Quantifying sound-graphic systematicity and application on multiple phonographs
Abstract
Do letter-shapes predict in any way the canonical sounds they represent? Does the letter a in any sense visually predictits canonical pronunciation //? We extended existing quantitative approaches to measuring systematicity between phonol-ogy and semantics. We quantified all pairwise visual distances between letters, using Hausdorff distance. We took thecorresponding canonical pronunciations of the letters and quantified all pairwise distances between their feature-level rep-resentations, using edit distance and Euclidean distance. We defined letter-sound systematicity as a correlation betweenthese two lists of distances. We confirmed Korean as the gold standard for letter-sound systematicity; it was designed in the15C to have exactly this characteristic. We found small but significant correlations in Arabic, Cyrillic, English, Finnish,Greek and Hebrew orthographies, with Courier New giving the most consistent correlations. Pitmans English shorthandand the Shavian shorthand alphabet also showed robust systematicity, and baseline fictitious orthographies showed nosystematicity, validating our approach.