Griffin, Nick; Schurger, Aaron and Garagnani, M.. 2025. 'Spontaneous emergence of slow ramping prior to decision states in a brain-constrained model of fronto-temporal cortical areas'. In: 34th Annual Computational Neuroscience Meeting (CNS 2025). Florence, Italy 5-9 July 2025. [Conference or Workshop Item] (Forthcoming)

No full text available

Text ShortAbstract.pdf - Accepted Version Permissions: Administrator Access Only until 5 July 2025. Download (59kB)

Abstract or Description

INTRODUCTION --
Currently an ongoing debate exists over two prevailing interpretations of the pre-movement ramping neural signal known as the readiness potential (RP): the “early-” and “late-decision” accounts [1]. The former holds that the RP reflects planning and preparation for movement – a decision outcome. The latter holds that it is pre-decisional, emerging because a commitment is made only after activity reaches a threshold. We used a fully brain-constrained neural-network model of six human frontotemporal areas to investigate this issue and the cortical mechanisms underlying the emergence of the RP and spontaneous decisions to act, extending the previous study that developed this neural architecture [2].

METHODS --
The network was trained via neurobiologically realistic learning mechanisms to induce formation of distributed perception-action cell assembly (CA) circuits. To replicate the experimental settings used to trigger the spontaneous emergence of volitional actions, we repeatedly reset its activity (trial start) and collected the resulting “WTs” (“wait time”: time steps elapsed between trial start and first spontaneous CA ignition) in absence of external stimulation, with neural activity driven only by uniform white noise. We then compared model and human data at both “behavioural” (WT distribution) and “neural activity” (RP index) level, where the simulated RP was defined simply as the total firing activity within the network’s model neurons.

RESULTS --
We found that, for select values of the parameters, the simulated WT distribution was statistically indistinguishable from the experimentally measured one. This result was replicated in eight out of ten repeated experiments, the variability being attributed to the noise inherently present in the network. We also found that the simulated RP, displaying the characteristic non-linear buildup, could be fitted to the experimental RP, with a mean square error that was minimal for the parameter set that produced the best-fitting simulated WT distribution. Finally, but importantly, individual trials also revealed sub-threshold fluctuations in CA activity insufficient by themselves for full ignition.

DISCUSSION --
We used a 6-area deep, brain-constrained model of frontotemporal cortical areas to simulate neural and behavioural indexes of the spontaneous emergence of simple, spontaneous decisions to act. The noise-driven spontaneous reverberation of activity within CA circuits and their subsequent ignition were taken as model correlates of the emergence of “free” volitional action intentions and conscious decisions to move, respectively. Replicating both behavioural and brain indexes of spontaneous voluntary movements, the present computational architecture and simulation results offer a neuro-mechanistic explanation for the emergence of endogenous decisions to act in the human brain, providing further support for a late, stochastic account of the RP.

Item Type:	Conference or Workshop Item (Poster)
Keywords:	voluntary action, readiness potential, neurobiologically realistic modelling, deep network, stochastic account, spontaneous thoughts, free will
Related URLs:	https://www.cnsorg.org/abstractsannual-m... https://link.springer.com/journal/10827
Departments, Centres and Research Units:	Computing
Dates:	Date Event 7 April 2025 Accepted
Event Location:	Florence, Italy
Date range:	5-9 July 2025
Item ID:	38859
Date Deposited:	27 May 2025 08:44
Last Modified:	27 May 2025 08:44
URI:	https://research.gold.ac.uk/id/eprint/38859

View statistics for this item...

Edit Record (login required)