# Perceptual similarity and clustering in braille letter recognition

## TL;DR

This study explores how braille letters are recognized through touch, revealing patterns in their perceptual similarity and structure.

## Contribution

The study introduces a novel method to analyze braille letter recognition using perceptual similarity matrices and hierarchical clustering.

## Key findings

- Braille letters form structured perceptual clusters based on dot features and configurations.
- Letter discriminability goes beyond simple dot overlap, involving complex tactile processing.
- The results offer a benchmark for tactile letter recognition models and instructional design.

## Abstract

Braille is a tactile writing system that enables individuals to read through the sense of touch. Although letter recognition research in the visual modality has informed reading instruction debates, the processes underlying braille letter recognition have received comparatively less attention which has led to little input from researchers toward educators. In this study, we first quantified the formal properties of braille dots using measures of cue validity and entropy-based informativeness, and we tested whether the 26 letters of the braille alphabet were linearly separable in the six-dimensional binary space defined by dot presence. We then examined letter discriminability in fluent Spanish braille readers using a same–different task that included all possible letter combinations. From participants’ accuracy and response time data, we constructed perceptual similarity matrices and applied hierarchical clustering to characterize the structure of braille letter similarity. The resulting clusters revealed a structured perceptual space that reflected both local dot features and global configurations. These results provide a characterization of the perceptual structure of the braille alphabet and show constraints on tactile letter recognition that extend beyond dot overlap, offering a benchmark to guide experimental control, instructional sequencing of letters, and computational models of tactile letter recognition.