Contoyiannis (2602.21848) generates biological-type spike trains from the superposition and coupling of intermittent chaotic maps — and the biological character survives increasing complexity.
The starting point: type I intermittency produces time series alternating between nearly periodic laminar phases and chaotic bursts. In earlier work, Contoyiannis showed that coupling two such intermittent dynamics — one tuned to a critical point, another to a tricritical point — produces temporal spike trains that share statistical properties with biological neural spikes. The question addressed here: what happens when you superpose multiple intermittencies before coupling them?
The answer is that biological-type spiking is robust. Superpose two critical intermittencies with different parameter values, and you get a modified time series that's still intermittent. Superpose three, four, or more. Now couple two such superposed time series and examine the resulting spike statistics. The spikes remain biological-type: their interspike interval distributions, burst durations, and fluctuation dynamics match the statistical signatures of real neural membrane potential recordings.
This robustness is not obvious. Superposition could wash out the specific statistical structure needed for biological-type spikes. Instead, the spiking character appears to be a structural property of coupled intermittent dynamics — not dependent on precise parameter tuning but emerging from the class of dynamics itself.
The framing is ambitious: if the dynamics of membrane potential fluctuations in biological neurons arise from coupled intermittent chaotic processes, then neurological decline from spike loss might be understood as parameter drift pushing the system away from the intermittent regime. The numerical framework allows systematic exploration of which parameter changes destroy spiking — a path from dynamical systems theory toward neurology that doesn't require detailed biophysical modeling.