{"paper_id":"d13b6db0-4172-4d09-a276-1fcfe18e55ce","body_text":"Rapid modulation of choice behavior by ultrasound on the \nhuman frontal eye fields \n \nSoha Farboud1*, Benjamin R. Kop1, Renée S. Koolschijn1, Solenn L.Y. Walstra1,2,  \nJosé P. Marques1, Andrey Chetverikov1,3, W. Pieter Medendorp1, Lennart Verhagen1,4, \nHanneke E.M. den Ouden1,4* \n \n1 Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, \nThe Netherlands \n2 Integrative Model-Based Neuroscience Research Unit, University of Amsterdam, \nAmsterdam, The Netherlands \n3 Department of Psychosocial Science, Faculty of Psychology, University of Bergen, \nNorway \n4 These authors contributed equally  \n \n* Correspondence: soha.farboud@donders.ru.nl & hanneke.denouden@donders.ru.nl  \n \n \nHighlights  \n \n• Focused ultrasound modulates saccades with high spatial and temporal \nprecision \n \n• Inhibitory circuits in the frontal eye fields shape choice computations \n \n• GABA levels predict individual variability in ultrasound-induced \nbehavioral changes \n \n• Ultrasound can be used to probe fast neural dynamics and individual \ndifferences \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nSummary  \nA fundamental challenge in neuroscience is establishing causal brain-function relationships \nwith spatial and temporal precision. Transcranial ultrasonic stimulation (TUS) offers a unique \nopportunity to modulate deep brain structures non-invasively with high spatial resolution, \nbut temporally precise effects and their neurophysiological foundations have yet to be \ndemonstrated in humans. Here, we develop a temporally precise TUS protocol targeting \nthe frontal eye fields (FEFs)  — a well -characterized circuit critical for saccadic eye \nmovements. We demonstrate that TUS  induces robust excitatory behavioral effects. \nImportantly, individual differences in baseline GABAergic inhibitory tone predict response \nmagnitude. These findings establish TUS as a  reliable tool  for chronometric circuit \ninterrogation and highlight the importance of neurophysiological state in \nneuromodulation. This work bridges human and animal research, advancing targeted TUS \napplications in neuroscience and clinical settings. \n \nKeywords  \ntranscranial ultrasonic stimulation (TUS), frontal eye fields (FEF), decision-making, non-\ninvasive brain stimulation, magnetic resonance spectroscopy (MRS), chronometry, GABA, \nsaccadic eye movements, state-dependency, interindividual variability  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nIntroduction \nIn recent years, transcranial ultrasonic stimulation (TUS) has emerged as a promising non -\ninvasive technique for brain stimulation, capable of targeting both cortical and subcortical \nregions with exceptional spatial resolution (Murphy, Nandi, et al., 2024) . This makes TUS \nhighly valuable for studying brain function and offers great potential for therapeutic \napplications. Much of our current understanding is derived from animal studies (Kubanek \net al., 2020; Menz et al., 2013; Mohammadjavadi et al., 2019; Murphy et al., 2022; Yoo et \nal., 2022), but there exists a translational gap to human application. The large majority of \nhuman studies to date focus on repetitive ‘offline’ protocols with temporally sustained \neffects (Riis et al., 2022; Yaakub et al., 2023), while those addressing immediate or ‘online’ \neffects remain limited and often marred by confounds (Kop et al., 2024; exception Butler et \nal., 2022). This leaves questions on the physiological mechanisms and temporal dynamics \nof TUS unanswered and calls for robust and replicable protocols in humans . In this study, \nwe introduce an effective online TUS protocol for humans  with immediate effects , by \nleveraging a well -established TUS protocol from non -human primates  (Kubanek et al., \n2020). To this end , we take advantage of an evolutionarily conserved brain circuit with a \nwell-characterized link to readily measurable behavior that acts as a model system for more \ncomplex decision-making – the frontal eye fields (FEFs).    \n \nThe role of the FEFs in the planning and generation of saccadic eye movements has been \nwell established in both humans and non-human primates (Paus, 1996; Vernet et al., 2014), \nand features a basic topographic representation that encodes both the direction and \namplitude of saccades in the opposite visual hemifield (Paus, 1996; Vernet et al., 2014) . \nFEF’s involvement in contralateral saccade generation has been further evidenced by lesion \nstudies (Gaymard et al., 1999; Guitton et al., 1985; Henik et al., 1994; Rivaud et al., 1994 ) \nand transcranial magnetic simulation experiments (Grosbras & Paus, 2002, 2003; Nagel et \nal., 2008; Nyffeler et al., 2006; Ro et al., 1997, 1999, 2002; Thickbroom et al., 1996) . This \nwell-characterized role of the FEF in contralateral saccades allows for precise \ncharacterization of TUS effects. For instance, in macaques online TUS of the FEF was found \nto bias saccades towards the contralateral side, which suggests that stimulation has a net \nexcitatory effect (Kubanek et al., 2020). However, it is not known whether these results can \nbe directly translated to humans, i.e. whether online TUS stimulation of the human FEF can \ninduce the same excitatory effect, or whether anatomical , physiological, and behavioral  \ndifferences between humans and non-human primates would instead result in  net \ninhibitory or perturbatory effects. Indeed, the effect of TUS has been found to vary from \nexcitatory to inhibitory to perturbatory depending on the specific stimulation protocol \nsettings (Nandi et al., 2024) , underscoring the need for caution when interpreting TUS -\ninduced behavioral changes. \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nConsidering that the effects of brain stimulation are highly dependent on brain states and \ntraits (Guerra, Asci, et al., 2020; Guerra, López -Alonso, et al., 2020; López -Alonso et al., \n2014; Pellegrini et al., 2018b, 2018a), it is expected that TUS effects vary not only between \nspecies, but  also between individuals. Consequently, it is pertinent to consider the \nindividual neurophysiological state when investigating the mechanisms and consequences \nof TUS. This interindividual variability may be influenced by factors such as an individual’s \ncortical inhibitory tone, which ha s been shown to impact the effects of other noninvasive \nstimulation methods (Stagg et al., 2011). Moreover, differences in cortical inhibitory tone in \nthe FEFs have been linked to individual variations in the capacity to resist distractions while \ngenerating saccades  (Sumner et al., 2010) . Therefore, it seems likely that the \nneuromodulatory effects of TUS on an individual may have different effects. Given that TUS \nmay modulate both excitatory and inhibitory neuronal populations in the brain, we \nhypothesize that the net effects of TUS could be shaped by individual differences in the \nexcitation/inhibition balance. To explore whether interindividual differences in the effects \nof TUS are similarly inhibitory tone -dependent, we measured individual level  \nconcentrations of the inhibitory neurotran smitter GABA + in the FEF  using magnetic \nresonance spectroscopy (MRS).  \n \nIn the present study, we tested the hypothesis that TUS applied to the human FEF has an \nimmediate, excitatory effect on saccade direction, and that this effect is mediated by local \ninhibitory tone. Participants completed a simple saccade choice task while receiving TUS \nduring stimulus presentation, applied to either their left or right FEF (‘stimulation’), or to the \nleft or right hand motor cortex (M1) (‘active control’). FEF TUS induced a significant increase \nin the selection of contralateral saccades, directly replicating findings from a previous study \nin macaques (Kubanek et al., 2020) and indicating that FEF TUS has net excitatory effects on \nsaccade selection in humans. Notably, participants’ characteristic inhibitory tone in FEF was \nfound to predict inter-individual differences in the effect of TUS, suggesting that TUS \nsusceptibility is linked to an individual’s inhibitory tone. Taken together, our findings pave \nthe way to use TUS as an effective and temporally specific tool to study the functional circuit \ndynamics of the human brain  and offer critical insights into the factors that drive \ninterindividual differences in response to this neuromodulation technique.  \n \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n2. Results \n \n2. 1 Baseline saccade task behavior \nThirty-five right-handed participants (Mage = 24.1, SDage = 3.2, range = 20 – 32; 15 females, \n20 males) performed a saccadic decision task in which two visual stimuli were presented \nasynchronously and equidistantly on either side of fixation  (Figure 1A). Participants were \ninstructed to saccade as quickly as possible to the stimulus that appeared first (i.e. target). \nWe examined  the probability of making a rightward saccade across all stimulus onset \nasynchronies (SOA, i.e. delay between target and distractor). Participants performed well \non the task i n the baseline (sham) condition : When the target is on the right, participants \nwere more likely to make a rightward saccade (b = 14.1, 95%-CI [13.0, 15.4], c 2 = 530, P < \n0.001, Figure 1B). At the group-level, there is a lack of a noticeable rightward or leftward \nbias in the sham condition, although within participants there is variability in baseline side \nbias (Figure 1B).   \nWe expected TUS effects to surface primarily in biasing responses on trials with short \nSOAs (hereafter referred to as the choice domain) , i.e. when sensory evidence is \nambiguous, instead of on trials with overwhelming sensory evidence. In the former case, \nboth FEFs compete to drive the saccade, and TUS could ‘nudge’ the participant’s response \nin the opposite direction. Therefore, we oversampled trials with shorter SOAs, and focused \nthe primary analysis on trials with SOAs where participants were <75% correct (Figure 1C).  \n \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \nFigure 1 | Study design and baseline behavioral results \n(A). Saccade task design. In each trial, two targets appeared on opposite sides of the screen after a variable \nstimulus onset asynchrony (SOA; 0–200 ms). Participants were instructed to look at the target that appeared first \nand received feedback based on performance: +10 points for correct responses, -10 points for incorrect \nresponses, and 0 points for correct but late responses (after 1500 ms). Transcranial ultrasonic stimulation (TUS) \nwas applied to either the left or right hemisphere, paired with an auditory masking sound, or a sham stimulation \nwith the same auditory stimulus. (B) Baseline behavioral  performance. Performance followed a typical \npsychometric sigmoid function, with lower accuracy for shorter SOAs.  We expected TUS effects to surface \nprimarily in biasing responses on trials with short SOAs, when sensory evidence is low, and TUS could ‘nudge’ \nthe participant’s response in the opposite direction. Therefore, we oversampled trials with shorter SOAs, and \nfocused the primary analysis on trials with delays where participants were <75% correct (marked in blue). Black \nline represents the group-level curve with error bars indicating standard error of the mean (S.E.M.).  Data are \nbinned for visual purposes into intervals of 0 to 1, 1 to 26, 26 to 50, 50 to 75, 75 to 100, 100 to 142, and 142 to \n200 ms; bins are symmetric for negative values.  Grey lines represent individual subject curves.  (C) SOA \ndistribution. The distribution of SOAs ranged from 0-200 ms. Shorter delays were oversampled, following our \nhypothesis that TUS would affect behavior in the choice domain, when uncertainty is high. The blue bars are a \nsimplified visual representation of SOAs that fall within the  choice domain. (D) Stimulation protocol. TUS was \ndelivered for 500 ms per trial, starting at onset of the first target. Each pulse followed a sinusoidal wave shape, \nramping up and down within 1 ms, with a pulse repetition frequency of 500 Hz. The intensity in free water (ISPPA) \nwas 25 W/cm², and the fundamental frequency was 250 kHz. Stimulation conditions included TUS applied to \nthe left or right frontal eye fields (FEF) with auditory masking, TUS applied to left or right motor cortex (M1) with \nbreak\n5 min\nbreak\n5 min\nbreak\n5 min\npractice trials (no TUS, no sound)\npadding trials (no TUS, no sound)\nrandomized left TUS (1/3), right TUS (1/3) and sham (1/3; no TUS) trials\n[-300, -600]\nfixate target 1 target 2 choice feedback\n-250 t = 0 [0, 200] 500 [0 - 1500]\ntime (ms)\nA B\n-200 -125 -58 0 58 125 200\nSOA delay bins\nC\nP(rightward saccade)\nleft FEF\nright FEF\nP(rightward saccade) P(rightward saccade) P(rightward saccade)\nF\nED\nsaccade task masking assessment\nintake session\ntask practice MRI\nTUS session 1\nno TUS\n left, right & shamTUS:\nblocks: FEF-M1-M1-FEF\nblocks: M1-FEF-FEF-M1\nleft, right & shamTUS:\nblocks: FEF-M1-M1-FEF\nblocks: M1-FEF-FEF-M1\nleft, right & shamTUS:\nblocks: FEF-M1\nblocks: M1-FEF\nsaccade task\nTUS session 2\nstructural: T1w, T2w, UTE\nMRS: left FEF & left M1\nHypotheses\nSOA\n-200 2000\n0.00\n1.00\nexcitatory\nSOA\n-200 2000\ninhibitory\nSOA\n-200 2000\nperturbatory\nStimulation protocol & conditions\nBaseline (sham) behaviour SOA distributionSaccade task\nStudy procedure\ntrial counts\nsham\n0.5 1.00.00.5 1.00.00.5 1.00.0\nTUS\nauditory mask\n1 ms\nIsppa\n 25\nW/cm2\n2 ms (500 Hz)\n500 ms\n250 kHz\nlocalizer: FEF & M1\nTUS blockTUS blockTUS block\nTUS block\n0.00\n0.25\n0.50\n0.75\n1.00\n−100 0 100\nSOA (ms)\nP(rightward saccade)\nchoice\ndomain\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nmasking, and sham stimulation with masking. (E) Hypothesized TUS effects. We assessed three potential effects \nof TUS: (i) Net excitatory effect, i.e. increased contralateral saccades, in line with previous findings (Kubanek et \nal., 2020); (ii) Net inhibitory effects, i.e. increased ipsilateral saccades; or (iii) Perturbation, i.e. overall reduced \naccuracy, resulting in increased variance . (F) Study design. Participants completed one intake and two TUS \nsessions. In the intake session, they practiced the saccade task for 20 minutes. They then entered the MRI \nscanner, where structural scans were obtained for neuronavigation and acoustic simulations, and functional \nlocalizers were used to identify individual FEF and M1 stimulation sites. Baseline GABA+ levels were measured \nin the left FEF and left M1 with MRS. In each TUS session, participants performed the saccade task for 60 minutes \n(4 blocks of 15 minutes). Each block involved stimulation of either FEF or M1. For each block, the distribution of \ntrials was 33% left TUS, 33% right TUS, 33% sham. All blocks were padded with trials where no auditory mask \nwas presented as wash-in/wash-out trials that were not of interest. Block order was counterbalanced across \nparticipants. At the end of the final session, participants completed a masking assessment to test the \neffectiveness of auditory masking. They received stimulation of either the left/right FEF, left/right M1, or sham \nand were asked to identify whether they were stimulated and, if so, on which side. \n \n2.2 FEF-specific TUS effects show robust contralateral bias dependent on GABA+ levels \nUltrasonic stimulation of both the left and right FEFs significantly increased contralateral \nsaccades (Figure 2C; b = -0.25, 95%-CI [-0.40, -0.10], c 2 = 10.3, p = 0.001). This finding \naligns with our hypothesis that the protocol induces excitatory behavioral effects , and \nreplicates prior findings observed in non -human primates (Kubanek et al., 2020) . This \nexcitatory behavioral effect on contralateral saccades was not observed for stimulation to \nleft versus right M1 (details reported below). These results highlight the specificity of the \neffects to the FEFs and provide robust evidence of direct TUS-induced behavioral changes \nin humans. \n There was substantial interindividual variability both in baseline (sham) directional \nbias ( Figure 1B) as well as in the susceptibility of saccade direction to TUS stimulation \n(Figure 2 C). Therefore, we next asked whether the baseline neural inhibitory tone in \nparticipants’ FEF could explain interindividual differences in TUS susceptibility . Note that \nwe measured only left hemispheric MRS (in FEF and M1, for details see methods). We found \nthat changes in saccade bias induced by left FEF TUS relative to sham were predicted by \nbaseline FEF GABA + levels (condition (left FEF/sham) x FEF GABA +: b = -0.21, 95%-CI [-\n0.39, -0.04], c 2 = 5.6, p = 0.017; Figure 2E). Specifically, higher baseline GABA+ levels in \nthe left FEF were associated with a stronger rightward bias on sham trials (sham x FEF \nGABA+: b = 0.1 4, 95%-CI [0.0 4, 0.24], c 2 = 7.0 , R2 = 0.096, p = 0.008; Figure 2F, top). \nImportantly, following TUS stimulation, this relationship of baseline GABA + and rightward \nbias disappeared (left FEF x FEF GABA +: b = -0.08, 95%-CI [-0.22, 0.07], c 2 = 1.1, p = 0.3; \nFigure 2 F, bottom). Thus, TUS increased contralateral responding predominantly in \nparticipants with lower baseline GABA+ levels in the FEF (voxel placement: Figure 2G).  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \nFigure 2 | Main behavioral TUS effects \n \n(A) Individual FEF localization. BOLD responses of left and right FEF for a single subject. Participants performed a \nP(rightward saccade)\n0.25 0.50 0.75\n*\n0 25\n0.54 0.90\n−0.50\n0.00\n0.50\n2 3 4 5\nP(rightward saccade)\nGABA+ left FEF (i.u.)\nP = 0.022\nL FEF TUS - sham\n0.3\n0.5\n0.7 sham\nP = 0.012\n0.3\n0.5\n0.7 L FEF TUS\nP = 0.3\n2 3 4 5\nGABA+ left FEF (i.u.)\nP(rightward saccade)\nleft FEF\nright FEF\nsham\n0.54 0.90\nTUS of the FEFs results in increased contralateral responses\nBaseline FEF GABA levels predict interindividual variability in FEF TUS effects\nBOLD (a.u.)\n2.50 5.00Isppa (W/cm2)\nBOLD (a.u.)\nparticipants\nC D\nE F G\nP(rightward saccade)\nSOA (ms)\n-100 0 100\n1.00\n0.75\n0.50\n0.25\n0.00\n0.80\n0.60\n0.40\n0.20\nleft FEF\nright FEF\n−40 −30 −20 −10 0 10 20 30 40\nSOA (ms)\nIndividual FEF localization and targeting\nA B\nleft FEF right FEF\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n  \n \n2.3 M1 TUS does not affect saccade choices, demonstrating FEF-specific modulation  \n \nTo ensure that the observed effects of TUS on saccade direction were specific to the FEF, \nin half of the stimulation trials the left and right M1 (hand area) were stimulated as control \nregions. Similar to the FEF stimulation, we used an fMRI functional localizer to determine \nthe participant specific target for M1 TUS ( Figure 3 A). Again, p ost hoc acoustic wave \npropagation simulations confirmed that we successfully targeted these regions (Figure 3B). \nAs expected, left and right M1 TUS did not induce significant differences in contralateral \nsaccades, further supporting the specificity of the observed effects to the FEF and ruling \nout potential confounds (b = -0.09, 95%-CI [-0.22, 0.04], c 2 = 1.8, p = 0.12; Figure 3C).  \nHowever, at delay 0, where both targets appear on the screen simultaneously, we \nobserved a qualitative difference in rightward saccade probability in M1 stimulation (Figure \n3D) that mirrors the pattern of FEF stimulation. This observation is further explored in the \nsupplementary documents (Supplementary Documents S.1). For consistency across all \nanalyses, we selected the choice domain data without these delay-0 trials. \nCritically, a formal side-by-region (FEF/M1) comparison revealed a significant interaction (b \n= 0.23, 95%-CI [0.03, 0.43], c 2 = 5.1, p = 0.025) between stimulation side (left/right) and \nregion (FEF/M1). This indicates that the TUS effects influencing choice bias and saccade \nbehavior are specific to FEF stimulation and not to M1 stimulation . This excludes  the \npossibility that these effects are driven by confounds such as auditory or somatosensory \nstimulation. This finding reinforces the conclusion that TUS selectively modulates behavior \nfunctional localizer during the intake session in which they alternated between left/right saccades and fixation, \nallowing for individual localization of the FEFs. (B) Acoustic simulation of TUS wave propagation. Acoustic simulations \nof left and right FEF for a single subject are shown. The simulation depicts the estimated intracranial intensity (Isppa) \nwith an intensity cutoff at the full width at half maximum (FWHM). (C) FEF TUS effects. Choice-domain average effects. \nGrey dots represent individual partic ipants’ mean saccadic directions within the choice domain. Colored dots \nrepresent group means, error bars indicate the standard error of the mean (S.E.M.), with a statistically significant \ndifference between left and right FEF stimulation (p = 0.001). (D) FEF TUS effects. Stimulation of the left and right \nfrontal eye fields (FEF) led to increased contralateral saccades, particularly within the choice domain (highlighted in \nlight blue, bottom). Compared to each other, left FEF stimulation produced more rightward saccades, w hile right \nFEF stimulation led to more leftward saccades. Data are binned for visual purposes into intervals of 0 to 1, 1 to 26, \n26 to 50, 50 to 75, 75 to 100, 100 to 142, and 142 to 200 ms; bins are symmetric for negative values. Dots represent \nthe group mean per bin, and error bars indicate the S.E.M. across participants.  (E) FEF GABA+ predicts FEF TUS \neffects. The relationship between baseline GABA + levels in the left FEF and the effect of TUS on saccadic bias, \ncalculated as the difference in probability of making a rightward saccade between left FEF TUS and sham conditions. \nHigher baseline FEF GABA+ levels correlate with a weaker TUS effect on rightward saccades (p = 0.022). The line is \na linear fit with a 95% confidence interval, e ach dot represents a participant.  (F) FEF GABA+ predicts baseline \nsaccade behavior. Top: baseline left FEF GABA + levels significantly correlate with rightward saccade probability \nduring sham stimulation alone ( p = 0.012). Bottom: under left FEF TUS, this correlation with rightward saccade \nprobability is absent (p = 0.3).  The line is a linear fit with a 95% confidence interval, each dot represents a participant, \neach dot represents a participant.  (G) MRS voxel placemen t. Magnetic resonance spectroscopy (MRS) voxel \nplacement for measuring GABA+ concentrations in the left frontal eye field (FEF). Color overlays represent GABA+ \nconcentration distributions in each region.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nvia its impact on the FEFs. Finally, between-variability in the effects of TUS in M1 in saccade \ndirection could not be explained by subjects’ baseline GABA+ levels in M1 (condition (left \nM1/sham) x M1 GABA+: b = 0.06, 95%-CI [-0.14, 0.26], c 2 = 0.3, R2 = 0.1, p = 0.5; Figure \n3E-G). These findings underscore the specificity of the TUS effects to the FEFs and provide \nadditional evidence against potential confounds in the study design. \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \nFigure 3 | Control analyses: M1 TUS effects \n \n(A) Individual M1 localization. BOLD responses of left and right M1 for a single subject. Participants performed \na functional localizer during the intake session in which they alternated between left - and right-hand tapping \nmovements (index finger and thumb), allowing for individual localization of the hand area in M1. (B) Acoustic \n0.54 0.90\nBOLD (a.u.)\n2.50 5.00Isppa (W/cm2)\nIndividual M1 localization and targeting (control region)\nA B\nP(rightward saccade)\nSOA (ms)\nM1 TUS does not influence saccade direction\nM1 GABA does not predict M1 TUS effects\nC D\nE F G\nP(rightward saccade)\n0.25 0.50 0.75\nns\nleft M1\nright M1\nsham\n−0.50\n0.00\n0.50\n2 3 4 5\nGABA+ left M1 (i.u.)\nP(rightward saccade)\n2 3 4 5\n0.3\n0.5\n0.7\n0.3\n0.5\n0.7\nGABA+ left M1 (i.u.)\nP = 0.5\nP = 1.0\nP = 0.4\nL M1 TUS\nshamL M1 TUS - sham\n0 25\n0.54 0.90\nBOLD (a.u.)\nparticipants\nP(rightward saccade)\nleft M1\nright M1\nleft M1 right M1\n-100 0 100\n1.00\n0.75\n0.50\n0.25\n0.00\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nsimulation of TUS wave propagation Acoustic simulations of left and right M1 for a single subject are shown. The \nsimulation depicts the estimated intracranial intensity (I sppa) with an intensity cutoff at the full width at half \nmaximum (FWHM).  (C) M1 TUS effects.  Choice-domain average effects. Grey dots represent individual \nparticipants' average saccadic direction within the choice domain. Colored dots represent group means, error \nbars indicate the standard error of the mean (S.E.M.) with no statistically significant group effects observed (p = \n0.12). (D) M1 TUS effects. Stimulation of the left and right M1 did not result in significant shifts in contralateral \nsaccades within the choice domain (highlighted in light blue, bottom). Compared to each other, left and right \nM1 stimulation showed no differences in saccadic direction. Data are binned for visual purposes into intervals \nof 0 to 1, 1 to 26, 26 to 50, 50 to 75, 75 to 100, 100 to 142, and 142 to 200 ms; bins are symmetric for negative \nvalues. Dots represent the group mean per bin, and error bars indicate the S.E.M. across participants. (E) M1 \nGABA+ does not predict TUS effects. Baseline GABA+ levels in the left M1 do not correlate with TUS-induced \nsaccadic bias, calculated as the difference in probability of making a rightward saccade between left M1 TUS \nand sham conditions ( p = 0.5). The line is a linear fit with a 95% confidence interval, each dot represents a \nparticipant. (G) M1 GABA+ does not predict baseline saccade behavior. Top: baseline left M1 GABA+ levels do \nnot significantly correlate with rightward saccade probability during sham stimulation alone (p = 1.0). Bottom: \nunder left M1 TUS, this correlation with rightward saccade probability remains absent ( p = 0.4). The line is a \nlinear fit with a 95% confidence interval, each dot represents a participant. (H) MRS voxel placement. Magnetic \nresonance spectroscopy (MRS) voxel placement for measuring GABA+ concentrations in the left motor cortex \n(M1). Color overlays represent GABA+ concentration distributions in each region. \n \n \n \n2.4 Control and follow-up analyses \nTUS does not globally perturb performance \nIn order to assess potential perturbatory effects of FEF TUS on performance, we completed \na regression with ‘correct response’ on TUS and sham trials in the FEF blocks as dependent \nvariable and side, region and delay as independent variables. Note that again zero-delay \ntrials are excluded from this analysis because no correct response can be defined. For a \nbinary choice task, sensory noise is directly reflected in the overall accuracy (i.e., the slope \nof the psychometric curve is inversely related to the va riance of the underlying signal \nprobability distribution). There was no perturbatory effect of TUS on performance (TUS vs. \nsham: b = 0.07, 95%-CI [-0.13, 0.27], c 2 = 0.5, p = 0.5). \nAdditionally, we performed supplementary analyses (Supplementary Documents \nS.2) examining estimation of bias, including effect size in decision curve shift (horizontal \nbias), slope, and lapse rate to confirm that the observed TUS effects were specific to bias \nand not confounded by changes in slope or lapse rate. \nOnline TUS effects are immediate and short-lived \nHaving demonstrated that TUS of FEF has an excitatory effect and that this effect is specific \nto stimulation of FEF, we  next assessed the duration and persistence of TUS effects on \nsaccade direction. This is critical to characterize the temporal dynamics of ultrasonic \nneuromodulation, in terms of how fast effects arise, and whether they persist into the next \ntrial. Slow and sustained effects suggest early -phase plasticity mechanisms to drive the \nobserved behavior, while fast and temporally precise effects suggest modulation of spiking \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nactivity. The latter would enable TUS to be used for cognitive chronometry: to disentangle \nthe functional contributions of brain regions and circuits across time.  \nFirst, we examined whether there were any carryover effects of stimulation on sham \ntrials that followed TUS trials.  First, when analyzing sham trials during FEF blocks, no \nsignificant carry-over effects were observed. More specifically, saccade direction was not \naffected by the side stimulated on the preceding trial (FEFt-1 (left/right): b = 0.15, 95%-CI [-\n0.07, 0.37], c 2 = 1.8, p = 0.18; BF01 = 2). However, when pooling together sham trials during \nFEF and M1 blocks, participants made significantly more ipsilateral saccades on these trials, \ndirecting their saccades toward the side stimulated on the preceding trial (sidet-1: b = 0.15, \n95%-CI [-0.08, 0.38], c 2 = 5.3, p = 0.021). Crucially, this ipsilateral bias did not differ between \nFEF and M1 stimulation (sidet-1 x region (FEF/M1): b = 0.09, 95%-CI [-0.23, 0.41], c 2 = 0.3, p \n= 0.6, BF01 = 12; Figure 4A), and can therefore not explain the observed specific effects on \ncontralateral saccades following FEF (but not M1) TUS. In Supplementary Documents S.3, \nwe will briefly further discuss the non-specific (potentially attentionally driven) after-effects.   \nFinally, to assess the immediacy of TUS effects relative to stimulation onset, we \nquantified TUS effects on the fastest saccades, defined as trials with a saccade reaction time \nbelow the median of 265 ms.  Even on this subset of trials where participants received less \nthan 26 5 ms of stimulation prior to saccade onset, TUS significantly shifted saccade \ndirection contralaterally (FEF (left/right): b = -0.32, 95%-CI [-0.59, -0.07], c 2 = 6.1, p = 0.013). \nIn contrast, no significant saccade bias was observed for left versus right M1 stimulation on \nfast trials (M1 (left/right): b = -0.15, 95%-CI [-0.38, 0.07], c 2 = 1.8, p = 0.19). Taken together, \nour results highlight the specificity and speed of TUS effects on saccade direction, \nreinforcing that they are immediate, fast and specific to the FEF. \nMasking assessment \nTo estimate the potential impact of auditory or somatosensory confounds (Braun et al., \n2020; Guo et al., 2018; Johnstone et al., 2021; Kop et al., 2024; Sato et al., 2018) , we \nincluded a masking assessment at the end of the second TUS session ( Figure 1F). This \nassessment allowed us to verify that potential confounds could not explain the observed \ndissociation of TUS effects over FEF versus M1. Participants received stimulation (or sham) \nrepeatedly either over FEF or M1 (in blocks) , all with an auditory mask , and reported i) \nwhether they perceived stimulation, and ii) on what side (forced choice, left vs. right).  First, \nparticipants reported perceiving stimulation more frequently on TUS trials compared to \nsham (stimulation (TUS/sham): b = 2.7, 95% -CI [1.9, 3.4], c 2 = 53.6, p < 0.001). Crucially, \nhowever, this ability to detect TUS versus sham did not differ between conditions (region \n(M1/FEF): b = 0.25, 95%-CI [-0.36, 0.86], c 2 = 0.7, p = 0.4, BF01 = 4.1; Figure 4B). Second, \non TUS trials, participants were biased to report perceiving stimulation contralaterally to the \nside of actual stimulation (side: b = -1.2, 95%-CI [-2.1, -0.3], c 2 = 7., p = 0.008, BF01 < 0.001; \nFigure 4C). However, again this contralateral reporting bias was not significantly different \nbetween FEF and M1 stimulation (side (left/right) x region (FEF/M1): b = -0.5, 95%-CI [-1.5, \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n0.5], c 2 = 1 .1, p = 0.3). In Supplementary Documents S.3 , we further discuss of this  \ncontralateral TUS perception in the context of lateralized non-specific aftereffects reported \nabove. Taken together, while putative confounding factors were present in our study, these \nwere identically for the FEF and M1 conditions, and thus crucially cannot explain our main \nfindings. More broadly, this masking assessment confirms the presence of putative \nconfounding factors and emphasize the importance of active control conditions for online \nTUS protocols.  \n \n \n \n \n \nFigure 4 | Effects of online TUS and masking assessment \n \n(A) TUS after-effect assessment on sham trials . Each sham trial was labeled based on the preceding trial's \nstimulation condition (e.g., \"L FEF à sham\" indicates a sham trial following left FEF stimulation). Dots represent \ngroup mean saccade bias on sham trials that followed stimulation trials; error bars represent the standard error \nof the mean (S.E.M.) . Participants made significantly more ipsilateral saccades on sham trials following \nstimulation (p = 0.021), but importantly and unlike the main TUS effect, this effect was not specific to FEF (side \nx region: p = 0.6). (B) Masking, perceived stimulation (yes/no). Probability of reported stimulation perception \nacross sham, FEF, and M1 conditions. Density clouds represent participant distributions, bars indicate group \nmeans, and error bars show the S.E.M. Participants were significantly more likely to perceive stimulation during \nTUS trials compared to sham trials (p < 0.001), highlighting that sham conditions alone may not fully account \nfor TUS effects. No significant difference was observed between FEF and M1 conditions (p = 0.4). (C) Masking, \nperceived stimulation side (left/right) . Probability of reported stimulation side perception in the masking \nassessment. Density clouds represent participant distributions, bars indicate group means, and error bars show \nthe S.E.M. Participants were more likely to perceive TUS contralateral to the actual stimulation site (p = 0.008). \nThis effect was consistent across FEF and M1 regions, as indicated by the absence of a significant side-by-region \ninteraction (p = 0.3), supporting the robustness of the active control design. \n \n \n \n \n  \nA\nL FEF - sham\nL M1 - sham\nR FEF - sham\nR M1 - sham\nsham - sham\nP(rightward saccade)\n0.25 0.50 0.75 ns\nP(left side response)\n0.00 0.25 0.50 0.75 1.00\nL FEF\nR FEF\nL M1\nR M1\n0.00\n0.25\n0.50\n0.75\n1.00P(stim response)\nFEF M1sham\nns\n**\n**\n* * nsB C\ncondition\nt       t+1\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nDiscussion \n \nThis study provides evidence for effective online transcranial ultrasound stimulation (TUS) \non saccadic decision-making in humans. These results advance our understanding of the \nunderlying neural mechanisms contributing to interindividual differences. We found that \nshort TUS pulse trains (500 ms) to the frontal eye fields (FEF) but not the primary hand motor \ncortex (M1) have immediate, short lived effects promoting contralateral saccades. \nImportantly, the effect of FEF TUS depends on individual inhibitory tone, as indexed with \nmagnetic resonance spectroscopy  (MRS). These findings provide  behavioral and  \nneurophysiological evidence for the direct effects of TUS on human brain function, \nestablishing its potential as temporally specific neuromodulatory tool for advancing \nfundamental neuroscience and enhancing our understanding of temporally dynamic brain-\nbehavior relationships. \n \nFEF-specific TUS effects are fast and show robust contralateral bias  \nIn recent years, evidence has emerged for sustained and early -phase plasticity effects of \nTUS in humans  (Nakajima et al., 2022; Riis et al., 2022; Yaakub et al., 2023) . Despite \nadvances in in -vitro and animal models demonstrating immediate neural effects of TUS \n(Kubanek et al., 2020; Mohammadjavadi et al., 2019; Yoo et al., 2022) , the translation of \nthese findings to humans has remained scarce (Butler et al., 2022) or potentially marred by \nconfounds (Kop et al., 2024) .  To address this, we adapted a well -established animal TUS \nprotocol for human application, targeting the left and right FEF while participants \nperformed a saccade choice task. This approach allowed us to assess the immediate \nbehavioral effects of TUS on saccadic choices. \nPrevious lesion and brain stimulation studies have demonstrated that the FEF \nmediates contralateral saccade generation (Gaymard et al., 1999; Grosbras & Paus, 2002, \n2003; Guitton et al., 1985; Henik et al., 1994; Nagel et al., 2008; Nyffeler et al., 2006; Rivaud \net al., 1994; Ro et al., 1997, 1999, 2002; Thickbroom et al., 1996). Here, we reveal that TUS \nover FEFs increased the selection of contralateral saccades, particularly during trials that \nrequired FEF-mediated resolution of saccade conflict (Figure 2C). Therefore, this protocol \nexerts a net facilitatory effect on FEF activity. Our findings replicate earlier work in non -\nhuman primates (Kubanek et al., 2020).  \nTo demonstrate the spatial specificity of  TUS, and to rule out the possibility that \nauditory or somatosensory confounds could drive the observed effects, we alternated \nblocks of FEF stimulation with blocks targeting an active control site – the hand area of the \nprimary motor cortex (M1). Importantly, we observed no significant changes in saccadic \nbehavior following M1 stimulation (Figure 3C).  \nOur results show that TUS biases responses on trials with short SOAs, where there \nwas high uncertainty about the correct response  (Figure 2D). On those trials, which we \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\npurposefully oversampled, we hypothesized FEF activity is more sensitive to be ‘nudged’ \nby TUS to tip the balance in the opposite direction. This effect supports the hypothesized \nrole of TUS in modulation of ongoing FEF computations, rather than the direct induction of \nsaccades. TUS employs sound waves that are beyond the audible spectrum to mechanically \nengage with neuronal tissue, influencing characteristics such as membrane capacitance and \nthe activity of mechanosensitive ion channels . Mechanistically, unlike magnetic or electrical \nstimulation techniques that can directly induce neural firing (MEP: Barker et al., 1985; \nVEP/phosphenes: Kammer, 1998; Murd et al., 2010) , TUS does not directly evoke neural \nfiring but rather modulates ongoing neural activity through subthreshold modulation and \nsubtly ‘nudges’ neural activity without inducing immediate or large-scale neural responses \n(Darmani et al., 2022).  \nFinally, we established the high temporal specificity of this TUS protocol by showing \nit has immediate effects that occur only during stimulation trials, and do not persist into \nfollow-up sham trials but ( Figure 4A). This highlights the specificity of TUS effects to the \nstimulation period itself. Notably, these effects emerge very rapidly: even in trials with the \nfastest reaction times, where participants made saccades before the TUS duration was \ncomplete (i.e. less than 265 ms), the TUS effect was still evident and specific to the FEF. This \nhigh temporal specificity is crucial for using TUS as a tool to probe brain function with \nmillisecond precision, aligning with principles of mental chronometry to better understand \nthe neural dynamics underlying rapid cognitive and motor processes. \n \nTUS effects are dependent on individual baseline GABA+ levels \nThe bias toward contralateral saccades suggests a net facilitatory effect of TUS on the FEF. \nHowever, this does not necessarily imply that TUS directly excites neuronal tissue or targets \nspecific neuron types. TUS can influence both excitatory and inhibitory neurons by altering \naction potential thresholds (Jerusalem et al., 2019; Yoo et al., 2022; Yu et al., 2021) . The \neffects of TUS may, therefore, depend on the baseline state of the neuronal populations \ninvolved (Heinen et al., 2014; Huang et al., 2017; Kamke et al., 2014, 2014; Massimini et al., \n2005; Siebner et al., 2022; Stagg et al., 2011) . Interestingly, our data indicate that such \nbaseline differences in neuronal state indeed modulate the extent of TUS effects.  \nAt baseline, participants exhibited individual biases in saccade direction, with some \nshowing a preference for leftward and others for rightward saccades ( Figure 1B). These \nbehavioral baselines were mirrored by neural differences, as individuals with  a lower \ninhibitory tone, as quantified by GABA+ levels in the left FEF , tended to have a stronger \nintrinsic leftward bias (Figure 2F, top). While this might appear counterintuitive, it aligns \nwith earlier findings from FEF MRS work (Sumner et al., 2010) , which demonstrated that \nhigher GABA+ levels are associated with better suppression of distractors. In this model, \ncortical regions execute actions via excitatory neurons while inhibiting competing cortical \nregions — a process regulated by local inhibitory interneurons. Furthermore, in other motor \nregions, GABA+ has previously been shown to play a critical role in behavior in its relation \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nto motor learning through action plan tuning and long-term potentiation (Kolasinski et al., \n2019; Stagg et al., 2011) and FEF GABA+ predicts saccade behavior (Sumner et al., 2010). \nOur findings support the framework proposing that local inhibitory interneurons \nplay a role in distraction suppression and ultimately saccade execution  (McSorley et al., \n2006; Sumner et al., 2010). Individuals with higher GABA+ levels in the left FEF were better \nable to inhibit competing right FEF projections, resulting in a stronger rightward bias at \nbaseline. Furthermore, baseline GABA + levels predicted the magnitude of the TUS -\ninduced behavioral effects such that individuals with lower GABA + levels and an intrinsic \nleftward bias showed the largest TUS -induced behavioral changes ( Figure 2 E). This \nsuggests that TUS acts to normalize interindividual differences in physiological states. This \nnormalization effect, which brings individuals closer to a common excitatory/inhibitory \nbalance, underscores the state-dependent nature of TUS and highlights the importance of \nconsidering baseline neural states when interpreting its effects. Similar to other non -\ninvasive brain stimulation methods, the variability in TUS effects appears to be modulated \nby the baseline excitatory/inhibitory balance of the targeted neural populations. \n \nFinal considerations and future directions \nNotably, the effects of TUS observed in humans are less pronounced than those \ndocumented in earlier animal work (Kubanek et al., 2020). Several reasons for this can be \nconceived. First, Macaques generally exhibit faster saccades, and consistent with this, we \nobserved strong and reliable TUS effects in trials with shorter reaction times. Second, the \nFEF circuitry in animals is more lateralized compared to humans  (Hutchison et al., 2012) , \nThird, human FEF is larger in absolute terms, reducing the relative territory of FEF reached \nby the TUS focus. This might lead to a reduced efficacy of stimulation, as the average TUS \nintensity across the entire FEF will be lower in humans than in macaques, even with \ncomparable peak intensities.  \nTUS is associated with auditory and somatosensory confounds, that could putatively \ndrive behavioral effects masquerading as effects of TUS  (Braun et al., 2020; Kop et al., \n2024). Therefore, in this study we included a number of careful controls to exclude this \npossibility. Importantly, the observed effects on contralateral saccades were specific to TUS \nover FEF, and not observed following M1 TUS. Crucially, however, confound effects were \nnot different between these conditions, as evidenced by a masking assessment following \nthe final session. First, while participant s were able to differentiate between sham and \nstimulation conditions, this did not differ between  M1 and FEF TUS (Figure 4B). Second, \nwhile participants were able to distinguish left versus right TUS stimulation , again this did \nnot differ between M1 and FEF TUS (Figure 4C). Taken together, these potential confound \neffects cannot explain the main finding. Nevertheless, these observations emphasize the \nimportance of including an active control condition rather than relying solely on a sham \ncondition in TUS studies.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nA final point of consideration is that while this study investigated baseline GABA + \nlevels, we did not record GABA+ as an outcome variable of TUS. Instead, our focus was on \nthe immediate effects of TUS and its relationship to individual differences. A key direction \nfor future research would involve assessing MRS-based changes in GABA+ to examine the \npotential for longer-lasting plasticity effects (Yaakub et al., 2023). \n \nIn summary, we demonstrated we can bias human choices using fast ultrasonic \nneuromodulation.  Transcranial ultrasound over the frontal eye fields, a model circuit for \nhuman decision -making, induced robust facilitatory effects  promoting contralateral \nsaccades. These effects were specific to FEF,  immediate, present only during stimulation \ntrials, and emerged rapidly. We showed that interindividual variability in TUS effects could \nbe explained by the inhibitory tone of participants’ FEF at baseline , with TUS modulating \nthis balance to bring individuals into a more uniform physiological state. These findings \ncontribute to advancing bra in research by demonstrating the feasibility and efficacy of \nimmediate TUS effects in humans. They emphasize the importance of considering \ninterindividual variability and active control conditions in study designs. Together, this work \nopens new avenues for  future studies aiming to explore causal brain function with high \ntemporal precision and to develop innovative therapeutic interventions. \nResource availability \nAll behavioral, fMRI and spectroscopy data, as well as the behavioral and fMRI task code \nand all analysis scripts, are available at https://doi.org/10.34973/drtg-kq58.  \nAcknowledgements \nThis experiment was supported by the Dutch Research Council (NW O), awarding VIDI \nfellowships to L.V. (18 919) and H.E .M.d.O. (452-17-016). We would like to acknowledge \nEdward J. Auerbach, Ph.D., and Mał gorzata Marjańska, Ph.D. (Center for Magnetic \nResonance Research and Department of Radiology, University of Minnesota, USA) for the \ndevelopment of the pulse sequences for the Siemens platform, which were provided by the \nUniversity of Minnesota under a C2P agreement. \nAdditionally, we thank Norbert Hermesdorf, Margely Cornelissen, Hubert Voogd, Sibrecht \nBouwstra, Gerard van Oijen, and Pascal de Water from the technical support group at the \nDonders Centre for Cognition, Faculty of Social Sciences, Radboud University, for  their \nexcellent technical assistance and support throughout this study. \nFinally, we would also like thank Marwan Engels (Donders Centre for Cognition, Radboud \nUniversity) for his substantial role during data acquisition. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nAuthor contributions \nS.F., L.V. and H.E.M.d.O. conceptualized and designed the experiment; A.C., S.L.Y.W. and \nS.F. designed and programmed the behavioral and functional localizer tasks; S.F. and \nS.L.Y.W. collected the data; J.P.M. set up the fMRI and MRS sequences; S.F. , L.V., and \nH.E.M.d.O. analyzed the behavioral, functional and spectroscopy data; B.R.K. contributed \nto behavioral data analysis; R.S.K. contributed to spectroscopy analysis; S.L.Y.W. \ncontributed to fMRI analysis; S.F., B.R.K., R.S.K., W. P.M., L.V. and H.E.M.d.O. wrote the \nmanuscript, J.P.M. and A.C. revised the manuscript.  \nDeclaration of interest \nThe authors declare no competing interests. \nSupplemental information \nDocument S1, S3, and S3: Additional analyses and discussions. \nFigures S1-S7: Supplementary figures. \nTable S1-S9: Additional analyses and statistical outcomes. \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nMethods \n \nParticipants \n \nWe preregistered (https://doi.org/10.17605/OSF.IO/K5P2M) a target sample size of 35 \nparticipants, based on a small to medium effect size of f ~ .35, with an alpha level of .05 and \na power of 80% (calculated using G*Power 3.1; Faul et al., 2009 ). Participants were \nscreened on medical history to exclude putative participants with a history of brain surgery, \nserious head trauma, epilepsy, convulsion, or seizure, as well as the presence of implanted \nmetal in the head or upper body, diagnosed neurological or psychiatric disorder s, and \nconsumption of either more than four alcoholic units within the preceding 24 hours or any \nrecreational drugs within the past 48 hours.  \nAccounting for technical issues, 39 participants were enrolled in the experiment, of \nwhich four participants were excluded due to poor eye-tracking quality (e.g., multiple task \nrestarts during stimulation due to loss of eye gaze) or low accuracy (below 60%) in the \nsaccade task (indicative of participants not understanding or focusing on the task). \n35 participants (Mage = 24.1, SDage = 3.2, range = 20 – 22; 15:20 female:male; right-handed) \nwere included in the final analysis. Written informed consent was obtained from all \nparticipants in accordance with the Declaration of Helsinki, and the experimental \nprocedures were approved by the local ethics committee (CMO2022 -15953, Commissie \nMensgebonden Onderzoek Arnhem-Nijmegen). \nSaccade task results include all 35 participants. Some participants were excluded \nfrom the following analyses: For 10 participants, magnetic resonance spectroscopy (MRS) \nGABA+ acquisition was of poor quality; hence, all MRS results are based on the data of 25 \nparticipants (Mage = 24.7, SDage = 3.1, range = 20 – 33, 11:14 female:male). One participant \ndid not complete the final stimulation session and was thus excluded from all masking \nassessment analyses; hence, all masking assessment analyses are based on 34 participants \n(Mage = 24.9, SD age = 3.1, range = 20 – 33, 15 :19 female:male). Importantly, since \nparticipants experienced all conditions in each TUS session, this participant was still \nincluded in the main analyses. \n \nStudy overview \n \nThe study comprised three double -blind, within -subject sessions, with an interval of \napproximately one week (and up to three months) between sessions ( Figure 1 F), \nscheduled at the same times of the day to reduce potential fluctuations in GABA+ induced \nby circadian rhythm. In the initial intake session, participants engaged in a practice of the \nsaccade task without TUS delivery. Subsequently, they entered the MRI scanner to acquire \nstructural scans. Additionally, participants completed FEF and M1 functi onal localizers \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n(described below), used to target TUS during the following brain stimulation sessions. \nFinally, we obtained measures of GABA+ concentration from the left hemispheric \nstimulation regions using single voxel MRS.  MRS measurements were limited to the left \nhemisphere due to time constraints. The left hemisphere was prioritized over the right, \ngiven the stronger lateralization in attentional processes of the right hemisphere in humans \n(Bartolomeo & Seidel Malkinson, 2019; Heilman & Abell, 1980). \n The two successive brain stimulation sessions incorporated a mix of sham and TUS \ntrials during the saccade task. Each session started with screening and a 45 -minute \npreparation phase (see Neuronavigation below). Participants started with a practice block \nwithout TUS delivery or auditory masking, to reacquaint themselves with the task. Instances \nof performance falling below 60% of the maximum score triggered an automatic repetition \nof the practice block. A padding block without TUS nor auditory mask bookended each \nTUS block. TUS transducer placement (either left and right FEF or left and right M1) was \ncontingent on the blocks within the sequence. Sequence order was counterbalanced \nacross participants in the two stimulation sessions (Figure S3). TUS blocks contained three \nconditions: 1) left and ii) right TUS paired with an auditory masking tone, and iii) a sham \ncondition with solely the auditory mask. The order of conditions within TUS blocks followed \na pseudorandom pattern limited to a maximu m of four consec utive trials of the same \ncondition. At the session’s conclusion, participants were queried about any symptoms they \nbelieved could be associated with TUS. This was only used for debriefing and is not further \nanalyzes. Only after the final stimulation session, the efficacy of blinding was assessed \nduring a short masking assessment.  \n \nTasks \nSaccade task \nEach trial started with fixation on a star-shaped stimulus (0.25 x 0.25 degrees of visual angle) \npresented at the center of the screen. After fixation, there was a delay of 300 -600 ms \n(jittered) before the first planet-shaped target (0.5 x 0.5 degrees of visual angle, acceptance \nwindow, 3 degrees) briefly appeared in either the left or right  hemifield (10 degrees  of \nvisual angle left and right from the center of the screen). Simultaneously with the \nappearance of this first target, TUS was delivered, lasting 500 ms. The auditory mask began \n250 ms before the first target appeared and lasted 1 second, fully padding the TUS delivery. \nThe delay between the first and second planet-shaped target ranged from 0 to 200 ms.  \nTarget delays exhibited a non-uniform distribution, with shorter delays clustered around \nthe central peak and the longer delays at the tails. This distribution was designed to \noptimize the potential for TUS -induced behavioral modulation at relatively short target \ndelays. At the same time, it allowed to make sure that TUS does not simply induce attention \nlapses, characterized by incorrect responses even with long delays.  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nParticipants were instructed to execute a saccadic eye movement to the first  appearing \ntarget in either the left or right hemifield. Trial completion was followed by feedback, which \nwas presented for 1 s indicating whether the correct target had been chosen within the \ndesignated time (Figure 1A). In this gamified task, participants could earn points and a \nmonetary bonus of up to €5 per stimulation session based on their overall performance. \nDuring stimulation sessions, they received mixed sham and TUS trial s while an auditory \nmask was played to blind them to the different stimulation conditions and to prevent \nauditory confounding. The auditory mask corresponded to the specific condition, either \nmasking or replicating the sound of stimulation (Figure S4). \nFunctional localizers \nTo prevent the risk of undershooting or missing the target due to the small ultrasound focus, \nwe employed functional localizers to identify each participant’s FEF  and M1  with high \nfidelity (Sack et al., 2009) . The individual coordinates of interest  determined using \nfunctional localizers  were used for neuronavigation in the following brain stimulation \nsessions.  \nThe FEF localizer (Amiez et al., 2006; Gagnon et al., 2002) consisted of alternating \n24-second blocks of saccadic eye movements and central fixation ( Figure S5A). \nParticipants followed and fixated on a target (visual angle, 1 x 1 degrees; white square; \nduration, 800 ms) presented at randomized screen positions located at the left, right, or \ncenter of the screen (target distance, 14 degrees). This eye movement and  fixation \nsequence repeated six times. Assessment of the contrast between active eye movement \nblocks and baseline fixation blocks allowed for localization of the left and right FEFs.  \nThe M1 localizer (Tzourio-Mazoyer et al., 2015) consisted of alternating 16 -second \nblocks of left and right finger movement (Figure S5B). Specifically, participants repetitively \npinched their index finger and thumb together within the 16 -second interval, alternating \nbetween left and right hands for six blocks per hand. This task enabled the establishment \nof contrasts between blocks of fin ger movement for each hand, providing information \nabout left and right M1 activation. \nMasking assessment \nFollowing the final stimulation session, participants experienced a shorter series of sham  \nand TUS trials involving both left and right FEF and M1. After each trial, they reported \nthrough button presses (up button for yes, down button for no) whether they believed they \nhad received stimulation and on which side (left button for left, right button for right) they \nbelieved the stimulation was applied (Figure S5C). The order of the three conditions was \nfully randomized. Additionally, the sequence of stimulatio n regions was counterbalanced \nacross participants. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nIntake session \nTask practice \nParticipants practiced the saccade task outside of the scanner for 15 minutes (198 trials) \nwithout delivery of TUS, to acquaint themselves with the task prior to the follow -up brain \nstimulation sessions.  \nStructural and functional MRI data acquisition \nMRI scanning was performed at the Donders Centre for Cognitive Neuroimaging using a \n3 Tesla Magnetom Skyra Scanner (Siemens AG, Erlangen, Germany) equipped with a 32- \nchannel head coil. During structural scan acquisition, participants kept their eyes closed. \nHigh-resolution T1w scans were acquired (sagittal plane; repetition time (TR), 2700 ms; \necho time (TE), 3.69 ms; flip angle, 9 degrees; voxel size, 0.9 x 0.9 x 0.9 mm; field of view, \n230 mm) for MRS voxel placement, co -registration with the functional data, and \nneuronavigation for TUS delivery during stimulation sessions. To capture detailed skull \nmorphology and tissue properties for acoustic simulations of ultrasonic wave propagation, \nT2w scans (sagittal plane; TR, 3200 ms; TE, 408 ms; flip angle, T2 var flip angle mode; voxel \nsize, 0.9 x 0.9 x 0.9 mm; field of view, 230 mm), and UTE scans (transversal plane; TR, 3.32 \nms; TE, 0.07 ms; flip angle, 2 degrees; voxel size, 0.8 x 0.8 x 0.8 mm; field of view, 294 mm) \nwere acquired.  \nTo functionally localize the stimulation regions, a Multi -Band sequence with an \nacceleration factor of four (MB4) was used (TR, 995 ms; TE, 32.8 ms; flip angle, 60 degrees; \nvoxel size, 2.5 x 2.5 x 2.5 mm; field of view, 210 x 210 x 130 mm acquired in axial direction). \nVisual stimuli of the localizer tasks were presented at the rear bore face on a flat panel \nscreen. \nMRS data acquisition \nMagnetic Resonance (MRS) Single Voxel Spectroscopy (SVS) of the left hemispheric target \nregions (FEF and M1) allowed for baseline GABA+ measures. For each ROI, after voxel \nplacement based on the participant’s T1 -weighted scan, shimming was performed using \nFASTEST map (Gruetter, 1993; Gruetter & Tkác, 2000) and a flip angle calibration process \nwas carried out. For the FEF, the voxel was placed for each participant based on anatomical \nlandmarks, at the intersect of the precentral gyrus, middle frontal gyrus and the superior \nfrontal gyrus in the left hemispher e (Amiez et al., 2006; Paus, 1996) . The M1 voxel was \nplaced at the left hemispheric precentral knob located posterior to the intersection of the \nsuperior frontal sulcus that divides the superior from the middle frontal gyrus, and the \nprecentral sulcus (Yousry et al., 1997) . Baseline level of GABA + was measured using the \npulse sequence MEshcher -GArwood Point RESolved Spectroscopy (MEGA -PRESS: TR, \n2000 ms; TE, 68 ms; voxel size, 2.0 x 2.0 x 2.0 cm; with VAPOR water suppression (Tkác et \nal., 1999) 128 averages and water unsuppressed reference 16 averages) as introduced by \nMescher et al (1996, 1998) . The baseline level of  glutamate and glutamine  (Glx) was \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nquantified using the pulse sequence Point RESolved Spectroscopy (PRESS: TR, 20000 ms; \nTE, 35 ms; voxel size, 2.0 x 2.0 x 2.0 cm; with VAPOR water suppression 64 averages) as \ndescribed by Marjańska et al  (2013).  This data was not analyzed in the present paper. \n \nTUS sessions \nNeuronavigation and hair preparation \nThe transducer was placed at the target location, and monitored throughout the session, \nusing frameless stereotaxic neuronavigation (Localite  Biomedical Visualization Systems  \nGmbH, Sankt Augustin, Germany). We used participant specific T1 w scans and x -, y -, z -\ncoordinates of the left and right FEF and M1 derived from functional localizers. A reference \ntracker, five fixed markers (nasian, left and right eye, left and right ear), and 350 – 400 head \nsurface markers were used to register the ana tomical image to the participant’s physical \nhead. The two TUS transducers were al so calibrated using a reference tracker and \ncalibration plate. Transducers positions for the four stimulation regions were registered and \nquantified for acoustic ultrasonic wave propagation simulations. \nUltrasound gel (Aquaflex Ultrasound Gel, Parker Laboratories) was applied to the \nparticipant’s head over stimulation regions, followed by placement of gel pads (Aquaflex \nUltrasound Gel Pad, Parker Laboratories) between the gelled head and gel-covered \ntransducers to eliminate air bubbles(Murphy, Nandi, et al., 2024). Refer to Figure S6 for a \nschematic set-up. \nTUS protocol \nUltrasonic stimulation was delivered using the NeuroFUS PRO system (Brainbox Ltd., \nCardiff, UK) with two two -element ultrasound transducers (CTX250-009 and CTX250-014, \n45 mm diameter, 250 kHz fundamental frequency, Sonic Concepts Inc., Bothell, WA, USA). \nWe utilized a two-channel transducer to maximize the stimulation focal area. Although the \nTUS focus is characterized by a cigar-shaped profile that may extend into the white matter, \nit does not extend into the gray matter territory of neighboring cortical regions. The TUS \nprotocol was adapted from Kubanek et al. (2020) (pulse duration, 2 ms; pulse ramp length, \n1 ms, pulse repetition frequency, 500 Hz; pulse train duration, 500 ms; duty cycle, 50%, Isppa \nin free water, 25 W/cm 2; Figure 1D). Our study employed ramped pulses in combination \nwith an auditory mask to minimize auditory co-stimulation. \nAlthough squared and sinusoidal ramped pulses have the same integral energy \ncontent, it is important to note that squared wave pulses have associated limitations. A \nsquared pulse encompasses a constant intensity peak for a longer duration due to their \nclear onset and offset, whereas a sinus -shaped pulse exhibits a gradually increasing and \ndecreasing peak that is never fully off. While low-intensity ultrasonic waves are beyond the \nrange of human hearing, the on -offset of the squared pulse is detectable by humans, \nincreasing the likelihood of auditory confounds, and thus contributing to a clearer temporal \nprofile of stimulation (Choi et al., 2023; Mohammadjavadi et al., 2019). Furthermore, since \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nhumans have a thicker skull than macaques, a higher free-field Isppa was applied (25 W/cm2) \nto match the realized intracranial intensity across species. Moreover, we adjusted the total \nstimulation duration to the average human saccade duration. \nThe temperature rise ( ΔT) remained below two degrees Celsius and the derated  \nintracranial mechanical index (MI) below 1.9 matching ITRUSST recommendations (Aubry \net al., 2024). During both sham and TUS trials, an auditory mask was played through bone \nconducting headphones (AfterShokz, New York, US).  TUS was delivered during the task \nthrough serial commands in a PsychoPy script (PsychoPy 2021.2.3; Peirce et al., 2019). \nBehavioral acquisition \nOculomotor behavior during the saccade task was tracked using Eyelink1000 PLUS (SR \nResearch). Specifically, saccadic eye movements of the dominant eye were tracked from a \ndistance of 80 cm between eye tracker and chinrest (Figure S6A).   \nPrior to the saccade task, a nine -target calibration and validation process was conducted. \nStimuli for the saccade task were programmed using PsychoPy 2021.2.3 (Peirce et al., 2019) \nand displayed on a 24 -inch BenQ monitor (resolution, 1920 x 1080; refresh rate, 120 Hz; \nQisda Corporation, Taipei, Taiwan). \n \nData analysis \nSaccade task \nData visualization and analyses were performed using R (version 2021.9.2.382; RStudio \nTeam, 2021). Trials on which participants made double saccades (M = 2.1%, SD = 1.6, range \n= 0.4% – 7.8%) and where response times exceeded 1 s (M = 2.7%, SD = 2.7, range = 0.1% \n- 11.0%), which may have indicated failed eyetracking , were excluded. The practice and \npadding trials were also excluded from the dataset. For all regression analyses  reported \nbelow, SOA was included as a z-scored covariate. To account for both between and within-\nsubject variability, saccade data were analyzed with logistic mixed-effects models using the \nlme4 package in R (Bates et al., 2015). Furthermore, p-values of fixed effects were acquired \nusing Type III conditional F -tests with Kenward -Roger approximation for degrees of \nfreedom, as implemented in the Anova function of the car package (Fox et al., 2001, 2024). \nFinally, in case of significant fixed effects, post hoc pairwise comparisons were performed \nusing the emmeans function of the emmean package (Lenth et al., 2024).  \n \nBaseline behavior  \nTo evaluate the efficacy of the saccade task by establishing a robust relationship between \ntarget delay onsets and the probability of saccades to certain directions. The dependent \nvariable is the probability of making a rightward saccade. The independent v ariable is \ntarget delay (continuous; range -200 to 200 ms). The model includes both within and \nbetween-subject factors for target delay. We hypothesize a higher probability of rightward \nsaccades at larger positive target delays (e.g., target on the right hemifield appeared first) \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nand a lower probability of rightward saccades at more negative target delays (e.g., target \non the left hemifield appeared first). \n \nThe following lme4 model syntax was used: \n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑑𝑒𝑙𝑎𝑦 + (1 + 𝑑𝑒𝑙𝑎𝑦\t|\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \nTUS effects \nTo assess the direction of TUS effects, we looked at the effects of TUS per condition on \nsaccade direction. We hypothesized TUS effects to surface primarily in biasing responses \non trials with higher uncertainty (i.e. trials with short SOAs ) and therefore focused on the \n‘choice domain’. Choice domains were defined at the individual level  by determining the \ndelay windows, i.e. SOAs, where participants showed a probability of making rightward \nsaccades between 0.25 and 0.75 (see Supplementary Tables S.1-S.9 for other choice \ndomain window results). This led to inclusion of on average 455 trials per participant (SD = \n127, range = 125 – 733) with an average of 65 trials per TUS condition (SD = 18, range = 15 \n– 108) and 130 trials per sham condition (SD = 37, range = 40 – 208). \nTo measure the effects of TUS on saccade behavior, we  first examined the effects of \nstimulation to the left versus right FEF and, separately, the left versus right M1. This step \nallowed us to investigate potential lateralized effects within each stimulated region. \nSubsequently, each of these conditions (left FEF, right FEF, left M1, right M1) was compared \nto sham to assess how TUS modulated saccade direction relative to baseline conditions. In \nthese analyses, we included target delas as a continuous predict or. For each participant \ntarget delays were scaled by subtracting each individuals mean target delay and dividing it \nby the delay range for each individual . This scaling ensured that delay effects were \nnormalized across participants. For example, a typical analysis model included predictors \nfor the stimulation condition (e.g., left versus right FEF) and scaled delay, as follows: \n\t\n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑙𝑒𝑓𝑡𝐹𝐸𝐹/𝑟𝑖𝑔ℎ𝑡𝐹𝐸𝐹 + 𝑑𝑒𝑙𝑎𝑦!\"#$%&\n+ 51 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑙𝑒𝑓𝑡𝐹𝐸𝐹/𝑟𝑖𝑔ℎ𝑡𝐹𝐸𝐹 + 𝑑𝑒𝑙𝑎𝑦!\"#$%& \t6\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \nFurthermore, to examine if the effects that we find are specific to FEF modulation, and not \na result of any other confounding factors, we looked at the TUS effect of stimulation side \n(left vs. right) and stimulation region (FEF vs. M1) on saccade direction .  \n\t\n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + \t 𝑠𝑖𝑑𝑒$%'(/*+,-( ∗ \t 𝑟𝑒𝑔𝑖𝑜𝑛././01 + \t𝑑𝑒𝑙𝑎𝑦!\"#$%&\n+ \t 51 + 𝑠𝑖𝑑𝑒$%'(/*+,-( ∗ \t 𝑟𝑒𝑔𝑖𝑜𝑛././01 + 𝑑𝑒𝑙𝑎𝑦!\"#$%& \t6\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \nFinally, in an exploratory analysis presented in Supplementary Documents S.1, we \nexamined the effects of TUS during trials where no correct choice could be made based on \nvisual cues alone, specifically when the two targets were presented simultaneously (zero-\ndelay trails). Here, we added the factor of zero-delay into the previous models:  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛$%'(././*+,-(./. ∗ 𝑑𝑒𝑙𝑎𝑦2%*3/43(2%*3 + 𝑑𝑒𝑙𝑎𝑦!\"#$%&\n+ 51 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛$%'(././*+,-(./. ∗ 𝑑𝑒𝑙𝑎𝑦2%*3/43(2%*3\n+ 𝑑𝑒𝑙𝑎𝑦!\"#$%& \t6\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛$%'(01/*+,-(01 ∗ 𝑑𝑒𝑙𝑎𝑦2%*3/43(2%*3 + 𝑑𝑒𝑙𝑎𝑦!\"#$%&\n+ 51 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛$%'(01/*+,-(01 ∗ 𝑑𝑒𝑙𝑎𝑦2%*3/43(2%*3 + 𝑑𝑒𝑙𝑎𝑦!\"#$%& \t6\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑠𝑖𝑑𝑒$%'(/*+,-( ∗ \t 𝑟𝑒𝑔𝑖𝑜𝑛././01 ∗ 𝑑𝑒𝑙𝑎𝑦2%*3/43(2%*3 + 𝑑𝑒𝑙𝑎𝑦!\"#$%&\n+ 51 + 𝑠𝑖𝑑𝑒$%'(/*+,-( ∗ \t 𝑟𝑒𝑔𝑖𝑜𝑛././01 ∗ 𝑑𝑒𝑙𝑎𝑦2%*3/43(2%*3\n+ 𝑑𝑒𝑙𝑎𝑦!\"#$%& \t6\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \nBiasing TUS effects  \nTo ascertain the specificity of TUS effects on saccade biasing rather than general \nperception, we investigated whether ultrasound modulates the decision curve's \ncharacteristics along different axes: horizontal shift (indicating choice bias), slope alteration \n(indicating impaired target discrimination), or changes in asymptotes/lapse (indicating a \nbias beyond the choice domain).  These exploratory analyses are discussed in \nSupplementary Documents S.2. \nTo estimate the slope and bias in milliseconds, we analyzed the interaction effect of \ncondition and delay on saccade direction, focusing specifically on a delay range of -75 to \n+75 ms to increase sensitivity for detecting any slope effects. This range was selected \nbecause it closely approximates the individual choice domain used in other analyses, \nensuring consistency and comparability across methods. Unlike previous analyses where \nindividual choice domains were used, we opted for a fixed delay range in this analysis. This \ndecision was made because we aimed to quantify the absolute value of the bias shift \n(horizontal shift of the curve) in milliseconds. Using the scaled individual choice domains \ndoes not provide the opportunity to calculate this fixed bias shift in absolute time units. By \nincluding condition and delay as random effects, we were able to estimate the random \nslopes and biases for each participant. This approach allowed us to determine whether TUS \ninduced horizontal shifts in the decision curve, indicative of a choice bias. \n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑠ℎ𝑎𝑚/𝑙𝑒𝑓𝑡𝐹𝐸𝐹/𝑟𝑖𝑔ℎ𝑡𝐹𝐸𝐹/𝑙𝑒𝑓𝑡𝑀1/𝑟𝑖𝑔ℎ𝑡𝑀1 ∗ 𝑑𝑒𝑙𝑎𝑦\n+ 51 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑠ℎ𝑎𝑚/𝑙𝑒𝑓𝑡𝐹𝐸𝐹/𝑟𝑖𝑔ℎ𝑡𝐹𝐸𝐹/𝑙𝑒𝑓𝑡𝑀1/𝑟𝑖𝑔ℎ𝑡𝑀1 ∗ 𝑑𝑒𝑙𝑎𝑦\t6\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \nFor lapse estimation, we aimed to understand whether TUS could evoke saccades at longer \ndelays, thus indicating an effect beyond mere biasing. We selected absolute delays \nbetween 75 and 200 ms and assessed whether choice accuracy depended on condition \nand absolute delay. By focusing on absolute delays, rather than distinguishing between \nnegative and positive delays, we prioritized analyzing overall accuracy rather than side -\nspecific biases. This choice was made because we do not expect side biases to play a role \nin this context; instead, we are interested in understanding general task performance and \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\naccuracy. Therefore, t his analysis focused on the asymptotes of the decision curve to \ndetermine if TUS influenced saccade behavior even when the delays were long, and the \ntask was easy. \n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/56789:9/;<=289:9/5678>?/;<=28>? + 𝑑𝑒𝑙𝑎𝑦3@1\n+ 21 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/56789:9/;<=289:9/5678>?/;<=28>? + 𝑑𝑒𝑙𝑎𝑦3@1\t3\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \n \nOnline TUS effects \nMoreover, to assess whether this protocol truly functions as a  short-lived protocol without \nproducing longer-lasting effects, each sham trial was labeled according to the preceding \ntrial (e.g., left_FEF–sham refers to a sham trial that followed a left FEF trial). However, note \nthat this analysis was conducted with a mean of only 22 trials per condition (SD = 7, range \n= 6 – 43).  We then ran the same side-by-region model to analyze these labeled sham trials. \nGiven the expectation that the protocol only exerts direct, immediate effects, we \nhypothesized that there would be no signif icant interaction effect observed. Furthermore, \nwe also performed a Bayesian ANOVA using the same model syntax, as this approach \nprovides a more robust assessment of evidence for the null hypothesis. \n \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑠𝑖𝑑𝑒1234(5678/;<=28) ∗ 𝑟𝑒𝑔𝑖𝑜𝑛1234(9:9/>?) + 𝑑𝑒𝑙𝑎𝑦1C356D\n+ 21 + 𝑠𝑖𝑑𝑒1234(5678/;<=28) ∗ 𝑟𝑒𝑔𝑖𝑜𝑛1234(9:9/>?) + 𝑑𝑒𝑙𝑎𝑦1C356D 3\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡)\t \n \nFunctional localizers \nTo accurately target the stimulation sites for each individual, the participants performed a \nFEF and M1 localizer during the intake session. This data was then pre -processed and \nanalyzed to obtain coordinates for each region per participant.  \nfMRI pre-processing and analysis were conducted using SPM12 in MATLAB R2023a, \nalong with MRIcroGL for result visualization. The initial steps included excluding the first five \nfMRI volumes to account for signal steady -state transition, converting IMA files  to DICOM \ncompatible format, and visually checking for artefacts. We performed both single subject \nand group level analyses (N = 35) to establish coordinates within native and standard space, \nrespectively (FEF: Figure 2A (single-subject) and 2G (group-level); M1: Figure 3A (single-\nsubject) and 3G (group-level)). \nRealignment and reslicing were performed for both levels, followed by coregistration with \nthe participant's T1w -image for single subject analysis and with Montreal Neurological \nInstitute (MNI) standard space for group level analysis. Data was smoothing with a six mm \nFWHM Gaussian kernel, and realignment parameters were inspected.  The blocks were \nconvolved with canonical hemodynamic response function, followed by voxel -wise fitting \nof a general linear model (GLM), resulting in the computation of statistical parametric maps \nfor the comparisons. Subsequently, beta weights for each condition were estimated to \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\ncreate contrast maps, enabling Family -Wise Error corrected cluster -level inferences (p < \n.05). For the FEF localizer, saccades minus fixation blocks were used to obtain coordinates \nfor the left and right FEF. The M1 localizer contrasts involved pinching blocks of right fingers \nminus left fingers and vice versa to identify the left and right M1, respectively. \nTo determine FEF targets for TUS delivery, we selected a peak voxel within the  \nsignificant cluster within FEF, specifically at the junction of the superior precentral sulcus  \nand the superior frontal sulcus. The FEF localizer required reflexive pro-saccades, activating \nboth medial and lateral FEF peaks. The medial peak, linked to higher order  cognitive \ncontrol (Cameron, Riddle, & D’Esposito, 2015; Curtis, 2006; Curtis & D’Esposito,  2006; \nGagnon et al., 2002; Neggers et al., 2012; McDowell et al., 2008), was selected for TUS \ntargets. This decision aligns with our hypothesis that TUS holds the highest potential  for \ninfluencing saccadic behavior at equal preference, requiring the execution of  voluntary \nsaccadic eye movements by FEF. The M1 localizer with pinching either the left  or right \nfinger elicited a distinctive activation cluster of significant voxels in both left and  right M1. \nWithin the activation cluster, the local maximum of peak voxel was selected for  the x-, y-, \nand z-coordinates. \nThe accuracy of selected coordinates within sulci branches was assessed with \nFSLeyes by means of visualizing effect sizes modulated by statistical significance with \ntransparent threshold. Once confirmed, established coordinates per stimulation region \nwere entered in the Localite software to plan and monitor TUS delivery. Group level analysis \ncalculated contrast estimates' standard error and mean, determining significance of the \naverage estimate.  \n \nMRS analysis \nTo investigate interindividual differences in TUS susceptibility, we quantified baseline \ninhibitory tone in the left hemisphere stimulation sites (left FEF and left M1) using MRS. \nGABA+ concentrations were quantified using Gannet version 3.1.4  (Edden et al., 2014) , \nwith water used as a reference. Gannet’s standard preprocessing pipeline was used, which \nincludes frequency and phase correction by spectral registration and line broadening. \nEdited spectra were generated by subtracting individual edit -ON spectra from ed it-OFF \nspectra. Notably, the editing approach not only targets GABA but also other \nmacromolecules at 3ppm, therefore the concentrations of GABA+ (GABA and \nmacromolecules) are reported. Grey matter, white matter and CSF tissue fractions for \ndetermining tissue-corrected concentrations were obtained for both voxels using SPM12. \nMetabolite concentrations were then relaxation and tissue -corrected (Gasparovic et al.  \nmethod).  \nTo ensure data quality, two independent researchers performed visual quality \nchecks of the data. Using the GannetLoad output, water frequency drift was assessed to \nidentify excessive movement artifacts. Next, Cr signal alignment was inspected to evaluate \nthe quality of frequency alignment. For participants with noticeable drift or misalignment, \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nthe affected averages were removed, and GABA+ quantification was reprocessed using \nGannet. Participants were excluded if more than 50% of their averages had to be removed \nor if drift and alignment remained insufficient despite reprocessing, as their GABA+ was \nunreliable or inestimable due to lipid contamination or low signal -to-noise ratio (SNR). \nReliable model fits were achieved for 25 out of 35 acquisitions. Data quality was further \nquantified using the signal -to-noise ratio (SNR) and full -width-at-half-max (FWHM) of N -\nacetylaspartate (NAA) and fit error of the GABA+ peak provided by Gannet (Figure S7). \nTo assess whether interindividual variability in saccade bias could be explained by \nbaseline inhibitory tone, we examined whether the probability of making a rightward \nsaccade in the left FEF and sham conditions (as well as in the control left M1 and sham \nconditions) was influenced by baseline GABA+ levels. Given that we only measured the left \nhemispheric target regions using MRS, we restricted our analyses to the left hemisphe re. \nSpecifically, we tested whether the interaction between condition (left FEF vs. sham) a nd \nbaseline GABA+ levels predicted saccade direction, with target delay included as a \nseparate predictor. We ran the same model for M1 GABA+, to assess if M1 GABA+ levels \nwere predictive of the M1 TUS effects or intrinsic bias.  \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/56789:9 ∗ 𝐹𝐸𝐹EFGFH + 𝑑𝑒𝑙𝑎𝑦1C356D\n+ \t 21 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/56789:9 + 𝑑𝑒𝑙𝑎𝑦1C356D \t3\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n𝑠𝑎𝑐𝑐𝑎𝑑𝑒\t𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/5678>? ∗ 𝑀1EFGFH + 𝑑𝑒𝑙𝑎𝑦1C356D\n+ 21 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/5678>? + 𝑑𝑒𝑙𝑎𝑦1C356D \t3\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡)\t \nMasking assessment \nTo evaluate the efficacy of participant blinding to different conditions, we investigated \nwhether participants could distinguish sham from TUS trials by analyzing if stimulation \nperception (yes/no) depended on the stimulation condition (sham/FEF/M1). Additionally, \nwe assessed whether stimulation and side perception differed between FEF and M1 \nconditions. Specifically, we analyzed if side perception (left/right) depended on the \nstimulation side (left/right) and region (FEF/M1) in the following mixed models: \n \n𝑠𝑡𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛\t𝑝𝑒𝑟𝑐𝑒𝑝𝑡𝑖𝑜𝑛\t\t~\t1 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/9:9/>? + \t 21 + 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛1234/9:9/>?\t3\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \n𝑠𝑖𝑑𝑒\t𝑝𝑒𝑟𝑐𝑒𝑝𝑡𝑖𝑜𝑛\t\t~\t1 + 𝑠𝑖𝑑𝑒5678/;<=28 ∗ 𝑟𝑒𝑔𝑖𝑜𝑛9:9/>?\n+ \t 21 + 𝑠𝑖𝑑𝑒5678/;<=28 ∗ 𝑟𝑒𝑔𝑖𝑜𝑛9:9/>?\t3\t𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡) \n \nFor both models, we also performed a Bayesian ANOVA to further assess the evidence for \nthe null hypothesis. We hypothesized that there would be no difference in stimulation \nperception between sham, FEF, and M1 conditions due to the delivery of the auditory mask. \nEven if differences in stimulation perception were found compared to sham, we expected \nthis not to be problematic due to the inclusion of the active control site (M1), where TUS \nwas also delivered to both the left and right hemispheres. We anticipat ed no significant \ndifferences between FEF and M1 under these conditions. While differing results in side \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nperception might be observed, this would not pose a problem since M1 stimulation is also \nlateralized. \n \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nReferences \n \n \nAmiez, C., Kostopoulos, P ., Champod, A.-S., & Petrides, M. (2006). Local Morphology Predicts \nFunctional Organization of the Dorsal Premotor Region in the Human Brain. The \nJournal of Neuroscience , 26(10), 2724 –2731. \nhttps://doi.org/10.1523/JNEUROSCI.4739-05.2006 \nAubry, J.-F., Attali, D., Schafer, M., Fouragnan, E., Caskey, C., Chen, R., Darmani, G., Bubrick, \nE. J., Sallet, J., Butler, C., Stagg, C., Klein-Flügge, M., Yoo, S.-S., Treeby, B., Verhagen, \nL., & Pauly, K. B. (2024). ITRUSST Consensus on Biophysical Safety for Transcranial \nUltrasonic Stimulation (arXiv:2311.05359). arXiv. \nhttps://doi.org/10.48550/arXiv.2311.05359 \nBarker, A. T., Jalinous, R., & Freeston, I. L. (1985). Non -invasive magnetic stimulation of \nhuman motor cortex. Lancet (London, England) , 1(8437), 1106 –1107. \nhttps://doi.org/10.1016/s0140-6736(85)92413-4 \nBartolomeo, P ., & Seidel Malkinson, T. (2019). Hemispheric lateralization of attention \nprocesses in the human brain. Current Opinion in Psychology , 29, 90 –96. \nhttps://doi.org/10.1016/j.copsyc.2018.12.023 \nBates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed -Effects Models \nUsing lme4. Journal of Statistical Software , 67, 1 –48. \nhttps://doi.org/10.18637/jss.v067.i01 \nBraun, V., Blackmore, J., Cleveland, R. O., & Butler, C. R. (2020). Transcranial ultrasound \nstimulation in humans is associated with an auditory confound that can be effectively \nmasked. Brain Stimulation , 13(6), 1527 –1534. \nhttps://doi.org/10.1016/j.brs.2020.08.014 \nButler, C. R., Rhodes, E., Blackmore, J., Cheng, X., Peach, R. L., Veldsman, M., Sheerin, F., & \nCleveland, R. O. (2022). Transcranial ultrasound stimulation to human middle \ntemporal complex improves visual motion detection and modulates \nelectrophysiological responses. Brain Stimulation , 15(5), 1236 –1245. \nhttps://doi.org/10.1016/j.brs.2022.08.022 \nChoi, M. H., Li, N., Popelka, G., & Butts Pauly, K. (2023). Development and validation of a \ncomputational method to predict unintended auditory brainstem response during \ntranscranial ultrasound neuromodulation in mice. Brain Stimulation , 16(5), 1362 –\n1370. https://doi.org/10.1016/j.brs.2023.09.004 \nDarmani, G., Bergmann, T. O., Butts Pauly, K., Caskey, C. F., de Lecea, L., Fomenko, A., \nFouragnan, E., Legon, W., Murphy, K. R., Nandi, T., Phipps, M. A., Pinton, G., \nRamezanpour, H., Sallet, J., Yaakub, S. N., Yoo, S. S., & Chen, R. (2022). Non-invasive \ntranscranial ultrasound stimulation for neuromodulation. Clinical Neurophysiology, \n135, 51–73. https://doi.org/10.1016/j.clinph.2021.12.010 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nEdden, R. A. E., Puts, N. A. J., Harris, A. D., Barker, P . B., & Evans, C. J. (2014). Gannet: A batch-\nprocessing tool for the quantitative analysis of gamma-aminobutyric acid–edited MR \nspectroscopy spectra. Journal of Magnetic Resonance Imaging: JMRI , 40(6), 1445–\n1452. https://doi.org/10.1002/jmri.24478 \nFaul, F., Erdfelder, E., Buchner, A., & Lang, A. -G. (2009). Statistical power analyses using \nG*Power 3.1: Tests for correlation and regression analyses. Behavior Research \nMethods, 41(4), 1149–1160. https://doi.org/10.3758/BRM.41.4.1149 \nFolloni, D., Verhagen, L., Mars, R. B., Fouragnan, E., Constans, C., Aubry, J.-F., Rushworth, M. \nF. S., & Sallet, J. (2019). Manipulation of  Subcortical and Deep Cortical Activity in the \nPrimate Brain Using Transcranial Focused Ultrasound Stimulation. Neuron, 101(6), \n1109-1116.e5. https://doi.org/10.1016/j.neuron.2019.01.019 \nFox, J., Weisberg, S., & Price, B. (2001). car: Companion to Applied Regression  (p. 3.1-3) \n[Dataset]. https://doi.org/10.32614/CRAN.package.car \nFox, J., Weisberg, S., Price, B., Adler, D., Bates, D., Baud-Bovy, G., Bolker, B., Ellison, S., Firth, \nD., Friendly, M., Gorjanc, G., Graves, S., Heiberger, R., Krivitsky, P., Laboissiere, R., \nMaechler, M., Monette, G., Murdoch, D., Nilsson, H., … R -Core. (2024). car: \nCompanion to Applied Regression  (Version 3.1 -3) [Computer software]. \nhttps://cran.r-project.org/web/packages/car/index.html \nGaymard, B., Ploner, C. J., Rivaud-Péchoux, S., & Pierrot-Deseilligny, C. (1999). The frontal \neye field is involved in spatial short -term memory but not in reflexive saccade \ninhibition. Experimental Brain Research , 129(2), 288 –301. \nhttps://doi.org/10.1007/s002210050899 \nGrosbras, M.-H., & Paus, T. (2002). Transcranial magnetic stimulation of the human frontal \neye field: Effects on visual perception and attention. Journal of Cognitive \nNeuroscience, 14(7), 1109–1120. https://doi.org/10.1162/089892902320474553 \nGrosbras, M.-H., & Paus, T. (2003). Transcranial magnetic stimulation of the human frontal \neye field facilitates visual awareness. European Journal of Neuroscience , 18(11), \n3121–3126. https://doi.org/10.1111/j.1460-9568.2003.03055.x \nGruetter, R. (1993). Automatic, localized in vivo adjustment of all first - and second-order \nshim coils. Magnetic Resonance in Medicine , 29(6), 804 –811. \nhttps://doi.org/10.1002/mrm.1910290613 \nGruetter, R., & Tkác, I. (2000). Field mapping without reference scan using asymmetric echo-\nplanar techniques. Magnetic Resonance in Medicine , 43(2), 319 –323. \nhttps://doi.org/10.1002/(sici)1522-2594(200002)43:2<319::aid-mrm22>3.0.co;2-1 \nGuerra, A., Asci, F., Zampogna, A., D’Onofrio, V., Petrucci, S., Ginevrino, M., Berardelli, A., & \nSuppa, A. (2020). Gamma -transcranial alternating current stimulation and theta -\nburst stimulation: Inter -subject variability and the role of BDNF. Clinical \nNeurophysiology, 131(11), 2691–2699. \nhttps://doi.org/10.1016/j.clinph.2020.08.017 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nGuerra, A., López -Alonso, V., Cheeran, B., & Suppa, A. (2020). Variability in non -invasive \nbrain stimulation studies: Reasons and results. Neuroscience Letters, 719, 133330. \nhttps://doi.org/10.1016/j.neulet.2017.12.058 \nGuitton, D., Buchtel, H. A., & Douglas, R. M. (1985). Frontal lobe lesions in man cause \ndifficulties in suppressing reflexive glances and in generating goal -directed \nsaccades. Experimental Brain Research , 58(3), 455 –472. \nhttps://doi.org/10.1007/BF00235863 \nGuo, H., Hamilton, M., Offutt, S. J., Gloeckner, C. D., Li, T., Kim, Y., Legon, W., Alford, J. K., & \nLim, H. H. (2018). Ultrasound Produces Extensive Brain Activation via a Cochlear \nPathway. Neuron, 98(5), 1020-1030.e4. \nhttps://doi.org/10.1016/j.neuron.2018.04.036 \nHeilman, K. M., & Abell, T. V. D. (1980). Right hemisphere dominance for attention. \nNeurology, 30(3), 327–327. https://doi.org/10.1212/WNL.30.3.327 \nHeinen, K., Feredoes, E., Weiskopf, N., Ruff, C. C., & Driver, J. (2014). Direct Evidence for \nAttention-Dependent Influences of the Frontal Eye -Fields on Feature -Responsive \nVisual Cortex. Cerebral Cortex (New York, NY) , 24(11), 2815 –2821. \nhttps://doi.org/10.1093/cercor/bht157 \nHenik, A., Rafal, R., & Rhodes, D. (1994). Endogenously generated and visually guided \nsaccades after lesions of the human frontal eye fields. Journal of Cognitive \nNeuroscience, 6(4), 400–411. https://doi.org/10.1162/jocn.1994.6.4.400 \nHuang, Y.-Z., Lu, M.-K., Antal, A., Classen, J., Nitsche, M., Ziemann, U., Ridding, M., Hamada, \nM., Ugawa, Y., Jaberzadeh, S., Suppa, A., Paulus, W., & Rothwell, J. (2017). Plasticity \ninduced by non -invasive transcranial brain stimulation: A position paper . Clinical \nNeurophysiology, 128(11), 2318–2329. \nhttps://doi.org/10.1016/j.clinph.2017.09.007 \nHutchison, R. M., Gallivan, J. P ., Culham, J. C., Gati, J. S., Menon, R. S., & Everling, S. (2012). \nFunctional connectivity of the frontal eye fields in humans and macaque monkeys \ninvestigated with resting-state fMRI. Journal of Neurophysiology, 107(9), 2463–2474. \nhttps://doi.org/10.1152/jn.00891.2011 \nJerusalem, A., Al -Rekabi, Z., Chen, H., Ercole, A., Malboubi, M., Tamayo -Elizalde, M., \nVerhagen, L., & Contera, S. (2019). Electrophysiological-mechanical coupling in the \nneuronal membrane and its role in ultrasound neuromodulation and general \nanaesthesia. Acta Biomaterialia , 97, 116 –140. \nhttps://doi.org/10.1016/j.actbio.2019.07.041 \nJohnstone, A., Nandi, T., Martin, E., Bestmann, S., Stagg, C., & Treeby, B. (2021). A range of \npulses commonly used for human transcranial ultrasound stimulation are clearly \naudible. Brain Stimulation , 14(5), 1353 –1355. \nhttps://doi.org/10.1016/j.brs.2021.08.015 \nKamke, M. R., Ryan, A. E., Sale, M. V., Campbell, M. E. J., Riek, S., Carroll, T. J., & Mattingley, \nJ. B. (2014). Visual Spatial Attention Has Opposite Effects on Bidirectional Plasticity \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nin the Human Motor Cortex. Journal of Neuroscience , 34(4), 1475 –1480. \nhttps://doi.org/10.1523/JNEUROSCI.1595-13.2014 \nKammer, T. (1998). Phosphenes and transient scotomas induced by magnetic stimulation of \nthe occipital lobe: Their topographic relationship. Neuropsychologia, 37(2), 191–\n198. https://doi.org/10.1016/S0028-3932(98)00093-1 \nKhalighinejad, N., Bongioanni, A., Verhagen, L., Folloni, D., Attali, D., Aubry, J.-F., Sallet, J., & \nRushworth, M. F. S. (2020). A Basal Forebrain-Cingulate Circuit in Macaques Decides \nIt Is Time to Act. Neuron, 105(2), 370 -384.e8. \nhttps://doi.org/10.1016/j.neuron.2019.10.030 \nKolasinski, J., Hinson, E. L., Divanbeighi Zand, A. P., Rizov, A., Emir, U. E., & Stagg, C. J. (2019). \nThe dynamics of cortical GABA in human motor learning. The Journal of Physiology, \n597(1), 271–282. https://doi.org/10.1113/JP276626 \nKop, B. R., Oghli, Y. S., Grippe, T. C., Nandi, T., Lefkes, J., Meijer, S. W., Farboud, S., Engels, \nM., Hamani, M., Null, M., Radetz, A., Hassan, U., Darmani, G., Chetverikov, A., Ouden, \nH. E. M. den, Bergmann, T. O., Chen, R., & Verhagen, L. (2024). Auditory confounds \ncan drive online effects of transcranial ultrasonic stimulation in humans. eLife, 12. \nhttps://doi.org/10.7554/eLife.88762.2 \nKubanek, J., Brown, J., Ye, P., Pauly, K. B., Moore, T., & Newsome, W. (2020). Remote, brain \nregion-specific control of choice behavior with ultrasonic waves. Science Advances, \n6(21), eaaz4193. https://doi.org/10.1126/sciadv.aaz4193 \nLenth, R. V., Banfai, B., Bolker, B., Buerkner, P ., Giné-Vázquez, I., Herve, M., Jung, M., Love, J., \nMiguez, F., Piaskowski, J., Riebl, H., & Singmann, H. (2024). emmeans: Estimated \nMarginal Means, aka Least -Squares Means  (Version 1.10.6) [Computer software]. \nhttps://cran.r-project.org/web/packages/emmeans/index.html \nLópez-Alonso, V., Cheeran, B., Río-Rodríguez, D., & Fernández-Del-Olmo, M. (2014). Inter-\nindividual variability in response to non-invasive brain stimulation paradigms. Brain \nStimulation, 7(3), 372–380. https://doi.org/10.1016/j.brs.2014.02.004 \nMahmoodi, A., Harbison, C., Bongioanni, A., Emberton, A., Roumazeilles, L., Sallet, J., \nKhalighinejad, N., & Rushworth, M. F. S. (2024). A frontopolar -temporal circuit \ndetermines the impact of social information in macaque decision making. Neuron, \n112(1), 84-92.e6. https://doi.org/10.1016/j.neuron.2023.09.035 \nMarjańska, M., Lehéricy, S., Valabrègue, R., Popa, T., Worbe, Y., Russo, M., Auerbach, E. J., \nGrabli, D., Bonnet, C., Gallea, C., Coudert, M., Yahia -Cherif, L., Vidailhet, M., & \nMeunier, S. (2013). Brain dynamic neurochemical changes in dystonic patients: A \nmagnetic resonance spectroscopy study. Movement Disorders , 28(2), 201 –209. \nhttps://doi.org/10.1002/mds.25279 \nMassimini, M., Ferrarelli, F., Huber, R., Esser, S. K., Singh, H., & Tononi, G. (2005). Breakdown \nof Cortical Effective Connectivity During Sleep. Science, 309(5744), 2228 –2232. \nhttps://doi.org/10.1126/science.1117256 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nMcSorley, E., Haggard, P ., & Walker, R. (2006). Time Course of Oculomotor Inhibition \nRevealed by Saccade Trajectory Modulation. Journal of Neurophysiology , 96(3), \n1420–1424. https://doi.org/10.1152/jn.00315.2006 \nMenz, M. D., Oralkan, Ö., Khuri-Yakub, P. T., & Baccus, S. A. (2013). Precise Neural Stimulation \nin the Retina Using Focused Ultrasound. Journal of Neuroscience , 33(10), 4550 –\n4560. https://doi.org/10.1523/JNEUROSCI.3521-12.2013 \nMescher, M., Merkle, H., Kirsch, J., Garwood, M., & Gruetter, R. (1998). Simultaneous in vivo \nspectral editing and water suppression. NMR in Biomedicine , 11(6), 266 –272. \nhttps://doi.org/10.1002/(sici)1099-1492(199810)11:6<266::aid-nbm530>3.0.co;2-\nj \nMescher, M., Tannus, A., Johnson, M. O., & Garwood, M. (1996). Solvent Suppression Using \nSelective Echo Dephasing. Journal of Magnetic Resonance, Series A , 123(2), 226–\n229. https://doi.org/10.1006/jmra.1996.0242 \nMohammadjavadi, M., Ye, P . P ., Xia, A., Brown, J., Popelka, G., & Pauly, K. B. (2019). \nElimination of peripheral auditory pathway activation does not affect motor \nresponses from ultrasound neuromodulation. Brain Stimulation , 12(4), 901 –910. \nhttps://doi.org/10.1016/j.brs.2019.03.005 \nMurd, C., Luiga, I., Kreegipuu, K., & Bachmann, T. (2010). Scotomas induced by multiple, \nspatially invariant TMS pulses have stable size and subjective contrast. International \nJournal of Psychophysiology , 77(2), 157 –165. \nhttps://doi.org/10.1016/j.ijpsycho.2010.06.001 \nMurphy, K. R., Farrell, J. S., Bendig, J., Mitra, A., Luff, C., Stelzer, I. A., Yamaguchi, H., \nAngelakos, C. C., Choi, M., Bian, W., DiIanni, T., Pujol, E. M., Matosevich, N., Airan, R., \nGaudillière, B., Konofagou, E. E., Butts -Pauly, K., Soltesz, I., & Le cea, L. de. (2024). \nOptimized ultrasound neuromodulation for non -invasive control of behavior and \nphysiology. Neuron, 0(0). https://doi.org/10.1016/j.neuron.2024.07.002 \nMurphy, K. R., Farrell, J. S., Gomez, J. L., Stedman, Q. G., Li, N., Leung, S. A., Good, C. H., Qiu, \nZ., Firouzi, K., Butts Pauly, K., Khuri -Yakub, B. P. T., Michaelides, M., Soltesz, I., & de \nLecea, L. (2022). A tool for monitoring cell type -specific focu sed ultrasound \nneuromodulation and control of chronic epilepsy. Proceedings of the National \nAcademy of Sciences of the United States of America , 119(46), e2206828119. \nhttps://doi.org/10.1073/pnas.2206828119 \nMurphy, K. R., Nandi, T., Kop, B., Osada, T., N’Djin, W. A., Lueckel, M., Caulfield, K. A., \nFomenko, A., Siebner, H. R., Ugawa, Y., Verhagen, L., Bestmann, S., Martin, E., Pauly, \nK. B., Fouragnan, E., & Bergmann, T. O. (2024). A Practical Guide to Transcranial \nUltrasonic Stimulation from the IFCN -endorsed ITRUSST Consortium  \n(arXiv:2407.07646). arXiv. https://doi.org/10.48550/arXiv.2407.07646 \nNagel, M., Sprenger, A., Lencer, R., Kömpf, D., Siebner, H., & Heide, W. (2008). Distributed \nrepresentations of the “preparatory set” in the frontal oculomotor system: A TMS \nstudy. BMC Neuroscience, 9(1), 89. https://doi.org/10.1186/1471-2202-9-89 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nNakajima, K., Osada, T., Ogawa, A., Tanaka, M., Oka, S., Kamagata, K., Aoki, S., Oshima, Y., \nTanaka, S., & Konishi, S. (2022). A causal role of  anterior prefrontal-putamen circuit \nfor response inhibition revealed by transcranial ultrasound stimulation in h umans. \nCell Reports, 40(7), 111197. https://doi.org/10.1016/j.celrep.2022.111197 \nNandi, T., Kop, B. R., Butts Pauly, K., Stagg, C. J., & Verhagen, L. (2024). The relationship \nbetween parameters and effects in transcranial ultrasonic stimulation. Brain \nStimulation, 17(6), 1216–1228. https://doi.org/10.1016/j.brs.2024.10.008 \nNeggers, S. F. W., Zandbelt, B. B., Schall, M. S., & Schall, J. D. (2015). Comparative diffusion \ntractography of corticostriatal motor pathways reveals differences between humans \nand macaques. Journal of Neurophysiology , 113(7), 2164 –2172. \nhttps://doi.org/10.1152/jn.00569.2014 \nNyffeler, T., Wurtz, P ., Lüscher, H.-R., Hess, C. W., Senn, W., Pflugshaupt, T., von Wartburg, R., \nLüthi, M., & Müri, R. M. (2006). Repetitive TMS over the human oculomotor cortex: \nComparison of 1 -Hz and theta burst stimulation. Neuroscience Letters, 409(1), 57–\n60. https://doi.org/10.1016/j.neulet.2006.09.011 \nPaus, T. (1996). Location and function of the human frontal eye -field: A selective review. \nNeuropsychologia, 34(6), 475–483. https://doi.org/10.1016/0028-3932(95)00134-4 \nPeirce, J., Gray, J. R., Simpson, S., MacAskill, M., Höchenberger, R., Sogo, H., Kastman, E., & \nLindeløv, J. K. (2019). PsychoPy2: Experiments in behavior made easy. Behavior \nResearch Methods, 51(1), 195–203. https://doi.org/10.3758/s13428-018-01193-y \nPellegrini, M., Zoghi, M., & Jaberzadeh, S. (2018a). Biological and anatomical factors \ninfluencing interindividual variability to noninvasive brain stimulation of the primary \nmotor cortex: A systematic review and meta-analysis. Reviews in the Neurosciences, \n29(2), 199–222. https://doi.org/10.1515/revneuro-2017-0048 \nPellegrini, M., Zoghi, M., & Jaberzadeh, S. (2018b). Cluster analysis and subgrouping to \ninvestigate inter-individual variability to non-invasive brain stimulation: A systematic \nreview. Reviews in the Neurosciences , 29(6), 675 –697. \nhttps://doi.org/10.1515/revneuro-2017-0083 \nRiis, T. S., Feldman, D. A., Losser, A. J., Okifuji, A., & Kubanek, J. (2022). Noninvasive targeted \nmodulation of pain circuits with focused ultrasonic waves. PAIN, \n10.1097/j.pain.0000000000003322. \nhttps://doi.org/10.1097/j.pain.0000000000003322 \nRivaud, S., Müri, R. M., Gaymard, B., Vermersch, A. I., & Pierrot -Deseilligny, C. (1994). Eye \nmovement disorders after frontal eye field lesions in humans. Experimental Brain \nResearch, 102(1), 110–120. https://doi.org/10.1007/BF00232443 \nRo, T., Cheifet, S., Ingle, H., Shoup, R., & Rafal, R. (1999). Localization of the human frontal \neye fields and motor hand area with transcranial magnetic stimulation and magnetic \nresonance imaging. Neuropsychologia, 37(2), 225 –231. \nhttps://doi.org/10.1016/s0028-3932(98)00097-9 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nRo, T., Farnè, A., & Chang, E. (2002). Locating the human frontal eye fields with transcranial \nmagnetic stimulation. Journal of Clinical and Experimental Neuropsychology, 24(7), \n930–940. https://doi.org/10.1076/jcen.24.7.930.8385 \nRo, T., Henik, A., Machado, L., & Rafal, R. D. (1997). Transcranial magnetic stimulation of the \nprefrontal cortex delays contralateral endogenous saccades. Journal of Cognitive \nNeuroscience, 9(4), 433–440. https://doi.org/10.1162/jocn.1997.9.4.433 \nSack, A. T., Cohen Kadosh, R., Schuhmann, T., Moerel, M., Walsh, V., & Goebel, R. (2009). \nOptimizing Functional Accuracy of TMS in Cognitive Studies: A Comparison of \nMethods. Journal of Cognitive Neuroscience , 21(2), 207 –221. \nhttps://doi.org/10.1162/jocn.2009.21126 \nSato, T., Shapiro, M. G., & Tsao, D. Y. (2018). Ultrasonic Neuromodulation Causes \nWidespread Cortical Activation via an Indirect Auditory Mechanism. Neuron, 98(5), \n1031-1041.e5. https://doi.org/10.1016/j.neuron.2018.05.009 \nSiebner, H. R., Funke, K., Aberra, A. S., Antal, A., Bestmann, S., Chen, R., Classen, J., Davare, \nM., Di Lazzaro, V., Fox, P . T., Hallett, M., Karabanov, A. N., Kesselheim, J., Beck, M. M., \nKoch, G., Liebetanz, D., Meunier, S., Miniussi, C., Paulus, W., … Ugawa, Y. (2022). \nTranscranial magnetic stimulation of the brain: What is stimulated? - A consensus \nand critical position paper. Clinical Neurophysiology: Official Journal of the \nInternational Federation of Clinical Neurophysiology , 140, 59 –97. \nhttps://doi.org/10.1016/j.clinph.2022.04.022 \nStagg, C. J., Bachtiar, V., & Johansen-Berg, H. (2011). The Role of GABA in Human Motor \nLearning. Current Biology , 21(6), 480 –484. \nhttps://doi.org/10.1016/j.cub.2011.01.069 \nSumner, P ., Edden, R. A. E., Bompas, A., Evans, C. J., & Singh, K. D. (2010). More GABA, less \ndistraction: A neurochemical predictor of motor decision speed. Nature \nNeuroscience, 13(7), 825–827. https://doi.org/10.1038/nn.2559 \nThickbroom, G. W., Stell, R., & Mastaglia, F. L. (1996). Transcranial magnetic stimulation of \nthe human frontal eye field. Journal of the Neurological Sciences, 144(1–2), 114–118. \nhttps://doi.org/10.1016/s0022-510x(96)00194-3 \nTkác, I., Starcuk, Z., Choi, I. Y., & Gruetter, R. (1999). In vivo 1H NMR spectroscopy of rat brain \nat 1 ms echo time. Magnetic Resonance in Medicine , 41(4), 649 –656. \nhttps://doi.org/10.1002/(sici)1522-2594(199904)41:4<649::aid-mrm2>3.0.co;2-g \nVernet, M., Quentin, R., Chanes, L., Mitsumasu, A., & Valero -Cabré, A. (2014). Frontal eye \nfield, where art thou? Anatomy, function, and non -invasive manipulation of frontal \nregions involved in eye movements and associated cognitive operations. Frontiers \nin Integrative Neuroscience, 8, 66. https://doi.org/10.3389/fnint.2014.00066 \nYaakub, S. N., White, T. A., Roberts, J., Martin, E., Verhagen, L., Stagg, C. J., Hall, S., & \nFouragnan, E. F. (2023). Transcranial focused ultrasound -mediated neurochemical \nand functional connectivity changes in deep cortical regions in humans. Nature \nCommunications, 14(1), 5318. https://doi.org/10.1038/s41467-023-40998-0 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nYoo, S., Mittelstein, D. R., Hurt, R. C., Lacroix, J., & Shapiro, M. G. (2022). Focused ultrasound \nexcites cortical neurons via mechanosensitive calcium accumulation and ion channel \namplification. Nature Communications, 13(1), 493. https://doi.org/10.1038/s41467-\n022-28040-1 \nYousry, T. A., Schmid, U. D., Alkadhi, H., Schmidt, D., Peraud, A., Buettner, A., & Winkler, P. \n(1997). Localization of the motor hand area to a knob on the precentral gyrus. A new \nlandmark. Brain: A Journal of Neurology , 120 ( Pt 1) , 141 –157. \nhttps://doi.org/10.1093/brain/120.1.141 \nYu, K., Niu, X., Krook -Magnuson, E., & He, B. (2021). Intrinsic functional neuron -type \nselectivity of transcranial focused ultrasound neuromodulation. Nature \nCommunications, 12(1), 2519. https://doi.org/10.1038/s41467-021-22743-7 \n  \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nSupplemental Materials \n \nS.1 Zero-delay trials \nOur analyses are tailored to the choice domain, where the SOA provides informative —but \nnot overwhelming—evidence in favor of one saccade direction over the other. When there \nis no meaningful choice to be made, i.e., when the SOA is long, performance approaches \nceiling, and we did not observe any modulatory effect of TUS (Supplementary Documents \nS.2).  \n \nIn contrast, exploratory analyses revealed that when no correct choice can be made (i.e., \nwhen the two cues are presented simultaneously ), TUS over both FEF and M1 biased \nsaccades. A three -way interaction trend between region (FEF/M1), stimulation side \n(left/right), and delay (non-zero/zero) suggested that any M1 effect was limited to the zero-\ndelay trials (b = -0.50, 95%-CI [-1.02, 0.02], c 2 = 3.6, p = 0.058; Figure S1). However, this \nanalysis is underpowered due to the inclusion of multiple interactions. To address this, we \nsplit the data into FEF and M1 subsets.  These subsequent tests confirmed that M1 TUS \neffects are specific to zero-delay trials (condition (left M1/right M1) x delay (non-zero/zero): \nb = -0.57, 95%-CI [-0.94, -0.20], c 2 = 9.0, p = 0.002; Figure S1). In contrast, FEF TUS effects \npersisted across both zero- and non-zero delay trials (condition (left FEF/right FEF) x delay \n(non-zero/zero): b = -0.08, 95%-CI [-0.45, 0.28], c 2 = 0.2, p = 0.6; Figure S1), indicating that \nthe observed effects of FEF TUS are robust and stable. \n \nThe M1 TUS effect on zero-delay trials was significant, both in statistical and absolute terms. \nWhen these trials are included in the choice domain, the shared direction of TUS bias across \nFEF and M1 obscures a putative interaction of stimulation side and region (side (left/right) \nx region (FEF/M1): b = 0.16, 95%-CI [-0.03, 0.36], c 2 = 2.9, p = 0.09). Instead, it reveals a \nmain effect of stimulation side (side (left/right): b = -0.26, 95%-CI [-0.41, -0.11], c 2 = 11.5, p \n= 0.007), and region (region (FEF/M1): b = -0.14, 95%-CI [-0.26, -0.01], c 2 = 4.6, p = 0.032).  \nThe highly specific biasing effect of TUS over M1 , observed only when the two visual cues \nare presented simultaneously, warrants further investigation. In these conditions, no correct \nchoice can be made based on the visual cues alone. One possible explanation is that  an \nM1 TUS bias arise s from true neuromodulation of M1. Indeed, M1 circuits anatomically \nconverge with downstream saccade circuits in the basal ganglia to support eye -hand \ncoordination (Neggers et al., 2015) . Perhaps when visual information is absent, motor \nbiases are propagated through these shared effector circuits. \n \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n \nS.2 Estimating the bias, slope and lapse rate \nTo ascertain the specificity of the TUS effects on biasing saccade direction, we explored \nwhether FEF TUS influences choice bias (a horizontal shift  in the decision curve ), target \ndiscrimination (a change in slope), or bias beyond the choice domain (a change in \nasymptotes/lapse).  \n \nBias and slope were analyzed using mixed effects logistic regression focusing on trials with \nshort delays (-75 to +75 ms).  Choice bias was assessed by examining the main effect of \nstimulation condition (c 2 = 22.2, p < 0.001), revealing that TUS induced a horizontal shift in \nthe decision curve. Specifically, left FEF stimulation shifted the curve by -3.67 ms, whereas \nright FEF stimulation shifted it by +3.77 ms. Post hoc tests showed that the shift between \nleft and right FEF stimulation was significantly different (p < 0.001; Figure S2A), unlike the \nshift between left and right M1 stimulation (p = 0.09; Figure S2A).  \n \nIn contrast, the slope of the decision curves, reflecting target discrimination was unaffected \nby TUS, as indicated by the non -significant interaction between stimulation condition and \ntarget delay (condition x SOA: c 2 = 2.715, p = 0.6; condition (left FEF) x SOA: b = 0.001, \n95%-CI [ -0.002, 0.003]; condition (right FEF) x SOA: b = 0.001, 95% -CI [ -0.002, 0.004]; \ncondition (left M1) x SOA: b = 0.001, 95%-CI [-0.002, 0.003]; condition (right M1) x SOA: b \n= 0.002, 95%-CI [-0.001, 0.005];  Figure S2B). \n \nFinally, to determine if TUS alters bias outside the choice domain, we analyzed trials with \nlonger absolute delays (75 to 200 ms). While there was a trend suggesting a condition effect \n(condition: c 2 = 9. 0, p = 0.060; condition (left FEF): b = 0.2 3, 95% -CI [ -0.01, 0.0.56]; \ncondition (right FEF): b = 0.34, 95%-CI [0.03, 0.64]; condition (left M1): b = 0.33, 95% -CI \n[0.02, 0.64]; condition (right M1): b = 0.35, 95%-CI [0.04, 0.65]; Figure S2C), post hoc tests \n \n \nFigure S1 | Differences in TUS effects for zero-delay and nonzero-delay trials in FEF and M1 \n \nTUS effects are expressed as the probability of making contralateral saccades (e.g., left hemisphere is \nstimulated and a rightward saccade is made). A contralateral saccade probability greater than 0.5 indicates \na TUS effect b eyond chance. Left and right FEF and M1 conditions were pooled to form FEF and M1 \ncategories, respectively. For M1, a significant difference in TUS effects was observed between zero-delay and \nnon-zero-delay trials (p = 0.002), with TUS effects being present only for zero-delay trials, indicating a delay-\ndependent modulation of M1. In contrast, FEF stimulation produced significant TUS effects for both zero -\ndelay and non -zero-delay trials, with no significant difference between the delays ( p = 0.6), suggesting a \nstable and consistent modulation of FEF regardless of delay.  Data are presented as group means with \nstandard error of the mean (S.E.M.). \nnonzero zero\nM1\n*\n0.0\n0.2\nns\nnonzero zero\nFEF\n.\n \nTUS effect\nsignificant \nTUS effects\nP(contralat. resp.)\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nconfirmed no significant impact of TUS on choice bias these trials (FEF (left/right): p = 1.0; \nM1(left/right): p = 1.0; Figure S2C). This reinforces the conclusion that TUS selectively \nbiases responses under conditions of response uncertainty and is unlikely to reverse or \nevoke responses under high certainty. \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \nFigure S2 | Estimating bias, slope and lapse rate following FEF and M1 TUS \n \n(A) Choice bias. Left: a visual representation of a horizontal shift in the decision curve, with the shaded blue \narea indicating the region of interest for the analysis. Right: estimated bias per participant derived from the \nmixed-effects model, which includes random effects (individual distributions with cloud and SEM) and fixed \neffects (group mean represented by the dot). Left FEF stimulation shifted the decision curve by -3.67 ms, \nwhile right FEF stimulation shifted it by +3.77 ms. A significant difference was found between left and right \nFEF (P < 0.001), but no significant shift was observed for M1 (p = 0.09), suggesting that TUS specifically affects \nchoice bias in FEF. (B) Target discrimination (slope). Left: a visual representation of how changes in slope \nwould look, with the blue shaded area marking the region of interest. Right: estimated slopes per participant \nderived from the mixed-effects model, which includes random effects (individual distributions with cloud and \nSEM) and fixed effects (group mean represented by the dot). No significant effect of TUS on slope was \nobserved, as no difference was found between these conditions ( p = 0.6), indicating that TUS does not \nsignificantly impact target discrimination. (C) Bias beyond the choice domain (lapse rate).  Left: a visual \nrepresentation of changes in lapse rate, with the shaded blue area marking the region of interest. Right: \nestimated lapse rates per participant derived from the mixed-effects model, which includes random effects \n(individual distributions with cloud and SEM) and fixed effects (group mean represented by the dot). While \nSOA (ms)\n-200 2000 75-75\nchoice\n0\n1\nSOA (ms)\n-200 2000 75-75\nchoice\n0\n1\nSOA (ms)\n-200 2000 75-75\nchoice\n0\n1\nA estimated delay bias\nB estimated slope\nC estimated lapse rate\n−10\n0\n10\n20\nL FEF R FEF L M1 R M1\ncondition\nestimated bias (ms)\n* ns\n0.025\n0.030\n0.035\n0.040\n0.045\nestimated slope\n0 L FEF R FEF L M1 R M1\nns ns\nns ns\n0.70\n0.75\n0.80\n0.85\n0.90\n0.95\nestimated lapse rate\n0 L FEF R FEF L M1 R M1\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n \n \na trend toward a condition effect was observed (p = 0.060), post hoc tests confirmed no significant impact of \nTUS on lapse rate (p = 1.0 for both FEF and M1), reinforcing the conclusion that TUS selectively biases choice \nbehavior without affecting performance in the lapse rate domain. \n \n \nS.3 Ipsilateral after-effects and stimulation perception \nWithin each block, participants received pseudorandomized left TUS, right TUS, and \nsham stimulation. To investigate potential longer -lasting TUS effects beyond the \nstimulation duration itself, we analyzed sham trials that directly followed a TUS trial. \nInterestingly, we observed a significant increase in ipsilateral responses during these \nsham trials—for example, if a sham trial followed a left TUS trial, participants were more \nlikely to make a leftward saccade (sidet-1: b = 0.15, 95%-CI [-0.08, 0.38], c 2 = 5.3, p = 0.021; \nFigure 4A). This ipsilateral bias on sham trials following TUS was in the opposite direction \nof the behavioral effects induced by FEF TUS, which increased contralateral saccades. \nOne could conceive that this reversal reflects compensatory carry -over effects of TUS . \nImportantly, however, these after-effects and stimulation perception biases did not differ \nbetween FEF and M1 conditions, where -as TUS effects were present only for FEF \nstimulation. Thus, the robust FEF TUS effects cannot be explained by these ipsilateral \ntendencies, nor do they support the presence of post-TUS compensatory mechanisms. \n \nThus, while neuromodulatory effects of TUS cannot explain the presence of ipsilateral \nafter-effects, this begs the question what does drive these effects. We speculate that the \nanswer to this question in the similar ipsilateral response pattern that emerge d in the \nmasking assessment. Here, participants performed a forced-choice task to report whether \nthey perceived the stimulation as originating from left or right TUS. Participants \nsignificantly misattributed stimulation to the ipsilateral side, independent of stimulation \nregion (side: b = -1.2, 95%-CI [-2.1, -0.3], c 2 = 7.0, p = 0.008; side (left/right) x region \n(FEF/M1): b = -0.5, 95%-CI [-1.5, 0.5], c 2 = 1.1, p = 0.3; Figure 4C).  \n \nThis ipsilateral and lateralized perception of TUS likely stems from specific properties of \nskull morphology. Variations in how flexural waves—vibrations traveling through the skull—\nare transmitted can cause the highest amplitude near the contralateral cochlea, \ninfluencing perceived sound location  (Braun et al., 2020) . We speculate that the \nregionally non-specific but lateralized after -effects observed in sham trials immediately \npost TUS may reflect an attention -orienting response. If participants subjectively \nperceived prior stimulation as originating from the left, they may have been bi ased \ntoward making leftward saccades afterward.  \n \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n \n \n \nFigure S3 | Schematic overview of the saccade task duration and block sequences \n \nThe task consists of two randomized and counterbalanced block sequences: X (FEF, M1, M1, and then FEF) \nand Y (M1, FEF, FEF, and M1), with the order of the blocks varying across participants. Each block \ncomprises 198 trials (66 left TUS, 66 right TUS, and 66 sham trials, all with masking sound), with additional \npadding of 16-17 trials before and after the block, where no TUS or masking sound is presented. At the \nstart of each session, participants complete a short practice block for task familiarization. A 5-minute break \nis provided between blocks for participants to stretch and rest before transducers are recoupled. FEF \nstimulation is represented in blue and M1 stimulation in green.  \n \nFigure S4 | Schematic overview of masking structure in TUS and sham trials \n \nThe trial begins with background white noise, delivered via bone-conducting headphones, lasting 1000 ms. \n150 ms after the onset of the white noise, a smooth wave (padding sound) consisting of frequencies 0.5, 10, \n12, and 14 kHz is delivered via bone -conducting headphones for 700 ms. For TUS trials (top), the 500 ms \nultrasonic stimulation is then delivered 250 ms after the white noise onset (100 ms after the padding sound). \nFor sham trials (bottom), the 500 ms of ultrasonic stimulation is replaced by a smooth wave of 10, 12, and 14 \nkHz, mimicking the TUS sound. \nSupplementary figure 1\nsaccade task\n60 min\nbreak\n5 min\nbreak\n5 min\nbreak\n5 min\nsequence X\nsequence Y\n M1 block\nM1 block\n M1 block\n FEF block\nM1 block\nFEF block\nFEF block\n FEF block\npadding: 16-17 trials (no TUS, no masking sound)\npractice: 16-17 trials (no TUS, no masking sound)\nTUS: 198 trials (66 left TUS, 66 right TUS, 66 sham; all with masking sound)\nTUS trial\n1000 ms\n150 ms 100 ms\n700 ms\n500 ms\nstimulation: TUS\npadding: smooth wave (0.5, 10, 12, 14 kHz)\nbackground: white noise\nsham trial\n1000 ms\n150 ms 100 ms\n700 ms\n500 ms\nstimulation: sham TUS (smooth wave, 10, 12, 14 kHz)\npadding: smooth wave (0.5, 10, 12, 14 kHz)\nbackground: white noise\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n \n \nFigure S5 | Schematic overview of functional localizers and masking assessment \n \n(A) FEF functional localizer. After a welcome screen, participants completed blocks of \"follow the target\" and \n\"fixate on the target\". In the 'follow the target' blocks, participants were presented with targets in a random \norder of left, center, and right positions. This sequence was r andomized and repeated 10 times per block, \nwith each target being shown for 800 ms. In the \"fixation\" block, the target was displayed in the center for 24 \nseconds. This sequence was repeated six times. The contrast was fixation versus follow the  target. (B) M1 \nfunctional localizer. After a welcome screen, two blocks were presented, repeated 10 times. In each block, \nparticipants were instructed to either pinch their right index finger and thumb as often as possible, or their \nleft index finger and thumb as often as possible for 16 seconds. The contrast was left versus right. (C) Masking \nassessment. Participants received TUS on the left/right FEF, left/right M1, and sham stimulation. For each \ntrial, participants were first asked to indicate whether they thought they had received stimulation by pressing \nthe up-arrow key for yes and the down-arrow key for no. Afterward, they were asked to guess whether the \nstimulation was on the left or right side by pressing the left-arrow key for left and the right-arrow key for right. \nThe order of the conditions/trials was randomized within each block, with the overall sequence determined \nby the block sequence (right panel). Each block comprises of 8 left TUS, 8 right TUS and 16 sham trials. \nSupplementary figure 3\npinch your right\nfingers together\nas often \nas possible\n16 s\npinch your left\nfingers together\nas often \nas possible\n16 s\nn=10\nwelcome\nscreen\n800 ms 800 ms 800 ms\n24000 ms\nn=10\nn=6\nwelcome\nscreen\nA\nB\nC\nDo you think\nyou were\nstimulated?\nIf you had\nto guess, \nleft / right?\nyes no left right\nmasking assessment\n5-10 min\nsequence A\nsequence B\n M1 block\nFEF block\n M1 block\nFEF block\nTUS: 32 trials (8 left TUS, 8 right TUS, 16 sham)\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \n \nFigure S7 | MRS-derived measurements for the left FEF and left M1 Voxels \n \n(A) Estimated GABA+ levels (in institutional units, i.u.) for each participant in the left FEF (blue) and left M1 \n(light blue) voxels, derived using Gannet. Each dot represents an individual participant, with bars showing \nthe group mean and error bars indicating the standard error of the mean (SEM). (B) N-Acetylaspartate (NAA) \nsignal-to-noise ratio (SNR) (%) for each participant in the left FEF (blue) and left M1 (light blue) voxels. The \nNAA SNR reflects the strength and quality of the NAA peak relative to background noise and serves  as an \nindicator of data quality. Each dot represents an individual participant, with bars showing the group mean \nand SEM. (C) GABA+ fit error (%) for each participant in the left FEF (blue) and left M1 (light blue) voxels. The \nfit error indicates the quality of the GABA+ estimation, with lower percentages reflecting better fits. Each dot \nrepresents an individual participant, with bars showing the group mean and SEM. \n \nSupplementary figure 5\nA B C\n0\n1\n2\n3\n4\n5\nGABA+ (i.u.)\nFEF M1\n0\n100\n200\n300\nNAA SNR (%)\nFEF M1\n0\n5\n10\n15GABA+ fit error (%)\nFEF M1\n \nFigure S6 | Overview of the study setup \n \n(A) Schematic overview of the TUS setup. Participants were seated 80 cm from the experimental task screen, \nwith their head stabilized on a chinrest at the center of the screen. Transducers were positioned using a \nVelcro headcap and guided by neuronavigation for precise targeting (coordinates derived from FEF and M1 \nfunctional localizers and entered into Localite software). Bone -conducting headphones for masking were \nplaced on participants, and an eye tracker was positioned below the screen to record eye movements. After \nneuronavigation, bone -conducting headphones were secured, and transducers were coupled to the \nparticipant’s head while they remained still. Eye-tracking calibration was performed at the start of each block, \nprior to initiating the saccade task. (B) Schematic overview of coupling materials and layers. The participant's \nhair and scalp were carefully prepared with ultrasound gel to ensure full coverage of hair follicles and \nminimize air pockets. A thin (~3 millimeters) gel pad was placed on top of the ultrasound gel layer, allowing \nfor visualization and removal of any remaining air bubbles. Another layer of ultrasound gel was applied to \nthe transducer surface, ensuring no air bubbles were present. The transducer was then carefully positioned \nat the stimulation site, guided by neuronavigation for precise targeting. \n \nneuronavigation camera\nneuronavigation screen\nSupplementary figure 4\nA B\nexperimental screen\neyetracker\ntransducer\ntransducer\ngelpad (~  3mm)\ngel\nskin\nboneconducting\nheadphones\n80 cm\n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nTables S1-S9: Additional analyses and statistical outcomes \nThis section provides a summary of key findings from supplementary statistical analyses. \n \nTable S1 | Additional analyses and statistical outcomes examining the effect of condition \nand delay on choice behavior. The model was tested across three different choice domains \n(25%-75%, 20%-80%, and 15%-85%) to assess the robustness of the effects. Bold cells  \nindicate column and row labels, while shaded blue cells  highlight the main comparison \noutcomes. \n \nModel: choice ~ condition(left FEF, right FEF) + delay + (condition(left FEF, right FEF) + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 10 1 0.8 0.14 0.1, 0.2 12 1 <0.001 0.1 0.1, 0.2 13 1 <0.001 0.14 0.1, 0.2 \nLFEF – RFEF 10 1 0.001 -0.25 -0.4, - 0.1  10 1 0.002 -0.2 -0.4, -0.1 8 1 0.005 -0.22 -0.4, -0.1 \nDelay 185 1 <0.001 1.62 1.4, 1.9 347 1 <0.001 2.2 2.0, 2.5 490 1 <0.001 2.82 2.6, 3.1 \nModel: choice ~ condition(left M1, right M1) + delay + (condition(left M1, right M1) + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 0 1 0.9 0.0 -0.1, 0.1 0 1 0.6 0.0 -0.1, 0.1 3 1 0.09 0.07 -0.0, 0.1 \nLM1 – RM1 2 1 0.17 -0.1 -0.2, 0.0 3 1 0.10 -0.1 -0.2, 0.0 4 1 0.044 -0.14 -0.3, -0.0 \nDelay 187 1 <0.001 1.64 1.4, 1.9 359 1 <0.001 2.3 2.0, 2.5 415 1 <0.001 2.98 2.7, 3.3 \n \n \nTable S2 | Additional analyses and statistical outcomes examining the effect of condition \nand delay on choice behavior, excluding trials where the target delay was 0 seconds. The \nmodel was tested across three different choice domains ( 25%-75%, 20%-80%, and 15%-\n85%) to assess the robustness of the effects. Bold cells indicate column and row labels, while \nshaded blue cells highlight the main comparison outcomes. \n \nModel: choice ~ condition(left FEF, right FEF) + delay + (condition(left FEF, right FEF) + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 7 1 0.007 0.12 0.0, 0.2 9 1 0.002 0.13 0.0, 0.2 11 1 <0.001 0.13 0.1, 0.2 \nLFEF – RFEF 8 1 0.004 -0.24 -0.4, - 0.1  8 1 0.005 -0.22 -0.4, -0.1 6 1 0.012 -0.20 -0.4, -0.0 \nDelay 175 1 <0.001 1.60 1.4, 1.8 339 1 <0.001 2.20 2.0, 2.4 481 1 <0.001 2.82 2.6, 3.1 \nModel: choice ~ condition(left M1, right M1) + delay + (condition(left M1, right M1) + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 1 1 0.4 -0.04 -0.1, 0.1 0 1 0.8 -0.01 -0.1, 0.1 1 1 0.3 0.04 -0.0, 0.1 \nLM1 – RM1 0 1 0.8 -0.02 -0.1, 0.1 1 1 0.5 -0.05 -0.2, 0.1 2 1 0.2 -0.09 -0.2, -0.0 \nDelay 175 1 <0.001 1.63 1.4, 1.9 347 1 <0.001 2.28 2.0, 2.5 390 1 <0.001 2.98 2.7, 3.3 \n \n \nTable S3 | Additional analyses and statistical outcomes examining the effect of stimulation \nside, stimulation region and delay on choice behavior, including an interaction between \nside and region. The model was tested across three different choice domains ( 25%-75%, \n20%-80%, and 15%-85%) to assess the robustness of the effects. Bold cells indicate column \nand row labels, while shaded blue cells highlight the main comparison outcomes. \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nModel: choice ~ side x region + delay + (side x region + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 9 1 0.003 0.14 0.0, 0.2 8 1 0.004 0.15 0.0, 0.2 6 1 0.011 0.14 0.0, 0.2 \nSide 12 1 <0.001 -0.26 -0.4, - 0.1  12 1 <0.001 -0.25 -0.4, -0.1 9 1 0.003 -0.22 -0.4, -0.1 \nRegion 5 1 0.032 -0.14 -0.3, -0.0  4 1 0.039 -0.12 -0.2, -0.0 2 1 0.14 -0.08 -0.2, 0.0 \nDelay 310 1 <0.001 1.63 1.4, 1.8 480 1 <0.001 2.27 2.1, 2.5 557 1 <0.001 2.93 2.7, 3.2 \nSide x Region 3 1 0.09 0.16 -0.0, 0.4 3 1 0.11 0.14 -0.0, 0.3 1 1 0.3 0.09 -0.1, 0.3 \n \n \nTable S4 | Additional analyses and statistical outcomes examining the effect of stimulation \nside, stimulation region and delay on choice behavior, including an interaction between \nside and region, excluding trials where target delay was 0 seconds. The model was tested \nacross three different choice domains ( 25%-75%, 20%-80%, and 15%-85%) to assess the \nrobustness of the effects. Bold cells indicate column and row labels, while shaded blue cells \nhighlight the main comparison outcomes. \n \nModel: choice ~ side x region + delay + (side x region + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 7 1 0.009 0.13 0.0, 0.2 7 1 0.009 0.13 0.0, 0.2 6 1 0.018 0.13 0.0, 0.2 \nSide 9 1 0.003 -0.25 -0.4, - 0.1  9 1 0.002 -0.24 -0.4, -0.1 7 1 0.008 -0.21 -0.4, -0.1 \nRegion 6 1 0.012 -0.18 -0.3, -0.0  5 1 0.020 -0.14 -0.3, -0.0 3 1 0.09 -0.10 -0.2, 0.0 \nDelay 288 1 <0.001 1.61 1.4, 1.8 468 1 <0.001 2.25 2.0, 2.5 534 1 <0.001 2.91 2.7, 3.2 \nSide x Region 5 1 0.025 0.23 0.0, 0.4 4 1 0.038 0.19 0.0, 0.4 2 1 0.15 0.13 -0.0, 0.3 \n \n \nTable S5 | Additional analyses and statistical outcomes examining the effect of stimulation \nside, stimulation region, presence of zero -delay trials and delay on choice behavior, \nincluding an interaction between side and region. The model was tested across three \ndifferent choice domains (25%-75%, 20%-80%, and 15%-85%) to assess the robustness of \nthe effects. Bold cells indicate column and row labels, while shaded blue cells highlight the \nmain comparison outcomes. \n \nModel: choice ~ side x region x delay0 + delay + (side x region + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nIntercept 7 1 0.008 0.13 0.0, 0.2 10 1 0.002 0.13 0.0, 0.2 5 1 0.020 0.13 0.0, 0.2 \nSide 10 1 0.001 -0.25 -0.4, - 0.1  9 1 0.003 -0.24 -0.4, -0.1 8 1 0.006 -0.21 -0.4, -0.1 \nRegion 6 1 0.014 -0.17 -0.3, -0.0  4 1 0.040 -0.14 -0.3, -0.0 3 1 0.09 -0.10 -0.2, 0.0 \nDelay0 0 1 0.5 0.08 -0.2, 0.4 0  0.5 0.08 -0.2, 0.3 0  0.5 0.08 -0.2, 0.3 \nDelay 309 1 <0.001 1.63 1.4, 1.8 476 1 <0.001 2.26 2.1, 2.5 554 1 <0.001 2.92 2.7, 3.2 \nSide x Region 5 1 0.033 0.22 0.0, 0.4 3 1 0.066 0.19 -0.0, 0.4 2 1 0.15 0.13 -0.0, 0.3 \nSide x Delay0 0 1 0.74 -0.06 -0.4, 0.3 0 1 0.6 -0.09 -0.4, 0.3 0 1 0.6 -0.10 -0.5, 0.3 \nRegion x \nDelay0 \n2 1 0.16 0.26 -0.1, 0.6 2 1 0.2 0.23 -0.1, 0.6 1 1 0.3 0.18 -0.2, 0.5 \nSid x Reg x \nDel0 \n4 1 0.058 -0.50 -1.0, 0.0 3 1 0.09 -0.45 -1.0, 0.1 2 1 0.14 -0.38 -0.9, 0.1 \n \nTable S6 | Additional analyses and statistical outcomes examining the interaction between \ncondition and baseline GABA+ levels in the FEF on choice behavior. The model was tested \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\nacross three different choice domains ( 25%-75%, 20%-80%, and 15%-85%) to assess the \nrobustness of the effects. Bold cells indicate column and row labels, while shaded blue cells \nhighlight the main comparison outcomes. \n \nModel: choice ~ condition x FEF_GABA + delay + (condition + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nCondition 6 1 0.011 0.64 0.1, 1.2  9 1 0.003 0.66 0.2, 1.2 9 1 0.003 0.65 0.2, 1.1 \nFEF GABA+ 2 1 0.12 0.12 0.0, 0.2  1 1 0.3 0.10 -0.0, 0.2 2 1 0.14 0.16 0.0, 0.3 \nDelay 269 1 <0.001 1.76 1.5, 2.0 433 1 <0.001 2.23 2.0, 2.4 509 1 <0.001 2.88 2.6, 3.1 \nCondition x \nFEF GABA+ \n3 1 0.070 -0.15 -0.3, 0.0 4 1 0.050 -0.15 -0.3, -0.0 4 1 0.037 -0.15 -0.3, -0.0 \nSham x FEF \nGABA+ \n5 1 0.019 0.12 0.0, 0.2 5 1 0.022 0.10 0.0, 0.2 4 1 0.050 0.15 0.0, 0.3 \nLFEF x FEF \nGABA+ \n0 1 0.6 -0.03 -0.2, 0.1 0 1 0.5 -0.05 -0.2, 0.1 0 1 0.9 0.01 -0.2, 0.2 \n \nTable S7 | Additional analyses and statistical outcomes examining the interaction between \ncondition and baseline GABA+ levels in the FEF on choice behavior, excluding trials where \ntarget delay was 0 seconds . The model was tested across three different choice domains \n(25%-75%, 20%-80%, and 15%-85%) to assess the robustness of the effects. Bold cells  \nindicate column and row labels, while shaded blue cells  highlight the main comparison \noutcomes. \n \nModel: choice ~ condition x FEF_GABA + delay + (condition + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nCondition 5 1 0.021 0.84 0.3, 1.4  8 1 0.005 0.84 0.3, 1.4 8 1 0.004 0.79 0.3, 1.3 \nFEF GABA+ 2 1 0.13 0.14 0.0, 0.2  1 1 0.4 0.11 0.0, 0.2 2 1 0.19 0.16 0.0, 0.03 \nDelay 274 1 <0.001 1.75 1.5, 2.0 432 1 <0.001 2.21 2.0, 2.4 546 1 <0.001 2.87 2.6, 3.1 \nCondition x \nFEF GABA+ \n6 1 0.017 -0.21 -0.4, -0.0 6 1 0.011 -0.21 -0.4, -0.0 6 1 0.011 -0.19 -0.3, -0.0 \nSham x FEF \nGABA+ \n7 1 0.008 0.14 0.0, 0.2 4 1 0.041 0.11 0.0, 0.2 4 1 0.035 0.16 0.0, 0.3 \nLFEF x FEF \nGABA+ \n1 1 0.3 -0.08 -0.2, 0.1 2 1 0.18 -0.09 -0.2, 0.0 0 1 0.7 -0.03 -0.2, 0.1 \n \n \nTable S8 | Additional analyses and statistical outcomes examining the interaction between \ncondition and baseline GABA+ levels in the M1 on choice behavior. The model was tested \nacross three different choice domains ( 25%-75%, 20%-80%, and 15%-85%) to assess the \nrobustness of the effects. Bold cells indicate column and row labels, while shaded blue cells \nhighlight the main comparison outcomes. \n \nModel: choice ~ condition x M1_GABA + delay + (condition + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nCondition 0 1 1.0 -0.24 -0.9, 0.4  0 1 0.7 -0.22 -0.9, 0.5 1 1 0.4 -0.01 -0.6, 0.6 \nFEF GABA+ 0 1 0.0 -0.03 -0.2, 0.1 1 1 0.3 -0.06 -0.2, 0.0 1 1 0.4 -0.09 -0.3, 0.1 \nDelay 191 1 <0.001 1.72 1.5, 2.0 264 1 <0.001 2.24 2.0, 2.5 369 1 <0.001 2.97 2.7, 3.3 \nCondition x M1 \nGABA+ \n1 1 0.5 0.07 -0.1, 0.3 1 1 0.5 0.07 -0.1, 0.3 0 1 0.8 0.02 -0.1, 0.2 \nSham x M1 \nGABA+ \n0 1 0.5 -0.04 -0.2, 0.1 1 1 0.3 -0.06 -0.2, 0.1 1 1 0.4 -0.08 -0.3, 0.1 \nLM1 x M1 \nGABA+ \n0 1 0.7 0.03 -0.1, 0.2 0 1 0.9 -0.01 -0.2, 0.2 1 1 0.4 -0.07 -0.2, 0.1 \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint \n\n \nTable S9 | Additional analyses and statistical outcomes examining the interaction between \ncondition and baseline GABA+ levels in the M1 on choice behavior, excluding trials where \ntarget delay was 0 seconds. The model was tested across three different choice domain s \n(25%-75%, 20%-80%, and 15%-85%) to assess the robustness of the effects. Bold cells  \nindicate column and row labels, while shaded blue cells  highlight the main comparison \noutcomes. \n \nModel: choice ~ condition x M1_GABA + delay + (condition + delay | sub)   \nData: choice domain 25% - 75% choice domain 20% - 80% choice domain 15% - 85% \n c2 Df p b 95%-CI c2 Df p b 95%-CI c2 Df p b 95%-CI \nCondition 0 1 0.6 -0.24 -1.0, 0.5  0 1 1.0 0.66 0.2, 1.2 0 1 0.7 0.01 -0.6, 0.6 \nM1 GABA+ 0 1 0.8 -0.01 -0.1, 0.1  0 1 0.8 0.10 -0.0, 0.2 1 1 0.5 -0.06 -0.2, 0.1 \nDelay 185 1 <0.001 1.71 1.5, 2.0 251 1 <0.001 2.23 2.0, 2.4 364 1 <0.001 2.96 2.7, 3.3 \nCondition x M1 \nGABA+ \n0 1 0.6 0.06 -0.1, 0.3 0 1 0.5 -0.15 -0.3, -0.0 0 1 1.0 0.01 -0.2, 0.2 \nSham x M1 \nGABA+ \n0 1 0.9 -0.01 -0.1, 0.1 0 1 0.6 -0.03 -0.2, 0.1 0 1 0.5 -0.06 -0.2, 0.1 \nLM1 x M1 \nGABA+ \n0 1 0.5 0.06 -0.1, 0.2 0 1 0.9 0.01 -0.2, 0.2 0 1 0.6 -0.05 -0.2, 0.1 \n \n.CC-BY 4.0 International licensemade available under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is \nThe copyright holder for this preprintthis version posted March 16, 2025. ; https://doi.org/10.1101/2025.03.16.643494doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}