Mnemonic Tracing: Using Eye Gaze to Search for Visual Memories

While modern image retrieval systems rely on natural language, some users struggle or are unable to verbalize specific visual memories. Drawing on gaze reinstatement research, we introduce mnemonic tracing: a non-verbal interaction where users intentionally externalize mental images by tracing them onto a blank space using eye gaze. We conducted a pilot study (N = 11) to evaluate retrieval of 30 images of people, objects, and locations using eye-tracking glasses. Without training data or prior user training, our system achieved Top-1 accuracy of 30.5% (compared to 3.3% chance level) and a Top-3 accuracy of 51.2%. We show performance is influenced by individual imagery vividness and task engagement, suggesting success is driven by active mental reconstruction. Highlighting its practical potential, participants reported moderately high perceived usefulness (M = 3.58/5, SD = 0.82) on the Technology Acceptance Model survey. The method offers a private, hands-free, and inclusive alternative for scenarios where a text- or voice-based search is inaccessible.