The oneirogen hypothesis: modeling the hallucinatory effects of classical psychedelics in terms of replay-dependent plasticity mechanisms

Colin Bredenberg, Fabrice Normandin, Blake A. Richards, Guillaume Lajoie

Abstract

Classical psychedelics induce complex visual hallucinations in humans, generating percepts that are coherent at a low level, but which have surreal, dream-like qualities at a high level. While there are many hypotheses as to how classical psychedelics could induce these effects, there are no concrete mechanistic models that capture the variety of observed effects in humans, while remaining consistent with the known pharmacological effects of classical psychedelics on neural circuits. In this work, we propose the “oneirogen hypothesis”, which posits that the perceptual effects of classical psychedelics are a result of their pharmacological actions inducing neural activity states that truly are more similar to dream-like states. We simulate classical psychedelics’ effects via manipulating neural network models trained on perceptual tasks with the Wake-Sleep algorithm. This established machine learning algorithm leverages two activity phases: a perceptual phase (wake) where sensory inputs are encoded, and a generative phase (dream) where the network internally generates activity consistent with stimulus-evoked responses. We simulate the action of psychedelics by partially shifting the model to the ‘Sleep’ state, which entails a greater influence of top-down connections, in line with the impact of psychedelics on apical dendrites. The effects resulting from this manipulation capture a number of experimentally observed phenomena including the emergence of hallucinations, increases in stimulus-conditioned variability, and large increases in synaptic plasticity. We further provide a number of testable predictions which could be used to validate or invalidate our oneirogen hypothesis.

Introduction

Classical psychedelics, such as LSD and psilocybin, are known to primarily target the 5-HT2a serotonin receptor. Despite this pharmacological understanding, a significant explanatory gap exists between receptor activation and the complex, surreal visual hallucinations experienced by humans. Current theories like the "Entropic Brain" or "REBUS" offer high-level descriptions but lack a mechanistic link between cellular changes and system-wide perceptual shifts. This paper introduces the "oneirogen hypothesis," proposing that psychedelics induce hallucinations by shifting neural activity from a sensory-driven "wake" state toward a top-down, generative "dream" state.

Methods and Modeling Framework

The study utilizes the Wake-Sleep algorithm, an unsupervised learning method for artificial neural networks (ANNs) that operates in two phases: a Wake Phase driven by bottom-up sensory input (inference), and a Sleep Phase driven by top-down generative connections (replay/dreaming).

Figure 1: Schematic of the Wake-Sleep model architecture and the biological mapping of the state variable α to apical dendrite activity.

A state variable, α, is introduced to modulate this balance: α = 0 represents normal waking (bottom-up dominance), while α = 1 represents a dreaming state (top-down dominance). Biologically, this shift corresponds to the excitation of apical dendrites in pyramidal neurons by 5-HT2a agonists, which enhances top-down feedback signals. The model was tested on datasets including MNIST, CIFAR10, and high-resolution faces (FFHQ-256) using various architectures like VDVAE.

Key Results

Emergence of Hallucinations

As α increases, the network's internal "priors" begin to dominate sensory input, leading to structured distortions. In the stimulus-conditioned case, digits and images deform into surreal versions of themselves. In the "eyes-closed" condition (zero sensory input), the model generates structured patterns purely from its internal generative pathways, replicating the phenomenology of closed-eye visuals.

Figure 4: Simulated hallucinations in MNIST and CIFAR10 datasets under stimulus-conditioned and eyes-closed conditions as the psychedelic dose (α) increases.

Neural Variability and Plasticity

The model successfully replicates the increased "stimulus-conditioned variability" observed in fMRI data under psychedelics. Furthermore, it predicts a massive surge in synaptic plasticity, peaking at intermediate doses (α ≈ 0.5). This "aberrant" plasticity is driven by the mismatch between hallucinated activity and sensory expectations, potentially explaining how psychedelics help the brain escape "local minima" in conditions like depression.

Discussion and Future Directions

The oneirogen hypothesis provides a biologically plausible bridge between serotonin receptor activity and complex hallucinations. While the current model uses a simplified version of the Wake-Sleep algorithm, it offers testable predictions, such as the effect of silencing apical dendrites on neural variability. Future work aims to incorporate more complex biological constraints, such as Dale’s Law and temporal dynamics, to better model the full spectrum of psychedelic experiences.

Summary updated

Abstract

Abstract Classical psychedelics induce complex visual hallucinations in humans, generating percepts that are coherent at a low level, but which have surreal, dream-like qualities at a high level. While there are many hypotheses as to how classical psychedelics could induce these effects, there are no concrete mechanistic models that capture the variety of observed effects in humans, while remaining consistent with the known pharmacological effects of classical psychedelics on neural circuits. In this work, we propose the “oneirogen hypothesis”, which posits that the perceptual effects of classical psychedelics are a result of their pharmacological actions inducing neural activity states that truly are more similar to dream-like states. We simulate classical psychedelics’ effects via manipulating neural network models trained on perceptual tasks with the Wake-Sleep algorithm. This established machine learning algorithm leverages two activity phases: a perceptual phase (wake) where sensory inputs are encoded, and a generative phase (dream) where the network internally generates activity consistent with stimulus-evoked responses. We simulate the action of psychedelics by partially shifting the model to the ‘Sleep’ state, which entails a greater influence of top-down connections, in line with the impact of psychedelics on apical dendrites. The effects resulting from this manipulation capture a number of experimentally observed phenomena including the emergence of hallucinations, increases in stimulus-conditioned variability, and large increases in synaptic plasticity. We further provide a number of testable predictions which could be used to validate or invalidate our oneirogen hypothesis.
BibTeX
@misc{bredenberg2024oneirogen,
  author = {Colin Bredenberg and Fabrice Normandin and Blake A. Richards and Guillaume Lajoie},
  title = {{The oneirogen hypothesis: modeling the hallucinatory effects of classical psychedelics in terms of replay-dependent plasticity mechanisms}},
  year = {2024},
  doi = {10.1101/2024.09.27.615483},
  howpublished = {bioRxiv},
}
Figure