It would be a mistake to think that the dynamical-systems account argues M1 activity has nothing to do with muscles or forces.

Imagine a neural population that follows a complex trajectory through state space during a reach.

Its activity might contain components that directly relate to the muscle commands being generated.

Other components might help organize those commands over time without being directly reflected in muscle activity at every instant.

In other words, the entire population trajectory can have a dynamical organization while certain dimensions of that activity provide effective motor output.

Because generating a movement requires muscle commands that change over time.

For example, during a reach, muscles may need to accelerate the arm, regulate its trajectory, and then decelerate it near the target. These operations require different patterns of muscle activity at different moments.

A recurrent neural network can produce structured sequences of activity through its own internal interactions.

In that sense, the dynamics of the population may help generate the temporally organized motor commands needed for a movement.

A recurrent neural network can produce structured sequences of activity through its own internal interactions.

Suppose researchers can reliably predict the direction of a monkey's reach from M1 activity.

It's tempting to conclude:

M1 represents movement direction, and downstream motor circuits must read this directional representation to generate the appropriate muscle commands.

But this conclusion does not necessarily follow.

Why? Because different movement variables are correlated. Example: Reaching to the right When a monkey reaches toward a target on its right, several things happen together: 1. The target occupies a particular spatial location. 2. The monkey produces particular patterns of muscle activation. 3. The arm generates particular forces. 4. The hand moves along a particular trajectory. 5. The motor cortical population follows particular patterns of neural activity. These variables all systematically covary with the direction of the reach.

Therefore, if M1 activity primarily reflects muscle activation or the dynamics of movement generation, a researcher may still be able to decode movement direction from it.

The fact that neural activity contains information about direction does not necessarily mean that representing direction is its primary computational function.

Differentiate between sensory feedback caused by its own movements and sensory changes that were not predicted from those movements

Motor cortical representations overlap, and the activity of M1 neurons relates to multiple movement variables. There's no "point" in M1 that corresponds to specific body parts.

The changing population activity can help generate the precisely timed muscle activation needed for the reach.

This is different from imagining that each neuron independently represents a fixed variable, such as one muscle or one movement direction.

one neuron's firing might be related to several of these variables simultaneously.

The experiment demonstrates that stimulating the site can cause a hand-to-mouth movement, but it does not establish that the site's normal function is to represent that entire action.

This is a general issue in neuroscience: what stimulation can make a circuit do is not necessarily what that circuit normally encodes or computes.

For example, stimulating a neural population might cause several connected circuits to become active together in a way that does not occur during natural movement.

Traditionally, researchers often used brief electrical stimulation to study the motor cortex.

For example, stimulating one location in M1 might produce a movement of the fingers, whereas stimulating another location might produce movement at the elbow.

These observations contributed to the familiar motor homunculus: different regions of M1 are broadly associated with different body parts.

But what happens if researchers stimulate for longer?

With long-train stimulation, rather than delivering a brief pulse or short train, researchers apply electrical stimulation over a longer period, sometimes on the order of hundreds of milliseconds.

Under these conditions, stimulating certain cortical sites can produce coordinated movements involving several joints.

The crucial insight is that descending pathways normally provide both excitatory and inhibitory influences on spinal circuits.

Even when you are not actively moving, neurons in the spinal cord receive ongoing descending influences that help maintain their excitability.

When those descending connections are suddenly interrupted, the spinal cord loses not only inhibitory control but also tonic excitatory drive.

Consequently, many spinal neurons become less responsive, and reflex activity is temporarily depressed.

This explains why severing descending pathways does not immediately produce exaggerated reflexes.

Higher motor centers do not need to generate the detailed pattern of muscle activation themselves. Instead, they can control movement by adjusting the activity of lower-level circuits that already have the machinery to generate those patterns.

Spinal circuits receive and combine proprioceptive information (limb position and muscle state), load-related information (whether the limb is supporting weight), and cutaneous information (contact with the obstacle).

The same sensory stimulus can have a different effect during stance, when the limb needs to support the animal's weight

Organizing which muscles and limbs become active at different points in that rhythm.

Rhythm generation answers: How does the nervous system produce repeated, alternating activity without receiving a separate command for every phase?

But notice that generating this alternation is not sufficient for walking. The cat needs to coordinate many muscles and multiple limbs.

In reality, spinal reflex circuits are embedded within larger motor-control networks. Their responses can be modified by ongoing locomotor activity, sensory feedback, and descending control.

The reflex is therefore context-dependent: the spinal network transforms sensory input differently depending on what the ongoing behavior requires.

In control systems, gain describes how strongly an input influences an output.

During swing: - The foot is moving forward through the air. - The touch might indicate that the foot is encountering an obstacle. - The spinal circuitry can respond by increasing flexion, lifting the foot higher to clear the obstacle.

During stance: - The foot is supporting body weight. - Producing the same strong flexion response could destabilize the body. - The spinal circuitry may instead reduce, suppress, or otherwise alter the response to preserve support.

Without proprioceptive feedback, spinal circuits lose information about the actual position and mechanical state of the limbs.

cut the connection between brain and spinal cord

Fictive locomotion demonstrates that the nervous system can generate alternating flexor–extensor motor commands even when the corresponding movements are absent. Therefore, the basic rhythm can be generated centrally rather than requiring each movement to trigger the next through sensory feedback.

Example: Reaching for a cup of water - Cortical systems: Use visual information to locate the cup and guide the reach toward it. - Basal ganglia: Help select and regulate the reaching action. - Cerebellum: Predict the movement’s sensory consequences and help calibrate its timing and accuracy. - Brainstem and spinal circuits: Coordinate posture, muscle activation, and rapid local adjustments. - Muscles and body: Execute the movement, generating continuous sensory feedback (e.g., proprioception and vision). - If the cup suddenly moves: Updated visual information allows cortical circuits to adjust the reach, while cerebellar and spinal circuits contribute to ongoing corrections. - Key principle: Motor control is hierarchical but recurrent. Higher systems influence lower controllers rather than specifying every muscle contraction, while feedback and predictive signals allow multiple levels to continuously adjust the movement.

1. The malleus and incus form a mechanical lever.

Because of their geometry, the movement of the tympanic membrane is transformed into a somewhat smaller displacement but greater force at the stapes.

This is the same general principle as using a lever to increase force at the expense of displacement.

1. P = F/A

sound travels through air, but the cochlea is filled with fluid. Transferring sound energy between these two media is inefficient unless the middle ear mechanically transforms the vibrations.

a neuron at one position in layer 1 and a neuron at a corresponding position in layer 2 represent approximately the same visual location, but their principal retinal inputs come from different eyes.

Similarly, neurons in magnocellular and parvocellular layers can represent that location while emphasizing different aspects of the visual signal.

For example, the left LGN represents the right visual hemifield, receiving information from: - The nasal retina of the right eye. - The temporal retina of the left eye.

There are two distinct types of comparison involved in the early processing of color:

First, comparison across time or adaptation state: How much has each cone class's excitation changed relative to its adapted background?

Second, comparison across cone classes: Are L, M, and S cones changing by similar proportions, or is one cone class changing more than the others?

The first produces the normalized cone contrasts. The second produces luminance-related and chromatic opponent signals.

because its receptive-field organization is approximately circular, it does not strongly prefer one edge orientation over another.

More precisely, when I say “light falls on its center,” I mean:

Light falls on the region of the retina corresponding to the center of that ganglion cell's receptive field.

It does not mean that light falls on the physical center of the ganglion cell's cell body or dendritic tree.

For example, suppose we're recording the firing of ganglion cell A: - Its receptive-field center is a particular region of the retina where light stimulation tends to increase its firing (assuming it's an ON-center cell). - Its receptive-field surround is a neighboring region where light stimulation tends to decrease its firing.

A ganglion cell doesn't simply measure illumination within one patch of the retina. Its response also depends on how that illumination compares with nearby regions.

A horizontal cell receives input from multiple neighboring photoreceptors.

For example, imagine three cones, A, B, and C. Cone B is connected to a bipolar cell that provides the main center signal. A horizontal cell also receives signals from A, B, and C.

The horizontal cell therefore carries information about illumination across a broader region, rather than from just one cone.

Instead of allowing only a narrow bundle of light to enter, the eye can admit light through a relatively large opening.

The lens refracts incoming light so that rays originating from the same external point converge toward approximately the same location on the retina.

Consequently, the eye can collect more photons without sacrificing as much spatial resolution.

As a result, different parts of the cup are preferentially illuminated by light coming from different directions.

For example: - Light coming from the left illuminates one pattern of receptors. - Light coming from the right illuminates a different pattern. - Light coming from directly in front produces another pattern. The deeper the cup, the more constrained the directions from which light can reach any particular receptor.

Pigment absorbs incoming light, preventing it from reaching receptors from certain directions.

This introduces directional shading

light arriving from different directions can stimulate the same photoreceptors

If the patch has no structures that provide directional information, it cannot reliably distinguish whether light came from the left, right, or directly above.

Consequently, light entering the center of the fovea travels through less intervening neural tissue before reaching the cones.

Notice that this doesn't mean the neurons serving those cones disappear. They are still present and connected to the cones, but many are positioned farther away from the central optical path.

Similar to biological optical fibers.

In the vertebrate retina, photoreceptors (rods and cones) are located toward the back of the retina, rather than at the front where light enters.

There is much more Na⁺ outside than inside → concentration gradient pushes Na⁺ inward.

The inside of the neuron is negative relative to outside → electrical gradient attracts positive Na⁺ inward.

To reach equilibrium, the electrical force must eventually point the other way. The inside therefore has to become positive enough that it repels Na⁺ outward → Equilibrium potential is positive

Suppose you contract a muscle quickly:

The whole muscle shortens.

If the spindle inside it simply shortened passively too, its central sensory region could go slack (the central region needs to be stretched so that the endings can be deformed to initiate the afferent signal to report rate at which muscle length is changing).

Then the sensory ending would no longer be stretched much, so its firing could drop.

Retina → LGN → V1 → higher visual cortex is primarily a modality-specific analysis pathway.

Retina → superior colliculus/tectum → downstream motor/attention systems is primarily a spatial selection-and-orienting pathway.

An area called “primary visual cortex,” “primary auditory cortex,” etc. is still mainly associated with one modality, but its activity can nevertheless be influenced by information from other senses, often through feedback or other cortical connections.

Early sensory cortex asks, “What is happening at this local patch?” The next stages increasingly ask, “What larger object, surface, sound source, or body configuration do these local signals jointly imply?”

But they are still modality-biased because those representations are still primarily visual, auditory, or somatosensory rather than fully abstract.

This example: "Cortisol can also act through receptors associated with the cell membrane, triggering intracellular signaling within seconds to minutes". Ligand binding activates intracellular signaling intermediaries

What this does is integration. What SON and PVN does is to send signals for hormone secretion and motivated behavior

Magnocellular (M) Pathway: Specialized for fast movement, depth perception, and large-scale contrast. It uses large cells with fast signal transmission, but it is colorblind and has low spatial resolution.

Parvocellular (P) Pathway: Specialized for high-resolution detail, color vision (such as red/green), and pattern recognition. It transmits signals more slowly using smaller cells and works best with stationary or slow-moving objects.

See chapter 6

Enables BOTH rapid autonomic reflexes AND higher-order interoception, homeostatic regulation and motivated behavior

That's why the nucleus ambiguus is located on the ventral plate.

Historically, anatomists often divided the brain according to visible anatomy—shape, location, cell appearance, boundaries seen under the microscope, and so on. Evo-devo gives scientists another kind of evidence: which developmental regulatory genes are expressed in which regions.

These are different ways allometry can manifest