Skip to main content
NCBI home page
Communications Biology logoLink to Communications Biology
. 2025 Nov 18;8:1550. doi: 10.1038/s42003-025-09068-7

Sustained visual signals in the primate cerebellar dentate nucleus drive associative learning

Yusuke Akiyama 1, Hiroshi Yamada 2,3, Masayuki Matsumoto 2,3,4, Jun Kunimatsu 2,3,✉
PMCID: PMC12627421  PMID: 41254153

Abstract

A number of studies have suggested that the cerebellum has cognitive functions; however, the underlying neuronal mechanisms remain unclear. In this study, we demonstrated that sustained visual signals in the cerebellar dentate nucleus represent the visuomotor associative information. We recorded neuronal activity from the dentate nucleus when monkeys performed a learning task involving the association between visual objects and saccade directions. We found that sustained visual activity was greater during learning than during memory retrieval. This enhancement disappeared under the uncertain reward condition, in which the monkeys did not engage in learning behavior. Furthermore, sustained visual signals changed the response to visual objects depending on the associated saccade direction. This direction selectivity was positively correlated with modulation during learning. These results suggest that sustained visual signals in the dentate nucleus reflect learning related motivation and drive learning by increasing the strength of discrimination among visual objects.

Subject terms: Cerebellum, Neurophysiology

Sustained visual signals in the cerebellar dentate nucleus, enhanced by learning-related motivation, facilitate visual discrimination and contribute to visuomotor associative learning.

Introduction

Several studies have suggested that the cerebellum contributes to various non-motor functions. Brain activity in the cerebellum has been linked to motivation1,2, attention3–5, social behaviors6,7, and various emotional behaviors8,9. The cerebellum has been implicated in psychiatric disorders such as autism spectrum disorder and attention-deficit/hyperactivity disorder10–14, suggesting that these disorders may be at least partially caused by impairments to its function. Therefore, understanding the neuronal processing of these non-motor cerebellar functions is critical. However, because the traditional focus in cerebellum research has been on its role in motor control and execution15,16, it remains unclear which signals related to non-motor cognitive functions are processed in the cerebellum.

It was recently reported that the cerebellum, cerebral cortex, and basal ganglia are interconnected and involved in a number of common functions17. Although reward-based learning has mainly been investigated in the basal ganglia circuit, reports have indicated that the input to the cerebellar cortex contains reward-related information as well18–22. Clinical studies have reported that non-motor associative learning is impaired in patients with cerebellar degenerative diseases23,24. Ipata et al. reported that neural activity in the mediolateral cerebellar cortex (crus I and II) changes in response to rewards during the process of visuomotor associative learning25. This activity has been shown to be connected to the medial prefrontal cortex (mPFC). However, previous studies on visuomotor associations in the lateral prefrontal cortex (lPFC) and striatum have reported that sustained activity during the delay period encodes information related to learning26,27. The cerebellar dentate nucleus, which has strong connections to lPFC, may also be involved in associative learning.

Kunimatsu et al. found that sustained signals in the dentate nucleus during motor preparation were related to motor strategies28. Similar preparatory activities also regulate higher-order control of movement, such as its timing29, self-generated movement30, and motor planning31. Previous neurophysiological reports have shown that sustained activity, rather than phasic, is related to cerebellar function, indicating that sustained signals regulate the processing of cognitive information through the cerebral cortex.

These observations prompted us to explore sustained signals in the dentate nucleus of monkeys during associative learning. In this study, we recorded neuronal activity from single neurons in the dentate nucleus of monkeys and observed sustained visual signals while they were performing learning tasks regarding the associations between visual objects and motor directions. These sustained visual signals encoded the learned visual objects, and were enhanced during learning but not during memory retrieval. This enhancement disappeared under the uncertain reward condition, in which the monkeys did not employ the appropriate learning strategy, indicating that the enhancement of sustained visual responses occurred specifically when the monkeys were motivated to learn the association. Thus, the enhancement of sustained visual signals may reflect the motivation or intention to learn, and facilitate learning by increasing the degree of discrimination between different visual objects.

Results

Performance of the visuomotor association learning task

Two monkeys performed a visuomotor associative learning task that required them to learn the association between two visual objects, involving both leftward and rightward saccades (Fig. 1a). When each monkey looked at the fixation point, one of two fractal objects was presented in the center of the monitor. The saccade targets were then presented at left and right positions on the monitor. The monkeys received a liquid reward if they chose the correct target associated with the fractal object. One of the two fractal objects was randomly selected for each trial. We used the three conditions; 1) the learning condition, 2) the over-trained condition, and 3) the uncertain condition (see Methods in detail). Monkeys performed these conditions as block manners in each recording session. Each block consisted of more than 80 trials. In the learning condition, we used the novel pair of objects in every block and monkey learned the visuomotor association (Fig. 1b, top). In the over-trained condition, we used the pair of objects which monkeys had previously learned the association at least 10 blocks (Fig. 1b, bottom). Different objects were used for the learning and over-trained conditions. In addition, to test the influence of reward uncertainty on the dentate neurons, we also used the uncertain condition in which a fractal object was not associated with any saccade direction (Fig. 1c, random object). The random object was mixed with a pair of learned objects wherein the correct saccade direction was fixed (learned objects), and one of the three objects was presented randomly on each trial in the uncertain condition. The monkeys had experienced this condition at least 10 blocks, as in the over-trained condition.

Fig. 1. Visuomotor associative learning task.

Fig. 1

Open in a new tab

a Sequence of events during the visuomotor associative learning task. Following the white fixation point, one of two fractal objects was presented in the center of the screen. Two white target spots appeared to the left and right at the time of the object’s offset. The monkey then chose one of the targets through its eye movements. The correct target was associated with the fractal object. b Under the learning condition, a pair of novel objects was used. Under the over-trained condition, a pair of learned objects that the monkey has experienced ≥ 10 blocks was used. c Under the uncertain condition, a pair of learned objects was used along with one random object. d Response accuracy in the learning and over-trained conditions. The mean response accuracy and 95% confidence interval for every 10 trials were plotted. The solid and dashed lines indicate the data from monkeys B and M, respectively. e Response accuracy for the learned and random objects.

The monkeys performed the task in a total of 53 sessions (learning block, 53; over-trained block, 53; uncertain block, 39). Figures 1d and 1e showed the performance of two monkeys in these conditions. In the learning condition, the response accuracy gradually increased with the number of trials (Fig. 1d, red lines; 18 and 35 blocks in monkeys B and M, respectively). By contrast, monkeys maintained a high response accuracy from the beginning of each session in over-trained condition (Fig. 1d, blue lines). In the uncertain condition, the response accuracy was consistently high for the learned objects, whereas it was at a random chance level rate for the random object, since the monkeys could not predict the correct saccade direction (Fig. 1e; 17 and 22 blocks in monkeys B and M, respectively).

Behavioral strategy during visuomotor associative learning

In order to perform the task correctly, the monkey need to adopt a behavioral strategy of making saccades in the same direction to the same object after a correct trial, and in the opposite direction after an error one (win-stay lose-shift strategy)32. To examine monkey’s motivation for the task, we calculated the difference in the probabilities of the monkeys’ making saccades in the same direction as the one it was originally presented in when the same object was presented, after both correct and error trials, and compared it between first and second halves of blocks in each condition. In the learning condition, the probability differed significantly between the first and second halves of blocks in both monkeys B (Fig. 2a left panel; Wilcoxon rank-sum test, p = 3.8 × 10-3) and M (right panel; p = 3.1 × 10-4), but no difference was observed in the over-trained condition (monkey B, p = 0.62; monkey M, p = 0.48), indicating that the monkeys’ behaviors adapted to a win-stay lose-shift strategy as the learning progressed. On the other hand, the behavioral strategy toward the random object was suboptimal across all phases (Fig. 2b; monkey B, p = 0.88; monkey M, p = 0.95). It is known that saccade reaction time also reflect task motivation33,34. For both monkeys, target acquisition times for the random object were significantly longer the novel objects in the learning and learned objects in the over-trained conditions (Fig. 2c). These results suggest that the monkeys adapted their strategy and exhibited high task motivation in the learning condition, whereas their motivation was lower when responding to random objects.

Fig. 2. Behavioral results and strategy in the visuomotor associative learning task.

Fig. 2

Open in a new tab

a The difference in probability of a saccade in the same direction when the same object was presented in the center after correct and error responses (center line, median value; box edges, 25th percentile [Q1] and 75th percentile [Q3]; whiskers, minimum and maximum values within 1.5 × [Q3 – Q1]; cross, outliers). Data were compared between the first and second halves of each block (Wilcoxon rank-sum test). b The difference in probability of a saccade in the same direction when the same object was presented in the center after correct and error responses. c Target acquisition time in the learning, over-trained, and uncertain condition (random and learned objects). Data were analyzed using one-way ANOVA followed by Scheffé’s method (one-way ANOVA, monkey B p = 8.7 × 10-5; monkey M p = 8.6 × 10-5).

Two types of visual signals in the dentate nucleus

We recorded the activities of 1008 neurons in the posterior region of the cerebellar dentate nucleus while the monkeys performed the visuomotor associative learning task (473 and 535 neurons in monkeys B and M, respectively). In this study, we focused on their response to a fractal object because the ability to integrate temporally separated stimuli and response information is particularly important in associative learning, and neural activity in response to cue stimuli is thought to play a role in this process within the prefrontal cortex (PFC)35. A previous study reported that activity in the cerebral cortex during the delay period following visual object presentation contributed to associative learning26,27. Principal component analysis (PCA) and k-means clustering revealed two components of the visual response: the early phasic and the late sustained response (Supplementary Fig. 1c). Based on this analysis, we defined the phasic-response neurons as those that exhibited activity during the early visual period (between 1–150 ms immediately after object presentation, see Methods) and the sustained-response neurons as those that exhibited activity during the late visual period (between 201–400 ms after object presentation). We found 56 phasic-response neurons and 181 sustained-response neurons in the ventral and dorsal regions of the dentate nucleus (Fig. 3 and Table 1). Because we did not find a clear difference in neuronal activity between the ventral and dorsal groups, the data were combined for our subsequent analyses. A total of 32 neurons were classified as both phasic- and sustained-response neurons but were plotted as sustained-response neurons.

Fig. 3. Locations of recording sites.

Fig. 3

Open in a new tab

Coronal sections showing the locations of visual neurons (circles, sustained-response; crosses, phasic-response). These sections were taken 9–10 mm posterior to the anterior-posterior (AP) and reconstructed via MR imaging (T2-weighted image, top-right panel). Neurons with both sustained and phasic response characteristics were grouped together as sustained-response neurons. The size of the circle represents the number of neurons. Some neurons from the AP-9 and AP-10 regions were also plotted alongside those from the AP-8 and AP-11 ones.

Table 1.

The number of phasic- and sustained-response neurons recorded from each monkey

Monkey B Monkey M Total
Phasic-response Sustained-response Phasic-response Sustained-response Phasic-response Sustained-response
Learning condition 27 121 29 60 56 181
Over-trained condition 27 121 29 60 56 181
Uncertain condition 24 119 14 33 38 152
All conditions 24 119 14 33 38 152
Open in a new tab

Activity of the sustained- and phasic-response neurons during learning

To examine the changes for each response type that take place during visuomotor associative learning, we compared the neuronal responses between the learning and over-trained conditions. Figures 4a and 4b show the firing rates for representative phasic- and sustained-response neurons. The phasic-response neuron was equally active for both novel and learned objects (Fig. 4a; unpaired-samples Student’s t-test, p = 0.54), whereas the sustained-response neurons were active only for novel objects (Fig. 4b; p = 1.1 × 10-14). Thirty of the 56 phasic-response neurons showed significant differences between the learning and over-trained conditions during the early visual period (unpaired-samples Student’s t-test, p < 0.05, filled circles in Fig. 4c), and there was no significant difference in the population (paired-samples Student’s t-test, p = 0.14; monkey B, p = 0.86; monkey M, p = 8.5 × 10-2). The time courses of the population activity for the 56 phasic-response neurons also showed no differences between the learning and over-trained conditions (Fig. 4e). By contrast, 97 of the 181 sustained-response neurons showed significant differences during the late visual period (p < 0.05, filled circles in Fig. 4d). The activity in this population was significantly greater during the learning condition vs the over-trained one (Fig. 4d; p = 7.1 × 10-7; monkey B, p = 1.0 × 10--4; monkey M, p = 2.2 × 10-3). The time courses of the population activity for the 181 sustained-response neurons confirmed that the activity during the learning condition was greater than that during the over-trained condition (Fig. 4f).

Fig. 4. The phasic and sustained visual responses under learning and over-trained conditions.

Fig. 4

Open in a new tab

a An example of a phasic-response neuron. Data are aligned on the object onset (vertical line). In the upper part of the figure, a raster plot is shown; each dot indicates the timing of a neuron spike, and each row corresponds to an individual trial. This includes the results of all trials in one session, with the bottom row being the first trial. However, it does not reflect the order of the learning and over-trained conditions. In the lower part, red solid and blue dash lines indicate the activity during the learning and over-trained conditions, respectively. Dashed vertical line indicates the onset of saccade target spots presentation. The timing of the saccade and the reward are not included, as these events mostly occurred outside the time window. b An example of a sustained-response neuron. c Each data point compares the mean firing rate during the learning condition with that during the over-trained one, for each phasic-response neuron. Mean firing rates were measured from the early visual period (black bar in a, 1–150 ms after object presentation). Filled symbols indicate the data, showing a statistically significant difference between the conditions (unpaired-samples Student’s t-test, p < 0.05). The circle and triangle indicate data from monkeys B and M, respectively. d Each data point compares the mean firing rates during the learning condition with those during the over-trained one, for each sustained-response neuron. Mean firing rates were measured from the late visual period (black bar in b, 201–400 ms after object presentation). One outlier point has been omitted (monkey B, 165.9, 162.3), but is included in the population activity. e, f Time courses of the population activity during the learning and over-trained conditions for the phasic- e and sustained-response neurons f. The traces indicate the mean and standard error of spike density for individual neurons.

Sustained visual signals during the uncertain condition

The sustained visual responses were enhanced under the learning condition. What does the sustained signal represent? Recent studies have reported that neurons in the septum and striatum show strong responses to uncertain reward cues, but not to the ones that indicate certain outcomes36,37. Sustained visual signals in the dentate nucleus may also encode the context of uncertain reward. To address this, we analyzed the activity of sustained-response neurons in the uncertain condition, which included a random object and a pair of learned objects (Fig. 1c). The behavioral data from this condition revealed that the monkeys did not attempt to learn the association the random object with saccade direction (Fig. 2b, c).

Among the 1008 recorded neurons, 761 were tested in the session with the uncertain condition. We examined the activity of the 152 sustained-response neurons during the uncertain condition, with a representative one shown in Fig. 5a (the same neuron as in Fig. 4b). This neuron showed significantly greater activity during the learning vs the over-trained condition (Fig. 4b); however, it exhibited no significant differences in activity between the random and over-trained objects (which were associated with the ipsilateral or contralateral saccade direction, respectively; Fig. 5a; one-way analysis of variance [ANOVA], p = 0.50). Among the 152 sustained-response neurons, 85 showed a significant difference between the three fractal objects (one-way ANOVA, p < 0.05, filled circles in Fig. 5b). We compared the responses to the three types of objects among the 152 sustained-response neurons quantitatively, but found no significant differences (Fig. 5b; one-way ANOVA, p = 0.66; monkey B, p = 0.64; monkey M, p = 0.97). This result was confirmed by a time course of the activity of the overall sustained-response neuron population (Fig. 5c). Furthermore, even when we restricted the analysis to the 85 neurons that showed significant differences in activity between objects, there were still no significant differences across the three objects (Supplementary Fig. 2; one-way ANOVA, p = 0.69; monkey B, p = 0.69; monkey M, p = 0.94). These results suggest that the sustained visual response does not reflect the probability of an uncertain reward, but is specifically enhanced during visuomotor associative learning.

Fig. 5. Activity of sustained-response neurons during the uncertain condition.

Fig. 5

Open in a new tab

a Example of the response of a representative sustained-response neuron (the same one shown in Fig. 4b) to a random object (green solid line), an ipsilateral saccade object (dark-blue dashed line), and a contralateral saccade object (light-blue dashed line). b The differences in responses among three objects. Each data point indicates the normalized response to three types of objects during the late visual period (black bar in a, see Method). Filled symbols indicate the data, showing a statistically significant difference between responses to three types of objects (one-way ANOVA p < 0.05). c Time courses of the population activity of the sustained-response neurons during the uncertain condition. Traces indicate the mean and standard error of spike density for individual neurons.

Sustained visual signals encode visual objects based on the associated saccade direction

If cerebellar dentate nucleus neurons are involved in associative learning, they may contribute by encoding task-relevant information. In the visuomotor associative learning task, the monkeys needed to determine whether to make saccades to the left or right direction when the object was presented. We calculated direction selectivity based on neuronal responses to objects associated with saccades in either the ipsilateral (ipsilateral saccade object, see Methods in detail) or contralateral direction (contralateral saccade object). Among the 56 phasic-response neurons, only two showed significant differences in their response to each object during the early visual period in correct trials under the learning condition (unpaired-samples Student’s t-test, p < 0.05). On the other hand, 117 of the 181 sustained-response neurons showed significant differences in their responses to each object during the late visual period (Fig. 6a, red; unpaired-samples Student’s t-test, p < 0.05). Comparable numbers of neurons preferred the ipsilateral and contralateral saccade object, and there was no bias toward a specific direction (one-sample Student’s t-test, p = 0.12). Therefore, we defined two types of responses: ipsilateral- and contralateral-object preferred response. These response classifications were based on whether the firing rate to the ipsilateral or contralateral saccade object was higher. Figure 6b shows the population activity of the ipsilateral- (left panel) and contralateral-object preferred neurons (right panel) during the correct (red traces) and error trials (black traces) in the learning condition. Notably, no significant difference was found in error trials (Ipsi-object preferred, paired-samples Student’s t-test, p = 0.27; monkey B p = 0.61; monkey M p = 0.27; contra-object preferred, p = 0.70; monkey B, p = 0.81; monkey M, p = 0.74).

Fig. 6. Direction selectivity of the sustained visual response.

Fig. 6

Open in a new tab

a Direction selectivity of sustained-response neurons in correct trials under the learning condition: differences in responses to the ipsilateral and contralateral saccade object during the late visual period (see Methods). The neurons showing significant direction selectivity are indicated by red (unpaired-samples Student’s t-test, p < 0.05). b Time courses of the population activity of the ipsilateral- (left panel) and contralateral-object preferred neurons (right panel) during the learning condition. Each line indicates the responses to the ipsilateral (solid) and contralateral saccade objects (dashed) in the correct (red) and error trials (black). c Direction selectivity of sustained-response neurons in correct trials under the over-trained condition. The neurons showing significant direction selectivity are indicated by blue. d Time courses of the population activity of the ipsilateral- and contralateral-object preferred neurons during the over-trained condition. e Correlation of direction selectivity within the sustained visual response between the learning and over-trained conditions. The color of each data point indicates neurons that exhibited significantly different responses to ipsilateral and contralateral saccade objects (one-sample Student’s t-test, p < 0.05, orange: both the learning and over-trained conditions; black: only during the learning condition; gray: only during the over-trained condition). The circle and triangle indicate data from monkeys B and M, respectively. The black line represents a linear regression line. f Changes of the magnitude of the direction selectivity in all trials in each condition. The mean magnitude and 95% confidence interval of direction selectivity were plotted for every 10 trials. The thick line represents a linear regression line.

In the over-trained condition, 115 neurons showed significant direction selectivity and exhibited preferences for either ipsilateral or contralateral saccade objects (Fig. 6c, d). As in the learning condition, a similar proportion of neurons showed preference for the ipsilateral and contralateral saccade object, with no bias toward a specific direction (one-sample Student’s t-test, p = 0.84). Note that the number of error trials during the over-trained condition was small and insufficient for this analysis. To examine the relationship of the direction selectivity between the learning and over-trained conditions, we plotted the strength of the direction selectivity during the late visual period for correct trials (Fig. 6e). We found a positive correlation between direction selectivity under the two conditions (Pearson’s r = 0.53, p = 1.8 × 10-14; monkey B, r = 0.68, p = 1.7 × 10-17; monkey M, r = 0.47, p = 1.8 × 10-4). These results indicate that the encoding of direction selectivity possessed by objects during learning was maintained after it had finished.

Next, we examined changes in direction selectivity during learning. In this analysis, both correct and error trials were included in the calculation of direction selectivity. Figure 6f shows changes in the magnitude of direction selectivity during the learning (red) and over-trained (blue) conditions. The magnitude did not change in the over-trained condition (one-sample Student’s t-test, p = 0.47), whereas in the learning condition it significantly increased as trials progressed (p = 1.5 × 10⁻²). This change in direction selectivity aligns with behavioral performance (Fig. 1d). Furthermore, we separately calculated the direction selectivity in trials with erroneous saccade directions (Supplementary Fig. 3). The magnitude of direction selectivity in the error trials showed no bias in the learning condition (one-sample Student’s t-test, p > 0.05) and did not change during learning (p = 0.085). These results suggest that the sustained visual signals encode object contingency information, rather than reflecting motor preparation for a specific direction.

Relation between direction selectivity and the enhancement of sustained visual signals

We proceeded to examine the role of sustained visual signal enhancement during associative learning, as this enhancement may affect the strength of direction selectivity. To examine this, we calculated the learning-modulation index, an index that indicates the degree to which firing rate in the over-trained condition changes in the learning condition (see Methods), and measured its correlation with direction selectivity. Among the 181 sustained-response neurons, 39 and 46 showed both enhancement of sustained responses (unpaired-samples Student’s t-test, p < 0.05) and clear direction selectivity (unpaired-samples Student’s t-test, p < 0.05) under the learning and over-trained conditions, respectively. In Fig. 7, we plotted the absolute value of the learning-modulation indexes and the magnitude of the direction selectivity of these neurons. They exhibited a positive correlation between them under both the learning (Fig. 7a; r = 0.59, p = 8.3 × 10-5; monkey B, r = 0.56, p = 5.6 × 10-3; monkey M, r = 0.62, p = 1.0 × 10-2) and over-trained conditions (Fig. 7b; r = 0.66, p = 5.2 × 10-7; monkey B, r = 0.76, p = 4.2 × 10-7; monkey M, r = 0.54, p = 4.5 × 10-2). These results suggest that direction selectivity is regulated by the enhancement of sustained visual responses during learning, and is maintained afterward.

Fig. 7. Correlations between learning-modulation index and direction selectivity in sustained-response neurons.

Fig. 7

Open in a new tab

a Correlation between the learning-modulation index and the magnitude of the direction selectivity in correct trials for each sustained-response neuron that showed significant learning-modulation activity and direction selectivity under the learning condition. Learning-modulation index is shown as an absolute value, indicating how much the response was modulated in the learning condition compared to the over-trained condition. The circle and triangle dots indicate the data from monkeys B and M, respectively. The black line represents a linear regression line. b Correlation between the learning-modulation index and the magnitude of the direction selectivity in correct trials for each sustained-response neuron that showed significant learning modulation activity and direction selectivity under the over-trained condition.

Discussion

In this study, we revealed that sustained visual signals in the dentate nucleus of monkeys represent information related to associative learning in two ways. First, the sustained visual signal represented the learning state (Fig. 4) but not the uncertain situation (Fig. 5). Second, the sustained response encoded visual objects associated with the saccade direction during and after learning (Fig. 6). Notably, this direction selectivity was stronger according to the strength of enhancement of sustained visual activity during learning (Fig. 7). These results suggest that the enhancement of visual signals modulates learning by increasing direction selectivity.

Then, how does the neuronal circuit comprising the learning signal affect the direction selectivity? Fig. 8 presents a diagram of a hypothetical neuronal network for visuomotor association learning. Our data suggested that the dentate neurons were activated by novel visual objects during the learning condition (Fig. 4). As we always used a novel object pair during the learning condition, the increment of activity change may represent a response to the object’s novelty. However, novelty-related responses are typically phasic and diminish over several trials38,39, making them unlikely to account for the observed activity in this experiment. However, the reward probability was low at the beginning of the learning phase, raising the possibility that this activity reflected the uncertainty of the reward, as reported in the septum and caudate nucleus. However, our results rejected this possibility (Fig. 5). Rather, the fact that neuronal activity tended to be greater during the over-trained condition (without reaching significance) suggests that the monkeys were still attempting to learn during this phase. Thus, our results suggest that late visual signals represent the intention or motivation to learn about the associations between novel objects and movements. Previous studies have demonstrated that motivation is essential for learning40,41, and clinical research has suggested that the cerebellum plays a role in learning-related motivation42. In rodents, cerebellar lesions have been reported to reduce learning motivation1. Based on these findings, we hypothesized that information indicating the learning state may originate from the cerebellar cortex (Fig. 8).

Fig. 8. Hypothetical diagram of the neuronal mechanism underlying visuomotor associative learning in the cerebellum.

Fig. 8

Open in a new tab

Neurons in the dentate nucleus receive saccade direction information from the prefrontal cortex and signals that encode the learning state, from the cerebellar cortex. Sustained visual activity increases during learning and enhances the difference between responses to ipsilateral and contralateral saccade objects. DN: dentate nucleus; PFC: prefrontal cortex.

Although direction-selective signals during saccade preparation have been widely reported in various brain regions43,44, our study found that such activity was not relevant to the dentate nucleus. Instead, our data reflected the direction of the goal associated with visual stimuli (Fig. 6). This suggests that the dentate nucleus represents integrated information that combines visual objects and movement direction, which represents an essential process for associative learning. To the best of our knowledge, this is the first report of such activity in the cerebellum. However, direction-selective activity during visuomotor learning has been reported in lPFC and caudate nucleus45. The activity described by Histed et al. does not rule out a motor-related component27, but its time course closely resembles what we observed in this study, suggesting a functional connection between them. Indeed, a subset of the caudal dentate nucleus (specifically, the dorsal region) has been shown to project to the lPFC46. Moreover, recent studies have reported that the dentate nucleus receives direct cortical projections47. It is therefore possible that the posterior dentate nucleus has a reciprocal connection with the lPFC, facilitating the integration of visual and movement-related information.

The increases in direction selectivity observed during learning (Fig. 7) suggest that the motivational signal modulates visuomotor integration in the dentate nucleus. Although motivation changes behavior, the neural mechanisms underlying how motivation facilitates learning remain unclear. The difference between the contralateral and ipsilateral saccade object responses increased in this study, through the enhanced neuronal activity, which correlated with improvements in visual discriminability. A similar modulation of the visual response was reported in the visual cortex by Treue and Trujillo48. They found that attention increased neuronal responses to all stimulus directions, not just to those moving in a preferred direction, which further improved discriminatory ability. These findings suggest common mechanisms by which internal states modulate the gain of the visual response. Notably, in our study, discriminability of the visual objects was maintained even after learning had been established (Figs. 6, 7). This result suggests that sustained visual signals alter synaptic weights to maintain visual discrimination ability. To verify this hypothesis, the time courses of neuronal activity in the frontal cortex and cerebellum must be explored in future studies. Furthermore, pathway-specific manipulation using pharmacological or optogenetic techniques would likely also significantly further the current understanding of these processes.

In behavioral tasks that require multiple processes, such as visuomotor learning, the brain often employs proactive control strategies. This is crucial for higher-order motor regulation, and several recent reports have described this type of sustained activity in the cerebellum49. Sustained activity has also been implicated in internal timing control30,50, visual attention5 and motor planning31, as well as in behavioral strategy activities such as the speed-accuracy trade-off28. In this study, we found that a sustained visual response was active when the monkeys performed a suitable learning strategy (Figs. 2, 4). Sustained activity in the cerebellum may contribute to the proactive control of behavioral strategies at the cognitive and motor levels. Recently, Sendhilnathan et al. found reward-related signals in the cerebellar crus I and II51 during visuomotor associative learning, and found that these signals were tonically active across trials. However, we did not observe such activity in the posterior dentate nucleus. This suggests that our study may have captured information from distinct neural pathways through crus I and II. Specifically, they demonstrate that crus I and II have strong connections with the mPFC52, whereas the posterior dentate nucleus sends signals to the lPFC46, suggesting that these signals are processed through different pathways. These visual and reward-related signals may be integrated within the frontal cortex to regulate visuomotor associative learning.

Thus, our study suggests that visuomotor association is achieved through the coordination of sustained activity between the dentate nucleus and PFC. To our knowledge, this represents the first proposed mechanism through which motivation or intention facilitates learning.

Methods

Animal preparation

Two Japanese monkeys (Macaca fuscata; monkey B male, 7.6 kg, 5 years old; monkey M, female, 6.6 kg, 5 years old) were used in this study. The care of the monkeys and all experimental procedures were approved in advance by the Animal Experimentation Committee of the University of Tsukuba (approval number: 20-266), and all experiments were performed in accordance with the protocol of the BioResource Project “Japanese Monkeys” of the MEXT, Japan. We have complied with all relevant ethical regulations for animal use. Prior to the study, we surgically placed a head-fixing device and recording chamber on each monkey’s skull, under general anesthesia (isoflurane, 0.5–2.0%). Behavioral task training and recording experiments were performed after the monkeys had fully recovered from surgery.

Behavioral task

We controlled the behavioral task using MonkeyLogic53, which is the MATLAB (Mathworks) application. Each monkey sat on a primate chair with its head fixed in place, within a dark and soundproofed room. The visual stimuli used in the task were presented on a 24-inch monitor (ProLite E2483HS, iiyama) placed 42 cm in front of the monkey’s eyes.

We used a visuomotor associative learning task in which subjects were required to learn the association between two visual objects and leftward or rightward saccades (Fig. 1). In each trial, a white square (0.5° × 0.5°) was presented in the center of the monitor, and the monkey was required to fixate on it for 501-800 ms. One of a pair of fractal objects (~ 6.0° × 6.0°) was then presented pseudo randomly. After fixating on the fractal object for 500 ms, two white targets (0.5° × 0.5°) were simultaneously presented in the left and right positions, at 13.5° from the center. The monkey was required to make a saccade to either the left or right target within 500 ms. If it fixated on the square in the correct direction, a high-pitched beep (2000 Hz) sounded and the monkey received a liquid reward. Otherwise (i.e., if the monkey made a saccade in an error direction, did not make a saccade within the time window, or did not fixate on the target for the required duration), a low-pitched beep (200 Hz) sounded without a reward. A single block consisted of ≥ 80 trials. The fractal objects were created using fractal geometry54.

We used three different conditions: (1) the learning condition (Fig. 1b, top panel), (2) the over-trained condition (Fig. 1b, bottom panel), and (3) the uncertain condition (Fig. 1c). In the learning condition, a pair of two novel fractal objects was used in each block. In the over-trained condition, a pair of learned objects that the monkeys had already been trained on for at least 10 blocks was used. The objects used in the learning and over-trained conditions were different. The uncertain condition consisted of three objects: one “random object,” for which the goal saccade direction was randomly changed on each trial, and a pair of learned objects. As in the over-trained condition, the uncertain condition was also trained on for at least 10 blocks. We tested the responses of individual neurons in sessions that included all three conditions, presented in a block-wise manner. Monkeys performed a single session per recording day. Either a learning or an over-trained block was always presented first, and the uncertain block was always presented last. Although some neurons were lost during the recording session, the analysis was restricted to neurons tested in both the over-trained and learning blocks.

Recording procedure

Based on the stereotaxic atlas55, the recording chamber was positioned above the cerebellum of each subject. Magnetic resonance (MR) images (3 T, Philips) were obtained along the direction of the recording chamber, which was visualized by gadolinium filling the grid holes. To record the activity of a single neuron, a stainless steel probe (S-probe, Plexon) was inserted through the guide tube. The insertion depth of the probe was controlled using a micromanipulator (MO-97, Narishige), and the recording position was controlled using a grid system fixed to the chamber. The probe was lowered into the cerebellar dentate nucleus (8–11 mm posterior to the interaural line and 6–9 mm lateral to the midline). The neuronal signals were amplified and filtered at a sampling rate of 40 kHz (OmniPlex, Plexon). The recorded neural activity data was separated into clusters using dedicated software (Offline Sorter, Plexon) to identify the activity of individual neurons.

Data acquisition

Eye movements were monitored using an infrared eye-tracking system (EyeLink, SR Research) with sampling at 1 kHz. The timestamps of task events were concurrently sampled at 1 kHz and saved as files during the experiment.

Statistics and Reproducibility

The subsequent analyses were performed using MATLAB. All statistical tests were performed with a significant level of 0.05. To examine the time course of learning during the visuomotor associative learning task, we calculated the mean response accuracy and 95% confidence interval for every 10 trials under both the learning and over-trained conditions (Fig. 1d). For the uncertain condition, we calculated the mean response accuracy and 95% confidence interval for every 10 trials when the learned and random objects were presented (Fig. 1e). Next, to examine the behavioral strategies of the monkeys, we analyzed their differences in behavior between trials, after the correct and error trials, when the same objects were presented (Fig. 2a, b). Specifically, for trials in which the same object was presented, the probability of same-direction saccades in trials following error ones was subtracted from that of subsequent correct trials. We then calculated and compared these values for the first and second halves of each block.

Next, we measured the target acquisition time, defined as the latency for the monkey to fixate either of the two saccade targets after they were displayed in each condition (Fig. 2c). In the uncertain condition, target acquisition times were analyzed separately for the random object and the learned objects.

To analyze the responses to visual stimuli, we performed PCA on all recorded dentate nucleus neurons (Supplementary Fig. 1 and Supplementary Table 1). First, we subtracted the mean firing rate over a period of 1–300 ms before the presentation of the fractal object from the firing rate, over a period of 1–400 ms after the presentation of the fractal object. Then, we divided the 400 ms corrected neural activity by the analysis window size of 50 ms and calculated the mean over the eight analysis windows. PCA was applied to a matrix of 1008 (number of neurons) × 8 (number of analysis windows), using the pca function in MATLAB. K-means clustering was then applied to the principal component scores of PC1–PC5, as their cumulative variance explained reached > 95% (Supplementary Fig. 1a). We attempted to determine the optimal number of clusters in the k-means clustering using the elbow method. The maximum second derivative method based on the within-cluster sum of squares (WSS) suggested that the optimal number of clusters was two. However, the overall decrease in WSS was smooth, and two clusters were insufficient to capture the diversity of neural responses. Then, we selected six clusters (Supplementary Fig. 1b), where the reduction in WSS per additional cluster dropped below 10%.

For quantification, we measured the firing rate during the following periods: (1) 1–300 ms before fractal object presentation (baseline period), (2) 1–150 ms after fractal object presentation (early visual period), and (3) 201–400 ms after fractal object presentation (late visual period). All trials from each session were included in the analysis. We calculated the mean firing rates using data from early and late visual periods, and compared them between ipsilateral and contralateral saccade objects. Based on this analysis, we selected the responses to the object that elicited the higher mean firing rate, and compared them during the baseline period, early visual period, and late visual period. Neurons that showed significant differences between neural activity during the baseline period and either the early or late visual periods (determined via one-way ANOVA followed by Scheffé’s method, p < 0.01) were used for subsequent analyses. The neurons with significant firing modulation during the early visual period in the learning or over-trained conditions were defined as phasic-response neurons (Fig. 4a, c, e), whereas those that fired during the late visual period were defined as sustained-response neurons (Fig. 4b, d, f). For both types, activity during the early or late visual period (respectively) was measured and used in subsequent quantitative analyses. The time course of neuronal activity for each condition was qualitatively examined by computing the spike density function using a Gaussian kernel (σ = 15 ms).

To analyze sustained-responses in the uncertain condition, we compared firing rates during the late visual period to the three types of objects: the random object and learned objects associated with ipsilateral and contralateral directions (Fig. 5 and Supplementary Fig. 2). Notably, we normalized the firing rate to the three objects (ipsilateral, contralateral, and random) during the late visual period by dividing it by their sum, to observe the response of individual neuron (Fig. 5b and Supplementary Fig. 2a, c).

To investigate the direction selectivity of the sustained-response neurons, we compared the visual response during the late visual period to ipsilateral and contralateral saccade objects in the correct trials (Fig. 6). Direction selectivity in correct trials was calculated by dividing the firing rate to the ipsilateral saccade object during the late visual period in correct trials, minus that to the contralateral saccade object in correct trials by the sum of them (Fig. 6a, c, e, f). In addition, the direction selectivity of each neuron was compared to zero to determine whether sustained-response neurons exhibited a significant directional bias in the learning and over-trained conditions. Neurons that exhibited higher mean firing rates to the ipsilateral or contralateral saccade objects during the late visual period in each condition were defined as “Ipsilateral-object preferred neurons” and “Contralateral-object preferred neurons” (Fig. 6b, d), respectively. We also calculated direction selectivity in bins of 10 trials across all trials. To quantify its magnitude, the direction selectivity of the contralateral-object preferred neurons was multiplied by –1. To assess changes over time, a regression line was fit, and its slope was tested against zero (Fig. 6f). Furthermore, we also analysis the direction selectivity on trials with correct and saccade direction error (accounting for 95.9% of all error trials, Supplementary Fig. 3). In this analysis we used the session included at least 10 saccades in the error direction for both the ipsilateral and contralateral directions.

To investigate the relationship between the enhancement of sustained-responses and the direction selectivity, we calculated the learning-modulation index and examined its correlation with the direction selectivity (Fig. 7). For this analysis, neurons were used that showed a significant difference in firing rate to ipsilateral saccade objects and contralateral ones (unpaired-sample Student’s t-test), as well as a significant difference between firing rates in the learning and over-trained conditions. Learning-modulation index was calculated by dividing the firing rate during the late visual period in the learning condition minus that in the over-trained condition by the sum of them.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Transparent Peer Review file (1MB, pdf)
Supplementary Information (459.7KB, pdf)
Reporting summary (1.4MB, pdf)

Acknowledgements

We would like to thank S. Nishino and N. Kajiwara for performing the animal care for this study, Y. Suwa and A. Yokoyama for their administrative support, K. Kobayashi, H. Numajiri, and Y. Ito for manufacturing the equipment, and all of the lab members for their comments and discussions. The Japanese monkeys were provided by the National Bio-Resource Project. This work was supported in part by grants from JST, PRESTO (grant numbers: JPMJPR21S4; MEXT, 21H05800 and 21H05036), and the Takeda Science Foundation.

Author contributions

Y.A. and J.K. designed the study, performed the experiments, and analyzed the data. Y.A. and J.K. wrote the manuscript, with contributions from H.Y. and M.M.

Peer review

Peer review information

Communications Biology thanks Yuhao Jin, Liya Ma and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Jacqueline Gottlieb and Jasmine Pan. A peer review file is available.

Data availability

Numerical data for main figures and figure supplements have been provided as source data files56.

Code availability

The code used to analyze the data collected in this study is available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-025-09068-7.

References

  • 1.D’Agata, V., Drago, F., Serapide, F. & Cicirata, F. Effects of cerebellectomy on motivation-related behavior: A time-course study. Physiol. Behav.53, 173–176 (1993). [DOI] [PubMed] [Google Scholar]
  • 2.Bauer, D. J., Kerr, A. L. & Swain, R. A. Cerebellar dentate nuclei lesions reduce motivation in appetitive operant conditioning and open field exploration. Neurobiol. Learn. Mem.95, 166–175 (2011). [DOI] [PubMed] [Google Scholar]
  • 3.Allen, G., Buxton, R. B., Wong, E. C. & Courchesne, E. Attentional Activation of the Cerebellum Independent of Motor Involvement. Science275, 1940–1943 (1997). [DOI] [PubMed] [Google Scholar]
  • 4.Le, T. H., Pardo, J. V. & Hu, X. 4 T-fMRI Study of Nonspatial Shifting of Selective Attention: Cerebellar and Parietal Contributions. J. Neurophysiol.79, 1535–1548 (1998). [DOI] [PubMed] [Google Scholar]
  • 5.Flierman, N. A. et al. Encoding of cerebellar dentate neuron activity during visual attention in rhesus macaques. eLife13, RP99696 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Van Overwalle, F., Baetens, K., Mariën, P. & Vandekerckhove, M. Social cognition and the cerebellum: A meta-analysis of over 350 fMRI studies. NeuroImage86, 554–572 (2014). [DOI] [PubMed] [Google Scholar]
  • 7.Carta, I., Chen, C. H., Schott, A. L., Dorizan, S. & Khodakhah, K. Cerebellar modulation of the reward circuitry and social behavior. Science363, eaav0581 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Schmahmann, J. D. & Sherman, J. C. The cerebellar cognitive affective syndrome. Brain121, 561–579 (1998). [DOI] [PubMed] [Google Scholar]
  • 9.Baumann, O. & Mattingley, J. B. Functional topography of primary emotion processing in the human cerebellum. NeuroImage61, 805–811 (2012). [DOI] [PubMed] [Google Scholar]
  • 10.Castellanos, F. X. et al. Quantitative Brain Magnetic Resonance Imaging in Girls With Attention-Deficit/Hyperactivity Disorder. Arch. Gen. Psychiatry58, 289–295 (2001). [DOI] [PubMed] [Google Scholar]
  • 11.Fatemi, S. H. et al. Purkinje Cell Size Is Reduced in Cerebellum of Patients with Autism. Cell. Mol. Neurobiol.22, 171–175 (2002). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Allen, G. & Courchesne, E. Differential Effects of Developmental Cerebellar Abnormality on Cognitive and Motor Functions in the Cerebellum: An fMRI Study of Autism. Am. J. Psychiatry160, 262–273 (2003). [DOI] [PubMed] [Google Scholar]
  • 13.Susan Mackie, B. S. et al. Cerebellar Development and Clinical Outcome in Attention Deficit Hyperactivity Disorder. Am. J. Psychiatry164, 647–655 (2007). [DOI] [PubMed] [Google Scholar]
  • 14.Stoodley, C. J. Distinct regions of the cerebellum show gray matter decreases in autism, ADHD, and developmental dyslexia. Frontiers in Systems Neuroscience8, 10.3389/fnsys.2014.00092 (2014). [DOI] [PMC free article] [PubMed]
  • 15.Friston, K. J., Frith, C. D., Passingham, R. E., Liddle, P. F. & Frackowiak, R. S. Motor practice and neurophysiological adaptation in the cerebellum: a positron tomography study. Proc. Biol. Sci.248, 223–228 (1992). [DOI] [PubMed] [Google Scholar]
  • 16.Manto, M. et al. Consensus paper: roles of the cerebellum in motor control-the diversity of ideas on cerebellar involvement in movement. Cerebellum11, 457–487 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Caligiore, D. et al. Consensus Paper: Towards a Systems-Level View of Cerebellar Function: the Interplay Between Cerebellum, Basal Ganglia, and Cortex. Cerebellum16, 203–229 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Ohmae, S. & Medina, J. F. Climbing fibers encode a temporal-difference prediction error during cerebellar learning in mice. Nat. Neurosci.18, 1798–1803 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Wagner, M. J., Kim, T. H., Savall, J., Schnitzer, M. J. & Luo, L. Cerebellar granule cells encode the expectation of reward. Nature544, 96–100 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Heffley, W. & Hull, C. Classical conditioning drives learned reward prediction signals in climbing fibers across the lateral cerebellum. eLife8, e46764 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Kostadinov, D., Beau, M., Blanco-Pozo, M. & Häusser, M. Predictive and reactive reward signals conveyed by climbing fiber inputs to cerebellar Purkinje cells. Nat. Neurosci.22, 950–962 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Larry, N., Yarkoni, M., Lixenberg, A. & Joshua, M. Cerebellar climbing fibers encode expected reward size. eLife8, e46870 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Drepper, J., Timmann, D., Kolb, F. P. & Diener, H. C. Non-motor associative learning in patients with isolated degenerative cerebellar disease. Brain122, 87–97 (1999). [DOI] [PubMed] [Google Scholar]
  • 24.Timmann, D. et al. Motor deficits cannot explain impaired cognitive associative learning in cerebellar patients. Neuropsychologia40, 788–800 (2002). [DOI] [PubMed] [Google Scholar]
  • 25.Ipata, A. E. et al. Purkinje cells in Crus I and II encode the visual stimulus and the impending choice as monkeys learn a reinforcement based visuomotor association task. bioRxiv, 2024.2009.2013.612926, 10.1101/2024.09.13.612926 (2024). [DOI] [PubMed]
  • 26.Asaad, W. F., Rainer, G. & Miller, E. K. Neural Activity in the Primate Prefrontal Cortex during Associative Learning. Neuron21, 1399–1407 (1998). [DOI] [PubMed] [Google Scholar]
  • 27.Histed, M. H., Pasupathy, A. & Miller, E. K. Learning Substrates in the Primate Prefrontal Cortex and Striatum: Sustained Activity Related to Successful Actions. Neuron63, 244–253 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Kunimatsu, J., Suzuki, T. W. & Tanaka, M. Implications of Lateral Cerebellum in Proactive Control of Saccades. J. Neurosci.36, 7066–7074 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Kunimatsu, J., Maeda, K. & Hikosaka, O. The Caudal Part of Putamen Represents the Historical Object Value Information. J. Neurosci.39, 1709–1719 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Ashmore, R. C. & Sommer, M. A. Delay activity of saccade-related neurons in the caudal dentate nucleus of the macaque cerebellum. J. Neurophysiol.109, 2129–2144 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Gao, Z. et al. A cortico-cerebellar loop for motor planning. Nature563, 113–116 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Sutton, R. S. & Barto, A. G. Reinforcement learning: An introduction. Vol. 1 (MIT press, 1998).
  • 33.Watanabe, K., Lauwereyns, J. & Hikosaka, O. Effects of motivational conflicts on visually elicited saccades in monkeys. Exp. Brain Res.152, 361–367 (2003). [DOI] [PubMed] [Google Scholar]
  • 34.Rothkirch, M., Ostendorf, F., Sax, A.-L. & Sterzer, P. The influence of motivational salience on saccade latencies. Exp. Brain Res.224, 35–47 (2013). [DOI] [PubMed] [Google Scholar]
  • 35.Miller, E. K. & Cohen, J. D. An Integrative Theory of Prefrontal Cortex Function. Annu. Rev. Neurosci.24, 167–202 (2001). [DOI] [PubMed] [Google Scholar]
  • 36.Monosov, I. E. & Hikosaka, O. Selective and graded coding of reward uncertainty by neurons in the primate anterodorsal septal region. Nat. Neurosci.16, 756–762 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.White, J. K. & Monosov, I. E. Neurons in the primate dorsal striatum signal the uncertainty of object–reward associations. Nat. Commun.7, 12735 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Xiang, J. Z. & Brown, M. W. Differential neuronal encoding of novelty, familiarity and recency in regions of the anterior temporal lobe. Neuropharmacology37, 657–676 (1998). [DOI] [PubMed] [Google Scholar]
  • 39.Foley, N. C., Jangraw, D. C., Peck, C. & Gottlieb, J. Novelty Enhances Visual Salience Independently of Reward in the Parietal Lobe. J. Neurosci.34, 7947–7957 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Deci, E. L., Ryan, R. M. & Williams, G. C. Need satisfaction and the self-regulation of learning. Learn. Individ. Differences8, 165–183 (1996). [Google Scholar]
  • 41.DePasque, S. & Tricomi, E. Effects of intrinsic motivation on feedback processing during learning. NeuroImage119, 175–186 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Thoma, P., Bellebaum, C., Koch, B., Schwarz, M. & Daum, I. The Cerebellum Is Involved in Reward-based Reversal Learning. Cerebellum7, 433–443 (2008). [DOI] [PubMed] [Google Scholar]
  • 43.Kim, J.-N. & Shadlen, M. N. Neural correlates of a decision in the dorsolateral prefrontal cortex of the macaque. Nat. Neurosci.2, 176–185 (1999). [DOI] [PubMed] [Google Scholar]
  • 44.Horwitz, G. D. & Newsome, W. T. Target Selection for Saccadic Eye Movements: Direction-Selective Visual Responses in the Superior Colliculus. J. Neurophysiol.86, 2527–2542 (2001). [DOI] [PubMed] [Google Scholar]
  • 45.Pasupathy, A. & Miller, E. K. Different time courses of learning-related activity in the prefrontal cortex and striatum. Nature433, 873–876 (2005). [DOI] [PubMed] [Google Scholar]
  • 46.Middleton, F. A. & Strick, P. L. Cerebellar Projections to the Prefrontal Cortex of the Primate. J. Neurosci.21, 700–712 (2001). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Shinoda, Y., Sugiuchi, Y., Futami, T. & Izawa, R. Axon collaterals of mossy fibers from the pontine nucleus in the cerebellar dentate nucleus. J. Neurophysiol.67, 547–560 (1992). [DOI] [PubMed] [Google Scholar]
  • 48.Treue, S. & Trujillo, J. C. M. Feature-based attention influences motion processing gain in macaque visual cortex. Nature399, 575–579 (1999). [DOI] [PubMed] [Google Scholar]
  • 49.Tanaka, M. et al. Roles of the Cerebellum in Motor Preparation and Prediction of Timing. Neuroscience462, 220–234 (2021). [DOI] [PubMed] [Google Scholar]
  • 50.Kunimatsu, J., Suzuki, T. W., Ohmae, S. & Tanaka, M. Different contributions of preparatory activity in the basal ganglia and cerebellum for self-timing. eLife7, e35676 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Sendhilnathan, N., Semework, M., Goldberg, M. E. & Ipata, A. E. Neural Correlates of Reinforcement Learning in Mid-lateral Cerebellum. Neuron106, 188–198.e185 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Sendhilnathan, N., Bostan, A. C., Strick, P. L. & Goldberg, M. E. A cerebro-cerebellar network for learning visuomotor associations. Nat. Commun.15, 2519 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Asaad, W. F. & Eskandar, E. N. A flexible software tool for temporally-precise behavioral control in Matlab. J. Neurosci. Methods174, 245–258 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Yamamoto, S., Kim, H. F. & Hikosaka, O. Reward Value-Contingent Changes of Visual Responses in the Primate Caudate Tail Associated with a Visuomotor Skill. J. Neurosci.33, 11227–11238 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Saleem, K. S. & Logothetis, N. K. A combined MRI and histology atlas of the rhesus monkey brain in stereotaxic coordinates. (Academic Press, 2012).
  • 56.Akiyama, Y. & Kunimatsu, J. Data for: Sustained visual signals in the primate cerebellar dentate nucleus drive associative learning [Data set]. figshare. 10.6084/m9.figshare.30326728.v1 (2025). [DOI] [PMC free article] [PubMed]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Transparent Peer Review file (1MB, pdf)
Supplementary Information (459.7KB, pdf)
Reporting summary (1.4MB, pdf)

Data Availability Statement

Numerical data for main figures and figure supplements have been provided as source data files56.

The code used to analyze the data collected in this study is available from the corresponding author upon reasonable request.

Articles from Communications Biology are provided here courtesy of Nature Publishing Group

RESOURCES