When a neural movement controller, called an “internal model,” is adapted to a novel environment, the movement error needs to be appropriately associated with the controller. However, their association is not necessarily guaranteed for bimanual movements in which two controllers—one for each hand—result in two movement errors. Considering the implicit nature of the adaptation process, the movement error of one hand can be erroneously associated with the controller of the other hand. Here, we investigated this credit-assignment problem in bimanual movement by having participants perform bimanual, symmetric back-and-forth movements while displaying the position of the right hand only with a cursor. In the training session, the cursor position was gradually rotated clockwise, such that the participants were unaware of the rotation. The movement of the right hand gradually rotated counterclockwise as a consequence of adaptation. Although the participants knew that the cursor reflected the movement of the right hand, such gradual adaptation was also observed for the invisible left hand, especially when the cursor was presented on the left side of the display. Thus the movement error of the right hand was implicitly assigned to the left-hand controller. Such cross talk in credit assignment might influence motor adaptation performance, even when two cursors are presented; the adaptation was impaired when the rotations imposed on the cursors were opposite compared with when they were in the same direction. These results indicate the inherent presence of cross talk in the process of associating action with consequence in bimanual movement.
- visual rotation
- credit assignment
- internal model
to perform a desired movement under a wide variety of environments, the brain constructs a neural feedforward controller, known as an “internal model” (Bastian 2006; Kawato 1999; Wolpert et al. 1998). In this learning process, information regarding movement errors needs to be appropriately assigned to the movement controller (in the present study, the “movement controller” refers to both the inverse model and the feedback controller comprised of the forward model). In contrast to unimanual movement, in which there is one-to-one correspondence between the movement controller and the resultant movement error, the problem of “credit assignment” (Berniker and Kording 2008; Chen-Harris et al. 2008; Kluzik et al. 2008) is not trivial for bimanual movement (Diedrichsen et al. 2004; Franz et al. 2001; Peters 1985), in which two movement controllers for both hands can generate two distinct movement errors (White and Diedrichsen 2010). Consider a task in which two cursors are controlled by moving handles bimanually. Intuitively, if the subject explicitly knows which cursor reflects which hand's movement, it is likely that the movement of each cursor will be appropriately assigned to the actual movement controller of the hand. However, considering a previous finding that implicit visuomotor learning can override even explicit strategy (Mazzoni and Krakauer 2006), such appropriate assignment is not necessarily guaranteed. Furthermore, the existence of neuronal interference or cross talk in movement planning and/or execution (Carson 2005; Heuer 1993; Marteniuk et al. 1984; Mechsner et al. 2001; Sherwood 1994; Swinnen and Wenderoth 2004) suggests that such cross talk also exists in the process of credit assignment.
A recent study reports that when visual rotation is applied to a cursor displayed at the middle position between both hands, both hands exhibit adaptation at the subsequent trial, indicating that the controllers of both hands refer to only one error information (White and Diedrichsen 2010). However, in their study, because of the redundant relationship between cursor movement and hand movements, the participants' natural reaction has been to move both hands to compensate for the cursor movement. Furthermore, since the participants were allowed to correct their movements during reaching, it is possible that their natural response to online correction worked as an explicit cue that could be used to teach the controllers how they should adapt to the error.
Therefore, we investigated the credit-assignment problem during bimanual movement more directly by gradually increasing the visual rotation such that the participants were unaware of it (Kagerer et al. 1997; Michel et al. 2007; Saijo and Gomi 2010). The participants performed a symmetric, bimanual movement, and apart from the study displaying a cursor at the middle position of both hands (White and Diedrichsen 2010), a single cursor moving with the right hand was displayed. We found that although the participants explicitly knew that the cursor reflected the movement of their right hand, the implicit adaptation of their movement to the visual rotation occurred not only for the right hand but also for the left hand, especially when the cursor was presented just above the left hand.
Even when two cursors were presented for both of the hands, the presence of such inherent cross talk in the credit assignment could substantially influence the performance of the adaptation to visual rotation during bimanual movement. Specifically, we predicted that the adaptation was impaired when the rotational directions imposed on both cursors were opposite compared with when they were the same, because in the former case, the movement correction of one hand to the relevant cursor should be interfered with by the opposite directional error of the irrelevant cursor associated with the other hand. Our additional experiment confirmed this prediction, which further supported the presence of inherent cross talk in the process by which movement errors are associated with movement controllers during bimanual movement.
MATERIALS AND METHODS
A total of 26 subjects participated in this study and were randomly assigned to one of two experimental groups. Each subject participated in only one experiment. Ten participants (nine men and one woman; aged 19–31 years) participated in experiment 1; 16 participants (14 men and two women; aged 21–28 years) participated in experiment 2. All participants had no cognitive or motor disorders and were naïve to the purpose of the experiments. All experiments were approved by the Ethical Committee of the Graduate School of Education, The University of Tokyo. All subjects provided written, informed consent before participating. The handedness of each participant was tested by the Edinburgh Inventory (Oldfield 1971), and all were judged as right-handed (laterality quotient = 86.7 ± 3.8).
The experiments were performed in a dark room. Participants sat in front of a white horizontal screen at chest height, where a computer screen (51 × 40 cm) was displayed using a projector (Fig. 1). Chair height was adjusted such that the elbows and wrists did not touch the working space or the upper horizontal plane. Participants were instructed to keep their head straight and face forward and to avoid twisting their head and body. In each hand, they held a handle of a custom-made manipulandum that could be moved in a horizontal plane under the screen. The participants could not directly see the handles and their hands. The vertical handle was attached to the end of the two linked bars (length of each bar = 47.5 cm) whose angular displacements were measured by potentiometers (Fig. 1A). The analog outputs of the potentiometers were sampled at 1,000 Hz (PXI-1042, National Instruments, Austin, TX) to display the positions of the handles by the cursors (blue and green for the left and right handles, respectively; diameter, 0.5 cm) on the screen in real time by a program written in LabVIEW 7.1 (National Instruments). The cursors were presented on the screen above the actual or the shifted handle position. Another computer was used for analog-to-digital conversion (200 Hz) and stored the data for later analysis using the MATLAB data acquisition toolbox (Mathworks, Natick, MA).
The experimental system also displayed starting position(s) and target(s) using red circle(s) (diameter, 0.9 cm). A target was presented for 2,000 ms every 5,000 ± 500 ms directly ahead of the starting position(s), whereas the starting position(s) were always visible. The distance from a starting position to a target was 7 cm. The starting positions of the left and right hands were separated by 32 cm.
The participants were instructed to move the cursor between the starting position and the target by performing straight, fast, and uncorrected out-and-back movements with a sharp reversal at the target (peak velocity was ∼250 mm/s). Before each trial, they needed to place the cursor at the starting position and wait for the appearance of a target. Prior to all experiments, the participants performed a sufficient amount of bimanual movement practice (150 trials) under normal visuomotor conditions.
Experiment 1 was designed to examine how visual errors are assigned to the left- and right-hand movement controllers during bimanual movements when the error information of one hand was only provided visually. In this experiment, the visual information of the right hand only (i.e., the starting position, target, and cursor) was displayed on the screen (Fig. 1B). The participants were instructed to simultaneously move both hands straight ahead so that the cursor hit the target. They explicitly knew that the movement of the cursor was associated with the movement of the right hand. Since it was difficult to return the invisible left hand to the starting position, a small tactile projection was placed there to help the participants place the invisible left hand.
Cursor rotation condition.
The cursor rotation (CR) condition consisted of four different phases: baseline (15 trials), adaptation (135 trials), unimanual catch trial (five trials), and washout (25 trials) (Fig. 2A). During the adaptation phase, the degree of imposed visual rotation gradually increased from 0° to 45° in a clockwise (CW) direction (0.6°/trial) and fixed at 45° for the remaining 60 trials so that the participants were not aware of the visual rotation. Such a gradual increase in visual rotation was adopted to examine how the participants implicitly adapt their movements to the novel environment (Kagerer et al. 1997; Michel et al. 2007; Saijo and Gomi 2010). In the unimanual catch trial phase, to examine if the invisible left hand actually adapts to the visual rotation, the after effect was measured when the participants performed unimanual movement with the invisible left hand from the starting position toward a target located directly ahead. In experiment 1, since every participant experienced two trials within 1 day, the washout phase was placed at the end of the trial of each condition to eliminate any remaining adaptation effect.
To investigate how the display location of visual information influences the performance of visuomotor learning, we set “left” and “right” conditions, in which the starting and target positions and the cursor were displayed above the left (CR-left) and right (CR-right) hands (Fig. 1B).
Target rotation condition.
In the CR experiment, the visible right hand should adapt to the visual rotation. The resultant change in the direction of movement of the right hand can influence the direction of movement of the invisible left hand. The target rotation (TR) condition was conducted as a control experiment to examine this possible effect. In this condition, the target location gradually shifted counterclockwise (CCW) around the starting position from 0° to 45° every trial (Fig. 1C). Otherwise, the experimental procedure was identical to that of the CR condition. The participants were instructed to move the left hand straight ahead regardless of the movement direction of the right hand.
Individuals participated in all four kinds of conditions (CR and TR conditions × left and right conditions) over 2 days. Participants were tested under both conditions for a given hand in 1 single day (i.e., CR- and TR-left on the same day). The order of the conditions (CR vs. TR) was reversed on the 1st and the 2nd day, and the order of the locations was randomized for each participant.
Experiment 2 was designed to examine whether the cross talk in the error-information assignment was present, even when visual information was appropriately provided for both hands. In contrast to experiment 1, in which only one cursor was displayed, in experiment 2, visual feedback (i.e., cursors) for both hands was provided (Fig. 1D). There were two conditions (n = 8 for each condition): the opposite rotation (OR) and same rotation (SR) conditions. In the OR condition, the directions of visual rotation, applied to the cursors of both hands, were opposite, whereas they were the same in the SR condition. The rotation direction (CW or CCW) was counterbalanced for each condition. If the cross talk were present, the adaptation performance was worse in the OR condition, because the left (or right)-hand movement correction was negatively affected by the cursor of the right (or left) hand in the opposite direction.
Each experiment comprised three phases (Fig. 2B): baseline (20 trials), adaptation (70 trials), and washout (10 trials). In the adaptation phase, the degree of visual rotations was gradually increased from 0° to 30° (0.5°/trial) so that the participants were not aware that it was occurring.
The position data of the handle were low-pass filtered with a zero-lag fourth-order Butterworth filter (cutoff frequency, 5 Hz). Then, the velocity of the handle was calculated by differentiating the position data with a three-point central difference equation. We obtained the position at which the outward velocity peaked. Movement direction was defined as the angle from a line connecting the starting position and the target to a line connecting the starting position and the position where the peak velocity was observed; CCW and CW directions were defined to take positive and negative value of the direction, respectively. The averaged value of the movement direction in the baseline phase was calculated; this value was subtracted from the data in the other phase for subsequent analyses. We also evaluated the movement direction at a fixed time (200 ms) after the movement onset, but the results did not differ. Thus we demonstrated only the data evaluated at the peak velocity.
Quantification of Adaptation
In experiment 1, to quantify the degree of adaptation to the visual rotation and the effect of TR, we calculated the mean movement direction of the last 60 trials of the adaptation phase, in which the amount of cursor or TR reached a constant angle (45°). We also calculated the correlation coefficient of the trial-by-trial changes in the movement directions between both hands to determine the similarity of the trial-by-trial movement corrections between them. Positive and negative correlations indicate that the movement directions of the hands are modified in the same or opposite direction in the extrinsic space, respectively. In the unimanual catch trial phase, only the movement direction of the first of five trials in a row was calculated and compared between conditions.
In experiment 2, the average movement direction of the last 10 trials in the adaptation phase was calculated to quantify the degree of adaptation. The correlation coefficient of the trial-by-trial changes in the movement directions between both hands was also obtained to determine the similarity of the trial-by-trial movement correction.
Data values are expressed as means ± SE, calculated using data from all participants. Repeated-measures ANOVA as well as paired and unpaired t-tests were performed to detect significant differences in the movement direction between hands, conditions, experimental types, and phases in each experiment. We also used a two-way factorial ANOVA to test the interaction between the positions of the visual information displayed (i.e., left vs. right) and the CR vs. TR condition. Regarding the values of the correlation coefficients, we calculated Fisher's z-scores and used them to compare different conditions by repeated-measures ANOVA. If a significant difference was detected by ANOVA, a post hoc Bonferroni test was used for multiple comparisons. The significance threshold was set at P < 0.05.
Experiment 1: Adaptation to Visual Rotation of a Single Cursor
None of the participants noticed the presence of the visual rotation due to the gradual increase. The averaged times to the peak velocity from the trial onset were 310.1 ± 14.3 ms (left hand) and 325.7 ± 24.1 ms (right hand). The averaged peak velocities were 256.6 ± 5.6 mm/s (left hand) and 232.9 ± 11.7 mm/s (right hand). In experiment 1, the participants performed two sessions within 1 day, but the influence was negligible, because the movement direction of the baseline phases was not statistically different between sessions, thanks to the washout phase performed at the end of each session.
Predictably, the movement direction of the visible right hand gradually rotated in the CCW direction and reached a plateau to compensate for the imposed CW visual rotation of the cursor in both the CR-left and CR-right conditions (Fig. 3). The average values of the last 60 trials of the adaptation phase were significantly different from zero [CR-left, 33.1 ± 1.4°, t(9) = 23.7, P < 0.01; CR-right, 34.5 ± 1.1°, t(9) = 30.3, P < 0.01]; however, they were not significantly different between the CR-right and CR-left conditions.
On the other hand, although the movement direction of the left hand did not change in the CR-right condition, it gradually shifted CCW with increasing the CW visual rotation imposed on the cursor of the visible right hand in the CR-left condition (Fig. 3). The average value in the last 60 trials of CR was not significantly different from zero in the CR-right condition [−0.9 ± 3.2°, t(9) = 0.29, P = 0.78]; however, that in the CR-left condition was significantly greater than zero [14.1 ± 3.0°, t(9) = 4.7, P < 0.01; Fig. 4].
In the TR conditions (TR-right and TR-left), the mean movement direction during the last 60 trials of the TR phase was significantly smaller than that of the baseline phase [TR-left, −8.9 ± 2.9°, t(9) = 3.1, P < 0.05; TR-right, −8.8 ± 2.5°, t(9) = 3.5, P < 0.01; Fig. 4]. This indicates that the movement direction of the invisible left hand gradually changed in the direction (CW) opposite of the visible right hand (Fig. 3).
The two-way factorial ANOVA also showed a significant interaction between the position of the visual information (left vs. right) and the CR and TR conditions [F(1,36) = 6.80, P < 0.05]. A post hoc test showed that there was a significant difference between visual positions only for the CR condition [F(1,36) = 13.3, P < 0.01], indicating that the change in the movement direction of the left hand after adaptation of the right hand to the visual rotation was influenced by the position of the visual information displayed only for the CR condition.
Correlation of the Trial-by-Trial Changes with Respect to the Movement Directions of Both Hands
The correlation coefficient between the trial-by-trial changes in the movement directions of both hands during the adaptation phase was positive only in the CR-left condition (r = 0.15 ± 0.06; Fig. 5A), indicating that the movement direction was corrected in the same direction for both hands. The correlation coefficient was negative in the other conditions (CR-right, r = −0.15 ± 0.07; TR-left, r = −0.12 ± 0.04; TR-right, r = −0.12 ± 0.03; Fig. 5A). There was a significant main effect of experimental types [F(3,36) = 7.24, P < 0.001]. Furthermore, the post hoc test indicated that there was a significant difference between the CR-left and other three conditions (P < 0.01; Fig. 5A).
Unimanual Catch Trials
In the movement direction in the first unimanual catch trial, we observed a significant positive after effect (i.e., in the CCW direction) for the CR-left condition [10.3 ± 2.8°, t(9) = 3.6, P < 0.01] but not for the CR-right condition [2.1 ± 2.6°, t(9) = 0.80, P = 0.45]. On the other hand, a significant negative after effect (i.e., in the CW direction) was observed for both the TR-left [−4.3 ± 1.0°, t(9) = 4.3, P < 0.01] and TR-right conditions [−4.3 ± 1.7°, t(9) = 2.6, P < 0.05; Fig. 5B].
Experiment 2: Adaptation to Visual Rotation Applied to Both Hands
In experiment 2, the averaged times to the peak velocity from the trial onset were 295 ± 30.6 ms (left hand) and 300.5 ± 30.0 ms (right hand), and the averaged peak velocities were 255.3 ± 20.0 mm/s (left hand) and 248.3 ± 18.4 mm/s (right hand). Figure 6 shows the changes in the movement direction in experiment 2. The rate of the changes appeared to be slower in the OR condition than in the SR condition (Fig. 6A). The average movement directions of the last 10 trials in the OR condition were 17.4 ± 2.0° and 20.0 ± 0.7° for the left and right hands, respectively. On the other hand, the average movement directions in the SR condition were 24.0 ± 1.6° and 24.0 ± 1.0° for the left and right hands, respectively (Fig. 6B). These changes in movement direction were significantly different between the two conditions for both hands [left hand, t(30) = 9.9, P < 0.01; right hand, t(30) = 7.9, P < 0.01, Fig. 6B].
Correlation of the Trial-by-Trial Changes in the Movement Directions of Both Hands
If the motor learning system refers to only one of the cursors in a given trial because of the limited amount of attention we have, movement of both of the hands should be corrected according to a common cursor's movement error; consequently, they should be corrected in the same direction. Thus we can expect that the trial-by-trial correction in a movement direction should be positively correlated between the hands for both the OR and SR conditions, as was observed in experiment 1 (Fig. 5A). However, the correlation coefficient between the trial-by-trial changes in both hands' movement direction during the adaptation phase averaged across participants was r = −0.07 ± 0.12, which is not significantly different from zero [t(7) = 0.58, P = 0.58; Fig. 6C]. In addition, a significantly positive correlation coefficient was observed in only two participants in the SR condition, indicating that the movement direction is not corrected in the same direction for both hands in most cases. In the OR condition, none of the participants exhibited significant positive correlation coefficients. Rather, the average across participants was significantly negative [r = −0.25 ± 0.07, t(7) = 3.9, P < 0.01; Fig. 6C]. Again, this indicates that they did not refer to only one of the cursors.
Previous studies demonstrate the existence of neural cross talk in the motor planning and execution during bimanual movements (Carson 2005; Heuer 1993; Marteniuk et al. 1984; Mechsner et al. 2001; Sherwood 1994; Swinnen and Wenderoth 2004). For example, cortical neural cross talk exists at a higher level of motor control through interhemispheric interaction via the corpus callosum (Eliassen et al. 2000; Franz et al. 1996; Kennerley et al. 2002). Bilateral brain activity related to the maintenance of unstable bimanual rhythmic movements is reported in the dorsal premotor cortex, a supplementary motor area (Meyer-Lindenberg et al. 2002), and primary motor cortex (Maki et al. 2008). Furthermore, the transition phase from unstable to stable in a bimanual, rhythmic finger-tapping task is also reported to be accompanied with activity in various brain areas, including the ventral premotor cortex though the corpus callosum (Aramaki et al. 2006), indicating the existence of the interhemispheric interaction during bimanual movements. Neural cross talk also exists at a lower level through the ipsilateral corticospinal tract (Cattaert et al. 1999; Kagerer et al. 2003), where motor signals are sent to the effector from both contralateral- and ipsilateral-descending corticofugal fibers. These intersecting interactions of neural pathways are thought to cause involuntary, bimanual interactions, such as mirror movements (Carson 2005).
We hypothesize that such cross talk during bimanual movement exists not only at the level of motor planning and motor execution but also at the level of motor learning. To test this hypothesis, we examined how left-hand movement was corrected during symmetric bimanual movements when visual information of only the right hand was provided (experiment 1). We also investigated that such implicit cross talk, if any, can influence the performance of motor learning, even when two cursors are displayed (experiment 2).
Single Visual Movement Errors are Implicitly Assigned to the Controllers of Both Hands
In the CR-left condition, in which the visual information was displayed just above the left hand, we observed a directional shift in hand movement along with the visual rotation both for the visible right hand and invisible left hand (Fig. 3), despite the fact that participants explicitly knew that the movement of the cursor was associated with their right hand alone. Considering that the participants were unaware of the presence of visual rotation in the adaptation phase, this result suggests that the unnoticeable movement error in the right hand is implicitly used to correct the motor output of the left hand. The significant, positive correlation in the trial-by-trial correction of the movement directions (Fig. 5A) also supports this idea.
In contrast, no such cross talk in the motor learning process was observed when the visual information was displayed just above the right hand (i.e., the CR-right condition). Such lateral bias suggests that the degree of error assignment is influenced by the position of the visual information in the workspace. We assumed that the degree of visual error assignment is related to the credit of visual information; that is, if visual information is on the left side, the left-hand controller tends to put more credit on it. One possible reason for this might be ascribed to the response characteristics of multisensory neurons. These neurons receive both proprioceptive and visual inputs; their response is positively correlated to the distance between the body part and visual stimulus (Bremmer et al. 2001; Duhamel et al. 1998; Fogassi et al. 1999; Graziano et al. 1994; Rizzolatti et al. 1998). If such neurons are involved in visuomotor learning, greater distances between proprioception and visual information in the CR-right condition would decrease the visuomotor learning performance of the left hand.
However, before these results can be ascribed to the cross talk in the motor learning process for certain, we need to consider other possibilities that explain the results of the present study. One of the possibilities is the influence of right-hand movement. Since the movement direction of the right hand gradually shifted with motor learning, such changes in movement direction could mechanically or neurally contribute to directional shift in left-hand movement. However, this is unlikely for the following two reasons. First, the directional shift was observed only in the CR-left condition. Second, in the TR conditions, the changes in the movement direction of the right hand induced the directional shift of the left-hand movement in the opposite direction (Fig. 3), rather than in the same direction, as observed in the CR-left condition. Such shifts toward the opposite direction can be partly explained by a preference for mirror-symmetric movements in bimanual movement when we perform movements with both hands simultaneously (Franz et al. 1996; Kelso 1984; Kelso et al. 1979; Lee et al. 2002; Swinnen et al. 1997; Swinnen and Wenderoth 2004).
Another possibility is that participants tried to maintain a constant distance between both hands, leading to the directional shift in left-hand movement. However, this explanation cannot explain why the shift was observed only in the CR-left condition. Furthermore, we observed significant after effects in the unimanual catch trials, indicating that directional changes in left-hand movement are not caused by the participants' intention to keep the distance between both hands constant but rather, by the adaptation of the left hand. The amount of after effect (10.3 ± 2.8°) was significantly [t(9) = 2.12, P < 0.05] smaller than that of the adaptation phase (14.1 ± 3.0°). This result was consistent with the partial motor learning transfer observed from bimanual to unimanual movement in previous studies (Nozaki et al. 2006), although almost full transfer was reported in visumotor rotation between unimanual and bimanual movements (Wang et al. 2010; Wang and Sainburg 2009).
Directional Shift of Left-Hand Movement in the TR Condition
In the previous section, the shift in the movement direction of the left hand in the TR condition was explained by the preference of mirror-symmetric bimanual movement. Here, we provide the alternative possibility that this shift can be explained by cross talk in the credit-assignment processes. In the TR condition, the movement of the right-hand cursor was gradually rotated CCW, according to the target position shift. If the movement of the cursor is implicitly assigned to the movement controller of the left hand, even in the TR condition, a discrepancy might arise between the controller's prediction (i.e., movement straight ahead) and the cursor's movement (i.e., CCW movement); this prediction error could correct the CW movement direction of the left hand. This idea is corroborated by the results of the TR condition (Figs. 3 and 4), although it could not perfectly explain why the CW shift of the left hand was not greater for the TR-left condition.
Cross Talk Exists Even When Two Cursors are Displayed
Experiment 2 examined the presence of implicit cross talk when cursors for each hand were displayed. In the OR condition, in which two cursors were rotated in opposite directions, the directions of errors between the left and right hands are opposite. Therefore, the cross talk, if it exists, should result in conflict in motor adaptation between the two controllers. In contrast, in the SR condition, such conflicting influence might be much smaller, because movement corrections in the same direction are needed. The results are consistent with the prediction when the presence of cross talk is assumed (Fig. 6, A and B). Similar results are reported by a recent study (Wang et al. 2010) using sudden visual rotation. In that study, the authors ascribe the impairment of visuomotor learning to the greater demands of visual information processing required for the opposite visual rotation. Our proposed cross-talk hypothesis provides a more concrete understanding of what the greater demand indicates. Our cross-talk scheme could also explain a recent finding by Casadio et al. (2010) that the adaptation to two opposite force fields to both of the arms was worse than the adaptation to the same force fields, because the adaptation to the opposite force fields necessarily accompanied the opposite directional visual error.
Although a study (Diedrichsen et al. 2004) reported that the participants did not find it difficult to pay attention to the two cursors when manipulating them with both hands, one might still consider that limited attention could cause the motor learning system to refer to only one of the cursors in a given trial. Furthermore, we did not constrain their eye movements, which would disperse their attention. However, without taking the effect of cross talk into account, such partial reference of visual error cannot explain the difference in motor learning performance between the OR and SR conditions (Fig. 6, A and B); although it might retard motor learning, as long as the credit assignment is appropriate, the degree of retardation should have been identical between these two conditions.
We also doubt that such partial reference itself is the case. If this were true, the trial-by-trial correction in movement direction should have been correlated between both hands, because the controllers referred to the visual error information of only one hand. However, the results are inconsistent with the prediction (Fig. 6C), indicating that the visual-error information of both hands simultaneously influences the motor learning of the movement controllers of both hands.
Possible Neuronal Mechanism of Cross Talk in Bimanual Motor Learning
Previous studies indicate that several brain areas are involved in adaptation to visual rotation during reaching movements, such as the primary cortex (Paz et al. 2003), cerebellum (Imamizu et al. 2000; Smith and Shadmehr 2005), posterior parietal cortex, and premotor cortex (Krakauer et al. 2004). Some of these areas are known to have bilateral, receptive field-like neurons in the Brodmann's Area 5 (BA5) (Iwamura 2000). Bilateral activation has been found in error information processing in response to visual rotation during reaching movements in the BA5 (Diedrichsen et al. 2005) and cerebellum (Diedrichsen et al. 2005; Imamizu et al. 2000). These findings suggest the possibility that some neurons are commonly used to process the information of both arms' movement errors and/or that the error-information processing is not specific to the arms, which might promote a cross-talk effect in bilateral visuomotor learning.
In summary, similar to the presence of cross talk in motor planning and execution during bimanual movement, the present study reveals that there is an implicit cross talk in credit-assignment processes in visuomotor learning. The cross talk is inevitable, despite explicit knowledge about the association of the movement controller and the resultant visual information.
Is such cross talk just a byproduct, or does it play any functional role in bimanual movement control? White and Diedrichsen (2010) recently demonstrated that when a visual rotation is applied to a cursor displayed at the middle position of both hands, both hands exhibit adaptation at the subsequent trial; this indicates that the controllers of both hands refer to only one visual error information. Such flexible construction of the reference to error information would require both hands' movement controllers to possess a mechanism that considers what the other hand is doing. Then, the implicit cross-talk effect found in this study might reflect the foundation of the flexible construction of visual error information. Our recent study also demonstrates that such interference between the controllers of both hands may play an important role in constructing and switching among internal models, depending on the kinematics of the opposite hand (Yokoi et al. 2011). Future studies are needed to elucidate the possible functional role of the interference.
This work was supported by Grants-in-Aid for Japan Society for the Promotion of Science Fellows (#2110411) to S. Kasuga and KAKENHI (#20670008) and the NEXT Program (#LS034) to D. Nozaki.
No conflicts of interest, financial or otherwise, are declared by the author(s).
S. Kasuga is a Research Fellow of the Japan Society for the Promotion of Science. The authors thank Yoshiharu Yamamoto, Gentaro Taga, and the members of the Nozaki lab for their helpful comments.
- Copyright © 2011 the American Physiological Society