Using caffeine as a chemical means to induce flow states

is an intrinsically rewarding


Introduction
Introduced by Mihaly Csikszentmihalyi in 1975, Flow is a cognitive state of peak performance that involves extreme focus, total task absorption, reduced self-referential thinking and enhanced motivation (Gold and Ciorciari, 2020;van der Linden et al., 2021a).A key requirement of flow is a match of the individual's skills and the challenge at hand, combined with achievement focus and the presence of, usually positive, action-stimulating emotions, such as enjoyment (Eisenberger et al., 2005;Fullagar and Kelloway, 2009;Keller et al., 2011;van der Linden et al., 2007).
Besides the challenge/skills balance, other known flow triggers are clear goals, immediate feedback, an interplay of novelty, unpredictability, complexity and insight, risk, deep embodiment and factors associated with intrinsic motivation, including passion/purpose, autonomy, mastery and curiosity (Kotler et al., 2022).Jointly, these flow triggers seem to be means to direct attention to the present moment or task.Flow typically occurs during challenging, interesting or intrinsically rewarding activities, including cognitive work, sports or leisure activities such as video gaming (Nakamura and Csikszentmihalyi, 2009).Once in flow, six key characteristics, comprising complete concentration on the present task, the merger of action and awareness, loss of self-reflective thoughts and awareness, a distorted sense of time, a feeling of 'complete control' that is accompanied by enhanced performance and positive affect (e.g.euphoria, intrinsic reward or an enhanced sense of meaning and purpose) (Kotler et al., 2022).It is not clear if all of these characteristics must be present to induce flow, however.
Flow correlates with the degree of intrinsic interest in an activity, or how goal-directed an activity is.In this context, it has been proposed that flow may be computed as a function of I(M; E).This function describes how strongly associated the desired end state or outcome (ends, E) and the means of attaining this goal (means, M) are (Melnikoff et al., 2022).For example, this could be 'add a section about the neural correlates of flow in order to convince the reviewers' (E) and 'summarize the paper at hand in a sentence' (M).This equation integrates various flow triggers such as clear goals, immediate feedback, unpredictability or risk (Kotler et al., 2022), because I(M; E) is maximised when receiving the reward (E; in this example acceptance of the paper) is uncertain prior to taking action (M), but the outcome is certain afterwards (the paper is accepted or rejected).Flow is poor, however, when the actions do not affect the desired outcome E (e.g.any revisions M do not influence whether the paper is accepted or rejected).
In addition, Flow is maximised when the difficulty of a task is matched to the skill level of the individual; in order to sustain taskrelated attention, the activity must neither be too easy, nor overwhelming (van der Linden et al., 2021b).If the challenge/skills balance is perfectly matched to induce flow, self-reported task absorption and performance, attention and focus, heart rate and objective mental effort are all increased compared to conditions of 'boredom' and 'overwhelm' (D.J. Harris et al., 2017a).Notably, flow states feature a feeling of 'effortlessness', which is not because the task-at-hand becomes easier in a flow state, but due to an altered perception of the invested effort (D.J. Harris et al., 2017b).
Importantly, Flow is deemed as a state of human peak performance (Gold and Ciorciari, 2020).In favour of this suggestion, a systematic analysis of the literature supports that flow is associated with many aspects of high performance, such as improvements in cognitive and physical performance, learning, the emotional state (e.g.excitement and more positive affect) as well as wellbeing (greater life and job satisfaction etc) (Peifer et al., 2022).Therefore, means to enter and sustain flow states are highly desirable.Recently, it has been alluded that caffeine might be an effective means to induce flow (van der Linden et al., 2021a).However, despite mechanistic links supporting that caffeine facilitates flow, no direct studies have been conducted.
This paper aims to summarize the behavioural and biological effects of caffeine relevant for inducing flow states.To set the stage, we initially discuss how flow is assessed in animals and humans, and we outline the brain networks involved in flow.We then introduce the biological effects of caffeine.Finally, we elaborate how specific aspects of flow could be enhanced, or negatively affected, by acute and chronic caffeine intake.We also investigate the role of caffeine in attention deficit hyperactivity disorder (ADHD), a condition that can both enhance, but also severely impair, flow proficiency.
Please note that, whilst we give an overview of flow research, in particular where potential interactions with caffeine may exist, we do not aim to extensively summarize the flow literature.For more insight, we refer to other reviews and systematic reviews in the field, such as D. J. Harris et al. (2017b), van der Linden et al. (2021a), Gold and Ciorciari (2020), Khoshnoud et al. (2020) and Peifer et al. (2022).

Measurement of flow in animals and humans
To measure general flow proneness or assess the depth of acute flow experiences of human subjects during various activities, for example video gaming, sports, music or other daily activities, flow questionnaires have been developed and employed (Abuhamdeh, 2020;Ottiger et al., 2021).However, these questionnaires are not always psychometrically validated and are somewhat inconsistent in their conceptualisation and operationalisation of flow (see Abuhamdeh, 2020).
Partly due to standard assessment with questionnaires, the state has exclusively been investigated in humans, but not animals.Flow-based animal research is possible, however, and aspects of the state that do not require self-reports may be investigated (reviewed in Hintze and Yee, 2023).Indeed, some flow characteristics are re-modellable in animals, such as a matching challenge/skills balance, clear goals, immediate feedback, focus and concentration (expressed as distractibility by noise, objects etc) and time distortion (e,g, with temporal bisection tasks).Given that correlations between dopamine D 2 receptor (D 2 R) density in the putamen (dorsal striatum), a reward-related brain area, and flow proneness have been reported (de Manzano et al., 2013), animals with targeted receptor upregulation in this area might be useful flow models.Similarly, animals might be employed to model motivational aspects of flow.For example, adult D2R-OE mice, with a virus-mediated, selective upregulation of dopamine D 2 Rs in the nucleus accumbens core (ventral striatum), exhibit greater motivation to invest effort for rewards, and could be used as a pro-flow model (described in (Simpson et al., 2022)).Inversely, if dopamine D 2 R overexpression is induced during development, D2R-OE mice display lowered motivation and cognitive deficits, thus paralleling schizophrenia, or an anti-flow model.

Brain networks and neurobiology of flow states
Generally, Flow is described as a positive, energetic and gratifying experience that occurs when the challenge level of the activity and skill of the individual are balanced, and involves the modulation of attention, executive function and reward-associated networks (Alameda et al., 2022;Weber et al., 2009).Research across the previous decade has provided more detailed insight into the neural correlates of flow.Broadly, flow involves the Default Mode Network (DMN), Central Executive Network (CEN) and Salience Network (SN), which are controlled by regional dopamine and NE release (reviewed in van der Linden et al., 2021a).We will further describe the impact of dopamine-signalling, which regulates reward-related and motivational behaviour, and the alertness-modulating locus coeruleus-norepinephrine (NE) system in the Sections 4.1 and 4.2.Regarding network dynamics, flow may be roughly summarized as a reduction in DMN activity, thus decreasing self-referential processing, autobiographic memory and mind-wandering, and a reciprocal elevation in CEN activity, which mediates task focus and suppression of irrelevant information (van der Linden et al., 2021a).The Salience Network (SN), consisting of the anterior insula cortex and anterior cingulate cortex (ACC), analyses the internal state and external (dopaminergic and/or adrenergic) stimuli, monitors performance and, by running a reward/futility assessment, switches between DMN and CEN activities to terminate or prolong task engagement and flow, respectively.
Non-invasive imaging techniques, including functional near-infrared spectroscopy (fNIRS) and functional magnetic resonance imaging (fMRI), have been employed to map regional brain activity in experimentally-induced flow, such as during video gaming and arithmetic tasks (Huskey, Wilcox et al., 2018;Klasen et al., 2012;Ulrich et al., 2016;Ulrich et al., 2014;Ulrich et al., 2022;Yoshida et al., 2014).Collectively, by using a skill-balanced, 'matched' flow condition that is contrasted with 'boredom' (too easy) or 'overload' (too difficult) conditions, these studies suggest that flow alters activity in brain areas associated with orienting attention (superior parietal lobes and precentral gyrus), attentional alerting (dorsoanterior insula), self-reference, success/goal evaluation, emotional and reward processing (all prefrontal cortex (PFC)-associated areas), coding of outcome probability (putamen), concentration and (visual) focus (cerebellum, visual system, bilateral intraparietal sulcus (IPS), rostral AAC and orbitofrontal cortex), setting of clear goals (IPS, fusiform face area, dorsal ACC and precuneus), feelings of control (inferior frontal gyrus, bilateral temporal lobes, bilateral angular gyrus and various visual, cerebellar, thalamic and motor-cortical regions) as well as negative arousal (amygdala).Please see D. J. Harris et al. (2017b) for a summary of flow-associated brain regions and their function.
Flow seems to maximise intrinsic motivation by improving connectivity between cognitive control (dorsolateral PFC) and rewardassociated (putamen) brain regions, reinforced by a simultaneous stimulation of brain regions responsible for orienting attention (superior parietal lobule, precentral gyrus) and attentional alerting (dorsoanterior insula) (Huskey, Craighead et al., 2018).A follow-up study verified these results, showing that flow states involve modular network topology, with increased synchrony amongst reward structures and enhanced flexibility in the fronto-parietal control network (Huskey et al., 2022).The latter has been associated with higher cognitive control and adaptability (Cole et al., 2013), which unarguably improve performance and, thus, engagement in challenging tasks capable of triggering flow (Flett, 2015;Kotler et al., 2022).Inversely, lower thickness in cortical control structures was linked to sensation-seeking, impulsivity and N. Reich et al. substance abuse, including caffeine (Holmes et al., 2016), supporting the idea that flow demands tight stimulus control exerted by fronto-parietal brain structures.The 'matched' flow condition during video gaming was further shown to produce a lower global efficiency value relative to 'boredom' or 'overload' conditions, indicating that flow is a metabolically efficient cognitive state (Huskey, Wilcox et al., 2018).Interestingly, flow-related depth of immersion, performance and loss of self-referential processing were positively correlated with the participants' cortical heart-evoked potential (HEP) amplitude, suggesting that flow enhances heart-brain interactions (Khoshnoud et al., 2022).In agreement with the features of flow, increased HEP has been linked to reward-associated motivation (Weitkunat and Schandry, 1990), attentional focus on an external stimulus (Montoya et al., 1993) and reduced pain perception (Shao et al., 2011), while depression, the opposite of a flow state, has been linked to decreased HEP (Terhaar et al., 2012).
In his transient hypofrontality hypothesis, Dietrich proposed that flow and other states of altered consciousness, such as (day)dreaming, meditation or hypnosis, involve inactivation of higher-order PFC structures (Dietrich, 2003).For example, a transient reduction in dorsolateral PFC activity may engender key characteristics of flow, including the loss of abstract thought, self-consciousness and working memory, with a reciprocal increase in automatic (implicit) processing (Dietrich, 2003;D. J. Harris et al., 2017b).However, imaging studies do not consistently support this hypothesis.Whilst lowered medial PFC activity was reported during flow (Ulrich et al., 2016;Ulrich et al., 2014), elevated ventrolateral anddorsolateral PFC (Yoshida et al., 2014), or no change in frontal cortex, activity (Harmat et al., 2015) have been demonstrated.This suggests that flow might only involve altered activity in selected PFC structures, or that the neural substrates of flow are task-specific.For example, transient hypofrontality during flow was observed during endurance running (Stoll and Pithan, 2016), a predictable and automatable activity, whereas a video game such as Tetris, which demands rapid (explicit) decision making, likely requires prefrontal cortical input (Harmat et al., 2015).

Background
Caffeine is a commonly consumed psychoactive compound and found in coffee, tea, cocoa-based products, soda, energy drinks and some medications (caffeine concentrations given in (McLellan et al., 2016)).Upon oral ingestion, caffeine is rapidly absorbed from the small intestines (~80%) and, partially, the stomach (~20%) into the blood stream (Arnaud, 1993;Chvasta and Cooke, 1971).Once in the circulatory system, the substance crosses the blood brain barrier (BBB) (McCall et al., 1982), with cerebral concentrations reflecting ~80% of those in the blood stream (Kaplan et al., 1989), to exert psychoactive effects (Cappelletti et al., 2015).Maximal caffeine plasma concentrations, and thus maximized effect, are usually reached after 30 -60 min (Cappelletti et al., 2015), as dependent on the formulation of caffeine, for example as coffee/cola, a capsule or gum (Kamimori et al., 2002;Liguori et al., 1997).Caffeine half-life times (t½) in the circulatory system typically range from 3 -5 h, but may be as long as 12 h (Cappelletti et al., 2015;McLellan et al., 2016), as influenced by certain lifestyle factors (Table 1) and genetics (Sachse et al., 1999).Ideally, to avoid side effects and standardize substance intake for inducing flow, the consumed caffeine source and surrounding lifestyle factors should be kept consistent.

Impact on cognitive and physical performance
Importantly, low (40 mg; ~0.5 mg/kg) to medium (300 mg; ~4 mg/ kg) caffeine doses offset sleep deprivation and enhance arousal, vigilance, attention, reaction times and cognitive performance for up to or over 10 h (reviewed in McLellan et al., 2016) (M.J. Jarvis, 1993).The impact of caffeine on memory, judgment and decision-making is debated, however (McLellan et al., 2016).Compared to cognitive performance, improvements in physical performance seem to require higher caffeine concentrations between 200 mg (~3 mg/kg) to 500 mg (~7 mg/kg), and sometimes benefit from up to 750 mg (~10 mg/kg).Forms of physical engagement that are boosted by caffeine include endurance-based activities to exhaustion that last multiple minutes to hours, high-intensity efforts lasting beyond 60 s or that necessitate repeated display of speed, power and agility, as well as sports demanding muscular strength and endurance (McLellan et al., 2016).These discrepancies between physical and mental performance suggest that the necessary caffeine dose to facilitate flow may vary depending on the nature of the performed activity.Moreover, because of adenosine accumulation during sleep deprivation, higher caffeine concentrations are necessary in the sleep -deprived to produce the same cognitive and physical improvements observed in a rested state (see McLellan et al., 2016).
On the other hand, high caffeine doses (> 400 mg; ~5.5 mg/kg) impair performance and trigger various adverse effects, such as gastrointestinal distress, nervousness, anxiety or sleep disturbances, if consumed close to bedtime (McLellan et al., 2016).Notably, frequent caffeine users appear to tolerate the drug's side effects better.For example, regular caffeine intake blunts the anxiogenic effects of the substance (Rogers et al., 2010), whilst, unlike infrequent caffeine users, chronic consumers seem to experience benefits in physical endurance at caffeine concentrations above 500 mg (Graham and Spriet, 1995;Spriet et al., 1992).
Because genetics determine coffee sensitivity and proneness to side effects (McLellan et al., 2016), optimal caffeine intake should be titrated.The effects of caffeine desensitize following regular consumption and withdrawal symptoms, such as headache and fatigue, occur after 24 -48 h (Cappelletti et al., 2015).It has been postulated that some benefits of caffeine are, in fact, linked to the relief of withdrawal symptoms (Cappelletti et al., 2015).
Notably, a matched challenge/skills balance is believed to be the most important flow trigger (Kotler et al., 2022).For example, Flett used this concept to define flow (Flett, 2015).Experienced college tennis players were asked to self-rate the perceived challenge (C) and their own skill level (S) in breaks during competitive matches.C and S were scored between 1 -10 each and summed up, then the difference between C and S subtracted from this value (to account for poor skill), in order to generate a challenge/skills balance index (CSBI) with a max value of 20.Values above 10 were defined as flow.As expected, the CSBI demonstrated a strong mutual correlation with functional performance, which dictates self-rated skill level, in this paradigm.
Caffeine is well known to enhance physical and cognitive performance in both a rested and sleep-deprived state (see Section 3.2) (McLellan et al., 2016).This strongly indicates that the substance facilitates engagement in more challenging activities, thus promoting flow.Furthermore, relative to a sleepy or bored state (low arousal), performance and flow are improved and optimized when alertness is increased to medium levels (Section 4.2.1)(van der Linden et al., 2021b).As such, caffeine is presumably particularly useful in facilitating flow when adenosine has accumulated after prolonged wakefulness or following physically and cognitively demanding work (Porkka-Heiskanen and Kalinchuk, 2011).

Dopamine receptors and mesolimbic dopamine release
The dopamine system invokes seeking and desire, while reinforcing feelings of reward, thus shaping behaviour.Dopamine transmission from dopamine-producing neurons in the ventral tegmental area (VTA) ensues across the mesocortical and mesolimbic pathways.Of these, the mesocortical pathway modulates verbal memory and executive function by projecting to the dorsal caudate nucleus and dorsal PFC, whereas mesolimbic dopamine-signalling induces motivation and reward by stimulating the nucleus accumbens and olfactory tubercle in the ventral striatum (Arias-Carrion et al., 2010;Hirano, 2021).Moreover, nigrostriatal dopamine transmission from the substantia nigra (SN) to the putamen and caudate nucleus (dorsal striatum) controls locomotion (Arias-Carrion et al., 2010;van der Linden et al., 2021a).
It is thought that dopamine-signalling mediates selective task attention, response (impulsivity) inhibition, intrinsic motivation, feelings of reward and positive mood, such as enjoyment or optimism, during flow (D.J. Harris et al., 2017b) (Weber et al., 2009).Importantly, brain imaging has demonstrated that experimentally induced flow involves increased neuronal activity in the caudate nucleus and putamen (Ulrich et al., 2016;Ulrich et al., 2014).The latter brain region calculates the value of rewards in order to navigate taking action (Hori et al., 2009), with flow proneness shown to correlate with dopamine D 2 R and D 3 R availability in the putamen (de Manzano et al., 2013).Notably, the dopamine system is also implicated in addiction, such as that related to drugs or gambling (Arias-Carrion et al., 2010).Highly immersive activities, for example: video gaming, have both been associated with addiction and flow, suggesting that common dopaminergic structures are activated by both (Hafeez and Kim, 2014;Seah and Cairns, 2008;van der Linden et al., 2021a).
Caffeine modulates dopamine-signalling in a manner that likely promotes flow states.In humans, 300 mg caffeine elevated the availability of dopamine D 2 Rs in the putamen and ventral striatum (nucleus accumbens and olfactory tubercle), but not caudate nucleus, which also mediated caffeine-evoked improvements in attention (Volkow et al., 2015).Generally, adenosine A 2A Rs and dopamine D 2 Rs are co-expressed in neurons across the striatum (Fink et al., 1992;M. F. Jarvis and Williams, 1989), and the heteromerization of both receptors, as induced by adenosine, inhibits dopamine-signalling (Ferre et al., 2007;Prasad et al., 2021).In turn, acute caffeine intake enhances dopamine D 1 R and D 2 R availability by antagonizing adenosine A 1 R and A 2A R activity, respectively, in striatal neurons (Fig. 2B) (Ferre et al., 2007;Holtzman et al., 1991;Manalo and Medina, 2018;Simola, 2010).
Biologically, dopamine-signalling across dopamine D 2 Rs in the striatum was shown to maintain wakefulness (Qu et al., 2010;Satoh et al., 1999;Wisor et al., 2001).Furthermore, dopamine-elevating drugs, such as modafinil, are clinically approved treatments for daytime sleepiness in e.g.Parkinson's disease (PD) and narcolepsy (Hogl et al., 2002;Minzenberg and Carter, 2008).Inversely, some receptor adenosine A 2A Rs polymorphisms in humans (for instance 1976 C>T, or formerly 1083 T > C, 2592 C>Tins and HT4) negate the vigilance-increasing effects of caffeine during sleep deprivation (Bodenmann et al., 2012), whilst aggravating caffeine-driven sleep disruption (Retey et al., 2007) and anxiety (Alsene et al., 2003;Childs et al., 2008;Rogers et al., 2010).Crucially, the presence of genetic polymorphisms increasing striatal dopamine D 2 R levels were demonstrated to improve flow proneness (Gyurkovics et al., 2016) and predict lower impulsivity and better emotional regulation (Blasi et al., 2009).Indeed, flow has been associated with better emotional stability (low neuroticism) (Peifer et al., 2022).Flow proneness was also correlated with the density of dopamine D 2 Rs in the putamen (de Manzano et al., 2013).This collectively suggests that caffeine-mediated upregulation of striatal dopamine D 2 R-signalling improves both wakefulness and flow proneness.Indeed, it has been suggested that caffeine ingestion may amplify other dopamine-associated behaviours besides flow, such as drug addiction or psychosis (Ferre, 2016;Simola, 2010).
Caffeine also stimulates mesolimbic dopamine release.Rodent studies have shown that, by blocking adenosine A 1 Rs, acute caffeine administration (medium dosages of 10 and 30 mg/kg, but not 3 or 100 mg/kg) heightened extracellular dopamine levels in the shell of the nucleus accumbens (ventral striatum) (Solinas et al., 2002).When chronically administered, caffeine increased tyrosine hydroxylase activity, the rate-limiting enzyme involved in dopamine synthesis, in the VTA and SN, the main sites of dopamine-producing neurons (Datta et al., 1996).Chronic caffeine ingestion also increased both dopamine levels and TH activity in the striatum in vivo (Hsu et al., 2010;Kirch et al., 1990).Interestingly, there appears to be a region-specific effect in the nucleus accumbens shell, where acute, but not chronic, caffeine intake stimulates glutamate and dopamine release (Quarta et al., 2004).
Importantly, in mice chronically treated with caffeine, the locomotor-enhancing effects of the substance were blunted, indicating desensitization (Holtzman et al., 1991;Karcz-Kubicha et al., 2003;Shi et al., 1994;Svenningsson et al., 1999).Generally, activation of both striatal dopamine D 1 Rs and D 2 Rs is motor-stimulating (reviewed in Zhou, 2020).Acute caffeine intake promotes dopamine D 1 R/D 2 R-mediated locomotion by antagonizing both adenosine A 1 Rs and A 2A Rs, but these motor activity-stimulating effects are lost in an A 1 R-associated manner following chronic caffeine administration (Karcz-Kubicha et al., 2003).Collectively, the available in vivo evidence suggests that chronic caffeine intake results in an increase in adenosine A 1 R levels in the striatum (Daval et al., 1989;Shi et al., 1994), but no effect was also reported (Svenningsson et al., 1999).The density of striatal adenosine A 2A Rs does not seem to be affected by chronic caffeine consumption in animals (Johansson et al., 1993;Shi et al., 1993), although some studies have reported downregulated receptor expression (Svenningsson et al., 1999;Tronci et al., 2006).Besides adenosine receptors, chronic caffeine consumption does not alter striatal dopamine D 1 R and D 2 R expression, or activation by agonists (Shi et al., 1993(Shi et al., , 1994;;Tronci et al., 2006).
It is well established that especially high and daily caffeine administrations induce desensitization (tolerance), and various mechanisms have been suggested (see Ferre, 2008).First, as discussed above, chronic caffeine intake induces a compensatory upregulation of adenosine A 1 Rs that likely dilutes the facilitatory effects of caffeine on dopamine D 1 R-signalling in the striatum (Fig. 2C).By contrast, congruent with the fact that adenosine A 2 R expression is not increased, studies using adenosine A 2 R antagonists in PD animal models suggest that there is no desensitization of this receptor (Pinna et al., 2001;Popoli et al., 2000).Second, adenosine A 1 R:A 2 R heteromers are affected.Such heteromers were shown to exist in glutamatergic cortico-striatal terminals, where adenosine can either block (at low concentrations) or stimulate (at high levels) striatal glutamate release (Fig. 2A) (Ciruela et al., 2006).The same group showed that chronic caffeine administration potentiates the inhibitory impact of adenosine A 2 Rs (activated by adenosine) on adenosine A 1 Rs, while affinity of adenosine A 2 Rs towards caffeine is lowered (Fig. 2C).Third, regular caffeine ingestion dose-dependently leads to a reversible increase in the plasma levels of adenosine, which might compete with caffeine (Conlay et al., 1997).
Interestingly, chronic or subchronic (every other day) caffeine intake followed by three days of caffeine withdrawal resulted in hyperaffinity of dopamine D 2 Rs, seemingly due to a lasting antagonism of adenosine A 2A Rs (Hsu et al., 2010;Simola, 2010;Simola et al., 2008;Simola et al., 2006).The latter studies suggest that sensitivity to caffeine was not only restored after withdrawal, but that locomotor stimulation by either caffeine or dopamine-releasing amphetamine was further potentiated.Thus, upscaling caffeine intake, in combination with strategic multi-day wash-out periods, might be employed to maintain and recover caffeine sensitivity, respectively.Moreover, dopamine hypersensitivity in response to re-introduction of caffeine may be exploited to carry out a deep flow session.

Dopamine-associated improvements in positive affect
By definition, Flow is autotelic experience and characterised by positive emotions, e.g.optimism and pleasure (Gold and Ciorciari, 2020).In flow-prone individuals with higher intrinsic motivation, positive affect also seems to be a mediating factor for preferring a higher challenge/skills balance, having a greater desire for accomplishments and showing elevated performance (Eisenberger et al., 2005).Fig. 2. Effects of acute and chronic caffeine intake in different striatal synapses.Striato-nigral and striato-pallidal GABAergic spiny projection neurons (SPNs) exclusively express dopamine D 1 Rs or dopamine D 2 Rs, respectively (shown are their dendrites/postsynaptic sites).Different combinations of adenosine A 1 Rs and A 2 Rs, including presynaptic receptor heteromers in cortico-striatal glutamatergic terminals, are also expressed in a synapse-specific manner, and can either be pre-or postsynaptic.Furthermore, adenosine A 1 Rs and A 2 Rs form heteromers with dopamine D 1 Rs and D 2 Rs, respectively.Notably, some of these, such as adenosine A 2 R: dopamine D 2 R heteromers, are constitutive and do not require ligand binding for their formation.(A) In the absence of caffeine, adenosine binds to adenosine A 1 Rs and A 2A Rs on postsynaptic terminals.This process inhibits dopamine D 1 R-signalling and alters D 2 R-signalling through noncanonical Gα q/11 -coupling.The result is a loss of motivation and drowsiness following adenosine accumulation via sleep withdrawal and cognitive/physical work.In cortico-striatal synapses, the amount of adenosine influences glutamate release: low adenosine concentrations mainly activate adenosine A 1 Rs, thus reducing neurotransmitter release below baseline, whereas high adenosine levels stimulate adenosine A 2A Rs to potentiate glutamate release.(B) If caffeine, a non-selective adenosine receptor antagonist, is ingested, binding of adenosine to its receptors is prevented, thus facilitating striatal dopamine D 1/2 R-signalling.Activation of both striato-nigral dopamine D 1 Rs and striatopallidal dopamine D 2 Rs induce locomotion.Moreover, stimulation of accumbal dopamine D 1 Rs results in pleasure ('linking'), while dopamine D 2 R enhance willingness to invest effort into obtaining rewards ('wanting'); these dopaminergic effects of caffeine likely facilitate flow states.Caffeine-mediated blockage of adenosine A 1 R:A 2A R heteromers leads to baseline glutamate release in cortico-striatal terminals.(C) Upon chronic caffeine intake, adenosine A 1 Rs are upregulated, which dilutes the effects of caffeine and nullifies the facilitatory impact on dopamine D 1 Rs.Adenosine A 2A R expression is unaltered or, in some studies, downregulated.Such a receptor downregulation likely leads to dopamine D 2 R hyperexcitability, if chronic caffeine is withdrawn for a few days.Regarding adenosine A 1 R:A 2A R heteromers in cortico-striatal presynaptic boutons, affinity of adenosine A 2A R towards caffeine is reduced, while the response to adenosine is strengthened.Jointly, this results in elevated glutamate release.Dopamine D 1 Rs and adenosine A 1 Rs also modulate neurotransmitter release, explaining why caffeine ingestion may induce the release of neurotransmitters, such as acetylcholine or glutamate (e.g. in the hippocampus, where the density of adenosine A 1 Rs is high) (Alasmari, 2020;Carter et al., 1995;Lopez-Cruz et al., 2018).Besides antagonising adenosine receptors, ingesting beyond 400 mg caffeine (toxic threshold in humans) elicits intracellular calcium accumulation and phosphodiesterase inhibition.Notably, the depicted receptor interactions are simplified.For example, the presence of adenosine A 2 R: dopamine D 2 R heterotetramers consisting of an adenosine A 2 R and a dopamine D 2 R homodimer has been suggested.In these heterotetramers, the Gα olf and Gα i/o proteins couple to their respective adenosine A 2 R and dopamine D 2 R homodimers, respectively (Bonaventura et al., 2015).Besides competitive inhibition of adenosine-binding via caffeine, there are also transmembrane antagonistic allosteric interactions within adenosine A 1 R:dopamine D 1 R and adenosine A 2 R:dopamine D 2 R heteromers that influence signalling (not shown), such as by switching G proteins.For more information regarding these complex receptor interactions, please see Borroto-Escuela and Fuxe (2019) and Ferre (2016).ATP = adenosine triphosphate, ADP = adenosine diphosphate, AMP = adenosine monophosphate, D 1/2 R = dopamine 1/2 receptor, A 1/2A R = adenosine 1/2A receptor, cAMP = cyclic adenosine 3′,5′-monophosphate, CaVs = voltage-gated calcium (Ca 2+ ) channels, PLC = phospholipase C, PKA = protein kinase A, DARPP-32 = dopamine and cAMP regulated phosphoprotein 32, AMPAR = α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor, MAPK = mitogen-activated protein kinase, CREB = cAMP response element-binding protein.
N. Reich et al.Mechanistically, positive affect and motivation are regulated by the dopamine system.Activation of the shell area of the nucleus accumbens enhances the enjoyment derived from a reward or activity ('liking'), whereas stimulation of the core region drives cravings and rewardobtaining behaviours ('wanting') (Berridge et al., 2009;Salamone et al., 2007).More specifically, dopamine receptor-signalling in specific neuronal subpopulations in the striatum play different reward/motivation-associated roles.Dopamine D 1 Rs are exhibited by striato-nigral spiny projection neurons (SPNs), while dopamine D 2 Rs are present on the cellular body of cholinergic interneurons (10%), somatodendritic and axonal areas of gamma aminobutyric acid (GABA)ergic striato-pallidal SPNs (50%) and afferent synaptic terminals of glutamatergic and dopaminergic cortical neurons (20%) (see Simpson et al., 2022).Induction of dopamine D 1 Rs was shown to mediate feelings of reward, whilst stimulation of dopamine D 2 Rs opposes effort aversion (Hikida et al., 2013;Ledonne and Mercuri, 2017).Particularly, deletion of dopamine D 2 Rs on striatal SPNs lowered action-initiation and willingness to engage in reward-seeking (Augustin et al., 2020).Inversely, virus-mediated upregulation of dopamine D 2 Rs on striato-pallidal SPNs in the nucleus accumbens was demonstrated to disinhibit ventral pallidum-located neurons, thus encouraging goal-directed motivation and willingness to invest effort to obtain a reward, but without influencing the reward value (reviewed in Simpson et al., 2022) (Carvalho Poyraz et al., 2016;Gallo et al., 2018;Trifilieff et al., 2013).
In the context of flow, it has been debated what role the nucleus accumbens core and shell play (van der Linden et al., 2021a).Activity in the nucleus accumbens shell ('liking') might mediate feelings of reward, satisfaction and positive affect in response to a flow experience, as induced by dopamine release during or after the activity.To initiate and maintain flow, however, the nucleus accumbens core ('wanting') is likely heavily engaged to generate drive and persistence in challenging tasks.The wanting aspect of flow is critical, given that flow seems to be more strongly associated with challenge and functional performance than experiencing positive emotions (Flett, 2015).Moreover, a well-adjusted challenge skills balance is considered to be one of the most important flow triggers (Kotler et al., 2022).
Congruent with its dopaminergic effects, caffeine is well known to promote positive mood, including wakefulness, vigor, clear-headedness and pleasure (Leathwood and Pollet, 1982;Solinas et al., 2002;Volkow et al., 2015;Warburton, 1995).Caffeine-induced positive affect, in turn, theoretically facilitates flow states (Eisenberger et al., 2005;van der Linden et al., 2021a).Notably, cognitive control (i.e.adaptability to tasks and management of reward stimuli) and, as such, better performance appear to be more important for inducing flow than the presence of positive emotions (Cole et al., 2013;Flett, 2015;Holmes et al., 2016;Huskey et al., 2022).It is not clear if positive affect is a requirement or by-product of the flow experience, however (van der Linden et al., 2021a).At the very least, positive emotions prevent the emergence of negative ones, such as depression, which are known to negatively affect performance and flow (Peifer et al., 2022).
Interestingly, a recent study showed that consuming caffeinated, but not decaffeinated, coffee prior to shopping increased impulsivity, as indicated by the higher number of 'high-hedonic', but not 'low-hedonic', items purchased (Biswas et al., 2023).This effect occurred despite the fact that the majority of participants (63.37%) believed that caffeine would not influence their spending.Furthermore, higher levels of energy, such as following caffeine intake, were demonstrated to reduce effort-aversion of consumers (Gibbs and Drolet, 2003).This resulted in a greater propensity to choose products that are more challenging to consume (i.e.movies with foreign language and subtitles instead of native language).Thus, the hedonic effects of caffeine do not seem to be placebo or belief-related, but are biological, and likely the consequence of upregulating dopamine DR availability in the nucleus accumbens (ventral striatum; Fig. 2B) (Volkow et al., 2015).
Since caffeine encouraged effort expenditure without increasing enjoyment of the consumed product (Gibbs and Drolet, 2003), this suggests that caffeine (174 mg in this study) did not appear to affect dopamine D 1 Rs, or the shell of the nucleus accumbens ('liking') (Berridge et al., 2009;Salamone et al., 2007).In support of this suggestion, it was shown that only acute, but not chronic, caffeine ingestion evokes glutamate and dopamine release in the nucleus accumbens shell, and that this effect is mediated by presynaptic dopamine D 1 Rs (see also Ferre, 2016) (Quarta et al., 2004).The latter receptor desensitizes in response to regular caffeine intake due to upregulation of adenosine A 1 Rs, however (Fig. 2C).Nevertheless, following substance withdrawal and re-sensitisation, lower doses of acute caffeine presumably boost the 'liking' aspect again.In fact, after re-sensitisation, caffeine might be more effective in facilitating flow because of the additional dopamine D 1 R-mediated dopamine release in the nucleus accumbens shell (Quarta et al., 2004;Solinas et al., 2002).Instead, when consumed regularly, caffeine likely promotes engagement in flow-trigger hedonic activities such as video gaming (Michailidis et al., 2018), by antagonizing adenosine A 2A Rs on SPNs in the nucleus accumbens (Huang et al., 2005;Lazarus et al., 2011;Simpson et al., 2022), including the core region responsible for 'wanting' and willingness to expend effort for rewards (Berridge et al., 2009;Salamone et al., 2007).Unlike that of adenosine A 1 Rs, striatal adenosine A 2A R expression does not increase, and might even slightly decrease, in response to chronic caffeine consumption (Johansson et al., 1993;Shi et al., 1993;Svenningsson et al., 1999;Tronci et al., 2006), indicating that the substance continues to enhance dopamine D 2 R-signalling (Fig. 2C).Therefore, caffeine seems to mainly influence striatal adenosine A 2A R:dopamine D 2 R heteromers, the 'wanting' aspect and effort aversion, if consumed regularly (Berridge et al., 2009;Gibbs and Drolet, 2003;Rossi et al., 2010;Salamone et al., 2007).
Notably, sleep deprivation for one day was shown to reduce dopamine D 2 R levels in the ventral striatum (nucleus accumbens) of humans, which, besides adenosine accumulation (Figs. 1 and 2A), contributes to the reduction in wakefulness (Volkow et al., 2012).Unsurprisingly, lack of sleep will likely impair flow by reducing dopamine D 2 R-associated motivation, activity engagement and performance; all adverse effects that can be ameliorated with a sufficiently high dose of caffeine (Fig. 2B) (McLellan et al., 2016).
Collectively, by antagonizing striatal dopamine D 2 Rs, caffeine enhances energy and motivation to invest effort into rewarding activities, both involving the endocannabinoid system (Section 4.3).and creating engagement and flow.(In agreement with these suggestions, participants ingesting 25 -200 mg caffeine showed greater energetic arousal, which is characterized by positive affect, excitement, clear-mindedness and lowered anxiety (Peeling and Dawson, 2007;P. Quinlan et al., 1997; P. T. Quinlan et al., 2000;Smit et al., 2004;Smit et al., 2006;Smit and Rogers, 2000).Caffeinated energy drinks also allowed study participants to maintain positive mood and performance during tiring and cognitively challenging tasks (Smit et al., 2004), whilst caffeine is a frequently used additive in analgesics due to its pain-relieving effects (especially ≥ 300 mg) (Myers et al., 1997;Temple et al., 2017).These properties of caffeine further support that the substance may prolong flow in strenuous cognitive or physical activities (Kotler et al., 2022).

Impact on negative affect
Previous studies have revealed that negative affect is an inverse predictor of flow (Peifer et al., 2022).In this context, neuroimaging has shown that flow exhibits inverted U-shape patterns in the DMN and amygdala, whose activities are lower during flow compared to 'boredom' and 'overload' conditions (Ulrich et al., 2016;Ulrich et al., 2014).It has been proposed that negative emotions stimulate flow-blocking DMN and amygdala activity, as reflected psychologically by heightened worrying and/or anxiety (van der Linden et al., 2021a).Therefore, the reduction of negative affect likely enhances flow proneness.
Interestingly, consistent caffeine intake was shown to increase hippocampal dopamine and serotonin levels, while reducing anxiety and N. Reich et al. depressive symptoms in chronically-stressed mice (Pechlivanova et al., 2012).In addition, adenosine A 2 R antagonists similar to caffeine have shown anti-depressive effects in vivo (El Yacoubi et al., 2001).In view of these qualities, caffeine has recently been reviewed for anti-anergic, motivation-boosting effects that may offset dopamine dysregulation in animal models and humans with depression (Lopez-Cruz et al., 2018).Drinking several cups of coffee per day is associated with reduced depression and suicidal thoughts in humans, although excessive caffeine can encourage psychosis and anxiety (Lara, 2010;Lucas et al., 2011;Lucas et al., 2014).Thus, regular caffeine ingestion, if not exorbitant, potentially facilitates flow by attenuating negative emotions.

Protection of dopaminergic neurons
As discussed, flow experiences involve activity in the dorsal striatum (putamen and caudate nucleus) (Ulrich et al., 2016;Ulrich et al., 2014).Furthermore, flow proneness is positively correlated with a genetically higher dopamine D 2 R density in the putamen and striatum (de Manzano et al., 2013;Gyurkovics et al., 2016).Interestingly, a cross-sectional study investigating healthy Japanese participants (mean age = 50.61)reported that grey matter volume in the right caudate mildly correlates with flow proneness (Niksirat et al., 2019).Inversely, PD patients exhibit degeneration of dopamine-producing neurons in the SN display low striatal dopamine levels, desensitization to reward, apathy and depression (Mele et al., 2020).These are all pathological features that oppose the high levels of intrinsic motivation and positive affect found in flow (Gold and Ciorciari, 2020;van der Linden et al., 2021a).This suggests that interventions to prevent age-associated neurodegeneration in the striatum may preserve the ability to enter flow states.
Additionally, caffeine may exert neuroprotective effects.In animal models of Alzheimer's disease (AD), caffeine was demonstrated to reduce the levels of soluble and aggregated amyloid beta (Aβ) 1-42 peptides, hyperphosphorylated Tau species, oxidative and endoplasmic reticulum stress, neuronal death and pro-apoptotic caspase 3 activity in the hippocampus, while improving mitochondrial function and memory (Chu et al., 2012;Dragicevic et al., 2012;Prasanthi et al., 2010).In a mouse model of PD, in an adenosine A 2 R-dependent manner, chronic caffeine administration was shown to decrease atrophy of dopaminergic neurons in the SN, a motor function-regulating brain region adjacent to the reward-associated VTA (Klein et al., 2019), and loss of striatal dopamine (J.F. Chen et al., 2001;Kachroo et al., 2010;Kalda et al., 2006;Sonsalla et al., 2012;Xu et al., 2010).Protective benefits have also been found when consumption began after neurotoxin injection (Sonsalla et al., 2012).Moreover, caffeine prevents BBB disruption in AD and PD animal models (see X. Chen et al., 2010).
These neuroprotective effects of the substance, as observed in vivo, seem to translate into humans.Regular caffeine consumption has been shown to improve motor symptoms in PD and lower cognitive decline in AD/dementia, whilst reducing the risk of developing neurodegenerative diseases (J.F. Chen and Chern, 2011;Eskelinen and Kivipelto, 2010;Eskelinen et al., 2009;Maia and de Mendonca, 2002;Palacios et al., 2012;Postuma et al., 2012).Considering the importance of dopamine circuits for flow (D.J. Harris et al., 2017b), protection of dopaminergic neurons by chronic caffeine consumption may preserve the ability to enter flow states in light of age-associated neurodegeneration.

Sympathetic and parasympathetic activity 4.2.1. Adrenergic and hypothalamic-pituitary-adrenal axis activation
The locus coeruleus-norepinephrine system regulates arousal and, under conditions of reward, maintains task focus (see van der Linden et al., 2021b).However, if a task is considered boring or overwhelming, concomitant changes in NE levels induce task disengagement and distraction.The sweet spot for ideal task engagement and, consequentially, flow is a moderate baseline of NE release (Fig. 3A).
Sympathetic arousal in response to stress is driven by the rapid release (a few minutes) of epinephrine (80%; also known as adrenaline) and NE (20%) by the adrenal glands to enhance action readiness.These noradrenergic effects include increases in heart rate, blood flow and Fig. 3. Impact of caffeine on sympathetic and parasympathetic activity, performance and flow.(A) Alertness, as dependent on the current circulatory (nor)epinephrine and cortisol levels, shows an inverted U-shape relationship with both performance and flow.Medium arousal is optimal, whereas both low and excessive sympathetic activity result in an adversely altered emotional state and impaired performance/flow.(B) Impact of caffeine on LF and HF components of HRV.Flow is linearly and positively correlated with HF HRV, which reflects parasympathetic activity.In non-regular caffeine users, ingestion of the substance elevates LF, a measure of both sympathetic and parasympathetic activities.When consumed by a regular user, however, caffeine (> 240 mg) transiently enhances HF HRV for up to 2 h, which may promote flow in this time period.If ingested post exercise, high caffeine doses (6 mg/kg) also enhance recovery of the exercise-associated reduction in parasympathetic activity (RMSSD; beat-to-beat variance of the heart) in athletes, but the substance has the opposite effect in untrained individuals (not shown).LF = low frequency, HF = High frequency, HRV = heart rate variability, RMSSD = root mean square of successive RR interval differences.Modelled after (van der Linden et al., 2021b) and HRV spectrum example from (Litscher et al., 2009).systolic blood pressure, improved eye sight and enhanced glucose and fatty acid metabolization in peripheral and cerebral tissue (Hinds and Sanchez, 2022;MacKenzie et al., 1976).Blood-borne (nor)epinephrine does not cross the BBB, however it may communicate with the central nervous system (CNS) via β-adrenoreceptors on efferent vagal fibers (Kostrzewa, 2007;Wong et al., 2012).Instead, NE-producing neurons in the locus coeruleus and nucleus tractus solitarius supply the CNS with this stress hormone (Wong et al., 2012).
Additionally, prolonged stress leads to hypothalamic-pituitaryadrenal (HPA) axis activation, evoking the circulatory release of the BBB-penetrant cortisol (~15 -20 min) to promote gluconeogenesis and adjust gene expression in the periphery and brain (Hinds and Sanchez, 2022).Notably, hypocortisolaemia, e.g. in Addison's disease, encourages symptoms of chronic fatigue syndrome (Tomas et al., 2013), suggesting that too low cortisol levels are not optimal for performance.Inversely, excessive cortisol levels deteriorate performance by inducing nervousness and anxiety (Hinds and Sanchez, 2022).As such, medium levels of arousal with optimal (nor)epinephrine and cortisol levels maximize performance, while boredom (low alertness) and anxiety (excessive arousal) are detrimental to performance (Peifer, Schulz, Schachinger et al., 2014;Yerkes and Dodson, 1908).
A similar relationship exists in flow, where the same u-shaped relationship between sympathetic arousal and performance characterized by the Yerkes-Dodson Law (Yerkes and Dodson, 1908) is found.For example, (Peifer, Schulz, Schachinger et al., 2014) subjected twenty-two male subjects to the Trier Social Stress Test, which involves a high stress job interview.Afterwards, subjects perform a complex computer task and rate their levels of absorption and fluency on the Flow Short Scale.In this experimental paradigm, moderate, but not low or high, levels of the HPA-related stress hormone cortisol resulted in more flow (Fig. 3A).
Caffeine is an alertness-enhancing somnolytic that is frequently used to treat drowsiness (Temple et al., 2017).In the periphery, the ingestion of typical caffeine doses lead to transient activation of the sympathetic nervous system, the subsequent adrenal secretion of catecholamines (NE/epinephrine and renin), a mild increase in both systolic and diastolic blood pressure, reduced heart rate and vasoconstriction (summarized in (Temple et al., 2017)).Caffeine evokes sympathetic catecholamine secretion in a dose-dependent manner (Papadelis et al., 2003), whilst higher caffeine doses (≥ ~250 mg) also stimulate the HPA axis and, thus, plasma cortisol release (Gavrieli et al., 2011;Lovallo et al., 2005).These adrenergic effects and cortisol secretion desensitize over time in habitual caffeine consumers, however (Corti et al., 2002;Lovallo et al., 2005).
Caffeine-induced sympathetic and HPA axis activation synergize with that induced by other sources of mental or physical stress (including at work and home), resulting in greater blood pressure and circulatory (nor)epinephrine and cortisol levels that exceed those elicited by caffeine (i.e.≥ 4.45 mg/kg / ~325 mg) or stress alone (Graham et al., 1998;Lane et al., 2002;Lovallo et al., 2006;Papadelis et al., 2003).In vivo studies specify that high, but not low or moderate, caffeine doses are necessary to see this additive effect of the substance and stress on HPA axis stimulation (Patz et al., 2006).In agreement with a dose-dependent effect, it has been suggested that the cardiovascular effects of caffeine are partially mediated by intracellular Ca 2+ accumulation (Temple et al., 2017), which would necessitate high caffeine doses close to the toxic threshold in humans (400 mg) (Fiani et al., 2021).
Notably, caffeine-released plasma NE does not cross the BBB (Kostrzewa, 2007;Wong et al., 2012).However, animal studies indicate a modulatory effect of caffeine or its metabolites on catecholamine levels in the brain.Injection of a methylxanthine derivate was shown to stimulate firing of noradrenergic neurons in the locus coeruleus (Grant and Redmond, 1982).Although acute caffeine or methylxanthine administrations did not affect cerebral NE levels, the synthesis, release and immediate turnover (tyrosine → dopamine → NE → normetanephrine) of NE were shown to be enhanced (Berkowitz et al., 1970;Corrodi et al., 1972;Karasawa et al., 1976;Schlosberg et al., 1981;Waldeck, 1971Waldeck, , 1975)).While the necessary caffeine dose to achieve this adrenergic effect in the human brain is not known, it was demonstrated that drinking two cups of coffee (1.5 mg/kg / ~225 mg caffeine each) reversed clonidine-induced reductions in central NE, alertness and cognitive performance (Smith et al., 2003).
By contrast, if given chronically, caffeine or methylxanthine administrations were shown to increase NE and epinephrine levels in the whole brain and frontal cortex in vivo (M.D. Chen et al., 1994;Kirch et al., 1990).Notably, both acute and chronic caffeine injections were demonstrated to enhance glucose metabolism in the locus coeruleus, suggesting that the stimulatory effects of caffeine on local NE neurons do not desensitize (Nehlig et al., 1986).However, because higher caffeine doses were demonstrated to lower β-adrenoreceptor levels in various brain areas (e.g.forebrain) 3 -4 h post injection in vivo (Goldberg et al., 1982;Lowenstein et al., 1982), there might be a transient reduction in NE receptor relay and, as such, alertness following caffeine consumption.Collectively, acute and chronic caffeine intake stimulate NE transmission from the locus coeruleus across the CNS, thus increasing sympathetic activation and elevating alertness, attention and wakefulness (Fig. 3A) (McLellan et al., 2016;Ranjbar-Slamloo and Fazlali, 2019).
In summary, high acute caffeine intake (≥ ~250 mg) synergizes with other stressors to elicit the plasma release of NE and cortisol via sympathetic and HPA axis activation (periphery), whilst elevating NE synthesis, release and turnover (CNS).Caffeine-induced sympathetic and HPA responses weaken following chronic use, however.Given that medium, but not low or excessive, alertness (NE) levels result in optimal performance and flow (Peifer, Schulz, Schächinger et al., 2014;van der Linden et al., 2021b;Yerkes and Dodson, 1908), caffeine is presumably the most effective for facilitating flow when under-aroused or sleep-deprived, but counterproductive when agitated (Fig. 3A) (McLellan et al., 2016).

Heart rate variability
A recent meta-analysis suggests that baseline heart rate variability (HRV) predicts the cardiac adaptability to stress (exercise, mental stress etc).Moreover, high basal HRV is linked to lowered cortical threat perception (involving e.g. the ventral PFC) (H.G. Kim et al., 2018).This greater stress tolerance imparted by high basal HRV levels likely facilitates self-regulation in stressful situations and challenging activities capable of inducing flow (H.G. Kim et al., 2018;Kotler et al., 2022;Thayer and Lane, 2000).Inversely, low tonic HRV levels are associated with anxiety disorders (Chalmers et al., 2014).
Interestingly, in a study by Peifer et al. (2014) (see previous section), the high frequency (HF) component of HRV (0.15 -0.4 Hz), which gauges parasympathetic vagal nerve activity, linearly correlated with flow.By contrast, low frequency (LF) HRV (0.04 -0.15 Hz), which measures both sympathetic and parasympathetic responses, showed an inverted U-shape relationship, with moderate LF optimising flow.This is in agreement with the fact that moderate cortisol levels, another marker for sympathetic (HPA axis) activation and arousal, were also associated with flow (Fig. 3A).Notably, the parasympathetic nervous system typically induces relaxation by quenching the activity of its sympathetic counterpart, although both systems may interact to finetune stress responses (Berntson et al., 1991;Porges, 1995;Thayer and Lane, 2000).This suggests that flow demands an appropriate level of sympathetic stress triggered by the task at hand, neither excessive nor too weak, in order to effectively engage into an activity (van der Linden et al., 2021b).Simultaneously, parasympathetic co-activation likely contributes to the characteristic 'sense of control' during flow (Kotler et al., 2022), and should presumably be as high as possible to blunt anxiety, overwhelm and cortical threat perception (H.G. Kim et al., 2018) when immersed into a challenging task.
Notably, in comparison to a 'boredom' condition, overall reduced HRV (root mean square of successive differences (RMSSD)); variability N. Reich et al. in the interval between two successive heart beats) and heightened stress indicated by increased salivary cortisol were shown for the skillmatched flow condition in a computer task (Keller et al., 2011).In this context, an acute reduction in HRV/RMSSD, which reflects reduced parasympathetic activity, is associated with increased vigilance and sustained attention, and symbolic of heightened mental workload (e.g. during busy periods at work) (Hjortskov et al., 2004;Jorna, 1992;Sloan et al., 1994;Thayer and Lane, 2000).Comparatively, sympathetic activity also seems to exceed that of its parasympathetic counterpart in flow, as indicated by a higher low frequency (LF) / HF HRV ratio during piano playing (flow was assessed via questionnaires (de Manzano et al., 2010)), or a reported reduction in the parasympathetic HF component in the 'matched' flow condition in a simulated car-racing game (D.J. Harris et al., 2017a).This suggests that the RMSSD and HF component of HRV decrease upon engagement in flow-inducing tasks, consequentially allowing the sympathetic nervous system activity to optimise attention and focus (van der Linden et al., 2021b).However, during the activity, maintenance of a greater parasympathetic HF HRV appears to balance sympathetic activation and facilitate flow (Peifer, Schulz, Schachinger et al., 2014).
Despite the fact that caffeine is a well-known stimulant and increased sympathetic activity would be expected (Cappelletti et al., 2015), caffeine promotes parasympathetic HF HRV after ingestion (reviewed in Koenig et al., 2013); (Fig. 3B).The effects of caffeine on HRV are context-dependent, however, and caffeine conditions that may promote flow are discussed below.
In non-habitual caffeine users, low doses of oral caffeine (≤ 200 mg) were shown to transiently increase the LF/HF HRV ratio, reflecting elevated sympathetic activity (Fig. 3B) (Rauh et al., 2006;Sondermeijer et al., 2002).The ingestion of 400 mg caffeine 30 min before sleep also increased the LF/HF HRV ratio during REM sleep, implying that caffeine can disrupt sleep and should be avoided at night (Bonnet et al., 2005).
On the other hand, higher caffeine doses (> 240 mg) were shown to promote parasympathetic HF HRV for up to 120 min post consumption (Fig. 3B) (Hibino et al., 1997;Zimmermann-Viehoff et al., 2016).Lower substance doses (i.e.75 mg) also seem to boost HF HRV following deliberate caffeine withdrawal and/or when the individual displays low sympathetic activity (i.e. at rest in a supine position) (Monda et al., 2009).Therefore, if caffeine is consumed regularly, strategic consumption of higher substance doses prior to task engagement might enhance subsequent parasympathetic activity and, consequentially, flow (de Manzano et al., 2010;Peifer, Schulz, Schachinger et al., 2014).
Notably, the current evidence suggests that consistent, regular caffeine intake (≥ 3 cups; ca.240 -360 mg caffeine) does not influence vagal tone, suggesting that caffeine can only transiently boost parasympathetic activity and flow (de Oliveira et al., 2017;Vansickle et al., 2020).
Interestingly, there are context-dependent effects of caffeine on HRV surrounding physical activity.Caffeine doses at or above 300 mg enhance physical readiness by increasing HF HRV before exercise, then both augment the exercise-induced drop in parasympathetic HF and increase in sympathetic LF HRV during exercise (Nishijima et al., 2002;Yeragani et al., 2005).Considering the performance-enhancing effects of ≥ 200 mg caffeine (McLellan et al., 2016), caffeine seems to improve HRV regulation and physical capability, thus likely resulting in more flow.
After physical activity, when parasympathetic HRV (e.g.measured as RMSSD) is typically transiently reduced, lower caffeine concentrations of < 3 mg/kg (~225 mg) had no effect (An et al., 2014;Kliszczewicz et al., 2018).However, a high dose of 6 mg/kg (corresponding to ~450 mg) caffeine accelerated normalisation of RMSSD post exercise compared to placebo and/or decaffeinated supplements, indicating that caffeine could shorten parasympathetic recovery after a flow state (Sarshin et al., 2020).By contrast, in untrained individuals, 300 -400 mg caffeine prolonged RMSSD rebound following exercise (Bunsawat et al., 2015;Gonzaga et al., 2019;Gonzaga et al., 2017).Notably, RMSSD is strongly correlated with HF HRV, and the latter is increased by caffeine consumption (see Shaffer and Ginsberg, 2017 for an overview of HRV measurements) (Bigger et al., 1989;Koenig et al., 2013).This suggests that caffeine is beneficial for parasympathetic recovery following physical activities which could help increase flow in athletes, but not non-athletes.

Endocannabinoid system
Retrograde endocannabinoid-signalling from the postsynaptic site to presynaptic cannabinoid receptors (CBs) finetunes the synaptic release of various neurotransmitters, while modulating anxiety and stress responses (Freund et al., 2003;Hillard, 2018).Generally, stress responses are bidirectionally driven by the levels of the endocannabinoids anandamide (AEA) and 2-arachidonoylglycerol (2-AG) (Morena et al., 2016).Acute stress triggers the fatty acid amide hydrolase (FAAH)-mediated clearance of AEA in the postsynaptic space, leading to reduced presynaptic CB1 receptor (CB1R) activation, disinhibition of basal lateral amygdala neurons, concomitant stimulation of cortisol release by the HPA axis, anxiety and, possibly, anhedonia.A temporally delayed and cortisol-induced increase in 2-AG levels in the medial PFC and paraventricular nucleus of hypothalamus then quenches and, eventually, terminates the stress response (Hill et al., 2010;Morena et al., 2016).During flow, the endocannabinoid system might act as a master regulator of the locus coeruleus-norepinephrine system that coordinates alertness and stress responses (Kotler et al., 2022).
Interestingly, the consumption of either 4 or 8 cups of coffee for a month each resulted in a reduction in four serum endocannabinoids, such as palmitoyl ethanolamide, whose accumulation competitively prevents AEA degradation (Bisogno et al., 1997;Clayton et al., 2021;Cornelis et al., 2018).Given that caffeine is an anxiogenic stimulant (Cappelletti et al., 2015), it is plausible that chronic intake of the substance lowers cerebral AEA levels similar to acute and chronic stress (Morena et al., 2016).Notably, inhibition of AEA-signalling in the basolateral amygdala leads to heightened activity in the locus coeruleus (Bedse et al., 2014), where NE is synthesised (Wong et al., 2012).Since chronic caffeine administration was shown to elevate NE release in the brain and frontal cortex in vivo (M.D. Chen et al., 1994;Kirch et al., 1990), caffeine-mediated downregulation of endocannabinoids might be involved in increasing activity in alertness and anxiety-associated brain networks, as seen in regular caffeine drinkers (Magalhaes et al., 2021).The impact of caffeine on cerebral AEA and 2-AG levels has not been investigated, however (Cornelis et al., 2018), and further studies are necessary to investigate how endocannabinoid levels in different brain areas are altered by caffeine.
Interestingly, there is evidence that caffeine modulates endocannabinoid-signalling in the brain, with implications on stress tolerance (Fig. 4).In the striatum, two types of principal GABAergic neurons exist, including dopamine D 1 R-expressing striato-nigral and dopamine D 2 R-exhibiting striato-pallidal SPNs (Simpson et al., 2022).Activation of striatal neurons is controlled by incoming glutamatergic projections (cortical origin) and dopaminergic terminals derived from the VTA and SNpc.The direct striato-nigral projection pathway involves striatal dopamine D 1 R-expressing SPNs that project to the internal globus pallidus (inhibition) and thalamus (excitation).The indirect striato-pallidal pathway is regulated by dopamine D 2 R-positive SPNs, involving projection across the external globus pallidus, subthalamic nucleus, internal globus pallidus and thalamus.Both pathways stimulate locomotion (Simpson et al., 2022;Wang et al., 2022).Additionally, relevant for the challenge aspect of flow, activation of dopamine D 1 Rs and dopamine D 2 Rs in the nucleus accumbens induces feelings of 'liking' and 'wanting', respectively (Section 4.1.2).
Besides dopamine D 1/2 Rs, other receptors located on somatodendrites or axonal boutons navigate neuronal activity across the direct and indirect projection pathways (region/synapse-specific receptor expression illustrated in (Wang et al., 2022)).This includes CB1Rs, which are presynaptically expressed by, for example, glutamatergic cortico-striatal neurons (Fig. 4A), or GABAergic striato-nigral and striato-pallidal SPNs (Fig. 4B for the latter) projecting to the internal or external globus pallidus, respectively.Interestingly, CB1Rs are capable of forming heteromers with adenosine A 2A Rs and dopamine D 2 Rs (Carriba et al., 2007;Marcellino et al., 2008).Globally, by activating CB1R, a pharmacological increase in AEA levels was shown to prevent the motor-stimulating effects of a dopamine D 2 R agonist, and reduce hyperactivity in an ADHD rodent model (Beltramo et al., 2000).Other studies confirm that striatal CB1 receptor activation antagonizes both dopamine D 1 R-and D 2 R-induced behaviours (see Wang et al., 2022 for a cannabinoid-focused review) (A.B. Martin et al., 2008).Some studies suggest that the motor-or synaptic transmission-depressing effects of CB1R agonists in the striatum were dependent on postsynaptic, but not presynaptic, adenosine A 2A Rs (Fig. 4A) (Carriba et al., 2007;Tebano et al., 2009).In cortico-striatal terminals, antagonism of adenosine A 2A R blocked the effects of metabotrophic glutamate receptor type 5 (mGlu5), whereas stimulation of mGlu5 further potentiated the synaptic Fig. 4. Influence of chronic caffeine consumption on striatal endocannabinoid-signalling.Chronic caffeine consumption leads to the upregulation of adenosine A 1 Rs, whereas adenosine A 2A R expression appears to be unaltered or downregulated.As such, only the latter receptors are modulated by regular substance intake.Functionally, presynaptic CB1Rs in the striatum inhibit neurotransmitter release, while administration of CB1 antagonists inhibits motor activity induced by dopamine D 1 R or D 2 R agonists.(A) In cortico-striatal glutamatergic neurons, CB1Rs inhibits adenosine A 2A R-signalling and glutamate release, when activated.CB1Rs further form presynaptic heteromers with adenosine A 2A Rs, which likely serve as an opposing switch.In this context, presynaptic adenosine A 2A R activation counteracts that of CB1Rs due to cAMP/PKA-mediated glutamate vesicle synthesis.On the postsynaptic level, activation of mGlu5 suppresses glutamate release by stimulating postsynaptic Gα q/11 -mediated endocannabinoid synthesis, followed by retrograde activation of presynaptic CB1Rs by AEA.Co-activation of postsynaptic adenosine A 2A Rs is necessary for the mGlu5-driven inhibition of CB1Rs.Notably, due to changes in adenosine A 1 R:A 2A R heteromers, chronic caffeine stimulates glutamate release (not shown; see Fig. 2).(B) In striato-pallidal GABAergic spiny projection neurons, the caffeine-mediated inhibition of presynaptic adenosine A 2A Rs facilitates CB1R-induced blockage of GABA release.Interestingly, chronic caffeine ingestion also sensitises CB1Rs to endocannabinoids.The underlying mechanism remains to be investigated, but could be a consequence of decreased adenosine A 2A R density in response to regular caffeine consumption.Postsynaptically, reduced GABA release disinhibits ventral pallidum neurons, resulting in enhanced motivation to work for rewards (as mediated by dendritic dopamine D 2 R activation on striato-pallidal SPNs; Fig. 2).Higher activity of ventral pallidum neurons, leading to increased postsynaptic Ca 2+ instream, also promotes the production and retrograde diffusion of AEA, thus strengthening presynaptic CB1R activation.Acute stress leads to the postsynaptic degradation of AEA by FAAH (not shown), indicating that chronic caffeine consumption may preserve CB1R activation.Moreover, chronic caffeine intake counteracts the downregulation of CB1Rs during chronic stress.This suggests that regular ingestion of the substance may maintain motivation in stressful or overwhelming situations, which likely facilitates activity engagement and flow.Notably, striato-nigral SPNs only express dopamine D 1 Rs and adenosine A 1 Rs, but not D 2 Rs and A 2A Rs, indicating that these neurons are not affected by chronic caffeine ingestion (Kreitzer, 2009).AEA = anandamide, A 2A R = adenosine 2A receptor, cAMP = cyclic adenosine 3′,5′-monophosphate, CaVs = voltage-gated calcium (Ca 2+ ) channels, CB1 receptor = cannabinoid receptor type 1, D 2 R = dopamine 2 receptor, KVs = voltage-gated potassium (K + ) channels, GABA = gamma aminobutyric acid, GABAR = gamma aminobutyric acid receptor, mGlu5 = metabotrophic glutamate receptor type 5, NAT = N-acyltransferase, NAPE = N-arachidonoyl phosphatidylethanolamine, PE = phosphatidylethanolamine, PKA = protein kinase A, PLD = phospholipase D. Modified from (Morena et al., 2016).
transmission-inhibiting effects of presynaptic CB1Rs (Tebano et al., 2009).In this context, postsynaptic mGlu(5)/Gα q/11 -signalling promotes induction of presynaptic CB1Rs by increasing endocannabinoid generation and retrograde release (Fig. 4A) (Xiang et al., 2019).Considering that an upregulation of striatal CB1Rs is found in Parkinson's disease, and that CB1R antagonists have been tested as treatments for this motor disease, this collectively suggests an anti-dopamine function of CB1Rs in the striatum (Wang et al., 2022).However, these effects are complex and both neuron (glutamatergic vs. GABAergic)-and region-specific.In fact, as discussed further below, CB1Rs might play a facilitatory role in specific synapses.
Importantly, caffeine appears to affect presynaptic CB1Rs, and thus neurotransmitter release.In cortico-striatal terminals of glutamatergic neurons, induction of CB1Rs inhibits glutamate release, while additional activation of presynaptic adenosine A 2A Rs counteracts this effect (Fig. 4A) (Martire et al., 2011).Similarly, in GABAergic neurons, inhibition of striatal GABA transmission following dopamine D 2 R agonism was partially mediated by CB1Rs (Fig. 4B) (Centonze et al., 2004).Moreover, striatal activation of dopamine D 2 Rs (likely on dendrites of striato-pallidal SPNs) was shown to stimulate synthesis, and inhibit degradation, of AEA.In agreement with Centonze et al. (2004), an in vitro study using GABAergic striatal neurons suggests that activation of CB1Rs by AEA or other cannabinoids decreases GABA release (Laprairie et al., 2013).Interestingly, an in vivo study has shown that chronic caffeine intake sensitizes CB1Rs towards endocannabinoids on GABAergic, but not glutamatergic, striatal neurons in a reversible manner (Fig. 4B) (Rossi et al., 2009).In terms of the resulting effect, the dopamine D 2 R-mediated reduction in GABA output from accumbal (striato-pallidal) SPNs to the external globus pallidus (indirect projection pathway), as well as towards SPNs of the direct pathway, were linked to increased motivation of mice to work for food rewards (Gallo et al., 2018).This suggests that regular caffeine intake selectively facilitates presynaptic CB1R-signalling in these striato-pallidal synapses, while CB1Rs amplify GABA inhibition and motivation evoked by accumbal dopamine D 2 R activation.
The basis of this facilitatory effect of chronic caffeine on CB1Rs in GABAergic striato-pallidal SPNs is not clear.When ingested chronically, the motor effects of caffeine desensitize due to an upregulation, or weaker activation, of adenosine A 1 Rs, whereas striatal A 2A Rs are not affected (Ferre, 2008).As such, regular caffeine intake predominantly affects adenosine A 2A Rs, and by extension dopamine D 2 Rs, on GABAergic striato-pallidal SPNs (Rossi et al., 2010;Simpson et al., 2022).Regarding CB1R sensitivity, caffeine could act on presynaptic adenosine A 2A Rs that heteromerise with, and thus inhibit, CB1Rs (Carriba et al., 2007;Martire et al., 2011).It is not known how chronic, but not acute, caffeine intake elicits a CB1R-sensitising effect.Because a reduction in striatal adenosine A 2A R receptors has been reported in response to (sub-)chronic ingestion of caffeine (Svenningsson et al., 1999;Tronci et al., 2006), such a downregulation of presynaptic adenosine A 2A Rs could, hypothetically, disinhibit CB1Rs.However, possibly due to the influence of other receptors, it is dubious that CB1R sensitivity is specifically improved by chronic caffeine in GABAergic, but not in glutamatergic cortico-striatal terminals.
In the context of motivation, dopamine D 1 R (adenosine A 1 R)-associated feelings of pleasure ('liking') are lost in response to chronic caffeine ingestion, but the dopamine D 2 R (adenosine A 2 R)-related impact on motivation ('wanting') persists (Section 4.1.2).Globally, AEA, or striatal CB1R activation, inhibits motor activity stimulated by dopamine D 1/2 R agonists, but facilitatory effects on dopamine D 2 R-induced locomotion by GABAergic SPNs have also been reported (see Wang et al., 2022) (Centonze et al., 2004).Therefore, endocannabinoids may either promote or suppress striatal dopamine-signalling in a synapse-specific manner.Strikingly, CB1R affinity was shown to be blunted by repeated psychoemotional stress (Rossi et al., 2008), whereas chronic caffeine intake prevented downregulation of CB1Rs following chronic social defeat stress (Rossi et al., 2009).Thus, CB1R modulation by chronic caffeine intake might protect from stress-induced alterations in the endocannabinoid system.As indicative of such a stress-buffering property of regular substance consumption, correlational studies have shown that caffeine ingestion is higher during stressful times (A.Harris et al., 2007;Steptoe et al., 2007).
According to the Yerkes-Dodson curve, performance and flow are ideal when alertness, reflected by NE levels in the brain, are neither low (sluggishness), nor excessive (overwhelm, anxiety, and eventually panic; Fig. 3A) (Peifer, Schulz, Schachinger et al., 2014;van der Linden et al., 2021b;Yerkes and Dodson, 1908).Acute caffeine consumption in an already aggravated state may result in excessive sympathetic activation (Fig. 3A).However, the energizing effect of caffeine may combat sluggishness, while enhancing the motivation to expend effort (Section 4.1.2).Regular substance intake additionally combats the stress-associated downregulation of CB1Rs.Therefore, caffeine might compensate for negative emotions that would otherwise inhibit task-induced flow (van der Linden et al., 2021a).A high caffeine dose could also allow to induce flow in an otherwise overwhelming activity that would exceed the individual's skill level (Kotler et al., 2022;van der Linden et al., 2021b).Thus, both acute and regular caffeine intake may be useful to tune the challenge/skills balance, which is considered to be the most important flow trigger (Kotler et al., 2022), in favour of more flow.

Brain wave and network activity
To generate the strong task focus during flow, the state is thought to involve reduced DMN, and a reciprocal elevation in CEN, activity (D.J. Harris et al., 2017b;van der Linden et al., 2021a).Network activity can also be represented by brain waves, or the intrinsic oscillatory activity of neuronal populations, that can be recorded via electroencephalogram (EEG) (Lopes da Silva, 1991).For example, although it is not fully understood what cognitive processes are represented by these brain waves, alpha power (8 -12 Hz band) suppression following a stimulus is thought to mediate selective visual-spatial attention (see Woodman et al., 2022).Inversely, EEG studies show that higher alpha band power is associated with DMN activity and mind-wandering (reviewed in Knyazev, 2013) (Knyazev et al., 2012).
Regarding flow, an EEG study utilising a mental arithmetic task demonstrated that the skill-matched flow condition is characterised by increased theta activity in frontal brain regions, which seems to reflect cognitive control and concentration (Cavanagh and Frank, 2014;Lagopoulos et al., 2009), and moderate alpha activities in frontal and central areas (Katahira et al., 2018).Interestingly, alpha power increased across the 'boredom', 'flow' and 'overload' conditions (Katahira et al., 2018).In this context, higher alpha activity specifically in the parieto-occipital sulcus was shown to represent visual disengagement in order to allocate resources to working memory (Tuladhar et al., 2007), suggesting that alpha band power could represent working memory load (Katahira et al., 2018).Notably, while flow is thought to involve decreased DMN activity (transient hypofrontality theory; see Dietrich, 2003 andD. J. Harris et al., 2017b), some working memory, as likely reflected by elevated regional PFC activity in some experimental flow studies, might be necessary for effective engagement in cognitively demanding activities (i.e. during video gaming) (Harmat et al., 2015;Yoshida et al., 2014).
Interestingly, after a 24 h caffeine abstinence period, modest caffeine intake (67 mg) improved neuropsychological test performance by reorganizing functional network connectivity in the brains of adults (H.Kim et al., 2021).In this context, a dose of either 50 mg or 200 mg caffeine decreased resting alpha wave activity in the frontal, central, occipital and parietal cortex of young adults 30 min post ingestion (Ajjimaporn et al., 2022;Siepmann and Kirch, 2002).Lowered global alpha band power was also observed 30 -60 min after the ingestion of 46 mg (coca cola), 50 mg, 80 mg or 250 mg caffeine by children, university students or adults (Barry et al., 2009;Barry et al., 2005;Foxe et al., 2012;Meng et al., 2017).Furthermore, caffeine-induced improvements in the attention-assessing trail-making test part B were inversely correlated with alpha power (Ajjimaporn et al., 2022;H. Kim et al., 2021), whilst positively associated with alpha small-worldness (H.Kim et al., 2021).Notably, alpha small-worldness was shown to be inversely correlated with alpha functional connectivity in the healthy brain (Paeske et al., 2020).Collectively, these studies suggest that acute caffeine intake enhances focus and attention, even in regular coffee drinkers, as reflected by reduced alpha power, or lowered mind-wandering, in cortical regions.
A recent study supports this suggestion, showing that more frequent caffeine consumption alters network properties in favour of alertness and action readiness, but also anxiety and emotionality (Magalhaes et al., 2021).These network-altering effects of chronic caffeine intake, which are likely linked to substance-induced activation of the locus coeruleus-norepinephrine system (Fig. 3A) (M.D. Chen et al., 1994;Kirch et al., 1990;Oken et al., 2006), might explain why the substance consistently enhances vigilance even at low doses from 32 mg (~0.5 mg/kg), including in regular users (McLellan et al., 2016;Nehlig et al., 1992).In agreement with increased functional connectivity, executive function and network efficiency bestowed by regular caffeine intake (H.Kim et al., 2021;Magalhaes et al., 2021), habitual substance consumers showed reduced connectivity in the posterior DMN, but enhanced connectivity in visual structures and the right executive control network (Pico-Perez et al., 2023).Furthermore, alpha power, as lowered by caffeine, seems to be inversely correlated with attention and focus (Knyazev, 2013;Woodman et al., 2022), while flow is an energy-efficient state almost undistinguishable from hyperfocus (see Ashinoff and Abu-Akel, 2021) (Huskey, Wilcox et al., 2018).Collectively, this suggests that chronic caffeine intake could prime brain networks for flow; provided that non-anxiogenic caffeine doses are used (Fig. 3A) (van der Linden et al., 2021b).

ADHD and its connection to flow proficiency
ADHD is characterised by dysfunctional mesolimbic dopamine transmission (originating from the VTA) and deregulated noradrenergic signalling (arising from the locus coeruleus) (Chandler et al., 2014;Faraone et al., 2015;Haenlein and Caul, 1987;Johansen et al., 2009).Specifically, projection of dopamine towards the nucleus accumbens stimulates feelings of reward, whereas NE release in the posterior cortex modulates the reactivity to stimuli (Chandler et al., 2014).Joint liberation of both dopamine and NE in the PFC improves attention and working memory.Glutamatergic deficits in corticostriatal regions and alterations in serotonergic and cholinergic pathways are also thought to contribute to ADHD (Cortese, 2012;Johansen et al., 2009;Lesch et al., 2013).Furthermore, genomic association studies support a causal link between gene polymorphisms in dopamine-, NE-and serotonin-related genes and ADHD (Faraone et al., 2015;Gizer et al., 2009).Indeed, ADHD has strong heritability (70 -80%) and is largely impacted by the aforementioned gene polymorphisms (present in 40% of patients) (Faraone et al., 2015).Additionally, ADHD commonly occurs with other comorbidities, such as autism, learning, anxiety and depression disorders etc (see Gnanavel et al., 2019).
Along with neurotransmitter deficits, ADHD patients exhibit an underactive and less connected executive control network (goal-directed behaviour, long-term planning, task inhibition and cognitive flexibility), an impaired alerting network and attenuated negative activity correlations between the DMN vs. frontoparietal control network (attention), as well as alterations in reward-associated brain structures (increased reward threshold, motivational shortcomings and difficulties in delaying gratification) (Faraone et al., 2015;Johansen et al., 2009).Three main ADHD subtypes exist, which include i) the display of inattention (poor focus on non-stimulating tasks, or those with only 'regular' reward, and distractibility), ii) hyperactive-impulsive behaviour (excessive motor activity), or iii) a joint representation of both (Association, 2013;Vazquez et al., 2022).Notably, children and teenagers are more commonly affected by the disorder than adults (~5% children vs. 2.5% adults worldwide).
ADHD presumably both positively and negatively affects the ability to enter flow states.Despite the presence of attentional and motivational impairments, ADHD patients may exhibit 'hyperfocus' on selected tasks that are deemed interesting or exciting.ADHD-associated task hyperfocus has been hypothesised to parallel flow, sharing common features such as intense concentration or elevated performance (reviewed in Ashinoff and Abu-Akel, 2021).On the other hand, inattentive and/or hyperactive-impulsive ADHD traits likely impair flow proficiency in 'normal' tasks (Association, 2013;Vazquez et al., 2022).

Caffeine as a flow-booster in ADHD
Caffeine might be useful to improve ADHD symptoms and, thus, access to flow, in individuals with this disorder.In vivo, a recent systematic review concluded that caffeine boosts attention, learning, olfactory accuracy and memory various animal models of ADHD (Vazquez et al., 2022).However, the effects of caffeine on the classical ADHD symptoms, hyperactivity and impulsivity, were deemed controversial.Mechanistically, animal research suggests that chronic caffeine consumption enhances ADHD symptoms by increasing striatal dopamine release (Franca et al., 2020), normalizing (the abnormally elevated) levels of synaptic DAT and reducing adenosine A 2A R densities in the frontal cortex and striatum (Pandolfo et al., 2013).Long-term caffeine intake by adolescent ADHD rodent models was also shown to heighten serotonin release as well as synaptosomal-associated-protein-25 (SNAP-25; a protein implicated in synaptic vesicle recycling, discharge of neurotransmitters and, in the case of gene polymorphisms, ADHD) and syntaxin 1 levels (a functional associate of SNAP-25) in the PFC and hippocampus (Franca et al., 2020;Gizer et al., 2009).Collectively, these studies indicate that chronic caffeine consumption may be useful to prime and 'train' attentional and motivational circuits in younger ADHD rodent models, but without improving behavioural ADHD symptoms.If also occurring in adolescents with ADHD, caffeine-induced re-arrangement of attention and motivation-associated brain networks, which are key players in flow (Peifer et al., 2022), will conceptually benefit the ability to enter the state in later life.
The in vivo benefits of caffeine might translate into humans with ADHD.Interestingly, some adenosine A 2A R gene polymorphisms have been associated with ADHD (Molero et al., 2013), whilst caffeine, a nonspecific adenosine receptor antagonist, was suggested to be useful for treatment of the disorder (Alasmari, 2020).Indeed, the presence of ADHD symptoms is associated with greater caffeine consumption in adolescents and young adults (twice as high relative to the healthy population (Walker et al., 2010)), possibly as an attempt of 'self-medication' (Kelly and Prichard, 2016;Marmorstein, 2016;C. A. Martin et al., 2008;Rapoport, 1986;Walker et al., 2010).Congruent with this suggestion, correlations between caffeine use disorder and ADHD have also been reported (Agoston et al., 2022;Cipollone et al., 2020).In favour of caffeine, 1 -2 cups of coffee daily (~175 -300 mg caffeine) improved behavioural symptoms, attention, reaction speed and accuracy in ADHD children (Firestone, Poitras-Wright et al., 1978;Garfinkel et al., 1981;Harvey and Marsh, 1978;Kupietz and Winsberg, 1977;Reichard and Elder, 1977;Schechter and Timmons, 1985;Schnackenberg, 1973).Less caffeine (158.6 mg) might be superior compared to higher doses (Garfinkel et al., 1981).On the other hand, the alkaloid triggered undesirable adverse effects, such as insomnia, in one of these studies (Schechter and Timmons, 1985).There are also reports of no (Arnold et al., 1978;Conners, 1975;Firestone, Davey et al., 1978;Garfinkel et al., 1975;Huestis et al., 1975), or even a negative impact (249 mg) (Gross, 1975), of caffeine on children with ADHD.As recently debated, caffeine might be recommended to treat deficits in cognitive and physical performance in ADHD children (Torres-Ugalde et al., 2020).By contrast, some adverse effects, such as sleep cycle disruptions and a greater proneness to anxiety and depression relative to adults, discourage caffeine ingestion by healthy kids.A challenge for future ADHD studies in children is the identification of an effective caffeine dose that boosts cognitive and physical function and, with it, flow.In addition, related to the stimulatory effects of caffeine (Fig. 3A), common ADHD comorbidities such as anxiety disorders must be considered (Gnanavel et al., 2019).
Concerning adult ADHD, a population study reported that soldiers with this condition benefitted from regular caffeine intake, leading to improved self-rated scores in orderliness, motivation, attention on repetitive tasks, duly task completion, behavioural inhibition and speedy driving (Cipollone et al., 2020).Therefore, the caffeine-evoked improvements observed in ADHD animal models and children also appear to be present in adults with ADHD, encouraging the frequent use of caffeine to facilitate flow states.

Limitations and drawbacks of using caffeine
There are caveats to using caffeine.Excessive caffeine intake beyond 500 mg may trigger anxiety, irritability, restlessness, tachycardia, tremors and psychosis (caffeinism), thus attenuating performance and the ability to enter flow (Cappelletti et al., 2015;Kaplan et al., 1997;Lara, 2010;McLellan et al., 2016;Peifer, Schulz, Schächinger et al., 2014;van der Linden et al., 2021a).Special care must be taken when using caffeine powder.While this is the most accurate manner of consuming the substance, ingestion of ~35 -40 mg/kg caffeine is fatal, as reported following the inadvertent consumption of caffeine powder (Jabbar and Hanly, 2013;McLellan et al., 2016).In the context of adverse effects, there is also a dose-dependent clearance effect; i.e. caffeine is eliminated at a slower rate if consumed at higher doses due to the plasma accumulation of its equally bioactive metabolite paraxanthine (Cappelletti et al., 2015;Denaro et al., 1990;Okuro et al., 2010;Snyder et al., 1981).The performance and, likely, the flow-enhancing properties of caffeine may also be lessened in genetically susceptible individuals with adenosine A and dopamine receptor polymorphisms.These individuals show a higher side effect profile, including aggravated anxiety, easier sleep disruption and headache post withdrawal, even at low caffeine concentrations from ~100 mg (Alsene et al., 2003;Childs et al., 2008;Retey et al., 2007;Rogers et al., 2010;Yang et al., 2010).While moderate caffeine use may reduce the risk of depression (Lucas et al., 2011), the substance should be avoided by individuals with generalized anxiety disorder and other neuropsychiatric conditions (Bruce et al., 1992;Lara, 2010;Peng et al., 2014).However, to improve the benefits and blunt the side effects of caffeine, more frequent caffeine use might be helpful (Rogers et al., 2010).
If consumed at high concentrations or close to bedtime, caffeine may inhibit sleep (Ali et al., 2015;Miller et al., 2014;Salinero et al., 2014).Specifically, caffeine alters GABA release and receptor expression in various brain regions (Alasmari, 2020;Ko et al., 2018;Roca et al., 1988), while delaying the circadian melatonin rhythm via the blockage of adenosine A 1 Rs (Burke et al., 2015).These caffeine-induced sleep disturbances are both hazardous for health and flow, which is an energy-demanding state.
Another drawback is that the effects of caffeine desensitize over time (Cappelletti et al., 2015).Therefore, a prudent re-sensitisation protocol, which may involve temporary caffeine withdrawal for up to three days, might be necessary (Hsu et al., 2010;Simola, 2010;Simola et al., 2008;Simola et al., 2006).To combat caffeine withdrawal symptoms, such as headache (Cappelletti et al., 2015), decaffeinated beverages may be used.Interestingly, an open-label study demonstrated that drinking decaffeinated coffee reduced perceived withdrawal symptoms relative to water, even when participants were informed about the type of beverage they consumed (Mills et al., 2023).
Although less addictive than nicotine, alcohol and other drugs of abuse, ~17% of caffeine consumers are dependent on this substance, suggesting that caffeine must be used responsibly (Cappelletti et al., 2015;Griffiths and Mumford, 1996).Furthermore, at least 85 commercially available drugs are known to interact with caffeine (listed on Drugbank website (Drugbank, 2023)), which may modify caffeine metabolism and lead to life-threatening complications, such as cardiac arrythmias and seizures (Carrillo and Benitez, 2000;Temple et al., 2017).Therefore, prior to the deliberate use of caffeine, potential interactive effects with other medications taken by the individual should be discussed with a physician.

Future directions
Collectively, we have highlighted various interaction between both acute and chronic caffeine consumption and flow, warranting mechanistic studies.In humans, the impact of different caffeine dosages on self-reported flow, ideally of non-users and regular consumers with or without substance withdrawal, may be assessed.This should be combined with assessment of flow-associated brain activity (e.g.fMRI, fNIRS and EEG) in a challenge/skill-matched flow condition and performance in a cognitive and/or physical task-at-hand (e.g.music or sports), with evaluation of sympathetic and parasympathetic nervous system activation.Given the motivation-enhancing and stress-buffering effects of caffeine, it would be interesting to investigate the impact of the substance on the C/S balance and subsequent flow, to determine if caffeine facilitates flow in an 'overwhelm' condition.These experiments may also include testing caffeine for inducing flow in ADHD patients.Modulation of the endocannabinoid-system, motivational and other aspects of caffeine, also those related to cerebral changes in neurotransmitter release and receptor-signalling related to NE, dopamine, adenosine, endocannabinoids etc, could be investigated in animal models (see Hintze and Yee, 2023).Finally, protocols for caffeine re-sensitisation might be established to induce flow long-term.
Besides using caffeine for flow, an interesting alternative might be the utility of selective adenosine A 2A R, and possibly even CB1R, antagonists.Adenosine A 2A R antagonists have been used to enhance locomotion in PD rat models, strikingly without desensitization (Pinna et al., 2001;Popoli et al., 2000).This suggests that an effective dose that enhances dopamine D 2 R-associated motivation ('wanting') could be determined and repeatedly used prior to physical/cognitive work, likely without the mood-related side effects from dopamine D 1 R desensitization (Section 4.1.2).
In principle, any drug that enhances performance, attention, concentration or other characteristics of flow could be useful to induce, facilitate or prolong this state.The possibilities are endless, and combination of such drugs for the purpose of flow poses an exciting field of research.

Conclusion
The available evidence supports that caffeine facilitates flow states in both cognitively and physically demanding tasks, in particular under sleep-deprived conditions and when under-aroused.Acute caffeine intake improves alertness, energetic arousal and positive affect.By blocking adenosine A 1 Rs (following caffeine withdrawal) and adenosine A 2A Rs (also in regular consumers), acute caffeine intake lowers effort aversion,.Specifically, by inhibiting adenosine A 2A Rs on striato-pallidal SPNs, caffeine enhances dopamine D 2 R affinity in the nucleus accumbens (ventral striatum), resulting in a greater desire ('wanting') to invest effort into intrinsically rewarding activities, thus facilitating flow.Furthermore, caffeine enhances alertness by stimulating cerebral NE and systemic cortisol release, and titration of caffeine is needed to optimize performance and flow by avoiding too high (anxiety) and too low (boredom) levels of these hormones.The ingestion of higher caffeine concentrations also transiently increases HF HRV for ~120 min, which leads to a reduction in cortical stress perception and may promote flow.Given that the dopamine D 1 R-associated effects of pleasure ('liking') desensitize due to upregulation of striatal adenosine A 1 Rs, strategic multi-day wash-out periods might be employed.
Chronic caffeine consumption potentially protects dopaminergic neurons from age-associated neurodegeneration, which is predicted to benefit flow with advanced age.Regular caffeine intake reduces systemic endocannabinoid levels, which might contribute to caffeineassociated increases in cerebral NE release and vigilance.Indeed, reflecting increased alertness, regular caffeine consumption reduces alpha power in cortical areas, decreases DMN activity and improves functional network connectivity, attention and executive function in humans.Interestingly, possibly through downregulation of local adenosine A 2A Rs, chronic caffeine intake seems to enhance presynaptic CB1R-signalling in striato-pallidal SPNs, thus potentiating dopamine D 2 R-induced motivation by reducing GABA release and disinhibiting ventral pallidum neurons.Furthermore, regular caffeine intake prevents the stress-induced downregulation of striatal CB1Rs, suggesting that the substance potentially enhances flow by improving stress tolerance.Frequent caffeine intake also appears to re-organise dopaminergic and adrenergic signalling in adolescents and adults with ADHD in favour of enhanced motivation and attention.
However, the potential development of substance addiction and side effects, including anxiety, sleep disturbances or even fatalities, must be considered when using caffeine.Due to genetic differences in caffeine metabolization or the presence of adenosine A 2A R polymorphisms or neuropsychiatric conditions that aggravate adverse effects, effective caffeine doses should be determined on an individual basis.

Literature research
To identify papers, PubMed, Web of Science and Scopus were mainly used.Publications in the English language were considered from 1980 up to the present time.A combination of search terms, as dependent on the respective section, was applied, usually including the terms 'caffeine', 'coffee', 'flow' and 'flow state'.

Fig. 1 .
Fig. 1.Adenosine production in neurons.(Presynaptic) Active neurons utilise ATP as part of their metabolism as well as to oxidise glucose and fatty acids.This results in dephosphorylation of ATP to ADP and AMP, whilst AMP may be converted to adenosine by 5'-nucleotidase (or its ecto-version extracellulary).Since the activity of adenosine-clearing enzymes, including adenosine kinase, adenosine deaminase and S-adenosylhomocysteine hydrolase, is regulated by the circadian rhythm and higher at night, adenosine accumulates during daytime and in response to prolonged neuronal activity.Intra-and extracellular adenosine concentrations are in equilibrium due to bidirectional transport by nucleoside transporters.Caffeine enters the brain via diffusion through the blood brain barrier.AMP = adenosine monophosphate, ADP = adenosine diphosphate, ATP = adenosine triphosphate.

Table 1
Lifestyle factors influencing caffeine absorption and plasma concentrations.