Linguistic systematicity between phonology, semantics, and orthography in English and Korean

Abstract

Language is a complex system with multiple hierarchical subsystems. The current thesis investigates the relation between sub-domains of language: phonology, semantics, and orthography. The dissertation consists of two parts: the first part introduces the concept of grapho-phonemic systematicity and demonstrates it in various orthographies; the second part investigates phono-semantic systematicity in English and in Korean.

Systematicity in Korean was the main motive of the dissertation. Hangeul, the Korean script was intentionally created based on a deep understanding of phonology. The consonants visually represent articulation. I conducted three experiments to confirm grapho-phonemic systematicity in Korean.

In the first experiment, I tested if naïve foreign participants could exploit cross-modal iconicity in hangeul. They had to choose a letter between two that they thought suited the given sound. The frontal phonemes with exposed articulation predisposed listeners to recruit more sensorimotor iconicity.

In the second experiment, I successfully quantified grapho-phonemic systematicity in hangeul, applying methods from previous phono-semantic systematicity research (Dautriche et al., 2017; Monaghan et al., 2014; Tamariz, 2008). I conducted correlation tests between the pairwise distances between written letters and the corresponding pairwise distances between their canonical pronunciations. A significant, positive correlation was found, indicating that similar letters tend to have similar pronunciations in Korean.

In a third experiment, I examined if this grapho-phonemic systematicity indeed facilitates learning. The participants had to learn Korean letter-sound associations without any instruction. The result showed no overall difference between correct letter-sound association and random, fake association. However, differences in performance were observed between consonants and vowels, which I attribute to stronger iconicity of the consonants.

I expanded the realm of grapho-phonemic research to many other languages and their orthographies: six contemporary orthographies (Arabic, Cyrillic, English, Finnish, Greek, and Hebrew); four ancient orthographies (Phoenician, Nabataean, Early Arabic, and Aramaic); two English artificial scripts (the Shavian alphabet and Pitman’s shorthand), and as methodological baselines, two fictitious scripts (Aurebesh from Star Wars and Klingon from Star Trek). I redefined phonemes as binary vectors based on articulatory features from the International Phonetic Alphabet and measured the distances between phoneme vectors. I applied Hausdorff distance (Huttenlocher et al., 1993) to measure the distances between letter shapes. I conducted correlation tests and a Monte-Carlo permutation test. Especially for English, 64 combinations of methods were examined to discover the maximum systematicity. Although lower than hangeul, all the contemporary orthographies returned significant correlations in a range of fonts. None of the four ancient orthographies and neither of the two fictitious orthographies returned significant systematicity. Observed systematicity is expected to increase when the metrics are particularly designed for the individual languages.

Two metrics I used to measure orthographical distances, Stroke Share Rate and Hausdorff distance were validated by being compared with confusion matrices—examples of higher-level letter recognition. With English uppercase letters, I conducted correlation tests between the inter-letter distances from my data and the inter-letter confusability from four confusion matrices (Gilmore et al., 1979; Lommis, 1982; Townsend, 1971; Van der Heijden et al., 1984). Substantial levels of correlation were found between low- and high-level data. Both Stroke Share Rate and Hausdorff distance partly capture letter confusability.

The second part of the dissertation includes two phono-semantic studies. Firstly, I replicated the existing research on phono-semantic (Monaghan et al., 2014) and grapho-semantic systematicity (Dautriche et al., 2017) in English using an identical corpus. Then I examined the grapho-semantic systematicity of spelling-reformed words. I found a high level of graphosemantic systematicity for normal spellings, and a lower level for spelling-reformed-words. I found exactly the same, very low but significant level of phono-semantic systematicity as in previous research (Monaghan et al., 2014).

I examined, for the first time, the phono-semantic systematicity of Korean. I extracted Korean monosyllabic words from a contemporary internet corpus and measured pairwise semantic distances based on contextual co-occurrence. I measured pairwise phonological distances based on articulatory features and conducted correlation tests between the two corresponding lists of distances. Korean demonstrated much higher phono-semantic systematicity than has been found for the five European languages that have previously been studied. It was consistently observed in etymologically and syntactically different subdomains of the language. I attribute this exceptional result to socio-cultural factors that make Korean a relatively ‘esoteric’ language.

I discuss the different findings under an overarching framework, that of the Principle of Least Effort (Zipf, 1949/2016). In language production in particular, high frequency of occurrence tends to make forms similar to each other in order to minimize effort. High frequency of occurrence also has predictable effects on semantic representations; these necessarily correlate with the form changes to generate systematicity. I also introduce a novel conceptual model that distinguishes the two different aspects of systematicity: higher-order systematicity mainly induced by etymological, syntactic or morphological regularities and distributed systematicity that is less intuitable and generalizable but still contributes to the systematicity of language (Distributed-Ostensive Systematicity Model).