The Oldest Chinese We Can Hear
Old Chinese (OC) is the ancestor of all Sinitic languages, spoken roughly 1250–200 BCE — the era of the Shijing (詩經) and oracle-bone inscriptions. No one recorded its sounds directly. Linguists reconstruct it by working backwards from four sources: Middle Chinese (MC) transcription systems, rhyme patterns in ancient poetry, loanwords borrowed by neighboring languages, and structural parallels with related tongues in the Sino-Tibetan family.
Karlgren's Trailblazing, and Its Limits
In Grammata Serica (1940), Swedish sinologist Bernhard Karlgren (高本漢) produced the first systematic OC reconstruction dictionary, built on the Middle Chinese system of his earlier Analytic Dictionary of Chinese and Sino-Japanese (1923). He used MC as his pivot and extracted proto-forms like *mi̯wən for 文. His framework was a breakthrough, but later scholars found two major limits. He built his system on phonetic series (the component shared by a cluster of characters hints at their shared OC initial), but bound them with a rigid homorganic principle — every character in a series had to share a place of articulation — which forced ad hoc patches wherever a series refused to comply; and his vowel system was under-differentiated.
The Cluster Revolution
The pivotal insight came when scholars noticed that characters sharing a phonetic element (xíngshēng 形聲) sometimes have radically different MC initials — e.g., 各 (MC kak) sharing a phonetic with 路 (MC lu). The only explanation: OC had consonant clusters that later simplified differently in different environments. Pulleyblank (1962) proposed preglottalized stops (*ʔb-, *ʔd-). Baxter's 1992 monograph A Handbook of Old Chinese Phonology rigorously formalized a cluster system that resolved hundreds of these puzzles.
Baxter-Sagart 2014
The landmark Old Chinese: A New Reconstruction (2014) added Laurent Sagart's morphological dimension. Key innovations:
- Sesquisyllables: OC had minor syllable prefixes (like Mon-Khmer languages today) — e.g., 水 'water' *s.turʔ, where the dot in s. marks a loosely attached presyllable, not part of the main syllable. A hyphen marks the opposite: in 山 *s-ŋrar the s- is fused tightly to the root initial.
- Morphological prefixes/suffixes: *s- causative, *N- stativizer, *-s nominalizer — turning verbs into nouns systematically.
- Pharyngealization: the long-disputed Type A/B syllable split is realised here as pharyngealization of the main-syllable initial — Jerry Norman's 1994 proposal, which Baxter and Sagart adopted and extended to every Type A syllable (五 *C.ŋˤaʔ).
Reading Two Characters
Take 一 'one': OC *ʔit → MC ʔjit (Baxter-Sagart). The glottal stop initial was already present in OC; Mandarin yī reflects the glottal weakening to /j/ plus tone change. Take 山 'mountain': OC *s-ŋrar — the s- preinitial together with the medial *-r- yielded the Middle Chinese retroflex sibilant sr- (ʂ-), the sh- of modern shān 山, while the root initial *ŋ- was lost. Its relatives keep the *ŋ- visible: Baxter and Sagart reconstruct 顏 'face' as *C.ŋˤrar.
Open Questions
Despite the 2014 system's elegance, debates continue. How many vowels did OC have — 4, 6, or more? Were the presyllables true syllables or mere consonantal prefixes? How do Tibetan and Burmese cognates constrain the vowel system? And fundamentally: how far back can the comparative method reach before Chinese characters' logographic nature obscures the sound history entirely?