HTI.350 Experimental Research · Tampere University
Does environmental noise affect how fast people react? We designed and ran a controlled experiment to find out — and the answer was more nuanced than we expected.
p = .687
No significant effect on reaction time
Environmental noise did not produce a statistically significant effect on raw reaction time across silent, moderate, and loud conditions.
6 ms
Largest RT difference across conditions
Silent (563ms) vs. loud (569ms) — a difference too small to perceive, equivalent to the blink of an eye.
p = .020
Noise DID affect self-reported focus
Participants reported significantly better focus in silence than in loud conditions — even though their actual performance was the same.
The most interesting finding: noise didn't slow people down, but it made them feel like they were performing worse. This gap between objective performance and subjective experience is a core challenge in interaction design.
Over 20% of Europeans are exposed to unhealthy noise levels. Does it actually slow them down?
"Noise affects performance and human error in relation to impairment in perception, memory, and attention processes." — Smith, 1991
Environmental noise is a documented health risk — linked to sleep disturbances, cognitive impairment, and cardiovascular disorders. But its effect on simple interaction tasks had not been clearly established.
Our hypothesis: exposure to high noise levels would lead to slower reaction times compared to silent conditions. We tested this with a controlled button-clicking task across three noise levels.
Within-subject design with noise level as the independent variable. Each participant completed all three conditions, reducing individual variation as a confounding factor.
Independent Variable
Environmental Noise Level
Silent (< 40 dB) — Noise-cancelling headphones only
Moderate (40 – 60 dB) — Traffic noise audio
Loud (60 – 80 dB) — Traffic noise audio, higher volume
Dependent Variable
Reaction Time (ms)
Measured in milliseconds using E-Prime software and the Chronos response device. Numbers 1–5 appeared on screen; participants pressed the corresponding button as fast and accurately as possible.
30 trials per condition · 90 total · Latin square counterbalancing
E-Prime 3.0
Stimulus presentation and reaction time recording with millisecond accuracy
Chronos Device
5-button response device — records button presses with ms precision
Sony WH-1000XM5
Noise-cancelling headphones for controlled noise delivery
8
Total participants
21–29
Age range
25.5 yrs
Mean age
7 / 1
Right / left-handed
Inclusion criteria: normal hearing, normal or corrected-to-normal vision, no discomfort with loud sounds, normal hand mobility. Signed informed consent obtained from all participants. Order of noise conditions counterbalanced using a Latin square design.
01 — Briefing & Consent
Participants were informed about the experiment, given the opportunity to ask questions, and provided signed informed consent. Background forms collected age, gender, dominant hand, and prerequisite checks.
02 — Practice Trial
10 familiarisation trials in the silent condition before the actual experiment began. This ensured participants understood the button-number mapping and were comfortable with the device.
03 — Experimental Trials
90 trials total — 30 per noise condition. Conditions were presented in counterbalanced order. 1–2 minute breaks between each condition. Total duration approximately 20 minutes.
04 — Post-experiment Questionnaire
5-item Likert-scale questionnaire after each condition. Items covered ability to focus, mental distraction, difficulty concentrating, perceived performance, and perceived reaction time speed.
Mean Reaction Time by Noise Condition
563 ms
Silent
591 ms
Moderate
569 ms
Loud
χ²(2) = 0.75, p = .687 — no statistically significant effect of noise on reaction time
Reaction Time — Objective
Friedman test
χ²(2) = 0.75, p = .687
Silent vs Moderate
Z = -1.68, p = .093
Moderate vs Loud
Z = -.560, p = .575
Silent vs Loud
Z = -.560, p = .575
Bonferroni-corrected Wilcoxon signed-rank tests, threshold p < .017
Subjective Focus — Self-reported
Ability to focus
F(2,14) = 5.25, p = .020 *
Silent vs Loud (focus)
p = .044 *
Mental distraction
F(2,14) = 1.78, p = .206
Hard to concentrate
F(2,14) = 2.12, p = .158
Perceived performance
F(1.11,7.77) = 0.54, p = .503
* Statistically significant. Greenhouse-Geisser correction applied where sphericity violated.
The null result on reaction time makes sense in hindsight. The task was cognitively simple — press a button matching a number. Tasks with low cognitive demand may not be sufficiently challenging for noise to produce a measurable effect. This aligns with Smith (1991), who noted noise primarily impairs memory and attention, not simple motor responses.
The more interesting finding is the subjective one. Participants felt significantly less focused in loud conditions — even though their actual performance was statistically identical. This is a classic example of the gap between perceived and actual usability.
UX Insight: perceived usability ≠ measured usability
If we only measured task performance, we'd conclude noise doesn't matter. But users feeling worse — even if they perform the same — is a real UX problem. It affects trust, satisfaction, and willingness to use a product again. This is exactly why subjective measures like SUS belong alongside performance metrics.
Limitation: N=8 is small, limiting statistical power and generalisability. A larger sample might reveal effects that were underpowered here.
What I Did (33% contribution)
Research plan writing and study design
Documentation and consent form preparation
Experimental setup and apparatus configuration
Participant recruitment
Data collection (running sessions)
Statistical analysis in SPSS
Report writing and proof reading
What I Learned
Running a controlled experiment from scratch — ethics documentation, Latin square counterbalancing, E-Prime scripting, SPSS analysis — gave me a rigorous understanding of what makes research valid and what makes it fragile.
The null result was the most educational outcome. It forced us to think about effect size, statistical power, and whether our task was sensitive enough to detect the phenomenon we were measuring. Most real research looks like this.
And the subjective/objective gap reinforced something I now apply to every UX project: always measure both what users do and how they feel about it.
This project sits outside my typical design work, and that's exactly why it matters. Most UX designers can run a usability test — fewer can design a controlled experiment, handle counterbalancing, apply non-parametric statistics, and interpret a null result correctly.
The finding that noise didn't affect objective performance but did affect subjective experience is something I now carry into every project. Users' feelings about their experience are as real as their measured performance — and sometimes more important for product decisions.