Predisposed and learned preferences for multipoint visual statistics in visually naive newly hatched chicks.

Area of Science:

• Visual neuroscience
• Comparative psychology
• Evolutionary biology

Background:

• Humans and rats can detect luminance distribution and pixel correlations in images.
• This sensitivity is linked to efficient coding in the visual system, retaining key features.
• It's unclear if this visual optimization is innate or learned.

Purpose of the Study:

• To investigate whether visual system optimization for image processing is innate or shaped by experience.
• To determine if newly hatched chicks exhibit preferences for specific visual correlation patterns.

Main Methods:

• Visually naive domestic chicks were presented with various visual patterns (luminance, 2-point, 3-point, 4-point correlations, and noise).
• Chicks' spontaneous approach behaviors towards these patterns were recorded.
• The effect of visual experience (natural visual imprinting) on pattern preference was also studied.

Main Results:

• Chicks preferred luminance, 2-point (lines), and 4-point (rectangular) correlation patterns over noise.
• No preference was observed for 3-point correlations compared to noise.
• Exposure to 3- and 4-point patterns led to a preference for noise, suggesting a learned reduction in complexity preference.

Conclusions:

• Innate biological predispositions significantly influence the efficient coding of natural images, as evidenced by chick behavior.
• Visual experience and learning can modify these preferences, potentially favoring stimuli with lower statistical complexity.