Neural population coding of multiple stimuli.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Neural population coding typically studies single objects, ignoring natural scenes with multiple co-occurring objects.
• Cortical responses to multiple objects often exhibit stimulus mixing, a combination of individual object responses.
• Existing models do not fully explain the impact of stimulus mixing on neural information processing.

Purpose of the Study:

• To theoretically analyze the consequences of common stimulus mixing rules in cortical responses.
• To investigate how stimulus mixing affects the brain's ability to decode individual objects.
• To explore the relationship between stimulus mixing, attention, and set size effects.

Main Methods:

• Theoretical analysis of neural population coding models.
• Mathematical modeling of linear and nonlinear stimulus mixing.
• Comparison of stimulus mixing costs with gain reduction and noise increase.

Main Results:

• Certain stimulus mixing rules significantly compromise neural decoding of individual objects.
• The detrimental effect of stimulus mixing exceeds that of reduced neural gain or increased variability.
• Stimulus mixing costs escalate with an increased number of encoded objects, potentially explaining set size effects.
• Specific neural correlations and heterogeneity can mitigate the negative impacts of stimulus mixing.

Conclusions:

• Stimulus mixing poses a significant challenge for neural object representation and decoding.
• Attention's benefits may stem primarily from stimulus selection (demixing) rather than solely gain increase or noise reduction.
• Stimulus mixing offers a novel mechanistic explanation for set size effects in perception.
• Understanding conditions for unharmful stimulus mixing is crucial for neural coding theories.