Mid-level perceptual features distinguish objects of different real-world sizes

Area of Science:

• Cognitive Science
• Visual Perception
• Object Recognition

Background:

• The connection between perceptual and conceptual representations is a key area in cognitive science.
• While basic object categories have clear perceptual correlates, broader distinctions like real-world size are less understood.
• It is unclear if early visual processing can reliably distinguish between large and small objects.

Purpose of the Study:

• To investigate whether real-world size is a perceptual feature that can be extracted by early visual processing.
• To determine if perceptual differences exist between large and small objects.
• To explore the influence of early perceptual size information on object recognition and categorization.

Main Methods:

• Visual search experiments were conducted using a wide range of large and small objects.
• Control measures included varying low-level features and image statistics.
• Object representations were simplified to texforms (textures preserving object form) to isolate mid-level features.

Main Results:

• Participants performed visual search tasks faster when distractor objects differed in real-world size from target objects.
• This size-based search advantage persisted across diverse object samples and controlled image properties.
• The effect was observed with texforms but not with basic textures, indicating reliance on mid-level perceptual features.

Conclusions:

• Large and small objects possess reliable mid-level perceptual differences.
• Early visual processing can extract this size information.
• This perceptual information about broad categories may impact subsequent object perception, recognition, and categorization.