Dynamics of Response to Perceptual Pop-Out Stimuli in Macaque V1

Matthew A. Smith, Ryan C. Kelly, Tai Sing LeeJournal of Neurophysiology 2007 · DOI

TL;DR

This study investigates how neurons in the primary visual cortex (V1) of macaques respond to "pop-out" stimuli, where a unique object stands out from its surroundings. They found that V1 neurons exhibit a delayed and distinct form of contextual modulation for these stimuli, suggesting a role for V1 in computing perceptual salience beyond simple feature detection.

What Problem Does This Paper Address?

The visual system is remarkably adept at quickly identifying salient objects, even in cluttered scenes. This phenomenon, known as "perceptual pop-out," is thought to involve contextual modulation, where the response of neurons in the visual cortex is influenced by stimuli outside their classical receptive field. While various forms of contextual modulation, like "surround suppression" (where stimuli in the surround reduce the response to a central stimulus), have been well-studied, their relationship to other types of modulation, particularly those not based on simple orientation, remained unclear. Specifically, researchers wanted to understand if the mechanisms underlying pop-out are distinct from those of surround suppression and how the timing and nature of these responses contribute to our ability to perceive these salient objects.

Methods

The researchers trained two macaque monkeys to perform a two-alternative forced-choice (2AFC) pop-out detection task. The stimuli used were arrays of shaded spheres, where one "oddball" sphere had a different lighting direction (either from above, LA, or from below, LB) compared to the surrounding "distractor" spheres. This shape-from-shading cue creates a perception of 3D shape and elicits pop-out.

Key methods included:

  • Behavioral Task: Monkeys had to fixate a point, then saccade to the location of the pop-out stimulus to receive a reward. This allowed for the measurement of behavioral performance (accuracy and contrast thresholds).
  • Electrophysiology: Single-unit recordings were made from V1 neurons while the monkeys performed the task.
  • Stimulus Design: The use of shape-from-shading stimuli was crucial, as it elicits pop-out without relying on orientation differences, allowing researchers to probe mechanisms beyond orientation-selective surround suppression.
  • Latency and Modulation Analysis: The study meticulously analyzed the timing of neuronal responses, distinguishing between the initial response to the stimulus (onset latency) and the later modulation caused by the surrounding distractors (modulation latency). They used statistical methods and bootstrap analysis to ensure the reliability of these measurements.

Datasets

Not applicable. This study involved novel behavioral experiments and electrophysiological recordings in live animals.

Key Insights and Ideas

  • Asymmetric Pop-Out Perception: Monkeys were significantly better at detecting pop-out stimuli where spheres were lit from below (LB) compared to those lit from above (LA). This behavioral asymmetry is consistent with human psychophysical findings and suggests our visual system has a bias towards assuming lighting from above.
  • Delayed Contextual Modulation: The contextual modulation in V1 neurons in response to pop-out stimuli occurred with a substantial delay (average of 145 ms after stimulus onset), much longer than previously reported for orientation-based surround suppression (around 8 ms). This suggests a different underlying neural circuitry.
  • Feature-Specific Modulation: The timing and magnitude of modulation differed between LA and LB pop-out stimuli, indicating that the contextual modulation is not a generic, non-specific enhancement but is sensitive to the specific features of the pop-out stimulus.
  • V1's Role in Salience Computation: The findings suggest that V1 neurons are not just passive receivers of sensory information but actively participate in computing perceptual salience, likely through recurrent interactions with higher visual areas.
  • Distinct Mechanisms: The long latency and feature specificity of pop-out modulation strongly argue against it being mediated by the same mechanisms as iso-orientation surround suppression.

Key Results

  • Behavioral Performance: Monkeys showed significantly lower contrast thresholds for detecting LB pop-out stimuli compared to LA pop-out stimuli.
  • Neuronal Latency: The average modulation latency for pop-out stimuli in V1 neurons was 145 ms (ranging from 70 to 230 ms). This is considerably longer than the latency for iso-orientation surround suppression (around 59 ms).
  • Modulation Magnitude: While a significant portion of V1 neurons (around 30%) showed contextual modulation to pop-out stimuli, the effect was generally smaller in magnitude compared to surround suppression.
  • Asymmetry in Neuronal Response: LB pop-out stimuli tended to elicit earlier and slightly larger contextual modulation responses in V1 neurons compared to LA pop-out stimuli, mirroring the behavioral asymmetry.
  • No Correlation with Choice Probability: Individual V1 neuron responses were not predictive of the animal's behavioral choice on a trial-by-trial basis, suggesting that the decision-making process relies on pooled activity from larger neuronal populations.

Limitations and Future Work

  • Spatial Origin of Modulation: While the study used stimuli designed to isolate the extraclassical surround, the precise spatial origin of the contextual modulation (e.g., within the classical receptive field vs. beyond) remains an area for further investigation. The authors acknowledge that stimulation of the outer regions of the RF could contribute.
  • Mechanism of Feedback: The study strongly implicates feedback from higher visual areas but does not definitively pinpoint the specific pathways or mechanisms involved. Future work could explore the role of specific feedback connections.
  • Perception of 3D Shape: While the stimuli are designed to evoke 3D shape perception, the study acknowledges that it's not definitively proven that macaques perceive these stimuli as 3D shapes in the same way humans do. However, they argue that the observed asymmetries and neuronal responses are consistent with such a perception.
  • Feature Specificity: The study found that neurons responding to LA pop-out did not necessarily respond to LB pop-out, suggesting distinct populations of neurons are involved. Further research is needed to understand what determines this feature specificity.
  • Attentional Role: While attention was considered, the results did not strongly support a simple, non-specific attentional enhancement as the sole driver of contextual modulation. The role of attention in feature-specific pop-out remains an open question.

Summary updated

Abstract

Contextual modulation due to feature contrast between the receptive field and surrounding region has been reported for numerous stimuli in primary visual cortex. One type of this modulation, iso-orientation surround suppression, has been studied extensively. The degree to which surround suppression is related to other forms of contextual modulation remains unknown. We used shape-from-shading stimuli in a field of distractors to test the latency and magnitude of contextual modulation to a stimulus that cannot be distinguished with an orientation-selective mechanism. This stimulus configuration readily elicits perceptual pop-out in human observers and induces a long-latency contextual modulation response in neurons in macaque early visual cortex. We found that animals trained to detect the location of a pop-out stimulus were better at finding a sphere that appeared to be lit from below in the presence of distractors that were lit from above. Furthermore, neuronal responses were stronger and had shorter latency in the condition where behavioral performance was best. This asymmetry is compatible with earlier psychophysical findings in human observers. In the population of V1 neurons, the latency of the contextual modulation response is 145 ms on average (ranging from 70 to 230 ms). This is much longer than the latency for iso-orientation surround suppression, indicating that the underlying circuitry is distinct. Our results support the idea that a feature-specific feedback signal generates the pop-out responses we observe and suggest that V1 neurons actively participate in the computation of perceptual salience.
BibTeX
@article{smith2007dynamics,
  author = {Matthew A. Smith and Ryan C. Kelly and Tai Sing Lee},
  title = {{Dynamics of Response to Perceptual Pop-Out Stimuli in Macaque V1}},
  journal = {Journal of Neurophysiology},
  year = {2007},
  doi = {10.1152/jn.00441.2007},
}
Figure