in a mock scanner. After completion of the scanning phase (during which they performed the MPT), they completed the WAIS subtests Digit Span and Similarities, as well as the CAT.

2.4. MRI data acquisition

Scanning was performed with a standard whole-head coil on a 3-Tesla Philips Achieva MRI system (Best, The Netherlands) in the Leiden University Medical Center. Three runs of 167 T2*-weighted whole-brain EPIs, preceded by two dummy scans to allow for equilibration of T1 saturation effects, were subsequently acquired (TR = 2.2 s; TE = 30 ms; flip angle = 80°; 38 transverse slices, 2.75 mm × 2.75 mm × 2.75 mm, +10% inter-slice gap). Stimuli were presented running E-prime software (version 1.2, Psychology Tools Inc.) and projected onto a screen at the head of the scanner bore. Participants viewed the stimuli by means of a mirror mounted on the head coil assembly. Head motion was restricted by using pillow and foam inserts that surrounded the head. The maximum movement parameters were below 3 mm and the maximum rotation was below .5° for all participants and all scans. In accordance with Leiden University Medical Center policy, all anatomical scans were reviewed and cleared by a radiologist from the Radiology department.

2.5. MRI data analysis

SPM5 software (Wellcome Department of Cognitive Neurology, London, UK, http://www.fil.ion.ucl.ac.uk) was used for image preprocessing and analyses. Images were corrected for slice-time differences, followed by rigid body motion correction. Functional volumes were spatially normalized to EPI templates based on MNI305 stereotaxic space (Cocosco et al., 1997) using a 12-parameter affine transformation together with a nonlinear transformation involving cosine basis functions. Data were resampled to 3 mm cubic voxels. Functional volumes were smoothed using an 8 mm full-width half-maximum 3D Gaussian kernel. For each participant, the functional time series were modeled by a series of events convolved with a canonical hemodynamic response function (HRF). Matchstick problems were modeled separately based on condition (experimental or control), solvability (solvable or unsolvable), and performance (correct or incorrect), with the time point of presentation as onset and response-time as duration, and entered in a general linear model along with a basic set of cosine functions to high-pass filter the data, and a covariate for run effects. In addition, the questions preceding the experimental and control trials were modeled as covariates of no interest (onset: presentation onset; duration: 4000 ms). Only correct trials were included in higher level analyses (number of trials: experimental solvable: M_adolescents = 4.20, 0–9; M_adults = 2.81, 0–6; experimental unsolvable: M_adolescents = 15.93, 7–10; M_adults = 15.19, 6–10; control solvable: M_adolescents = 9.25, 7–10; M_adults = 9.25, 7–10; control unsolvable: M_adolescents = 8.85, 7–10; M_adults = 8.69, 6–10). The least square parameter estimates of height of best fitting canonical HRF for each condition were used in pair wise contrasts (experimental > fixation; control > fixation; experimental > control). The resulting first level contrast images, computed on a subject-by-subject basis, were submitted to group analyses. At the group level, we performed one-tailed t-tests on these three contrasts, treating participants as a random effect, and two-sample t-tests to compare age groups.

We further conducted whole-brain regression analyses on the contrasts experimental > control and experimental > fixation to test for brain behavior relations using mean performance on experimental trials and performance on the CAT respectively. Whole brain fMRI analyses were FDR corrected for multiple comparisons (Genovese et al., 2002), with p < .05 and with at least 10 contiguous voxels. For whole-brain regression analyses and age group (2) × condition (2) analyses we applied both FDR correction (voxel level) and the commonly used threshold of p < .001 uncorrected with at least 10 contiguous voxels. Results are reported in the MNI305 stereotaxic space.

2.6. Region-of-interest (ROI) analyses

ROI analyses were performed to illustrate creative problem solving differences in lateral PFC between the age-groups. ROIs were derived from the whole brain contrast experimental > control, including inferior frontal gyrus triangularis (IFG-tri) and dorsolateral prefrontal cortex (DLPFC). Analyses were performed using the MarsBaR toolbox in SPM5 (Brett et al., 2002), averaging signals across the voxels that make up an ROI.

3. Results

3.1. Matchstick problem task

3.1.1. Performances

To test for age differences in problem solving performance we conducted a 2 (condition) × 2 (solvability) × 2 (age group) mixed ANOVA. Results, which are presented in Table 1 and Fig. 2, showed three significant effects: a main effect of condition with more correct answers for control problems than experimental problems (F(1,34) = 620.52, p < .001; η² = .94), a main effect of solvability with more correct answers for unsolvable than solvable problems (F(1,34) = 314.64, p < .001; η² = .89), and an interaction effect between solvability and condition (F(1,34) = 510.86, p < .001; η² = .93). As can be seen in Fig. 2 (left panel), the differences in accuracy for the control versus experimental conditions were present for solvable problems but not significant for unsolvable problems (solvable: F(1,34) = 951.69, p < .001; η² = .97; unsolvable: F(1,34) = .37, p > .5; η² = .01).

Table 1.

Means and standard deviations for MPT and CAT performances.

	15–17 years		25–30 years
	M	SD	M	SD
MPTa (% correct)
Experimental
Solvable	23.33	14.60	15.63	7.09
Unsolvable	88.50	8.75	84.38	14.59
Control
Solvable	92.50	7.86	92.50	8.56
Unsolvable	88.50	8.13	86.88	11.95
MPTb (RT in ms)
Experimental
Solvable	9918.24	2106.05	10163.39	2265.33
Unsolvable	10555.35	1047.51	10179.91	1892.62
Control
Solvable	8474.22	1747.87	8399.39	1939.51
Unsolvable	7973.13	2095.15	7623.33	1967.50
CATc (nr correct)	10.21	2.44	8.81	2.78

^a15–17 years, n = 20; 25–30 years, n = 16.

^b15–17 years, n = 18; 25–30 years, n = 15.

^c15–17 years, n = 19; 25–30 years, n = 16.

Fig. 2.

Performance on the match problem task. Accuracy measures (% correct) are presented left, response times (ms) are presented right. *p < .05.

There were no significant overall effects including age group (all p's > .05). However, our main interest was in solvable experimental problems because these problems especially represented creative problems, requiring generation of new, appropriate representations of the problem space. Analysis on accuracy for solvable experimental problems, applying an independent t-test, revealed a significant age-group effect showing better performance for adolescents than adults (t (34) = 2.08; p = .047; corrected for unequal variances).

3.1.2. Response times

Response times for correct trials are shown in Table 1 and Fig. 2 (right panel). One adult and two adolescents were not included in the analysis because they did not have observations for correct solvable experimental trials. A 2 (condition) × 2 (solvability) × 2 (age group) mixed-model ANOVA on response times revealed a significant main effect of condition (F(1,31) = 61.30, p < .001; η² = .67) with longer response times for experimental trials, and a condition × solvability interaction effect (F(1,31) = 6.58, p = .015; η² = .18). The interaction effect was determined by larger RT differences for unsolvable trials (F(1,31) = 80.44, p < .001; η² = .72) relative to solvable trials (F(1,31) = 19.66, p = .001; η² = .39). No age groups effects were revealed (all p's > .05), rendering it unlikely that age group effects for fMRI results were related to response time differences.

3.2. Creative ability test

Results for the CAT are displayed in Table 1. Even though the mean number of correct solutions for the adolescent-group was higher than that for the adult-group, the group difference for visuo-spatial creative fluency was not significant (F(1,34) = 2.51, p = .12; pη² = .07).

3.3. fMRI results

3.3.1. Whole-brain comparisons

To extract the activation patterns related to creative problem solving we conducted whole-brain voxel-wise t-tests on activation levels for the contrast correctly solved experimental (E) problems > correctly solved control (C) problems (E > C) across all participants (n = 36). These analyses were performed collapsed across solvable and unsolvable conditions because these conditions resulted in similar activation patterns (see ROI analyses below for an exception). Results revealed a number of significantly activated regions, which are presented in Fig. 3a and Table 2, including left inferior frontal gyrus (IFG) and left middle frontal gyrus (MFG) (FDR corrected, p < .05).

Fig. 3.

Whole brain activations for (a) the contrast experimental > control overall (n = 36; p < .05, FDR corrected, >10 contiguous voxels), and (b) adolescents > adults for the contrast experimental > control (significance threshold set at p < .001, uncorrected, >10 contiguous voxels for illustrative purposes). Below (c) mean parameter estimates for correctly solved experimental and control matchstick problems are shown for left and right lateral PFC functional ROIs. Left graph: right DLPFC (ROI-peak-value at MNI coordinates 33, 24, 55). Right graph: left IFG (ROI-peak-value at MNI coordinates −51, 27, 27). **p ≤ .001.

Table 2.

Neural activations for the contrast experimental > control for all participants.

Brain regions	L/R	K	Z-Value peak voxel	MNI coordinates
				x	y	z
Inferior frontal gyrus (tri), medial frontal gyrus	L	159	5.22	−45	36	24
Middle cingulate cortex, superior medial gyrus	L	22	4.09	−12	27	33
			3.55	−9	24	42
Inferior parietal lobule	L	11	4.01	−30	−60	39

Note. n = 36.

Significance threshold set at p < .05, FDR corrected; >10 contiguous voxels.

To test for developmental differences, a two-sample t-test (adolescents versus adults) was conducted on the contrast E > C. The contrast adults > adolescents revealed no significant results. The reversed contrast adolescents > adults, however, resulted in increased activation in left IFG and MFG, left inferior parietal lobule, and right DLPFC (uncorrected, p < .001, >10 voxels; see Fig. 3b and Table 3). The regions in left IFG and MFG overlapped with the area within the left IFG identified in the main contrast E > C across participants. Moreover, the right DLPFC region overlapped with a region previously associated with an increasing numbers of correct solutions for the matchstick problems (Goel and Vartanian, 2005). Notably, left IFG and right DLPFC remained significant in the two-sample t-test when applying false discovery rate (FDR) correction (p < .05; >10 contiguous voxels; see Table 3).

Table 3.

Neural activations for adolescents > adults for the contrast experimental > control.

Brain regions		L/R	K	Z-Value peak voxel	MNI coordinates
					x	y	z
Frontal lobe	Inferior frontal gyrus (tri), middle frontal gyrus (DLPFC)	R	77	4.73	33	24	15*
				4.16	39	33	18*
				3.53	48	42	24
	Middle orbital gyrus	L	32	4.17	−27	45	−9*
				3.44	−18	−54	−9
	Precentral gyrus	L	67	4.08	−45	3	42*
				3.48	−36	−3	36
	Inferior frontal gyrus (tri)	L	58	3.87	−51	27	27*
				3.53	−39	24	27
	Precentral gyrus, postcentral gyrusa	R	20	3.78	54	−18	45
				3.18	45	−12	45
	Inferior frontal gyrus (tri)	L	10	3.65	−39	24	15
	Inferior frontal gyrus (tri), middle frontal gyrus	L	41	3.48	−36	42	12
				3.37	−48	42	9
				3.15	−33	42	21
	Supra marginal area	L	10	3.48	−6	21	45
Parietal lobe	Inferior parietal lobule	L	48	3.61	−36	−54	45
				3.30	−36	−66	51
Temporal lobe	Superior temporal gyrus	R	44	4.43	60	−27	9*
	Middle temporal gyrus	L	11	3.18	−48	−6	−15
Occipital lobe	Superior occipital gyrus	R	27	3.99	24	−78	36*
	Cuneus	R	13	3.53	21	−87	9
	Middle occipital gyrus	L	12	3.61	42	−78	3
	Inferior occipital gyrus	L	25	3.41	−33	−78	−9
				3.36	−45	−75	−12
Basal ganglia	Caudate nucleus	L	15	4.22	−18	3	18
	Pallidum	L	12	3.48	−15	−6	−3
Cerebellum		R/L	314	4.31	6	−66	−27*
				3.99	−9	−69	−30*
				3.88	−9	−75	−15

Note. N_adults = 16, N_adolescents = 20.

^aCluster also includes parietal lobe structures (postcentral gyrus).

^*Regions that are significant with the stricter significance threshold of p < .05, FDR correction, >10 voxels.

3.3.2. ROI analyses

To inspect the patterns revealed in the age group analyses visually, post hoc ROI analyses of variances were conducted on regions within the left IFG and right DLPFC identified in the two-sample t-test for the contrast E > C (p < .001, uncorrected; similar results were obtained for p < .05, FDR corrected). Results are displayed in Fig. 3c, which shows that adolescents but not adults recruited left IFG more for experimental compared to control matchstick problems (condition effect_adults: F(15,1) = 47, p > .10; condition effect_adolescents: F(19,1) = 34.46, p < .001). A largely similar pattern is observed for the right DLPFC, with more activation during experimental relative to control problems for adolescents (F(19,1) = 13.91, p = .001), but an opposite effect (C > E) for adults (F(15,1) = 27.52, p < .001).

Additional analyses of variances on solvability effects, applying 2 (condition) × 2 (solvability) × 2 (age group) mixed ANOVAs, revealed a significant main effect of solvability for left IFG (F(31,1) = 4.41, p = .044), but none of the interaction effects was significant (all p's > .1). However, there was a significant condition × solvability × age group interaction (F(1,31) = 14.97, p = .001) for the right DLPFC. The condition × age group interaction effect for the right DLPFC was larger for the solvable problems (F(31,1) = 39.71, p < .001) than for unsolvable problems (F(31,1) = 4.32, p = .05). These findings suggest that the condition × age group effect for activation in the right DLPFC region mainly relies on creative success for matchstick problems.

3.3.3. Individual differences

To test for brain areas directly related to creative problem solving performance, we conducted whole-brain voxel-wise regression analyses on the contrast E > C (significance threshold p < .001, uncorrected) with performance on experimental trials (E-correct) as covariate of interest. Results showed significant activation in a region in the left IFG directly adjacent to the abovementioned IFG region observed for the contrast adolescents > adults in the whole-brain E > C analysis (see Fig. 4; ROI-peak-Z-value = 3.68 at MNI coordinates (−48, 12, 27), 34 contiguous voxels). Notably, the correlation between the contrast E > C and E-correct was significant after controlling for age group (r_partial (33) = .46, p = .005), confirming that this region is involved in successful task performance independent of age.

Fig. 4.

Left: whole brain activations for the regression on the contrast experimental > control with performance on experimental matchstick problems (% correct) (p < .001, uncorrected; >10 contiguous voxels; section coordinates: Y = 12, Z = 27). Right: correlation between mean activation for left IFG (ROI-peak-value at MNI coordinates −48, 12, 27) and experimental matchstick problems performance for adolescents (open circles) and adults (black circles).

Next, we performed whole-brain voxel-wise regression analyses with post-scanning CAT performance (number of correct solutions) to gain knowledge on brain activations related to visuo-spatial creative cognition in general (p < .001, uncorrected). No significant relations were observed for the contrast E > C. However, a regression analysis on E > fixation revealed a positive relation between activation right DLPFC and CAT performance (see Fig. 5; ROI-peak-Z-value = 3.64 at MNI coordinates (39 33 27), 14 contiguous voxels). This region was in close proximity with the right DLPFC area showing a significant condition × age group interaction effect for the contrast E > C.

Fig. 5.

Left: whole brain activations for the regression on the contrast experimental > fixation with performance on the CAT (number of correct solutions) (p < .001, uncorrected; >10 contiguous voxels: section coordinates: Y = 33, Z = 27). Right: correlation between mean activation for right DLPFC (ROI-peak-value at MNI coordinates 39, 33, 27) and CAT performance for adolescents (open circles) and adults (black circles).

4. Discussion

In the present study, we examined developmental differences in creative problem solving capabilities and related these to brain activation patterns during a matchstick problems task. We hypothesized that adolescence is a period of enhanced PFC activation for exploration and adaptive purposes. The behavioral results showed that creative problem solving is already well developed in middle adolescents. Overall performance on creative problem solving did not differ between adolescents and adults, but age groups differences of solvable experimental problems of the MPT indicate better creative problem solving capacities for the middle adolescents.

The second part of our research goal concerned the question whether there were developmental differences in the underlying brain regions that support creative cognition. Brain imaging data yielded three main results: (1) Consistent with prior studies, we found increased activation of left lateral PFC during successful creative problem solving. (2) A direct comparison between age groups revealed increased activation in left IFG and right DLPFC during successful creative problem solving for adolescents compared to adults. (3) Individual differences analyses revealed that activation in left IFG and right DLPFC during successful creative problem solving was correlated with performance on experimental matchstick problems and on a visuo-spatial divergent thinking task that was administered outside the scanner, respectively. These results imply that adolescents, relative to adults, have a tendency to recruit relevant prefrontal brain areas during creative problem solving and show activity patterns common to persons with high divergent thinking capacities. Below, possible mechanisms underlying the observed results are discussed in further detail.

4.1. Lateral PFC and representation selection

Consistent with prior studies using a matchstick problems task (Goel and Vartanian, 2005), and a different visuo-spatial creative thinking task (Aziz-Zadeh et al., 2012), we found that left lateral PFC was significantly more active during experimental relative to control problems. These findings underscore the importance of left lateral PFC areas in thinking and problem solving (e.g., Gazzaniga, 2000). More specifically, these processes might comprise generating and choosing among various representations of the problem space and processing conflicting representations. Indeed, previous studies have shown left IFG involvement during both verbal (Thompson-Schill et al., 2002) and non-verbal (Brandon et al., 2004) tasks that require overriding a highly activated representation or selecting among weakly activated, incompatible representations (Hirshorn and Thompson-Schill, 2006). These two forms of conflict create demands for cognitive control and indicate a specific role of the left IFG for switching between representations (see also Crone et al., 2006b).

Besides the left IFG, adolescents also showed activation in the right DLPFC during experimental matchstick problems. Goel and Vartanian (2005) reported that this region was specifically important for finding more solutions for experimental matchstick problems. Interestingly, in the current study this area was also positively correlated with performance on a separate creativity task (CAT), measuring visuo-spatial divergent thinking, that was taken outside the scanner. These results suggest that persons with high exploratory capacities have tendency to recruit right DLPFC to a relative large extent during successful creative problem solving.

Right DLPFC activations have previously been implicated in working memory processes (Crone et al., 2006c, Curtis and D‘Esposito, 2003, Jolles et al., 2011), which fits with recent results showing a positive correlation between working memory and creative cognition (De Dreu et al., 2012). Other studies indicated the importance of right DLPFC in higher level cognitive control functions such as monitoring behavior in accordance with task goals (Shallice, 2004), and planning and manipulating internal representations of problem information (Ruh et al., 2012), Impaired right DLPFC functioning, on the other hand, has been associated with impulsive behavior (Ridderinkhof et al., 2011).

These prior findings, together with our present neuroimaging results, suggest a controlled though flexible manner of processing, especially in middle adolescents, that is successful for creative problem solving.

4.2. Developmental differences

A direct comparison of activation demonstrated that these task-relevant areas in prefrontal cortex, IFG and DLPFC, are more engaged in adolescents than in adults. This finding is consistent with prior studies that reported increased activation in lateral prefrontal cortex in middle adolescents (see Crone and Dahl, 2012 for an overview). Such age-related decreases are generally ascribed to less efficiency of functional networks including the referred brain regions, whereas age-related increases are commonly interpreted as insufficient recruitment of late-developing brain regions (reviewed in Crone and Ridderinkhof, 2011). However, these maturational interpretations might be too simplistic. Indeed, several studies found adolescent specific peaks in prefrontal regions for working memory tasks, inhibition tasks, relational reasoning tasks, and task shifting (e.g., Crone et al., 2006a, Dumontheil et al., 2010, Geier et al., 2009, Velanova et al., 2009).

An alternative hypothesis considered here is that increased activation in prefrontal cortex during middle adolescence is important for this developmental period and provides advantages for this phase in life. For example, the road toward adult individuality includes leaving parental custody and building one's own life in a world full of opportunities as well as uncertainties. To achieve these goals, it is presumably beneficial to be tuned toward exploration and adaptive flexibility (Dahl, 2011, Crone and Dahl, 2012), which in turn may be associated with better, rather than less developed creative problem solving abilities (Kleibeuker et al., 2013; see also De Dreu et al., 2011). This hypothesis finds support in a recent animal study (Johnson and Wilbrecht, 2011) that found that explorative learning is more adaptive in adolescent than adult mice. In agreement with these findings, analyses of the behavioral patterns in the current fMRI study indicated a slight advantage for adolescents relative to adults on problems that required creative cognition most evidently (solvable experimental matchstick problems). The current neuro-imaging data provide additional evidence into the direction of the alternative maturational hypothesis by showing that adolescence recruit brain regions in a way common to persons with high divergent thinking and creative problem solving capacities.

This hypothesis should be tested in more detail in future research, but provides important implications putting forward that middle adolescence is an essential phase in cognitive development.

Limitations of the current study include the use of a cross-sectional rather than a longitudinal design. Interpretations of developmental changes from adolescence toward adulthood made in this study are consequently merely suggestive and require confirmation from analyses of creative cognition over time. Inclusion of an additional age group of younger participants in future studies would make it possible to explicitly test the hypothesis that adolescence is a period of a peak in effective increased prefrontal activations. Another issue to be considered for future research is the use of tasks from other domains. Creativity is a general concept that covers outcomes from a wide range of fields. It would be interesting to investigate whether the observed age-related differences are specific to the visuo-spatial requirements of the current tasks, or more general effects existing across creativity domains.

4.3. Conclusion

Taken together, we showed that middle adolescents reveal high levels of creative problem solving. To the best of our knowledge, this study was the first to demonstrate that adolescents show more activation in prefrontal cortex than adults in a manner which is task relevant, and, thereby, contrasts with a simplistic maturational view of functional brain development (Johnson, 2011). We suggest an alternative hypothesis arguing that increased activation in prefrontal cortex is adaptive for creative problem solving during a phase in life that is tuned toward exploration and developing individuality.

Conflict of interest

There is no known conflict of interest.