Questions: Cerebellum: Motor Learning and Coordination

Question 1

A patient with cerebellar damage tries to pick up a glass but repeatedly overshoots, correcting only after touching the table. Their arm muscles are not weak and they feel normal sensation. What is the most likely explanation?

• A. The motor cortex cannot generate movement plans without cerebellar input, so movements are initiated randomly
• B. Sensory feedback from the arm cannot reach the brain because cerebellar damage disrupts ascending pathways
• C. The cerebellum can no longer apply error corrections to motor commands, so movements are inaccurate and fail to improve with practice
• D. Dopamine reward signals required for motor learning are disrupted by cerebellar damage

Answer: (C)

Cerebellar damage does not cause paralysis or sensory loss — the motor cortex still initiates movements and spinal circuits still execute them. What is lost is the cerebellum's error-correction function. The patient can feel the error (intact sensation) and understand what happened (intact cognition), but the neural machinery that detects the error and adjusts future motor commands is damaged. Crucially, the overshoot also fails to improve with practice — the learning algorithm itself is broken. This distinguishes cerebellar ataxia from weakness (which impairs force production) or sensory loss (which impairs feedback).

Question 2

Which event triggers long-term depression (LTD) at a parallel fiber-Purkinje cell synapse in the cerebellum?

• A. Repeated parallel fiber firing alone, without any climbing fiber input
• B. Parallel fiber activity occurring simultaneously with climbing fiber firing
• C. The absence of climbing fiber input during repeated parallel fiber activity
• D. High-frequency firing of the Purkinje cell itself during movement

Answer: (B)

The coincidence rule is the key to cerebellar learning: LTD occurs specifically when a parallel fiber is active at the same time the climbing fiber fires (signaling error). Parallel fiber firing alone (without error) does not produce LTD — it may even produce LTP. The climbing fiber is the teacher; the parallel fibers are the inputs. Only the input pathways active during an error get weakened. This specificity is what allows the cerebellum to target and correct only the motor commands that contributed to the mistake, leaving correct patterns intact.

Question 3

Damage to the cerebellum causes paralysis because the cerebellum is required to generate the voluntary movement commands that initiate limb movements.

• T. True
• F. False

Answer: False

The cerebellum does not initiate voluntary movements — that is the job of the motor cortex and associated structures. Cerebellar damage causes ataxia: movements are poorly coordinated, inaccurate, and jerky, but the patient can still move. The classic symptoms include dysmetria (overshooting targets), intention tremor (tremor during movement, not at rest), and dysdiadochokinesia (inability to perform rapid alternating movements). The muscles work, the intent is there, but the fine-tuning and error-correction system is gone.

Question 4

The climbing fiber from the inferior olive functions as a teaching signal in cerebellar learning by firing specifically when a movement error has occurred.

• T. True
• F. False

Answer: True

This is the core of the Marr-Albus-Ito theory of cerebellar learning. The climbing fiber from the inferior olive fires in response to unexpected or erroneous movement outcomes — it is the error signal. When the climbing fiber fires, it produces a powerful complex spike in the Purkinje cell. Any parallel fiber synapses that were active just before or during this error signal undergo LTD — they are selectively weakened because they were 'active during the error.' The climbing fiber provides the supervised learning teaching signal; the parallel fibers provide the input representation.

Question 5

In what sense is the cerebellum described as a 'supervised learning machine'? Identify what plays the role of the input, the teaching signal, and the synaptic update rule.

Model answer: The cerebellum implements supervised learning: parallel fibers provide the input (current motor command and sensory context), the climbing fiber provides the teaching signal (error — fires when movement goes wrong), and LTD at active parallel fiber-Purkinje cell synapses is the update rule (weakens the synapses that were active during the error, adjusting future output to reduce the mistake).

The analogy to machine learning is precise: the parallel fibers are the input layer carrying a high-dimensional representation of motor context (each Purkinje cell receives ~200,000 parallel fiber inputs). The climbing fiber is the supervisor, providing a binary error signal. LTD implements gradient descent by selectively weakening the weights (synaptic strengths) of connections that were active during erroneous outputs. Unlike unsupervised learning (which needs no teacher) or reinforcement learning (which uses reward signals), the cerebellum has a clear, dedicated error-signal channel — making it supervised in the strict sense.