Abstract
The responses of neocortical cells to sensory stimuli are variable and state dependent. It has been hypothesized that intrinsic cortical dynamics play an important role in trial-to-trial variability; the precise nature of this dependence, however, is poorly understood. We show here that in auditory cortex of urethane-anesthetized rats, population responses to click stimuli can be quantitatively predicted on a trial-by-trial basis by a simple dynamical system model estimated from spontaneous activity immediately preceding stimulus presentation. Changes in cortical state correspond consistently to changes in model dynamics, reflecting a nonlinear, self-exciting system in synchronized states and an approximately linear system in desynchronized states. We propose that the complex and state-dependent pattern of trial-to-trial variability can be explained by a simple principle: sensory responses are shaped by the same intrinsic dynamics that govern ongoing spontaneous activity.
Introduction: The Challenge of Trial-to-Trial Variability
A fundamental observation in neuroscience is that identical sensory stimuli often elicit markedly different neural responses across trials. Traditionally, this trial-to-trial variability has been treated as stochastic noise. However, an alternative hypothesis suggests that these variations are deterministic consequences of the brain's internal state at the moment of stimulation. While it is well-established that global brain states (e.g., arousal levels) modulate sensory processing, a quantitative mathematical framework linking pre-stimulus spontaneous activity to subsequent evoked responses has remained elusive. This study investigates whether a simple dynamical model can predict these responses on a trial-by-trial basis in the rat auditory cortex.
Methodology: Modeling Population Dynamics
The researchers recorded multiunit activity (MUA) from the primary auditory cortex of urethane-anesthetized rats. To capture the population dynamics, they defined two primary variables:
- Activation ($v$): The instantaneous average firing rate of the recorded population.
- Inactivation ($w$): A hidden variable representing the integrated history of past activity, acting as a recovery or adaptation mechanism.
The core of the approach involved fitting the FitzHugh-Nagumo (FHN) equations—typically used for single-neuron action potentials—to the entire population's activity. Crucially, the model parameters were not fixed; instead, they were re-estimated for each trial using the three seconds of spontaneous activity immediately preceding the stimulus. This "snapshotting" allowed the model to adapt to the brain's fluctuating internal state. The stimulus was then modeled as a brief mathematical perturbation (an alpha-function "kick") to this fitted system.
Results: State-Dependent Predictability
The study found that sensory responses are not merely added to ongoing activity but are shaped by the same intrinsic dynamics governing spontaneous fluctuations. In synchronized states (resembling slow-wave sleep), the cortex behaves as a nonlinear, self-exciting system where stimuli can trigger transitions between "up" and "down" states. Conversely, in desynchronized states (resembling wakefulness), the dynamics become approximately linear, leading to more consistent responses.
Key findings include:
- Quantitative Prediction: The model successfully predicted the population response structure for the initial 100–200 ms following a stimulus on a trial-by-trial basis.
- State-Dependent Linearity: A strong correlation ($R = 0.75$) was observed between the level of cortical synchrony and the nonlinearity of the fitted model. As the cortex desynchronizes, the complex phase-space landscape flattens into a simpler linear regime.
- Role of Past Activity: In synchronized states, the persistent response was strongly anticorrelated with the pre-stimulus "past activity" ($w$), confirming that internal recovery variables dictate response duration.
Discussion and Future Directions
This research suggests that much of the observed "noise" in sensory responses is actually the deterministic result of a stimulus interacting with the brain's current dynamical trajectory. While the model's predictive power is high for early responses, it decays after approximately 200 ms, likely due to complex circuit loops or feedback mechanisms not captured by the two-variable FHN model. Furthermore, while these findings in anesthetized rats provide a robust proof-of-concept, future work must validate this unified dynamical principle in awake, behaving animals where cognitive factors like attention further modulate cortical states.
Summary updated