Backpropagation is biologically plausible in neural networks
While a minority of recent theoretical models propose that cortical microcircuits might approximate error propagation, the overwhelming consensus in neuroscience and machine learning literature is that standard backpropagation is biologically implausible due to requirements like weight transport and global error signals.
The retrieved literature heavily emphasizes that standard backpropagation is biologically implausible, driving researchers to develop alternative local learning rules (e.g., papers 0 through 9). Although papers 10 and 11 propose models attempting to realize error backpropagation in cortical microcircuits, the vast majority of the provided papers explicitly state that backpropagation lacks biological plausibility and is not used by the brain. Therefore, the claim is evaluated as refuted or at best highly contested, with the weight of evidence leaning against biological plausibility.
Max K, Jaras I, Granier A, Wilmes KA, Petrovici MA. ‘Backpropagation and the brain’ realized in cortical error neuron microcircuits. 2025. https://doi.org/10.1101/2025.07.11.664263
Paper [10] presents a cortical microcircuit model arguing that the brain implements a form of error backpropagation.
B. Illing, W. Gerstner, Johanni Brea. Biologically plausible deep learning - but how far can we go with shallow networks?. 2019. https://doi.org/10.1016/j.neunet.2019.06.001
Paper [0] states that error backpropagation is considered biologically implausible, prompting research into local learning rules.
See more details
Max K, Jaras I, Granier A, Wilmes KA, Petrovici MA. 'Backpropagation and the brain' realized in cortical error neuron microcircuits.. 2026. https://doi.org/10.1371/journal.pcbi.1014164
Paper [11] similarly demonstrates a biologically motivated circuit model approximating error backpropagation.
Changze Lv, Jingwen Xu, Yiyang Lu, Xiaohua Wang, Zhenghua Wang, Zhibo Xu, Di Yu, Xin Du, Xiaoqing Zheng, Xuanjing Huang. Dendritic Localized Learning: Toward Biologically Plausible Algorithm. 2025. https://doi.org/10.48550/arXiv.2501.09976
Paper [1] highlights that backpropagation's biological plausibility is challenged by weight symmetry, global error signals, and dual-phase training.
Masataka Konishi, Kei M. Igarashi, Keiji Miura. Biologically plausible local synaptic learning rules robustly implement deep supervised learning. 2023. https://doi.org/10.3389/fnins.2023.1160899
Paper [2] notes that prevailing backpropagation learning rules are not necessarily biologically plausible and cannot be implemented in the brain.
Chia-Hsiang Kao, B. Hariharan. Counter-Current Learning: A Biologically Plausible Dual Network Approach for Deep Learning. 2024. https://doi.org/10.48550/arXiv.2409.19841
Paper [3] indicates that error backpropagation faces criticism for a lack of biological plausibility due to issues like the backward locking problem.
Q. Liao, Liu Ziyin, Yulu Gan, Brian Cheung, Mark T. Harnett, T. Poggio. Self-Assembly of a Biologically Plausible Learning Circuit. 2024. https://doi.org/10.48550/arXiv.2412.20018
Paper [4] confirms the neuroscience consensus that backpropagation is unlikely to be used by the brain.
David Rotermund, Klaus R. Pawelzik. Biologically plausible learning in a deep recurrent spiking network. 2019. https://doi.org/10.1101/613471
Paper [5] contrasts global optimization methods like backpropagation with local synaptic mechanisms used in biological brains.
M. Nakajima, Katsuma Inoue, Kenji Tanaka, Yasuo Kuniyoshi, Toshikazu Hashimoto, K. Nakajima. Physical Deep Learning with Biologically Plausible Training Method. 2022. https://doi.org/10.48550/arXiv.2204.13991
Paper [6] notes that neural network training still relies on methods like backpropagation that are optimized for digital rather than biological processing.
Anirudh Apparaju, Ognjen Arandjelovíc. Towards New Generation, Biologically Plausible Deep Neural Network Learning. 2022. https://doi.org/10.3390/sci4040046
Paper [7] describes end-to-end backpropagation as presenting a major limitation when contrasted with biological neural network processes.
Zhang C, Rolls ET, Feng J. Invariant visual object and face learning in the ventral cortical visual pathway: A biologically plausible model.. 2026. https://doi.org/10.1371/journal.pcbi.1013959
Paper [8] develops a biologically plausible model contrasting with artificial networks that do not use local synaptic learning rules.
Galloni AR, Peddada A, Chennawar Y, Milstein AD. Cellular and subcellular specialization enables biology-constrained deep learning.. 2026. https://doi.org/10.1016/j.celrep.2026.117159
Paper [9] points out that most modern artificial neural network algorithms are not compatible with fundamental principles of neuroscience.
The paper trail · every fact has a biography
Challenge the receipt
Citation formatting by citeproc-js (Frank Bennett) and the Citation Style Language project. Source and licenses.
Terms · Privacy · How verdicts work · Dispute this receipt