Spontaneous Neurons: Mechanisms & Implications

• Spontaneous neurons are characterized by generating action potentials without external inputs, driven by intrinsic stochastic synaptic and channel mechanisms.
• They play a crucial role in encoding information, synchronizing network activity, and stabilizing population codes in both biological tissues and artificial systems.
• Network topology and substrate properties modulate firing patterns, offering insights into energy-efficient and robust neuromorphic design.

Spontaneous neurons are defined by their ability to generate action potentials in the absence of explicit, time-varying external inputs. In both biological and artificial neural systems, such spontaneous firing emerges intrinsically—either from stochasticity in synaptic or channel mechanisms, from structured pacemaker currents, or from dynamical instabilities in network architecture. This behavior is not merely epiphenomenal: in diverse contexts, spontaneous activity underpins information encoding, synchronization, assembly formation, and homeostatic stability in both living neural tissue and neuromorphic hardware.

1. Mechanisms of Spontaneous Neuron Firing

In biological circuits, spontaneous neuronal firing arises from intrinsic noise sources such as stochastic synaptic vesicle release and channel-gating noise. In dissociated rat-cortical microcircuits, for example, the dominant contributors are stochastic vesicle fusion and channel fluctuations that can trigger suprathreshold depolarizations, causing irregular "background" spiking. This process operates even when no explicit external drive is present, with up to 80% of cortical energy consumption being devoted to background spiking that supports functions such as stochastic resonance and exploratory dynamics in learning (Hasani et al., 2017).

In artificially realized neuromorphic systems, spontaneous firing is modeled by explicitly injecting noise currents into silicon neurons engineered with subthreshold integrate-and-fire or Hodgkin–Huxley dynamics. In these devices, both white (flat spectral density) and pink ($1/f$ spectral density) noise are synthesized on-chip, typically by leveraging amplified resistor thermal noise, and added directly to the membrane equation: $$C_{m}\,\frac{dV_{m}}{dt} = -g_{L}(V_{m} - E_{L}) + I_{\text{noise}}(t) + I_{\text{syn}}(t),$$ allowing precise control over the rate and statistics of spontaneously generated spikes (Hasani et al., 2017).

2. Influence of Network Topology and Connectivity

The manifestation and correlation structure of spontaneous firing are strongly shaped by network connectivity. In both living microcircuits and corresponding silicon analogs, organization into distinct modules (e.g., "islands") with variable inter-module connectivity reveals the key role of cross-talk and synaptic fan-out.

Topology	Intra-island Correlation ($\rho$)	Inter-island Correlation ($\rho$)	# Connections Required
Isolated	0.8–0.9	~0	0
Single microtunnel	High	0.3–0.4	1
Triple microtunnel	0.7–0.8	0.7–0.8	3
CMOS, 8 point-to-point	0.8	0.3–0.4	8
CMOS, 3×8 multi-synaptic	0.8	0.7	3

In these systems, a single physical interneuron (microtunnel or multi-synaptic connection) can elicit substantial correlation between modules, but increasing the number of synaptic contacts per fiber (multi-contact) is markedly more efficient than merely raising the number of single point-to-point connections (Hasani et al., 2017). This suggests that biological networks likely exploit divergence in synaptic targeting to regulate the degree of spontaneous signal correlation across compartments.

3. Measurement and Quantification of Spontaneous Correlation

Synchrony and correlation in spontaneous firing are quantified using the Pearson coefficient ($\rho_{ij}$)—computed over binned spike counts in extended recordings. For a population of $N$ neurons, full correlation matrices are constructed, and intra-module and inter-module block averages are compared: $$\rho_{ij} = \frac{1}{N-1} \sum_{k=1}^N \bigg( \frac{x_i[k] - \mu_i}{\sigma_i} \bigg) \bigg( \frac{x_j[k] - \mu_j}{\sigma_j} \bigg).$$ (Hasani et al., 2017)

Empirically, in engineered scaffolds, modules connected by one or three microtunnels display inter-module correlation coefficients of ~0.4 and ~0.7–0.8, respectively. In artificial circuits, the degree of synchronization can be titrated by tuning the number and architecture of inter-module synapses.

4. Modulation of Spontaneous Activity via Substrate and Microenvironment

Spontaneous firing and correlated bursting in neural populations are not fixed properties but are sensitive to the cellular microenvironment. Cultures plated on ultrasoft PDMS substrates (elastic modulus ≈0.5 kPa), which approximate brain tissue, exhibit significant reduction in spontaneous EPSC amplitude (~–27%), frequency (~–35%), and particularly in cross-cell correlation ($\bar r$), compared to standard glass substrates. The suppression of hypersynchronous bursting on ultrasoft supports is mediated by decreased activation of stretch-activated cation channels (SACs), as demonstrated by the reduced dose of the SAC antagonist GsMTx-4 required for bursting suppression compared to stiffer substrates (Sumi et al., 2019). This mechanosensitivity underscores the importance of the physical cellular context in modulating spontaneous network-level activity and its translation to realistic models.

5. Functional and Computational Roles of Spontaneous Neurons

Spontaneous neuronal firing is not regarded as "noise" in the pejorative sense, but as substrate for important computational and homeostatic functions:

• Combinatorial exploration: Stochastic spiking supports network exploration of state space, enhancing the network's ability to encode new patterns or transition between attractor states.
• Stabilization of representations: Baseline spontaneous activity can stabilize population codes, especially in regimes subject to strong sparsification or dynamic pruning—an insight extended to artificial systems in the development of "spontaneous neurons" for LLMs (Xu et al., 14 Dec 2025).
• Energy-efficient computation: In neuromorphic systems, low-power subthreshold firing driven by engineered noise sources and flexible synaptic architectures facilitates energy-efficient, noise-driven computation, leveraging principles observed in biological microcircuits (Hasani et al., 2017).
• Graded control of correlation: Both connection multiplicity and noise amplitude can be titrated to modulate the degree of synchrony and correlation, providing a design axis for both biological and artificial architectures.

6. Implications for Neuromorphic Engineering and Network Design

The comparative analysis of spontaneous activity in biological and CMOS neural microcircuits demonstrates that:

• Compact white/pink noise generators can engender realistic, distributed tonic firing in silicon architectures.
• Control over the number of synaptic contacts per interneuron provides a powerful lever for modulating inter-population correlation, more so than simply the number of discrete connections.
• Low-power, subthreshold analog blocks can recapitulate the modular, spontaneously active behavior—the "Lego blocks" of computation—found in biological systems, with implications for scalable, noise-driven, and energy-efficient architectures (Hasani et al., 2017).

Moreover, the requirement for multiple synaptic contacts in artificial circuits to recapitulate biological correlation levels implies that neurons in real circuits may utilize substantial synaptic divergence to achieve robust, distributed spontaneous coupling—an experimentally testable prediction.

7. Future Directions and Experimental Extensions

Emerging evidence identifies the control of spontaneous neuron activity as a substrate for manipulating network-level synchrony, coding efficiency, and resilience. Future work leveraging synthetic biology, soft materials, and advanced CMOS platforms promises to further dissect and engineer the interplay between noise, microcircuit topology, and emergent spontaneous dynamics.

Experimental paradigms incorporating brain-mimetic scaffolds, direct measurement of SAC activity, or controlled modulation of connection divergence will further clarify the quantitative links between microenvironment, network architecture, and spontaneous firing statistics (Hasani et al., 2017, Sumi et al., 2019). The translation of these principles to artificial neural systems and neuromorphic hardware remains a promising avenue for building robust, efficient, and adaptive computational systems inspired by biological precedent.