Change in sensitivity to visual error in superior colliculus during saccade adaptation

The superior colliculus projection upon the macaque inferior olive

Saccade accommodation is a productive model for exploring the role of the cerebellum in behavioral plasticity. In this model, the target is moved during the saccade, gradually inducing a change in the saccade vector as the animal adapts. The climbing fiber pathway from the inferior olive provides a visual error signal generated by the superior colliculus that is believed to be crucial for cerebellar adaptation. However, the primate tecto-olivary pathway has only been explored using large injections of the central portion of the superior colliculus. To provide a more detailed picture, we have made injections of anterograde tracers into various regions of the macaque superior colliculus. As shown previously, large central injections primarily label a dense terminal field within the C subdivision at caudal end of the contralateral medial inferior olive. Several, previously unobserved, sites of sparse terminal labeling were noted: bilaterally in the dorsal cap of Kooy and ipsilaterally in the C subdivision of the medial inferior olive. Small, physiologically directed, injections into the rostral, small saccade portion of the superior colliculus produced terminal fields in the same regions of the medial inferior olive, but with decreased density. Small injections of the caudal superior colliculus, where large amplitude gaze changes are encoded, again labeled a terminal field located in the same areas. The lack of a topographic pattern within the main tecto-olivary projection suggests that either the precise vector of the visual error is not transmitted to the vermis, or that encoding of this error is via non-topographic means.

Comparison of adaptation characteristics between visually and memory-guided saccades

Saccade adaptation plays a crucial role in maintaining saccade accuracy. The behavioral characteristics and neural mechanisms of saccade adaptation for an externally cued movement, such as visually guided saccades (VGS), are well studied in nonhuman primates. In contrast, little is known about the saccade adaptation of an internally driven movement, such as memory-guided saccades (MGS), which are guided by visuospatial working memory. As the oculomotor plant changes because of growth, aging, or skeletomuscular problems, both types of saccades need to be adapted. Do both saccade types engage a common adaptation mechanism? In this study, we compared the characteristics of amplitude decrease adaptation in MGS with VGS in nonhuman primates. We found that the adaptation speed was faster for MGS than for VGS. Saccade duration changed during MGS adaptation, whereas saccade peak velocity changed during VGS adaptation. We also compared the adaptation field, that is, the gain change for saccade amplitudes other than the adapted. The gain change for MGS declines on both smaller and larger sides of adapted amplitude, more rapidly for larger than smaller amplitudes, whereas the decline in VGS was reversed. Thus, the differences between VGS and MGS adaptation characteristics support the previously suggested hypothesis that the adaptation mechanisms of VGS and MGS are distinct. Furthermore, the result suggests that the MGS adaptation site is a brain structure that influences saccade duration, whereas the VGS adaptation site influences saccade peak velocity. These results should be beneficial for future neurophysiological experiments.NEW & NOTEWORTHY Plasticity helps to overcome persistent motor errors. Such motor plasticity or adaptation can be investigated with saccades. Thus far our knowledge is primarily about visually guided saccades, an externally cued movement, which we can make only when the object is visible at the time of saccade. However, as the world is complex, we can make saccades even when the object is not visible. Here, we investigate the adaptation of an internally driven movement: the memory-guided saccade.

The olivary input to the cerebellum dissociates sensory events from movement plans

Neurons in the inferior olive are thought to anatomically organize the Purkinje cells (P-cells) of the cerebellum into computational modules, but what is computed by each module? Here, we designed a saccade task in marmosets that dissociated sensory events from motor events and then recorded the complex and simple spikes of hundreds of P-cells. We found that when a visual target was presented at a random location, the olive reported the direction of that sensory event to one group of P-cells, but not to a second group. However, just before movement onset, it reported the direction of the planned movement to both groups, even if that movement was not toward the target. At the end of the movement if the subject experienced an error but chose to withhold the corrective movement, only the first group received information about the sensory prediction error. We organized the P-cells based on the information content of their olivary input and found that in the group that received sensory information, the simple spikes were suppressed during fixation, then produced a burst before saccade onset in a direction consistent with assisting the movement. In the second group, the simple spikes were not suppressed during fixation but burst near saccade deceleration in a direction consistent with stopping the movement. Thus, the olive differentiated the P-cells based on whether they would receive sensory or motor information, and this defined their contributions to control of movements as well as holding still.

Activity of the Substantia Nigra Pars Reticulata during Saccade Adaptation

When movements become inaccurate, the resultant error induces motor adaptation to improve accuracy. This error-based motor learning is regarded as a cerebellar function. However, the influence of the other brain areas on adaptation is poorly understood. During saccade adaptation, a type of error-based motor learning, the superior colliculus (SC) sends a postsaccadic error signal to the cerebellum to drive adaptation. Since the SC is directly inhibited by the substantia nigra pars reticulata (SNr), we hypothesized that the SNr might influence saccade adaptation by affecting the SC error signal. In fact, previous studies indicated that the SNr encodes motivation and motivation influences saccade adaptation. In this study, we first established that the SNr projects to the rostral SC, where small error signals are generated, in nonhuman primates. Then, we examined SNr activity while the animal underwent adaptation. SNr neurons paused their activity in association with the error. This pause was shallower and delayed compared with those of no-error trial saccades. The pause at the end of the adaptation was shallower and delayed compared with that at the beginning of the adaptation. The change in the intertrial interval, an indicator of motivation, and adaptation speed had a positive correlation with the changes in the error-related pause. These results suggest that (1) the SNr exhibits a unique activity pattern during the error interval; (2) SNr activity increases during adaptation, consistent with the decrease in SC activity; and (3) motivational decay during the adaptation session might increase SNr activity and influence the adaptation speed.

The cost of correcting for error during sensorimotor adaptation

Learning from error is often a slow process. In machine learning, the learning rate depends on a loss function that specifies a cost for error. Here, we hypothesized that during motor learning, error carries an implicit cost for the brain because the act of correcting for error consumes time and energy. Thus, if this implicit cost could be increased, it may robustly alter how the brain learns from error. To vary the implicit cost of error, we designed a task that combined saccade adaptation with motion discrimination: movement errors resulted in corrective saccades, but those corrections took time away from acquiring information in the discrimination task. We then modulated error cost using coherence of the discrimination task and found that when error cost was large, pupil diameter increased and the brain learned more from error. However, when error cost was small, the pupil constricted and the brain learned less from the same error. Thus, during sensorimotor adaptation, the act of correcting for error carries an implicit cost for the brain. Modulating this cost affects how much the brain learns from error.

An implicit memory of errors limits human sensorimotor adaptation

During extended motor adaptation, learning appears to saturate despite persistence of residual errors. This adaptation limit is not fixed but varies with perturbation variance; when variance is high, residual errors become larger. These changes in total adaptation could relate to either implicit or explicit learning systems. Here, we found that when adaptation relied solely on the explicit system, residual errors disappeared and learning was unaltered by perturbation variability. In contrast, when learning depended entirely, or in part, on implicit learning, residual errors reappeared. Total implicit adaptation decreased in the high-variance environment due to changes in error sensitivity, not in forgetting. These observations suggest a model in which the implicit system becomes more sensitive to errors when they occur in a consistent direction. Thus, residual errors in motor adaptation are at least in part caused by an implicit learning system that modulates its error sensitivity in response to the consistency of past errors.

The Substantia Nigra Pars Reticulata Modulates Error-Based Saccadic Learning in Monkeys

The basal ganglia have long been considered crucial for associative learning, but whether they also are involved in another type of learning, error-based motor learning, is not clear. Error-based learning has been considered the province of the cerebellum. However, learning to use a robotic arm and saccade adaptation, which use error-based learning, are facilitated by motivation, which is a function of the basal ganglia. Additionally, patients with Parkinson’s disease, a basal ganglia deficit, show slower saccade adaptation than age matched controls. To further investigate whether the basal ganglia actually influence error-based learning, we reversibly inactivated the oculomotor portion of the substantia nigra pars reticulata (SNr) in two monkeys and tested saccade adaptation. Here, we show that nigral inactivation affected saccade adaptation. In particular, the inactivation facilitated the amplitude decrease adaptation of ipsiversive saccades. Consistent with previous studies, no effect was seen on the amplitude of the ipsiversive saccades when we did not induce adaptation. Therefore, the facilitated adaptation was not caused by inactivation directly modulating ipsiversive saccades. On the other hand, the kinematics of corrective saccades, which represent error processing, were changed after the inactivation. Thus, our data suggest that the oculomotor SNr assists saccade adaptation by strengthening the error signal. This effect indicates the basal ganglia influence error-based motor learning, a previously unrecognized function.

Population coding in the cerebellum: a machine learning perspective

The cere resembles a feedforward, three-layer network of neurons in which the "hidden layer" consists of Purkinje cells (P-cells) and the output layer consists of deep cerebellar nucleus (DCN) neurons. In this analogy, the output of each DCN neuron is a prediction that is compared with the actual observation, resulting in an error signal that originates in the inferior olive. Efficient learning requires that the error signal reach the DCN neurons, as well as the P-cells that project onto them. However, this basic rule of learning is violated in the cerebellum: the olivary projections to the DCN are weak, particularly in adulthood. Instead, an extraordinarily strong signal is sent from the olive to the P-cells, producing complex spikes. Curiously, P-cells are grouped into small populations that converge onto single DCN neurons. Why are the P-cells organized in this way, and what is the membership criterion of each population? Here, I apply elementary mathematics from machine learning and consider the fact that P-cells that form a population exhibit a special property: they can synchronize their complex spikes, which in turn suppress activity of DCN neuron they project to. Thus complex spikes cannot only act as a teaching signal for a P-cell, but through complex spike synchrony, a P-cell population may act as a surrogate teacher for the DCN neuron that produced the erroneous output. It appears that grouping of P-cells into small populations that share a preference for error satisfies a critical requirement of efficient learning: providing error information to the output layer neuron (DCN) that was responsible for the error, as well as the hidden layer neurons (P-cells) that contributed to it. This population coding may account for several remarkable features of behavior during learning, including multiple timescales, protection from erasure, and spontaneous recovery of memory.

Persistent grasping errors produce depth cue reweighting in perception

When a grasped object is larger or smaller than expected, haptic feedback automatically recalibrates motor planning. Intriguingly, haptic feedback can also affect 3D shape perception through a process called depth cue reweighting. Although signatures of cue reweighting also appear in motor behavior, it is unclear whether this motor reweighting is the result of upstream perceptual reweighting, or a separate process. We propose that perceptual reweighting is directly related to motor control; in particular, that it is caused by persistent, systematic movement errors that cannot be resolved by motor recalibration alone. In Experiment 1, we inversely varied texture and stereo cues to create a set of depth-metamer objects: when texture specified a deep object, stereo specified a shallow object, and vice versa, such that all objects appeared equally deep. The stereo-texture pairings that produced this perceptual metamerism were determined for each participant in a matching task (Pre-test). Next, participants repeatedly grasped these depth metamers, receiving haptic feedback that was positively correlated with one cue and negatively correlated with the other, resulting in persistent movement errors. Finally, participants repeated the perceptual matching task (Post-test). In the condition where haptic feedback reinforced the texture cue, perceptual changes were correlated with changes in grasping performance across individuals, demonstrating a link between perceptual reweighting and improved motor control. Experiment 2 showed that cue reweighting does not occur when movement errors are rapidly corrected by standard motor adaptation. These findings suggest a mutual dependency between perception and action, with perception directly guiding action, and actions producing error signals that drive motor and perceptual learning.

Behavioral training of marmosets and electrophysiological recording from the cerebellum

The common marmoset (Callithrix jacchus) is a promising new model for study of neurophysiological basis of behavior in primates. Like other primates, it relies on saccadic eye movements to monitor and explore its environment. Previous reports have demonstrated some success in training marmosets to produce goal-directed actions in the laboratory. However, the number of trials per session has been relatively small, thus limiting the utility of marmosets as a model for behavioral and neurophysiological studies. In this article, we report the results of a series of new behavioral training and neurophysiological protocols aimed at increasing the number of trials per session while recording from the cerebellum. To improve the training efficacy, we designed a precisely calibrated food regulation regime that motivates the subjects to perform saccade tasks, resulting in ~1,000 reward-driven trials on a daily basis. We then developed a multichannel recording system that uses imaging to target a desired region of the cerebellum, allowing for simultaneous isolation of multiple Purkinje cells in the vermis. In this report, we describe 1) the design and surgical implantation of a computer tomography (CT)-guided, subject-specific head post, 2) the design of a CT- and MRI-guided alignment tool for trajectory guidance of electrodes mounted on an absolute encoder microdrive, 3) development of a protocol for behavioral training of subjects, and 4) simultaneous recordings from pairs of Purkinje cells during a saccade task.NEW & NOTEWORTHY Marmosets present the opportunity to investigate genetically based neurological disease in primates, in particular, diseases that affect social behaviors, vocal communication, and eye movements. All of these behaviors depend on the integrity of the cerebellum. We present training methods that better motivate the subjects, allowing for improved performance, and we also present electrophysiological techniques that precisely target the subject's cerebellum, allowing for simultaneous isolation of multiple Purkinje cells.

Reward Prediction Error Modulates Saccade Vigor

Movement vigor, defined as the reciprocal of the latency from availability of reward to its acquisition, changes with reward magnitude: movements exhibit shorter reaction time and increased velocity when they are directed toward more rewarding stimuli. This invigoration may be due to release of dopamine before movement onset, which has been shown to be modulated by events that signal reward prediction error (RPE). Here, we generated an RPE event in the milliseconds before movement onset and tested whether there was a relationship between RPE and vigor. Human subjects (both sexes) made saccades toward an image. During execution of the primary saccade, we probabilistically changed the position and content of that image, encouraging a secondary saccade. On some trials, the content of the secondary image was more valuable than the first image, resulting in a positive RPE (+RPE) event that preceded the secondary saccade. On other trials, this content was less valuable (-RPE event). We found that reaction time of the secondary saccade was affected in an orderly fashion by the magnitude and direction of the preceding RPE event: the most vigorous saccades followed the largest +RPE, whereas the least vigorous saccades followed the largest -RPE. Presence of the secondary saccade indicated that the primary saccade had experienced a movement error, inducing trial-to-trial adaptation. However, this learning from movement error was not modulated by the RPE event. The data suggest that RPE events, which are thought to transiently alter the release of dopamine, modulate the vigor of the ensuing movement.SIGNIFICANCE STATEMENT Does dopamine release in response to a stimulus serve to invigorate the ensuing movement? To test this hypothesis, we relied on the fact that reward prediction error (RPE) is a strong modulator of dopamine. Our innovation was a task in which an RPE event occurred precisely before onset of a stimulus-driven movement. We probabilistically produced a combination of large or small, negative or positive RPE events and observed that saccade vigor carried a robust signature of the preceding RPE event: high vigor saccades followed +RPE events, whereas low vigor saccades followed -RPE events. This suggests that in humans, vigor is partly controlled through release of dopamine in the moments before onset of a movement.

A neuronal process for adaptive control of primate saccadic system

In 1980, Dr. Optican established the existence of an adaptive plasticity of saccades and its dependence on the cerebellum with Dr. Robinson. The advantage of saccades is that the neuronal mechanisms underlying their generation have been well established. This knowledge allows us to identify the neuronal elements that participate in saccade adaptation. Briefly, the superior colliculus (SC) produces a saccade command signal, which reaches motoneurons in the abducens nucleus via the brainstem burst generator. The SC saccade command also is sent to the oculomotor vermis (OMV), a saccade-related area of the cerebellar cortex, and finally converges on the same motoneurons via the caudal fastigial nucleus (cFN) and inhibitory burst neurons (IBN). During adaptation, the saccade-related burst of SC neurons does not change; however, the activity of the cerebellum and its downstream targets do. We demonstrate that the SC is the source of the error signal to the OMV, and the error signal increases the probability of complex spike occurrence and decreases simple spike activity in the OMV. This decrease, in turn, is delivered through the cFN and IBN neurons to decrease motoneuron activity and hence saccade amplitude.

Elimination of the error signal in the superior colliculus impairs saccade motor learning

Theories of cerebellar-dependent motor learning use the error between the desired and actual movement to correct the erroneous movement. To support this idea, several studies have tried to eliminate the error signal to the cerebellum and demonstrate an impairment of learning. However, such former approaches have not been successful because blocking the error signal also affected the movement to be learned. In this study, we selectively block an error signal for saccade adaptation, a type of cerebellar motor learning, by inactivating the source of the error signal in the superior colliculus without affecting the movement to be learned. Saccade adaptation was impaired. Thus, our study provides the first experimental evidence that an error signal is required for cerebellar motor learning.

When movements become dysmetric, the resultant motor error induces a plastic change in the cerebellum to correct the movement, i.e., motor adaptation. Current evidence suggests that the error signal to the cerebellum is delivered by complex spikes originating in the inferior olive (IO). To prove a causal link between the IO error signal and motor adaptation, several studies blocked the IO, which, unfortunately, affected not only the adaptation but also the movement itself. We avoided this confound by inactivating the source of an error signal to the IO. Several studies implicate the superior colliculus (SC) as the source of the error signal to the IO for saccade adaptation. When we inactivated the SC, the metrics of the saccade to be adapted were unchanged, but saccade adaptation was impaired. Thus, an intact rostral SC is necessary for saccade adaptation. Our data provide experimental evidence for the cerebellar learning theory that requires an error signal to drive motor adaptation.

Signals driving the adaptation of saccades that require spatial updating

Saccades adapt to persistent natural or artificially imposed dysmetrias. The characteristics and circuitry of saccade adaptation have been revealed using a visually guided task (VGT) where the vectors of the target step and the intended saccade command are the same. However, in real life, another saccade occasionally intervenes before the saccade to the target occurs. This necessitates an updating of the intended saccade to account for the intervening saccadic displacement, which dissociates the visual target signal and the intended saccade command. We determined whether the adaptation process is similar for VGT and updated saccades by studying the transfer of adaptation between them. The ultimate visual target was dissociated from the intended saccade command with double-step saccade tasks (DSTs) in which two targets are flashed sequentially at different locations while the monkey maintains fixation. The resulting saccades toward the first and second targets occur in the dark. The transfer of visually guided saccade adaptation to the second saccades of a DST and vice versa depended on the eccentricity of the second visual target, and not the second saccade command. If a target with the same eccentricity as the adapted target appears briefly during the intersaccadic interval of a DST, more adaptation transfers. Because a brief appearance of the visual target either before the first saccade or during the intersaccadic interval influences how much adaptation transfer the second saccade will express, the processing of adaptation and DST updating may overlap. NEW & NOTEWORTHY Adaptation and the spatial updating of saccades are thought to be independent processes. When we dissociate the visual target and the intended saccade command, the transfer of visually guided saccade adaptation to the saccades of the double-step saccade tasks (DST) and vice versa is driven by a visual not motor error. The visual target has an effect until the second saccade of a DST occurs. Therefore, the processing of adaptation and the spatial updating of saccades may overlap.