Abstract
Memory-based predictions influence perception, but how they are instantiated in naturalistic environments is unclear. Across five experiments, we used head-mounted virtual reality (VR) to test whether memory for a 360° environment facilitates rapid perceptual judgements of views from within that environment across head-turns. In Experiment 1, participants (N=26) studied 18 real-world panoramas in VR. On each trial of a subsequent Priming Test, participants head-turned left or right toward a snapshot from a studied panorama (target; 110º width) and made a perceptual (open/closed) judgement. Before target onset, another view from the same panorama was briefly presented to prime memory of the target image (prime; 110º width). Before prime onset, an arrow (left/right; valid/invalid) indicated which direction to plan a head-turn in to see the target. We found an interaction between prime condition (Same-scene/Neutral) and arrow validity (p=0.04). On trials with valid arrows (which correctly predicted where the target would land), Same-scene primes improved reaction times (RTs) compared to Neutral primes (p=<.001), but this effect was absent in the invalid arrow condition (p=0.52), showing that memory-based priming occurs in 360° environments and is skewed in the direction of a planned head-turn. Four additional experiments support these results. We replicate the RT advantage for Same-scene < Neutral primes (Experiment 2: N=18, p<.001), and demonstrate that unfamiliar panoramas produce no RT difference between prime conditions, indicating that the priming effect hinges on memory (Experiment 3: N=20, p=0.61). Additionally, by manipulating the spatial structure of studied panoramas, we show evidence that priming respects both the broad spatial structure of studied panoramas (Experiment 4: N=18, p<0.001) and the local spatial structure of upcoming scene views (Experiment 5: N=22, p<0.01). Together, these results indicate that memory of a visuospatial environment influences ongoing perception, priming perceptual judgments of scene views across head-turns in service of efficient perception.