Eye-hand coordination beyond vision: preceding gaze history shapes reaching memories

Eye-hand coordination beyond vision: preceding gaze history shapes reaching memories | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Eye-hand coordination beyond vision: preceding gaze history shapes reaching memories Naotoshi Abekawa , Hiroaki Gomi doi: https://doi.org/10.1101/2025.11.11.687771 Naotoshi Abekawa 1 Communication Science Laboratorie s, NTT, Inc. 3-1 Morinosato Wakamiya, Atsugi, Kanagawa, 243-0198, Japan Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: naotoshi.abekawa{at}ntt.com Hiroaki Gomi 1 Communication Science Laboratorie s, NTT, Inc. 3-1 Morinosato Wakamiya, Atsugi, Kanagawa, 243-0198, Japan 2 Faculty of Science and Engineering, Waseda University , 3-4-1 Okubo, Shinjuku-ku, Tokyo, 169-8555, Japan Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Skilled reaching behaviors, such as grasping a cup or catching a ball, require adaptive changes in response to error signals. Our natural reaches are usually coordinated with preceding eye movements. Despite this behavioral relevance, the potential linkage between preceding eye movements and memory formation for reaching has been largely overlooked. We show that two opposing visuomotor maps for reach, which normally prevent learning due to interference, can be learned when each is associated with a different immediately preceding eye movement. We further show that the generalization pattern of reach learning across different gaze conditions can be explained by the gaze state’s distance computed with temporal weighting. These findings indicate that reaching memories are encoded in tight association with the recent history of preceding gaze states. Our results expand the functional role of eye-hand coordination beyond visual guidance, revealing its importance in orchestrating flexible motor learning. Significance In everyday behavior, goal-directed arm movements are typically preceded by eye movements, and eye-hand coordination has been regarded mainly as a functional mechanism for visual guidance of action. This study expands that view by showing that preceding gaze states also shape how the brain learns and retrieves motor memories of reaches. We demonstrate that conflicting visuomotor maps, which are usually difficult to learn because of memory interference, can be learned simultaneously when they are associated with different preceding eye movements. Furthermore, generalization experiments show that reach learning depends on the recent history of gaze states within a brief time window before reach onset. These findings indicate that preceding gaze history plays a significant role in shaping flexible reach memories. Introduction In an ever-changing world, our skillful arm-reaching movements rely on motor learning, which involves the formation and updating of sensorimotor maps called internal models 1 . A key issue in motor learning is understanding how the brain represents internal models, especially regarding their association with various body states 2 , 3 . Reaches are typically preceded by eye movements toward the target, hence much research has been conducted on how the brain spatiotemporally coordinates eye and hand movements, both in daily activities and laboratory tasks, particularly in relation to target representation 4 , 5 . In real-world scenarios, reach learning often occurs under coordinated eye-then-reach movements 6 . In tennis, for example, players track a moving ball with their eyes, hitting it while foveating it. Sometimes, their gaze shifts from the net to the ball just before taking the shot. Novel reaching skills, such as forehand strokes and backhand volleys, are acquired through repeated practice under dynamic eye-hand chains. This raises an unexplored question that spans both fields of eye-hand coordination and motor learning: Does the preceding dynamic gaze state inherently contribute to the encoding of motor memory for subsequent reaching movements? There would be two opposing views on this matter. One view highlights the functional aspect of eye-hand coordination as target representation and deems it unrelated to the reach learning. From this view, gaze states serve two primary functions: 1) providing accurate localization of the reach target via foveation and 2) serving as a coordinates reference for encoding target location. During reach planning, target locations are initially represented in a gaze-centered frame and updated through eye movements. 7 – 9 Such gaze-dependent signals are transformed into gaze-independent signals (e.g., hand-to-target trajectories), subsequently used with internal models to generate reach motor commands. Thus, in this sequential processing scheme, preceding gaze states are generally considered irrelevant to motor learning that primarily occurs downstream of target representation. An alternative view focuses on the direct coupling of sensorimotor signals in the interwoven mechanisms of the eye and hand motor systems. Such motor-coordination mechanisms enable multiple effectors to work cooperatively, which is observed at various levels of the control hierarchy (e.g., bimanual coordination and multi-joint movements) 10 . From this view, motor commands to one effector depend heavily on the state of the other effector. Thus, in closely linked eye-hand coordination 11 – 13 , such as a single motor sequence, preceding gaze states may play a role in representing reach internal models. Consistent with the latter view, motor-learning studies have demonstrated that learning reaching movements with one arm depends on the state of the other arm 14 – 17 . Several of these studies used context-dependent learning tasks in which two opposing perturbations (e.g., clockwise and counterclockwise velocity-dependent force fields) randomly switched across trials. Such opposing fields are usually difficult to learn due to memory interference but can be learned simultaneously when each is uniquely associated with a different movement pattern in the other arm. In unimanual and bimanual tasks, simultaneous learning is facilitated even when contextual movements occur before the onset of movement for adaptation 16 – 19 . Studies suggest that the prior and current hand states have a contextual effect on the separation of motor memories, thus are crucial components in the representation of reach internal models. In particular, the prior contextual effect is consistent with recent theories that emphasize preparatory neural states in motor-related cortices in motor learning 20 – 22 . Most studies, however, have been limited to focusing on hand-related states as the learning context, and it remains untested whether similar parallel conclusions can be drawn for other body states. Although our recent study has demonstrated that the gaze location relative to the target (i.e., reaching towards a central or peripheral target) can separate visuomotor memories, the investigation was confined to specific static gaze conditions 23 . Despite its relevance for natural and common human behaviors, it is yet unknown whether/how preceding dynamic gaze states contribute to reach learning. We address this issue and demonstrate, through a series of experiments, that participants can learn two opposing visuomotor maps when each is associated with a distinct preceding dynamic gaze state. This eye-dependent learning effect is sensitive to the time interval between eye and reaching movements. Our findings provide evidence that memory encoding for reaching is tightly linked to the recent history of gaze states. Preliminary findings from Experiment 1 were previously reported in a conference proceeding 24 . Results Experiment 1 In Experiment 1, we investigated whether humans are able to learn two opposing internal models when each was associated with either a static or dynamic gaze condition. Participants moved a stylus on a digitizing tablet, and the position of the stylus was presented as a cursor on the monitor ( Figure 1A ). They performed forward reaching movements toward a single target under three gaze conditions (two static and one dynamic) in random order ( Figure 1B ). Under the two static conditions, participants fixated a cross throughout a trial and reached toward a peripheral target (PER) or foveal target (FOV). Under the dynamic condition (SACCADE), the trial started as in PER, after which the fixation cross jumped to the reach target. Participants had to make a saccade, followed by reaching toward the foveal target. After saccade ended, fixating the target was required until the completion of the reach. In the learning session, participants learned visuomotor rotations of the cursor feedback. The direction of rotation, clockwise (CW) and counterclockwise (CCW), changed across trials, but was associated with either a static or dynamic gaze condition (i.e., CW for PER and FOV; CCW for SACCADE; Figure 1B ). This association manner was counterbalanced across participants. Download figure Open in new tab Figure 1: Simultaneous learning based on static and dynamic gaze states (Experiment 1). A, Experimental apparatus. Participants performed reaching movements on a digitizing tablet to move a visual cursor toward a target on a monitor. B, Gaze conditions. Under static conditions, reaching was performed with gaze fixed on the target (FOV) or elsewhere (PER). Under the dynamic condition (SACCADE), participants made a saccade toward the reach target then reached toward the foveal target. Three gaze conditions were interleaved randomly across trials. Visuomotor rotations were applied, with their directions associated with each condition (curved arrows; CW for static and CCW for dynamic; See main text for details). C, Trial sequence. In PER trials, the trial started when participants maintained fixation and the reach target appeared. Reaching was then initiated following a go beep. FOV trials were identical to the PER trials except that fixation was on the reach target. In SACCADE trials, participants made a saccade in response to a fixation jump then reached toward the target at the go beep. The go beep timing depended on the gap-duration group (Short, Middle, or Long) to manipulate the gap duration between saccade end and reach initiation. Each participant was assigned to one group. D, Learning curves for each gaze condition across three gap-duration groups. Initial reach direction was used as the learning measure, with its sign adjusted to match the rotation direction (positive values indicate learning). Data were averaged within each block and plotted as mean ± SEM across participants. The angle of visuomotor rotation gradually increased up to 20° (black line). E, Final learning levels of initial reach direction (averaged over the last four blocks in the learning session), shown for a two-way mixed ANOVA with gaze condition (PER, FOV, and SACCADE; within-participants) and Gap duration (Short, Middle, and Long; between-participants). Data are presented as mean ± SEM across participants. Asterisks indicate significant differences (Tukey post hoc tests; * p<0.05, ** p<0.01, *** p<0.001). F, Final learning levels (same data as in E) replotted for each Gap duration, with individual participant data shown as gray lines. G, H Aftereffect magnitudes (averaged over the four blocks of the post-learning session), shown in the same format as in E and F. Given that distinct static gaze conditions (i.e., PER and FOV) serve as effective contextual cues for simultaneous reach learning under conflicting conditions 23 , we designed three gaze conditions to test the contextual effect of eye movements. Even in the SACCADE condition, static gaze phases exist both before and after eye movements. In a two-condition paradigm (e.g., PER and SACCADE, or FOV and SACCADE), these static phases could serve as contextual cues, making it difficult to isolate the effect of dynamic eye movements. In contrast, our three-condition design enables a clear dissociation between static and dynamic gaze states. In the current three-condition design, SACCADE trials started in the same way as PER, so pre-saccade gaze states were identical between the two conditions. Because different visuomotor rotations were applied to both conditions, pre-saccade static gaze states could not separate motor memories. Likewise, post-saccade static gaze states under SACCADE were identical to those under FOV, resulting in memory interference when relying on these post-saccade static gaze states. Thus, to enable simultaneous learning across the three gaze conditions, the brain must rely on the contextual differences between static and dynamic gaze conditions. Throughout the experiment, three gaze conditions (PER, FOV, and SACCADE) were randomly presented within a block of 16 trials. The experiment consisted of three sessions: a pre-learning session with normal visual feedback, learning session with visuomotor rotations, and post-learning session with normal vision. In the learning session, the rotation angle gradually increased up to 20 degrees. Further details are provided in Methods. To assess learning, we quantified the initial direction of reaching movements. The sign of reach directions was adjusted for each gaze condition so that a positive value indicates the compensating direction against each visuomotor rotation (CW or CCW). Such simultaneous learning would be impacted by the temporal duration between saccade termination and reach initiation under SACCADE since the longer this duration is, the more similar SACCADE and FOV become. To manipulate this gap duration, 36 participants were assigned to one of three timing groups (12 participants per group): Short, Middle, and Long. The “Go” beep, cuing participants to initiate the reach, was presented 0.7, 1.0, and 1.6 s after the target presentation in the Short, Middle, and Long groups ( Figure 1C , right). The actual time between saccade termination and reach onset obtained in the experiment was 264 ± 97 ms (91–408 ms) for the Short, 541 ± 49 ms (466–638 ms) for the Middle, and 1213 ± 95 ms (109 –1449 ms) for the Long (mean ± SD and range across participants). To match the time course between conditions, the timing of the “Go” beep also varied according to the three timing groups under the static gaze conditions (PER and FOV; Figure 1C , left). See Methods and Figure 1C for details on the time procedure for the task. We expected that inserting a long gap duration (i.e., Long group) would impair learning in SACCADE and FOV. If we assume that the representation of dynamic gaze state decays over time, the longer interval would make SACCADE equivalent to simple reaching toward a foveal target, thereby rendering SACCADE and FOV indistinguishable in terms of dynamic gaze state. In contrast, PER—reaching to a peripheral target—provides a distinct context, as reported in our previous study 23 . Therefore, learning in PER would remain unaffected by the gap duration. As shown in Figure 1D (left), participants in the Short group exhibited clear learning progress under all three gaze conditions. In the post-learning session, reach directions remained positive despite the restoration of normal visual feedback, indicating clear aftereffects. Consistent with our prediction, in the Middle and Long groups ( Figure 1D , center and right), learning under FOV and SACCADE (green and pink curves) appeared to be impaired, whereas learning under PER remained largely unchanged. To quantify these observations, the final learning level was calculated from the last four blocks of the learning session. A two-way mixed ANOVA was conducted with Gaze condition (PER, FOV, and SACCADE; within participants) and Gap duration (Short, Middle, and Long; between-participants) as factors. As shown in Figure 1E , the analysis revealed significant main effects of Gaze condition (F(2,66)=14.78, p<0.001) and Gap duration (F(2,33)=3.63, p=0.038), as well as a significant interaction between Gaze condition and Gap duration (F(4,66)=8.71, p<0.001). Specifically, this interaction indicates that the effect of gap duration on learning differed across gaze conditions. Returning to the results of the Short group ( Figure 1D , left), learning progressed similarly across all three gaze conditions. The final learning level was significantly different from 0 under all conditions ( Figure 1F , Short; PER: t(11)=9.5, p-adj.=3.5×10⁻⁶; FOV: t(11)=9.7, p-adj.=3.0 × 10⁻⁶, SACCADE: t(11)=15.7, p-adj.=2.1 × 10⁻⁸; Bonferroni-corrected t-tests) and did not differ significantly between gaze conditions ( Figure 1E ; simple main effect of Gaze: F(2,22)=0.037, p=0.96). These results indicate that static and dynamic gaze contexts can serve as contextual cues for separating the learning of conflicting visuomotor rotations. For the Middle group ( Figure 1D , center), learning still progressed under all conditions, as the final learning levels remained significantly greater than 0 ( Figure 1F , Middle; PER: t(11)=9.9, p-adj.=2.5×10⁻⁶; FOV: t(11)=6.8, p-adj.=8.7×10⁻ 5 , SACCADE: t(11)=7.1, p-adj.=5.7×10 -5 ). However, the final learning levels differed significantly among gaze conditions (simple main effect of Gaze: F(2,22)=7.68, p=0.003). As shown in Figure 1E , Tukey post-hoc tests revealed that learning for PER was significantly greater than for FOV (t(11)=4.43, p=0.003), whereas no significant difference was observed between PER and SACCADE (t(11)=1.15, p=0.50) or between FOV and SACCADE (t(11)=−2.39, p=0.09). These results indicate that the inserted gap duration induced learning interference for FOV and SACCADE, leading to a partial failure of simultaneous learning. The trend observed in the Middle group was more pronounced in the Long group ( Figure 1D , right). The final learning level differed significantly between gaze conditions (simple main effect of Gaze: F(2,22)=19.0, p<0.001). As shown in Figure 1E , Tukey post-hoc tests revealed that learning for PER was significantly greater than for both FOV (t(11)=5.89, p<0.001) and SACCADE (t(11)=5.32, p<0.001). In contrast, no significant difference was observed between FOV and SACCADE (t(11)=1.78, p=0.22). In addition, the final learning level was significantly different from zero for PER and FOV, but not for SACCADE ( Figure 1E , Long; PER: t(11)=12.9, p-adj.=1.6×10 -7 ; FOV: t(11)=5.1, p-adj.=0.001, SACCADE: t(11)=2.1, p-adj.=0.18). These results indicate that when a longer gap duration was inserted, FOV and SACCADE became similar as gaze contexts, leading to severe learning interference. We next examined how the inserted gap duration affected learning. As shown in Figure 1E , the final learning levels for FOV and SACCADE were reduced when a gap duration was introduced (simple main effect of Gap duration, FOV: F(2,33)=4.46, p=0.019, SACCADE: F(2,33)=8.94, p<0.001), whereas learning for PER was unaffected (F(2,33)=0.29, p=0.75). Post-hoc tests further revealed that learning in the Long group was reduced relative to the Short group for FOV (t(33)=2.8, p=0.02) and relative to both the Short (t(33)=3.8, p=0.002) and Middle (t(33)=3.5, p=0.004) groups for SACCADE. These results suggest that inserting a gap duration reduced the distinctiveness of the FOV and SACCADE contexts, leading to learning interference. In contrast, PER remained a distinct contextual cue, allowing learning to proceed largely unaffected by the gap duration. The patterns observed in the final learning level were generally reflected in the aftereffects estimated from the four blocks of the post-learning session ( Figure 1G, H ). A two-way mixed ANOVA revealed significant main effects of Gaze condition (F(2,66)=4.42, p=0.016) and Gap duration (F(2,33)=7.22, p=0.003), as well as a significant interaction (F(4,66)=4.21, p=0.004). Notably, in the Short group, aftereffect magnitudes significantly differed from 0 for all gaze conditions (p-adj.<2.0×10 -5 ) and did not differ across conditions (simple main effect of gaze: F(2,22)=0.85, p=0.44). These results indicate successful context-dependent retrieval of visuomotor mappings corresponding to the immediately preceding gaze states. Consistent with the results for the final learning level, when a gap duration was inserted (i.e., in the Middle and Long groups), aftereffects for FOV and SACCADE were reduced, whereas those for PER remained largely unaffected. In particular, in the Long group, aftereffects differed between gaze conditions (simple main effect of gaze: F(2,22)=7.55, p=0.003; Figure 1G ) and the aftereffect for SACCADE was not significantly different from 0 (t(11)=2.58, p-adj.=0.08; Fig.1 H ). Before drawing conclusions from Experiment 1, we considered the possibility that differences in reaching parameters across gaze conditions, rather than gaze behavior, might have served as learning contexts. To test this, we analyzed reaching kinematics during the pre-learning session in the Short group. However, no parameters significantly differed between SACCADE and PER and between SACCADE and FOV ( Supplementary Table 1 ). These results rule out a kinematics-based explanation and suggest a specific contribution of static and dynamic gaze contexts to reach learning. These results rule out a kinematics-based explanation and suggest the contribution of distinct static and dynamic gaze contexts to reach learning. Notably, the contextual effect of dynamic gaze states decays over time. View this table: View inline View popup Download powerpoint Supplementary Table 1. Reaching kinematics for short group during pre-learning session (normal visual feedback) in Experiment 1 Each movement-related parameter was compared between PER and SACCADE and between FOV and SACCADE using paired T-tests. A positive reach-direction value indicates a clockwise deviation from a straight path to the target. For each participant, measures were averaged across the four blocks of the pre-learning session. The table shows the mean ± SD across participants. No measure differed significantly in both the PRE-SACCADE and the FOV-SACCADE comparisons. Thus, the simultaneous learning observed under the three gaze conditions in Experiment 1 cannot be explained by a contextual effect of different reach-related parameters across conditions. Control Experiment 1 We further conducted a control experiment to test if the visual cue of fixation jump, rather than eye movements, could facilitate learning. The paradigm was the same as in Experiment 1, except that SACCADE was replaced with the fixation jump condition (FIX-JUMP) ( Suppl. Fig.1A ). Under FIX-JUMP, the fixation cross jumped to the target, but participants were instructed to maintain their gaze at the original fixation location without making any eye movements. As in Experiment 1, one direction of visuomotor rotation was associated with two static conditions (i.e., PER and FOV), and the other direction was associated with FIX-JUMP. If visual cues are sufficient for simultaneous learning, comparable learning should be observed across all three conditions, as in the Short group of Experiment 1. In contrast, if gaze states, rather than visual cues, are crucial, learning should fail for PER ad FIX-jump, as these conditions share the same gaze states (i.e., reaching to a peripheral target) while being associated with opposing visuomotor rotations (CW and CCW; See Suppl. Fig.1A ). Meanwhile, learning should progress for FOV, as it has a unique gaze state (i.e., reaching to a foveal target). Consistent with latter prediction, the results showed that learning was evident only under FOV, whereas PER and FIX-JUMP showed minimal improvement, indicating a failure of simultaneous learning ( Supplementary Figure 1B ). This pattern was statistically confirmed by quantifying the final learning level and aftereffects (see Supplementary Figure 1C ). These results indicate that visual cues alone were insufficient, and that static versus dynamic gaze states are crucial for separating motor memories. Download figure Open in new tab Supplementary Figure 1: Control Experiment 1. A. Trial conditions. PER and FOV were identical to those in Experiment 1. In FIX-JUMP, the fixation cross jumped to the target, but participants were required to maintain fixation at the original location and reach toward the target in response the go beep, resulting in gaze behavior identical to PER. The direction of rotation was associated with static and visual dynamic conditions (i.e., CW for PER and FOV, CCW for FIX-JUMP). If visual cues alone were sufficient, simultaneous learning across conditions would be expected. B. Learning curves under each condition. Initial reach direction is shown (mean ± SE across participants). Figure formats are the same as in Figures 1D . Learning progressed only in FOV (pink curve), whereas PER and FIX-JUMP showed little improvement. These results are consistent with our prediction: PER and FIX-JUMP shared the same gaze states (reaching to a peripheral target), while two opposing rotations (CW and CCW) were applied, leading to learning interference. The finding that learning progressed only for FOV likely reflects differences in static gaze state, consistent with our recent work 23 : foveal reaching under FOV versus peripheral reaching under PER and FIX-JUMP. C. Final learning level and aftereffect magnitudes (estimated over the last four blocks of the learning phase and the four blocks of the post-learning phase, respectively). Bar graphs show the mean across participants (± SE), with gray lines indicating individual participant data. A one-way repeated-measures ANOVA revealed a significant effect of condition on final learning levels (F(2,22)=10.7, p<0.001). Post hoc Tukey tests confirmed that FOV was greater than both PER (t(11)=4.9, p=0.001) and FIX-JUMP (t(11)=3.9, p=0.006). Aftereffect magnitudes showed a similar pattern: a significant effect of condition (F(2,22)=6.8, p=0.005), with FOV greater than both PER (t(11)=4.6, p=0.002) and FIX-JUMP (t(11)=3.3, p=0.02). The minimal learning effects under PER and FIX-JUMP suggest that dynamic visual cues alone are insufficient as learning contexts. Experiment 2 Experiment 1 demonstrated that static versus dynamic gaze conditions can facilitate simultaneous learning. In Experiment 2, we further investigated whether distinct patterns of eye movements also serve as effective learning contexts. In each trial, participants performed one of two types of sequential saccades, followed immediately by reaching toward the target ( Figure 2A , top). Each trial began with gaze fixation at F0, after which the fixation cross jumped to F1 then to F2. Participants were asked to make successive saccades from F0 to F1 then F1 to F2 as quickly and smoothly as possible. The final fixation point (F2) always coincided with the reach target. Therefore, participants performed reaching toward a foveal target while maintaining fixation on F2 throughout the reach. The two saccade conditions were labeled by initial saccade direction (i.e., from F0 to F1): Left-side and Right-side. Download figure Open in new tab Figure 2: Simultaneous learning based on different movement patterns of saccades (Experiment 2). A, Trial conditions. In the learning session, participants performed either Left-side or Right-side sequential saccade (random order across trials), followed immediately by a reach toward the foveal target (upper panels). The direction of visuomotor rotation was associated with each condition. In the learning and probe sessions, two types of probe trials were additionally introduced (lower panels). In the saccade probe , participants executed the same eye-hand movements as in the learning trials, whereas in the fixation probe , they performed the reaching movement while maintaining fixation on an intermediate via point. In both probe trials, no cursor feedback was provided, enabling assessment of the retrieval of learned motor outputs under different gaze patterns. B – C, Learning curves for two saccade conditions (B) and final learning levels and aftereffect magnitudes (C). Data formats are the same as in Figures 1D and 1F , respectively. D, Learning effects in the learning and probe sessions. Initial reach direction averaged in this session was shown for learning trials, saccade probe, and fixation probe (gray lines: individual data; circles: group mean ± SEM). Asterisks denote statistically significant differences (** p<0.01 and *** p<0.001), and “ns” denotes non-significance. In the learning session, the direction of visuomotor rotation was pseudo-randomly selected across trials, but uniquely associated with the saccade condition (i.e., CW for Left-side saccade and CCW for Right-side saccade; Figure 3A ). The results indicate that participants gradually changed the reach direction over trials for both saccade conditions ( Figure 2B ). Final learning levels were significantly greater than 0 for both conditions (p-adj.<8.9×10 -5 ) and did not differ between them (paired t-test, t(11)=0.16, p=0.88; Figure 2C ). Similarly, aftereffect magnitudes were greater than 0 (p-adj.<0.0013 for both conditions) and did not differ between conditions (t(11)=0.89, p=0.39). Download figure Open in new tab Figure 3: Learning generalization across static and dynamic gaze states (Experiment 3) A, Session design. In the learning session, participants made a saccade to the reach target, followed immediately by a reach toward the foveal target. They adapted to a visuomotor rotation (single direction; either CW or CCW). In the subsequent probe session, generalization of learning was tested across various gaze states, with no cursor feedback provided during reaching. To prevent memory decay, learning and probe trials were consistently presented in series. B, Fourteen gaze states tested in the probe session. Probe 1 was identical to the learning condition, in which reaching was preceded by a rightward saccade of 10 cm. Information about eye movements for dynamic probes (Probe 1–9) is denoted in brackets, indicating the horizontal start and end gaze positions relative to the reach target. For example, [-10, 0] in Probe 1 indicates that gaze was initially fixed 10 cm to the left of the reach target then shifted to the target location. Under the static probe conditions (Probes 10–14), values in brackets denote fixation location relative to the reach target. C, Example pattern of eye-hand movements. Solid lines show horizontal eye position (scale on the left), and dashed lines show forward hand position (scale on the right), both plotted relative to reach onset. D, Learning curve in the learning session. Initial reach direction (mean ± SE across participants) is plotted across blocks. The thin line indicates the angle of visuomotor rotation. E, Generalization in the probe session. Initial reach direction was measured in the probe trials without cursor feedback, reflecting retrieval of learned motor outputs. See panel B and the main text for details of the 14 probe conditions. Data are presented as mean ± SE across participants. F, Schematic for computing gaze-state distance. The gaze-state distance for each probe condition was calculated relative to the gaze state obtained in the learning trial. The difference in horizontal eye position (time-aligned to reach onset) was first computed and squared at each time point. These differences were then temporally weighted, with the weight decreasing as the time from reach onset increased. Finally, the temporally weighted differences were integrated over time. G, Generalization against gaze-state distance. Data for the 14 probe conditions (shown in panel B) were replotted as a function of gaze-state distance with a fitted curve. H, Estimated temporal weighting function (Gaussian). The function is shown with 95% confidence intervals based on bootstrap resampling. The arrows indicate three levels of the standard deviation of the Gaussian function. See main text for details In this task, the initial and final fixation points (F0 and F2 in Figure 2A ) were identical across the two saccade conditions, indicating that memory separation had to rely on different eye-movement patterns. An alternative possibility, however, was that the static gaze state at the intermediate point (F1) during the two sequential saccades might act as a contextual cue. To test this, we introduced a learning-and-probe session following the learning session, in which two learning and two probe conditions were interleaved across trials ( Figure 2A ). The learning conditions replicated the trials in the learning session (i.e., Left-side and Right-side saccade). The probe conditions included a saccade probe, in which participants executed sequential saccades followed by reaching, and a fixation probe, in which participants maintained fixation on F1 before and during reaching ( Figure 2A , bottom). Under both probe conditions, no visual feedback of the cursor was provided, enabling assessment of memory retrieval. If learning relied on dynamic gaze states, reach directions comparable to learning trials would appear in the saccade probes; if it relied on static gaze states they would appear in the fixation probes. The results indicate that reach directions for the saccade probe were comparable to those for the learning trial, whereas those for the fixation probe were smaller ( Figure 2D ). One-way ANOVA revealed a significant main effect of condition (learning trials, saccade probe, fixation probe; F(2,22)=18.7, p<0.001). Post hoc Tukey comparisons showed that the saccade probe was not different from learning trials (t(11)=0.03, p=1.0), whereas the fixation probe was different from both learning trials (t(11)=4.3, p=0.003) and the saccade probe (t(11)=5.2, p<0.001). These results indicate that the preceding dynamic gaze states (i.e., different eye movement patterns) can separate two opposing motor memories for reaching. Control Experiment 2 Regarding the results of Experiment 2, one possibility remains: learning may be facilitated by different visual patterns of fixation jumps. To test the contextual effect of such visual information, rather than eye movements, we conducted Control Experiment 2, in which a fixation cross jumped either to the left or right, as in Experiment 2 ( Suppl. Fig. 2A ). Regardless of the direction of the cross jump, participants were instructed to make a single saccade toward the target, immediately followed by reaching to the foveal target. The direction of visuomotor rotation was associated with the direction of cross movement. The result showed that participants did not exhibit clear learning for either condition ( Suppl. Fig. 2B ), likely due to memory interference. Crucially, the final learning level was not significantly different from 0 in either conditions ( Suppl. Fig. 2C ; Left-side; t(11)=0.49,p-adj.=1, Right-side; t(11)=2.25, p-adj.=0.09). As in Control Experiment 1, these results suggest that visual information about fixation cross jumps is not sufficient for simultaneous learning. Rather, eye movements, not visual information, provide the contextual cue necessary to separate motor memories. Download figure Open in new tab Supplementary Figure 2: Control Experiment 2. A. Task design and trial conditions. The movement of the fixation cross was identical to that in Experiment 2, and two conditions were defined by the direction of the first movement of the cross (left vs. right). Importantly, regardless of condition, participants were instructed to make a single saccade to the reach target upon fixation appearance, followed immediately by a reaching movement. The direction of rotation was associated with conditions (i.e., CW for left-side displacement and CCW for right-side displacement). If visual cues alone were sufficient, simultaneous learning across conditions would be expected. B. Learning curves under each condition. Initial reach direction is shown (mean ± SE across participants). Figure formats are the same as in Figures 1D or 2B . As in other figures, the sign of reach direction was adjusted so that positive values indicate compensation for the visuomotor rotation. In the graph, the Right-side condition appeared to show slight learning (red curve), whereas the Left-side condition did not (blue curve). C. Final learning levels and aftereffect magnitudes (estimated over the last two blocks of the learning phase and the first two blocks of the post-learning phase, respectively). Bar graphs show the mean across participants (± SE), with gray lines indicating individual participant data. Post hoc comparisons with Bonferroni correction showed that final learning levels did not differ significantly from 0 in either the Left-side (t(11)=0.49, p-adj.=1.0) or Right-side condition (t(11)=2.25, p-adj.=0.09), indicating a failure of learning. Regarding aftereffect magnitudes, the Right-side condition differed from 0 (t(11)=2.9, p=0.03), whereas the Left-side condition did not (t(11)=0.50, p=1.0), again indicating a failure of simultaneous learning. These results suggest that, when eye movement states were identical, visual cues from fixation cross jumps alone were insufficient to establish context-dependent simultaneous learning. Experiment 3 Finally, we investigated which factors in dynamic-gaze contexts contribute to reaching adaptation, using a generalization paradigm. Participants adapted to a single visuomotor rotation (20° CW or CCW; counterbalanced across participants). In each learning trial, they executed a rightward saccade toward the reach target, followed immediately by a reach to the foveal target ( Figure 3A ). In the subsequent probe session, we tested learning generalization to 14 possible gaze conditions ( Figure 3B ), with no cursor feedback. For the dynamic probes (probes 1–9 in Fig.3B ), reaching movements were preceded by saccades, the locations and amplitudes of which varied across probes. For the static probes (probes 10-14), reaches were performed while fixation was maintained at a certain location. Reach target remained fixed at the position used during the learning session. To prevent rapid decay of memory, each probe trial was preceded by a learning trial identical to those in the learning session (20° rotation; Figure 3A ). For clarity, we plotted typical eye (horizontal position, solid lines) and hand (forward position, dashed lines) movements from the learning and selected probe conditions to illustrate the temporal eye-hand relationship ( Figure 3C ). As in Experiments 1 and 2, we measured the sign-adjusted initial reach direction. A positive value indicated appropriate compensation for the rotation in the learning session and retrieval of learned movements in the probe session. As shown in Figure 3D , reach directions gradually increased over the course of the learning session, indicating clear progress in learning. In the subsequent probe session, the largest reach direction was observed for the learning probe, which had the same gaze state as under the learning condition (probe 1; pink square in Figure 3E ). We regarded this learning probe as a reference and compared it with the remaining 13 probe conditions using Dunnett’s test. Large reach directions were found for dynamic probes 2 and 3, which did not differ significantly from the learning probe (p=0.99 for both; Figure 3E ). These probes included saccades with the same endpoint and direction as in the learning condition ( Figure 3B ). Other dynamic gaze probes showed significantly lower reach directions, indicating limited transfer of learning. Specifically, these probes shared only endpoint (probe 4; p<0.001), start point (probes 5, 6, and 7; p<0.05, p<0.001, and p<0.001, respectively), or amplitude (probes 8 and 9; p<0.001 for both) with the learning condition. These results indicate that learning of reaching movements is robustly transferred between dynamic gaze conditions when the preceding saccade shares both direction and endpoint. Experiment 1 suggested that dynamic and static gaze states may activate distinct neural states for separating motor memories. Consistent with this view, we found that learning transfer from the saccade-accompanied learning condition to static probes requiring fixation was limited ( Figure 3B , probes 10–14). As shown in Figure 3E , the reach directions for these static probes were significantly smaller than for the learning probe (p<0.001 for all). Notably, under the learning condition, the eye fixated on the reach target after saccade, and its gaze state was identical as that for static probe 13 (see green and blue solid lines in Figure 3C ). Thus, the limited transfer at probe 13 further supports the contribution of eye movements preceding each reach during reach learning. In a comparison between static probes, the reach direction for probe 13 (fixating the target) appeared greater (i.e., greater learning transfer) than for other static probes (probes 10, 11, 12, and 14). For the dynamic probes, as already noted, saccade endpoint and direction—rather than start point or amplitude—were crucial for learning transfer. Taken together, these results indicate that learning transfer depends on the recent history of gaze states before reach onset. To estimate such periods and capture the generalization pattern, we developed a model to compute gaze-state distance between the learning condition and each probe condition. As shown in Figure 3F , for each probe condition, the squared difference from the learning condition was calculated at each time point, using the horizontal eye position aligned to reach onset. The squared difference was temporally weighted by a Gaussian function with mean 0 and standard deviation σ t : where x P ( t ) and x L ( t ) denote the horizontal eye position at time t for the probe P and learning condition, respectively. For temporal weighting, only the first half of the Gaussian was used, with the gain set to 1 at time 0 (i.e., reach onset) and decreased with earlier time points. The time-weighted distance of d P ( t ) was then integrated over time, and the square root of this integral was defined as the gaze-state distance of probe P ( D P ) from the learning condition, as follows: Due to the data-recording window, time integration was executed between -1.5 and 0 s. Figure 3G shows the reach direction plotted against the calculated gaze-state distance for each probe condition. The data were fitted with a Gaussian function as follows: where Y P is the estimated initial reach direction for probe P , A is the gain factor, B is the overall offset, and σ d is the standard deviation of the Gaussian. The center of the Gaussian was fixed at 0, and the latter half of the function was used. In this model approach, four free parameters were estimated: the width of the temporal weighting ( σ t ) and three parameters for the generalization curve (A , B , and σ d ), using a nonlinear least-squares fitting. The model fit to the group data is shown over the observed data in Figure 3G (coefficient of determination, R 2 =0.87). Our main interest in this model approach was to estimate the duration of preceding gaze states involved in the encoding of motor memories. Such duration is expressed as σ t , which was estimated at 0.19 s. Figure 3H shows the temporal weighting function with 95% confidential intervals based on bootstrap resampling (N = 2000). Considering a threshold of three times the standard deviation, our results indicate that the temporal history of gaze states until 0.57 s (confidence interval: 0.24–0.69 s) before reaching onset contributed to the learning of reaching movements. The fitting results and 95% confidence intervals for the four parameters were as follows: A =4.5 [3.5-5.6], B =14.6 [13.0-15.9], σ d =24.6 [2.4-32.3], and σ t =0.19 [0.08-0.23]. Discussion Reaching movements are typically preceded by coordinated eye movements. Such coordination has primarily been studied in relation to the roles of retinal and extraretinal signals in target representation and reaching accuracy 25 – 29 . Other studies have focused on the initiation of eye and hand movements (i.e., reaction times) to elucidate the interconnected mechanisms between both motor systems 13 , 30 – 33 . Surprisingly little attention has been paid to how preceding eye movements impact reach-memory encoding, even though reach learning often occurs under eye-hand coordination in daily life. We used two learning paradigms—context-dependent learning and generalization learning—and demonstrated that preceding dynamic gaze states contribute to the formation and updating of internal models for reaching. Several studies using sequential reaching tasks have reported improvement in reach learning when participants were allowed free eye movements compared to when gaze was fixaed 34 , 35 . However, these effects may be attributed to the rich visual information, such as target location and error signals, obtained by eye movements. This view is supported by findings from a hand-tracking adaptation task 36 . Thus, prior studies could not dissociate whether the effect was due to the inherent linkage of gaze states to the reach internal model or to the visual effect on the online reach control. Our previous study showed that different gaze conditions, reaching toward foveal and peripheral targets, can lead to the formation of distinct motor memories for reaching 23 . While this suggests a close link between gaze states and reach learning, gaze remained stationary throughout each trial, leaving open the possibility that learning was driven by differences in visual inputs (i.e., foveal vs. peripheral vision) during and/or at the end of the reaching movement. In contrast to prior studies, we ensured that gaze was directed at the reach target during reach execution, even under different gaze conditions associated with opposing visuomotor rotations (i.e., FOV and SACCADE in Experiment 1, Left-side and Right-side in Experiment 2). Thus, the success of simultaneous learning cannot be explained by the difference in the quality of online visual feedback control. In Experiment 3, learning transfer was limited in probes 4 and 13, even though gaze was on the reach target during movement execution, just as under the learning condition ( Fig. 3E ). These results indicate that the preceding dynamic gaze states, rather than differences in visual feedback, play a significant role in the formation and expression of reaching memories. The contextual encoding of multiple motor memories is a central issue in motor neuroscience 37 , 38 . Arbitrary visual cues unrelated to reaches generally fail to facilitate learning 39 , 40 , as confirmed in Control Experiment 1 and Control Experiment 2 (but, see Refs. 27 and 28). Previous studies have shown that effective contextual cues are typically related to the state of the hand 14 , 39 , 40 , 43 – 45 or hand-associated objects 46 – 48 . Nozaki et al. 14 demonstrated that, in a bimanual reaching task, force-field learning in one arm was impacted by concurrent movements of the other arm but not by ankle flexion, suggesting that contextual states should be somatotopically close to the trained effector. In contrast, we, along with our previous study 23 , show that gaze states—despite being somatotopically distant from the arm—can contribute to reach learning. This challenges the conventional view of somatotopic proximity and instead highlights functional relevance to reach planning as the critical determinant of context-dependent reach learning. Gaze locations relative to the reach target provide essential information for target representation, and dynamic gaze states update such gaze-centered representations in reach planning 28 , 49 – 51 , which might contribute to learning. The importance of functional relevance in contextual learning is also supported by recent findings showing that decision uncertainty can serve as a learning context 52 . The impact of prior movements on motor learning suggests that the adapted motor output at time t depends not only on the corresponding body state at time t ([x_t]) but also on its recent history within a certain time window, expressed as [x_t,x_(t-1),⋯,x_(t-Δt)] 16 – 19 . This indicates that the learning effect is not determined by binary information (i.e., presence or absence of eye movements) but should be sensitive to the time interval between the contextual and adaptation movements. Supporting this view, in Experiment 1 ( Fig.1D ), the contextual learning effect decreased from the Short (∼260-ms interval) to the Middle group (∼540 ms) and disappeared in the Long group (∼1200 ms). In Experiment 3, the generalization pattern was well accounted for by the state distance of eye positions across gaze conditions, computed using a temporally weighting ( Fig. 3G ). This analysis showed that learning incorporated gaze history up to approximately 600 ms prior to reach onset ( Fig. 3H ), a temporal window broadly consistent with the findings of Experiment 1. Several explanations may account for the temporal decay of prior contextual effects. One possibility is that the limited time window reflects a common mechanism across different learning paradigms. In sequential reaching tasks, lead-in movements serve as contextual learning cues for subsequent movements in force fields 18 , 19 . Howard et al. 18 found that this effect decayed progressively as the interval between two sequential movements increased, disappearing by ∼600 ms, a time range consistent with our findings. A more recent study 53 also showed that associative learning between preceding contextual stimuli (sound or light) and reaching movements occur when the stimulus-reach interval is around 300 ms but not beyond 1000 ms. These findings, together with ours, support the hypothesis that prior context impact memory encoding for subsequent reaching within a constrained temporal window, regardless of the modality or effector serving as the contextual cue. One potential mechanism underlying this temporal window is cerebellar processing, as has been frequently investigated in eye-blink conditioning, a well-established cerebellum-dependent learning paradigm that is highly sensitive to the timing between a conditioned stimulus (e.g., tone or light as context) and unconditioned stimulus (e.g., air puff eliciting a blink response). Consistent with previous motor learning studies and our findings, optimal conditioning learning occurs when the conditioned stimulus precedes the unconditioned stimulus by approximately 500 ms (e.g., Ref. 54 ). Alternatively, the temporal window in gaze-dependent learning may stem from eye-movement-specific processes. To separate motor memories by preceding gaze states, the brain needs to retain saccade-related information for a certain period during and after the saccade, ensuring its availability for reach planning and/or execution. According to this view, neurophysiological studies have shown that the dorsolateral prefrontal cortex and lateral intraparietal cortex (LIP), both involved in saccade control, exhibit post-saccadic activity lasting approximately 500 ms after saccade termination 55 – 60 . Within the first 300 ms, this post-saccadic activity carries information about the visual location of the saccade target, saccade direction 55 , 59 , and movement vector in depth 56 . Eye position and saccade direction were also accurately decoded by population activities in the LIP after saccade termination 60 . Thus, post-saccadic activity carrying saccade-related information could provide contextual signals for reaching memories, with its impact decaying as more time elapses after the saccade. Consistent with our findings, a recent behavioral study suggested that preceding saccades are used as learning contexts in a sequential saccades task 61 . Together, these results raise an interesting question for future research: whether the temporal features of contextual learning are mediated by cue-specific processes linked to gaze or by more general mechanisms of motor learning such as cerebellum-dependent processes. Several behavioral studies 16 – 20 , along with the present study, have highlighted the importance of contextual cues during reach planning for reach learning. These findings align with an emerging framework that views motor cortices as a dynamical system 62 . Preparatory neural activities in the premotor and primary motor cortices are tightly linked to reach learning 21 , 22 . Human imaging 63 and brain stimulation 64 studies further supported the critical role of motor cortices in context-dependent learning. According to this framework, our results suggest that preceding static and dynamic gaze states induce different preparatory neural states for reaching, enabling the formation of multiple motor memories. Consistent with this interpretation, gaze-dependent neural signals have been identified along the dorsal visuomotor pathway, from parietal to motor areas. Premotor neural activities vary with static and dynamic gaze conditions in the eye-hand movement task 65 – 71 . Future work is needed to clarify how interplays among the parietal cortex, motor areas, and cerebellum contribute to contextual encoding and the formation of motor memories in context-dependent learning. In daily situations such as playing tennis, individuals are often faced with the need to learn and switch different reaching skills (e.g., forehand and backhand shots) while tracking a fast-moving ball with their gaze. Given the limitations of external cues in contextual learning, it is possible that specific patterns of preceding eye movements are linked to particular internal models of reaching, facilitating efficiency of motor-skill acquisition and expression, especially during the early stages of skill training. Considering the highly variable eye-hand coordination in daily life, gaze-independent learning would of course be more appropriate than a gaze-dependent manner in certain situations. Important questions remain as to how the brain transitions between gaze-dependent and gaze-independent learning and how these mechanisms can be effectively applied in real-life settings, including sports and rehabilitation. Methods Participants A total of 84 healthy right-handed individuals participated in Experiment 1 (including three experimental groups), Experiment 2, Experiment 3, Control Experiment 1 and Control Experiment 2, each with 12 participants. The sample size was chosen on the basis of our previous similar experiments and the current counterbalanced experimental design. All participants were naïve to the purpose of the study. They gave informed consent to participate, which was approved by the NTT Communication Science Laboratories Research Ethics Committee. Apparatus Participants moved a stylus with their right hand on a digitizing tablet (Wacom Intuos 2) placed horizontally on the table ( Figure 1A ). The position of the stylus captured at 100 Hz was presented as a hand cursor on a vertical CRT monitor (Mitsubishi RDF223G; 100 Hz vertical refresh rate) placed approximately 50 cm from participants’ eyes. Participants were asked to move the cursor (black circle with 0.2-cm radius) from a start point (circle with 0.45-cm radius) to a target (white circle with 0.5-cm radius), where the cursor was moved upward or rightward by the forward or rightward stylus motion correspondingly. The hand and stylus were not directly visible. Participants’ heads were stabilized by a chin rest on which a video-based eye tracker (Eyelink 2000) was mounted to record gaze position (500 Hz). According to the trial type, participants were asked to keep fixating on a fixation cross (0.8-cm long line) or to make saccades. Stimulus presentation and experimental procedure were controlled using MATLAB (R2012a, MathWorks) and Psychophysics Toolbox (PTB-3) Experiment 1 Experiment 1 investigated whether two opposing visuomotor rotations can be learned when each was associated with dynamic and static gaze context. A total of 36 participants were assigned to one of three timing groups (12 individuals per group): Short (8 females; age 33.5 ± 10.3, mean ± SD), Middle (6 females; age 30.2 ± 11.2, mean ± SD), and Long (8 females; age 35.9 ± 11.6, mean ± SD). The timing of the “go” cue to initiate the reach varied across groups to manipulate the relative timing of eye and hand movements. Details on this timing group are described later. In each trial, participants performed the reaching task under one of three different gaze conditions ( Figure 1B ). Under the two static conditions (PER and FOV), participants reached toward the target while maintaining gaze fixation throughout the trial. Under the dynamic condition (SACCADE), they made a saccadic eye movement to the target, followed by reaching toward the foveal target. The reach target was always located 10 cm above the start point. For half the participants, the fixation cross under PER and SACCADE was presented on the left side of the target, 10 cm from the start point (i.e., 7.66 cm left and 3.57 cm below the reach target), as shown in Figure 1B . Thus, in this setting, participants reached toward the target in the right visual field under PER and made rightward saccades under SACCADE. To rule out the potential confounding effect of the saccade direction, the location of initial fixation was vertically mirrored (i.e., lower right to the reach target) for the other half of participants. Each trial started with participants placing the cursor at the start point and directing the gaze to the fixation cross ( Figure 1C ). After maintaining fixation for 0.7 s, a reach target appeared. Under the static conditions (i.e., PER and FOV; Figure 1C , left), after a delay according to the timing group, a beeping sound was given for initiating the reach. Thus, participants reached toward the peripheral target under PER and foveal target under FOV and terminated the reach at a second beep presented 0.3 s after reach onset. Under the dynamic condition (i.e., SACCADE; Figure 1C , right), the trial started in the same way as under PER, but the fixation cross jumped to the target 0.6 s after the target presentation. Participants had to make a saccade to the new fixation position (i.e., reach target) as fast as possible and initiate the reach in response to the subsequent beep. In this time course, the reach began after saccade termination; thus, under SACCADE, participants first made a saccade then performed the reach toward the foveal target. To manipulate the interval between saccade termination and reach onset, the start beep for reach was presented 0.7, 1.0, and 1.6 s after the target presentation in the Short, Middle, and Long groups, respectively ( Figure 1C , right). The actual intervals are reported in Results. To match the time course across conditions, the timing of the beep also varied according to the three timing groups under PER and FOV, which did not require saccades ( Figure 1C , left). Throughout the experiment, the three gaze conditions (PER, FOV, and SACCADE) were randomly presented within each block of 16 trials, consisting of 4 PER, 4 FOV, and 8 SACCADE trials. The experiment comprised three sessions. After familiarization, participants completed a pre-learning session (4 blocks/64 trials) with normal visual feedback. In the subsequent learning session (75 blocks/1200 trials), a visuomotor rotation was applied to the cursor movements. The rotation angle increased by 0.8° per block over 25 blocks to reach 20° then was maintained for the remaining 50 blocks. The rotation direction (CW or CCW) varied pseudo-randomly across trials but was uniquely associated with the gaze conditions. For half the participants (n=6), CW was applied for the static gaze conditions (i.e., PER and FOV) and CCW for the dynamic gaze condition (i.e., SACCADE), and this association was reversed for the other half. To ensure an equal number of CW and CCW trials in the simultaneous learning paradigm, the number of trials for each gaze condition within a block (4 PER, 4 FOV, and 8 SACCADE) was set. Finally, participants completed a post-learning session (4 blocks/64 trials) identical to the pre-learning session (i.e., normal feedback). Trials that violated any of the following criteria were specified as errors with a warning beep and repeated at the end of the block: 1) reach movement time was less than 0.15 s or more than 0.7 s, 2) the reach reaction time was over 1 s, 3) saccade reaction time was over 0.4 s, 4) eye movements were detected under PER and FOV, and 5) reach initiated before saccade termination under SACCADE. Control Experiment 1 Control Experiment 1 was conducted to rule out the possibility that the visual cues of fixation jumps, rather than eye movements, served as a contextual effect in Experiment 1. Twelve participants (9 females; age 33.4 ± 9.4, mean ± SD) took part in this experiment. The task was identical to that of the Short group in Experiment 1, except that SACCADE was replaced with FIX-JMUP ( Supplementary Figure 1A ). Thus, there were three gaze conditions: PER, FOV, and FIX-JUMP. Under FIX-JUMP, the trial began as under PER, but 0.6 s after target presentation the fixation cross jumped to the target. Participants had to maintain gaze fixation at the original location of the cross prior to the jump. The direction of visuomotor rotation was associated with static and visual dynamic conditions (i.e.., CW for PER and FOV, CCW for FIX-JUMP), with the assignment counterbalanced across participants. As in Experiment 1, PER, FOV, and FIX-JUMP were randomly presented within each block of 16 trials (4 trials each for PER and FOV and 8 trials for FIX-JUMP). Participants completed a pre-learning session (4 blocks/64 trials), learning session (75 blocks/1200 trials), and post-learning session (4 blocks/64 trials). In the pre- and post-learning sessions, normal visual feedback was presented. During the learning session, a visuomotor rotation was applied: the rotation angle gradually increased by 0.8° per block until it reached 20° then was maintained for the remainder of the session. The spatial configurations of the reach target and fixation cross were the same as in Experiment 1. Under PER and FIX-JUMP, the fixation was placed either left or right of the target, counterbalanced across participants. Experiment 2 Experiment 2 investigated whether two opposing visuomotor rotations were able to be learned when each was associated with a different sequential eye movement. Twelve participants (9 females; age 25.8 ± 8.7, mean ± SD) took part in this experiment. In each trial, participants performed one of two sequential saccade patterns, followed immediately by a reach toward the foveal target ( Figure 2A ). Participants began each trial by placing the cursor at the start point and directing their gaze to a fixation cross located 1 cm above the start point (F0, Figure 2A ). After maintaining fixation for 0.7 s, the reach target appeared 10 cm above the start position. At 0.6 s after target presentation, the fixation jumped to 2 cm upward and 6 cm to the left or 2 cm upward and 6 cm to the right (F1). Participants were asked to make a saccade to this new fixation position as fast as possible. Once this eye movement was detected, the fixation cross jumped again to the reach-target location (F2). Thus, participants executed the first saccade to F1 then second saccade to F2 (i.e., reach target) and maintained fixation on the target until the subsequent reaching movement was completed. Depending on the direction of the first saccade, two saccade conditions were defined: Left-side and Right-side. A beep was presented 1 s after target appearance, signaling participants to initiate the reach. This timing was set so that reaching would begin immediately after the termination of sequential saccades. The actual time between saccade end and reach initiation was 221 ± 66 ms for Left-side saccade and 222 ± 65 ms for Right-side saccade (mean ± SD across participants). In the learning session, the direction of visuomotor rotation was associated with the saccade conditions. For half the participants (n=6), CW was applied to the Left-side saccade, and CCW to Right-side saccade; this assignment was reversed for the other half. In the subsequent learning-and-probe session, two types of probe trials were interleaved with the learning trials to exclude the possibility that the static gaze state at F1, rather than dynamic sequential saccades, has a contextual effect on learning. The first, saccade probe , was identical to the learning trial, in which participants performed either Left-side or Right-side saccades, followed by reaching ( Figure 2A ). In the second, fixation probe , participants maintained gaze fixation at F1 before and during reaching movements. In both probe trials, no cursor feedback was provided. The type of eye movements (Left-side or Right-side saccade) in the saccade probe and the fixation location (left or right of the reach target) in the fixation probe were counterbalanced across participants. The experiment was divided into four sessions. The first pre-learning session included two saccade conditions (Left-side saccade and Right-side saccade; normal cursor feedback) and two probe conditions (saccade probe and fixation probe; no cursor feedback). These four conditions were presented randomly within each block of 22 trials, consisting of 16 trials of the saccade conditions (8 trials for each) and 6 trials of the probe conditions (3 trials for each). The block was repeated four times (4 blocks/88 trials). In the subsequent learning session, two saccade conditions were presented randomly within each block of 16 trials (8 trials for Left-side saccade and 8 trials for Right-side saccade). The block was repeated 45 times (45 blocks/720 trials). As in Experiment 1, the angle of visuomotor rotation increased gradually by 0.8° per block up to 20° over the first 20 blocks and maintained 20° for the remaining 25 blocks. As mentioned above, the direction of rotation (CW or CCW) was uniquely associated with the saccade conditions. In the following learning-and-probe session, the block structure was identical to the pre-learning session: each block consisted of 22 trials, with 16 trials for the two saccade conditions (±20° rotation) and 6 trials for the two probe conditions (no-cursor feedback). The block was repeated three times (3 blocks/66 trials). The last post-learning session included only two saccade conditions with normal cursor feedback. Each block consisted of 16 trials (8 trials for each saccade condition) and was repeated 3 times (3 blocks/48 trials). Control Experiment 2 Control Experiment 2 (twelve participants; 5 females; age 26.3 ± 6.1, mean ± SD) was conducted to rule out the possibility that visual cues from fixation jumps, rather than eye movements, served as contextual signals in Experiment 2. As in Experiment 2, each trial began when participants placed the cursor at the start point and fixated on the cross (F0; Supplementary Figure 2A ). After 0.6 s of target presentation, the fixation cross jumped to F1 and then to the reach target. Two conditions were defined based on the direction of the first jump: leftward and rightward displacement. The direction and amplitude of fixation jumps were identical to those in Experiment 2. Regardless of the direction of the fixation cross jump, participants were instructed to make a single saccade to the reach target as soon as the fixation cross appeared on it, followed immediately by a reaching movement. The actual time between saccade end and reach initiation was 154 ± 70 ms for Left-side and 152 ± 69 ms for Right-side conditions (mean ± SD across participants). The direction of visuomotor rotation was associated with the displacement condition (CW for leftward, CCW for rightward; shown in Suppl. Fig. 2A ), and this assignment was counterbalanced across participants. If visual cues from fixation jumps are sufficient for contextual learning, participants would exhibit simultaneous learning under these conditions. Two conditions were randomly presented within each block of 16 trials (8 trials per condition). Participants completed a pre-learning session (4 blocks/64 trials), learning session (45 blocks/720 trials), and post-learning session (3 blocks/48 trials). In the pre- and post-learning sessions, normal visual feedback was provided. During the learning session, the rotation angle gradually increased by 1° per block until it reached 20°, after which it was maintained for the remainder of the session. A gap duration of 0.18 s was inserted between the first and second fixation cross jumps. This interval was set to match the median interval between the first and second fixation jumps in Experiment 2, where the second jump was triggered by detection of the first saccade. Experiment 3 Experiment 3 investigated the degree of learning transfer from a single training gaze state to other novel gaze states. Twelve participants (7 females; age 24.7 ± 8.7, mean ± SD) reached toward a target under 14 different gaze states ( Figure 3B ). Under the five static gaze conditions (blue 10–14 in Figure 3B ), participants maintained the fixation throughout the trial, which was located at -15, -10, -5, 0, or 10 cm to the reach target along the horizontal direction. These fixation positions are indicated with blue dots in Figure 3B , with values in brackets. Under the nine dynamic gaze conditions (pink 1–9 in Figure 3B ), participants made a horizontal saccade, followed immediately by reaching toward the target. In Figure 3B , the fixation positions before and after saccade are indicated with pink dots with their values. The experiment began with a baseline session in which participants reached toward a single target located 10 cm above and 5 cm to the right of the start point, under 14 gaze conditions, as described above. Each block consisted of these 14 different gaze conditions, performed with and without cursor feedback, for a total of 28 trials in random order. The block was repeated 12 times (12 blocks/336 trials). In the feedback trials, veridical online cursor feedback was provided. In the subsequent learning session, participants adapted to a single visuomotor rotation (CW or CCW, counterbalanced across participants), as shown in Figure 3A . In all trials, participants performed a rightward saccade of 10 cm toward the reach target, followed immediately by a reach toward the foveal target (learning condition in Figure 3B ). Each block consisted of five trials. In most trials (3/5), the reach target was placed at the same position as in the baseline session, with the initial fixation point located 10 cm to the left of the reach target (learning condition in Figure 3B ). To prevent participants from making habitual movements, in the remaining trials (2/5), both the reach target and initial fixation point were shifted 2.5 cm to the left or right of the original learning condition (i.e., saccade amplitude was maintained at 10 cm). These five trials were randomly presented within each block, and the block was repeated 30 times (30 blocks/150 trials). The angle of visuomotor rotation gradually increased by 0.5° per trial until reaching 20° then was maintained at 20° for the remainder of the learning session. The last probe session consisted of a learning trial and probe trial (168 trials each; for a total of 336 trials), always in succession ( Figure 3A ). The learning trial was identical to the original learning condition in the learning session (i.e., 20° rotation; learning condition in Figure 3B ) and was inserted before every probe trial to prevent rapid decay of adaptation. In the probe trials, participants performed reaching movements to the same target position used under the learning condition, under 14 gaze conditions ( Figure 3B ) without cursor feedback. These probe trials were identical to the non-cursor feedback trials in the baseline session. Data analysis The initiation and termination of reaching movements were detected online during the experiment and defined as the first time point at which tangential velocity exceeded or fell below 2 cm/s for at least 30 ms. For offline analysis, the obtained position data were aligned at the reach onset and low-pass filtered at 20 Hz (4th-order Butterworth). Velocity profiles were also obtained by three numerical time differentiations. To assess learning, the initial direction of reaching movements (initial reach direction) was quantified on each trial as the angular difference between the movement trajectory and a line connecting the start point to the target. The initial reach direction was computed as the mean direction of velocity vectors during the 100–120-ms window after reach onset. To collapse data across different rotation directions (CW and CCW), the sign of the reach direction was adjusted so that positive values indicated compensatory movements for the corresponding rotation. To remove movement biases across gaze conditions, we first averaged the initial reach direction under each condition in the pre-learning session. In Experiment 3, such biases were estimated across different gaze and visual conditions (cursor visible or invisible). These baseline biases were then subtracted from the corresponding trials throughout the experiment. This process was conducted for each participant. The sign-adjusted reach direction, normalized by baseline, was used in subsequent analyses, including plotting learning curves and quantifying final learning levels, aftereffect magnitudes, and outputs of probe trials. In Experiment 1 and Control Experiment 1, the final learning level was defined as the mean reach direction in the last four blocks of the learning session, and the aftereffect magnitude as the mean reach direction in the four blocks of the post-learning session. In Experiment 2 and Control Experiment 2, these variables were estimated from the last two blocks of the learning session and from the first two blocks of the post-learning session. This difference in the number of blocks used for evaluation reflects differences in the number of trials per condition within each block across experiments, thereby ensuring a roughly comparable number of trials for evaluation. ACKNOWLEDGMENTS We thank T. Isobe for the data collection and T. Kawanishi, A. Kimura, and F. Naya for their support and encouragement. Footnotes We clarified the design, hypotheses, and predicted outcomes of Experiment 1 and incorporated these into the manuscript. We also revised the analytical approach for Experiment 1, particularly the statistical methods, leading to updates in the Results section and Figure 1. In addition, we included a new control experiment (Control Experiment 2) related to Experiment 2. References 1. ↵ Kawato , M . ( 1999 ). Internal models for motor control and trajectory planning . Curr. Opin. Neurobiol . 9 , 718 – 727 . OpenUrl CrossRef PubMed Web of Science 2. ↵ Krakauer , J.W. , Hadjiosif , A.M. , Xu , J. , Wong , A.L. , and Haith , A.M . ( 2019 ). Motor Learning . Comprehensive Physiology , 9 ( 2 ), 613 – 663 . OpenUrl CrossRef PubMed 3. ↵ Shadmehr , R . ( 2004 ). Generalization as a behavioral window to the neural mechanisms of learning internal models . Hum Mov Sci 23 , 543 – 568 . doi: 10.1016/j.humov.2004.04.003 . OpenUrl CrossRef PubMed Web of Science 4. ↵ de Brouwer , A.J. , Flanagan , J.R. , and Spering , M. ( 2021 ). Functional use of eye movements for an acting system . Trends in Cognitive Sciences . 5. ↵ Land , M.F . ( 2009 ). Vision, eye movements, and natural behavior . Vis. Neurosci 26 , 51 – 62 . doi: 10.1017/S0952523808080899 . OpenUrl CrossRef PubMed Web of Science 6. ↵ Davids , K. , Williams , A.M. , and Williams , J.G. ( 2005 ). Visual perception and action in sport (Routledge) . 7. ↵ Andersen , R.A. , and Cui , H . ( 2009 ). Intention, action planning, and decision making in parietal-frontal circuits . Neuron 63 , 568 – 583 . OpenUrl CrossRef PubMed Web of Science 8. Crawford , J.D. , Medendorp , W.P. , and Marotta , J.J . ( 2004 ). Spatial transformations for eye-hand coordination . J. Neurophysiol . 92 , 10 – 19 . doi: 10.1152/jn.00117.2004 . OpenUrl CrossRef PubMed Web of Science 9. ↵ Crawford , J.D. , Henriques , D.Y.P. , and Medendorp , W.P . ( 2011 ). Three-dimensional transformations for goal-directed action . Annu. Rev. Neurosci . 34 , 309 – 331 . doi: 10.1146/annurev-neuro-061010-113749 . OpenUrl CrossRef PubMed Web of Science 10. ↵ Diedrichsen , J. , Shadmehr , R. , and Ivry , R.B . ( 2010 ). The coordination of movement: optimal feedback control and beyond . Trends Cogn. Sci. (Regul. Ed .) 14 , 31 – 39 . doi: 10.1016/j.tics.2009.11.004 . OpenUrl CrossRef PubMed Web of Science 11. ↵ Ariff , G. , Donchin , O. , Nanayakkara , T. , and Shadmehr , R . ( 2002 ). A real-time state predictor in motor control: study of saccadic eye movements during unseen reaching movements . J. Neurosci . 22 , 7721 – 7729 . OpenUrl Abstract / FREE Full Text 12. Miall , R.C. , Reckess , G.Z. , and Imamizu , H . ( 2001 ). The cerebellum coordinates eye and hand tracking movements . Nature neuroscience 4 , 638 – 644 . OpenUrl CrossRef PubMed Web of Science 13. ↵ Yttri , E.A. , Liu , Y. , and Snyder , L.H . ( 2013 ). Lesions of cortical area LIP affect reach onset only when the reach is accompanied by a saccade, revealing an active eye-hand coordination circuit . Proc. Natl. Acad. Sci. U.S.A . 110 , 2371 – 2376 . doi: 10.1073/pnas.1220508110 . OpenUrl Abstract / FREE Full Text 14. ↵ Nozaki , D. , Kurtzer , I. , and Scott , S.H . ( 2006 ). Limited transfer of learning between unimanual and bimanual skills within the same limb . Nature neuroscience 9 , 1364 – 1366 . OpenUrl CrossRef PubMed Web of Science 15. Yokoi , A. , Hirashima , M. , and Nozaki , D . ( 2011 ). Gain field encoding of the kinematics of both arms in the internal model enables flexible bimanual action . Journal of Neuroscience 31 , 17058 – 17068 . OpenUrl Abstract / FREE Full Text 16. ↵ Howard , I.S. , Franklin , S. , and Franklin , D.W . ( 2024 ). Kernels of Motor Memory Formation: Temporal Generalization in Bimanual Adaptation . J. Neurosci . 44 . doi: 10.1523/JNEUROSCI.0359-24.2024 . OpenUrl Abstract / FREE Full Text 17. ↵ Gippert , M. , Leupold , S. , Heed , T. , Howard , I.S. , Villringer , A. , Nikulin , V.V. , and Sehm , B . ( 2023 ). Prior Movement of One Arm Facilitates Motor Adaptation in the Other . J. Neurosci . 43 , 4341 – 4351 . doi: 10.1523/JNEUROSCI.2166-22.2023 . OpenUrl Abstract / FREE Full Text 18. ↵ Howard , I.S. , Ingram , J.N. , Franklin , D.W. , and Wolpert , D.M . ( 2012 ). Gone in 0.6 seconds: the encoding of motor memories depends on recent sensorimotor states . J. Neurosci . 32 , 12756 – 12768 . doi: 10.1523/JNEUROSCI.5909-11.2012 . OpenUrl Abstract / FREE Full Text 19. ↵ Wainscott , S.K. , Donchin , O. , and Shadmehr , R . ( 2005 ). Internal Models and Contextual Cues: Encoding Serial Order and Direction of Movement . Journal of Neurophysiology 93 , 786 – 800 . doi: 10.1152/jn.00240.2004 . OpenUrl CrossRef PubMed Web of Science 20. ↵ Sheahan , H.R. , Franklin , D.W. , and Wolpert , D.M . ( 2016 ). Motor Planning, Not Execution, Separates Motor Memories . Neuron 92 , 773 – 779 . doi: 10.1016/j.neuron.2016.10.017 . OpenUrl CrossRef PubMed 21. ↵ Sun , X. , O’Shea , D.J. , Golub , M.D. , Trautmann , E.M. , Vyas , S. , Ryu , S.I. , and Shenoy , K.V . ( 2022 ). Cortical preparatory activity indexes learned motor memories . Nature , 1 – 6 . 22. ↵ Vyas , S. , O’Shea , D.J. , Ryu , S.I. , and Shenoy , K.V . ( 2020 ). Causal Role of Motor Preparation during Error-Driven Learning . Neuron 106 , 329 – 339 .e4. doi: 10.1016/j.neuron.2020.01.019 . OpenUrl CrossRef PubMed 23. ↵ Abekawa , N. , Ito , S. , and Gomi , H . ( 2022 ). Gaze-specific motor memories for hand-reaching . Current Biology 32 , 2747 – 2753 .e6. doi: 10.1016/j.cub.2022.04.065 . OpenUrl CrossRef PubMed 24. ↵ Abekawa , N. , and Gomi , H . ( 2024 ). A Computational Mechanism for Forming Arm Motor Memories Differentiated by Dynamic Gaze States . In 2024 IEEE International Conference on Systems, Man, and Cybernetics (SMC) ( IEEE ), pp. 2305 – 2310 . 25. ↵ Prablanc , C. , Echallier , J.F. , Komilis , E. , and Jeannerod , M . ( 1979 ). Optimal response of eye and hand motor systems in pointing at a visual target. I. Spatio-temporal characteristics of eye and hand movements and their relationships when varying the amount of visual information . Biol Cybern 35 , 113 – 124 . OpenUrl CrossRef PubMed Web of Science 26. Soechting , J.F. , Engel , K.C. , and Flanders , M . ( 2001 ). The Duncker illusion and eye-hand coordination . J. Neurophysiol . 85 , 843 – 854 . OpenUrl PubMed Web of Science 27. Binsted , G. , and Elliott , D . ( 1999 ). Ocular perturbations and retinal/extraretinal information: the coordination of saccadic and manual movements . Exp Brain Res 127 , 193 – 206 . OpenUrl CrossRef PubMed Web of Science 28. ↵ Henriques , D.Y. , Klier , E.M. , Smith , M.A. , Lowy , D. , and Crawford , J.D . ( 1998 ). Gaze-centered remapping of remembered visual space in an open-loop pointing task . J. Neurosci . 18 , 1583 – 1594 . OpenUrl Abstract / FREE Full Text 29. ↵ Buneo , C.A. , Jarvis , M.R. , Batista , A.P. , and Andersen , R.A . ( 2002 ). Direct visuomotor transformations for reaching . Nature 416 , 632 – 636 . doi: 10.1038/416632a . OpenUrl CrossRef PubMed Web of Science 30. ↵ Dean , H.L. , Martí , D. , Tsui , E. , Rinzel , J. , and Pesaran , B . ( 2011 ). Reaction time correlations during eye-hand coordination: behavior and modeling . J. Neurosci . 31 , 2399 – 2412 . doi: 10.1523/JNEUROSCI.4591-10.2011 . OpenUrl Abstract / FREE Full Text 31. Fisk , J.D. , and Goodale , M.A . ( 1985 ). The organization of eye and limb movements during unrestricted reaching to targets in contralateral and ipsilateral visual space . Exp Brain Res 60 , 159 – 178 . OpenUrl CrossRef PubMed Web of Science 32. Sailer , U. , Eggert , T. , Ditterich , J. , and Straube , A . ( 2000 ). Spatial and temporal aspects of eye-hand coordination across different tasks . Exp Brain Res 134 , 163 – 173 . OpenUrl CrossRef PubMed Web of Science 33. ↵ Armstrong , I.T. , Judson , M. , Munoz , D.P. , Johansson , R.S. , and Flanagan , J.R . ( 2013 ). Waiting for a hand: saccadic reaction time increases in proportion to hand reaction time when reaching under a visuomotor reversal . Front Hum Neurosci 7 , 319 . doi: 10.3389/fnhum.2013.00319 . OpenUrl CrossRef PubMed 34. ↵ Massing , M. , Blandin , Y. , and Panzer , S . ( 2016 ). Magnifying visual target information and the role of eye movements in motor sequence learning . Acta Psychologica 163 , 59 – 64 . doi: 10.1016/j.actpsy.2015.11.004 . OpenUrl CrossRef PubMed 35. ↵ Vieluf , S. , Massing , M. , Blandin , Y. , Leinen , P. , and Panzer , S . ( 2015 ). The role of eye movements in motor sequence learning . Human Movement Science 40 , 220 – 236 . doi: 10.1016/j.humov.2015.01.004 . OpenUrl CrossRef PubMed 36. ↵ Gouirand , N. , Mathew , J. , Brenner , E. , and Danion , F.R . ( 2019 ). Eye movements do not play an important role in the adaptation of hand tracking to a visuomotor rotation . Journal of neurophysiology 121 , 1967 – 1976 . OpenUrl CrossRef PubMed 37. ↵ Heald , J.B. , Lengyel , M. , and Wolpert , D.M . ( 2021 ). Contextual inference underlies the learning of sensorimotor repertoires . Nature 600 , 489 – 493 . doi: 10.1038/s41586-021-04129-3 . OpenUrl CrossRef PubMed 38. ↵ Heald , J.B. , Lengyel , M. , and Wolpert , D.M . ( 2023 ). Contextual inference in learning and memory . Trends in Cognitive Sciences 27 , 43 – 64 . doi: 10.1016/j.tics.2022.10.004 . OpenUrl CrossRef PubMed 39. ↵ Gandolfo , F. , Mussa-Ivaldi , F.A. , and Bizzi , E . ( 1996 ). Motor learning by field approximation . Proceedings of the National Academy of Sciences 93 , 3843 – 3846 . OpenUrl Abstract / FREE Full Text 40. ↵ Howard , I.S. , Wolpert , D.M. , and Franklin , D.W . ( 2013 ). The effect of contextual cues on the encoding of motor memories . J. Neurophysiol . 109 , 2632 – 2644 . doi: 10.1152/jn.00773.2012 . OpenUrl CrossRef PubMed Web of Science 41. Addou , T. , Krouchev , N. , and Kalaska , J.F . ( 2011 ). Colored context cues can facilitate the ability to learn and to switch between multiple dynamical force fields . Journal of Neurophysiology 106 , 163 – 183 . doi: 10.1152/jn.00869.2010 . OpenUrl CrossRef PubMed Web of Science 42. Osu , R. , Hirai , S. , Yoshioka , T. , and Kawato , M . ( 2004 ). Random presentation enables subjects to adapt to two opposing forces on the hand . Nature neuroscience 7 , 111 – 112 . OpenUrl CrossRef PubMed Web of Science 43. ↵ Ghahramani , Z. , and Wolpert , D.M . ( 1997 ). Modular decomposition in visuomotor learning . Nature 386 , 392 – 395 . OpenUrl CrossRef PubMed Web of Science 44. Howard , I.S. , Wolpert , D.M. , and Franklin , D.W . ( 2015 ). The value of the follow-through derives from motor learning depending on future actions . Curr. Biol . 25 , 397 – 401 . doi: 10.1016/j.cub.2014.12.037 . OpenUrl CrossRef PubMed 45. ↵ Hirashima , M. , and Nozaki , D . ( 2012 ). Distinct motor plans form and retrieve distinct motor memories for physically identical movements . Current Biology 22 , 432 – 436 . OpenUrl CrossRef PubMed 46. ↵ Heald , J.B. , Ingram , J.N. , Flanagan , J.R. , and Wolpert , D.M . ( 2018 ). Multiple motor memories are learned to control different points on a tool . Nat Hum Behav 2 , 300 – 311 . doi: 10.1038/s41562-018-0324-5 . OpenUrl CrossRef PubMed 47. Hwang , E.J. , Donchin , O. , Smith , M.A. , and Shadmehr , R . ( 2003 ). A gain-field encoding of limb position and velocity in the internal model of arm dynamics . PLoS Biol . 1 , E25 . doi: 10.1371/journal.pbio.0000025 . OpenUrl CrossRef PubMed 48. ↵ Proud , K. , Heald , J.B. , Ingram , J.N. , Gallivan , J.P. , Wolpert , D.M. , and Flanagan , J.R . ( 2019 ). Separate motor memories are formed when controlling different implicitly specified locations on a tool . Journal of neurophysiology 121 , 1342 – 1351 . OpenUrl CrossRef PubMed 49. ↵ Batista , A.P. , Buneo , C.A. , Snyder , L.H. , and Andersen , R.A . ( 1999 ). Reach plans in eye-centered coordinates . Science 285 , 257 – 260 . OpenUrl Abstract / FREE Full Text 50. Medendorp , W.P. , Goltz , H.C. , Vilis , T. , and Crawford , J.D . ( 2003 ). Gaze-centered updating of visual space in human parietal cortex . J. Neurosci . 23 , 6209 – 6214 . OpenUrl Abstract / FREE Full Text 51. ↵ Khan , A.Z. , Pisella , L. , Vighetto , A. , Cotton , F. , Luauté , J. , Boisson , D. , Salemme , R. , Crawford , J.D. , and Rossetti , Y . ( 2005 ). Optic ataxia errors depend on remapped, not viewed, target location . Nat. Neurosci . 8 , 418 – 420 . doi: 10.1038/nn1425 . OpenUrl CrossRef PubMed Web of Science 52. ↵ Ogasa , K. , Yokoi , A. , Okazawa , G. , Nishigaki , M. , Hirashima , M. , and Hagura , N . ( 2024 ). Decision uncertainty as a context for motor memory . Nature Human Behaviour 8 , 1738 – 1751 . OpenUrl PubMed 53. ↵ Avraham , G. , Taylor , J.A. , Breska , A. , Ivry , R.B. , and McDougle , S.D . ( 2022 ). Contextual effects in sensorimotor adaptation adhere to associative learning rules . eLife 11 , e75801 . doi: 10.7554/eLife.75801 . OpenUrl CrossRef PubMed 54. ↵ Kjell , K. , Löwgren , K. , and Rasmussen , A . ( 2018 ). A Longer Interstimulus Interval Yields Better Learning in Adults and Young Adolescents . Front. Behav. Neurosci . 12 . doi: 10.3389/fnbeh.2018.00299 . OpenUrl CrossRef PubMed 55. ↵ Funahashi , S. , Bruce , C.J. , and Goldman-Rakic , P.S . ( 1991 ). Neuronal activity related to saccadic eye movements in the monkey’s dorsolateral prefrontal cortex . Journal of Neurophysiology 65 , 1464 – 1483 . doi: 10.1152/jn.1991.65.6.1464 . OpenUrl CrossRef PubMed Web of Science 56. ↵ Genovesio , A. , Brunamonti , E. , Giusti , M.A. , and Ferraina , S . ( 2007 ). Postsaccadic Activities in the Posterior Parietal Cortex of Primates Are Influenced by both Eye Movement Vectors and Eye Position . J. Neurosci . 27 , 3268 – 3273 . doi: 10.1523/JNEUROSCI.5415-06.2007 . OpenUrl Abstract / FREE Full Text 57. Funahashi , S . ( 2014 ). Saccade-related activity in the prefrontal cortex: its role in eye movement control and cognitive functions . Front. Integr. Neurosci . 8 . doi: 10.3389/fnint.2014.00054 . OpenUrl CrossRef PubMed 58. Goldman-Rakic , P.S. , Funahashi , S. , and Bruce , C.J . ( 1990 ). Neocortical memory circuits . Cold Spring Harb Symp Quant Biol 55 , 1025 – 1038 . doi: 10.1101/sqb.1990.055.01.097 . OpenUrl Abstract / FREE Full Text 59. ↵ Takeda , K. , and Funahashi , S . ( 2002 ). Prefrontal Task-Related Activity Representing Visual Cue Location or Saccade Direction in Spatial Working Memory Tasks . Journal of Neurophysiology 87 , 567 – 588 . doi: 10.1152/jn.00249.2001 . OpenUrl CrossRef PubMed Web of Science 60. ↵ Graf , A.B. , and Andersen , R.A . ( 2014 ). Inferring eye position from populations of lateral intraparietal neurons . eLife 3 , e02813 . doi: 10.7554/eLife.02813 . OpenUrl CrossRef PubMed 61. ↵ Azadi , R. , and McPeek , R.M . ( 2022 ). Contextual saccade adaptation induced by sequential saccades . J Neurophysiol 127 , 746 – 755 . doi: 10.1152/jn.00221.2021 . OpenUrl CrossRef PubMed 62. ↵ Shenoy , K.V. , Sahani , M. , and Churchland , M.M . ( 2013 ). Cortical control of arm movements: a dynamical systems perspective . Annual review of neuroscience 36 , 337 – 359 . OpenUrl CrossRef PubMed Web of Science 63. ↵ Ogawa , K. , and Imamizu , H . ( 2013 ). Human sensorimotor cortex represents conflicting visuomotor mappings . Journal of Neuroscience 33 , 6412 – 6422 . OpenUrl Abstract / FREE Full Text 64. ↵ Nozaki , D. , Yokoi , A. , Kimura , T. , Hirashima , M. , and de Xivry , J.-J.O. ( 2016 ). Tagging motor memories with transcranial direct current stimulation allows later artificially-controlled retrieval . Elife 5 , e15378 . OpenUrl CrossRef PubMed 65. ↵ Fujii , N. , Mushiake , H. , and Tanji , J . ( 2000 ). Rostrocaudal Distinction of the Dorsal Premotor Area Based on Oculomotor Involvement . Journal of Neurophysiology 83 , 1764 – 1769 . doi: 10.1152/jn.2000.83.3.1764 . OpenUrl CrossRef PubMed Web of Science 66. Jouffrais , C. , and Boussaoud , D . ( 1999 ). Neuronal activity related to eye-hand coordination in the primate premotor cortex . Exp Brain Res 128 , 205 – 209 . doi: 10.1007/s002210050837 . OpenUrl CrossRef PubMed Web of Science 67. Pesaran , B. , Nelson , M.J. , and Andersen , R.A . ( 2010 ). A relative position code for saccades in dorsal premotor cortex . J. Neurosci 30 , 6527 – 6537 . doi: 10.1523/JNEUROSCI.1625-09.2010 . OpenUrl Abstract / FREE Full Text 68. Boussaoud , D. , Jouffrais , C. , and Bremmer , F . ( 1998 ). Eye position effects on the neuronal activity of dorsal premotor cortex in the macaque monkey . Journal of neurophysiology 80 , 1132 – 1150 . OpenUrl CrossRef PubMed Web of Science 69. Pesaran , B. , Nelson , M.J. , and Andersen , R.A . ( 2006 ). Dorsal premotor neurons encode the relative position of the hand, eye, and goal during reach planning . Neuron 51 , 125 – 134 . OpenUrl CrossRef PubMed Web of Science 70. Batista , A.P. , Santhanam , G. , Yu , B.M. , Ryu , S.I. , Afshar , A. , and Shenoy , K.V . ( 2007 ). Reference frames for reach planning in macaque dorsal premotor cortex . Journal of neurophysiology 98 , 966 – 983 . OpenUrl CrossRef PubMed Web of Science 71. ↵ Cisek , P. , and Kalaska , J.F . ( 2002 ). Modest gaze-related discharge modulation in monkey dorsal premotor cortex during a reaching task performed with free fixation . Journal of neurophysiology 88 , 1064 – 1072 . OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted May 04, 2026. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Eye-hand coordination beyond vision: preceding gaze history shapes reaching memories Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Eye-hand coordination beyond vision: preceding gaze history shapes reaching memories Naotoshi Abekawa , Hiroaki Gomi bioRxiv 2025.11.11.687771; doi: https://doi.org/10.1101/2025.11.11.687771 Share This Article: Copy Citation Tools Eye-hand coordination beyond vision: preceding gaze history shapes reaching memories Naotoshi Abekawa , Hiroaki Gomi bioRxiv 2025.11.11.687771; doi: https://doi.org/10.1101/2025.11.11.687771 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neuroscience Subject Areas All Articles Animal Behavior and Cognition (7635) Biochemistry (17697) Bioengineering (13894) Bioinformatics (41951) Biophysics (21455) Cancer Biology (18592) Cell Biology (25507) Clinical Trials (138) Developmental Biology (13380) Ecology (19903) Epidemiology (2067) Evolutionary Biology (24321) Genetics (15610) Genomics (22509) Immunology (17737) Microbiology (40398) Molecular Biology (17182) Neuroscience (88618) Paleontology (667) Pathology (2833) Pharmacology and Toxicology (4825) Physiology (7641) Plant Biology (15158) Scientific Communication and Education (2046) Synthetic Biology (4296) Systems Biology (9825) Zoology (2271)