Semantic Expectation Effects on Object Detection: Using Figure Assignment to Elucidate Mechanisms

Area of Science:

• Cognitive Psychology
• Visual Perception
• Neuroscience

Background:

• Previous research suggests semantic labels improve object detection.
• However, it was unclear if this effect involved object or feature detection, or the underlying mechanisms.

Purpose of the Study:

• To investigate whether object detection is influenced by semantic information.
• To differentiate between object detection and feature detection mechanisms.
• To explore the role of semantic representations in object recognition.

Main Methods:

• Objects were presented briefly (90-100 ms) on masked bipartite displays.
• Participants performed object detection by segmenting figures from backgrounds.
• Valid and invalid semantic labels were used, with control conditions for comparison.

Main Results:

• Valid labels enhanced accuracy and reduced response times, particularly for upright objects.
• Invalid labels (different or same superordinate category) decreased accuracy for upright objects only.
• Orientation dependency suggests object representations, not invariant features, are key.

Conclusions:

• Object detection is not merely affected by semantics; it intrinsically involves semantic processing.
• Semantic interference, especially from same-category labels, delays object detection.
• The findings highlight the crucial role of semantic memory in visual object recognition.