Module 6

Neural Implementation

The Bayesian-brain hypothesis predicts specific laminar and oscillatory signatures in cortex. Bastos 2012 integrated anatomical and electrophysiological evidence into a canonical microcircuit: superficial layers signal prediction errors in gamma frequencies, deep layers send predictions in alpha/beta. This module reviews the empirical case and its residual uncertainties.

1. The Canonical Microcircuit

Bastos 2012 (Neuron) integrates anatomy (Douglas & Martin 1991) with electrophysiology to propose a consistent mapping: L4 and L2/3 receive feedforward input and compute residuals against L5 predictions; L5 and L6 house the slower predictive units that project back to lower areas. Superficial layers exhibit gamma-band (40–80 Hz) synchrony during feedforward signalling; deep layers show alpha/beta (10–30 Hz) for feedback.

2. Mismatch Negativity

MMN is an auditory ERP component elicited by rare deviant tones embedded in streams of standard tones. Wacongne 2012 built a predictive-coding model of MMN that reproduces its latency, amplitude, and pharmacological modulation (NMDA antagonism reduces MMN, consistent with NMDA underlying precision-weighted prediction errors).

Simulation: Laminar Oscillations

Python

script.py27 lines

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

# Bastos 2012 canonical microcircuit: gamma (superficial, bottom-up)
# vs alpha/beta (deep, top-down)
t = np.linspace(0, 1.0, 5000)
gamma = np.sin(2*np.pi*60*t) * (1 + 0.2*np.sin(2*np.pi*2*t))
beta  = np.sin(2*np.pi*20*t) * (1 + 0.3*np.cos(2*np.pi*1*t))
alpha = np.sin(2*np.pi*10*t) * (1 + 0.2*np.sin(2*np.pi*1*t))

fig, ax = plt.subplots(figsize=(11, 6), facecolor='#0a0a1a')
ax.set_facecolor('#111827'); ax.tick_params(colors='#cbd5e1')
for s in ax.spines.values(): s.set_color('#334155')
ax.plot(t, gamma + 6, color='#fbbf24', lw=1, label='Gamma 60 Hz (L2/3 PE)')
ax.plot(t, beta + 3, color='#60a5fa', lw=1, label='Beta 20 Hz (L5/6 predictions)')
ax.plot(t, alpha, color='#a78bfa', lw=1, label='Alpha 10 Hz (L5/6 expectations)')
ax.set_xlim(0, 0.5)
ax.set_xlabel('Time (s)', color='#cbd5e1')
ax.set_ylabel('Simulated LFP amplitude', color='#cbd5e1')
ax.set_title('Bastos 2012 laminar oscillations',
             color='#c4b5fd', fontweight='bold')
ax.legend(facecolor='#1e293b', edgecolor='#334155', labelcolor='#cbd5e1')
plt.tight_layout()
plt.savefig('output.png', dpi=120, bbox_inches='tight', facecolor='#0a0a1a')
print('Superficial gamma: feedforward prediction errors')
print('Deep alpha/beta: feedback predictions/expectations')

Click Run to execute the Python code

Code will be executed with Python 3 on the server

3. Open Questions

The neural implementation of Bayesian brain theories has several unresolved questions: are prediction errors literally separate neurons, or are they encoded in firing-rate deviations? How is precision-weighting implemented — through synaptic gain, neuromodulatory tone (ACh, NE), or dendritic compartmentalisation? How do canonical microcircuits implement non-linear generative models? Active empirical programmes are testing each.

Key References

• Bastos, A. M. et al. (2012). “Canonical microcircuits for predictive coding.” Neuron, 76, 695–711.

• Douglas, R. J. & Martin, K. A. C. (1991). “A functional microcircuit for cat visual cortex.” J. Physiol., 440, 735–769.

• Wacongne, C. et al. (2012). “A neuronal model of predictive coding accounting for the mismatch negativity.” J. Neurosci., 32, 3665–3678.

• Arnal, L. H. & Giraud, A.-L. (2012). “Cortical oscillations and sensory predictions.” Trends Cogn. Sci., 16, 390–398.

Share:X Reddit LinkedIn

← Module 5 Module 7: Psychiatry →