Correction to: Detecting distortions of peripherally presented letter stimuli under crowded conditions

Thomas S. A. Wallis, Saskia Tobias, Matthias Bethge, Felix A. WichmannAttention 2019 · DOI · OpenAlex

Introduction

Visual crowding is a fundamental limitation of peripheral vision where objects that are easily identified in isolation become indistinguishable when surrounded by nearby "flankers." To quantify this effect, researchers often employ Radial Frequency (RF) distortions—sinusoidal modulations of a shape's contour. A 2017 study investigated how crowding impacts the detection of these distortions in Sloan letters. However, the authors subsequently identified a mathematical error in their stimulus generation code. This paper presents a formal correction, demonstrating that while the bug introduced unintended visual artifacts, the original conclusions regarding the robustness of crowding remain valid.

The "Swirl" Bug and Stimulus Correction

The original study intended to use a phase offset ($\phi$) to rotate distortion patterns around a letter's center. Due to a coding error, the phase offset incorrectly affected both the angular and radial components of the distortion equation. This resulted in "swirls"—distortions whose intensity varied based on the interaction between the letter's geometry and the randomized phase.

To address this, the researchers corrected the software to produce "true" RF distortions. They then conducted replication experiments using both an original author and a new naïve observer to determine if the "swirl" artifact had influenced the initial results.

Examples of Radial Frequency (RF) distortions applied to Sloan letters across different amplitudes.

Experimental Methods

The replication utilized an interleaved design, where "buggy" original images and "bug-fixed" images were presented under identical conditions. This allowed for a direct comparison of detection thresholds. Data were analyzed using Bayesian Weibull psychometric functions (via the brms R package) to estimate thresholds and credible intervals accurately. The researchers also corrected a minor typographical error in the reported amplitudes for their second experiment. The full dataset and corrected stimuli are now available via Zenodo.

Results and Discussion

The primary finding of the re-evaluation is the remarkable robustness of the crowding effect. Despite the change in the nature of the distortion, the core phenomenon—that flanking letters significantly increase detection thresholds—was consistently observed across both stimulus types.

Comparison of detection thresholds for original and bug-fixed stimuli across different distortion frequencies.

As shown in the figure above, the relationship between distortion frequency and detection threshold followed the same pattern for both original and corrected stimuli: sensitivity increased (thresholds decreased) linearly on a log-log scale as frequency rose. While the absolute thresholds for the bug-fixed images were generally lower (making them easier to detect), the relative magnitude of the crowding effect remained stable. Interestingly, because the bug created a distinct type of distortion, the researchers inadvertently demonstrated crowding for three types of distortions rather than two, broadening the scope of their original conclusions.

Furthermore, the study re-confirmed the lack of "distortion pop-out." Detecting a distorted target among undistorted flankers was not significantly easier than the reverse, reinforcing the idea that crowding disrupts the integration of shape features. The authors noted significant individual differences in sensitivity, particularly with the new naïve observer, suggesting that factors like fixation stability and physiology warrant further study.

Conclusion

This correction highlights the importance of scientific transparency and the robustness of visual crowding as a perceptual phenomenon. Although the original stimuli contained unintended artifacts, the fundamental "laws" of crowding—specifically the frequency-dependent sensitivity and the disruptive effect of flankers—hold true. The updated resources ensure the published record is mathematically consistent and provides a solid foundation for future research into peripheral vision.

Summary updated

BibTeX
@inproceedings{wallis2019correction,
  author = {Thomas S. A. Wallis and Saskia Tobias and Matthias Bethge and Felix A. Wichmann},
  title = {{Correction to: Detecting distortions of peripherally presented letter stimuli under crowded conditions}},
  booktitle = {Attention},
  year = {2019},
  doi = {10.3758/s13414-019-01855-9},
}
Figure