A Paradox:Merger or Compounding — Which comes First? Comments on Geoffrey Sampson’s Article Matthew Y Chen Geoffrey Sampson’s paper highlights a peculiar fact about Chinese: the prevalence of coordinate compounds consisting of two near-synonyms. To illustrate this point, in a randomly chosen article of about 670 characters, I found as many as 74 such tokens, depending on how loosely one defines “near-synonyums”. That is almost 1/5 of the running text! Here etc.1 Word-formation of this type is rare, for example in English. The closest examples I can think of are expressions like first and foremost, rules and regulations, and twists and turns. This observation raises the intriguing question of whether such redundant compounds are motivated by the avoidance of massive homophony created by phonological mergers (such as the neutralization between -m and -n endings) or, vice-versa, the mergers arise from the profusion of such pleonastic compounds that render finer phonological distinctions unnecessary. At the heart of Sampson’s argument is the relative chronology of merger (e.g. -m and -n) and compounding, e.g. (cf. M.Chen 1976, Relative chronology: three methods of reconstruction, Journal of Linguistics, 12:209-258). There are three logical possibilities: [End Page 692] A. merger > compounding (x > y means x precedes y) B. compounding > merger C. merger & compounding “co-evolved” Scenario (A) constitutes a counterexample to Gilliéron-Martinet’s hypothesis, which “outlaws” mergers resulting in massive homophony. Therein lies the enigma of Chinese compounds, so maintains Sampson. To make this case, ideally one would like to be able to date the onset of mergers and the rise of disyllabic compounds. Unfortunately such philological evidence is spotty at best. For one thing we don’t have anything close to the venerable, magisterial Oxford English Dictionary, that documents the first occurrence of each lexical entry (more precisely, ). That being the case, Sampson does not offer a direct proof of (A), instead he arrives at (A) by a process of elimination. He rejects (B) by noting that in this case (i.e. before the onset of massive phonological mergers) compounding would be “pointlessly redundant” (e.g. friend + friend = friend). It goes without saying that not all compounds are pleonastic. Prof. Sampson could strengthen his case by providing some statistical estimates of compounds made up of near-synonyms vs. other types of disyllabic words (e.g. ). Next, the author dismisses (C) by saying that such a scenario “seems to reduce homophony avoidance to an unfalsifiable doctrine with no predictive power” (p.5, fn. 3). Indeed, (C) gives too much wiggle room for Gilliéron-Martinet’s hypothesis, and introduces unwelcome wrinkles into the tidy world of logical positivism, where falsifiability is the golden yardstick of truth value. But in the real world, the timeline and relationship between cause and effect may not be as neat as we would like them to be. Specifically: 1. The massive homophony in modern Chinese was the result of multiple sound changes, including nasal ending mergers, final consonant loss, devoicing, palatalization (hence the neutralization between k,kh,x and ts, tsh, s), and so forth — processes that need not happen all at once. It is entirely conceivable that, for instance, nasal merger took place at time x, creating pressure for disambiguation by means of compounding at time y, which in turn facilitates devoicing at time z, and so forth. [End Page 693] 2. Even within one single sound change (e.g. nasal merger), there is the logical possibility of gradual lexical diffusion instead of an instantaneous sweep across the vocabulary as conceived by the Neogrammarians. (For the latest discussion on this topic, see A dialogue on Sound Change between William Labov and William S-Y. Wang, ed. by Sh.L. Feng and Virginia Yip, Peking University Press, 2014). Apart from the core argument, there are a number of related issues. • First of all, Gilliéron-Martinet’s hypothesis bears some scrutiny. Almost half a century ago, as a graduate student I noted in a long-forgotten memo (Monthly Internal Memos, Phonology Laboratory, University of California, Berkeley, circa 1968) that all phonological change is an instance of neutralization. X > Y / Z means there is no contrast...