Visual working memory and eye movements in context: How we make use of the external world

Hoogerbrugge, Alex Jan · 2025 · Crossref

DOI: 10.33540/2960

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This dissertation investigates the dynamic interplay between visual working memory (VWM) and eye movements, specifically how humans balance internal memory storage with external information sampling in real-world contexts. The research is motivated by the cognitive costs associated with maintaining VWM representations versus the time and energy required for oculomotor sampling. The central problem addressed is understanding the trade-offs that govern when individuals rely on internal templates versus repeatedly inspecting external sources, such as instruction manuals, to guide visual search and assembly tasks. The study comprises six empirical chapters divided into two parts. Part I utilizes copying tasks and visual search paradigms to examine VWM usage. In Chapter 1, participants performed a copying task where external information was either constantly available, intermittently occluded, or subject to gaze-contingent delays. Chapters 2 and 3 employed a visual search paradigm where participants memorized templates and searched for matches among distractors, manipulating the availability of templates and repeating trials up to twenty-five times to test the persistence of resampling behavior. Chapter 4 investigated multi-target search, determining whether participants used sequential or concurrent search modes when given free choice. Part II focuses on individual and state-dependent influences on eye movements. Chapter 5 analyzed free-viewing eye movements of 1,600 museum visitors across diverse demographics to evaluate the predictive accuracy of 21 saliency models. Chapter 6 examined the physiological coupling between heart rate and oculomotor metrics during movie viewing, using blink frequency and saccade velocity as indicators of arousal. The findings reveal that humans strongly prefer offloading VWM by just-in-time sampling of external information when it is continuously available. Any disruption to this availability, whether through occlusion or delay, significantly increases VWM load, regardless of whether the disruption is predictable. In visual search, participants frequently reinspected external templates even after extensive repetition, a behavior that persisted despite the ability to perform the task from memory. This persistent resampling was found to boost metacognitive confidence rather than improve memory accuracy. Furthermore, participants flexibly switched between sequential and concurrent search modes based on task demands. Regarding eye movement modeling, saliency models performed best for young adults (18–29 years) but poorly for children, highlighting demographic biases in current models. Finally, oculomotor metrics, particularly blink frequency and saccade velocity, were significantly linked to heart rate, demonstrating a physiological coupling between arousal and eye movements. The significance of this work lies in its detailed characterization of the storage-sampling trade-off, demonstrating that human cognition is highly adaptive and context-dependent. The results challenge assumptions about memory efficiency, showing that individuals often prioritize confidence and low-cost sampling over optimal memory encoding. Additionally, the findings underscore the need for demographic validation in computational vision models and provide a non-invasive method for assessing arousal through eye-tracking metrics. These insights contribute to a more integrated understanding of how visual attention, memory, and physiological states interact to guide behavior in complex environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

StageOutcomeToolModelPromptAttemptsCompleted
discover success Crossref 1 2026-06-17
archive success canonical_url 1 2026-06-25
extract success cached 2 2026-06-26
clean success clean 1 2026-06-18
chunk success chunk 1 2026-06-18
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-18
promote success 1 2026-06-17
summarize success llm qwen3.6-27b-prismaquant summ-v5 1 2026-06-26
tag success vector_similarity 6 2026-06-18
verify success 1 2026-06-26

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.