Linguistic synesthesia in language contact: Sino-Korean vs. native Korean synesthetic compounds

The paper investigates whether the directionality of linguistic synesthesia is universal or varies by language and culture, particularly under conditions of language contact. Synesthesia in language is framed as metaphorical mapping between sensory domains (e.g., touch to sight in warm color). Classical accounts (Ullmann, Williams) propose a hierarchical, often unidirectional mapping (typically from lower senses like touch to higher senses like hearing/vision). Recent corpus studies, especially in Mandarin Chinese, challenge universality by revealing alternative or bidirectional patterns. To test universality under contact, the study compares synesthetic directionality in Sino-Korean compounds (loan-based or character-derived lexicon) against indigenous Korean compounds, hypothesizing that if universality holds, borrowing should not alter mapping direction; differences would point to cultural/language-specific variation.

The research background surveys foundational and contemporary work on linguistic synesthesia. Ullmann (1963[1957]) proposed a hierarchical, largely unidirectional mapping (Touch → Heat → Taste → Smell → Sound → Sight), with touch commonly a source and hearing a frequent target, based on European poetic data. Williams (1976) extended support using diachronic English adjective semantics, reinforcing unidirectionality. Numerous studies across languages (Hebrew, Italian, Indonesian, Japanese, French, German, Hungarian, Korean) generally support a directional tendency, sometimes attributed to biological or cognitive factors. Counter-evidence includes Mandarin studies (Zhao & Huang, 2015; Zhao et al., 2019) showing rule- and frequency-based biases and bidirectionality, and Turkish (Kumcu, 2021) where some backward mappings are frequent. Korean research is relatively sparse; Jo & Jhang (2019) and Jo (2019) report patterns consistent with the universal tendency using poetic and corpus data. The paper also reviews Sino-Korean as Chinese-origin or character-based lexicon in Korean, historically studied mainly phonologically, and outlines research questions comparing Sino-Korean vs. indigenous Korean synesthetic compounds. Theoretical stances on linguistic synesthesia (metaphorical, neurological, literal) are noted; this study adopts the metaphorical view while allowing cross-linguistic variation.

Materials and Procedure: The study targets synesthetic mappings within Korean compound words, comparing Sino-Korean compounds with indigenous Korean compounds. Data were collected using the KAIST Korean WordNet (http://wordnet.kaist.ac.kr/), a lexical-semantic network comprising 9714 synsets, 8270 words, 20,415 senses, 5752 definitions, 7126 examples, and 4157 case frames. Initial seed items (compound synesthesia examples and sensory adjectives) were drawn from prior literature and expanded via WordNet queries; entries were manually checked for synesthetic status. The Standard Korean Grand Dictionary (423,182 headwords) was used to verify meanings and identify grammaticalized auxiliary-verb compounds for exclusion. Auxiliary-verb compounds (e.g., -보다 ‘see/try’, -대다 ‘do repeatedly’) were excluded due to grammaticalization and loss of literal sensory semantics. Counting was by type rather than token to avoid over-representation of productive derivatives sharing the same root (e.g., 높은음 ‘high sound’ vs. its derived compounds). Examples illustrate expansion (e.g., 경음 ‘fortis’ → 강음 ‘strong sound’, 약음 ‘weak sound’; 높은음 ‘high sound’ → 높은음자리, 높은음자리표). Analysis: The distribution of source→target sensory mappings across compound types was tabulated separately for Sino-Korean, indigenous Korean, and mixed compounds. Given sparse expected counts, the Fisher–Freeman–Halton Exact Test was applied to assess association between source and target modalities (forward vs. backward directions and among mapping types).

• Dataset: 55 synesthetic compound types (160 tokens) were identified: 26 Sino-Korean, 25 indigenous Korean, and 4 mixed Indigenous–Sino-Korean types (Table 1).
• Mapping distribution by type (Table 2): • Touch → Taste: Sino-Korean 1; Indigenous 1; Mixed 1; Total 3 • Touch → Sight: Sino-Korean 2; Indigenous 1; Mixed 0; Total 3 • Touch → Hearing: Sino-Korean 6; Indigenous 5; Mixed 0; Total 11 • Taste → Smell: Sino-Korean 1; Indigenous 5; Mixed 0; Total 5 • Taste → Sight: Sino-Korean 1; Indigenous 2; Mixed 0; Total 3 • Taste → Hearing: Sino-Korean 2; Indigenous 1; Mixed 0; Total 3 • Sight → Smell: Sino-Korean 2; Indigenous 0; Mixed 0; Total 2 • Sight → Hearing: Sino-Korean 12; Indigenous 10; Mixed 3; Total 25
• Overall unidirectionality: Both Sino-Korean and indigenous Korean compound synesthesia exhibit unidirectional mappings with touch as the most basic source and hearing as the highest target domain.
• Distinct olfactory behavior: • Sino-Korean: Olfaction appears only as a target from vision (Sight → Smell present; Taste → Smell present at low frequency), and smell does not function as a source; the presence of Sight → Smell (e.g., 미향 ‘fine/weak fragrance’, 농향 ‘dark/strong scent’) is notable and atypical in Indo-European-based models. • Indigenous Korean: Olfaction is modified by taste (Taste → Smell) and not by sight; aligns with the canonical route Touch → Taste → Smell → Sight → Hearing.
• Statistical test: Fisher–Freeman–Halton Exact Test showed a significant association between source and target categories (p < 0.001), indicating non-random, directionally biased mappings and significant differences between forward and backward transfers and among mapping types.
• Alignment with other languages: Sino-Korean pattern partially aligns with Mandarin (e.g., vision-to-olfaction and lack of smell as source), differing from Indo-European universality claims; indigenous Korean follows Ullmann/Williams-style hierarchy.

The findings indicate that language contact influences synesthetic directionality. Sino-Korean compounds exhibit a rule-like unidirectional hierarchy with touch as the lowest source and hearing as the highest target, but crucially include vision-to-olfaction mappings and lack olfaction as a source. This pattern resembles Mandarin results (Zhao et al., 2019) more than Indo-European models, suggesting that borrowed or character-based lexicon retains source-language conceptualization patterns. Indigenous Korean compounds, by contrast, follow the canonical Touch → Taste → Smell → Sight/Color → Hearing trajectory (when dimension/color are subsumed under sight), supporting universal tendencies reported across many languages. The divergence centers on olfactory mappings: Sino-Korean allows Sight → Smell, whereas indigenous Korean favors Taste → Smell. These outcomes challenge strong universality claims while preserving a general unidirectionality tendency. Differences between Sino-Korean and Mandarin (e.g., Sino-Korean’s stricter unidirectionality and fewer olfactory targets beyond vision) may reflect dataset composition (compound lexemes vs. broader phrasal data) and sample size. Overall, results support a model where universal biases coexist with language- and culture-specific variations shaped by lexical source and contact history.

The study shows that synesthetic compounds in Korean display consistent mapping directionality, but Sino-Korean compounds differ from indigenous Korean particularly in olfactory mappings. Sino-Korean patterns are closer to Mandarin (e.g., Sight → Smell, absence of smell as source), whereas indigenous Korean aligns with the Ullmann/Williams universal hierarchy (Touch → Taste → Smell → Sight/Color → Hearing). Thus, while unidirectionality appears robust, contact-induced variation complicates claims of strict universality. Future work should expand datasets beyond KAIST WordNet, develop or apply automated extraction for compound synesthesia, conduct broader cross-linguistic comparisons (including more non-Indo-European languages), and probe theoretical motivations behind cross-modal hierarchies, especially the role of olfaction in contact lexicon vs. native lexicon.

• Small dataset: 55 types (160 tokens), potentially constrained by the coverage of KAIST Korean WordNet and the inherent rarity of compound synesthesia compared with phrasal instances.
• Data focus: Compound lexemes only; exclusion of phrasal/sentential data may limit generalizability relative to studies using broader corpora (e.g., Mandarin).
• Methodological exclusions: Auxiliary-verb compounds were excluded due to grammaticalization; while justified, this reduces the pool of candidate items.
• Inferential scope: Although the Fisher–Freeman–Halton test indicates significant associations, limited cell counts and sample size caution against overgeneralization.
• Unexplored factors: The study does not analyze underlying motivations for Sino-Korean vs. Mandarin differences or the cognitive/cultural mechanisms driving olfactory mapping differences.