¹ This paper was begun as a research project in a graduate seminar at the University of California. To Mary It. Haas I owe special gratitude for developing my interest in descriptive linguistics. I have benefited from discussions of this paper with Bernard Bloch, Elinor Clark Horne, and especially Itulon S. Wells.

² Cf. B. Bloch, Lg. 26.87–8 (1950). I have heard the speech of three informants from Seoul: Miss Young Sook Chang, Mrs. Blanche C. Lim, and Mrs. Sang Soon Yun; and of two informants from Phyongyang: Mr. Frank Lee and Mr. Eung Pal Yun; as well as that of Miss Pauline Kim, of a Korean family living in China. (Since this article was written I have heard the speech of two more informants from Seoul: Mr. Sung Un Chang and Mr. Peter H. Lee.)

I have benefited from the studies made by Elinor Clark Home, Introduction to spoken Korean, 2 vols. (New Haven, 1950–1), and by Fred Lukoff, Spoken Korean, 2 vols. (New York, 1947), as well as from unpublished studies by Rulon S. Wells.

A partial bibliography of other secondary sources consulted: Hakhwe [Korean Language Society], phyocunmal [Outline of standard Korean language] (Seoul, 1939). Id., mal [Unabridged Korean dictionary], Vol. 1, (Seoul, 1947). Id., Vol. 2 (Seoul, 1949), through Korean letter m. P. A. Eckardt, Koreanische Konversations-Grammatik (Heidelberg, 1923). H. H. Figulla, Prolegomena zu einer Grammatik der koreanischen Sprache, Berlin-Universität Mitteilung des Seminars für orientalische Sprachen 38.101–21 (1935). J. S. Gale, A Korean-English dictionary (Seoul, 1897). Id., Korean grammatical forms (Seoul, 1894). Sun-Gi Gim (= Sunki Kim), Korean, Le maître phonétique III.15.21–2 (1937). M. Haguenauer, Système de transcription de l'alphabet coréen, Journal asiatique 222.145–61 (1933). I (= Li), nonkoŋ [Articles on Korean linguistic research] (Seoul, 1947). I (= Li), tohε hak [A study of Korean phonetics, with diagrams] (Seoul, 1947). Yuncε I (= Li), Phyocun mal [Dictionary of standard Korean] (Seoul, 1947); Insub Jung, Romanization of Korean, Actes du 4^e congrès international de linguistes 210 (Copenhagen, 1936). A. A. Kholodovich, O latinizacii korejskogo pis'ma, Sovetskoje Jazykoznanije 1.147 ff. (1935). Kim, [An exposition of Korean orthographic rules] (Seoul, 1946). G. M. McCune and E. O. Reischauer, The romanization of the Korean language, Transactions of the Korean Branch of the Royal Asiatic Society (Seoul, 1939). Sinpei Ogura, The outline of the Korean dialects, Memoirs of the Research Department, Tôyô Bunko 12 (1940). Id., Tyoosengo-hoogen no kenkyuu [Research on Korean dialects], 2 vols. (Tokyo, 1944). Id., Tyoosengogaku-si [A history of Korean language study] (Seoul, 1921). E. W. Pae, Conversational Korean (Washington, D. C, 1944). K. D. Park, Oral Korean for beginners (Honolulu, 1945). E. D. Polivanov, Glasnye korejskogo jazyka (Petrograd, 1916). G. J. Ramstedt, A Korean grammar, Mémoires de la Société Finno-ougrienne 82 (Helsinki, 1939). James Scott, A Corean manual² (Seoul, 1893). H. W. Sunoo, A standard colloquial Korean text book for university students (Seattle, 1944). Tyoosengo-Daiziten [Unabridged Korean–Japanese dictionary] (Seoul, 1920). H. G. Underwood, An introduction to the Korean spoken language² (Seoul, 1914).

The romanization systems appearing in these works are many and diverse; the most consistent are those in the manuals of Home and Lukoff. All of the systems are somewhat influenced, as for teaching purposes they should be, by morphophonemic considerations.

I am indebted to Thomas A. Sebeok for calling my attention to a lengthy bibliography of works mostly pertaining to Korean historical linguistics; this appears in a review of Ramstedt's grammar by Andrej Rudnev, Finnisch-Ugrische Forschungen, Anzeiger 27.55–68 (1941).

³ A component cluster is itself a coextensive sequence of components which is under consideration with respect to its sequential relation with still other components.

⁴ In other words, pause is defined as a phoneme uniquely characterized by the absence of distinctive components and perceived as the interruption of a componential sequence. Many instances which other analysts might consider ‘open juncture’ are thus included as occurrences of the pause phoneme. For example, the expression meaning ‘both’ (literally 'two all') is pronounced sometimes as / # tul # ta # / and sometimes as / # tulta # /; it is also pronounced as / # tultqa # /. The occurrence or non-occurrence of pause in a given utterance or set of utterances—like the occurrence of any other phoneme—is a problem of morphophonemic description. Some analysts seem to perceive pause only when the sequential interruption is accompanied by some pitch change, usually described as 'intonational'; this is encouraged by the frequent interrelation of pause and pitch phonemes on the morphophonemic level (where the analysis of intonations is usually appropriate). On pause as a phoneme, cf. Bloch, Lg. 26.97 fn. 23. On individual variation in pause perception, cf. J. M. Cowan and B. Bloch, An experimental study of pause in English grammar, American speech 23.89–99 (1948).

⁵ To the contrary, with respect to /u/, Haguenauer 148: ‘prononcé presque sans arrondir les lèvres’. This seems more like a description of /ə/.

⁶ Somewhat loosely, the color of /ə/ and // may be described as unrounded [u] and unrounded [o] respectively.

⁷ Cf. Haguenauer 150: ‘Quant à , dans certains cas ... il donne à l'oreille l'impression d'un yee.‘

⁸ The phonetic nature of /pq, tq, cq, sq, kq/ has been variously interpreted. Haguenauer (159 ff.) calls them ‘quasi-geminées’ and Polivanov considers them ‘long’ (La perception des sons d'une langue étrangère, TCLP 4.82 [1931]). On the other hand, the Korean phonetician Jung calls the sounds ‘implosives’ and the late eminent Japanese Koreanist Ogura writes them preglottalized (McCune 29 ff.). All observers seem to agree on the tenseness of the articulation, but differ in their evaluation of its nature and significance. Ramstedt (5–6) points out that the tenseness continues throughout the syllable, even raising the perceived pitch of the vowel. It is true that these sounds are of longer duration than a single consonant phoneme; so are /ph, th, ch, kh/. Although the sounds are not so heavily glottalized as many American Indian consonants, I hear distinct glottal tension. The interpretation as geminate consonants is perhaps influenced by the automatic alternations frequent in the morphophonemic structure.

⁹ Cf. J. Lotz, Journal of the Acoustical Society of America 22.716 (1950).

¹⁰ On the lax nature of these phonemes preceding /h/, cf. Haguenauer 159: ‘Les aspirées du coréen sont en général sonores et plus douces qu'en chinois.‘ (The Peking aspirates are generally considered tense. So are the voiceless aspirates of most Chinese dialects which have hitherto been reported.)

¹¹ The phoneme /c/ is considered an affricate by most writers; Jung, however, evidently thinks it is a palatal plosive (McCune 25)—unless, like Gim, he is using the IPA symbol for a palatal plosive to represent an affricate in a ‘broad’ transcription. Haguenauer (258–8) describes the consonant as follows: 'le dos de la langue touche le bas des dents de la mâchoire supérieure pendant que l'apex est placé derrière les dents d'en bas. Il paraît bien y avoir occlusion synchronique du praedorsum. ... certains sujets placent la point de la langue à la base de la région alvéolaire de la mâchoire supérieure ...' He describes /s/ (156) as 'une fricative douce qui est prononcée la langue reposant à plat et sans tension; l'apex est en arrière de la face interne des incisives inférieures. L'expiration est moyenne; la bouche n'est qu'entr'ouverte.'

¹² For contrary findings, cf. E. D. Polivanov, Zur Frage der Betonungsfunktion, TCLP 6.80 ff. (1938): 'Und schliesslich hat im Koreanischen ein Wort innerhalb des Satzes Endbetonung, doch im Satzschluss sowie in isolierter Aussprache Anfangsbetonung.! This was paraphrased by N. S. Trubetskoy, Grundzüge der Phonologie, TCLP 7.246 (1939): 'Nach E. D. Polivanov soll im Koreanischen jedes Wort die Schlussilbe betonen, und nur das letzte Wort im Satze betont die Anfangsilbe.' This would seem to present a phonological criterion for defining the word. Unfortunately, my own data do not confirm Polivanov's findings, and accordingly my criteria for words must be primarily non-phonological.

¹³ On phonetic and phonemic syllables, cf. JAOS 64.151–5 (1944), and SIL 3.46–50 (1945).

¹⁴ These are the principal cases of fully automatic alternation in the morphophonemic structure. (Other cases involve similar reductions of consonant clusters which do not occur phonemically). By fully automatic, I mean what Wells calls ‘narrow static alternation’, Lg. 25.107 (1949). The somewhat more complicated alternations of /l/ and /ny, ni/ are also cases of narrow static alternation after pause but not in other environments (e.g. word-initial within an utterance), which exhibit what Wells calls ‘wide static alternation’ (105–6). The difference is basically whether the automatic character is phonemically determined ('narrow alternation') or grammatically conditioned ('wide alternation').

¹⁵ I.e. a syllable ending with a vowel.

¹⁶ I.e. a syllable beginning with a vowel or with /y/ or /w/.

¹⁷ It would be possible to eliminate the quality Mid by regarding the mid quality of /e, ö, o/ as absence of both high and low qualities (just as the quality Back is treated as absence of front quality). This would define the phoneme // as having no distinctive qualities at all. While such a view is possible, it seems wiser to reserve the definition by total absence of qualities for the phoneme of pause /#/.

¹⁸ Cf. R. Jakobson and J. Lotz, Notes on the French phonemic pattern, Word 5.151–8 (1949); R. Jakobson, On the identification of phonemic entities, Travaux du Cercle linguistique de Copenhague 5.205–13 (1949). For clarity, where Jakobson and Lotz use the terms ‘higher and lower saturation’ I speak of ‘back and front’; where they use the terms ‘gravity and acuteness’ I speak of ‘low and high’.

¹⁹ Cf. J. Lotz, Journal of the Acoustic Society of America 22.715–6 (1950): ‘Thus it seems that speech in its distinctive aspect is built up entirely of features in accordance with a binary principle.‘

²⁰ Ibid.:‘ ... in such a case we have to introduce either a complex middle term or allow trinary oppositions.‘ As a matter of fact, Jakobson and Lotz seem to admit still another contrast, that of applicability of a specific opposition to a particular phoneme. This may then be conceived as a double-level binary contrast: for a given phoneme, a specific opposition either occurs (i.e. is applicable) or does not occur; if it occurs, either (1) both members occur together, or (2) a secondary level of contrast is involved, that of the polarity of the two members. In other words, occurrence of either member or both members of the opposition is itself opposed to non-occurrence, and the differentiation within the opposition is relegated to a secondary level. I suspect this concept of the opposition theory may be in conflict with unstated assumptions on the part of some who support the theory.

²¹ This principle of economy in the selection of positive entities is implied but usually not explicitly stated in analyses of phonemic systems.

²² This represents a departure from the technique applied to French by Jakobson and Lotz (see fn. 18). Whereas, in their treatment, certain oppositions are not represented in certain phonemes at all (so that the number of distinctive features is not necessarily the same for all phonemes), I here assume that every one of the six oppositions is represented in every Korean phoneme, by its positive or its negative member, or by both together.