Seven Questions about Creativity and Creative Thinking

3. Is creative thinking domain-specific, a more general skill – or both?

Can students be creative no matter what they do? Or is their creativity bound to specific areas of expertise? Understanding whether creativity is domain-general or domain-specific has been a subject of ongoing research over the past century (Miravete and Tricot, 2024[1]). Early research considered creative abilities to be influenced by a set of enduring personality traits and thinking patterns that support creativity across many kinds of tasks. More recently, research has shifted towards a domain-specific understanding of creativity, acknowledging that creative work in different domains draws upon different internal resources including knowledge, skills and attitudes. Some researchers argue that the domain-general vs. domain-specific dichotomy in creativity research is unhelpful as it obscures the interplay of both domain-general (“context-free”) and domain-specific (“context-dependent”) factors that support the creative thinking process (Plucker and Beghetto, 2006[2]).

The PISA 2022 assessment reported students’ performance on the creative thinking test as part of a unidimensional scale, following the application of an Item-Response-Theory (IRT) model to students’ test-item scores (OECD, 2024[3]). The IRT model indicated weakly correlated residuals between test items, supporting the construction of a unidimensional scale and suggesting that student performance on the PISA test largely captured an underlying latent creative thinking trait. Nonetheless, this latent scale obscured some differences in performance across task types and domain contexts. Item-level analyses found that students across countries and economies showed different strengths and weakness in creative thinking (OECD, 2024[4]). The PISA 2022 results lend support to the idea that creative thinking is supported by both context-free and context-dependent factors, both of which may be amenable to development through education.

What can the PISA 2022 Rescoring project data contribute to this debate? As described in the Introduction of this report, the PISA 2022 scoring method evaluated the extent to which students could successfully engage in idea generation, but it did not evaluate the holistic creativity of students’ ideas nor identify the most creative responses within each country and economy. The PISA CT Rescoring project allows us to do this. To what extent are students capable of coming up with very creative ideas across different tasks and domain contexts? Who are these high-scoring creative all-rounders, and what factors might support consistent creative performance across different types of tasks?

3.1. Did students perform consistently across creative thinking tasks?

Of the seven tasks included in the PISA 2022 Rescoring project, students in the sample completed between two and four tasks each (see Annex A1). Table 3.1 presents the correlations of residuals between the seven tasks included in the study, on average across country-language groups, illustrating how student scores across items relate after accounting for a shared underlying trait. The generally weak correlations support the idea that performance in the PISA test was driven primarily by a single dimension (i.e. skill in creative thinking) rather than task- or domain-specific factors. Nonetheless, some weak clustering between tasks is apparent – particularly within the problem-solving tasks asking students to generate multiple ideas (Tasks 5 to 7; see Figures B-H in the Introduction for a presentation of the tasks included in the study). This may be expected to some extent, as the task demands and format in this group (generate one or more solution ideas of any quality) are more similar than those of the remaining task types. Indeed, task design and instructions have been found to affect performance and cognitive processes in creative thinking tasks (Acar and Shen, 2025[5]).

Table 3.1. Average residual correlations between tasks included in the PISA CT Rescoring study

	Creative expression tasks			Creative problem-solving tasks
	Task 1	Task 2	Task 3	Task 4	Task 5	Task 6	Task 7
Task 1 – Science Fair Poster	1	0.04	-0.02	-0.09	- 0.04	-0.06	0.02
Task 2 – Illustration Titles		1	0.06	0.04	0.07	0.06	0.13
Task 3 – 2983			1	0.03	0.11	0.11	-0.01
Task 4 – Save the Bees				1	0.07	0.06	0.15
Task 5 – Food Waste					1	0.10	0.17
Task 6 – Save the River						1	0.12
Task 7 – The Exhibit							1

Note: Bolded values show significant correlations at the average country-language group level. A Generalized Partial Credit Model (GPCM) was modelled for each country-language group, using the (unstandardised) holistic creativity scores. Higher values of residual correlation can be interpreted as evidence of other shared factors, beyond a shared underlying creative thinking skill, that influences performance in task pairs.

Source: Table B3.1

Another way to consider the extent to which students can think creatively across different tasks is to examine their success at the individual level. The PISA CT Rescoring project sample included 5530 students who completed at least three different tasks (around 300-400 students per country-language group) (Annex A1). Figure 3.1 shows that while just over half of all students (52%) that sat at least three tasks were high-scoring in one or two of them – with “high-scoring” defined as scoring in the top quartile of the standardised holistic creativity scores relative to their country-language-task group peers (see Box 3.1) – only 8% of students were high-scoring in three or more tasks. Moreover, over 1 in 3 students were not able to come up with a single high-scoring response relative to their country-language group peers in any of the tasks included in the study.

Box 3.1. Which score data are used in this chapter?

The analyses in this chapter mainly draw upon the holistic creativity scores (Annex A2). Each student response was scored based on judges’ evaluations of the overall creative quality of the response. The holistic creativity score reflects a relative ranking of creative quality compared to all other responses in that country-language-task group. To account for differences in the distribution of scores across country-language-task groups, the analyses in this chapter standardised and transformed the raw holistic creativity scores onto a 1–10-point scale, with a mean of 5 score points.

Terms used in this chapter

• Jittered standardised holistic score: For some of the analyses in this chapter, the standardised holistic creativity scores were jittered. This involves adding a small random error, to the magnitude of 1e-6, onto each standardised holistic creativity score. The jittered scores ensure that there is an even distribution of students across quantiles.

• High-scoring answer: A student response was classified as high scoring if its jittered standardised holistic score fell within the top quartile of scores (i.e. at or above the 75^th percentile) relative to all other responses within its country-language-task group. Note that the creative quality of high-scoring answers across country-language groups may vary due to the relative ranking nature of the holistic scoring method.

• Creative all-rounders: Students that responded to three or four tasks in the PISA CT Rescoring project and who achieved at least three high-scoring answers. Students with three or four high-scoring answers were considered together given the small sample sizes (see Annex A1).

This chapter focuses on examining the relative shares of creative all-rounders across country-language groups and understanding the factors that influence students’ capacity to consistently achieve high holistic creativity scores across different types of task.

Figure 3.1. Few students across country-language groups manage to be creative all-rounders

Share of students that score in the top quartile of standardised holistic creativity scores in multiple tasks (on average across country-language groups)

Note: Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). Results shown are at the average country-language group level and are rounded to the nearest integer. “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability.

Source: Table B3.2

This relatively low share of “creative all-rounders” across country-language groups – defined as students who achieved at least three high-scoring responses (see Box 3.1) – might be affected by the types of tasks included in our study. The four tasks classified as ‘generate diverse ideas’ tasks only instructed students to provide ideas that were as different to each other as possible, but to achieve a high holistic creativity score, students may have needed to suggest one or more creative ideas within the same response. The relative ranking of the holistic scoring method mitigates this issue to some extent, as responses are only compared to others in the same country-language-task group, but these task types may nonetheless place relatively high demands on students and thus make it harder to consistently achieve high-scoring responses across tasks.

Figure 3.2 shows that the share of creative all-rounders within each country-language group varies slightly, with around 10% of students in Portugal and Brazil achieving high-scoring responses compared to their country-language peers across three or more tasks, compared to around 7% of students who performed well across multiple tasks in Italy, Saudi Arabia and Hong Kong (China). In countries with fewer creative all-rounders compared to the country-language average, there is a more varied spread of students that come up with creative ideas across the different tasks in the study.

Figure 3.2. Some country-language groups have more creative all-rounders than others

Percentage of students in each country-language group that achieved three or more high-scoring responses (relative to their country-language-task group)

Note: Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). Value labels shown in the figure are rounded to the nearest integer. “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results show the average of 20 replications of the analysis to account for sampling and jitter-induced variability.

Source: Table B3.2

3.2. Who are the creative all-rounders and what characteristics do they share?

Students who achieved high holistic creativity scores across three or more tasks were relatively rare, representing less than 1 in 10 students on average across country-language groups and task combinations. What factors could account for the variation in student performance across tasks and differences in the proportion of creative all-rounders across country-language groups?

3.2.1. Students’ socio-demographic background (gender, socio-economic status) and academic performance

Table 3.2 shows the share of boys and girls amongst students who provided different numbers of high-scoring responses. The share of female students increases within each group as the number of high-scoring responses increases, with the gender gap being largest amongst the creative all-rounders group (59% of creative all-rounders are girls compared to 41% boys, on average across country-language-task groups). As reported in the PISA 2022 results (OECD, 2024[4]) and explored later in Chapter 6 of this report, girls tended to perform better in the PISA 2022 creative thinking tasks overall – so finding that they also managed to achieve high-scoring responses more consistently than boys is no surprise.

Table 3.2. More girls than boys are creative all-rounders

Percentage of girls and boys (on average across country-language groups), by number of high-scoring responses

High-scoring answers	Share of girls	Share of boys	Difference (Girls-boys)
0	47%	53%	-6%
1	49%	51%	-2%
2	54%	46%	9%
3 or more	59%	41%	18%

Note: Significant values are highlighted in bold. Percentages are rounded to the nearest integer. Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability. Results shown are at the average country-language group level.

Source: Table B3.3

Table 3.3 similarly shows that socio-economic advantage – which is typically associated with higher scores in PISA, in general, but also found to be positively associated to performance in creative thinking (OECD, 2024[4]) – is associated with students’ capacity to consistently achieve high-scoring responses across different types of tasks. On average across country-language groups, around 4 in 10 creative all-rounders were amongst the top quartile of their country-language socio-economic index, while only 1 in 10 creative all-rounders were in the lowest quartile of their country-language socio-economic index.

Table 3.3. More advantaged students than disadvantaged students are creative all-rounders

Percentage of students in the top / bottom quartile of the country-language socio-economic index (on average across country-language groups), by number of high-scoring responses

High-scoring answers	Advantaged students	Disadvantaged students	Difference (Disadvantaged-advantaged)
0	16%	29%	12%
1	24%	21%	-3%
2	33%	15%	-18%
3 or more	41%	10%	-31%

Note: Significant values are highlighted in bold. Percentages are rounded to the nearest integer. Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability. Results shown are at the average country-language group level.

Source: Table B3.4

In addition to students’ background (gender and socio-economic status), students’ performance on reading and mathematics likely influence the probability that they are a creative all-rounder. Indeed, the PISA 2022 results showed that differences in the performance of advantaged and disadvantaged students on the PISA test were largely accounted for (albeit not completely) by students’ mathematics and reading scores (OECD, 2024[4]). Unsurprisingly, the mean mathematics and reading scores of students increases as the number of high-scoring responses increases, on average across task combinations and country-language groups (Table B3.5). The mean score difference between students with no high-scoring answers and students that are creative all-rounders is more than the mean standard deviation (OECD countries) for both mathematics and reading in PISA 2022. These differences were statistically significant for both subjects in all country-language groups.

Figure 3.3 summarises the associations between gender, socio-economic status and academic performance, and the number of high-scoring responses achieved by students. While these factors are all significantly associated with differences between being a creative all-rounder and achieving no high-scoring responses, how important are they in determining whether or not students score highly across tasks in the PISA test? Only around 20% of the variance in high-scoring responses is accounted for by these factors, on average across task combinations and country-language groups, suggesting that there are other important factors influencing students’ capacity to consistently score highly across creative tasks (Table B3.6).

Figure 3.3. Creative all-rounders are more likely to be girls, be from an advantaged socio-economic background, and score more highly in reading than non-all-rounders

Share of girls / advantaged students (left y-axis) and mean PISA reading score (right y-axis), on average (across country-language groups), by number of high-scoring responses

Note: Significant differences between girls and boys, and advantaged and disadvantaged students, within each high-scoring response group are shown by markers in a darker shade. Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). Advantaged students refer to students in the top quartile of their country-language ESCS index. “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability. Results shown are at the average country-language group level.

Source: Tables B3.3-B3.5

3.2.2. By personality traits and attitudes

Certain enduring personality traits, like openness and extraversion, have typically been associated with creative individuals (Puryear, Kettler and Rinn, 2017[6]). Other attitudes, like seeing oneself as imaginative or creative self-efficacy, have also been found to support creative outcomes in prior research (Abdullah, Omar and Panatik, 2016[7]). Some authors argue that these traits and attitudes can support creative work across domains (Van Broekhoven, Cropley and Seegers, 2020[8]; Miravete and Tricot, 2024[1]).

Figure 3.4 shows that creative all-rounders do generally hold more positive beliefs about their capacity to engage in creative tasks (creative self-efficacy), and they also see themselves as more curious, more imaginative, and more open to intellect and the arts than their peers – particularly those who did not come up with any or only one high-scoring response across tasks. Certain socio-emotional skills might also support students’ performance in creative thinking tasks, for example, by supporting students to continue working on open-ended problems that require multiple responses. Figure 3.5 shows that creative all-rounders also tended to self-report higher levels of assertiveness and perseverance, but interestingly, much lower stress resistance than all other groups of students.

Figure 3.4. Creative all-rounders report stronger attitudes and beliefs associated with creativity

Mean index of creative attitudes (on average across country-language groups), by number of high-scoring responses

Note: Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). Some scales from the PISA 2022 student questionnaire were suppressed from some country-language groups, and therefore they are not included in the average country-language group results; for more information, see the PISA 2022 Technical Report (OECD, 2024[3]). “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability.

Source: Table B3.7

Coming up with creative answers across several tasks in the PISA test requires that students remain engaged with the test and sustain a certain level of investment throughout; it may be that those who are less resistant to stress are more inclined to put in the effort to perform well, in order to reduce feelings of anxiety. For example, while some affective states might be considered as “deactivating states” that can decrease arousal and negatively impact creativity, stress – although triggered by negative emotions – might be considered an “activating state” for creativity that serves to enhance creative engagement by its association with persistence (Nijstad et al., 2010[9]; De Dreu and Nijstad, 2008[10]). The optimal level of arousal model also suggests that overstimulation from one’s environment might promote an increase in creative thinking skills aimed at creating simpler input stimuli (i.e. convergent thinking), whereas understimulation might lead to inverse effects and rather promote an increase in exploratory behaviours and cognitive processes aimed at novelty and increasing complexity in one’s environment (i.e. divergent thinking) (Gustafsson, 2023[11]). Several studies have investigated the relationship between mood, affect and creativity, with the varied findings across studies demonstrating the complex interactions between an individual’s environment, personality and cognitive preferences (Baas, De Dreu and Nijstad, 2008[12]).

Figure 3.5. Creative all-rounders report higher levels of assertiveness and perseverance, but lower levels of stress resistance

Mean index of socio-emotional skills (on average across country-language groups), by number of high-scoring responses

Note: Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). Some scales from the PISA 2022 student questionnaire were suppressed from some country-language groups, and therefore they are not included in the average country-language group results; for more information, see the PISA 2022 Technical Report (OECD, 2024[3]). “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability.

Source: Table B3.8

Although several attitudes and socio-emotional skills clearly demonstrate a positive association with the number of high-scoring responses that students achieve, are these factors specific to supporting students’ creative thinking performance or do they rather reflect a more general association with student performance in PISA? After accounting for students’ mathematics and reading scores on the PISA test, several of the beliefs, attitudes and socio-emotional skills discussed in this chapter remain small yet significant predictors of being a creative all-rounder. However, they each explain only a marginal proportion of the variation in students’ number of high-scoring responses across tasks before accounting for students’ mathematics and reading scores compared to the 17-18% of the variation explained after accounting for mathematics and reading (Table B3.9 and Table 3.10). Comparing the difference in the mean number of high-scoring responses between students in the top and bottom quarter of the self-reported indices, even after accounting for mathematics and reading scores, highlights the associations between all-round performance in creative thinking and certain attitudes and beliefs (Figure 3.6).

Figure 3.6. Students that report strong attitudes and socio-emotional skills associated with creativity are more likely to score highly across creative tasks

Difference in the mean number of high-scoring responses between students in the top / bottom quartile of select attitudinal and socio-emotional indices (on average across country-language groups), before and after accounting for mathematics and reading scores

Note: Bars show differences (top quartile – bottom quartile) before accounting for mathematics and reading, and markers show differences after accounting for mathematics and reading. Significant differences after accounting for mathematics are shown by filled markers. All differences before accounting for mathematics and reading and significant. Sample includes all students who took at least 3 tasks included in the PISA CT Rescoring project (N = 5530). Some scales from the PISA 2022 student questionnaire were suppressed from some country-language groups, and therefore they are not included in the average country-language results; for more information, see the PISA 2022 Technical Report (OECD, 2024[3]). “High scoring” responses are jittered standardised holistic scores that fall within the top quartile (75^th percentile) of each country-language-task group. The standardised holistic creativity scores were jittered, i.e. a random small digit (1e-6) was added to each score, so that equal quartiles in each country-language group could be produced. Results shown are the average of 20 replications of the analysis to account for sampling and jitter-induced variability.

Source: Table B3.9-B3.10

Key findings and implications

• Many students that took the PISA test demonstrated that they are capable of being creative, with around 6 in 10 students scoring highly in at least one task relative to their country-language group peers.

  • Successful creative thinking, on a task-by-task basis, does not appear limited to a particular student group.

• Few students in all country-language groups, and only around 8% of students on average, are creative “all-rounders”, i.e. students that consistently score highly in creative thinking tasks. Consistent creative performance across different types of task therefore appears much more exclusive and difficult to achieve than successful performance on individual tasks.

  • Students’ gender, socio-economic status, and mathematics and reading performance were all significantly associated with scoring highly across multiple tasks, together accounting for close to 20% of the total variation in high-scoring outcomes.

  • Students who believed in their own capacity to be creative, who saw themselves as imaginative, open and curious, and who reported higher perseverance and assertiveness, also tended to achieve more high-scoring responses than those who did not.

• Our results are consistent with theories that suggest creative thinking is supported by both domain-general (“context-free”) and domain-specific (“context-specific”) factors (Plucker and Beghetto, 2006[2]).

  • While “context-free” factors include socio-demographic background, academic performance, and attitudes and beliefs, much of the variation in consistent creative thinking performance remains unexplained by these factors.

  • It may be that other general factors support performance across domains, like general intelligence or conscientiousness, or that context-specific factors like domain knowledge and skills or opportunities to practice creative thinking are indeed more important for supporting consistent performance and account for the rest of the variation in creative all-round performance.

  • Teachers and school environments can play a key role in fostering both types of supporting factors, for example by providing positive experiences that reinforce student beliefs about creativity, by developing their knowledge and skills, and by enabling them to practice applying creative thinking in domain-relevant ways.

References

[7] Abdullah, I., R. Omar and S. Panatik (2016), “A literature review on personality, creativity and innovative behavior”, International Review of Management and Marketing, Vol. 6/1, pp. 177-182.

[5] Acar, S. and Y. Shen (2025), “Along the convergent–divergent continuum: The role of task structure in the PISA Creative Thinking assessment”, The Journal of Creative Behavior, Vol. 59/2, https://doi.org/10.1002/JOCB.70029.

[12] Baas, M., C. De Dreu and B. Nijstad (2008), “A meta-analysis of 25 years of research on mood and creativity: Hedonic tone, activation, or regulatory focus?”, Psychological Bulletin, Vol. 134/6, pp. 739-756, https://doi.org/10.1037/a0012815.

[10] De Dreu, C. and B. Nijstad (2008), “Conflict and creativity: Threat-rigidity or motivated focus?”, Journal of Personality and Social Psychology, Vol. 95, pp. 648-661, https://doi.org/10.1037/0022-3514.95.3.648.

[11] Gustafsson, E. (2023), “How can contextual variables influence creative thinking? Contributions from the optimal-level of arousal model”, Journal of Creative Behavior, Vol. 57/1, pp. 96-108, https://doi.org/10.1002/jocb.565.

[1] Miravete, S. and A. Tricot (2024), “Are some people generally more creative than others? A systematic review of fifty years’ research”, Educational Psychology Review 2024 36:4, Vol. 36/4, pp. 1-39, https://doi.org/10.1007/S10648-024-09926-6.

[9] Nijstad, B. et al. (2010), “The dual pathway to creativity model: Creative ideation as a function of flexibility and persistence”, European Review of Social Psychology, Vol. 21/1, pp. 34-77, https://doi.org/10.1080/10463281003765323.

[4] OECD (2024), PISA 2022 Results (Volume III): Creative Minds, Creative Schools, PISA, OECD Publishing, https://doi.org/10.1787/765ee8c2-en.

[3] OECD (2024), PISA 2022 Technical Report, PISA, OECD Publishing, https://doi.org/10.1787/01820d6d-en.

[6] Puryear, J., T. Kettler and A. Rinn (2017), “Relationships of personality to differential conceptions of creativity: A systematic review”, Psychology of Aesthetics, Creativity, and the Arts, Vol. 11/1, pp. 59-68, https://doi.org/10.1037/ACA0000079.

[2] Sternberg, R., E. Grigorenko and J. Singer (eds.) (2006), Why creativity is domain general, why it looks domain specific, and why the distinction does not matter., American Psychological Association, https://doi.org/10.1037/10692-009.

[8] Van Broekhoven, K., D. Cropley and P. Seegers (2020), “Differences in creativity across Art and STEM students: We are more alike than unalike”, https://doi.org/10.1016/j.tsc.2020.100707.