Author ORCID Identifier
Date of Award
Summer 2026
Document Type
Thesis (Ph.D.)
Department or Program
Cognitive Neuroscience
First Advisor
Caroline E. Robertson
Abstract
As we look around the world, a large portion of our surroundings sit outside our field of view. To build a complete picture of our surroundings, we constantly exchange our current view for new ones through eye, head, and body movements. How do we maintain a unified sense of place despite these frequent exchanges? In this dissertation, I propose that prediction of upcoming scene views during head and body movements supports our seamless experience of the visual world.
Building on evidence that visual prediction occurs across eye movements, Aim 1 uses a novel head-mounted virtual reality (VR) paradigm to test whether predictions also arise across head and body movements, drawing on observers' memory for familiar immersive scenes. I show that observers make faster perceptual judgments after turning toward a familiar scene view when they are primed with a contiguous view from the same scene. This provides evidence that predictions of upcoming scene views are generated across head turns. In three additional experiments, I show that (i) priming relies on memory for the studied scene, with no priming observed in unfamiliar scenes, (ii) priming only occurs in the direction of observers’ planned actions, consistent with a role in supporting active vision, and (iii) priming only occurs for views presented in their expected spatiotopic locations, indicating that primed content depends on observers’ actions in a scene. In Aims 2 and 3, I examine the neural substrates of scene view prediction through two central elements of hierarchical predictive coding: prediction representation (Aim 2) and prediction error tracking (Aim 3). Participants first learned an immersive scene in VR before watching sequences of views from the studied scene during an fMRI scan. In Aim 2, I tested whether the brain prospectively represents upcoming scene views before they appear. I found that predicted-but-omitted views were selectively represented in an anterior, memory-guided scene network – the place memory areas (PMAs) rather than in a closely neighboring posterior scene-perception network – the scene perception areas (SPAs). In Aim 3, I found that the same memory-guided scene network also tracked violations of predicted scene views.
Together, this work provides evidence that memory-based predictions are generated across head turns in immersive, real-world scenes, and that perceptual continuity across scene views relies on a memory-guided scene network representing both predictions and prediction errors of upcoming views. More broadly, these findings suggest that prediction is a key mechanism supporting our seamless experience of our immersive visual world.
Original Citation
Chapter 1 of this dissertation, Memory-based predictions prime perceptual judgments across head turns in immersive, real-world scenes, was originally published in Current Biology (2025); 35(1), 121-130, and is reproduced here with minor edits for formatting consistency.
Recommended Citation
Mynick, Anna, "EXAMINING PREDICTION ACROSS VIEWS IN IMMERSIVE, REAL-WORLD SCENES" (2026). Dartmouth College Ph.D Dissertations. 544.
https://digitalcommons.dartmouth.edu/dissertations/544
