Module 7

Neuroscience & Behavior

Sleep architecture, vestibular precision, predatory neural circuits, and social cognition in the feline brain

1. Sleep Architecture

Domestic cats are among the most prolific sleepers in the mammalian world, spending 12–16 hours per day asleep. Unlike human monophasic sleep, feline sleep is polyphasic — distributed across multiple bouts throughout the 24-hour cycle. This pattern reflects their evolutionary heritage as ambush predators: brief bursts of high-energy activity punctuated by extended rest periods to conserve metabolic resources.

Sleep Stages & EEG Signatures

Feline sleep cycles last approximately 104 minutes (compared to ~90 minutes in humans) and comprise two main stages:

NREM (Slow-Wave Sleep)

Approximately 70% of total sleep. Characterised by high-amplitude, low-frequency delta waves (0.5–4 Hz) on EEG. Muscle tone reduced but not abolished. Heart rate and respiration slow. Critical for physical restoration and immune function. Cats often sleep in a curled posture during NREM to minimise heat loss.

REM (Paradoxical Sleep)

Approximately 30% of total sleep (higher than human ~25%). Characterised by low-amplitude, high-frequency theta waves (4–8 Hz). Complete muscle atonia (except eyes and diaphragm). Rapid eye movements, whisker twitching, paw flexion. Jouvet (1962) first described REM in cats and demonstrated that lesioning the locus coeruleus removes atonia, causing cats to “act out” dreams — stalking and pouncing at invisible prey.

EEG Signal Decomposition via Fourier Transform

The electroencephalogram (EEG) signal \(x(t)\) recorded from feline cortex is a superposition of oscillatory components at different frequencies. The power spectral density (PSD) reveals which frequency bands dominate in each sleep stage:

\[\hat{X}(f) = \int_{-\infty}^{\infty} x(t) \, e^{-2\pi i f t} \, dt\]

The power spectral density is:

\[S(f) = |\hat{X}(f)|^2 = \hat{X}(f) \cdot \hat{X}^*(f)\]

For a discrete signal sampled at rate \(f_s\) with \(N\) samples, we use the discrete Fourier transform:

\[X_k = \sum_{n=0}^{N-1} x_n \, e^{-2\pi i k n / N}, \quad k = 0, 1, \ldots, N-1\]

In practice, the PSD is computed using Welch's method: the signal is divided into overlapping segments, each windowed (e.g., Hann window) and FFT-transformed, then the resulting periodograms are averaged. By Parseval's theorem, the total power in the time domain equals the integrated PSD:

\[\sum_{n=0}^{N-1} |x_n|^2 = \frac{1}{N}\sum_{k=0}^{N-1} |X_k|^2\]

EEG Frequency Bands in Cats

• Delta (0.5–4 Hz): Dominant in NREM. Amplitude 50–200 \(\mu\)V
• Theta (4–8 Hz): Dominant in REM, also prominent during alert stalking
• Alpha (8–13 Hz): Relaxed wakefulness, eyes half-closed
• Beta (13–30 Hz): Active wakefulness, visual tracking
• Gamma (30–100 Hz): Predatory attention, sensory binding

2. Vestibular System & Balance

The extraordinary balance of cats — walking along narrow fences, landing on their feet from any orientation — depends on an exquisitely tuned vestibular system. The three semicircular canals in each inner ear detect angular acceleration in three orthogonal planes, while the otolith organs (utricle and saccule) detect linear acceleration and gravity.

The Torsion Pendulum Model

Each semicircular canal can be modelled as a torsion pendulum. The endolymph fluid within the canal has angular displacement \(\theta\) relative to the canal wall. When the head rotates, the fluid lags due to inertia, deflecting the cupula (a gelatinous membrane) and bending the embedded hair cells:

\[I \frac{d^2\theta}{dt^2} + b \frac{d\theta}{dt} + k\theta = \tau(t)\]

where:

• \(I\) is the moment of inertia of the endolymph fluid ring
• \(b\) is the viscous damping coefficient (fluid friction against canal walls)
• \(k\) is the cupula restoring stiffness
• \(\tau(t)\) is the applied torque from head rotation

The natural frequency and damping ratio are:

\[\omega_0 = \sqrt{\frac{k}{I}}, \quad \zeta = \frac{b}{2\sqrt{kI}}\]

For a step angular velocity input \(\Omega\) (sudden head turn), the torque is \(\tau = I \dot{\Omega} \delta(t)\). The transfer function in the Laplace domain is:

\[H(s) = \frac{\Theta(s)}{T(s)} = \frac{1}{Is^2 + bs + k} = \frac{1/I}{s^2 + 2\zeta\omega_0 s + \omega_0^2}\]

Feline vs Human Vestibular Parameters

• Canal natural frequency: Cat ~5 Hz vs Human ~0.2 Hz
• Damping ratio: Cat \(\zeta \approx 5\text{--}8\) (overdamped) vs Human \(\zeta \approx 3\text{--}5\)
• Time constant \(\tau_1 = b/k\): Cat ~4 s vs Human ~5–7 s
• Canal radius: Cat ~2.5 mm vs Human ~3.2 mm (but higher sensitivity per unit volume)

The heavily overdamped feline semicircular canal acts as an angular velocity sensor in its working range: when \(\zeta \gg 1\), the cupula deflection is proportional to angular velocity rather than angular displacement. The high natural frequency means the cat's vestibular system responds rapidly to sudden perturbations — essential for the righting reflex (Module 1), which requires vestibular input to initiate within ~100 ms of the onset of a fall.

Vestibulo-Ocular Reflex (VOR)

The VOR stabilises the retinal image during head movement by generating compensatory eye rotations equal and opposite to head rotation. The gain of the feline VOR is:

\[G_{\text{VOR}} = \frac{\dot{\theta}_{\text{eye}}}{\dot{\theta}_{\text{head}}} \approx -0.95 \text{ to } -1.0\]

This near-unity gain means the eyes compensate almost perfectly for head movements, crucial for maintaining visual acuity during rapid head movements and the final phase of a predatory pounce.

3. Predatory Neural Circuits

The feline predatory sequence is a stereotyped motor programme comprising five phases: orient → stalk → pounce → grab → kill bite. Each phase is governed by distinct neural circuits, with the sequence orchestrated by the hypothalamic predatory attack area (PAA) and modulated by cortical input.

Orient Phase

Superior colliculus generates saccadic eye movements and head turns toward prey. Visual motion detected in area 17/18 of visual cortex feeds into the tectum. Auditory localisation via inferior colliculus also contributes. Latency: 50–80 ms from stimulus to orientation.

Stalk Phase

Slow, low-profile approach controlled by the cerebellum (coordination) and basal ganglia (movement sequencing). Pupils dilate maximally. Hindlimb muscles load elastically. Respiratory rate decreases to reduce motion detection by prey.

Pounce Phase

Ballistic launch from crouched position. Motor cortex sends the go signal to spinal pattern generators. Hindlimb extensors fire in coordinated burst. Once initiated, the pounce is largely pre-programmed (open-loop ballistic trajectory) with limited mid-flight correction.

Grab & Kill Bite

Forepaw grab controlled by proprioceptive feedback. The kill bite targets the nape of the neck, guided by mechanoreceptors in the canine teeth that detect the gap between cervical vertebrae. The trigeminal nerve provides precise bite-force modulation. Hypothalamic PAA drives bite execution.

Optimal Pounce Ballistics

The pounce can be modelled as projectile motion. For a cat launching at velocity \(v_0\)and angle \(\alpha\) from horizontal, the range is:

\[R = \frac{v_0^2 \sin(2\alpha)}{g}\]

Maximum range occurs at \(\alpha = 45°\). However, cats rarely launch at 45° because prey can move during the flight time. The optimal angle must account for prey evasion probability, which increases with flight duration \(t_f\):

\[t_f = \frac{2v_0 \sin\alpha}{g}\]

Modelling success probability as a decreasing function of flight time and an increasing function of range matching:

\[P_{\text{success}}(\alpha, d) = \exp\!\left(-\frac{(R(\alpha) - d)^2}{2\sigma_R^2}\right) \cdot \exp\!\left(-\frac{t_f(\alpha)}{\tau_{\text{evade}}}\right)\]

where \(d\) is the distance to prey, \(\sigma_R\) is the positional uncertainty, and \(\tau_{\text{evade}}\) is the prey evasion time constant. This analysis shows the optimal launch angle is typically 30–38° rather than 45°, consistent with high-speed video observations of hunting cats.

Energy Budget of Predation

Cats are ambush predators with a success rate of approximately 30–50% per pounce attempt (higher for indoor/barn cats hunting mice, lower for feral cats hunting birds). The energy economics are:

\[E_{\text{net}} = P_{\text{success}} \cdot E_{\text{prey}} - E_{\text{stalk}} - E_{\text{pounce}} - (1 - P_{\text{success}}) \cdot E_{\text{recovery}}\]

A typical mouse provides ~125 kJ. A pounce costs ~2–5 kJ. An extended stalk may cost 10–20 kJ. With a 40% success rate, the expected energy gain per attempt is:

\[E_{\text{net}} \approx 0.4 \times 125 - 15 - 4 - 0.6 \times 3 \approx 29.2 \text{ kJ per attempt}\]

This positive energy balance explains why the sit-and-wait ambush strategy is viable, and why cats spend so much time resting — they need only a few successful hunts per day to meet their ~1,000 kJ daily energy requirement.

4. Social Cognition

Despite their reputation as solitary creatures, domestic cats display sophisticated social cognitive abilities that have been increasingly documented in controlled behavioural studies.

Object Permanence

Cats demonstrate Piaget Stage 6 object permanence — the highest level, involving invisible displacement. In experiments by Dore & Dumas (1987), cats correctly searched for a toy hidden in a container that was then moved behind a screen and emptied. The cat inferred the toy's final location without seeing the transfer — a feat also achieved by great apes but not by dogs in most studies.

The Slow Blink

Humphrey et al. (2020) demonstrated that slow-blink sequences function as an affiliative communication signal between cats and humans. In controlled experiments, cats were significantly more likely to approach a human who had slow-blinked at them compared to a neutral-expression control. fMRI-analogue studies suggest reduced amygdala activation in response to slow-blink stimuli, consistent with a threat-reduction signal.

Slow Blink Kinematics

A feline slow blink comprises: (1) half-close of eyelids over ~0.5 s, (2) hold for 0.5–1.5 s with eyes narrowed, (3) slow reopen over ~0.5 s. Total duration ~1.5–2.5 s. This contrasts with a threat stare (unblinking, direct gaze, dilated pupils) and a startle blink (~0.1 s, bilateral, full closure).

Vocal Adaptation to Humans

Adult feral cats rarely meow at each other. The meow is primarily a human-directed vocalisation that has been shaped by thousands of years of cohabitation. Nicastro & Owren (2003) showed that domestic cats produce meows with higher fundamental frequencies and more varied spectral content than wild felids.

McComb et al. (2009) discovered the “solicitation purr” — a purr embedded with a high-frequency cry component (300–600 Hz) that resembles a human infant's cry and is perceived as more urgent by human listeners. This represents a remarkable case of interspecific vocal manipulation: cats have evolved to exploit a perceptual bias in the human auditory system.

Spatial Memory & Mental Maps

Cats maintain detailed spatial maps of their territory. Studies show that cats can remember the location of hidden food after delays of up to 16 hours (Fiset & Dore, 2006), suggesting robust hippocampal-dependent spatial memory. The feline hippocampus is proportionally large relative to brain volume, consistent with the demands of maintaining large home ranges (up to 10 km\(^2\) for feral males).

5. Feline Brain Anatomy

The cat brain weighs approximately 30 g (0.9% of body mass) and contains an estimated 760 million cortical neurons — more than twice the ~530 million in the dog cortex (Jardim-Messeder et al., 2017). Below is a labelled sagittal view.

6. Simulations

Predatory Pounce Optimization & EEG Power Spectrum

This simulation produces two panels: (1) a contour map of pounce success probability as a function of launch angle and distance, identifying the optimal strategy, and (2) synthetic EEG power spectra for feline NREM vs REM sleep stages.

Python

script.py156 lines

import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec

# ================================================================
# Panel 1: Predatory Pounce Success Probability
# ================================================================
g = 9.81
v0 = 5.5  # typical launch velocity (m/s)
sigma_R = 0.3  # positional uncertainty (m)
tau_evade = 0.35  # prey evasion time constant (s)

angles = np.linspace(5, 80, 200)  # degrees
distances = np.linspace(0.3, 3.5, 200)  # metres
A, D = np.meshgrid(np.radians(angles), distances)

# Range and flight time
R = v0**2 * np.sin(2 * A) / g
t_f = 2 * v0 * np.sin(A) / g

# Success probability
P = np.exp(-(R - D)**2 / (2 * sigma_R**2)) * np.exp(-t_f / tau_evade)

# Find optimal angle for each distance
opt_angles = []
opt_probs = []
for i, d in enumerate(distances):
    j_max = np.argmax(P[i, :])
    opt_angles.append(angles[j_max])
    opt_probs.append(P[i, j_max])

# ================================================================
# Panel 2: Synthetic Feline EEG Power Spectra
# ================================================================
np.random.seed(123)
fs = 256  # sampling rate Hz
duration = 30  # seconds
t = np.arange(0, duration, 1/fs)
N_pts = len(t)

# NREM: dominated by delta (0.5-4 Hz)
nrem = (3.0 * np.sin(2 * np.pi * 1.5 * t) +
        2.0 * np.sin(2 * np.pi * 2.5 * t) +
        1.5 * np.sin(2 * np.pi * 0.8 * t) +
        0.5 * np.sin(2 * np.pi * 6 * t) +
        0.3 * np.random.randn(N_pts))

# REM: dominated by theta (4-8 Hz)
rem = (0.5 * np.sin(2 * np.pi * 2 * t) +
       2.5 * np.sin(2 * np.pi * 5.5 * t) +
       2.0 * np.sin(2 * np.pi * 6.5 * t) +
       1.5 * np.sin(2 * np.pi * 7 * t) +
       0.8 * np.sin(2 * np.pi * 12 * t) +
       0.4 * np.random.randn(N_pts))

# Compute PSD using Welch-like approach (manual for clarity)
from numpy.fft import rfft, rfftfreq

def compute_psd(signal, fs, nperseg=1024):
    step = nperseg // 2
    n_segs = (len(signal) - nperseg) // step + 1
    freqs = rfftfreq(nperseg, 1/fs)
    psd = np.zeros(len(freqs))
    window = np.hanning(nperseg)
    for i in range(n_segs):
        seg = signal[i*step:i*step+nperseg] * window
        fft_seg = rfft(seg)
        psd += np.abs(fft_seg)**2
    psd /= n_segs
    psd = 10 * np.log10(psd / np.max(psd) + 1e-12)
    return freqs, psd

freqs_nrem, psd_nrem = compute_psd(nrem, fs)
freqs_rem, psd_rem = compute_psd(rem, fs)

# ================================================================
# Plot
# ================================================================
fig = plt.figure(figsize=(14, 7))
fig.patch.set_facecolor('#0a0a1a')
gs = gridspec.GridSpec(1, 2, wspace=0.3)

# Panel 1: Pounce contour
ax1 = fig.add_subplot(gs[0])
ax1.set_facecolor('#0a0a1a')
cf = ax1.contourf(angles, distances, P, levels=20, cmap='inferno')
ax1.plot(opt_angles, distances, color='#22c55e', linewidth=2.5, linestyle='--',
         label='Optimal angle')
cbar = plt.colorbar(cf, ax=ax1, fraction=0.046, pad=0.04)
cbar.ax.tick_params(colors='#6b7280', labelsize=7)
cbar.set_label('Success probability', color='#9ca3af', fontsize=9)

ax1.set_title('Predatory Pounce: Success Probability', color='#fbbf24',
              fontsize=12, fontweight='bold', pad=10)
ax1.set_xlabel('Launch angle (degrees)', color='#9ca3af', fontsize=10)
ax1.set_ylabel('Distance to prey (m)', color='#9ca3af', fontsize=10)
ax1.tick_params(colors='#6b7280', labelsize=8)
ax1.legend(fontsize=9, loc='upper right', facecolor='#1a1a2e',
           edgecolor='#334155', labelcolor='#d1d5db')
ax1.spines['bottom'].set_color('#334155')
ax1.spines['left'].set_color('#334155')
ax1.spines['top'].set_visible(False)
ax1.spines['right'].set_visible(False)

# Annotate optimal zone
ax1.annotate('Optimal zone:\n30-38 degrees', xy=(34, 1.5), fontsize=9,
             color='#22c55e', ha='center',
             bbox=dict(boxstyle='round,pad=0.3', facecolor='#0a0a1a',
                       edgecolor='#22c55e', alpha=0.8))

# Panel 2: EEG spectra
ax2 = fig.add_subplot(gs[1])
ax2.set_facecolor('#0a0a1a')

mask = freqs_nrem <= 40  # show up to 40 Hz
ax2.plot(freqs_nrem[mask], psd_nrem[mask], color='#60a5fa', linewidth=2,
         label='NREM (slow-wave)', alpha=0.9)
ax2.plot(freqs_rem[mask], psd_rem[mask], color='#f59e0b', linewidth=2,
         label='REM (paradoxical)', alpha=0.9)

# Band shading
ax2.axvspan(0.5, 4, alpha=0.12, color='#60a5fa')
ax2.axvspan(4, 8, alpha=0.12, color='#f59e0b')
ax2.axvspan(8, 13, alpha=0.08, color='#a78bfa')

ax2.text(2.25, -5, 'Delta', color='#60a5fa', fontsize=9, ha='center', fontweight='bold')
ax2.text(6, -5, 'Theta', color='#f59e0b', fontsize=9, ha='center', fontweight='bold')
ax2.text(10.5, -5, 'Alpha', color='#a78bfa', fontsize=9, ha='center', fontweight='bold')

ax2.set_title('Feline EEG Power Spectra by Sleep Stage', color='#fbbf24',
              fontsize=12, fontweight='bold', pad=10)
ax2.set_xlabel('Frequency (Hz)', color='#9ca3af', fontsize=10)
ax2.set_ylabel('Power (dB, normalized)', color='#9ca3af', fontsize=10)
ax2.set_xlim(0, 40)
ax2.legend(fontsize=9, loc='upper right', facecolor='#1a1a2e',
           edgecolor='#334155', labelcolor='#d1d5db')
ax2.tick_params(colors='#6b7280', labelsize=8)
ax2.spines['bottom'].set_color('#334155')
ax2.spines['left'].set_color('#334155')
ax2.spines['top'].set_visible(False)
ax2.spines['right'].set_visible(False)
ax2.grid(True, alpha=0.15, color='#475569')

fig.suptitle('Feline Neuroscience: Predatory Ballistics & Sleep EEG Analysis',
             color='#f59e0b', fontsize=14, fontweight='bold', y=0.98)

plt.savefig('output.png', dpi=150, bbox_inches='tight',
            facecolor='#0a0a1a', edgecolor='none')
print("Simulation complete.")
print("Left: Pounce success probability contour map.")
print("  Optimal launch angle is 30-38 degrees (not 45!) due to prey evasion time.")
print("Right: EEG power spectra showing delta dominance in NREM, theta in REM.")
print("  Cats spend ~30% of sleep in REM (higher than humans at ~25%).")

Click Run to execute the Python code

Code will be executed with Python 3 on the server

References

Jouvet, M. (1962). “Recherches sur les structures nerveuses et les mecanismes responsables des differentes phases du sommeil physiologique.” Archives Italiennes de Biologie, 100, 125–206.
Jardim-Messeder, D. et al. (2017). “Dogs have the most neurons, though not the largest brain: Trade-off between body mass and number of neurons in the cerebral cortex of large carnivoran species.” Frontiers in Neuroanatomy, 11, 118.
Humphrey, T., Proops, L., Forman, J., Sherwood, R., & McComb, K. (2020). “The role of cat eye narrowing movements in cat–human communication.” Scientific Reports, 10, 16503.
McComb, K., Taylor, A. M., Wilson, C., & Charlton, B. D. (2009). “The cry embedded within the purr.” Current Biology, 19(13), R507–R508.
Nicastro, N. & Owren, M. J. (2003). “Classification of domestic cat (Felis catus) vocalizations by naive and experienced human listeners.” Journal of Comparative Psychology, 117(1), 44–52.
Dore, F. Y. & Dumas, C. (1987). “Psychology of animal cognition: Piagetian studies.” Psychological Bulletin, 102(2), 219–233.
Fiset, S. & Dore, F. Y. (2006). “Duration of cats' (Felis catus) working memory for disappearing objects.” Animal Cognition, 9, 62–70.
Zepelin, H., Siegel, J. M., & Tobler, I. (2005). “Mammalian sleep.” In Principles and Practice of Sleep Medicine (4th ed.), pp. 91–100.
Wilson, D. E. & Mittermeier, R. A. (2009). Handbook of the Mammals of the World, Vol. 1: Carnivores. Lynx Edicions.
Bradshaw, J. W. S. (2013). Cat Sense: The Feline Enigma Revealed. Allen Lane.

← Module 6: Coat & Pigmentation Module 8: Evolution & Domestication →