Letters and their sounds are not perfectly arbitrary : exploring grapho-phonemic systematicity in multiple orthography systems

Language, as a complex system, suggests coordination between subsystems. Recent studies demonstrated that semantically similar words tend to have similar pronunciation (Blasi et al., 2016; Dautrich et al., 2017; Jee, Tamariz, & Shillcock, 2022; Monaghan et al., 2014; Tamariz, 2008). The current research, for the first time, quantified mapping between letters and their canonical pronunciations, or grapho-phonemic systematicity. We examined naturally developed phonograms (Arabic, English, Greek, and Hebrew), consciously designed phonograms (Korean, Shavian alphabet, and Pitman's shorthand), a logographic orthography (Chinese) and fictitious orthography systems (Aurebesh and Klingon). We measured all the pairwise phonological distances between phonemes in the respective alphabet system, and the corresponding pairwise orthographical distances between letters. We then tested Pearson's r between these two lists of pairwise distances. The positive correlation coefficient means that similar letter-shapes have similar canonical pronunciation. In contrast, the negative correlation means that similar letter-shapes have more distinct sounds, or vice versa. We verified the significance of the correlations by conducting Monte-Carlo permutation tests. For the phonological distance, phonemes were encoded into vectors according to the articulatory features and the distance between the vectors were calculated in various ways. We applied three methods to measure the pairwise distances between letter-shapes. Pixel count simply defines the distances as the difference in the number of pixels between two characters. Perimetric complexity is defined 354 This paper is distributed under a Creative Commons BY-NC-ND 3.0 license