Dual Inhibitory Mechanisms for Definition of Receptive Field Characteristics in a Cat Striate Cortex

A. B. BondsNeurIPS 1991

Introduction

Visual cortical cells in the striate cortex (V1) are remarkably selective. While their inputs from the Lateral Geniculate Nucleus (LGN) are "broadband"—responding to a wide range of orientations and spatial frequencies—cortical cells act as highly tuned filters, firing only for specific angles or levels of detail. The neuroscientific community has long debated how this tuning is achieved. A popular theory, "cross-orientation inhibition," suggested that a cell is inhibited specifically by stimuli at a 90-degree angle to its preference. However, experimental data has often been inconsistent.

This paper seeks to clarify the nature of cortical inhibition by asking: Is it a single process, or are there multiple systems at play? By analyzing the temporal structure of neural spike bursts, the author demonstrates that the cat's visual cortex utilizes two distinct inhibitory mechanisms: a global, contrast-driven "gain control" and a local, spatially-tuned "filter refiner."

Experimental Methods

The study relies on electrophysiological recordings from single cells in the cat striate cortex. Across various experiments, the author analyzed data from dozens of cells (e.g., 52 for orientation dependence, 42 for spatial frequency, 36 for contrast hysteresis, and 59 for burst structure analysis). To tease apart the inhibitory effects, several experimental designs were employed:

  • Dual-Grating Stimulation: A "base" grating (at the cell's preferred orientation) and a "mask" grating (varying in orientation or frequency) were presented simultaneously. By drifting these at different temporal frequencies, Fourier analysis could separate the cell's response to each stimulus.
  • Contrast Hysteresis: To study gain control, sequences of contrast were presented in ascending and descending steps, revealing how a cell's stimulation history affects its current sensitivity.
  • Burst Analysis: Instead of relying solely on average firing rates, the author analyzed the structure of spike trains. An adaptive algorithm identified "bursts" (clusters of spikes with intervals ≤ 8ms) to measure burst frequency and duration.
Orientation tuning and suppression responses to dual-grating stimulation

Key Findings: Global vs. Local Inhibition

The paper argues against a narrow "cross-orientation" model. Instead, it identifies two distinct components of inhibition:

  1. Global Inhibition: A spatially non-selective component that triggers regardless of the stimulus orientation.
  2. Local Inhibition: A spatially-tuned component that refines the edges of the tuning curve.

Hysteresis as Automatic Gain Control

The cortex possesses a powerful, rapid gain control system. When a cell is exposed to high contrast, its sensitivity drops almost instantly (within 50ms). This prevents the system from saturating, allowing it to remain sensitive to changes across a wide dynamic range of light. The average response hysteresis corresponded to a significant shift in sensitivity—about 0.36 log units of contrast—based purely on recent visual history.

Contrast hysteresis demonstrating automatic gain control shifts

Burst Structure and the Violation of Univariance

The most significant finding is that the average firing rate alone obscures underlying neural dynamics. The burst analysis revealed a clear "violation of univariance": two different stimuli (e.g., low contrast versus a non-optimal orientation) might produce the same average firing rate, but the temporal pattern of the spikes is fundamentally different.

If a cell's firing rate is lowered by reducing contrast, the bursts become fewer and shorter. However, if the firing rate is lowered by showing a "wrong" spatial orientation, the bursts specifically become much shorter. Using a metric called "gamma" to measure these differences, the author found that while burst frequency showed little difference (gamma near zero), burst duration showed a gamma of 0.15 to 0.25. This proves that spatial inhibition exerts a unique physiological effect on neural communication. Preliminary pharmacological tests suggest this spatial refinement is specifically mediated by GABAergic inhibitory interneurons.

Limitations and Future Directions

Because the study was conducted entirely on cats, the exact balance of these inhibitory mechanisms might differ in primates or humans, despite the cat visual system being a standard model. Furthermore, while the paper identifies the physiological signatures of these two mechanisms, it does not map the specific physical circuits (i.e., which specific interneurons connect to which pyramidal cells).

However, the findings offer a promising path forward. The author suggests that spike train burst analysis can serve as a "non-invasive" method to observe GABA-mediated inhibition in action. This could open new doors for studying inhibitory deficits in various neurological conditions without the need for direct chemical interventions.

Summary updated

BibTeX
@inproceedings{bonds1991dual,
  author = {A. B. Bonds},
  title = {{Dual Inhibitory Mechanisms for Definition of Receptive Field Characteristics in a Cat Striate Cortex}},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {1991},
}
Figure