Switching, fast and slow: Deciphering the dynamics of memory search, its brain connectivity patterns, and its role in creativity

doi:10.21203/rs.3.rs-3826172/v1

Download PDF

Article

Switching, fast and slow: Deciphering the dynamics of memory search, its brain connectivity patterns, and its role in creativity

https://doi.org/10.21203/rs.3.rs-3826172/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Creative ideas emerge from searching, reorganizing, and combining ideas or concepts within memory. This involves an interplay between associative and controlled processes. How these processes occur during memory search varies between individuals and how they relate to creative abilities remain unclear. Here, we explored the neurocognitive correlates of semantic memory search by integrating concepts and methods from two distinct approaches: the clustering-switching characterization of responses typically explored in fluency tasks, and the principles of optimal foraging as proposed by the marginal value theorem. We used an associative fluency task involving polysemous words that enabled us to identify clusters and switches among responses with respect to the different meaning of the cue words. We additionally captured the reaction times of the retrieved words during the task, and explored individual patterns of memory search at the cognitive and brain level. Our results indicate that search in semantic memory follows a pattern consistent with optimal foraging. Furthermore, when measuring the time intervals between consecutive responses, we observed that switches during memory search occurred on average as predicted by the marginal value theorem. However, individual patterns of fast or slow clustering and switching related to creativity. Participants with more frequent slow-clustering during retrieval exhibited higher divergent thinking ability, whereas participants with more frequent fast-switching were better able to combine remote associates. Finally, patterns of slow clustering and fast switching were predicted by brain functional connectivity and mediated the brain connectivity-creativity relationship. Overall, we developed new measures of semantic search, identified neurocognitive correlates of semantic search patterns, and related them to creative abilities. Our findings uniquely highlight the significance of the type of search (clustering vs. switching), as well as its temporal modulation (slow vs. fast), in relation to individual differences in creativity.

Biological sciences/Neuroscience/Cognitive neuroscience

Biological sciences/Psychology/Human behaviour

Creativity

semantic foraging

fluency

executive control

semantic memory

functional connectivity

Generating creative ideas involves processes and strategies to search, reorganize, and combine ideas or concepts in memory in a novel and useful way (Benedek et al., 2023; Kenett, 2023). Two main categories of cognitive processes are involved when people come up with creative ideas: associative and controlled processes (Benedek et al., 2023; Benedek & Jauk, 2018; Volle, 2018). Associative processes have been explored since the 60s, inspired by Mednick’s associative theory of creativity (Beaty & Kenett, 2023; Mednick, 1962). This theory hypothesizes that creativity depends on the organization of the concepts in memory (He et al., 2021), allowing more creative individuals to easily connect unrelated associative elements in memory compared to less creative individuals. It has been supported by several findings showing that more creative individuals have better associative abilities, more uncommon word associations, and a more flexible organization of semantic associations in memory (Beaty & Kenett, 2023; Beaty et al., 2014; Bendetowicz et al., 2018; Benedek, Könen, et al., 2012; Kenett & Faust, 2019; Marron et al., 2018; Ovando-Tellez et al., 2023).

Besides associative processes, which depend on memory structure, producing creative ideas also involves controlled processes that allow efficient retrieval and navigation within memory (Beaty et al., 2014; Benedek et al., 2023; Benedek & Jauk, 2018; Volle, 2018) and the inhibition of common and unoriginal ideas (Beaty et al., 2017; Camarda et al., 2018; Lloyd-Cox et al., 2021). The role of executive functions involving controlled processes has been suggested by the relationship between executive function abilities and creative abilities (Beaty et al., 2017; Benedek & Fink, 2019; Chrysikou, 2019; Edl et al., 2014; Jauk et al., 2013; Lee & Therriault, 2013). For instance, better performance in divergent thinking creativity tasks has been associated with inhibition (Benedek, Franz, et al., 2012; Benedek et al., 2014; Camarda et al., 2018; Krumm et al., 2018), task shifting (Krumm et al., 2018; Pan & Yu, 2016), and broad retrieval abilities (Benedek, Franz, et al., 2012; Benedek et al., 2017; Forthmann et al., 2019; Miroshnik et al., 2023; Silvia et al., 2013). The latter involves retrieving information from long-term memory fluently and flexibly (Schneider & McGrew, 2018), which is fundamental for creative thinking (Benedek, Franz, et al., 2012; Forthmann et al., 2019). These findings suggest a critical role of both associative and controlled processes in creative production, that are involved when searching for ideas in memory. However, how associative and controlled processes interact dynamically during searching for creative ideas in memory is still unclear. In particular, the patterns of memory search processes that allow the generation of creative ideas remains unknown.

The interplay between associative and controlled processes during memory search has been explored using classical fluency tasks. In fluency tasks, participants are asked to retrieve as many members of a category as possible (e.g., animals) in a limited time. Previous studies (Troyer et al., 1997; Troyer et al., 1998) have categorized the responses in these tasks as clustering and switching and proposed that they rely on associative and controlled processes, respectively. Clustering was characterized by successive responses that belong to the same subcategory (e.g., pets, wild animals, or sea animals) and switching as responses that belong to a different category than the previous one. Troyer’s research suggested that clustering relies on memory representations supported by temporal regions, while switching involves controlled processes supported by frontal brain regions (Troyer et al., 1997).

Recently, the framework of clustering-switching in fluency tasks has been reframed in the context of an exploitation-exploration trade-off (Hills et al., 2012; Hills et al., 2015; Todd & Hills, 2020). Hills and collaborators showed that memory search parallels the optimal foraging of animals in space when searching for food in a trade-off between local (exploitation) and global (exploration) search. Using classical fluency tasks, the authors demonstrated that switching to a different cluster (i.e., a subcategory in the fluency task) occurs when the retrieval rate within the current cluster falls below a threshold, as predicted by the optimal foraging theory framework described by the marginal value theorem (MVT, (Charnov, 1976). These switches took longer (longer inter-response times – IRT) compared to staying within the same cluster. Therefore, based on the semantic foraging approach, participants leave a cluster when the current retrieval rate equals the long-term average time of retrieval over the whole task. In addition, the authors showed that the semantic similarity between two successive responses (inter-response semantic similarity – IRS) is larger during clustering (within-cluster transitions) than between-cluster transitions (switching). A recent fMRI study (Lundin et al., 2023) provided evidence supporting distinct neural patterns for clustering and switching. The authors demonstrated increased activation in the hippocampus and posterior cerebellum during switching compared to clustering. Moreover, these regions increased their activity as participants got closer to a switch in a classical fluency task.

In a previous study, we extended the MVT approach to semantic memory by developing a novel free association task that asked participants to generate associates in response to polysemous word cues (Polysemous Fluency task; PolyFT; Ovando-Tellez, Benedek, et al., 2022). The advantage of using polysemous words as cues rather than classical categories (e.g., animals) in semantic fluency tasks is that the different meanings of polysemous words allow us to easily identify clusters in the individual’s responses. Using PolyFT, we examined the neurocognitive patterns of clustering and switching behaviors during semantic retrieval. We found that clustering was linked to higher divergent thinking abilities (measured with the alternative uses task) and involved higher brain functional connectivity between executive control networks (ECN) and attentional networks. This component was proposed to capture attentional focus and persistence when searching in memory. Conversely, switching correlated to the ability to combine remote associates (measured with the remote associate task), semantic memory structure, and executive abilities, as well as interactions between the default mode network (DMN) and ECN. This component was proposed to capture the interplay between associative and controlled processes allowing more flexible searching in memory, supported by DMN and ECN, respectively (Bendetowicz et al., 2018; Ovando-Tellez et al., 2019). These findings suggest that measures of clustering and switching during semantic retrieval relate to individual differences in creativity and reflect distinct neurocognitive processes.

Overall, the two components related to clustering and switching identified in our previous study (Ovando-Tellez, Benedek, et al., 2022) partially converge with the flexibility and persistence pathways of the dual pathway to creativity model (Nijstad et al., 2010). According to this theory, creative performance can be the result of both a persistence pathway related to more executive control and a flexibility pathway related to more associative search processes. However, in contrast to assumptions of the dual pathway to creativity model, the switching/flexibility component described in Ovando-Tellez, Benedek et al. (2022) correlated with controlled processing (at the cognitive and neural level). Likewise, the neurocognitive patterns of these two components were only partially aligned with the processes of clustering/associative and switching/controlled during the semantic search described in the work of Troyer et al. (1997). Specifically, our clustering component seemed to also involve control processes as suggested by the link between brain connectivity within control and attention-related networks and participants' clustering performance. Overall, these previous findings suggest that our components reflect different cognitive processes involved in search within or between clusters during the exploitation–exploration trade-off (Hills et al., 2012; Hills et al., 2015). However, the partial inconsistency between these results (Ovando-Tellez, Benedek, et al., 2022) and previous studies (Nijstad et al., 2010; Troyer et al., 1997; Troyer et al., 1998) prompt further exploration of the processes that occur during memory search related to creativity.

To further understand the memory search behavior that correlates with creativity and reconcile it with previous results from different approaches and theories (Hills et al., 2012; Nijstad et al., 2010; Troyer et al., 1997), in the current study, we re-analyzed data collected in our previous work (Ovando-Tellez, Benedek, et al., 2022) in light of the MVT (Charnov, 1976; Hills et al., 2012; Hills et al., 2015). We aimed to better characterize the clustering and switching processes that occur during semantic search and to identify individual patterns of semantic retrieval while participants perform the PolyFT task.

Based on previous findings, we hypothesized that the principles of the MVT could be used to characterize switching during search and retrieval in memory during the PolyFT task (see also Ackerman, 2014). We expected semantic retrieval patterns during the PolyFT task to be consistent with the MVT for semantic memory (Hills et al., 2012). To explore the clustering and switching trade–off occurring during semantic retrieval in light of the MVT we first explored the responses successively produced by the participants during the PolyFT in terms IRT and also IRS (Hills et al., 2012; Hills et al., 2015; Todd & Hills, 2020). Consistent with optimal foraging in semantic memory, we expected that 1) it takes more time to retrieve word associations when moving between clusters (switching) than within clusters as defined by the different meanings of the polysemous cue words; 2) the words within a cluster are semantically more similar to each other than words between clusters; 3) participants leave a cluster when they reach their average retrieval rate (i.e., long-term IRT during the whole task); 4) participants who switch when reaching their average rate of retrieval produce more responses during the task than participants who switch before or after reaching their value.

Furthermore, we expected that time and similarity covary during clustering and switching. Specifically, we hypothesized that responses with longer IRT would generate more semantically dissimilar words during both clustering and switching. During memory search through the semantic space, then more distant items should take longer to retrieve.

Lastly, we also identified clustering and switching responses that occurred either faster or slower than the average retrieval rate, as predicted by the MVT. We categorized these types of responses as Fast-Clustering, Slow-Switching (consistent with the MVT prediction), and Fast-Switching, Slow-Clustering (unexpected based on MVT). We then tested the applicability of the marginal value theorem separately for slow and fast switches.

We next quantified these types of responses for each individual and examined how they relate to executive function and creative abilities. We expected that Slow-Clustering and Slow-Switching responses should be related to controlled processes, whereas Fast-Clustering and Fast-Switching responses would correspond to more spontaneous responses. More precisely, consistent with the work of Troyer et al. (1997), we hypothesized that Fast-Clustering responses reflect more automatic/associative processes showing the role of memory representation in clustering, while Fast-Switching responses relate to the flexibility pathway of creativity described in Nijstad et al. (2010). Regarding slower responses, we expected that Slow-Switching responses reflect controlled processes and relate to executive abilities, as predicted by Troyer et al. (1997; 1998), and Slow-Clustering reflect attentional control efforts to stay focused in a cluster and maximally exploit this cluster, which is consistent with Ovando-Tellez, Benedek et al. (2022) and with the persistence pathway of creativity (Nijstad et al., 2010).

Regarding creativity, based on the dual pathways for creativity (Nijstad et al., 2010), we hypothesized that Fast-Switching (flexibility pathway) and Slow-Clustering (persistence pathway) responses would relate to creative abilities. However, drawing from previous findings (Mastria et al., 2021; Ovando-Tellez, Benedek, et al., 2022) and on the established role of both associative and control processes in creativity (Beaty et al., 2014; Benedek & Jauk, 2018; Volle, 2018), one could also argue for the relevance of Slow-Switching as well as Fast-Clustering in creative performance, that should correspond respectively to the executive control and semantic memory representation involved in fluency tasks (Troyer et al., 1997).

Finally, we used a connectome-based predictive modeling (CPM) approach to predict Fast-Clustering, Fast-Switching, Slow-Clustering, and Slow-Switching from individual brain functional connectivity patterns. Similarly, to Ovando-Tellez, Benedek et al. (2022) and following the dual pathways for creativity, we expected to reveal distinct functional connectivity patterns predictive of Fast-Switching and Slow-Clustering and that these patterns would relate to creative abilities.

Participants

We reanalyzed the sample of 86 participants (43 women, mean age = 25.5 years, SD = 3.48 years) collected by Ovando-Tellez, Benedek et al., (2022) who performed the PolyFT, tests of executive control functions, a battery of creativity tasks, and underwent a brain MRI. All participants were French native speakers, right-handed, with normal or corrected-to-normal vision and no neurological disorder, cognitive disability, or medication affecting the central nervous system. Participants gave informed written consent and received monetary compensation for their participation. The French ethics committee “CPP Sud Mediterranee IV” approved the study.

General Procedure

Participants first completed an MRI scan session during which they performed a task of relatedness judgments (RJT; see Supplementary Methods in Supplementary Material), followed by a rest session. Then, outside the scanner, participants completed the PolyFT and a set of creativity and executive function tasks.

The Polysemous word-fluency task (PolyFT)

In the PolyFT, participants were presented with a cue word and were instructed to freely generate all the single-word associations they could think of in relation to the given cue word for one minute (Ovando-Tellez, Benedek, et al., 2022). We presented three cue words in total. The cue words were three ambiguous French polysemous words: SOMME, GLACE, and RAYON. The different meanings of the cue words are presented in English in Table 1. All the cue words have high lexical frequency (> 20 occurrences per million in a large corpus according to the French lexicon project, http://www.lexique.org/) and at least five different meanings according to the French dictionary and the French linguistic resource for research (Centre National de Ressources Textuelles et Lexicales; https://www.cnrtl.fr/). The cue words were presented visually on a paper sheet and read aloud by the experimenter. Participants gave their responses aloud, which were audio recorded and written down on a computer by the experimenter. After manually cleaning the responses (i.e., typing errors), we categorized each produced response as belonging to one of the meanings of each cue word. This step allowed us to identify clustering responses, as the responses in the same category as the one preceding it, and switching responses that belong to a different category than the one preceding it. We quantified the total number of responses by summing the unique responses given by the participants for the three cue words (PolyFT_fluency).

Table 1

**Different meanings of the French ambiguous words used in the PolyFT**. The different meanings of the three cue words used in the PolyFT are presented separately in each column. Each word has at least five different meanings. In parenthesis is presented the total number of meanings for each cue word. A few words could not be categorized and were labeled as undetermined (< 2% overall responses across participants).
Somme (12)	Glace (5)	Rayon (12)
amount of money	ice	radius
French region	ice cream	ray, beam
nap	mirror, window, looking-glass	radiation
to summon	emotional coldness	shelf
we are (verb to be)	icing or frosting	wheel spoke
work		honeycomb
summary (In sum)		domain
A mountain in France		range
lexical or phonetic proximity		diverge, spread
a river in France		joy
pack animal		light
sum (maths), amount, quantity		Department, row in a shop

Searching in memory during PolyFT in the context of optimal foraging theory

The inter-response time between retrieved words

In addition to the clustering and switching categorization, we analyzed the response times of the participants during their performance in the PolyFT. Using the software Audacity 3.1 (https://www.audacityteam.org), we manually marked the initial and final time of each response. We then calculated the IRTs for all consecutive generated by the participants as the difference between the end of a given response and the beginning of the next one. We also calculated the long-term IRT as the mean IRT for each participant and cue word in the task.

Semantic similarity between retrieved responses

We explored the semantic similarity between all consecutive items (IRS) using pre-trained word2vec models for French. Word2vec is a word embedding algorithm based on neural networks. It learns vector representations of the words in a text so that words that share similar contexts are represented by close numerical vectors (Mikolov et al., 2013). We downloaded a pre-trained word2vec model for French from https://fauconnier.github.io/. This model was built from a 1.6 billion-word corpus constructed from websites with .fr domains, as described in Baroni et al. (2009). Word2vec assigns values between − 1 (low similarity) and 1 (high similarity), indicating the semantic similarity between two words. For each participant and each cue word, we also calculated the long-term similarity as the mean word2vec similarity in the whole task.

Exploration of optimal foraging in semantic memory in terms of inter-response time and similarity within and between clusters

To explore the processes occurring during semantic retrieval in light of the semantic foraging policies as in (Hills et al., 2012), we first examined whether the retrieval time (IRT) and the semantic similarity (IRS) differs between clustering and switching items. For each participant and cue word, we averaged the IRT and IRS, respectively, for the items classified as clustering (i.e., items belonging to the same cluster as the previous item; clustering-IRT and clustering-IRS) and switching (the first item of a different cluster; switching-IRT and switching-IRS). The switching-IRT reflects the time it takes to move from one cluster to another, while clustering-IRT reflects the retrieval time within the same cluster. The switching-IRS reflects the semantic similarity when moving from one cluster to another, while clustering-IRS reflects the semantic similarity when retrieving items within the same cluster. For each response and each cue word, we calculated the ratio between the IRT of the actual response and the individual’s long-term IRT (IRTr). Hence, a value of 1 indicates that the IRT of a given response corresponded to the individual's average retrieval rate, values above and below 1 indicate longer or shorter than average IRTs, respectively. Similarly, we calculated the ratio between the actual response’s similarity (IRS) and the mean similarity in the whole task (IRSr). Thus, values above or below 1 indicate higher or lower than average similarity, respectively.

We then examined whether participants leave a cluster when they reach their average retrieval rate (Hills et al., 2012; Hills et al., 2015) by exploring the IRTr and IRSr values in relation to their position relative to a switching response. We identified the responses preceding and following the switches as follows: -2 and − 1 are the second last and the last responses (pre-switch response), respectively, of the same cluster before switching, 1 is the first response in a new cluster (after a switch), 2 and 3 are the second and third responses, respectively, in the cluster after switching.

Finally, we explored if participants who switch when reaching their average retrieval rate generate more responses during the task than participants who switch before or after reaching their average retrieval value. We calculated the absolute difference between the IRTs before switching (response preceding switching) and the average retrieval rate. We regressed this value on the number of total responses produced during the task.

The computations of clustering-IRT, switching-IRT, clustering-IRS, switching-IRS, IRTr, and IRSr were conducted first for each response and then averaged across the three cue words and participants. All analyses were performed with Matlab (Matlab R2021b, The MathWorks, Inc.,USA) and plots were performed using Matlab and R studio Version 2022.07.1.

Identifying individual patterns of clustering and switching

We characterized PolyFT responses generated by participants in relation to their IRTr. Responses with IRTr values lower than one were responses with IRTs below the individual’s average retrieval rate and were classified as “fast.” Conversely, responses with IRTr values higher than one were responses with IRTs exceeding the individual’s average retrieval rate and were classified as “slow.” Combined with the clustering and switching coding, IRTs allowed us to categorize responses into four types: fast clustering (Fast-Clustering) and fast switching (Fast-Switching), slow clustering (Slow-Clustering), and slow switching (Slow-Switching). We then quantified the number of responses generated by each participant for each of these four response types, as well as their mean IRT and IRS. Finally, we computed the average semantic distance of each type from the cue word using word2vec.

Creativity tasks

Alternative Uses Task (AUT)

In the AUT, participants were instructed to generate as many alternatives uses as possible for three common objects: tire, bottle and knife (Acar & Runco, 2019). The experimenter presented the instructions orally, and then participants read them on the computer screen before starting the task. After reading the instructions, the object's name was displayed on the computer screen and remained until the time ran out. Participants were given three minutes for each object to write all their responses on the computer using a keyboard. At the end of the 3 minutes, participants were asked to select their two most creative ideas for each object (top-2) (Benedek et al., 2013; Silvia et al., 2008). All participants' responses were cleaned (i.e., corrected for typo errors and homogenizing singular and plurals of the same idea). We computed three scores for each object: AUT-fluency, AUT-uniqueness, and AUT-ratings. We calculated the AUT-fluency as the total number of unique responses generated by the participants. The AUT-uniqueness was quantified as the number of ideas generated by less than 5% of the participants. The AUT ratings were the evaluations provided by five experienced, external raters to the top 2 creative ideas (Benedek et al., 2013; Silvia et al., 2008). The raters were provided written instructions translated into French from https://osf.io/vie7s. They were asked to rate the creativity of the ideas using a Likert scale from 0 (not creative) to 4 (highly creative). The inter-rater reliability was good as evidenced by an intraclass correlation coefficient of .74. The database to compute the AUT-uniqueness included eight additional participants who performed the set of creativity tasks but not the PolyFT. The final data for each score was the sum of the scores of the three AUT objects.

The Combination of Associates Task (CAT)

The CAT is an adaptation of the classical creativity task named the Remote Associates Test (Mednick, 1962). In the CAT, participants were presented with 100 trials composed of triplets of cue words that, at first glance, seem unrelated. For each trial, participants were asked to find a fourth word (i.e., the word solution) that relates to all the cue words within 30 seconds. The trials of the CAT vary on the semantic distance between the cue words and the solution, as described in previous articles (Bendetowicz et al., 2017; Bendetowicz et al., 2018; Ovando-Tellez et al., 2023). For each trial, the three cue words were displayed on the screen, and participants were instructed to press the space bar immediately when the solution came to their minds and to write it down on the computer using the keyboard. Then, participants were asked to report if the solution was found with a feeling of insight or Eureka, i.e., when the solution comes to mind suddenly and effortlessly (Kounios & Beeman, 2014; Topolinski & Reber, 2010). They were given five seconds to press the “v” key if they found the solution with an insight or the “n” key otherwise. We quantified the CAT_CR as the participants' accuracy in the task, computed as the percentage of total correct responses. We estimated the CAT-eureka score as the percentage of reported Eureka among correct trials. This score represents the ability of the participants to solve trials via insight versus a more analytical process.

Executive function tests

Digit span test

We used the digit span test of the Wechsler Adult Intelligence Scale (WAIS; Wechsler, 1981) to assess working memory abilities. We used the digit span test's performance criteria based on the WAIS IV manual. This task comprises two parts, in which participants are given a string of numbers that increase in size and are asked to repeat them aloud. In the first part, participants were given 16 different strings of numbers going from 2 to 9 digits consecutively and were instructed to repeat them in the same order. In the second part, they were given 16 other strings from 2 to 8 digits and were instructed to repeat them reversely. Participants performed two trial examples before the actual tasks. We quantified the performance in the second part (Backward-span) by giving 1 point for each correct response. The sum of these points was used as a final score.

Trail-making test

The trail-making test (TMT) assesses set-shifting (Reitan, 1992). This task comprises two parts. In the first part (A), participants were presented with numbers (1 to 25) distributed randomly on a paper. They were instructed to link the numbers in increasing order with a pen as fast as possible. In the second part (B), participants were presented with numbers (1 to 13) and letters (A to L) distributed randomly on a paper. They were instructed to alternate numbers and letters in increasing order as fast as possible. We recorded the time participants took to complete each part of the task. We quantified the difference between the time to complete the second minus the first part as TMT-shifting. Therefore, higher values reflect lower abilities in shifting.

Stroop test

The Stroop test (Stroop, 1935) was used to measure inhibition. We used the French version (Chatelois et al., 1993). In the first part, participants were presented with 100 written names of different colors and were instructed to read them aloud. In the second part, participants were presented with 100 color squares and were instructed to name the color of the ink. In the third part, participants were presented with 100 written names of colors with different color ink and were instructed to name the color of the ink aloud. We recorded the time participants took to complete each part of the task. We quantified the interference effect (Stroop-interference) as the difference in time to complete the third (ink naming with interference) and the second part (color naming) (Chatelois et al., 1993).

Statistical analysis

To explore if the semantic retrieval patterns during the PolyFT are consistent with the MVT, we ran paired Wilcoxon tests to compare the mean IRT and mean IRS between clustering and switching, and to compare the mean IRTr and IRSr between retrieval positions in relation to a cluster switch. We also regressed the absolute difference between the IRT in position − 1 and the long-term IRT on the total number of items retrieved.

To explore the relationship between IRT and IRS, we ran mixed models for the items classified as clustering and switching in the coding based on meanings and separately for the items classified as Fast-Switching, Slow-Switching, Slow-Clustering, and Fast-Clustering by combining our coding with foraging time. Each mixed model explored whether the IRT values predict the IRS values. We included the rank of each word in the model to control for time on task, and included the subjects and cue words as random effects.

To examine how our four types of responses relate to creative abilities and executive function abilities, we ran Spearman correlations between executive function tests and creative abilities tests and Fast-Clustering, Fast-Switching, Slow-Clustering, and Slow-Switching. We used false discovery rate (FDR) to correct for multiple comparisons.

Brain data analysis

MRI data acquisition and preprocessing

The acquisition and preprocessing of the MRI data are the same as described in Ovando-Tellez, Kenett et al. (2022). We collected whole-brain imaging on a 3T scanner (Siemens Prisma, Germany) with a 64-channel head coil while participants performed the RJT task (six runs). Data was collected using multi-echo echo-planar imaging (EPI) sequences. Each run consisted of 335 volumes acquired with a repetition time (TR) = 1600 ms, echo times (TE) for echo 1 = 15.2 ms, echo 2 = 37.17 ms and echo 3 = 59.14 ms, flip angle = 73°, 54 slices, slice thickness = 2.50 mm, isotropic voxel size 2.5 mm, Ipat acceleration factor = 2, multi-band = 3 and interleaved slice order. Next, we acquired a T1-weighted structural image using TR = 2300 ms, TE = 2.76 ms, flip angle = 9°, 192 sagittal slices with a 1mm thickness, isotropic voxel size 1 mm, Ipat acceleration factor = 2, and interleaved slice order. At the end of the session, we acquired resting-state data for 15 minutes (not used in this study) with the same characteristics as those described for the fMRI data. No volume was discarded from the fMRI data since the recording did not contain dummy scans. Functional volumes of each run were first despiked, slice timing corrected, and realigned to the first volume (computed on the first echo) using the afni_proc.py pipeline from the Analysis of Functional Neuroimages software (AFNI; https://afni.nimh.nih.gov). We then denoised the data using the TE-dependent analysis of multi-echo fMRI data (TEDANA; https://tedana.readthedocs.io/en/stable/), version 0.0.9150. The TEDANA pipeline performs an optimal combination of the echo time series followed by a PCA to reduce the data and ICA to decompose the multi-echo BOLD data. The BOLD components were then classified as BOLD or non-BOLD, and the latter was removed, allowing the removal of thermal and physiological noise, including artifacts generated by motion, respiration, and cardiac activity. The last step included the co-registration of the denoised data on the T1-weighted structural image using the Statistical Parametric Mapping (SPM) 12 package running in Matlab (Matlab R2017b, The MathWorks, Inc., USA) and normalization to the Montreal Neurological Institute (MNI) template. To spatially normalize the fMRI data, we used the transformation matrix computed from the normalization of the T1-weighted structural image using the default settings of the computational anatomy toolbox (CAT 12; http://dbm.neuro.unijena.de/cat/) implemented in SPM 12. No participant was removed due to excessive head motion within a single run (> 2 mm translation or > 3° rotation), and no participant had mean Framewise Displacement > 0.5 mm (Power et al., 2014). The denoised and normalized fMRI data were entered in a general linear model in SPM to covary out the task-related signal from each run. We regressed out of the BOLD signal 24 motion parameters (standard motion parameters, first temporal derivatives, standard motion parameters squared, and first temporal derivatives squared) and the onsets and durations of each task-related event (reflection period, response period, inter-trial interval, cross-fixation periods, and change of the cross-fixation color). We then standardized and detrended the residuals of the GLM for each run and concatenated the residuals of the six runs, removing the between-runs rest periods. This approach follows the background connectivity approach (Cole et al., 2019), as intrinsic networks during task performance compared to rest facilitate the prediction of individual traits and differences in brain-behavior relationships (Cole et al., 2021; Greene et al., 2018; Jiang et al., 2020). These preprocessed data (i.e., concatenated residuals of the six RJT runs) were used in the subsequent analyses.

Predicting behavior from the individual’s brain functional connectivity patterns

Building functional connectivity matrices

For each participant, we built a 200 x 200 functional connectivity matrix. To this aim, we first defined our brain regions of interest (ROI). We used the Schaefer atlas (Schaefer et al., 2018), composed of 200 brain regions distributed in 17 functional networks. Using Nilearn v0.3 (Abraham et al., 2014) in Python 2.7, we explored the BOLD signal in these regions and correlated them across time. The 200 ROIs were used as rows and columns to build the individual matrices, and the Pearson correlation coefficients between each pair of regions were used to fill the corresponding matching cells. Therefore, each matrix represents the individual brain network, with the nodes represented by the 200 ROIs connected by links represented by the correlation coefficients.

Connectome brain predictive modeling (CPM)

To predict individual patterns of foraging behavior, we used the CPM approach with leaving-one-out cross-validation (Shen et al., 2017). We ran four independent CPM analyses to predict Fast-Clustering, Fast-Switching, Slow-Clustering, and Slow-Switching, respectively. The CPM consisted of three steps. In the first step, we selected the significant features related to the foraging behavior patterns. For N-1 participants, we correlated each link of their brain network (i.e., each value in the individual's brain connectivity matrix) with the behavior to be predicted. The links that correlated positively or negatively with the behavior (p < .01) were saved as positive and negative masks, respectively. In the second step, for each participant (i.e., N-1), we summed the weights of all the links that composed the positive and negative mask, which resulted in one unique value for each mask for each participant (i.e., mean connectivity strength). In the third step, we used these values to fit a linear regression model as predictors of behavior. In addition, we added a third predictor, which was the participants' mean framewise displacement (Power et al., 2014) during the acquisition. This parameter controls the head movements inside the scanner during the data acquisition and ensures that the prediction is not biased by noise. In the last step, we tested this model in the left-out subject to predict its behavior. We ran these steps N times (N = 86). For each loop, the positive and negative masks were slightly different, so in the end, the final positive and negative masks were composed of brain links that were significantly correlated to the behavior in all the loops. We used the BrainNet toolbox implemented in MATLAB to visualize the predictive networks.

Mediation analysis

To test whether the patterns of functional connectivity that predict individual patterns of memory search are also relevant for creativity, we ran mediation analyses. For significant CPM predictions, we tested whether the frequency of the types of responses during semantic retrieval mediates the relationship between their predictive patterns of brain functional connectivity and creativity. The mediation analyses focused on the types of responses that were predicted by the CPM and that correlated to creativity scores. The mediation analyses explored an indirect effect of functional brain connectivity on creativity through the semantic memory search patterns. The mediation analysis (Hayes, 2017) consisted in calculating the product of (i) the regression coefficient of the regression analysis on the independent variable (i.e., brain functional connectivity of the positive model network) to predict the mediator (i.e., frequency of the response type) and (ii) the regression coefficient of the regression analysis on the mediator to predict the dependent variable (i.e., creativity score) when controlling for the independent variable. We also calculated the regression coefficient of the regression analysis on the independent variable to predict the dependent variable without controlling for the mediator (total effect) and when controlling for it (direct effect). The indirect effect was calculated as the product of path a and path b. All the variables entered in the mediation analyses were normalized. In all analyses, we used the positive predictive brain network because we expected positive correlations with the creativity score. We tested the significance of the indirect effect using a bootstrapping method, computing unstandardized indirect effects for each 5000 bootstrapped samples, and the 95% confidence interval was computed by determining the indirect effects at the 2.5th and 97.5th percentiles. The mediation analyses were performed using MATLAB codes.

Inter-response times and semantic similarity within and between cluster transitions

We explored semantic retrieval patterns during the PolyFT in relation to the predictions of the marginal theorem value. The mean IRT across participants and cue words was 3.6 s (SD = 1.4 s). We identified a mean of 3.7 switches across participants (SD = 1.2, representing 28% of the mean retrieved associations). We first examined whether retrieving items during between-cluster rather than within-cluster transitions takes more time.

On a global average, the switching-IRT was 4.87 s (SD = 1.97 s), while the mean clustering-IRT was 2.89 s (SD = 1.49 s). The Wilcoxon test showed that IRT was lower when people stayed in a cluster than switching to a different cluster. This result was significant for the three cue words (cue1, glace: W = 351, p < .001; cue2, rayon: W = 461, p < .001; cue 3, somme: W = 542, p < .001; Supplementary Figure S1A), and for the mean of the three cue words (W = 312, p < .001; Fig. 1A). We also examined whether semantic similarity when retrieving items during between-cluster transitions differs from within-cluster transitions. Words related to the same meaning are expected to be closer in terms of semantic distance, and we aimed to capture this effect. On a global average, the switching-IRS was 0.06 (SD = 0.05), while the mean clustering-IRS was 0.25 (SD = 0.06). The Wilcoxon test showed that IRS was lower when people switch to a different cluster than when staying in a cluster. This result was significant for the three cue words (cue1, glace: W = 3664, p < .001; cue2, rayon: W = 3357, p < .001; cue 3, somme: W = 2541, p < .001; Supplementary Figure S1B), and for the mean of the three cue words (W = 3740, p < .001; Fig. 1B).

Second, we explored whether participants leave a cluster when they reach their long-term IRT (i.e., when the ratio between the IRT for retrieving an item and the long-term IRT [IRTr] equals 1) as predicted by the marginal value theorem. Consistent with this theory, the mean IRTr was greater than 1 for switches (IRTr = 1.44; Fig. 2A), while it was lower than 1 for non-switching items (IRTr were 0.78 and 0.84 for responses in positions − 2 and − 1 before switching, and 0.63 and 0.83 for the position 2 and 3 after switching, respectively. The IRTr was significantly different for switches than non-switching responses (p < .001; Fig. 2A). The results of statistical tests are reported in Supplementary Table S1.

Third, we explored whether a comparable pattern is observed for IRSr, but in the opposite direction, by comparing if responses generated during within-cluster transitions are more similar to each other than the ones in between-cluster transitions. We observed higher IRSr in responses during within-cluster transitions compared to between-cluster transitions. The mean IRSr during switching (position = 1; Fig. 2B) was 0.35, while for responses before switching, IRSr was higher (1.27 for position − 2 and 1.21 for position − 1). IRSr were also higher for responses after switching (1.65 and 1.18 for positions 2 and 3, respectively). The IRSr for switching was different from all other values (p < .001). Thus, as expected, the first response of a new cluster is highly dissimilar (< 1) to the last response of the previous cluster (pre-switch response) but highly similar (> 1) to the next one in the same cluster (Fig. 2B). The results of statistical tests are reported in Supplementary Table S1.

Finally, following Hills et al. (2012), we explored whether participants with IRT close to reaching their long-term IRT before switching generated more responses during the task (higher PolyFT_fluency), as it would be expected from the marginal value theorem. The value of the IRTr before switching (responses in position − 1 in Fig. 2A) reflects how close the participants are to reaching their threshold or long-term IRT. We regressed the absolute difference between the IRT in position − 1 and the long-term IRT on the total number of responses retrieved (for each PolyFT cue word and each participant). The results show that the closer participants were to their long-term IRT before switching, the more responses they retrieved during the PolyFT, R² = 0.108, F(1,81) = 9.83, p < .01 (Fig. 2C).

We further examined the link between IRS and IRT for clustering and switching responses. Mixed regression models showed that time and similarity correlated significantly for clustering and switching responses (clustering: R² = 0.08, b = − .018, p < .001; switching: R² = 0.07, b = − .004, p < .01). This result indicates that time and similarity covary during clustering and switching; it takes longer to retrieve distant responses in both clustering and switching. Together, these results indicate that memory search during the PolyFT follows the MVT predictions.

Patterns of inter-response time and similarity of the responses categorized as clustering and switching

While IRT and IRS supported MVT on average, participants varied in their adherence to the MVT policies. As shown in Fig. 3, some switching responses occurred faster than expected by the MVT (IRTr < 1), whereas some clustering responses occurred later than expected (IRTr > 1). Therefore, we further subdivided our clustering and switching responses to take into account this incongruency. We used the IRTr of each clustering and switching response to classify them either as Fast-Switching or Fast-Clustering (when IRTr was lower than 1), or as Slow-Switching or Slow-Clustering (IRTr larger than 1) (Supplementary Figure S2-S3). At the group level, participants generated a mean of 21.24 (SD = 9.6) responses corresponding to Fast-Clustering, 5.17 (SD = 2.2) responses corresponding to Fast-Switching, 8.62 (SD = 4.8) responses corresponding to Slow-Clustering and 5.84 (SD = 2.3) responses corresponding to Slow-Switching. In total, Fast-Clustering represented 49.98% (SD = 9.95%) of all responses, Fast-Switching 14.43% (SD = 7.82%), Slow-Clustering 20.17% (SD = 7.17%), and Slow-Switching15.42% (SD = 6.95%).

To explore whether fast and slow switching involved distinct mechanisms, we compared the IRT and IRS for responses in positions − 1 and − 2 for both Fast-Switching vs. Slow-Switching, separately. (Fig. 4). The IRTr and IRSr of responses in position − 1 and − 2 before switching did not differ significantly from each other in Fast-Switching (IRTr: W = 1842, p > .05 and IRSr: W = 1358, p > .05). For Slow-Switching responses in position − 1 and − 2 before switching did not differ for IRSr (W = 2063, p > .05) but these responses were significantly different for IRTr (W = 1411, p < .05). However, as for the global switching responses described above (see Fig. 2), the IRTr and IRSr of the pre-switch responses (position = -1) were significantly different from those of the switching-response (position = 1) in either Slow-Switching (IRTr: W = 9, p < .001 and IRSr: W = 3543, p < .001). and Fast-Switching (IRTr: W = 2493, p < .01; IRSr: W = 3366, p < .001). Finally, the IRTr and IRSr of pre-switch responses (position = -1) in Slow-Switching equaled or were close to 1, while it was farther from 1 before Fast-Switching. Additional statistical analyses confirmed that participants reached their marginal value before Slow-Switching but this was not the case before Fast-Switching (See Supplementary Results S1). The results of statistical tests are reported in Supplementary Table S2. Altogether these findings suggest that fast and slow switching rely on distinct mechanisms.

To further characterize fast and slow clustering and switching responses, we explored the relationship between IRT and IRS separately for each of these types of responses. We ran mixed models to predict the IRS from the IRT values for each type of response and found that IRS was predicted from the IRT values for both Fast-Clustering (R² = .07, b = − .05, p < .001) and Slow-Clustering (R² = .05, b = − .01, p = .007) responses. Thus, it took longer to retrieve more semantically distant concepts within a cluster during both fast and slow clustering. We did not observe this pattern in the Fast-Switching (R² = .09, b = − .01, p = .11) and Slow-Switching (R² = .03, b = -0.00, p = .39) responses (Supplementary Figure S2).

Finally, to explore whether slow clustering and switching responses relate to controlled processes, we ran Spearman correlations between executive function abilities and the number of Fast-Clustering, Fast-Switching, Slow-Clustering and Slow-Switching responses generated by the participants. We found significant correlations between Slow-Switching and higher Backward-span (r_s = .261, p = .015), faster TMT-shifting (r_s = − .232, p = .032) and lower Stroop-interference (r_s = − .257, p = .018). No significant correlations were found between executive function abilities and Fast-Clustering, Fast-Switching and Slow-Clustering. The Spearman correlation coefficients are presented in Supplementary Table S3. Although no correlations remained significant after FDR correction for multiple comparisons, these trends may suggest that individuals with more Slow-Switching responses had better executive function abilities.

Individual patterns of brain functional connectivity predict Fast-Switching and Slow-Clustering

To explore the brain correlates of clustering and switching behaviors, we performed four independent CPM analyses to determine whether individual patterns of task-based brain functional connectivity predict Fast-Clustering, Fast-Switching, Slow-Clustering, and Slow-Switching. Fast-Clustering (r_s = .23, p = .036), Fast-Switching (r_s = .33, p = .002) and Slow-Clustering (r_s = .33, p = .002) were significantly predicted by brain functional connectivity but not Slow-Switching (r_s = − .18, p = .09). After permutation testing, only Fast-Switching (p = .019) and Slow-Clustering (p = .024) remained significant. We went on to characterize the predictive networks of these two response types.

Characterization of the brain networks predicting Fast-Switching and Slow-Clustering

The positive model network predicting Fast-Switching was composed of 19 brain regions connected by 18 links (Fig. 5A). Most of these connections were localized between brain regions of DMN and visual networks (5 links), DAN and visual networks (4 links), DMN and somatomotor networks (3 links), and DMN and salience networks (3 links). The highest degree node was localized in the left inferior parietal lobule (k = 12), region that belong to the DMN network. This brain region connects to brain regions of the visual, somatomotor, salience, limbic and temporoparietal networks.

The model network predicting lower Fast-Switching was composed of 37 brain regions connected by 34 links. Most of these connections were localized between brain regions within DMN (9 links), within DAN (4 links), between DAN and salience networks (3 links), and between DMN and temporoparietal networks (3 links). The highest degree nodes were localized in the left inferior parietal lobule (k = 6), region that belong to the DMN network, and the right post central region of the DAN network (k = 6). The left inferior parietal lobule had connections to DMN brain regions, and one connection to the lateral PFC of the ECN. The right post central region connected to regions of the DAN, ECN and salience networks.

The positive network model predicting Slow-Clustering was composed of 28 brain regions connected by 26 links (Fig. 5B). Most of these connections were localized between brain regions of DAN and somatomotor networks (6 links), ECN and somatomotor networks (4 links), DAN and visual networks (3 links), and DMN and somatomotor networks (3 links). The highest degree nodes were localized in the right superior parietal lobule (k = 8), region that belong to the DAN network, and the left intraparietal sulcus of the ECN (k = 6). These nodes connected to brain regions of the visual, somatomotor and salience networks.

The positive network model predicting lower Slow-Clustering was composed of 24 brain regions connected by 29 links. Most of these connections were localized between brain regions within the DMN (13 links), between DMN and temporoparietal (8 links) and between ECN and DMN networks (4 links). The highest degree nodes were localized in the right temporoparietal regions of the temporoparietal network (k = 9), and brain regions of the DMN (left temporal: k = 8; right anterior temporal: k = 6; right dorsal PFC: k = 4). All these regions had mostly within network connections to the rest of the brain.

Relationships between PolyFT responses, brain connectivity, and creativity

To explore the relationship between PolyFT responses, brain connectivity and creativity, we first explored how individual patterns of memory search during PolyFT relate to creative abilities. We ran Spearman correlations between the number of Fast-Switching, Fast-Clustering, Slow-Clustering and Slow-Switching items generated by the participants and the different measures of the creativity. For the CAT task, we found significant correlations between CAT_CR and Fast-Clustering (r_s = − .244, p = .024) and between CAT_CR and Fast-Switching (r_s = .260, p = .015). No significant correlation was observed for CAT_eureka. For the AUT task, we found significant correlations between AUT-fluency and Slow-Switching (r_s = .213, p = .049), Fast-Clustering (r_s = .455, p < .001) and Slow-Clustering (r_s = .315, p = .003), and between AUT-uniqueness and Fast-Clustering (r_s = .381, p < .001) and Slow-Clustering (r_s = .370, p < .001). No significant correlation was found for AUT-ratings. The correlations between AUT-fluency and AUT-uniqueness and Fast-Clustering and Slow-Clustering remained significant after correction for multiple comparisons. These results suggest that individuals with more Fast-Clustering and Slow-Clustering responses had higher divergent thinking abilities. In addition, our findings may also suggest that individuals with more Fast-Switching have better associative combination abilities, although the correlation lost statistical significance after applying FDR correction for multiple comparisons. Results for statistical tests are reported in Supplementary Table S2.

Since the frequency to which participants produce Fast-Switching and Slow-Clustering responses were predicted by individual patterns of brain functional connectivity, and these frequencies also related to creative abilities, we lastly examined the relationship between brain connectivity patterns and creative abilities. Specifically, we explored whether the relationship between brain functional connectivity (i.e., mean connectivity strength of the positive model network) and creative abilities was mediated by the individual patterns of clustering and switching behavior during semantic retrieval (i.e., semantic foraging patterns). We conducted mediation analyses that focused on the indirect effect of brain connectivity predicting Fast-Switching and Slow-Clustering on creative abilities. In all mediation analyses, we calculated the indirect effect as the product of path a (i.e., the regression coefficient between brain functional connectivity and foraging patterns) and path b (i.e., the regression coefficient between foraging patterns and creativity). We tested the significance of the indirect effect using bootstrapping methods.

Since Fast-Switching was related to the ability to combine remote associates evaluated by the CAT task, we explored the mediating role of Fast-Switching on the relationship between the brain connectivity patterns predicting it and CAT_CR. As shown in the previous analyses, the regression coefficient between the brain connectivity pattern and Fast-Switching (b = .371, p < .001), and between Fast-Switching and CAT_CR (b = .247, p < .05) were significant. The total effect represented by the regression coefficient between the brain connectivity patterns and CAT_CR was significant (b = .201, p < .05), and the direct effect was not significant (b = .109, p = .294). The bootstrapped indirect effect was (0.371) × (0.247) = .092, and the 95% confidence interval ranged from 0.019 to 0.206. Hence, Fast-Switching mediated the relationship between the positive predictive network model of Fast-Switching and CAT performance: The higher the strength of connectivity in the positive network model that predicts Fast-Switching, the higher the number of fast switches in semantic retrieval, and the higher the abilities to combine remote associates in memory.

We also explored the mediation analyses for Slow-Clustering. Because Slow-Clustering was predicted by the fluency and originality of the participants in the AUT, we explored the mediating role of Slow-Clustering on the relationship between the brain connectivity patterns predicting it and AUT_fluency and AUT_uniqueness in independent analyses. As shown in the previous analyses, the regression coefficient between the brain connectivity pattern and Slow-Clustering (b = .50, p < .001) was significant. The regression coefficient between Slow-Clustering and AUT_fluency (b = .284, p < .01) and between Slow-Clustering and AUT_uniqueness (b = .324, p < .001) were significant. The total effect and the direct effect between the brain connectivity patterns were not significant for AUT_fluency (total effect: b = .056, p = .549; direct effect: b = − .087, p = .390) or for AUT_ uniqueness (total effect: b = .036, p = .691; direct effect: b = − .126, p = .200). For AUT_fluency, the bootstrapped indirect effect was (0.50) × (0.284) = .142, and the 95% confidence interval ranged from 0.06 to 0.29, while for AUT_originality, the bootstrapped indirect effect was (0.50) × (0.324) = .162, and the 95% confidence interval ranged from 0.072 to 0.32. Hence, Slow-Clustering mediated the relationship between the positive predictive network model of Slow-Clustering and AUT performance: The higher the strength of connectivity in the positive network model that predicts Slow-Clustering, the higher the frequency of Slow-Clustering during semantic retrieval and the more participants were fluent and original in the divergent thinking task.

Processes occurring during searching in memory are critical for creativity but remain to be clarified (Kenett, 2023). This study characterized distinct foraging behaviors during semantic retrieval, explored their brain correlates, and their relationships with creative abilities and executive function abilities. We show that during PolyFT, individuals follow an optimal foraging policy consistent with the marginal value theorem. However, clustering and switching can also occur fast and slow, and this provides new hypotheses (and measures) for investigating the processes involved in semantic foraging. The differentiation between these fast and slow processes allows us to reconcile prior theories and approaches. Fast-Clustering and Slow-Switching are consistent with the principles of the marginal value theorem and with the classical distinction of clustering and switching in fluency tasks (Troyer et al., 1997). Fast-Switching and Slow-Clustering are in line with the flexibility and persistence pathways of the 'two pathways to creativity' model (Nijstad et al., 2010). Both were relevant to creativity. Furthermore, we identified specific functional connectivity patterns predictive of Fast-Switching and Slow-Clustering, and examined their relationship with creative abilities. Our results indicate that fast-switching between clusters (which might be related to spontaneous processes) mediates the relationship between brain connectivity and the ability to combine remote associates in memory, whereas slow-clustering (which might be related to controlled processes) mediates the relationship between brain connectivity and divergent thinking abilities.

Although different accounts of how people search in memory for concepts exist (Abbott et al., 2012; Todd & Hills, 2020; Troyer et al., 1997), studies using classical fluency tasks have shown that memory retrieval involves the interaction between automatic and controlled processes, possibly involved during clustering and switching, respectively (Troyer et al., 1997) (Hills et al., 2013; Hills & Pachur, 2012; Lundin et al., 2023; Rosen & Engle, 1997) The MVT describes these interactions as local-global transitions related to processes of exploitation-exploration in which global transitions between clusters occurs when the retrieval rate within the current cluster falls below a threshold (Charnov, 1976; Hills et al., 2012; Hills et al., 2015). In our study, we combined the clustering- switching approach with the MVT approach during an associative fluency task. We explored the clustering and switching trade–off as a function of inter-response time and semantic similarity. As expected, we observed a pattern of search behavior consistent with the MVT, in which switches between clusters occur when the time to retrieve a response reaches the individuals' retrieval time in the entire task. Transitioning to another cluster (i.e., switching to a different meaning) took longer than staying within the same cluster, and the retrieved elements were more dissimilar to each other than within cluster responses. It is important to note that the clusters were assigned by design. We expected words related to the same meaning to be closer in terms of semantic distance, and we aimed to capture this effect when comparing semantic similarity within and between clusters. The organization of words in semantic memory based on their meaning or category is a debated question, making our findings particularly noteworthy. Furthermore, the closer participants were to their long-term IRT (their individual threshold) before switching, the more responses they retrieved during the task. Altogether, these findings demonstrated that semantic retrieval in the PolyFT follows the MVT, as previously demonstrated with classical fluency tasks (Hills et al., 2012; Hills et al., 2015; Todd & Hills, 2020). This broader applicability of the MVT principles beyond traditional category-based tasks suggests their wider and more general relevance in understanding how people access and retrieve information from their semantic memory.

Though the MVT prediction was appropriate on average, we also observed considerable variation in whether switches were made or not at any given point during the PolyFT task. Specifically, we discovered that both clustering and switching can occur slow and fast (slower or faster than the MVT would predict). Fast-Clustering is consistent with the MVT prediction, and also converges with the clustering process during classical fluency tasks described in Troyer et al., (1997), which proposed that clustering involves spontaneous memory processes. Conversely, Slow-Clustering is inconsistent with the MVT—participants are staying too long before switching—and the longer latency may suggest the involvement of more controlled processes, in terms of deliberate exploitation of semantic clusters rather than passive associations. We observed that participants with more frequent Fast-Clustering and Slow-Clustering behavior also showed higher ideational fluency and originality. Fast-Clustering aligns with the classically explored clustering behavior in the literature, and its relationship with creative abilities is consistent with the role of memory structure in creativity (Beaty & Kenett, 2023; Benedek et al., 2023; Kenett & Faust, 2019; Mednick, 1962). The relationship between Slow-Clustering and creativity resonates with previous studies suggesting that longer response latency in divergent thinking tasks is related to increased originality and greater dissimilarity between consecutive responses (Acar et al., 2019; Hass, 2017; Mastria et al., 2021). These findings highlight the significance of persistence and thorough exploitation of a topic as a means to generate more and more creative ideas, which is in agreement with the persistence pathway of the dual pathway to creativity model (De Dreu et al., 2008; Nijstad et al., (2010). According to this model, the persistence pathway requires a sustained and focused cognitive effort to achieve creative ideas. Supporting this hypothesis, the brain network predicting Slow-Clustering was characterized by the interactions between DAN and ECN networks in conjunction with somatomotor networks. Closing the loop, we found that the predictive network of higher Slow-Clustering was related to higher divergent thinking ability, which was mediated by higher Slow-Clustering. We observed that highly connected nodes were localized in the right superior parietal lobule of the DAN and the left intraparietal sulcus of the ECN. Previous studies have shown the role of intraparietal sulcus in control-demanding semantic tasks (Noonan et al., 2013) and multiple-demand tasks (Fedorenko et al., 2013), which is consistent with the proposed role of control-related networks in Slow-Clustering. Altogether, these findings suggest that more persistent memory search behavior supported by higher functional connectivity in attentional and control related networks, allows better performance in divergent thinking task.

In contrast to clustering, there was no relationship between time and semantic similarity of items during Slow-Switching and Fast-Switching. The IRT when switching to a new cluster was not related to the similarity between the retrieved concepts, indicating that some switches between meanings occur more quickly, whereas others take longer. Moreover, Slow-Switching and Fast-Switching showed different cognitive patterns. The behavior of Slow-Switching is coherent with the description of switching in Troyer et al. (1997) during semantic retrieval, and its cognitive pattern follows the policies of the MVT. We expected switching to be related to executive abilities and to reflect more controlled processes. Slow-Switching was correlated to executive function abilities related to working memory (Backward-span), inhibition (Stroop-interference) and set-shifting (TMT-shifting), which are key components of executive control (Miyake et al., 2000). Previous work exploring processes of clustering and switching during the generation of creative ideas has shown that, compared to clustering, switching requires a larger investment of cognitive resources during creative ideation and involves longer searching times (Hass, 2017; Mastria et al., 2021). Despite several studies showing the role of inhibition, updating and working memory in creativity (Beaty et al., 2017; Benedek et al., 2014; Edl et al., 2014), we did not find a relationship between Slow-Switching and creative abilities. In a recent publication, Ivancovsky, Baror and Bar (2023) suggest that the ability to flexibly switch between inhibition/disinhibition and exploration/exploitation plays a key role in optimal novelty-seeking performance, and these two factors may differently impact creative abilities (see also Kenett et al., 2023). However, further investigation is needed to understand our non-significant correlation between Slow-Switching and creativity that may result from a lack of power in our sample. Nevertheless, the cognitive patterns of switching described in previous studies (Hass, 2017; Mastria et al., 2021; Ovando-Tellez, Benedek, et al., 2022) in relation to creativity, align with the Slow-Switching behavior identified in the current study.

Contrary to Slow-Switching, Fast-Switching was not predicted by the MVT and was relevant for creative abilities. We found that individuals with more Fast-Switching had better performance in the CAT, a creativity task requiring the combination of unrelated associates in memory (Ovando-Tellez et al., 2023). The link between switching and CAT is plausible as more creative individuals have been shown to be more flexible and are more likely to activate multiple lexico-semantic meaning representations (Beaty & Kenett, 2023; Benedek et al., 2023; Mednick, 1962). The CAT involves activating different semantic representations of each of the three cue words to find the one that is shared across cue words. Switching between different representations of the cue words is thus essential to the performance of this task. Hence, Fast-Switching may reflect the flexibility pathway of the dual pathway to creativity model (Nijstad et al., 2010), which proposes that creative ideas can be produced by spontaneous flexible thinking. The ability of flexible switching between categories or semantic representations may facilitate connecting more remote associates within memory and thus, lead to a better performance in CAT.

However, we were surprised that Fast-Switching and not Slow-Switching correlated with the CAT, as this task has been shown to relate to controlled processes (Bendetowicz et al., 2018; Lee & Therriault, 2013). Nevertheless, if Fast-Switching reflects a spontaneous flexibility that is relatively free from effortful control, its correlation with the CAT is consistent with the associative theory of creativity. Indeed, Mednick (1962) hypothesized that more creative people had flatter associative hierarchies enabling them to connect distant concepts. Fast-Switching responses were produced fast and were remote from the previous item. We thus hypothesize that fast switches reflect semantic shortcuts with respect to the typical memory organization, which may still reflect close associations in the individual semantic memory. This hypothesis remains to be tested more directly. Schilling (2005) proposed that shortcuts in an individual's network of representations occur during insight, triggered by atypical associations. While this idea aligns with our results, we did not observe correlations between Fast-Switching and the insight phenomenon measured in the CAT. Overall, more frequent Fast-Switching may suggest a more flexible exploration in memory that facilitates associative combination abilities.

At the brain level, Fast-Switching was predicted from individuals’ pattern of functional connectivity and mediated the relationship between brain connectivity and creative abilities. The predictive network of higher Fast-Switching behavior predicted the ability to combine remote associates. The brain network predicting Fast-Switching was dominated by DMN connections and was characterized by an inferior parietal lobule region of the DMN overlapping the left angular gyrus highly connected to brain regions of all other functional networks, with no links to DMN and ECN brain regions. Interestingly, the brain network predicting a lower behavior of Fast-Switching was characterized by the same node connecting with other regions within the DMN or with ECN. Previous research indicates that the angular gyrus is a critical brain region for semantic cognition, as it is considered a hub region for semantic integration (Binder et al., 2009; Ralph et al., 2017) and it has been shown to be critical for the automatic retrieval of concepts in memory (Davey et al., 2015; Humphreys & Lambon Ralph, 2015; Jackson, 2021; Jefferies et al., 2020). The angular gyrus has also been associated to different creativity tasks and domains across studies (Gonen-Yaacovi et al., 2013). Moreover, the literature has widely supported the role of the DMN regions in the spontaneous generation of associates (Davey et al., 2016; Teige et al., 2019) and in creative abilities (Beaty et al., 2015; Bendetowicz et al., 2018; Marron et al., 2018). Hence, in agreement with behavioral results, the brain prediction concur to suggest that our measured Fast-Switching may related to associative processes. However, whether Fast-Switching represents purely automatic processes remains to be directly tested.

A recent fMRI study by Lundin et al. (2023) provided further support for the hypothesis that individuals engage in foraging behavior during memory search. Their research revealed that certain brain regions, including the hippocampus and cerebellum, exhibited greater activity when participants were exploring as opposed to exploiting a cluster in a category fluency task. They also showed that the hippocampus and cerebellum displayed increased activity leading up to switches, suggesting a potential cognitive monitoring process. However, their study did not distinguish fast and slow clustering and switching, which constitutes an avenue to future fMRI studies. In the current work, we were unable to predict Fast-Clustering and Slow-Switching from brain connectivity patterns, which prevented us from comparing the brain patterns of functional connectivity characterizing these behaviors with the ones observed in Slow-Clustering and Fast-Switching.

We had previously explored how classical clustering and switching relate to semantic memory structure, executive control abilities and creative abilities (Ovando-Tellez, Benedek, et al., 2022). By analyzing the data using a different approach, in light of the MVT, we were able to delve deeper into this idea. In the current study, we found that the cognitive patterns of Fast-Clustering and Slow-Switching converge with the clustering and switching processes during searching in memory described in Troyer et al. (1997). Fast-Clustering likely involves associative processes, while Slow-Switching relates to controlled processes and follows the policies of MVT. On the other hand, we introduce here Slow-Clustering and Fast-Switching and show that they represent cognitive patterns that play a critical role in creative abilities. These patterns largely converge with the persistent and flexible pathways of the dual pathway to creativity model (Nijstad et al., 2010), and mediate the relationship between brain connectivity and creative abilities. Slow-Clustering relates to a persistent search that is critical for divergent thinking, and is supported by ECN and attentional brain networks, while Fast-Switching involves a more flexible search in memory related to associative combination abilities (measured by the CAT) and supported mainly by the DMN.

We acknowledge some limitations of our study. First, our analyses, based on data from 86 participants, were exploratory and should be replicated using different cue words to validate our findings. Similarly, our results were obtained using polysemous words in the French language. Future research is required to determine if these results can be generalized to other languages or if language-specific factors play a role. Second, we did not account for the possible variations in IRTs along the task. We used the mean IRT across the entire task to define a marginal value for each participant, but this value may change over time. Therefore, our analyses lack dynamic measures related to time during the PolyFT task. It might be important to consider this parameter in future studies as participants may adapt their strategies or behavior with progression of the task.

Our findings demonstrate that the processes occurring during memory search differ between individuals and relate to creativity. Characterizing semantic search behaviors as fast or slow, allowed us to capture more nuanced aspects of foraging behavior, which are represented in the brain patterns of functional connectivity and mediate the creativity-brain relationship. Overall, the new distinction of fast and slow clustering and switching helps clarify the discrepancies between previous approaches and theories on memory search, including Troyer et al. hypothesis, the MVT, and to relate memory search processes with the dual pathway to creativity model. Our results also highlight that the dual process models separating associative and controlled processes are not simply aligned with clustering and switching during memory search. Among these processes, a more persistent memory search reflected in more Slow-Clustering is supported by attentional and control networks and relates to better divergent thinking abilities, whereas higher flexibility to switch faster, possibly by associative processes, is supported by DMN connectivity with other networks (except ECN), and relates to better abilities to combine remote associations in memory. Altogether these findings conciliate different theories and shed light on new measures to explore memory search behavior that may help to better understand the processes underlying creativity.

Abbott, J. T., Austerweil, J. L., & Griffiths, T. L. (2012). Human memory search as a random walk in a semantic network. Advances in neural information processing systems, 25, 3050-3058.
Abraham, A., Pedregosa, F., Eickenberg, M., Gervais, P., Mueller, A., Kossaifi, J., Gramfort, A., Thirion, B., & Varoquaux, G. (2014). Machine learning for neuroimaging with scikit-learn. Frontiers in neuroinformatics, 8, 14.
Acar, S., & Runco, M. A. (2019). Divergent thinking: New methods, recent research, and extended theory. Psychology of Aesthetics, Creativity, and the Arts, 13(2), 153-158. https://doi.org/10.1037/aca0000231
Acar, S., Runco, M. A., & Ogurlu, U. (2019). The moderating influence of idea sequence: A re-analysis of the relationship between category switch and latency. Personality and Individual Differences, 142, 214-217.
Ackerman, R. (2014). The diminishing criterion model for metacognitive regulation of time investment. Journal of Experimental Psychology: General, 143(3), 1349.
Baroni, M., Bernardini, S., Ferraresi, A., & Zanchetta, E. (2009). The WaCky wide web: a collection of very large linguistically processed web-crawled corpora. Language resources and evaluation, 43, 209-226.
Beaty, R. E., Benedek, M., Kaufman, S. B., & Silvia, P. J. (2015). Default and Executive Network Coupling Supports Creative Idea Production [Article]. Scientific Reports, 5, 10964. https://doi.org/10.1038/srep10964
Beaty, R. E., Christensen, A. P., Benedek, M., Silvia, P. J., & Schacter, D. L. (2017). Creative Constraints: Brain Activity and Network Dynamics Underlying Semantic Interference During Idea Production. NeuroImage, 148, 189-196. https://doi.org/10.1016/j.neuroimage.2017.01.012
Beaty, R. E., & Kenett, Y. N. (2023). Associative thinking at the core of creativity. Trends in Cognitive Sciences, 27(7), 671-683. https://doi.org/10.1016/j.tics.2023.04.004
Beaty, R. E., Silvia, P. J., Nusbaum, E. C., Jauk, E., & Benedek, M. (2014). The roles of associative and executive processes in creative cognition. Memory & Cognition, 42(7), 1-12. https://doi.org/10.3758/s13421-014-0428-8
Bendetowicz, D., Urbanski, M., Aichelburg, C., Levy, R., & Volle, E. (2017). Brain morphometry predicts individual creative potential and the ability to combine remote ideas. Cortex, 86, 216-229.
Bendetowicz, D., Urbanski, M., Garcin, B., Foulon, C., Levy, R., Bréchemier, M.-L., Rosso, C., Thiebaut de Schotten, M., & Volle, E. (2018). Two critical brain networks for generation and combination of remote associations. Brain, 141(1), 217-233.
Benedek, M., Beaty, R. E., Schacter, D. L., & Kenett, Y. N. (2023). The role of memory in creative ideation. Nature Reviews Psychology, 2(4), 246-257. https://doi.org/10.1038/s44159-023-00158-z
Benedek, M., & Fink, A. (2019). Toward a neurocognitive framework of creative cognition: the role of memory, attention, and cognitive control. Current Opinion in Behavioral Sciences, 27, 116-122. https://doi.org/https://doi.org/10.1016/j.cobeha.2018.11.002
Benedek, M., Franz, F., Heene, M., & Neubauer, A. C. (2012). Differential effects of cognitive inhibition and intelligence on creativity. Personality and Individual Differences, 53(4), 480-485. https://doi.org/10.1016/j.paid.2012.04.014
Benedek, M., & Jauk, E. (2018). Spontaneous and Controlled Processes. In The Oxford handbook of spontaneous thought: Mind-wandering, creativity, and dreaming (pp. 285).
Benedek, M., Jauk, E., Sommer, M., Arendasy, M., & Neubauer, A. C. (2014). Intelligence, creativity, and cognitive control: The common and differential involvement of executive functions in intelligence and creativity. Intelligence, 46, 73-83. https://doi.org/http://dx.doi.org/10.1016/j.intell.2014.05.007
Benedek, M., Kenett, Y. N., Umdasch, K., Anaki, D., Faust, M., & Neubauer, A. C. (2017). How semantic memory structure and intelligence contribute to creative thought: a network science approach. Thinking & Reasoning, 23(2), 158-183. https://doi.org/10.1080/13546783.2016.1278034
Benedek, M., Könen, T., & Neubauer, A. C. (2012). Associative abilities underlying creativity. Psychology of Aesthetics, Creativity and the Arts, 6(3), 273-281.
Benedek, M., Mühlmann, C., Jauk, E., & Neubauer, A. C. (2013). Assessment of divergent thinking by means of the subjective top-scoring method: Effects of the number of top-ideas and time-on-task on reliability and validity. Psychology of Aesthetics, Creativity and the Arts, 7(4), 341-349. https://doi.org/10.1037/a0033644
Binder, J. R., Desai, R. H., Graves, W. W., & Conant, L. L. (2009). Where Is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies. Cerebral Cortex, 19(12), 2767-2796. https://doi.org/10.1093/cercor/bhp055
Camarda, A., Borst, G., Agogué, M., Habib, M., Weil, B., Houdé, O., & Cassotti, M. (2018). Do we need inhibitory control to be creative? Evidence from a dual-task paradigm. Psychology of Aesthetics, Creativity, and the Arts, 12(3), 351-358.
Charnov, E. L. (1976). Optimal foraging, the marginal value theorem. Theoretical population biology, 9(2), 129-136.
Chatelois, J., Pineau, H., Belleville, S., Peretz, I., Lussier, I., Fontaine, F. S., & Renaseau-Leclerc, C. (1993). Batterie informatisée d'évaluation de la mémoire inspirée de l'approche cognitive. [A computerized memory test battery based on the cognitive approach.]. Canadian Psychology / Psychologie canadienne, 34(1), 45-63. https://doi.org/10.1037/h0078803
Chrysikou, E. G. (2019). Creativity in and out of (cognitive) control. Current Opinion in Behavioral Sciences, 27, 94-99. https://doi.org/https://doi.org/10.1016/j.cobeha.2018.09.014
Davey, J., Cornelissen, P. L., Thompson, H. E., Sonkusare, S., Hallam, G., Smallwood, J., & Jefferies, E. (2015). Automatic and controlled semantic retrieval: TMS reveals distinct contributions of posterior middle temporal gyrus and angular gyrus. The Journal of Neuroscience, 35(46), 15230-15239. https://doi.org/10.1523/jneurosci.4705-14.2015
Davey, J., Thompson, H. E., Hallam, G., Karapanagiotidis, T., Murphy, C., De Caso, I., Krieger-Redwood, K., Bernhardt, B. C., Smallwood, J., & Jefferies, E. (2016). Exploring the role of the posterior middle temporal gyrus in semantic cognition: Integration of anterior temporal lobe with executive processes. NeuroImage, 137, 165-177. https://doi.org/https://doi.org/10.1016/j.neuroimage.2016.05.051
Edl, S., Benedek, M., Papousek, I., Weiss, E. M., & Fink, A. (2014). Creativity and the Stroop interference effect. Personality and Individual Differences, 69, 38-42. https://doi.org/http://dx.doi.org/10.1016/j.paid.2014.05.009
Fedorenko, E., Duncan, J., & Kanwisher, N. (2013). Broad domain generality in focal regions of frontal and parietal cortex. Proceedings of the National Academy of Sciences, 110(41), 16616-16621.
Forthmann, B., Jendryczko, D., Scharfen, J., Kleinkorres, R., Benedek, M., & Holling, H. (2019). Creative ideation, broad retrieval ability, and processing speed: A confirmatory study of nested cognitive abilities. Intelligence, 75, 59-72.
Gonen-Yaacovi, G., de Souza, L. C., Levy, R., Urbanski, M., Josse, G., & Volle, E. (2013). Rostral and caudal prefrontal contribution to creativity: A meta-analysis of functional imaging data [Original Research]. Frontiers in Human Neuroscience, 7, 465. https://doi.org/10.3389/fnhum.2013.00465
Hass, R. W. (2017). Semantic search during divergent thinking. Cognition, 166, 344-357. https://doi.org/http://dx.doi.org/10.1016/j.cognition.2017.05.039
Hayes, A. F. (2017). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. Guilford publications.
He, L., Kenett, Y. N., Zhuang, K., Liu, C., Zeng, R., Yan, T., Huo, T., & Qiu, J. (2021). The relation between semantic memory structure, associative abilities, and verbal and figural creativity. Thinking & Reasoning, 27(2), 268-293. https://doi.org/10.1080/13546783.2020.1819415
Hills, T. T., Jones, M. N., & Todd, P. M. (2012). Optimal foraging in semantic memory. Psychological Review, 119(2), 431-440. https://doi.org/10.1037/a0027373
Hills, T. T., Mata, R., Wilke, A., & Samanez-Larkin, G. R. (2013). Mechanisms of age-related decline in memory search across the adult life span. Developmental Psychology, 49(12), 2396-2404.
Hills, T. T., & Pachur, T. (2012). Dynamic search and working memory in social recall. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38(1), 218-228.
Hills, T. T., Todd, P. M., & Jones, M. N. (2015). Foraging in semantic fields: How we search through memory. Topics in Cognitive Science, 7(3), 513-534. https://doi.org/10.1111/tops.12151
Humphreys, G. F., & Lambon Ralph, M. A. (2015). Fusion and fission of cognitive functions in the human parietal cortex. Cerebral Cortex, 25(10), 3547-3560.
Ivancovsky, T., Baror, S., & Bar, M. (2023). A shared novelty-seeking basis for creativity and curiosity. Behavioral and Brain Sciences.
Jackson, R. L. (2021). The neural correlates of semantic control revisited. NeuroImage, 224, 117444.
Jauk, E., Benedek, M., Dunst, B., & Neubauer, A. C. (2013). The relationship between intelligence and creativity: New support for the threshold hypothesis by means of empirical breakpoint detection. Intelligence, 41(4), 212-221. https://doi.org/http://dx.doi.org/10.1016/j.intell.2013.03.003
Jefferies, E., Thompson, H., Cornelissen, P., & Smallwood, J. (2020). The neurocognitive basis of knowledge about object identity and events: dissociations reflect opposing effects of semantic coherence and control. Philosophical Transactions of the Royal Society B, 375(1791), 20190300.
Kenett, Y. N. (2023). Creatively searching through semantic memory structure: a short integrative review. In L. J. Ball & F. Vallée-Tourangea (Eds.), The Routledge International Handbook of Creative Cognition (pp. 160-179). Routledge.
Kenett, Y. N., & Faust, M. (2019). A semantic network cartography of the creative mind. Trends in Cognitive Sciences, 23(4), 271-274. https://doi.org/10.1016/j.tics.2019.01.007
Kenett, Y. N., Humphries, S., & Chatterjee, A. (2023). A thirst for knowledge: Grounding curiosity, creativity, and aesthetics in memory and reward neural systems. Creativity Research Journal, 35(3), 412-426. https://doi.org/10.1080/10400419.2023.2165748
Kounios, J., & Beeman, M. (2014). The cognitive neuroscience of insight. Annual review of psychology, 65, 71-93. https://doi.org/10.1146/annurev-psych-010213-115154
Krumm, G., Filippetti, V. A., & Gutierrez, M. (2018). The contribution of executive functions to creativity in children: What is the role of crystallized and fluid intelligence? Thinking Skills and Creativity, 29, 185-195.
Lee, C. S., & Therriault, D. J. (2013). The cognitive underpinnings of creative thought: A latent variable analysis exploring the roles of intelligence and working memory in three creative thinking processes. Intelligence, 41, 306-320.
Lloyd-Cox, J., Christensen, A. P., Silvia, P. J., & Beaty, R. E. (2021). Seeing outside the box: Salient associations disrupt visual idea generation. Psychology of Aesthetics, Creativity, and the Arts, 15(4), 575.
Lundin, N. B., Brown, J. W., Johns, B. T., Jones, M. N., Purcell, J. R., Hetrick, W. P., O’Donnell, B. F., & Todd, P. M. (2023). Neural evidence of switch processes during semantic and phonetic foraging in human memory. Proceedings of the National Academy of Sciences, 120(42), e2312462120.
Marron, T. R., Lerner, Y., Berant, E., Kinreich, S., Shapira-Lichter, I., Hendler, T., & Faust, M. (2018). Chain Free Association, Creativity, and the Default Mode Network. Neuropsychologia. https://doi.org/https://doi.org/10.1016/j.neuropsychologia.2018.03.018
Mastria, S., Agnoli, S., Zanon, M., Acar, S., Runco, M. A., & Corazza, G. E. (2021). Clustering and switching in divergent thinking: Neurophysiological correlates underlying flexibility during idea generation. Neuropsychologia, 158, 107890.
Mednick, S., A. (1962). The associative basis of the creative process. Psychological Review, 69(3), 220-232. http://www.ncbi.nlm.nih.gov/pubmed/14472013
Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013). Distributed representations of words and phrases and their compositionality. Advances in neural information processing systems, 26, 1-9.
Miroshnik, K. G., Forthmann, B., Karwowski, M., & Benedek, M. (2023). The relationship of divergent thinking with broad retrieval ability and processing speed: A meta-analysis. Intelligence, 98, 101739. https://doi.org/https://doi.org/10.1016/j.intell.2023.101739
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive psychology, 41(1), 49-100.
Nijstad, B. A., De Dreu, C. K., Rietzschel, E. F., & Baas, M. (2010). The dual pathway to creativity model: Creative ideation as a function of flexibility and persistence. European review of social psychology, 21(1), 34-77.
Noonan, K. A., Jefferies, E., Visser, M., & Lambon Ralph, M. A. (2013). Going beyond inferior prefrontal involvement in semantic control: evidence for the additional contribution of dorsal angular gyrus and posterior middle temporal cortex. Journal of Cognitive Neuroscience, 25(11), 1824-1850. https://doi.org/10.1162/jocn_a_00442
Ovando-Tellez, M., Benedek, M., Kenett, Y. N., Hills, T., Bouanane, S., Bernard, M., Belo, J., Bieth, T., & Volle, E. (2022). An investigation of the cognitive and neural correlates of semantic memory search related to creative ability. Communications Biology, 5(1), 604. https://doi.org/10.1038/s42003-022-03547-x
Ovando-Tellez, M., Kenett, Y. N., Benedek, M., Bernard, M., Belo, J., Beranger, B., Bieth, T., & Volle, E. (2022). Brain connectivity–based prediction of real-life creativity is mediated by semantic memory structure. Science Advances, 8(5), eabl4294. https://doi.org/doi:10.1126/sciadv.abl4294
Ovando-Tellez, M., Kenett, Y. N., Benedek, M., Bernard, M., Belo, J., Beranger, B., Bieth, T., & Volle, E. (2023). Brain Connectivity-Based Prediction of Combining Remote Semantic Associates for Creative Thinking. Creativity Research Journal, 35(3), 522-546. https://doi.org/10.1080/10400419.2023.2192563
Ovando-Tellez, M. P., Bieth, T., Bernard, M., & Volle, E. (2019). The contribution of the lesion approach to the neuroscience of creative cognition. Current Opinion in Behavioral Sciences, 27, 100-108.
Pan, X., & Yu, H. (2016). Different effects of cognitive shifting and intelligence on creativity. The Journal of Creative Behavior, n/a-n/a. https://doi.org/10.1002/jocb.144
Power, J. D., Mitra, A., Laumann, T. O., Snyder, A. Z., Schlaggar, B. L., & Petersen, S. E. (2014). Methods to detect, characterize, and remove motion artifact in resting state fMRI. NeuroImage, 84, 320-341.
Ralph, M. A. L., Jefferies, E., Patterson, K., & Rogers, T. T. (2017). The neural and computational bases of semantic cognition. Nature reviews neuroscience, 18(1), 42-55.
Reitan, R. M. (1992). Trail Making Test: Manual for administration and scoring. Reitan Neuropsychology Laboratory.
Rosen, V. M., & Engle, R. W. (1997). The role of working memory capacity in retrieval. Journal of Experimental Psychology: General, 126(3), 211-227.
Schaefer, A., Kong, R., Gordon, E. M., Laumann, T. O., Zuo, X.-N., Holmes, A. J., Eickhoff, S. B., & Yeo, B. T. (2018). Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cerebral Cortex, 28(9), 3095-3114.
Schilling, M. A. (2005). A "small-world" network model of cognitive insight. Creativity Research Journal, 17(2-3), 131-154. https://doi.org/10.1080/10400419.2005.9651475
Schneider, W. J., & McGrew, K. S. (2018). The Cattell-Horn-Carroll theory of cognitive abilities. In D. P. Flanagan & E. M. McDonough (Eds.), Contemporary intellectual assessment: Theories, tests, and issues (pp. 73-163). The Guilford Press.
Shen, X., Finn, E. S., Scheinost, D., Rosenberg, M. D., Chun, M. M., Papademetris, X., & Constable, R. T. (2017). Using connectome-based predictive modeling to predict individual behavior from brain connectivity. nature protocols, 12(3), 506-518.
Silvia, P. J., Beaty, R. E., & Nusbaum, E. C. (2013). Verbal fluency and creativity: General and specific contributions of broad retrieval ability (Gr) factors to divergent thinking. Intelligence, 41(5), 328-340. https://doi.org/10.1016/j.intell.2013.05.004
Silvia, P. J., Winterstein, B. P., Willse, J. T., Barona, C. M., Cram, J. T., Hess, K. I., Martinez, J. L., & Richard, C. A. (2008). Assessing creativity with divergent thinking tasks: Exploring the reliability and validity of new subjective scoring methods. Psychology of Aesthetics, Creativity, and the Arts, 2(2), 68-85.
Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643-662. https://doi.org/10.1037/h0054651
Teige, C., Cornelissen, P. L., Mollo, G., Alam, T. R. d. J. G., McCarty, K., Smallwood, J., & Jefferies, E. (2019). Dissociations in semantic cognition: Oscillatory evidence for opposing effects of semantic control and type of semantic relation in anterior and posterior temporal cortex. Cortex, 120, 308-325.
Todd, P. M., & Hills, T. T. (2020). Foraging in Mind. Current Directions in Psychological Science, 29(3), 309-315. https://doi.org/10.1177/0963721420915861
Topolinski, S., & Reber, R. (2010). Gaining insight into the “Aha” experience. Current Directions in Psychological Science, 19(6), 402-405.
Troyer, A. K., Moscovitch, M., & Winocur, G. (1997). Clustering and switching as two components of verbal fluency: Evidence from younger and older healthy adults. neuropsychology, 11(1), 138-146.
Troyer, A. K., Moscovitch, M., Winocur, G., Alexander, M. P., & Stuss, D. (1998). Clustering and switching on verbal fluency: the effects of focal frontal- and temporal-lobe lesions. Neuropsychologia, 36(6), 499-504. http://www.ncbi.nlm.nih.gov/pubmed/9705059
Volle, E. (2018). Associative and controlled cognition in divergent thinking: Theoretical, experimental, neuroimaging evidence, and new directions. In R. E. Jung & O. Vartanian (Eds.), The Cambridge handbook of the neuroscience of creativity (pp. 333-362). Cambridge University Press.
Wechsler, D. (1981). Wechsler adult intelligence scale. Frontiers in Psychology.

There is NO Competing Interest.

20231130SupplementaryMaterialmot.docx
Supplementary Material

Download PDF

Version 1

posted

You are reading this latest preprint version

Switching, fast and slow: Deciphering the dynamics of memory search, its brain connectivity patterns, and its role in creativity

Status:

Version 1

Abstract

Figures

Introduction

Material and Methods

Statistical analysis

Building functional connectivity matrices

Connectome brain predictive modeling (CPM)

Results

Discussion

Conclusions

References

Additional Declarations

Supplementary Files

Status:

Version 1