Under revisionEye-tracking

Task success dropped 55% when visual complexity crossed a threshold.

An eye-tracking study identified the precise threshold at which visual complexity collapses user performance — with direct implications for dense dashboard and UI design.

Role: Lead researcher Org: Brain, Attention & Reality Lab, UBC Methods: Eye-tracking, within-user Status: Under revision
Task success didn't decline gradually — it dropped 55% once complexity crossed a threshold.

Below a certain level of complexity, users cope fine. Above it, task success collapses rather than declining slowly. Knowing where that threshold sits for your product is more useful than knowing that complexity is bad.

The problem

Every team building a dashboard, a search results page, or an information-dense interface faces a version of the same question: how much is too much? Visual complexity and cognitive load are easy to discuss qualitatively and hard to measure. The design stakes are concrete: at what point does adding information start subtracting from usefulness?

The approach

I ran an eye-tracking experiment (N=42) with within-user manipulations of display visual complexity and task load, so each participant served as their own baseline. Rather than assume a smooth relationship, I used LLM-assisted scripting to run threshold-detection analyses — searching for the inflection point where performance changes character, not just degrades. That threshold detection is the methodological contribution.

Methodology

Design
Eye-tracking, within-user complexity manipulation (N=42)
Analysis
Threshold detection, gaze-pattern analysis, LLM-assisted scripting
Tools
Eye-tracker, Python, R
Status
Under revision (peer-reviewed)
Code
github.com/jacobgerlofs/h_s_complex_displays

Manuscript

Under peer review — a full citation and link will be added once published.

What we found

Task success was not a linear function of complexity. There were identifiable thresholds at which performance collapsed, and users' gaze patterns changed qualitatively at those same points — attention dispersed rather than degrading smoothly. In the most complex condition, task success fell by up to 55%.

Impact

The results directly inform information-density and visual-hierarchy decisions for dense UIs. Under revision for peer-reviewed publication, with data and code available on GitHub for replication.

So what for product

Visual complexity doesn’t degrade performance linearly — it fails at thresholds. Knowing where those thresholds sit for your specific product and users is far more actionable than general guidance to “avoid clutter.” Research like this gives design and product teams an empirically grounded line to defend: this much complexity maintains task success; this much collapses it. That’s the difference between a design heuristic and a design criterion.