ARGIRA V: Cross-Corpus Evidence for Predictor Inversion in Visual-to-Acoustic Mapping

ARGIRA V investigates whether the same visual variables predict roughness across visual and acoustic domains.

The study combines five independent visual corpora (n = 319 images) and one acoustic corpus (n = 72 sonifications), yielding a total of 391 analyzed cases.

Across visual corpora, edge density consistently predicts visual roughness (Spearman ρ = 0.49–0.84), while luminance contrast shows stable secondary effects (ρ = 0.56–0.74). In contrast, acoustic roughness is primarily predicted by hue entropy (ρ = 0.595), with edge density remaining only a secondary predictor (ρ = 0.428).

This produces a predictor inversion between domains: the dominant predictor of visual roughness is not the dominant predictor of acoustic roughness.

Feature-correlation analysis further indicates that edge density and hue entropy are nearly orthogonal within the visual feature space, suggesting that they capture independent dimensions rather than a common latent factor.

The study identifies three operational regimes:

1. Stable chromatic regime (reproducible mappings).
2. Grayscale-collapse regime (artefactual saturation effects).
3. High-edge-density structural regime (valid extreme cases).

Removal of grayscale-collapse cases eliminates the apparent negative saturation–roughness relationship, demonstrating that previous saturation effects were artefacts generated by grayscale outliers.

The results support a multilayer sonification architecture in which structural information (edge density) and chromatic information (hue entropy) should be represented independently because they contribute non-redundant information to different roughness domains.

Files included:

• multilayer_features.csv
  Visual feature dataset used for cross-feature analysis.

• unified_acoustic_entropy_v4.csv
  Acoustic roughness and entropy measurements.

• argira5_correlation_results.csv
  Spearman correlation outputs reported in the study.

• argira5_multilayer_feature_analysis.py
  Statistical analysis pipeline.

• argira_multilayer_v1-2.py
  ARGIRA multilayer sonification implementation.

• README.md
  Documentation and reproducibility notes.

DOI: 10.5281/zenodo.20534328 License: CC BY-NC 4.0