Creative minds at rest: Creative individuals are more associative and engaged with their idle thoughts

Abstract

Despite an established body of research characterizing how creative individuals explore their external world, relatively little is known about how such individuals navigate their inner mental life, especially in unstructured contexts such as periods of awake rest. Across two studies, the present manuscript tested the hypothesis that creative individuals are more engaged with their idle thoughts and more associative in the dynamic transitions between them. Study 1 captured the real-time conscious experiences of 81 adults as they voiced aloud the content of their mind moment-by-moment across a 10-minute unconstrained baseline period. Higher originality scores on a divergent thinking task were associated with less perceived boredom, more words spoken overall, more freely moving thoughts, and more loosely-associative (as opposed to sharp) transitions during the baseline rest period. In Study 2, across 2,612 participants, those who reported higher self-rated creativity also reported less perceived boredom during the COVID-19 pandemic, a time during which many people experienced unusually extended periods of unstructured free time. Overall, these results suggest a tendency for creative individuals to be more engaged and explorative with their thoughts when task demands are relaxed, raising implications for resting state functional MRI and societal trends to devalue idle time.

Introduction

An intriguing feature of human cognition is the stream of consciousness that unfolds when individuals are left to their own musings. Most people are familiar with the dynamic experience of thoughts that arise whenever they are “doing nothing” – i.e., when waiting for the bus, taking a shower, or simply sitting on their bed not doing anything. The characteristics and dynamics of such thoughts lead some to be attracted to (and others to avoid) situations where they are left alone with nothing but their thoughts. While the average person finds this experience only somewhat enjoyable and is more likely to prefer engaging in a number of alternative activities (Alahmadi et al., 2017; Buttrick et al., 2018; Westgate et al., 2017, 2021; Wilson et al., 2014), creative individuals are anecdotally believed to enjoy being left alone with their thoughts (Long & Averill, 2003). As one example, Descartes described the importance of sitting alone with his thoughts for his most creative thinking (Schultz & Schultz, 2011). Beyond anecdotal links, however, experimental evidence for the association between creative potential and the experience of being alone with one’s thoughts remains largely unexplored. This study thus provides one of the first empirical tests of how creativity relates to our thought stream when task demands are relaxed. We explore whether individuals’ creative potential relates to their levels of engagement with their thoughts during an unconstrained baseline state, and whether the dynamics of their stream of consciousness differs in terms of mental exploration and associative thinking.

Why might creative individuals value idle time? One possibility stems from a growing body of research linking creativity to the inquisitive thinking trait, curiosity (reviewed in Gross et al., 2020). This relationship has been shown for both self-reported creativity and to a lesser degree to creative problem solving assessed by external raters (for a meta-analysis, see Schutte & Malouff, 2020). If individuals who have an inherent desire for knowledge tend to exhibit higher creative potential, they may also have a unique relationship with their idle thoughts making them better able to use unstructured periods as a seed for creative thinking.

Related to curiosity, prior research has found that openness to experience, need for cognition, and mindfulness – constructs that have also been related to one or more forms of creativity (Da Costa et al., 2015; Lebuda et al., 2016; Watts et al., 2017) – are among the strongest predictors of enjoyment of being left alone with one’s thoughts (Buttrick et al., 2018; Wilson et al., 2014). Further support comes from studies suggesting that trait mindfulness and meditation practice – contemplative phenomena whereby individuals have a trait-like disposition towards or practice being at peace with their inner thoughts – can promote both insight and divergent thinking (for a meta-analysis, see Lebuda et al., 2016). The most direct evidence linking creativity to thought engagement during idle time comes from a study reporting that people with better divergent thinking performance were less anxious, less tense, and less tired following an hour of isolation spent in either a floating tank or a dimly-lit room, compared to individuals with lower divergent thinking scores (Forgays & Forgays, 1992).

Conversely, less creative individuals appear to be less at ease in situations where external stimulation is low and they have no particular task to perform. For instance, for the average young adult, doing nothing increases the likelihood of experiencing boredom, second only to studying (Chin et al., 2017). Along a similar vein, societal trends have seen a rise in technological devices in recent years used as a means to alleviate boredom (Kil et al., 2021). The difficulty that some people find being left alone with their thoughts came to extreme levels during the COVID-19 pandemic, when government lockdowns, unemployment, and virtual schooling disrupted the structured social contexts of numerous life domains. These sudden changes coincided with spikes in mental health issues (Hossain et al., 2020), substance abuse (Abdo et al., 2020; Dumas et al., 2020; Slavova et al., 2020), and boredom (Brodeur et al., 2021; Droit-Volet et al., 2020). One study in particular linked the combination of loneliness, boredom, and repetitive negative thinking to depression in the context of the COVID-19 pandemic (Hager et al., 2022). Yet, creativity appeared to mitigate some of these negative outcomes. For instance, increased engagement in creative activities during the pandemic (as compared to before its onset) was associated with decreased boredom levels (Mercier et al., 2021) and increases in well-being (Tang et al., 2021). Overall, these diverse results suggest creative individuals would be less likely to be bored in situations where they are left alone with their thoughts. If that is indeed the case, a follow-up question is whether their stream of consciousness differs in ways that makes these situations less boring and more propitious to having a creative mind.

Beyond content, a key component of the stream of consciousness is its dynamic qualities (Girn et al., 2020; James, 1890), including the nature of the dynamic transitions that occur from one thought to the next. According to the associative theory of creativity, forming associations between loosely connected concepts is considered a key component of the idea generation phase of the creative process (Kenett, 2019; Kenett & Faust, 2019; Mednick, 1962). These findings raise the question of whether more creative individuals experience idle thoughts that are more associatively linked. There is evidence of a link between divergent thinking and associative thinking in the context of lab-based tasks requiring conscious engagement. Creative individuals, as assessed by divergent thinking (Beaty et al., 2021; Gray et al., 2019; Upmanyu et al., 1996) or evaluation by peers (Gough, 1976), have been reported to produce more unusual associations during word association tasks, suggesting that the language networks of creative individuals may be more associative overall (e.g. show a greater number of associations between words, including associative links that are more atypical; Kenett & Faust, 2019). In addition, inducing a loose association mode of thinking by dampening inhibition has been shown to lead to improvement on divergent thinking and insightful problem solving tasks (Radel et al., 2015; Wieth & Zacks, 2011).

Arguably phenomenologically closer to idle thinking, mind wandering – a phenomenon that periods of awake rest tend to facilitate (Wilson et al., 2014) – has been considered by some researchers to involve the mental exploration of connections between loosely related constructs and ideas (Christoff et al., 2016; Sripada, 2018; Zadra & Stickgold, 2021), paralleling features of the creative process (Fox & Beaty, 2019). Thus, individuals who are best at engaging with and harnessing this mode of thinking could hypothetically show an increased creative potential. Accordingly, many correlational studies have shown a positive relationship between the propensity to daydream and divergent thinking (Preiss et al., 2016) and the creativity of participants’ spontaneous daydreams and storytelling as assessed by independent raters (Singer & Schonbar, 1961). Here we considered the degree of associations between thoughts, speculating that creative individuals may be more apt at using alone time to explore associations in a way that is analogous to engagement in creative thinking.

Considering that the stream of consciousness is an inherently personal experience, it is challenging to continuously assess its content. A commonly used research paradigm in which participants are left alone with their thoughts in the absence of a constrained task is the so-called “resting state” paradigm, widely used in the context of brain imaging to reveal distributed temporally-correlated networks of brain activity (Fox & Raichle, 2007). Individuals who are better able to think “outside the box”, i.e., divergent thinking, have been shown to exhibit more dense functional connections between default, frontoparietal and salience brain networks during resting state contexts (Beaty et al., 2014; Beaty, Kenett, et al., 2018) as well as divergent thinking tasks (e.g. Beaty, Kenett, et al., 2018). These findings point to an overall more associative pattern of brain connectivity characteristic of creative individuals, in line with associative theories of creativity. However, despite a large body of research into resting state connectivity (and an albeit smaller amount of research intersecting with aspects of creativity), relatively little is known about the stream of consciousness that unfolds during resting state contexts. Using retrospective self-report questionnaires following resting state periods, recent work found evidence that periods of rest tend to be filled with a diversity of internally-guided thoughts that can take many forms (e.g., Andrews-Hanna et al., 2010; Delamillieure et al., 2010; Diaz et al., 2013; Gorgolewski et al., 2014; Karapanagiotidis et al., 2020). More recently, a technique called the “think aloud paradigm” has been applied to resting state contexts, in which participants are trained to voice aloud the content of their minds as their thoughts enter into their conscious awareness (Li et al., 2021; Raffaelli et al., 2021; Sripada & Taxali, 2020). An advantage of this technique is that it yields a rich array of data that researchers can use to quantify the real-time content and dynamics of resting state cognition. For example, across just 10 minutes of rest, we were previously able to track the emergence and persistence of ruminative thoughts (repetitive self, negative, past-oriented thoughts) in adults with higher trait rumination (Raffaelli et al., 2021). Aside from its utility in illuminating individual differences in the thinking styles of individuals, the resting state is also a useful experimental context with which to explore how thoughts are associatively linked within one’s mind. Collectively, despite these hints in the literature that more creative individuals are more associative with their resting state thoughts, direct support for the hypothesis that individuals with higher divergent thinking have more associative idle thoughts is lacking.

To address these gaps, Study 1 examined the relationship between idle thoughts and divergent thinking ability – one of the most commonly used indicators of creative potential (Runco & Acar, 2012). Participants were trained to voice their thoughts aloud in real-time across a 10-minute unconstrained baseline, and separately responded to the divergent thinking prompt, “How would you make money with 100 rubber bands?”. The divergent thinking task was scored by an independent set of trained raters for originality and other indicators of divergent thinking. We conducted a series of linear regressions with originality scores as the dependent variable, testing 4 independent hypotheses:

1. Participants who are more engaged by their idle thoughts – as measured by less self-reported boredom and more content overall – will score higher on originality metrics of divergent thinking.

2. Across participants, more curious individuals – as assessed with a trait curiosity questionnaire – will score higher on originality metrics of divergent thinking.

3. Individuals who are more associative in their idle thoughts – as measured by a higher propensity to associatively link one thought to the next, a higher semantic similarity between thoughts, and a higher self-report rating of having experienced freely moving thoughts – will score higher on originality metrics of divergent thinking.

4. Individual differences in measures of overall engagement and movement in thought during the baseline Think Aloud task will positively associate with individual differences on parallel metrics during the divergent thinking task. Furthermore, individuals who self-report more overall engagement and freely moving thought during the divergent thinking task will score higher on originality metrics of divergent thinking.

Study 2 aimed to complement Study 1 by examining whether any observed associations between creativity and engagement with one’s thoughts (Hypothesis #1) extended to the COVID-19 pandemic, a societal health challenge that forced most people to spend much longer periods of time with their thoughts than usual. Across a larger and more heterogeneous group of individuals, Study 2 tested whether individuals who reported being less bored during the COVID-19 pandemic were also more likely to score higher on a brief self-report scale of trait creativity.

Study 1

Method

Participants.

The sample included 90 participants recruited from the University of Arizona SONA subject pool (compensated with research credits, n = 78), or through advertisement for paid compensation (n = 12). The mean age of the sample was 19.5 years (SD = 2.3) and 58.9% of the sample was female (see Supplementary Table S1 for more details). Written informed consent was obtained from all participants, and procedures were performed in accordance with regulations approved by the University of Arizona’s Institutional Review Board. The participant sample for this study also participated in Study 1 and 2 from Raffaelli et al. (2021), although the analyses presented here are non-overlapping, and additionally includes data from a creativity task not reported in Raffaelli et al. (2021). For a variety of reasons, some participants needed to be removed from subsequent analyses involving the divergent thinking task, the Think Aloud task, or both, and some of the remaining participants had missing data for select measures (see Supplementary Materials). Since all available data were included in relevant analyses, some analyses have a different sample size: n=81 participants for the creativity task, n=80 for the post divergent thinking task Think Aloud questionnaire, n=79 for both the post-unconstrained baseline Think Aloud questionnaire and curiosity and exploration inventory-II, and n=78 for the Think Aloud task.

Overview of Study Procedures.

Upon arrival, participants first completed the Curiosity and Exploration Inventory-II survey (see below for details about this and other tasks), along with other surveys not analyzed in this manuscript. Participants then completed an unconstrained baseline Think Aloud task during which they voiced their ongoing thoughts aloud (as described in Raffaelli et al., 2021), followed by retrospective self-report questions pertaining to their experience of and thoughts during the unconstrained baseline Think Aloud task. Next, participants completed a variation of the divergent thinking task, followed by a retrospective questionnaire asking the same questions as those following the unconstrained baseline Think Aloud task. Participants answered additional questions and completed additional tasks which will be the focus of future manuscripts.

Unconstrained baseline Think Aloud.

The method for this task has been described in details elsewhere (Raffaelli et al., 2021). Briefly, participants were left alone in a room with minimal external stimulation after disabling their phones and setting aside their belongings. Participants were trained to continually voice aloud whatever the content of their mind was in real time for 10 minutes. No prompt was given to participants as to what to think about. Most participants completed the tasks while undergoing physiological recording, data that may be the focus of future manuscripts, but has not been analyzed to date.

Each audio file was transcribed by experimenters and parcellated into individual thoughts by three separate human raters (2 males, 1 female, mean age = 25.7, SD = 5.9). A transition marks a shift in theme; the main concept or percept that is the current focal point of attention. In addition to demarcating switches in theme within participants’ transcripts to tease thoughts apart, raters also had to name each theme. This served the double purpose of providing additional information and reducing the likelihood that long thoughts spanning multiple themes would go unnoticed. In addition, raters categorized the manner in which a transition between thoughts occurred as either strong or associational. Strong transitions were characterized by an abrupt shift in topic where nothing seemed to link the previous thought from the current thought, conceptually similar to a Venn diagram where both circles do not overlap in terms of content (e.g., “Um, hopefully I’ll get the new job ‘cause I really want it. I’m tired of my old job. [strong transition] Um, I miss my dog. I haven’t seen her in a long time.”). On the contrary, associational transitions were transitions wherein an aspect of one thought appeared to trigger a shift to a new thought, as determined by a change in overarching theme. Associational transitions are best exemplified by the expression “This reminds me of” as a transition point and is conceptually akin to a shift from one circle in a Venn diagram to the other after an exploration of a common area (e.g., “Yeah I had Japanese subtitles. I preferred Japanese rather than English. Just sounds very whitewashed when it’s in English [associational transition] But speaking of Japanese, I’m glad I was able to choose what I wanted to do regarding my other language.”).

Subsequent to this coding, total word count was computed, as well as the ratio of associational to total transitions (number of associational transitions / (number of strong transitions + number of associational transitions)), which is inversely related to the ratio of strong transitions. Two other metrics of thought dynamics were computed and analyzed exploratively for descriptive and explorative purposes only: the total number of thoughts and mean number of words per thought. The relationship between those two variables was strong (r(72) = −.844, p < .001) as they represent a trade-off between quantity of themes explored and depth of exploration of each theme. We did not have strong hypothesis for these variables as we reasoned that their relationship with creativity may be driven by balances between a propensity for more exploration between themes (i.e., by more frequent associational transitions between thought, resulting in a higher number of thoughts) and more exploration within themes (i.e., higher mean number of words per thought) (see Girn et al., 2020, for a related discussion). The inter-rater reliability for those metrics were all good or excellent with the exception of the associational transitions which was moderate (see Supplementary Table S2).

A final variable of interest derived from the unconstrained baseline Think Aloud task was the degree of semantic similarity between thoughts in each transcript. As described in Raffaelli et al. (2021), the spaCy library in Python (https://spacy.io/) was used to compute the average similarity between a thought and each subsequent thought (i.e., average [similarity t1 - t2, similarity t1 - t3, … similarity t1 - tfinalthought]). It computed semantic similarity metrics using a large model in order to derive a “distance” between two thoughts using their mathematical likeness (e.g. “truck” is semantically similar to “car”), after removing non-meaningful “stop” words which are not relevant for semantic similarity (the, and, a, etc.). The large model, which contains a massive corpus of written text (blogs, news, comments, etc.) was chosen because it includes vectors (or mathematical word embeddings) along with vocabulary and syntax. The likeness of two different ‘thoughts’ is then determined based on previously established relationships derived from the word embeddings, and is recommended when a non-specific vocabulary is relevant, as in the case of our thoughts. Indeed, the idea of using semantic similarity (e.g, cosine) as a meaningful indicator of thought processes is becoming popular in creativity research (Beaty & Johnson, 2021). The semantic similarity metrics ranged from 0 to 1, with 0 representing extreme levels of variability in the content of thought and 1 representing extreme levels of similarity between the thoughts’ content.

Divergent thinking task.

Creativity was assessed with a divergent thinking task consisting of an open-ended question intended to mimic prompts from the Alternate Uses Tasks from the Torrance Test of Creativity (Guilford, 1967) in which participants are asked to think of alternate uses for a common object. The probe was “How would you make money with 100 rubber bands?” (found in Seelig & Stevens, 2009). Participants were given 5 minutes to generate answers to the questions. Participants’ transcripts were coded by a set of three raters. Following this process, each rater scored each new idea generated by participants for their originality on a scale from “1 - Not at all original” to “5 - Extremely original” for originality. Each idea was named so that any idea mentioned several times would only be counted once. In addition, the number of unique ideas was compiled in order to compute a metric of idea fluency. All of the coded metrics had good to excellent inter-rater reliability (see Supplementary Table S2). Although the main outcome of interest for this study was mean originality (sum of all originality ratings / number of unique ideas) given the manuscript’s primary interest in the quality of novel ideas, findings using fluency are also reported in the Supplemental Results and Supplementary Figure 1.

Retrospective self-report questions.

Following the unconstrained baseline and divergent thinking Think Aloud tasks, participants were asked retrospective questions about their phenomenological experiences during these tasks. All questions were answered using a sliding scale with the following anchors on both extreme ends of the scale: “Not at all - Extremely.” The default position of the sliding scale was at the midpoint and the final position of the cursor was automatically converted into a two-decimal number between 0.00 and 1.00. One question captured participants’ level of engagement with the task: “I felt bored,” and another captured the manner with which their thoughts unfolded over time: “To what extent did your thoughts flow freely from one to another?”. There were also two quality control questions about the nature of the task: “How similar were your thoughts to those you experience in your day-to-day life?” and “To what extent did you censor yourself during the task?”. Other questions assessing content characteristics (e.g. valence, self-focus, etc.) are not assessed in the present manuscript (see Raffaelli et al., 2021).

Trait curiosity.

Participants’ trait curiosity was assessed with the Curiosity and Exploration Inventory-II (Kashdan et al., 2009), a 10-item questionnaire divided in two subscales of 5 items each. The first subscale, stretch, captures one’s propensity to seek new knowledge and experiences (e.g. “I actively seek as much information as I can in new situations”). The second, embrace, captures ones’ ability to embrace uncertainty (e.g. “I am the kind of person who embraces unfamiliar people, events, and places.”). Both subscales had good reliability (ICC = .826, CI₉₅ = [.756; .880] for stretch and ICC = .829, CI₉₅ = [.760; .882] for embrace) according to Koo and Li’s (2016) guidelines.

Statistical analysis.

Threshold for significance was set at p<.05. Though described in the results with their original scale, all variables were Box-Cox transformed and subsequently z-scored for ease of statistical analysis and interpretation. All variables but the self-report questions about boredom (following both unconstrained baseline and divergent thinking task) and the experience of freely moving thoughts (divergent thinking task only) achieved satisfactory levels of normality following Box-Cox transformation. When assessing the relationship involving these non-normally distributed variables, Spearman correlations were run. For all analyses involving mean originality scores on the divergent thinking task, linear regressions with mean originality scores as the dependent variable were used. The remaining analyses were performed with Pearson correlations. Data points having high leverage as determine by a Cook distance value above .5 were discarded from analysis.

A handful of participants had data missing completely at random on the retrospective questionnaires that followed the unconstrained baseline and divergent thinking Think Aloud tasks. Multiple imputations were implemented with the MICE R package in order to impute the missing data. Fifty imputed datasets were created (predictive mean matching method, max iteration = 5) and the imputed data were inferred Rubin’s law (i.e., average across all imputations).

Results

Quality control analyses.

All descriptive statistics for the unconstrained baseline Think Aloud task, curiosity questionnaire, and mean originality scores are reported in Supplementary Table S3. During the baseline Think Aloud task, participants generated a mean of 28 thoughts (SD = 13), of a mean length of 63 words (SD = 49). Exploratory analyses revealed that there were no associations between mean originality scores and the total number of thoughts (r(71) = .030, p = .80) or mean number of words per thoughts (r(71) = .181, p = .13). Transcripts had a mean total word count of 1272 (SD = 366). The mean ratio of associational transitions was 31% (SD = 18%).

With regards to the experience of the task, the mean level of reported boredom was .40 (SD = .30) and the mean experience of freely moving thoughts was .63 (SD = .22), consistent with prior descriptive results of these constructs during resting state tasks (Alahmadi et al., 2017; Westgate et al., 2017, 2021; Wilson et al., 2014).

On the divergent thinking task, participants generated a mean of 5.12 ideas (SD = 2.65) across the 5-minute period, with a mean total word count of 612.06 (SD = 152.56). Participants’ mean originality scores were 2.42 (SD = .60). Participants’ mean level of reported boredom and experience of having free flowing thought during the divergent thinking task were .53 (SD = .28) and .67 (SD = .25), respectively.

As also described in Raffaelli et al. (2021), participants reported only somewhat censoring their spoken thoughts during the unconstrained baseline Think Aloud task (mean = .36, SD = .24), and experienced a moderately high degree of similarity between their spoken thoughts and thoughts in daily life (mean = .72, SD = .21, n = 81). Neither of these variables significantly associated with participants’ mean originality scores (level of censorship: beta = −.04, t(78) = −.350, p = .73, daily thought similarity: beta = −.039, t(80) = −.343, p = .73). In addition, the level of censorship (mean = .34, SD = .30) and daily thought similarity (mean = .42, SD = .32) experienced during the divergent thinking task was not related to mean originality scores either (level of censorship: Spearman’s ρ (78) = −.089, p = .43, daily thought similarity: Spearman’s ρ (78) = .−.104, p = .36).

Participants with higher originality scores displayed signs of higher levels of engagement with their idle thoughts.

In line with Hypothesis #1 outlined in the Introduction, individuals with higher mean originality scores were less likely to report being bored while sitting alone with their thoughts (Spearman’s ρ(77) = −.225, p = .046) (Figure 1A). Additionally, participants with higher mean originality scores generated more words across the 10 minute unconstrained baseline Think Aloud task, i.e., had a higher total word count (beta = .330, r(72) = .325, p = .005, r² = 10.57%) (Figure 1B). Analyses of baseline Think Aloud metrics using fluency instead of originality are reported in the Supplementary Results. Overall, findings using fluency tend to preserve a similar (albeit weaker) overall pattern as originality (see Supplementary Figure 1).

Figure 1.

Scatterplots representing relationships between mean originality scores on the divergent thinking task and variables related to engagement with the unconstrained baseline Think Aloud task (A & B) and associational thinking at rest (C & D). The shaded grey area represents the 95% confidence region for the regression fit.

Trait curiosity was positively associated with higher originality scores and less bored during the baseline Think Aloud.

The mean stretch-curiosity subscale score was 3.54 (SD = .87) and the mean embrace-curiosity score was 3.10 (SD = .95, see Supplementary Table S3). Higher mean originality scores were associated with higher scores on both the stretch (beta = .282, t(78) = 2.552, p = .013, r² = 7.78%) and embrace (beta = .317, t(77) = 2.993, p = .004, r² = 10.54%) subscales of the curiosity questionnaire. Consistent with hypothesis #2, both forms of curiosity were associated with lower levels of self-reported boredom during the unconstrained baseline Think Aloud task (Spearman’s ρ (75) = .298, p = .008 for stretch curiosity and Spearman’s ρ (75) = 290, p = .011 for embrace curiosity). However, neither forms of curiosity predicted a higher total word count (beta = .115, t(70) = .987, p = .33 for stretch and beta = −.013, t(70) = −.109, p = .92).

Participants with higher originality scores displayed signs of more associative thinking during the baseline Think Aloud task.

Participants with higher mean originality scores reported that their thoughts moved more freely from one to the next (beta = .273, r(78) = 2.466, p = .016, r² = 7.32%) (Figure 1C). They also had a higher ratio of transitions qualified as associational (beta = .234, t(72) = 2.079, p = .041, r² = 5.74%) (Figure 1D). Examining each type of transition individually, mean originality scores were significantly correlated with the number of associational (beta = .314, t(72) = 2.869, p = .005, r² = 10.39%), but not strong transitions (beta = −.023, t(72) = −.232, p = .82). Finally, contrary to predictions, participants with higher mean originality scores were not more likely to explore semantically-related content during the unconstrained baseline Think Aloud task, as measured by the semantic similarity metric (r(71) = .151, p = .20). A correlation matrix displaying the relationship between the multiple variables used in Study 1 is available in Supplementary Table S3 for informational and meta-analytic purposes.

Parallels between unconstrained baseline and divergent thinking.

Supporting Hypothesis #4, the variables available for both tasks were all predictive of mean originality scores on the divergent thinking task. More specifically, participants with higher originality scores displayed higher levels of task engagement during the divergent thinking task, as mean originality scores were positively associated with the total number of words generated (beta = .262, t(78) = 2.410, p = .018, r² = 6.85%), and a lower propensity to report being bored during the divergent thinking task (Spearman’s ρ(78) = −.343, p = .002). In addition, they were also more likely to report experiencing freely moving thoughts during the divergent thinking task (Spearman’s ρ(78) = .243, p = .030).

Furthermore, each of the variables in one task was positively related to its counterpart in the other task. There were strong associations between the total word count generated during the baseline Think Aloud task and the total word count generated during the divergent thinking task (r(72) = .710, p < .001). Similarly, individuals who reported being less bored and experiencing more freely moving thoughts during the unconstrained baseline Think Aloud task were also more likely to experience the same feelings during the divergent thinking task (Spearman’s ρ(76) = .339, p = .002, and Spearman’s ρ(76) = .259, p = .022, respectively).

Discussion

In line with the study hypotheses, variability in thoughts during the unconstrained baseline Think Aloud task, as captured both by self-report and external coders, was associated with higher originality on a divergent thinking task (for a schematic of the main results, see Figure 2). As originality on the divergent thinking task increased across individuals, overall engagement with idle thoughts did as well, as suggested by lower self-report of boredom and a tendency to generate more content across the 10 minutes of the think aloud task. In addition, in line with prior studies (e.g. Schutte & Malouff, 2020), individuals with higher originality scores were also more curious. Individuals with higher originality scores also showed signs of being more associative in their idle thinking, as evidenced by a stronger tendency to transition between thoughts in an associative manner and to self-report experiencing their thoughts as more freely flowing. Contrary to expectations, however, originality scores did not significantly relate to metrics of semantic similarity. Finally, we observed several parallels between cognition emerging during the baseline think aloud task and the divergent thinking task. The three variables common to both tasks (total number of words generated, self-reported boredom and experience of freely moving thought) were predictive of originality scores on the divergent thinking task for both tasks, and each positively correlated with its counterpart in the other task. In Study 2, to complement results from our first hypothesis, the possibility that individuals self-rated as more creative better handle longer periods of unstructured free time was tested by assessing whether self-rated trait creativity was associated with lower levels of boredom during the COVID-19 pandemic.

Figure 2.

Schematic of the main results from Study 1. Higher originality scores on the divergent thinking task were significantly related to metrics derived from analysis of the unconstrained baseline Think Aloud task suggestive of more engagement with idle thoughts (i.e., less bored and higher total word count), higher levels of exploratory and associative thinking (i.e., more freely moving thought, higher percentage of associative transitions), and higher levels of curiosity (CEII: Curiosity & Exploration Inventory II). Semantic similarity (in grey), another variable theoretically related to associative thinking, was not significantly related to originality (though see conclusion). Bold arrows represent significant relationships between mean originality scores on the divergent thinking task and other metrics.

Study 2

Method

Participants.

A total of 2,612 participants participated in Study 2 (mean age = 52 yrs, SD = 20.70, age range = 18 – 89 yrs, % female = 70.8). See Supplementary Table S1 for more details about study demographics. A little more than a third of participants (37.2%) were recruited through the University of Arizona’s Psychology undergraduate student pool and participated in this experiment in exchange for course credit. The remaining participants downloaded and used the Mind Window app (see below) on a voluntary basis. Participants were required to be at least 18 years of age and fluent enough in English to understand the study questions. Data collection began in August 2020 during the height of the COVID pandemic and continued (for the purpose of Study 2) until December 31, 2021. Informed consent and study procedures were approved by the University of Arizona’s Institutional Review Board.

The Mind Window app.

Mind Window is a free, cross-platform ecological momentary assessment smartphone app developed by our lab. Although Mind Window is part of a larger ongoing study, analyses for the purposes of Study 2 focused on questions asked only once after participants initially downloaded the app, and did not involve any of the ecological momentary assessment surveys. The first questionnaire of interest was the 2-item Kumar and Holman’s Global Measure of Creativity Capacity scale from the revised Creativity Styles Questionnaire (Kumar et al., 1997). The two items are ‘I consider myself to be a creative person’ and ‘I am engaged in creative type of work on a regular basis’ which were answered in the Mind Window app (see below) using a 5-item Likert-scale from “strongly agree” to “strongly disagree.” The two items had moderate reliability in our sample (ICC = .735, CI₉₅ = [.521; .833]). The second questionnaire of interest was designed by our lab to assess participants’ experience during the COVID pandemic. Of relevance to Study 2 was a question inquiring about participants’ levels of boredom experienced during COVID: “Since becoming aware of or impacted by the global spread of coronavirus / COVID-19, how BORED have you been?”. The question was answered on a sliding scale with 5 anchors (not at all, minimally, somewhat, quite a bit, extremely) which was converted into a two-decimal number from 0 to 1. Finally, as part of the demographics information, participants indicated their age and household size (“How many people do you currently live with?”), and participants’ socioeconomic status (SES) was assessed with the MacArthur Scale of Subjective Social Status – Adult Version (Adler et al., 2000). The decision to control for household size was based on the previously reported relationship between loneliness, boredom, and well-being (Hager et al., 2022; Tutzer et al., 2021).

Statistical analysis.

All variables were Box-Cox transformed and then scaled. Next, the direct relationship between levels of boredom experienced during COVID-19 and self-reported creativity on the one hand, as well as the control variables on the other hand, were assessed using Spearman correlations as some of the variables remained non-normally distributed (e.g., level of boredom experienced during COVID-19 and age) and this method provides better estimate of the true correlation in case of large normality distributed samples (De Winter et al., 2016). Finally, this relationship was reassessed using a regression analysis, with levels of boredom experienced during COVID-19 as the outcome variable, while age, SES, and household size were included as predictors in addition to creativity. The metric used to assess how much of the variance was accounted for in the multiple regression model was the adjusted r².

Results

Higher self-rated creativity associated with less boredom during the COVID pandemic.

The mean creativity score across the sample was 6.76 (SD = 1.99) on a range from 2 to 10. The mean level of boredom experienced during COVID-19 was .47 (SD = .32) on a scale from 0 to 1. The mean subjective SES score was 6.35 (SD = 1.63) on a scale from 1 to 10. The mean household number was 1.92 people (SD = 1.44).

In line with our predictions based on Study 1, the correlation between levels of creativity and levels of boredom experienced during COVID-19 was negative and significant, and indicated a small effect size (Spearman’s ρ(2610) = −.120, p < .001; see Figure 3). Levels of boredom experienced during COVID-19 were also negatively associated with being older (Spearman’s ρ(2610) = −.342, p < .001), belonging to a higher SES (Spearman’s ρ(2610) = −.159, p < .001), and positively associated with having a larger household size (Spearman’s ρ(2610) = .133, p < .001). Importantly, however, the relationship between creativity and boredom remained significant even when including age, SES, and household size in the model (b_creativity = −.112, p < .001). The overall model was significant (F(4, 2607) = 104.6, p < .001) and accounted for 13.70% (adjusted r²) of the variance in the levels of boredom experienced during COVID-19 (b_creativity = −.112, p < .001; b_age = −.337, p < . 001, b_SES = −.052, p = .005; b_{household size} = −.011, p = .57).

Figure 3.

Scatterplot representing the relationship between self-rated creativity (assessed by the Kumar and Holman’s Global Measure of Creativity Capacity scale) and self-rated boredom during COVID. The shaded grey area represents the 95% confidence region for the regression fit. Note that for visual purposes, the data points have been scattered using the jitter option of ggplot so as to minimize overlapping data points.

Discussion

Complementary to the findings from Study 1 that individuals with higher originality scores on the divergent thinking task were less bored when left alone with their thought, analyses from the Mind Window data also revealed a negative relationship between participants’ self-rated creativity and feelings of boredom during the COVID-19 pandemic, a time that disrupted the structured social contexts of educational, occupational and leisure life. Consistent with prior reports investigating boredom across a number of daily activities (Chin et al., 2017), levels of boredom also declined as age and subjective SES increased. Though complementary, these results should be interpreted cautiously in light of the differences in method and operationalization of boredom and creativity, and the small effect size which may be tied to our very large sample size and the many likely sources of variability that were not assessed in Study 2.

General Discussion

Across two studies employing different methodologies, independent participant samples, and varying environmental contexts, individuals displaying higher divergent thinking scores demonstrated more overall engagement with their idle thoughts and were more tolerant of unstructured idle time. In Study 1, individuals who generated more original answers on a divergent thinking task supplied lower ratings of boredom across the 10-minute unconstrained baseline Think Aloud task and generated more overall thought content (as measured by total word count). In an independent sample of >2,500 participants assessed during the height of the COVID-19 pandemic, Study 2 complemented these results by showing a negative relationship across participants between feelings of boredom during the pandemic and self-rated trait creativity. Consistent with the higher levels of engagement with idle thoughts, we also found that higher divergent thinking scores were associated with being more curious. As expected, individuals scoring higher on the divergent thinking task displayed signs of having more associational idle thoughts, reporting experiencing their thoughts as flowing more freely from one to another and exhibiting thoughts that were more likely to be associatively linked. Finally, we observed some interesting parallels between idle and divergent thinking, as individuals with higher divergent thinking scores were more engaged with and experienced more freely moving thoughts on both tasks, and scores on these metrics on one task was predictive of scores on the other task. Interestingly, a similar but weaker pattern of results was observed when assessing idea fluency metrics of creativity, consistent with the idea that fluency and originality are conceptually related but not the same (Runco & Acar, 2012). Collectively, these findings have important implications as they expand our understanding of the relationship that creative individuals have with their own minds and offer a source of variability in resting state-like cognition with possible relevance to resting state neuroimaging.

Individuals with higher divergent thinking scores are more engaged with their unconstrained baseline thoughts.

Previous studies exploring resting state contexts have reported that, on average, young adults find the experience of being left alone with their thoughts somewhat boring (Alahmadi et al., 2017; Buttrick et al., 2018; Hatano et al., 2022; Westgate et al., 2017, 2021; Wilson et al., 2014). Study 1 replicated these findings, but additionally found that increases in divergent thinking scores predicted less boredom across both short (Study 1) and extended (Study 2) periods of unstructured idle time. Although earlier work has linked individual variability in the experience of the resting state to constructs related to creativity, including openness to experience, need for cognition, positive daydreaming, and mindfulness (Agnoli et al., 2018; Colzato et al., 2012; Da Costa et al., 2015; Gold & Henderson, 1990; Lebuda et al., 2016; Watts et al., 2017), Study 1 extended prior work by using a Think Aloud approach to measure resting state-like thoughts in real-time, and by quantifying individual differences in creativity with an externally-scored divergent thinking Task.

Study 2 complemented Study 1’s findings with a longer period of unstructured time, a larger and more heterogeneous participant sample, and a different means of assessing creativity. The COVID-19 pandemic, with its accompanying lengthy periods of lockdown and restricted social interactions, provided people with an unprecedented amount of unstructured time, which coincided with an increase in boredom among adult samples (Brooks et al., 2020; Droit-Volet et al., 2020). In a much larger sample than in Study 1, individuals in Study 2 who were higher in self-reported trait creativity were less bored during the COVID-19 pandemic. Both results corroborate prior research linking engagement in creative endeavors during the COVID-19 pandemic to lower levels of boredom (Mercier et al., 2021) and higher well-being (Tang et al., 2021). Considering that boredom is a state in which one feels under-stimulated (Raffaelli et al., 2018), the notion that creative people were less bored during the unconstrained baseline Think Aloud task suggests they found their idle thoughts adequately stimulating.

Another significant associate of higher divergent thinking – the overall amount of thought content generated (i.e., total word count) – is also consistent with the notion that creative participants are more engaged with their resting state thoughts. Although neither the Wilson et al. (2014) and Buttrick et al. (2018) studies used a Think Aloud approach, participants’ retrospective written descriptions of their thoughts at rest revealed that the number of words participants wrote was a significant predictor of higher enjoyment of the thinking period. Though these prior reports could have been confounded by factors such as retrospective memory abilities or other factors influencing how much individuals recall their resting state thoughts, the present results suggest that the association persists when controlling for these potential confounds by having participants voice aloud their thoughts in real time. Another possible confound of these prior and present results is that they are dependent on higher verbal fluency in creative individuals. This would, however, be hard to disentangle as verbal fluency is associated with higher levels of creativity (Lee & Therriault, 2013) and may be a direct by-product of facets related to creativity such as intelligence, knowledge, and more associative thinking.

Individuals with higher divergent thinking scores may be more prone to associative thinking.

Individuals with higher divergent thinking scores displayed a greater number of thoughts triggered by an associative mental link, as well as a higher ratio of associative to total transitions. Transitioning associatively suggests that a participant has drawn a connection between the theme of a given thought and the theme of its successor. Indeed, individuals with higher divergent thinking scores have been shown to more readily generate associations between remote concepts, and are more likely to generate more loosely associative words on association tasks (Beaty et al., 2021; Gray et al., 2019; Upmanyu et al., 1996). The ability to notice connections between concepts may facilitate novel idea generation and problem solving in creativity, consistent with findings that creative individuals have high connectivity and short path between concepts in their semantic network (Kenett et al., 2014; Kenett & Faust, 2019).

Interestingly, individuals with higher divergent thinking scores also reported their thoughts to more freely flow from one to the next, and across participants, the ratio of associationally-linked thoughts was positively related to feelings of thought flow. Thoughts perceived as more freely-moving are theoretically less constrained and more spontaneous in nature (Christoff, Irving, Fox, & Spreng, 2016; Mills et al., 2018; but see Smith et al., 2022) and thus propitious to the exploration of loose associations. Lowering inhibition can improve divergent thinking (Radel et al., 2015; Wieth & Zacks, 2011), arguably because it increases the chances that unlikely connections between concepts are considered and consolidated. Having a natural inclination to explore one’s semantic network with low constraints, i.e., with less inhibition, would favor the development of semantic network promoting highly divergent thinking and creativity.

Unexpectedly, the relationship between semantic similarity and higher divergent thinking scores was not significant. Upon closer qualitative inspection, individuals with extreme levels of semantic similarity (2SD above the mean (mean = .69 and SD = .10), >.89 on a scale ranging from 0 to 1) show a sharp decline in divergent thinking scores (see Supplemental Figure S2). As a follow up exploratory analysis, we thus re-examined the relationship between semantic similarity and higher divergent thinking scores, after excluding these three extreme semantic similarity values, and observed a significant positive relationship (b = 3.022, t(69) = −2.330, p = .023, r² = 7.34%). A highly speculative possibility is that extreme levels of semantic similarity may reflect negative traits that are not conducive to divergent thinking such as obsessiveness or ruminative tendencies. Future work could assess whether the relationship between semantic similarity and divergent thinking abilities is impacted by such traits.

Differences in cognitive style of creative individuals may be observed across thinking contexts: From creative cognition to resting state cognition.

Interestingly, individuals with higher divergent thinking scores experienced less boredom, more subjective flow of thought, and greater verbal fluency of thought content and ideas both at rest and during the divergent thinking task. While there is a widely established link between creativity and both curiosity and openness to new external experiences (Da Costa et al., 2015; Schutte & Malouff, 2020), Study 1 also extends this exploratory behavior and increased motivation of creative individuals to their internal experiences in the absence of a task and contributes to the small but growing literature providing evidence of a link between curiosity and divergent thinking as assessed by a task (Koutstaal et al., 2022; Schutte & Malouff, 2020). Although speculative, this propensity to explore both the external world and the inner mind with curiosity and may have benefited creative individuals during the COVID-19 pandemic when many found themselves left alone with their thoughts for more extended periods of time, consistent with correlational findings from Study 2. While our results were restricted to divergent thinking, future work should examine relationships between metrics on the Think Aloud task and other aspects of creative thinking, particularly convergent thinking measures.

The observed differences in thinking across contexts between individuals displaying high and low levels of divergent thinking abilities have implications for resting state neuroimaging studies examining patterns of brain activity and/or connectivity in the absence of a task at hand. Resting state neuroimaging studies rarely assess spontaneous cognition emerging during rest, yet the higher engagement and more exploratory thinking patterns observed in individuals with higher divergent thinking scores may manifest as different patterns of functional network connectivity. Of particular relevance to resting state cognition is the default mode network (DMN), which has been linked to imaginative processes (Raffaelli et al., 2020), including those important for creative endeavors: the simulation of new or alternative scenarios (Van Hoeck et al., 2015), the recombination of existing knowledge (Abraham, 2016; Ellamil et al., 2012), associative thinking (Bar et al., 2007; Beaty et al., 2021; Kenett & Faust, 2019; Marron et al., 2018), and insight (Kounios & Beeman, 2014). Higher openness to experience has been associated with individual differences in higher global efficiency of the DMN at rest (Beaty et al., 2016) – a measure considered to indicate more efficient information processing (Achard & Bullmore, 2007). Beyond the DMN, participants with higher trait openness display greater brain network involvement in a metastate exhibiting positive coupling between regions within the DMN, the executive network, and the salience network (Beaty, Chen, et al., 2018). This pattern of increased integration between brain networks has also been linked to higher levels of divergent thinking (Beaty et al., 2015; Beaty, Kenett, et al., 2018). In consideration of these collective findings, we recommend that future studies employing resting state neuroimaging consider assessing the cognitive correlates of resting state cognition, as well as individual differences in divergent thinking abilities. A Think Aloud approach may be beneficial towards these endeavors (Li et al., 2022).

Limitations and Future Directions

Some of the limitations of this work should be noted, as well as possible avenues for future work. First, although we sought to assess the stream of consciousness unfolding in the absence of an experimental task (i.e., during resting state contexts), the instructions to “think aloud” created a task context that may have increased constraints on cognition. Promisingly, participants reported a moderately high degree of similarity in their spoken thoughts to those experienced in daily life, but an important avenue for future research will be to develop more covert markers of the stream of consciousness.

Second, the unconstrained baseline Think Aloud task was completed before the divergent thinking task, raising the possibility that specific types of cognitive operations at rest, or mood states (e.g., being annoyed or tired by the Think Aloud task), may have causally impacted performance on the divergent thinking task. Conversely, it is also possible that having the think aloud task before the divergent thinking task facilitated performance on the latter. If idle periods tend to promote loose associative thinking, there may have been a carryover effect. Considering these tasks were part of a larger dataset that included multiple goals, it was important for the baseline Think Aloud task to be a true baseline and thus to always come first. Future studies should counterbalance baseline and divergent thinking tasks, or assess more stable metrics of creativity by averaging task performance across multiple days to tease apart the possible order effects.

Third, coding other people’s thoughts – as our experimenter coders have done – is still a subjective process. This is especially relevant for the classification of transition type, as what may seem disconnected to an outside observer (and thus labelled a strong transition) may actually be associatively connected in the mind of the thinker. In this sense, the proportion of associational transitions estimated by the current method may be an underestimate. Future work comparing trained raters’ partitioning of the transcripts to that of research participants’ own partitioning of their thoughts would be necessary to further support the validity of this methodology. Along similar lines, we acknowledge that our scoring procedure needs improvement with regards to identifying associational transitions as the inter-rater reliability is only moderate. Future work could also consider adopting a more precise coding procedure, such as a continuous rating scale from “more gradual transition” to “less gradual transition” or further classifying associative transitions into distinct subtypes.

Collectively, despite these limitations, the present study sheds light on our understanding of how individuals high in divergent thinking abilities think outside of creativity tasks, and opens interesting new avenues of future exploration with applications to neuroimaging and clinical psychology.