Effects of caffeine intake and exercise intensity on executive and arousal vigilance

During physical efforts and sport practice, vigilance is responsible for maintaining an optimal state of activation, guaranteeing the ability to quickly respond and detect unexpected, but critical, stimuli over time. Caffeine and physical exercise are able to modulate the activation state, affecting vigilance performance. The aim of the present work was to assess the specific effects and modulations of caffeine intake and two physical intensities on vigilance components. Participants performed an attentional task (ANTI-Vea) to measure the executive and arousal components of vigilance, in six double-blinded counterbalanced sessions combining caffeine, placebo, or no-ingestion, with light vs. moderate cyclergometer exercise. Exercise at moderate intensity improved executive vigilance with faster overall reaction time (RT), without impairing error rates. Instead, caffeine intake generally improved arousal vigilance. In conclusion, caffeine and acute exercise seems to moderate executive and arousal vigilance in different ways.

on neuroelectric findings, LC-NE paradigm and information processing theories, we expect to see a positive modulation of moderate exercise intensity, exploring an exercise-caffeine interaction, on executive vigilance. (d). Complementarily, we expect 5 mg/kg caffeine intake to reduce phasic alertness in low consumers during exercise, whereas we wanted to further explore caffeine and exercise effects attentional networks, extending previous research with ANT and ANT-I tasks.

Methods
Participants. Eligible participants were all young active adults, between 18 and 26 y/o, enrolled in the Sport Sciences Degree at Catholic University of Valencia "San Vicente Mártir". Exclusion criteria were cardiac pathologies, diabetes, vision problems, allergies to caffeine o cellulose, and existing musculoskeletal injury. A 24 participants sample size was estimated a priori with G*Power 3.1.9 (Universität Kiel, Germany) 34 to reach d = 0.4 effect size, 1-β = 0.8 power, and p = 0.05 statistical significance for a repeated measures ANOVA. This effect size is reasonable according to a meta-analysis regarding caffeine effects on physical performance 35 and a similar research involving caffeine, acute exercise and attentional performance 36 . 27 participants were selected to participate in the study to prevent possible sample loss. Indeed, three of them dropped out before study completion of all experimental sessions: two of them due to schedule breach and one due to an injury. Then, 24 men; (age: M = 22.6; SD = 2.9), non-habitual and low caffeine consumers (see Table 1) completed the study. All of them signed an informed consent before their inclusion in the experiment and were fully informed about the study protocol, experimental sessions, relevance of commitment, and rigorous following of pre-session indications. All experimental protocols were approved by the Catholic University of Valencia Ethics Committee (UCV/2016-2017/02). Furthermore, methods described in this study were performed following the requirements of the declaration adopted by the 18th World Medical Assembly (Helsinki, 1964), revised in Fortaleza (Brazil, 2013), and the Taipei declaration (2016).

ANTI-Vea.
This new version of the attentional networks test is suitable to assess, within a single session, the independence and interactions of the classic attentional functions (e.g., phasic alertness, orienting, and executive control), simultaneously with the decrement across time on task of the executive and the arousal vigilance components 4 . The procedure, stimuli sequence, timing, and correct responses for the three type of trials of the task (e.g., ANTI, EV, and AV) are depicted in Fig. 1. The task was completed on a laptop (HP Pro Book, EEUU) at approximately 60 cm from its monitor, and all stimuli were presented at eye level and in a dimly lit room. E-Prime software (Psychology Software Tools, Inc.) was used for stimuli presentation and response registration.
The ANTI trials (48 per experimental block) follow the procedure of the ANTI task, wherein participants have to perform a flanker task -which is useful to assess the executive control network -by responding to the direction the target (e.g., a central arrow) points to, while ignoring the direction pointed by the surrounding flanker arrows (e.g., two distracting arrows at each side of the target). Importantly, as shown in Fig. 1, the response stimuli: (a) could appear below or above the fixation point; (b) could be anticipated or not by a warning signal (which provides an effective measure of phasic alertness); and (c) its location on the screen regarding the fixation point could be cued either correctly, incorrectly, or not cued at all, by a visual cue (which provides a measure of attentional orienting). Therefore, the executive control effect was established as the congruency (congruent) vs. incongruency (incongruent) of the target and flankers' arrows direction; the phasic alertness effect was defined as the presence (tone) vs. absence (no tone) of the auditive warning signal; and finally, the orienting effect was defined as the presence of a correct cued location (valid), incorrect cued location (invalid), or no spatial cue at all (no cue).
In a smaller proportion of trials (16 by block), the participants performed a typical signal detection task (similar to the SART), which is suitable to assess the EV component: they had to remain vigilant for detecting an infrequent displacement of the target (e.g., either upwards or downwards) from its central position, ignoring in these cases the direction pointed by the target (see Fig. 1) and pressing the middle button. Finally, on AV trials (16 by block, see Fig. 1) participants performed a task similar to the PVT, which is useful to obtain a direct and independent measure of the AV component: they had to stop a millisecond counter as fast as possible by pressing any available response button. Further details on stimuli properties and design can be reviewed in Luna et al. 4 .

Preliminary testing. A sub-maximal incremental ramp test was performed on a cyclergometer Cardgirus
Medical Pro (G&G Innovation, Álava, Spain) to obtain ventilatory thresholds (VT 1 and VT 2 ) and their associated power output. Ventilatory thresholds represent two key points of metabolic change due to increase in exercise intensity, and have been used as references for exercise intensity prescription 37  Exercise protocol. Experimental sessions were conducted following a constant sub-maximal exercise protocol. Participants started with a warm-up based on Schmit et al. 41 : 180 s at 50% of VT 2 , 120 s of progressive increase/decrease intensity until session target intensity was reached, and 90 s at session target intensity for VO 2 stabilization. Afterwards, intensity remained constant at 80% of VT 1 or VT 2 , with no interruptions during 33 min and 45 s. Cadence had to be maintained above 60 rpm. Treatment ingestion. Participants received 5 mg/kg of either anhydride caffeine, or placebo capsules with cellulose, or no-treatment at all, as control condition. Caffeine dosage was slightly greater than in previous research 31,42,43 , aiming at obtaining a clear modulation of caffeine intake. Capsules were ingested with ad libitum water and were identical (number, color, tact, and weight) in both treatments. In all experimental sessions, Caffeine and Placebo treatment was always administered 45 min prior attentional assessment and exercise protocol, following Huertas et al. 31 .
Procedure. Participants completed eight sessions at the Catholic University of Valencia. Following previous research 31 , sessions were separated by 48-96 h to avoid the influence of exercise on attentional performance, to minimize physical or physiological changes, and to ensure caffeine removal 44 . Also following previous research 45 , www.nature.com/scientificreports www.nature.com/scientificreports/ instructions to restrict caffeine or stimulants ingestion 12 h before each session, vigorous physical activity during previous 24 h, and doing last meal at least 2 h before experimental session were given to all participants. In the first session, participants were familiarized with the sub-maximal incremental test and experimental instruments by completing the ANTI-Vea task while pedaling on the cyclergometer at an auto-selected intensity. Importantly, only in the first session of the study participants gradually received extensive instructions to perform correctly each type of trial of the ANTI-Vea and completed several practice blocks with visual feedback regarding incorrect responses, as in Luna et al. (see also Fig. 1). Next, during the second session, the sub-maximal incremental ramp test was performed to assess ventilatory thresholds and to establish, individually for each participant, the intensity of the experimental sessions. In the following six sessions, participants only completed one practice block prior to the experimental task (see Fig. 1), and were randomly assigned through a double-blinded and counterbalanced Latin-Square design to one of the following treatments: (1) 5 mg·kg Caffeine at 80% VT 1 ; (2) 5 mg·kg Caffeine at 80% VT 2 ; (3) Placebo at 80% VT 1 ; (4) Placebo at 80% VT 2 ; (5) Control at 80% VT 1 ; (6) Control at 80% VT 2 . In the Caffeine and Placebo sessions, participants received the corresponding treatment and waited in the laboratory facilities. In Control sessions, they received no treatment. After 45 min, participants performed the ANTI-Vea task in the cyclergometer. To collect responses to stimuli while pedalling, three response buttons were placed in the left, middle, and right side of the cyclergometer's handlebar. Buttons positions were adjusted individually for each participant, allowing responding with the index fingers to the left and right buttons, and with the dominant hand thumb to the middle button (see Fig. 1).
Statistical analysis. Data analysis was performed in Statistica Dellsoft 12.0 (StatSoft, Inc.). Breath-by-breath ventilatory variables (VCO 2 , VO 2 , V E , EqO 2 , EqCO 2 y RER) were analyzed for extreme and outliers' values by box-plot method (1.5 coefficient interquartile range), recoding those values with two-sided Tukey. Then, mean values were calculated for every treatment session and participant. No HR extreme and outliers' values were found with Polar Protrainer v.5 (Polar Electro) software. To assess exercise and treatment influence on ventilatory variables and HR, six repeated measures ANOVAs were conducted, including Exercise (Light/Moderate) and Treatment (Caffeine/Placebo/Control) as within-participant' factors, and VCO 2 , VO 2 , V E , EqO 2 , EqCO 2 , RER, and HR, as dependent variables. ANTI-Vea outliers and extreme values were handled following the same procedure as Luna et al. 4 .
For the RT analysis (both for ANTI and EV trials), responses with RT faster than 200 ms or slower than 1500 ms (0.44%), and those with incorrect responses (7.47%), were excluded 4 . Then, two repeated measures ANOVAs were conducted, one with RT and the other with error rate (% of errors) as dependent variable, and including Phasic Alertness (No Tone/Tone), Orienting (Invalid/No Cue/Valid), Executive Control (Congruent/ Incongruent), Exercise (Light/Moderate) and Treatment (Caffeine/Placebo /Control), as within-participants factors. Classic attentional networks indexes of RT and error rates performance were calculated by means' subtraction in specific conditions, following previous studies 4 : Phasic Alertness (No Tone -Tone, only considering No Cue trials), Orienting (Invalid -Valid) and Executive Control (Incongruent -Congruent). Repeated measures ANOVAs were conducted then for each index as dependent variable, including Treatment (Caffeine/Placebo/ Control) and Exercise (Light/Moderate) as within-participants factors.
Additionally, EV and AV measures were calculated per block of trials in each experimental session, in order to assess vigilance decrement across time on task. Therefore, for the analysis of EV trials, data from Phasic Alertness, Orienting and Executive Control was not considered. Following Luna et al. 4 , non-parametric indexes of sensitivity (A') and response bias (B") were calculated from Hits (e.g., correct detection on infrequent target) and False Alarms (FA, e.g., responses to the frequent stimuli with the infrequent target response) 46 . Finally, mean RT and variability (SD) of RT on Hits were also obtained. Four independent repeated measures ANOVAs were conducted, one for each dependent variable (A' , B", mean and SD of RT), including Treatment (Caffeine/Placebo/ Control), Exercise (Light/Moderate) and Block (six levels), as within-participants factors.
Furthermore, regarding AV data, mean and SD of RT, and percentage of Lapses (e.g., responses equal or higher than 600 ms) per block of trials and experimental session were calculated. The consideration of Lapses as responses with RTs equal or greater than 600 ms differs from the 500 ms threshold usually used 9 , due to the greater task demands and the higher mean RT observed in the ANTI-Vea in comparison with much simpler tasks like PVT. Then, three repeated measures ANOVAs were conducted including Treatment (Caffeine/Placebo/Control), Exercise (Light/Moderate) and Block (six levels) as within-participants factors, one for each dependent variable: mean RT, SD of RT, and percentage of Lapses. Statistical significance level and confidence intervals were set at 0.05 and 95%, respectively, for all analysis.

Results
Participants' basic characteristics and ventilatory values obtained in the incremental test are shown in Table 1 and RER were also observed, but not for EqCO 2 (see Table 2). Exercise × Treatment interaction was not observed for any of the physiological variables (all Fs < 1).
ANT-I RT. The typical main effects and interactions, usually observed with ANT-I task, for RT measures can be found as Supplementary Material S1. This pattern of results demonstrated the correct functioning of the task and therefore its suitability to assess the functioning of the classic attentional networks (see Table 3). ANT-I Percentage of errors. The typical main effects and interactions (see Table 4), usually observed with ANT-I task, for the errors rate performance can be also found in the Supplementary Material S1. The main effects of Exercise, Treatment, and the interaction Exercise × Treatment were not significant (all Fs < 1, ps > 0.05), thus demonstrating that participants respond faster to ANT-I trials during ME, but without committing more errors. The modulations of Exercise intensity and Treatment on the attentional networks' indexes are reported in the Supplementary Material S1.  Fig. 2). In other words, the main effect of exercise, with faster RT for the Moderate than the Light exercise condition, increased across blocks of trials. Last, no interaction for Treatment was observed as significant: Exercise × Treatment (F < 1), Treatment × Blocks (F = 1.00, p = 0.440) or Exercise × Treatment × Blocks (F = 1.05, p = 0.399).    Fig. 3). www.nature.com/scientificreports www.nature.com/scientificreports/ SD of RT. The main effect of Exercise was not significant (F < 1, p = 0.649), but it was marginal for Treatment [F (2, 46) = 3.06, p = 0.056, η = .

Discussion
The present study aimed at assessing the effects of caffeine intake during exercise, mainly on executive and arousal vigilance, but also in the classic attentional networks functions. To do so, two different exercise intensities were used to compare the impact of acute exercise and caffeine intake. Behavioral performance was assessed with a task (the ANTI-Vea) suitable to measure in a single session phasic alertness, orienting, and executive control functions, along with the decrement of two vigilance components (arousal and executive vigilance) across time on task 4 . To our knowledge, this is the first study assessing the effects of exercise and caffeine on executive and arousal vigilance in the same experimental session, providing new additional information about the effects of 5 mg/kg caffeine ingestion on the functioning of the attentional networks during exercise.
Our results suggest that moderate exercise prevent an executive vigilance decrement (but not arousal vigilance) during a ~30 min physical effort. On the contrary, while caffeine ingestion seems to have no effect on executive vigilance during exercise, it improved arousal vigilance compared to the placebo and control treatments, reducing reaction time and interacting with moderate exercise to reduce the number of lapses committed across time on task. It is worth mentioning that the reaction time modulation was independent of exercise intensity, suggesting that caffeine ingestion would be capable of improving arousal vigilance despite the attentional benefit of moderate exercise, but also would particularly reduce lapses during the most challenging exercise condition.
Regarding the classic attentional networks, previous studies had shown faster overall responses at the expense of accuracy (known as tradeoff effect) during ANT-I tasks and moderate-to-high exercise intensities (e.g., 80-90% VT 2 ), in comparison to a rest condition 31,32 . Complementary to these reports, our research found the same global RT facilitation effect, in this case compared to light exercise (80% VT 1 ), but without any increase in error rate. Despite these differences, the absence of a tradeoff effect in our study seems to be in line with several studies suggesting that moderate exercise intensity may not be "stressful" enough to produce sympathetic-adrenal changes that impair error rate compared to lower intensities [47][48][49] .
Nevertheless, contrary to our hypothesis, no significant modulation of caffeine ingestion over global RT on the ANT-I sub-task was observed, at difference with previous studies 29,38,50 . Note, however, that Huertas et al. 31 observed that caffeine ingestion facilitated RT at rest in moderate caffeine consumers, but not during moderate exercise. The basis for caffeine attentional effects relays on its role as an adenosine antagonist (blocking A 1 and A 2a receptors, highly present in striatum, hippocampus, cortex, cerebellum, and hypothalamus brain regions) therefore promoting the release of neurotransmitters (e.g., dopamine, noradrenaline, acetylcholine or serotonin) and increasing arousal 24 . Still, following the Locus Coeruleus -Norepinephrine (LC-NE) hypothesis 51 , and as Huertas et al. 31 emphasized, both exercise and caffeine intake would have equally stimulated the LC-NE system, explaining the absence of results in our study. Therefore, as in the present study caffeine treatment was always analyzed in conditions of at least light exercise, this could have caused an arousal ceiling effect of exercise, masking the effects of caffeine on the classic attentional networks. Regarding error rates, our results are in line with previous research 29,31 in which it has been shown that after caffeine ingestion (150-450 mg), either an extremely low error rate or even improved performance is observed in complex attentional tasks 24 .
Regarding arousal vigilance, we observed a decrement in arousal vigilance across time on task, with lapses increasing over time independently of exercise intensity, as in previous research at rest 4 and during exercise 19 . Rationale for this behavioral pattern seems to follow the overload hypothesis 52 : cognitive resources were progressively depleted during a high demanding cognitive task (e.g., the ANTI-Vea) and physical effort (light and moderate), impairing attention and increasing lapses. However, this result implied that the moderate exercise condition had no effect on arousal vigilance, which was one of our main hypotheses. Limited research had previously shown Scientific RepoRtS | (2020) 10:8393 | https://doi.org/10.1038/s41598-020-65197-5 www.nature.com/scientificreports www.nature.com/scientificreports/ a reduction in mean RT by light-to-moderate exercise (40-80% VT 2 ), but also conflicting results during high intensities, and no effect on vigilance decrement 17,18 .
Arousal vigilance has not been so far assessed specifically along with other attentional networks in a complex task, being a factor that influences vigilance performance 33 . Therefore, the complexity of the ANTI-Vea task, due to the simultaneous assessment of the executive vigilance component (and the other attentional functions), at difference with previous research only using the PVT 53 , could explain why, contrary to our initial hypothesis and also to Gonzalez-Fernández et al. work 18 , moderate exercise (80% VT 2 ) intensity produced no beneficial effect compared with light exercise (80% VT 1 ). Altogether, the results by Gonzalez-Fernández et al. 18 seems to be in line with the reduction on executive vigilance decrement by the moderate exercise intensity observed in our experiment, that will be discussed latter. Nevertheless, our work could complement Huertas et al. 32 proposal, suggesting that vigilance improvement may only occur when comparing rest conditions with moderate exercise, and not between two exercise intensities due to the activation induced by the minimal exercise.
On the contrary, arousal vigilance was modulated by caffeine ingestion, with 5 mg·kg of caffeine leading to faster responses and fewer lapses during moderate exercise intensity. These results provide new insights into caffeine effects on arousal vigilance during a concomitant physical effort. Our results agree with McLellan et al. 28 work, who described a reduction in the number of lapses committed during a military specific vigilance exercise after a 200 mg caffeine intake, and with the solid body of evidence in resting condition, with or without sleep deprivation 22,54,55 . In this regard, it is well known that caffeine blocks vigilance related adenosine receptors in the brain (e.g., hippocampus, cortex, cerebellum) causing a general effect of vigilance 56 . Additionally, the placebo ingestion during light exercise conditions (not in moderate intensity) mildly reduced RT variability, which supports placebo's significant cognitive 57 and psychomotor 58 effects through the release of dopamine in prefrontal cortex, but highlights that it could be limited to low activation states.  www.nature.com/scientificreports www.nature.com/scientificreports/ Regarding executive vigilance, we expected the decrement across time on task on this component to be observed as a change in response bias towards committing fewer errors. This would had major implications in a real situation, as the tendency to be more "conservative" during a critical and unexpected event could lead, for example, to confound a quick turn to avoid a crash of a cyclist with a normal movement of the peloton during a cycling ride (e.g., the cyclist has to see the danger very clearly to respond to it). In this regard, while the great majority of scientific studies haven't found significant changes in response bias over time (mostly due to limited task complexity 7 ), our data showed, partially confirming our hypothesis, a significant reduction in hits accompanied by a reduction in sensitivity and a significant increase in response bias across blocks, independent of exercise intensity. These results differ from data collected by Luna et al. 4 at rest, and contrast with Thomson et al. 7 hypothesis. To explain these results, we suggest that dual task demands in our study exaggerate the executive vigilance decrement due to the reallocation of cognitive resources 59 , allowing us to see both a sensitivity loss and an increase in response bias during the ANTI-Vea, not observed by Luna et al. 4 .
Indeed, other studies assessing the executive component during exercise have also found a decrement of vigilance performance with time on task 15,29 , that might be magnified when heavy loaded physical efforts are performed 15 . Still, based on neuroelectric findings with light intensity efforts (improved attentional resource allocation: increased α1 and β1 power measures of an electroencephalogram) 13 and exercise-related norepinephrine release, we hypothesized that higher exercise intensities would enhance executive vigilance, and possibly reduce vigilance decrement. We observed that moderate exercise intensity prevented the increase in RT observed during light exercise, thus avoiding a significant vigilance decrement. Although this result opposes a reported increase in vigilance decrement during "loaded walking" vs. "unloaded walking" with a Go/NoGo task 15 , physical effort (120 min of complex, heavy loaded and variable inclination walking vs. constant light and moderate aerobic exercise in a cyclergometer) could explain the differences with our vigilance outcomes.
Moreover, our results agree with one report of improved brain activation, but no behavioral performance changes, in a SART task after light exercise (120-150 heart beats per minute) compared to rest 60 and support the role of LC-NE 51 , possibly explaining exercise modulation of executive vigilance through a better maintenance of excitation during the ANTI-Vea task. Following the LC-NE hypothesis, maintenance of excitation throughout a vigilance task depends on tonic activation of the neurons associated to dorsolateral pre-frontal cortex, frontal eye fields, intraparietal sulcus, thalamus and anterior cingulate cortices 46,61 ; and any decrease in their activation frequency would lead to the decrease of performance observed in vigilance tasks 62 . Indeed, it has been recently shown that stimulation with transcranial direct current stimulation over the posterior parietal or the dorsolateral pre-frontal cortex reduces the executive vigilance decrement 63 . In this scenario, low-to-moderate exercise would stimulate the synthesis and release of norepinephrine from the locus coeruleus to the rest of the brain, thus maintaining tonic activation and preventing performance decrease 14 .
Therefore, based on Smit et al. 13 work and our results, a moderate intensity exercise could be necessary (e.g., 80% VT 2 ) to induce a sufficient increase in catecholamines to activate and maintain executive vigilance related brain areas and observe a behavioral change. Nevertheless, our findings suggest that intensity selection can be crucial in a wide range of sport modalities that aim at maintaining executive control during an exercise period of 30-35 min. In these particular cases, a moderate intensity effort can contribute to reduce executive vigilance decrements and enhance global attentional performance. However, future research should confirm this hypothesis in specific sport contexts (e.g., outside the lab).
Lastly, regarding caffeine, no significant modulation of executive vigilance by caffeine or placebo ingestion, neither an interaction between caffeine ingestion and exercise, was observed. These results contradict available evidence regarding caffeine effects on vigilance during 29,38 or after exercise 27 29,38 , but mostly the attentional task and participant characteristics could explain this controversy. One possible explanation is that low prevalence of infrequent signals in previous studies 29,38 (10% vs. 20% in our task), as well as the fact that the ANTI-Vea demands responses to both frequent and infrequent stimuli, may have influenced the effects of caffeine. A second, and perhaps more solid explanation, is that habitual caffeine intake could influence acute caffeine effects on executive vigilance, as suggested for the effect on the classic attentional networks 31 . In this regard, while Shulder et al. 16 and our sample were low (<100 mg/day) or non-habitual (<50 mg/day) caffeine consumers, participants habitually consumed 100-225 mg/day in studies that have found a positive impact of caffeine ingestion on executive vigilance, during or after exercise 27,29,38 .
Still, the absence of a significant interaction between exercise and caffeine on executive vigilance and opposed caffeine effects on executive and arousal vigilance remain to be explained. Following information processing theories, a psychomotor dual-task would reduce the amount of cognitive resources allocated to the cognitive task, impairing its performance, but caffeine has been proposed to counteract this reduction with an increase of brain activity 38 . From this, one would expect higher caffeine effects on vigilance as physical demands of dual task increases (e.g., increasing exercise intensity from light to moderate). Since our work is the first to analyze the interaction between caffeine and sub-maximal physical exercise in executive vigilance, it is difficult to draw solid conclusions, but, in order to explain the absence of caffeine-exercise interaction, it is likely that the two vigilance components trigger the brain differently, both on location and intensity, determining caffeine effects. Arousal vigilance has been associated to right hemisphere activation 64,65 , while there is evidence of how activation of the left hemisphere (e.g., lower frontal gyrus) is critical for inhibition of motor response (a critical ability in executive vigilance tasks 66 ). As caffeine appears to increase brain activity, especially in the right hemisphere 67 , we suggest that executive vigilance could be less responsive to caffeine effects. Nevertheless, this hypothesis needs to be proven in future research with neuro-physiological measures like fMRI, fNIRS or EEG. The fact that an on-line version of the ANTI-Vea task is freely available (https://www.ugr.es/~neurocog/ANTI/) in different languages might help to extend research in this field with consistent measures in different studies.
Finally, we expected caffeine intake to reduce phasic alertness in low consumers during exercise and explore the effects of caffeine and exercise over the functioning of the attentional networks, extending previous research with ANT and ANT-I tasks. In this sense, caffeine has been shown to improve neuronal activity and neurotransmitters release 68 , thus improving phasic alertness 69 , executive control 70 and orienting 71,72 at rest, in a similar way to exercise arousal effects through catecholamine's increase and dopaminergic system stimulation 73 . As happened with global RT, exercise influence on dopaminergic system would have caused a ceiling effect and masked the effects of caffeine on the RT phasic alertness index. Still, we did find a significant improvement in the error rate phasic alertness index after caffeine ingestion, that was only present in light exercise condition and agrees with a superior increase of plasma catecholamines during moderate intensity exercise 74 . Limitations. Several limitations of the study are worth noting. Firstly, individual caffeine assimilation was not assessed. Secondly, time of day was not controlled as an independent variable. Therefore, circadian rhythm might have influenced physiological outcomes. Future studies should overcome these limitations.

Conclusions
To conclude, our results reveal that executive and arousal vigilance are affected differently by caffeine and exercise. While caffeine improved arousal vigilance during light and moderate exercise, additionally reducing lapses committed during moderate exercise, it had no effect on executive vigilance. Quite the reverse, moderate exercise improved reaction time on executive vigilance, while arousal vigilance and the attentional networks functions remained unaffected by the exercise manipulation. Additionally, supporting previous theorizing 7 , our results show that vigilance decrement during aerobic exercise also occurs mainly due to a change in response strategy towards more conservative responses, rather than by a loss of sensitivity across time on task.

Data availability
The dataset generated and analyzed during the current study is available in the OSF repository, https://osf.io/ kbypa/.