How do cortisol temporal memory timescales shape prefrontal-amygdala emotion regulation dynamics?

Cortisol as memory rather than a fixed input

The paper asks how the timing of cortisol's influence changes the dynamics of a prefrontal-amygdala circuit, not how much cortisol is present. Cortisol is not modeled as its own dynamical variable or as a direct generator of neural oscillations. Instead, its effect appears as history-dependent modulation of two stress-sensitive interactions: recurrent excitation within the prefrontal pyramidal population, written w_PP, and inhibition of basolateral amygdala principal cells, written w_CB.

That choice sets up the central question. If past neural activity feeds back into current coupling strengths, then the circuit does not experience fixed coupling strengths at fixed baseline parameters. The effective recurrent excitation and amygdala inhibition vary in time according to recent activation history. The paper's target is therefore the shape of the temporal memory through which that history acts, and how that shape reorganizes transitions between stable activity and infra-slow oscillations.

A four-population circuit with two stress-sensitive weights

The model is a Wilson-Cowan system with four interacting variables. Pyramidal cortical neurons (P) provide recurrent glutamatergic excitation and the main cortical output. Somatostatin-expressing cortical interneurons (S) inhibit pyramidal dendrites and shape dendritic integration. Cortico-amygdala principal neurons (C) are glutamatergic basolateral amygdala cells that receive cortical input and send reciprocal excitation back to prefrontal cortex. Basolateral amygdala interneurons (B) provide local GABAergic inhibition onto those principal cells.

The two couplings that cortisol modulates map onto this architecture. Recurrent prefrontal excitation w_PP captures sustained cortical activity and top-down control. Amygdala inhibition w_CB captures the inhibitory control of basolateral amygdala principal cells. In the paper's interpretation, w_CB represents relatively fast cortisol-related modulation of amygdala inhibition, while w_PP captures slower effects on recurrent prefrontal recruitment. The bifurcation diagrams therefore plot baseline coupling strengths, while cortisol acts on their effective values through temporally filtered representations of recent amygdala activity.

Weak, strong, and discrete-delay memory kernels

To separate the effect of having temporal memory from the effect of how that memory is distributed, the authors compare three representations. A weak Gamma kernel (n_f = n_s = 1) places the greatest weight on the recent past and decays monotonically with time. A strong Gamma kernel (n_f = n_s = 2) shifts the contribution of past activity away from the present and organizes it around a delayed maximum. The discrete delay is treated as the limiting case of increasingly concentrated Gamma distributions at a fixed mean delay.

The simulations used fast and slow average delays of τ_f = 120 s and τ_s = 1000 s, described as within the baseline physiological range for the fast and slow average delays specified in the paper's Table 1. These are model parameters, not fitted measurements. The comparison is designed so that differences between the weak and strong kernels reflect the distribution of memory, while the discrete-delay limit represents the most concentrated form of that memory.

Memory delays oscillation onset and extends its range

Without temporal memory, the baseline system organizes oscillations in two bounded regions of the (w_PP, w_CB) plane. Low recurrent cortical excitation is insufficient to destabilize the equilibrium, and the system remains non-oscillatory. As w_PP increases, oscillations emerge through a supercritical Hopf bifurcation. The oscillatory window then terminates within the observed parameter range for all values of w_CB, with restabilization to an equilibrium. The termination mechanism depends on amygdala inhibition: weaker inhibition ends the window through a limit point of cycles, mid-range values end it through a SNIC-like transition with a blowup of the period, and stronger inhibition ends it through a limit point of cycles. Notably, the w_PP level at which oscillations terminate stays relatively unchanged across these mechanisms.

Introducing weak Gamma memory changes this picture in two directions. With w_CB fixed at −3.2, the transition into sustained cycling occurs at a higher value of recurrent prefrontal excitation than in the no-delay case: at w_PP = 3 and w_PP = 3.5 the trajectories show damped oscillatory transients toward a stable equilibrium rather than a stable periodic orbit, and sustained oscillations appear only after w_PP is increased further. By w_PP = 4.43 the system has entered a large-amplitude oscillatory regime with a clear separation between low- and high-activity phases, and oscillations remain organized at w_PP = 4.5. The continuation analysis shows the oscillations extending over a broader range of w_PP than without delays.

The effect of amygdala inhibition is also altered. In the no-delay system, stable cycles persist as inhibition is strengthened across the explored range. In the weak-delay system, the periodic branch eventually terminates via a homoclinic bifurcation at a finite value of strongly negative w_CB. The authors interpret this as a lower cutoff in w_CB that was not a dominant feature of the instantaneous system: moderate inhibitory strength supports the periodic regime, but sufficiently strong inhibition eventually suppresses it.

Kernel shape barely changes the global picture

The most striking comparison is between the weak and strong Gamma kernels. Because both kernels share the same equilibria, the authors expected no general guarantee that bifurcations and oscillatory regions would be robust. In the model, however, changing the delay distribution from the weak to the strong Gamma kernel had very little effect on the global picture. At the same mean delays, the major bifurcation boundaries and the region supporting stable periodic activity remained almost unchanged.

This matters because the two kernels weight history differently: the weak kernel emphasizes the immediate past, while the strong kernel suppresses the present and places greater weight around a delayed peak. Those differences in temporal weighting produced only minor changes in the principal transitions between attracting states. Within the range examined, the model therefore appears locally robust to moderate changes in the shape of distributed cortisol memory. The authors conclude that the principal dynamical effect arises from introducing history dependence itself, while the precise distribution of that history matters less.

The discrete-delay limit produces quasiperiodic activity

The discrete-delay limit behaves differently. It generates a substantially richer bifurcation structure than either Gamma kernel, including quasiperiodic activity. In this regime, the timing and shape of successive loops become modulated, so recurrent prefrontal excitation can destabilize the regular rhythm without simply suppressing oscillations altogether.

The side panels show that these quasiperiodic dynamics are themselves strongly affected by w_PP. As recurrent prefrontal excitation increases, the trajectory takes substantially longer to complete a full loop around the torus. The authors describe this as progressively longer modulation timescales in the interaction between prefrontal and amygdala populations: the circuit may continue to alternate between regulatory and reactive configurations, but the return to a comparable network state occurs over increasingly long intervals. In this regime, changes in recurrent prefrontal excitation affect not only whether the circuit oscillates, but also how regularly and how quickly cortico-limbic states are revisited.

The authors connect this to human electrophysiology, where coherent quasi-periodic fluctuations centered near 0.015 Hz have been observed. They suggest that quasi-periodicity may not necessarily reflect a distinct intrinsic circuit architecture, but may instead arise from the system's temporal integration, particularly when the relevant history is concentrated around a more sharply localized past time.

A phenomenological framework, not a fitted physiology

The model's main limitation is its phenomenological nature. The authors state that their aim was not a quantitatively fitted representation of cortico-limbic physiology but a proof-of-principle framework for asking how temporal integration of stress memory alters effective interactions. The functional forms were chosen to be as simple as possible while reproducing the qualitative mechanisms of interest. Parameterization was guided by known relationships, relative strengths, and physiological constraints reported in the empirical literature, but the exact numerical values were not fitted to data and should not be read as direct physiological measurements. The specific numerical boundaries of the oscillatory and non-oscillatory regimes are therefore model-dependent, as are the transitions and mechanisms identified as the framework's main predictions.

The circuit is also intentionally incomplete. The minimal prefrontal-amygdala system isolates the consequences of stress-sensitive modulation of recurrent cortical excitation and amygdala inhibition, but emotion regulation emerges from a larger network. The hippocampus, thalamic and hypothalamic structures, neuromodulatory systems, and the HPA axis itself can all contribute. The authors propose coupling this cortico-limbic model to their previously developed HPA-axis model so that cortisol emerges from an explicit endocrine feedback system rather than an imposed history-dependent modulation.

The shape of memory, not only its timescale

The paper's central claim is a comparison, not a single number. Introducing temporal memory reorganizes the parameter region supporting infra-slow oscillations and alters their persistence, and this happens across the distributed-delay cases with average delays consistent with free cortisol courses. The finer details of the dynamics depend on the shape of the delay kernel. Weak and strong Gamma kernels produce remarkably similar global dynamics, whereas the discrete-delay limit generates a substantially richer bifurcation structure, including quasiperiodic activity.

For readers weighing what cortisol timing does to prefrontal-amygdala regulation, the practical priority follows from that split. The presence of history dependence shifts the onset of stable cycling toward stronger recurrent prefrontal excitation and makes the periodic regime considerably more persistent as w_PP increases; strong amygdala inhibition can then introduce a lower cutoff in w_CB that the instantaneous system lacks. The shape of the memory matters most when it becomes sharply concentrated, as in the discrete-delay limit, where quasiperiodic regimes and slower revisiting of network states appear. The authors' interpretation is that different forms of temporal integration may provide a mechanism by which neural systems implement distinct functional dynamics, while changes in this integration could promote atypical or pathological rhythmic regimes.

Sources used in this answer