Table 2.

Neuroimaging technique, key findings, and limitations.

Study	Neuroimaging techniques used	Key findings	Limitations
Ulrich et al. (2014)	Perfusion fMRI (Arterial Spin Labeling) during mental arithmetic task. Within-subjects: boredom vs. flow vs. overload conditions (3-min blocks).	– Flow vs. others: Increased activity in left IFG (Brodmann 45) and left putamen; Decreased activity in mPFC and amygdala. – Interpretation: IFG activation = greater executive control; Putamen = reward/outcome anticipation. mPFC down = less self-referential thought, Amygdala down = lower arousal/negative emotion.	– Sample: N = 27 healthy students; limited diversity (possible gender imbalance not reported). – Task: Arithmetic may not generalize to all flow activities (e.g., creative arts). – Connectivity: Only regional activation measured; functional connectivity between regions was not analyzed (inferred DMN involvement only by deactivation pattern). – Flow measurement: Relied on task design and post-task ratings; real-time flow fluctuations are not captured.
Ulrich et al. (2016)	Task fMRI (BOLD) with short blocks (30 s) of boredom, flow, and overload in arithmetic task. fMRI contrasts + physiological measures (skin conductance).	– Flow activation: Significant increases in bilateral anterior insula, bilateral inferior frontal gyrus, basal ganglia (putamen/caudate), and midbrain during flow vs. non-flow. – Flow deactivation: Decreased BOLD in mPFC, PCC (DMN core nodes), and in medial temporal lobe, including amygdala. – Replicated U-shaped pattern: intermediate arousal (skin conductance) in flow, higher in overload, lower in boredom (supporting flow’s distinct physiological profile).	– Sample: N = 23, all male young adults, which limits generalizability to females or older populations. – Design: Task order possibly counterbalanced but not explicitly stated (if not, order effects could influence results). – Analysis: Still activation-based; no direct connectivity analysis (though authors discuss possible network interactions). – Flow validity: Short 30 s “flow” blocks may not fully capture sustained flow; however, subjective and EDA data confirmed participants did experience something akin to flow.
Ulrich et al. (2022a)	Task fMRI (BOLD) during a challenge-skill task, inducing flow, boredom, overload (block design). PPI connectivity analysis centered on right anterior insula (rAI).	– Connectivity (Flow vs. others): Increased functional coupling of rAI with left and right DLPFC (ECN) during flow. No flow-related coupling increase with mPFC or amygdala. – Decreased coupling of rAI with ventral striatum (ventral caudate/nucleus accumbens) during flow relative to boredom/overload. – Activation: rAI itself showed an “inverted U” activation (highest in flow, lower in boredom/overload); ventral striatum showed a U-shaped pattern (least active during flow). – Interpretation: The anterior insula orchestrates network switching: in flow, it strongly engages ECN (DLPFC) and disengages reward input (striatum) relative to non-flow. This supports the salience network driving focus in flow.	– Sample: N = 41, all males. Gender-specific effects are unknown. – Generality: Used a lab task; unclear if findings generalize to real-world flow (e.g., sports or arts). – Regions of interest: Focused on insula connectivity; may have missed other network interactions (e.g., direct DMN–ECN coupling) by not doing whole-brain connectivity analysis beyond rAI. – Subjective flow: Assessed (presumably) via questionnaires but not reported in detail in results—assuming flow manipulation was effective, but individual differences in experienced flow are not analyzed against connectivity.
Huskey et al. (2018)	Task fMRI (BOLD) using an open-source video game with varying difficulty. PPI connectivity analysis between cognitive control regions (e.g., DLPFC) and reward regions. Also measured self-reported intrinsic reward and flow.	– Balanced-difficulty (highest engagement) condition showed the highest self-reported intrinsic reward/engagement. – In PPI analyses using nucleus accumbens as the seed, connectivity differences were observed with frontal control-related regions when contrasting balanced vs. low/high difficulty. – Basal ganglia findings (including putamen) included an additional reaction-time component indexing engagement	– Multi-study design with a moderate fMRI sample, limiting precision/generalizability. – Naturalistic first-person game boosts ecological validity but may not generalize to other flow domains/non-gamers. – In parts of the paper, intrinsic reward/engagement is used as a proxy for flow, which is related but not identical. – PPI results are seed-dependent: effects were reported with nucleus accumbens as seed; when DLPFC was the seed, a significant DLPFC–accumbens link was not reported, so findings should not be framed as bidirectional control–reward coupling. – The added reaction-time task and large RT differences raise a dual-task confound, meaning basal ganglia findings may reflect control demands rather than reward per se. – Multiple analyses/contrasts increase interpretive complexity.
de Sampaio Barros et al. (2018)	NIRS (near-infrared spectroscopy) monitoring oxygenation in prefrontal and parietal regions + physiological measures (heart rate, HRV, breathing) during gameplay. Within-subjects: easy, optimal, hard, and self-selected difficulty conditions.	– Optimal (flow) vs. Easy/Hard: Greater self-reported flow and higher oxygenated hemoglobin concentration in right lateral PFC and right inferior parietal lobule (frontoparietal network) during optimal difficulty. Indicates increased activation of attentional control regions in flow. – Autonomy (self-chosen level): Did not increase flow beyond optimal but showed even greater physiological arousal (higher breathing rate, lower HRV) and high PFC/parietal activation. Suggests choice adds effort/stress without boosting flow feelings. – Attentional focus: Fewer task-unrelated thoughts (“mind-wandering”) were reported in the flow condition. – Conclusion: Flow is supported by mobilizing attentional resources in frontoparietal regions, enabling absorption. Autonomy can engage those resources but may introduce performance pressure, not necessarily increasing subjective flow.	– Technique: NIRS has limited depth (cortical surface only) and spatial resolution, so deeper DMN nodes (mPFC, PCC) were not directly measured—only lateral PFC and parietal regions. – Sample: N = 18, relatively small; all young adults. – Flow generalization: Used simple games (Pong, Tetris) —flow dynamics might differ in complex or long-term tasks. – No direct network measure: Although results imply frontoparietal (executive/attention) network engagement, no connectivity between regions or other networks was analyzed. – Autonomy condition: Order of conditions not clear—if autonomy was always last, effects could partly be due to fatigue or expectation.
Beaty et al. (2016)	Task fMRI (BOLD) during a creative divergent thinking task (“Alternate Uses” for common objects) vs. a control task (object characteristics listing). Analyzed whole-brain functional connectivity (using network-based analyses and seed-to-voxel) to identify network coupling related to creativity.	– Creative ideation vs. control: Increased functional connectivity between DMN and ECN regions during creative thinking. Notably, stronger coupling between medial prefrontal (DMN) and dorsolateral prefrontal (ECN) regions, and between inferior parietal (DMN) and inferior frontal (ECN) regions, was observed. This indicates cooperation between introspective and executive processes during creativity. – Resting-state follow-up: In a separate sample, resting connectivity analysis confirmed that the regions active in the creative task belong to distinct networks (default, executive, salience) and that individuals with higher creativity had more integrated networks at rest. – Divergent thinking ability: The efficiency of the identified creative brain network (including DMN and ECN nodes) correlated with participants’ divergent thinking scores. – Interpretation: The ability to generate creative ideas relies on concurrent engagement of spontaneous thought (DMN) and controlled processing (ECN)—a neural parallel to the flow experience of being freely imaginative yet task-focused.	– Flow not measured: Did not explicitly induce or measure “flow,” though participants were performing a creative task. Relevance to flow is inferred (creative immersion likely engages similar networks). – Sample: n = 163 across multiple analyses (including validation sample), which is robust. However, all were young adults; needs replication in other age groups. – Task design: Block design might allow mind-wandering even in control blocks; however, contrasts were carefully done. – Causality: Correlational fMRI; cannot determine if DMN–ECN connectivity causally improves creativity or is a byproduct of creative effort.
Beaty et al. (2018)	Resting-State fMRI connectivity + Creative ability assessment. Used connectome-based predictive modeling to link intrinsic network connectivity with divergent thinking scores across participants.	– Identified creative network: A whole-brain network spanning default, executive, and salience systems predicted creativity scores. Key hubs included medial prefrontal (DMN), lateral prefrontal (ECN), and insula (salience), among others. – Individuals with stronger connectivity within this tri-network circuitry had higher creative performance (measured by divergent thinking tasks scored for originality). – Provides additional evidence that default executive network coupling (with salience network mediation) is a trait marker of creativity. This complements state-based findings of network coupling during creative tasks.	– Not task-based: Did not involve a flow task, so direct conclusions about flow states cannot be drawn. Included here to support general link between network integration and creativity. – Sample: Adequate size (two samples, n ≈ 90 each). – Method: Predictive modeling is complex; results are robust, but the specific network connections contributing to the model are numerous and not all are easily interpretable (beyond noting DMN/ECN/Salience hubs). – Generality: Focus on divergent thinking; other kinds of creativity (artistic, interpersonal) are not directly tested.
Rosen et al. (2024)	EEG (64-channel) recorded during real-time jazz guitar improvisation. Compared neural oscillations in expert vs. less experienced musicians. Analyzed spectral power changes associated with self-rated flow levels.	– Expert musicians: Achieved flow more frequently and intensely than novices. Their EEG showed patterns of “optimal processing”: specifically, decreased beta power in frontal regions (less executive control interference) and increased alpha/theta synchronization in auditory and visual areas (heightened sensory immersion) during high-flow improvisation (based on press description) TECHEXPLORIST. COM. This reflects the idea of transient hypofrontality—experts in flow showed lower engagement of frontal control circuits, allowing creative processes to unfold more automatically. – Novice musicians: Required more cognitive control (likely higher frontal beta activity) and experienced fewer periods of true flow. Their performance was more effortful and accompanied by more self-monitoring (as inferred from less frontal deactivation). – DMN: The default mode regions were not directly measured (EEG limitation), but indirectly, the high-flow state in experts was linked with “letting go” of self-focused thinking—consistent with reduced DMN influence (as seen in fMRI studies). – Emotional state: High-flow performances were subjectively rated as more enjoyable and creatively satisfying.	– Ecological validity: High (real musical creation in lab), but at cost of experimental control. Each improvisation was unique; it was hard to ensure consistency of “challenge-skill balance” across individuals. – EEG source localization: Difficult to pinpoint deeper sources like DMN nodes; conclusions about DMN are indirect. – Flow measurement: Flow was likely assessed via self-report after each improvisation. Real-time changes had to be inferred rather than continuously rated, which could introduce some inaccuracy in aligning EEG data to flow states. – Group differences: Experts vs. non-experts differ in many ways (years of training, perhaps confidence, etc.), so some EEG differences may not solely be due to flow but expertise generally. However, within experts, clear neural changes from low- to high-flow improvisation strengthen the flow-specific interpretation.
Ulrich et al. (2022b)	Task-based BOLD fMRI (3 T; block design) during a mental arithmetic flow paradigm with three conditions (boredom, flow, overload) and adaptive difficulty; whole-brain task-activation GLM testing quadratic (inverted-U and U-shaped) contrasts across conditions; replication Bayes factor approach to quantify replication evidence for neural effects. Concurrent electrodermal activity (EDA) recording as a psychophysiological marker of the flow manipulation.	Reported strong replication evidence for the inverted-U-shaped EDA effect (highest sympathetic arousal in flow relative to boredom/overload). For neural activation, the study found decisive replication evidence for both canonical quadratic flow effects: (1) an inverted U-shaped activation pattern (greater activation during flow than boredom/overload) in regions including dorsolateral prefrontal cortex, anterior insula, and parietal cortex; and (2) a U-shaped activation pattern (lower activation during flow than boredom/overload) in regions including medial prefrontal cortex, ventral striatum, amygdala, and cingulate cortex (subgenual, middle, posterior).	Men-only sample (N = 41) limits generalizability; authors explicitly note the need for replication in women-only samples under hormonal control. Flow experience ratings could not be fully analyzed as a replication outcome due to an item implementation error (only 2/3 intended items were valid), reducing comparability on subjective flow indices. The replication Bayes factors were derived using a p-value-based upper-bound approximation to Bayes factors, which may overestimate absolute BF magnitudes (though the authors argue this should minimally affect the replication BF ratio). Finally, the paradigm uses mental arithmetic, so some inverted-U activations may partly reflect task demands rather than flow-specific processes.