The Problem

Accurately detecting traffic signals is essential to safe driving. Drivers must continuously scan their environments for signals and hazards, while factors such as color vision deficiency (CVD) and distracted driving can alter visual perception. This experiment examined whether simulated deutan deficiency severity and shortened glance time affected response accuracy in a non-immersive traffic-signal detection task.

The completed study used two independent experimental conditions. In the glance-time condition, simulated deutan deficiency was held at 50% while signal duration varied from 200 to 516.7 ms. In the deutan-deficiency condition, signal duration was held at 350 ms while simulated severity varied from 24% to 100%.

Practical application: Data collected from this experiment can inform machine learning models embedded in automotive systems, enabling better accessibility features for CVD drivers and safety interventions for distracted drivers.

What We Set Out to Answer

RQ1 — Deutan Deficiency

Does increasing simulated deutan deficiency severity reduce the proportion of correct traffic-signal responses?

RQ2 — Glance Time

Does increasing signal glance time from 200 ms to 516.7 ms improve the proportion of correct traffic-signal responses?

Null Hypothesis (H₀₁)

Deutan deficiency severity will not be significantly related to response accuracy.

Null Hypothesis (H₀₂)

Glance time will not be significantly related to response accuracy.

Alternative Hypothesis (H₁₁)

Increasing deutan deficiency severity will reduce response accuracy.

Alternative Hypothesis (H₁₂)

Shorter glance times will reduce response accuracy.

Two Independent Conditions

ConditionIndependent VariableHeld ConstantStimuli Per Condition
Condition 1Glance time (200–516.7 ms)50% simulated deutan deficiency40 trials: 20 red + 20 green
Condition 2Deutan deficiency severity (24–100%)350 ms glance time40 trials: 20 red + 20 green

Dependent variable: proportion of participants who responded correctly on each trial. Red signals were classified as signal present (Brake) and green signals as signal absent (Go).

Participants, Stimuli & Procedure

Participants

Six Virginia Tech students participated. All had normal or corrected vision and no color vision deficiency. Four participants were motivated by partial course requirements and two participated voluntarily.

Stimuli

Eighty visual stimuli were created from 20 original real-world traffic-signal images in Ultralytics' Traffic Light Dataset: 10 red and 10 green signals. Images were processed using the RGBlind Color Blindness Simulator. In Condition 1, all signals were presented with 50% deutan deficiency while duration varied from 200 ms (12 frames) to 516.7 ms (31 frames). In Condition 2, duration was fixed at 350 ms (21 frames) while deutan severity varied from 24% to 100% in 4% increments.

Procedure

01
Randomized presentation
02
Record Go / Brake
03
Complete 80 trials
04
Analyze accuracy

The experiment was self-paced and completed in an artificially lit classroom. Microsoft PowerPoint on a 2020 M1 Apple MacBook Air at full screen brightness was used. Conditions were counterbalanced. Participants completed four practice trials, then viewed all 80 randomized stimuli. After each stimulus, they recorded “G” for Go or “B” for Brake before proceeding to the next trial.

Deutan Deficiency Condition — Sample Clips

All 40 trials in the deutan-deficiency condition used a constant glance time of 350 ms. The clips below illustrate representative signal images at different levels of simulated deutan deficiency. Participants responded G (Go) or B (Brake).

Green Signal Stimuli

Green — Mild Deuteranomaly 24% severity · 350ms · 21 frames
Green — Moderate Deuteranomaly 68% severity · 350ms · 21 frames
Green — Deuteranopia 100% severity · 350ms · 21 frames

Red Signal Stimuli

Red — Mild Deuteranomaly 28% severity · 350ms · 21 frames
Red — Moderate Deuteranomaly 56% severity · 350ms · 21 frames
Red — Deuteranopia 100% severity · 350ms · 21 frames

Glance Time Condition — Sample Clips

These samples hold simulated deutan deficiency constant at 50% while varying glance time. The experimental range was 200 to 516.7 ms (12–31 frames at 60 fps). Participants responded G (Go) or B (Brake).

Glance Time Stimuli

Green — 250ms Glance 50% deuteranomaly · 250ms · 15 frames
Green — 350ms Glance 50% deuteranomaly · 350ms · 21 frames
Red — 500ms Glance 50% deuteranomaly · 500ms · 30 frames

Experimental Results

Alternative hypotheses were partially supported. For each trial, the dependent variable was the proportion of participants who responded correctly. Spearman rank-order correlations were used to test the relationships between each experimental predictor and response accuracy.

Glance time — all participants

ρ(38) = −.07, p = .651 — No significant relationship; accuracy did not improve as glance time increased from 200 to 516.7 ms.

Glance time — excluding Participant 5b

ρ(38) = −.13, p = .419 — The relationship remained non-significant after excluding Participant 5b.

Deutan deficiency — all participants

ρ(38) = −.55, p < .001 — Greater simulated deutan deficiency severity was associated with lower response accuracy.

Deutan deficiency — excluding Participant 5b

ρ(38) = −.24, p = .134 — The relationship was no longer significant after excluding Participant 5b, who accounted for 18 of 19 Condition 2 misses.

Participant Accuracy by Condition

Overall response accuracy for each participant under the glance time and deutan deficiency conditions, with mean and 95% confidence interval

Figure 1. Overall response accuracy for each participant in Condition 1 (shorter glance time) and Condition 2 (greater deutan deficiency). Open circles are individual participants, with lines connecting the same participant across conditions. Black points and error bars are the mean ± 95% CI. This is a descriptive comparison only; the two predictors are not levels of one common factor.

ConditionPredictorTrials (n)ParticipantsSpearman rsp
1Glance time (ms)40All 6−.07.651
1Glance time (ms)40Excluding 5b−.13.419
2Deutan deficiency (%)40All 6−.55< .001
2Deutan deficiency (%)40Excluding 5b−.24.134

Note. The plot summarizes participant-level overall accuracy. Correlations use 40 stimuli/trials within each condition and relate predictor values to proportion correct. Spearman rank-order coefficients are reported. The plot and table summarize complementary analyses.

Interpretation, Limitations & Implications

What the Results Suggest

Glance time was not significantly related to response accuracy in this experiment, so the shortened-glance hypothesis was not supported. In contrast, greater simulated deutan deficiency severity was significantly associated with lower response accuracy when all six participants were included. However, the deutan relationship became non-significant when Participant 5b was excluded, making the finding sensitive to an individual participant's performance.

Signal Detection Theory Context

The task was framed as signal detection: red signals represented signal-present trials and green signals represented signal-absent trials. Brake responses to red signals were treated as hits, while Go responses to green signals represented correct rejections. The report also showed individual variability in false-alarm rate (1.2–23%) and sensitivity, with d′ values ranging from 2.37 to 4.50.

Limitations & Future Directions

A ceiling effect limited the task's ability to reveal performance differences. The overall hit rate was 91%, and three participants had d′ = 4.5, suggesting that discriminating signal from noise was relatively easy. The non-immersive simulation also limits direct generalizability to naturalistic driving. Future studies should use a more immersive medium- or high-fidelity driving simulator and broader, more widely spaced ranges of glance time and deutan deficiency.

Implications

The findings provide support for considering individual variability in color vision when designing roadways, traffic signals, signs, and automobile safety applications. The present experiment did not show a significant effect of glance time, but prior literature cited in the report indicates that distracted-driving glance behavior remains relevant to driving safety.

Why Signal Detection Theory Matters

Signal Detection Theory (SDT) provides the analytical framework for distinguishing perceptual sensitivity from response tendencies. In this task, red traffic signals were defined as signal-present events and green signals as signal-absent events; participant responses could therefore be classified as hits, misses, false alarms, or correct rejections.

The completed experiment also demonstrated individual differences in SDT measures. The report identified false-alarm rates ranging from 1.2% to 23% and d′ values from 2.37 to 4.50, showing that participants varied in sensitivity and decision criteria even within the same task.

Sources

  • Atchison, D. A., Pedersen, C. A., Dain, S. J., & Wood, J. M. (2003). Traffic signal color recognition is a problem for both protan and deutan color-vision deficients. Human Factors, 45(3), 495–503. https://doi.org/10.1518/hfes.45.3.495.27247
  • Bitkina, O. V., Park, J., & Ryu, D. H. (2025). Color vision deficiency recognition based on eye-tracking metrics using machine learning approaches. International Journal of Human–Computer Interaction, 41(14), 8928–8942. https://doi.org/10.1080/10447318.2024.2415764
  • Fitch, G. A., Soccolich, S. A., Guo, F., McClafferty, J., Fang, Y., Olson, R. L., Perez, M. A., Hanowski, R. J., Hankey, J. M., & Dingus, T. A. (2013). The impact of hand-held and hands-free cell phone use on driving performance and safety-critical event risk (Report No. DOT HS 811 757). National Highway Traffic Safety Administration.
  • O’Brien, K. A., Cole, B. L., Maddocks, J. D., & Forbes, A. B. (2002). Color and defective color vision as factors in the conspicuity of signs and signals. Human Factors, 44(4), 665–675.
  • RGBlind. (n.d.). Color blindness simulator. Retrieved September 20, 2026, from https://rgblind.com/color-blindness-simulator
  • Thorslund, B., & Strand, N. (2016). Vision measurability and its impact on safe driving: A literature review. Scandinavian Journal of Optometry and Visual Science, 9(1), 1–9.
  • Ultralytics. (n.d.). Traffic lights dataset. Retrieved September 19, 2026, from https://platform.ultralytics.com/elk-penguin/datasets/traffic-lights-dataset
  • Wang, Y., Bao, S., Du, W., Ye, Z., & Sayer, J. R. (2017). Examining drivers' eye glance patterns during distracted driving: Insights from scanning randomness and glance transition matrix. Journal of Safety Research, 63, 149–155. https://doi.org/10.1016/j.jsr.2017.10.006
  • Wickens, C. D., Helton, W. S., Hollands, J. G., & Banbury, A. (2022). Engineering psychology and human performance (5th ed.). Routledge.