Marie Guiraud is currently a PhD student working with Lars Chittka. She is going to present her research on bee cognition.
Bumblebees have remarkable visual learning capacities and are a simple model for understanding visual processing in a miniature network. The mechanisms underlying bee visual discrimination are still poorly understood. Biederman’s theory on human visual object recognition suggests that objects can be described by a combination of their simple components. Global visual perception in humans may be derived by eye saccades moving between local features. Accordingly, critical questions in bee vision are whether and how bees process localised elemental information within stimuli to generate a global perception. Here, we explore insights from Biederman’s theory studying which visual features (vertices or edges) may be important for bees in the visual recognition and which strategy they may use to infer a holistic perception. After training bumblebees to discriminate between a square and a triangle, bees were tested to identify these 2D objects with deletion of vertices, edges or in other modified configurations. We show that bees can still recognise these visual stimuli depending on distinctive preferences in using edges or vertices for pattern recognition. Using 3D video tracking, we discover that bees mainly scan the bottom of the presented shapes and generally focus on one of the side of the stimuli (lateralization), which was also found in other sensory modalities in vertebrates. Our results suggest that bumblebees perceive the natural environment via local features as humans do. They then integrate these features to produce a global perception by flying between these elemental features.