Impaired perception of temporal contiguity between action and effect is associated with disorders of agency in schizophrenia.

Delusions of control in schizophrenia are characterized by the striking feeling that one's actions are controlled by external forces. We here tested qualitative predictions inspired by Bayesian causal inference models, which suggest that such misattributions of agency should lead to decreased intentional binding. Intentional binding refers to the phenomenon that subjects perceive a compression of time between their intentional actions and consequent sensory events. We demonstrate that patients with delusions of control perceived less self-agency in our intentional binding task. This effect was accompanied by significant reductions of intentional binding as compared to healthy controls and patients without delusions. Furthermore, the strength of delusions of control tightly correlated with decreases in intentional binding. Our study validated a critical prediction of Bayesian accounts of intentional binding, namely that a pathological reduction of the prior likelihood of a causal relation between one's actions and consequent sensory events-here captured by delusions of control-should lead to lesser intentional binding. Moreover, our study highlights the import of an intact perception of temporal contiguity between actions and their effects for the sense of agency.

Closed-Loop Optogenetic Perturbation of Macaque Oculomotor Cerebellum: Evidence for an Internal Saccade Model.

Internal models are essential for the production of accurate movements. The accuracy of saccadic eye movements is thought to be mediated by an internal model of oculomotor mechanics encoded in the cerebellum. The cerebellum may also be part of a feedback loop that predicts the displacement of the eyes and compares it to the desired displacement in real time to ensure that saccades land on target. To investigate the role of the cerebellum in these two aspects of saccade production, we delivered saccade-triggered light pulses to channelrhodopsin-2-expressing Purkinje cells in the oculomotor vermis (OMV) of two male macaque monkeys. Light pulses delivered during the acceleration phase of ipsiversive saccades slowed the deceleration phase. The long latency of these effects and their scaling with light pulse duration are consistent with an integration of neural signals at or downstream of the stimulation site. In contrast, light pulses delivered during contraversive saccades reduced saccade velocity at short latency and were followed by a compensatory reacceleration which caused gaze to land on or near the target. We conclude that the contribution of the OMV to saccade production depends on saccade direction; the ipsilateral OMV is part of a forward model that predicts eye displacement, whereas the contralateral OMV is part of an inverse model that creates the force required to move the eyes with optimal peak velocity for the intended displacement.

Developmentally Unique Cerebellar Processing Prioritizes Self- over Other-Generated Movements.

Animals must distinguish the sensory consequences of self-generated movements (reafference) from those of other-generated movements (exafference). Only self-generated movements entail the production of motor copies (i.e., corollary discharges), which are compared with reafference in the cerebellum to compute predictive or internal models of movement. Internal models emerge gradually over the first three postnatal weeks in rats through a process that is not yet fully understood. Previously, we demonstrated in postnatal day (P) 8 and P12 rats that precerebellar nuclei convey corollary discharge and reafference to the cerebellum during active (REM) sleep when pups produce limb twitches. Here, recording from a deep cerebellar nucleus (interpositus, IP) in P12 rats of both sexes, we compared reafferent and exafferent responses with twitches and limb stimulations, respectively. As expected, most IP units showed robust responses to twitches. However, in contrast with other sensory structures throughout the brain, relatively few IP units showed exafferent responses. Upon finding that exafferent responses occurred in pups under urethane anesthesia, we hypothesized that urethane inhibits cerebellar cortical cells, thereby disinhibiting exafferent responses in IP. In support of this hypothesis, ablating cortical tissue dorsal to IP mimicked the effects of urethane on exafference. Finally, the results suggest that twitch-related corollary discharge and reafference are conveyed simultaneously and in parallel to cerebellar cortex and IP. Based on these results, we propose that twitches provide opportunities for the nascent cerebellum to integrate somatotopically organized corollary discharge and reafference, thereby enabling the development of closed-loop circuits and, subsequently, internal models.

Universal structural requirements for maximal robust perfect adaptation in biomolecular networks.

Adaptation is a running theme in biology. It allows a living system to survive and thrive in the face of unpredictable environments by maintaining key physiological variables at their desired levels through tight regulation. When one such variable is maintained at a certain value at the steady state despite perturbations to a single input, this property is called robust perfect adaptation (RPA). Here we address and solve the fundamental problem of maximal RPA (maxRPA), whereby, for a designated output variable, RPA is achieved with respect to perturbations in virtually all network parameters. In particular, we show that the maxRPA property imposes certain structural constraints on the network. We then prove that these constraints are fully characterized by simple linear algebraic stoichiometric conditions which differ between deterministic and stochastic descriptions of the dynamics. We use our results to derive a new internal model principle (IMP) for biomolecular maxRPA networks, akin to the celebrated IMP in control theory. We exemplify our results through several known biological examples of robustly adapting networks and construct examples of such networks with the aid of our linear algebraic characterization. Our results reveal the universal requirements for maxRPA in all biological systems, and establish a foundation for studying adaptation in general biomolecular networks, with important implications for both systems and synthetic biology.

Interchangeable Role of Motor Cortex and Reafference for the Stable Execution of an Orofacial Action.

Animals interact with their environment through mechanically active, mobile sensors. The efficient use of these sensory organs implies the ability to track their position; otherwise, perceptual stability or prehension would be profoundly impeded. The nervous system may keep track of the position of a sensorimotor organ via two complementary feedback mechanisms-peripheral reafference (external, sensory feedback) and efference copy (internal feedback). Yet, the potential contributions of these mechanisms remain largely unexplored. By training male rats to place one of their vibrissae within a predetermined angular range without contact, a task that depends on knowledge of vibrissa position relative to their face, we found that peripheral reafference is not required. The presence of motor cortex is not required either, except in the absence of peripheral reafference to maintain motor stability. Finally, the red nucleus, which receives descending inputs from motor cortex and cerebellum and projects to facial motoneurons, is critically involved in the execution of the vibrissa positioning task. All told, our results point toward the existence of an internal model that requires either peripheral reafference or motor cortex to optimally drive voluntary motion.SIGNIFICANCE STATEMENT How does an animal know where a mechanically active, mobile sensor lies relative to its body? We address this basic question in sensorimotor integration using the motion of the vibrissae in rats. We show that rats can learn to reliably position their vibrissae in the absence of sensory feedback or in the absence of motor cortex. Yet, when both sensory feedback and motor cortex are absent, motor precision is degraded. This suggests the existence of an internal model able to operate in closed- and open-loop modes, requiring either motor cortex or sensory feedback to maintain motor stability.

Moving without sensory feedback: online TMS over the dorsal premotor cortex impairs motor performance during ischemic nerve block.

The study investigates the role of dorsal premotor cortex (PMd) in generating predicted sensory consequences of movements, i.e. corollary discharges. In 2 different sessions, we disrupted PMd and parietal hand's multisensory integration site (control area) with transcranial magnetic stimulation (TMS) during a finger-sequence-tapping motor task. In this TMS sham-controlled design, the task was performed with normal sensory feedback and during upper-limb ischemic nerve block (INB), in a time-window where participants moved without somatosensation. Errors and movement timing (objective measures) and ratings about movement perception (subjective measures) were collected. We found that INB overall worsens objective and subjective measures, but crucially in the PMd session, the absence of somatosensation together with TMS disruption induced more errors, less synchronized movements, and increased subjective difficulty ratings as compared with the parietal control session (despite a carryover effect between real and sham stimulation to be addressed in future studies). Contrarily, after parietal area interference session, when sensory information is already missing due to INB, motor performance was not aggravated. Altogether these findings suggest that the loss of actual (through INB) and predicted (through PMd disruption) somatosensory feedback degraded motor performance and perception, highlighting the crucial role of PMd in generating corollary discharge.

Temporal Information Processing in the Cerebellum and Basal Ganglia.

Temporal information processing in the range of a few hundred milliseconds to seconds involves the cerebellum and basal ganglia. In this chapter, we present recent studies on nonhuman primates. In the studies presented in the first half of the chapter, monkeys were trained to make eye movements when a certain amount of time had elapsed since the onset of the visual cue (time production task). The animals had to report time lapses ranging from several hundred milliseconds to a few seconds based on the color of the fixation point. In this task, the saccade latency varied with the time length to be measured and showed stochastic variability from one trial to the other. Trial-to-trial variability under the same conditions correlated well with pupil diameter and the preparatory activity in the deep cerebellar nuclei and the motor thalamus. Inactivation of these brain regions delayed saccades when asked to report subsecond intervals. These results suggest that the internal state, which changes with each trial, may cause fluctuations in cerebellar neuronal activity, thereby producing variations in self-timing. When measuring different time intervals, the preparatory activity in the cerebellum always begins approximately 500 ms before movements, regardless of the length of the time interval being measured. However, the preparatory activity in the striatum persists throughout the mandatory delay period, which can be up to 2 s, with different rate of increasing activity. Furthermore, in the striatum, the visual response and low-frequency oscillatory activity immediately before time measurement were altered by the length of the intended time interval. These results indicate that the state of the network, including the striatum, changes with the intended timing, which lead to different time courses of preparatory activity. Thus, the basal ganglia appear to be responsible for measuring time in the range of several hundred milliseconds to seconds, whereas the cerebellum is responsible for regulating self-timing variability in the subsecond range. The second half of this chapter presents studies related to periodic timing. During eye movements synchronized with alternating targets at regular intervals, different neurons in the cerebellar nuclei exhibit activity related to movement timing, predicted stimulus timing, and the temporal error of synchronization. Among these, the activity associated with target appearance is particularly enhanced during synchronized movements and may represent an internal model of the temporal structure of stimulus sequence. We also considered neural mechanism underlying the perception of periodic timing in the absence of movement. During perception of rhythm, we predict the timing of the next stimulus and focus our attention on that moment. In the missing oddball paradigm, the subjects had to detect the omission of a regularly repeated stimulus. When employed in humans, the results show that the fastest temporal limit for predicting each stimulus timing is about 0.25 s (4 Hz). In monkeys performing this task, neurons in the cerebellar nuclei, striatum, and motor thalamus exhibit periodic activity, with different time courses depending on the brain region. Since electrical stimulation or inactivation of recording sites changes the reaction time to stimulus omission, these neuronal activities must be involved in periodic temporal processing. Future research is needed to elucidate the mechanism of rhythm perception, which appears to be processed by both cortico-cerebellar and cortico-basal ganglia pathways.

Validating model-based Bayesian integration using prior-cost metamers.

There are two competing views on how humans make decisions under uncertainty. Bayesian decision theory posits that humans optimize their behavior by establishing and integrating internal models of past sensory experiences (priors) and decision outcomes (cost functions). An alternative hypothesis posits that decisions are optimized through trial and error without explicit internal models for priors and cost functions. To distinguish between these possibilities, we introduce a paradigm that probes the sensitivity of humans to transitions between prior-cost pairs that demand the same optimal policy (metamers) but distinct internal models. We demonstrate the utility of our approach in two experiments that were classically explained by Bayesian theory. Our approach validates the Bayesian learning strategy in an interval timing task but not in a visuomotor rotation task. More generally, our work provides a domain-general approach for testing the circumstances under which humans explicitly implement model-based Bayesian computations.

The attention schema theory in a neural network agent: Controlling visuospatial attention using a descriptive model of attention.

In the attention schema theory (AST), the brain constructs a model of attention, the attention schema, to aid in the endogenous control of attention. Growing behavioral evidence appears to support the presence of a model of attention. However, a central question remains: does a controller of attention actually benefit by having access to an attention schema? We constructed an artificial deep Q-learning neural network agent that was trained to control a simple form of visuospatial attention, tracking a stimulus with an attention spotlight in order to solve a catch task. The agent was tested with and without access to an attention schema. In both conditions, the agent received sufficient information such that it should, theoretically, be able to learn the task. We found that with an attention schema present, the agent learned to control its attention spotlight and learned the catch task. Once the agent learned, if the attention schema was then disabled, the agent's performance was greatly reduced. If the attention schema was removed before learning began, the agent was impaired at learning. The results show how the presence of even a simple attention schema can provide a profound benefit to a controller of attention. We interpret these results as supporting the central argument of AST: the brain contains an attention schema because of its practical benefit in the endogenous control of attention.

People adapt a consistent center-of-mass trajectory in a novel force field.

During human walking the whole body center-of-mass (COM) trajectory may be a control objective, a goal the central nervous system uses to plan and regulate movement. Our previous observation, that after practice walking in a novel laterally directed force field people adapt a COM trajectory similar to their normal trajectory, supports this idea. However, our prior work only presented data demonstrating changes in COM trajectory in response to a single force field. To evaluate whether this phenomena is robust, in the present study we present new data demonstrating that people adapt their COM trajectory in a similar manner when the direction of the external force field is changed resulting in drastically different lower limb joint dynamics. Specifically, we applied a continuous, left-directed force field (in the previous experiment the force field was applied to the right) to the COM as participants performed repeated trials of a discrete walking task. We again hypothesized that with practice walking in the force field people would adapt a COM trajectory that was similar to their baseline performance and exhibit aftereffects, deviation of their COM trajectory in the opposite direction of force field, when the field was unexpectedly removed. These hypotheses were supported and suggest that participants formed an internal model to control their COM trajectory. Collectively these findings demonstrate that people adapt their gait patterns to anticipate consistent aspects of the external environment. These findings suggest that this response is robust to force fields applied in multiple directions that may require substantially different neural control.NEW & NOTEWORTHY With experience people adapted a predictive internal model to control their whole body center-of-mass walking trajectory that anticipated the disruptive laterally directed forces of a novel and consistent external environment. Collectively these findings demonstrate that adaptation of gait to anticipate consistent aspects of the external environment is a response that is robust to force fields in multiple directions that require substantially different lower limb dynamics and neural control.

Reinforcement Motor Learning After Cerebellar Damage Is Related to State Estimation.

Recent work showed that individuals with cerebellar degeneration could leverage intact reinforcement learning (RL) to alter their movement. However, there was marked inter-individual variability in learning, and the factors underlying it were unclear. Cerebellum-dependent sensory prediction may contribute to RL in motor contexts by enhancing body state estimates, which are necessary to solve the credit-assignment problem. The objective of this study was to test the relationship between the predictive component of state estimation and RL in individuals with cerebellar degeneration. Individuals with cerebellar degeneration and neurotypical control participants completed two tasks: an RL task that required them to alter the angle of reaching movements and a state estimation task that tested the somatosensory perception of active and passive movement. The state estimation task permitted the calculation of the active benefit shown by each participant, which is thought to reflect the cerebellum-dependent predictive component of state estimation. We found that the cerebellar and control groups showed similar magnitudes of learning with reinforcement and active benefit on average, but there was substantial variability across individuals. Using multiple regression, we assessed potential predictors of RL. Our analysis included active benefit, somatosensory acuity, clinical ataxia severity, movement variability, movement speed, and age. We found a significant relationship in which greater active benefit predicted better learning with reinforcement in the cerebellar, but not the control group. No other variables showed significant relationships with learning. Overall, our results support the hypothesis that the integrity of sensory prediction is a strong predictor of RL after cerebellar damage.

Respective Involvement of the Right Cerebellar Crus I and II in Syntactic and Semantic Processing for Comprehension of Language.

The right posterolateral portions of the cerebellum (crus-I/II) are involved in language processing. However, their functional role in language remains unknown. The cerebellum is hypothesized to acquire an internal model that is a functional copy of mental representations in the cerebrum and to contribute to cognitive function. In this research, based on the cerebellar internal model hypothesis, we conducted task-based and resting-state functional magnetic resonance imaging (fMRI) experiments to investigate the role of the cerebellum in the syntactic and semantic aspects of comprehension of sentences. In a syntactic task, participants read sentences with center-embedded hierarchical structures. The hierarchical level-dependent activity was found in the right crus-I as well as Broca's area (p < 0.05, voxel-based small volume correction (SVC)). In a semantic task, the participants read three types of sentences for investigation of sentence-level, phrase-level, and word-level semantic processing. The semantic level-dependent activity was found in the right crus-II as well as in the left anterior temporal lobe and the left angular gyrus (p < 0.05, voxel-based SVC). Moreover, the right crus-I/II showed significant activity when the cognitive load was high. Resting-state fMRI demonstrated intrinsic functional connectivity between the right crus-I/II and language-related regions in the left cerebrum (p < 0.05, voxel-based SVC). These findings suggest that the right crus-I and crus-II are involved, respectively, in the syntactic and semantic aspects of sentence processing. The cerebellum assists processing of language in the cerebrum when the cognitive load is high.

The (in)effectiveness of anticipatory vibrotactile cues in mitigating motion sickness.

The introduction of (fully) automated vehicles has generated a re-interest in motion sickness, given that passengers suffer much more from motion sickness compared to car drivers. A suggested solution is to improve the anticipation of passive self-motion via cues that alert passengers of changes in the upcoming motion trajectory. We already know that auditory or visual cues can mitigate motion sickness. In this study, we used anticipatory vibrotactile cues that do not interfere with the (audio)visual tasks passengers may want to perform. We wanted to investigate (1) whether anticipatory vibrotactile cues mitigate motion sickness, and (2) whether the timing of the cue is of influence. We therefore exposed participants to four sessions on a linear sled with displacements unpredictable in motion onset. In three sessions, an anticipatory cue was presented 0.33, 1, or 3 s prior to the onset of forward motion. Using a new pre-registered measure, we quantified the reduction in motion sickness across multiple sickness scores in these sessions relative to a control session. Under the chosen experimental conditions, our results did not show a significant mitigation of motion sickness by the anticipatory vibrotactile cues, irrespective of their timing. Participants yet indicated that the cues were helpful. Considering that motion sickness is influenced by the unpredictability of displacements, vibrotactile cues may mitigate sickness when motions have more (unpredictable) variability than those studied here.

Simple spike dynamics of Purkinje cells in the macaque vestibulo-cerebellum during passive whole-body self-motion.

Theories of cerebellar functions posit that the cerebellum implements internal models for online correction of motor actions and sensory estimation. As an example of such computations, an internal model resolves a sensory ambiguity where the peripheral otolith organs in the inner ear sense both head tilts and translations. Here we exploit the response dynamics of two functionally coupled Purkinje cell types in the vestibular part of the caudal vermis (lobules IX and X) to understand their role in this computation. We find that one population encodes tilt velocity, whereas the other, translation-selective, population encodes linear acceleration. We predict that an intermediate neuronal type should temporally integrate the output of tilt-selective cells into a tilt position signal.

A control-based observer approach for estimating energy intake during pregnancy.

Gestational weight gain outside of Institute of Medicine guidelines poses a risk to both the mother and her unborn child. Behavioral interventions such as Healthy Mom Zone (HMZ) that aim to regulate gestational weight gain require self-monitoring of energy intake, which is often significantly under-reported by participants. This paper describes the use of a control systems approach for energy intake estimation during pregnancy. It relies on an energy balance model that predicts gestational weight based on physical activity and energy intake, the latter treated as an unmeasured disturbance. Two control-based observer formulations relying on Internal Model Control and Model Predictive Control, respectively, are presented in this paper, first for a hypothetical participant, then on data collected from four HMZ participants. Results demonstrate the effectiveness of the method, with generally best results obtained when estimating energy intake over a weekly time period.

[Adaptive repetitive control of wrist tremor suppression based on functional electrical stimulation].

Tremor is an involuntary and repetitive swinging movement of limb, which can be regarded as a periodic disturbance in tremor suppression system based on functional electrical stimulation (FES). Therefore, using repetitive controller to adjust the level and timing of FES applied to the corresponding muscles, so as to generate the muscle torque opposite to the tremor motion, is a feasible means of tremor suppression. At present, most repetitive control systems based on FES assume that tremor is a fixed single frequency signal, but in fact, tremor may be a multi-frequency signal and the tremor frequency also varies with time. In this paper, the tremor data of intention tremor patients are analyzed from the perspective of frequency, and an adaptive repetitive controller with internal model switching is proposed to suppress tremor signals with different frequencies. Simulation and experimental results show that the proposed adaptive repetitive controller based on parallel multiple internal models and series high-order internal model switching can suppress tremor by up to 84.98% on average, which is a significant improvement compared to the traditional single internal model repetitive controller and filter based feedback controller. Therefore, the adaptive repetitive control method based on FES proposed in this paper can effectively address the issue of wrist intention tremor in patients, and can offer valuable technical support for the rehabilitation of patients with subsequent motor dysfunction.

Increased upper-limb sensory attenuation with age.

The pressure of our own finger on the arm feels differently than the same pressure exerted by an external agent: the latter involves just touch, whereas the former involves a combination of touch and predictive output from the internal model of the body. This internal model predicts the movement of our own finger, and hence the intensity of the sensation of the finger press is decreased. A decrease in intensity of the self-produced stimulus is called sensory attenuation. It has been reported that, because of decreased proprioception with age and an increased reliance on the prediction of the internal model, sensory attenuation is increased in older adults. In this study, we used a force matching paradigm to test whether sensory attenuation is also present over the arm and whether aging increases sensory attenuation. We demonstrated that, although both young and older adults overestimate a self-produced force, older adults overestimate it even more, showing an increased sensory attenuation. In addition, we also found that both younger and older adults self-produce higher forces when activating the homologous muscles of the upper limb. Although this is traditionally viewed as evidence for an increased reliance on internal model function in older adults because of decreased proprioception, proprioception appeared unimpaired in our older participants. This begs the question of whether an age-related decrease in proprioception is really responsible for the increased sensory attenuation observed in older people.NEW & NOTEWORTHY Forces generated externally (by the environment on the participant) and internally (by the participant on her/his body) are not perceived with the same intensity. Internally generated forces are perceived less intensely than externally generated ones. This difference in force sensation has been shown to be higher in elderly participants when the forces were applied on the fingers because of their impaired proprioception. Here we replicated this finding for the arm but suggest that it is unlikely to be linked to impaired proprioception.

Predictive steering: integration of artificial motor signals in self-motion estimation.

The brain's computations for active and passive self-motion estimation can be unified with a single model that optimally combines vestibular and visual signals with sensory predictions based on efference copies. It is unknown whether this theoretical framework also applies to the integration of artificial motor signals, such as those that occur when driving a car, or whether self-motion estimation in this situation relies on sole feedback control. Here, we examined if training humans to control a self-motion platform leads to the construction of an accurate internal model of the mapping between the steering movement and the vestibular reafference. Participants (n = 15) sat on a linear motion platform and actively controlled the platform's velocity using a steering wheel to translate their body to a memorized visual target (motion condition). We compared their steering behavior to that of participants (n = 15) who remained stationary and instead aligned a nonvisible line with the target (stationary condition). To probe learning, the gain between the steering wheel angle and the platform or line velocity changed abruptly twice during the experiment. These gain changes were virtually undetectable in the displacement error in the motion condition, whereas clear deviations were observed in the stationary condition, showing that participants in the motion condition made within-trial changes to their steering behavior. We conclude that vestibular feedback allows not only the online control of steering but also a rapid adaptation to the gain changes to update the brain's internal model of the mapping between the steering movement and the vestibular reafference.NEW & NOTEWORTHY Perception of self-motion is known to depend on the integration of sensory signals and, when the motion is self-generated, the predicted sensory reafference based on motor efference copies. Here we show, using a closed-loop steering experiment with a direct coupling between the steering movement and the vestibular self-motion feedback, that humans are also able to integrate artificial motor signals, like the motor signals that occur when driving a car.

Estimating the time structure of descending activation that generates movements at different speeds.

In targeted movements of the hand, descending activation patterns must not only generate muscle activation but also adjust spinal reflexes from stabilizing the initial to stabilizing the final postural state. We estimate descending activation patterns that change minimally while generating a targeted movement within a given movement time based on a model of the biomechanics, the muscle dynamics, and the stretch reflex. The estimated descending activation patterns predict human movement trajectories quite well. Their temporal structure varies across workspace and with movement speed, from monotonic profiles for slow movements to nonmonotonic profiles for fast movements. Descending activation patterns at different speeds thus do not result from a mere rescaling of invariant templates but reflect varying needs to compensate for interaction torques and muscle dynamics. The virtual attractor trajectories, on which active muscle torques are zero, lie within reachable workspace and are largely invariant when represented in end-effector coordinates. Their temporal structure along movement direction changes from linear ramps to "N-shaped" profiles with increasing movement speed.NEW & NOTEWORTHY The descending activation patterns driving movement must be integrated with spinal reflexes, which would resist movement if left unchanged. We estimate the descending activation patterns at different movement speeds using a model of the stretch reflex and of muscle and limb dynamics. The descending activation patterns we find are temporally structured to compensate for interaction torques as predicted by internal models but also shift the reflex threshold, solving the posture-movement problem.

Cerebellar Representations of Errors and Internal Models.

After decades of study, a comprehensive understanding of cerebellar function remains elusive. Several hypotheses have been put forward over the years, including that the cerebellum functions as a forward internal model. Integrated into the forward model framework is the long-standing view that Purkinje cell complex spike discharge encodes error information. In this brief review, we address both of these concepts based on our recordings of cerebellar Purkinje cells over the last decade as well as newer findings from the literature. During a high-dimensionality tracking task requiring continuous error processing, we find that complex spike discharge provides a rich source of non-error signals to Purkinje cells, indicating that the classical error encoding role ascribed to climbing fiber input needs revision. Instead, the simple spike discharge of Purkinje cells carries robust predictive and feedback signals of performance errors, as well as kinematics. These simple spike signals are consistent with a forward internal model. We also show that the information encoded in the simple spike is dynamically adjusted by the complex spike firing. Synthesis of these observations leads to the hypothesis that complex spikes convey behavioral state changes, possibly acting to select and maintain forward models.