In this blog post, we wish to highlight the following article that was recently published in the Journal of Robustness Reports:
Wallis, T. S. A. & Martin, J. M. (2026). No evidence that late-sighted individuals rely more on color for object recognition: A Bayesian generalized mixed effects model analysis. Journal of Robustness Reports, 1. https://doi.org/10.21468/JRobustRep.1-rr1.
The target article is:
Vogelsang, M., et al. (2024). Impact of early visual experience on later usage of color cues. Science, 384, 907. https://doi.org/10.1126/science.adk9587.
The one-sentence abstract from the Wallis and Martin article:
A Bayesian mixed-effects analysis found no evidence that patients treated with cataract surgery rely more on color cues for object recognition compared to controls.
The Wallis and Martin discussion section:
Vogelsang et al. [1] reported a significant group difference using t-tests, suggesting that Prakash patients improved more than controls when color was available. In contrast, our Bayesian mixed-effects analysis found evidence in favor of a model with no interaction between group and image type, indicating that Prakash patients were, if anything, equally impacted by the removal of color information in an image recognition task. Expressing performance changes as raw percentages (assuming a linear probability model) results in the color and greyscale conditions in the control group being compressed closer together (Figure 1A). This compression directly creates the interaction effect observed using t-tests. We argue that assuming a linear probability model with Normal measurement error is therefore inappropriate here. Using a generative model that treats this performance data on its natural scale (log odds and binomial errors) mitigates the effects of performance saturation by attributing greater weight to changes at or near ceiling. Therefore, the conclusion that late-sighted individuals rely more on color information than individuals with normal visual development from these data hinges on which modeling approach one views as most appropriate for the underlying data-generating process.
Author
The JRR Team
References
Vogelsang, M., et al. (2024). Impact of early visual experience on later usage of color cues. Science, 384, 907. https://doi.org/10.1126/science.adk9587.
Wallis, T. S. A. & Martin, J. M. (2026). No evidence that late-sighted individuals rely more on color for object recognition: A Bayesian generalized mixed effects model analysis. Journal of Robustness Reports, 1. https://doi.org/10.21468/JRobustRep.1-rr1.
Featured image from the Wallis and Martin paper.