Search

This article presents the results of an acoustic study of nasal assimilation and gestural overlap at word boundaries in Korean and Korean-accented English. Twelve speakers of Seoul Korean recorded phrases containing obstruent#nasal and obstruent#obstruent sequences in both Korean and English. Nasalization of the word-final obstruent, predicted by the rules of Korean phonology, occurred in 93% of obstruent#nasal sequences in Korean and in 32% of such sequences in Korean-accented English, a rate of application higher than that reported in most other studies of external sandhi alternations in nonnative speech. Acoustic analysis found categorical nasalization in the L1 Korean productions, but both categorical and gradient nasalization, along with a high degree of inter- and intraspeaker variation, in the L2 English productions. For a subset of speakers, there was a significant correlation between quantitative measures of nasalization in English and measures of consonant overlap in the English obstruent#obstruent sequences. An analysis in terms of articulatory gestures and the coupled-oscillator model of speech planning is supported. The analysis is based on the articulatory phonology model (Browman & Goldstein 1990a,b, 1992, 2000, Goldstein et al. 2006), though with modifications. Implications for phonetic and phonological representations, and for speech planning in both L1 and L2, are explored.

Comaltepec Chinantec tone sandhi normally consists of rightward high tone spreading from syllables with a low-high tone pattern, and almost all sandhi outputs are non-neutralizing. This sound pattern may be understood by considering articulatory, aerodynamic, acoustic, and auditory principles, in necessary combination with phonetically rooted historical forces, and the principles of contrast maintenance, economy of effort, and what is termed pattern coherence.*

The study contains a synchronical description of the San Giorgio variety of the Slovene dialect spoken in the Resia valley (Val Resia/Rezijanska dolina) situated in north-eastern Italy. The following linguistic levels are analysed: phonology, morphonology and morphology. Apart from this some remarks on syntax and a lexicon have been included. The first chapter contains an overview of existing descriptive publications on Resian. Taking this overview as a starting point the choice of exactly the San Giorgio variety as the topic of this study is accounted for and the need for not only phonological, but also morphological analysis is made pointed out. The chapter further contains information on the native speakers whose speech is analysed and on the various methods used to obtain the dialect material. In the second chapter the phoneme inventory is presented, along with information on realisations, (optional) neutralisations and sandhi phenomena. Notwithstanding the considerable amount of phonetic detail given, the first and foremost aim of this chapter remains the quest for phonological oppositions and their functioning. In the third chapter the morphonological alternations that occur in the substantive, adjective and verb categories are being treated. Instead of dividing this information over the respective chapters on these categories, the alternations are presented together in a separate chapter, because some of the more frequent of them occur in all these word classes. However, through a classification by accent classes alternations concerning the location of stress are treated together with the word class they occur in. The third through seventh chapter inclusive contain the morphology of the substantive (chapter 4), the adjective (chapter 5), the pronoun, the numeral and the article (chapter 6) and the verb (chapter 7), respectively. In each chapter, together with an inventory of the attested desinences, an overview is given of rare desinences, of irregular declinations/conjugations and of the distribution of alternative desinences.

and smelt, and Gk. riyos and arTyos, both meaning 'roof'. Several decades ago s movable was much discussed. Some scholars explained it by a sandhi alternation in which initial s before a consonant was lost after a final consonant of the preceding word, with later leveling and perhaps analogical spread of s to roots which had not originally had it. Others thought the s the remnant of a prefix whose meaning had become obscured. Our reference books usually leave the matter there, as one of the minor uncertainties of IE phonology. On the whole, the champions of the prefix theory, like Siebs,1 had the stronger case; but there was one major difficulty. If s movable represents a prefix, one would expect to find it not only before stops and sonants but also before vowels. For this, Siebs could find few plausible examples.2 Today of course the question appears in a different light. Hittite examples like istanta(e)'tarry' beside titta'cause to stand'3 seem to prove what was probable anyhow, namely that the alternation existed in Indo-Hittite-or, if one prefers, before the general disappearance of the laryngeals in the non-Hittite branches of Indo-European. Where Siebs expected simply s plus a vowel, we must therefore reckon with a sequence s plis laryngeal plus vowel; for we may be certain that no IH word began with a vowel. Unless a laryngeal after initial s gives exactly the same result as initially, Siebs' failure to find substantial evidence may conceivably be due to the fact that he was unprepared to deal with the effects of laryngeals on their environment. Consider the following etymologies: (1) Skt apn6ti 'reaches'; Lat. apiunt 'they fasten', aptus 'handy', co-jpi 'I began'; Arm. unim 'possess'; Hitt. epzi 'he takes, begins' Av. hapti 'holds in his hands'; Skt. sdpati 'worships, is attached to'; Gk. 4-rw 'get done', 6irXov 'tool, weapon'. (2) Skt. r with drna'flowing, flood'; OHG rinnan 'flow'; Skt. pf. r Hitt. arer'arrive' Skt. sdrati 'hastens, flows', sard'liquid'; Gk. (Ion.?) op6s and Lat. serum 'whey'. (3) Gk. apros 'protection', apKow 'aid, suffice'; Lat. arx 'fortress', arce6 'ward off'; Arm. argel 'obstacle'; Hitt. larkanzi 'they have' Gk. fpKos 'wall, pro tection', by some connected with Lat. sarcia 'mend'. (4) OHG ano 'ancestor', ana 'ancestress'; OPr. ane 'grandmother'; Lat. anus

In the treatment of IE s,1 the insular Celtic languages share in general the peculiarities of Iranian and Greek, where the s-sound was characteristically changed into an aspirate in certain specified positions, and was further liable to disappear. In Irish, however, initial s remained (lenited 4, eventually h).2 The rule for insular Celtic may be stated briefly as follows: (1) Initially before a vowel, IE s remained in Irish and became h in Brittonic. (2) Between vowels, it was eventually dropped in Irish and became h in Brittonic, but was lost here also if the preceding vowel was stressed (cf. further below). (3) In final position, it probably resulted in aspiration, as appears from later sandhi forms, although no trace of it appears in the orthography of the older insular dialects.3 In other cases, such as before or after another consonant, the s-sound as a rule remained or was fused with such a consonant into a new sound; e.g. -st> -ss-, -sl> -ll-. No other development of IE s has, to my knowledge, been accepted by anyone writing on Celtic phonology; I disregard at the moment such cases as the change of s to lenited d (later written dh, Welsh dd) before certain voiced consonants, such as g and b (OIr. medg, odb, etc., see TH' 217). It is, however, quite apparent that in a number of cases s has undergone a very special evolution, or at least has had a distinct effect on a preceding vowel. This is evident especially in Welsh. Already Pedersen derives Welsh oedd (Cornish o, Breton oa) 'was' from *esat (cf. Latin erat; VG I, p. 73), and, in the same way, the plural suffix Welsh -oedd (nef-oedd 'heavens', glynn-oedd 'valleys') from Celtic *-esa (cf. Latin gen-era, Greek, 'yv-ea; VG 442).4 According to Pedersen (VG I, p. 73), es in the Celtic penult before a would be treated like ij in the Celtic antepenult, but on account of many irregularities this general statement is of but little help. In Irish the Celtic long , except in the archaic

Syntactical juncture, the pronunciation of words in a close-knit phrase with the allophones appropriate to the pronunciation of a single word, is a characteristic and well-known phenomenon in Spanish. However, so far as we know, there is no precise or extensive statement about which elements of an utterance are usually connected by close juncture. Tomas Navarro Tom6s, it is true, has written on the subject: El grupo f6nico como unidad mel6dica, Revista de Filologia Hispinica 1.3-19 (1939); Manual de entonaci6n espafiola 41-59 (New York, 1944); Estudios de fonologia espafiola 77-101 (Syracuse, 1946); and his remarks, stimulating and valuable as always, will be considered later. But his approach to the phenomenon in question differs from ours. We should like to suggest a special technique, the application of objective criteria to the Spanish phrase system, and to supply some basic conclusions.' Objective criteria rather than mere auditory impression should determine whether two elements are in close or in open juncture. Spanish offers such criteria in its system of sandhi. One example will show what we mean: whether the two words unamadre 'a mother' are in close juncture is merely a question of one's auditory impression; but that the words umbaso 'a vessel' are in close juncture is shown by objective criteria: the labial nasal before the labial, and the stop after the nasal. If we can prove that in all or in almost all instances where objective criteria exist, the members of two given form-classes are in close juncture, we may safely draw the conclusion that other members of the same two form-classes are also in close juncture in those more numerous instances where objective criteria are lacking. Our procedure was briefly the following. We marked in a Mexican play all places where close juncture is theoretically possible. We then had a Mexican woman from Oaxaca, of average education but of good intelligence, read the whole play aloud, and checked her pronunciation for signs of open or close juncture. (This procedure, dictated by practical considerations, has the obvious disadvantages connected with reading as against spontaneous talk, and with the use of only one informant as against the use of several.) The criteria of close juncture established for Castilian Spanish turned out to be not quite identical with those in the speech of our informant. In her speech, [e] is not open in the diphthong [ej]; [o] is not open in a post-tonic syllable; the labiodental nasal is not used before [f]; [s] is not voiced before [1]; and of course the changes connected in Castilian with the voiceless interdental fricative [0] (replaced in Mexican Spanish by [s]) do not appear. The criteria that could be applied are the following

If we reconstruct a pronoun of the first person singular as *H1eml or *Hlm (with zero vocalism), such as may be attested by the Greek ernme or the Old Irish mo(both with a vocalic suffix), and put it into various possible syntagmatic phonetic environments, the regular phonetic developments will account for the following forms, all of which are attested as verbal endings of the first person singular: m, m, H1, o:m, o:, m:m, om, and e. This hypothesis, namely that the verbal endings of the first person singular are derived from sandhi forms of an originally distinct morpheme of the personal pronoun, may be most easily understood in the form of a chart. (Column A: formula for Word or Stem 1; column B: pronoun of the first person; column C: Word or Stem 2; Column D: later development of the pronoun as an ending of Word 1; column E: forms by which the pronoun is attested in various languages.) A B C D E (1) CCVC H1m CCVC m Skt., Hitt., Gk. -m(i) (2) CCVC H,m mCVC mi Gk. -a (3) CCVC H1m mVCC H1 Hitt. -h(t) (4) CCV Hlm CCVC o:m Aryan -am(i) (5) CCV Him mVCC o: Gk., Lat. -o, Goth. -a (6) CVCm H1m CCVC m:m Gk. -man (7) CCVC Hle/om CCVC om Gk. -on (8) CCVC Hle/om mVCC e Skt. -e < *-e(i) Column A represents the word or stem which precedes the pronoun. All the examples are given as Benveniste's Themes I and II; lines 4 and 5 may be regarded as Theme II with zero consonant in final position. They are forms well known to Indo-Europeanists, and as such are generally acceptable. Actually, only the last two phonemes of the word or stem preceding the pronoun are relevant, since they determined the treatment of the phonemes of the pronoun. The assumption of different types of morphemes preceding the pronoun would not change the central hypothesis. Column C represents the word or stem which follows the pronoun. Here again, although I have indicated Benveniste's Themes I and II, only part of the word is relevant, in this case the first two phonemes. Column D gives the regular form which the pronoun would take in the phonetic environment represented, if word boundaries were disregarded.

The linguist who works with older stages of a language is faced with serious problems in applying descriptive techniques. Perhaps one of the greatest difficulties lies in the lack of adequate information concerning stress, pitch, and juncture.l It is then of considerable interest if on rare occasions we find some trace of these suprasegmental features in a manuscript or a printed text. The material to be presented here deals with one such trace in Old Icelandic. Manuscript NRA 52 is one of the earliest in Old Icelandic. Unfortunately it is preserved only in six fragments;2 but enough text is available to permit analysis, though the limitations placed upon it by extralinguistic conditions must be borne in mind. The Old Icelandic preposition which is usually normalized as um shows the variant forms um and umnb. Since both forms occur in several of the oldest manuscripts, this is probably an allomorphic variation. In NRA 52, as I wish to show, the variation is phonologically conditioned, and one of the conditioning factors is a type of juncture. As the basic form of the morpheme I choose umb. Historically umb (the b-form, as opposed to the m-form) is the older; cf. OHG umbi, OS umbi, OE ymb(e). Only the b-form is found in the earliest manuscripts; in slightly later ones the m-form occurs beside the b-form or to the exclusion of it. In NRA 52 umb occurs seven times. I give the occurrences in unnormalized form, and number them for ease of reference. The translations are fairly literal.3 (1) ... .oc r%Sba vm me) s.r. ... 'and (they) speak about (it) among themselves' (6.7-8) (2) ... .oc ma tat verb?a enn til vmbdta nacqvarra of hitt malit. 'and that can then still be some improvement concerning this affair' (11.24-5) (3) ... bibi vmb orb oc atgon[go ...] 'both concerning (his) words and (his) help' (5.10-1) (4) [bat e]r vmb at r,Sba ... 'that is to be spoken about' (9.1) (5) ... er sigldi vmb oss fyrr ... 'who sailed past us before' (10.10) (6) ... .oc ma ecci of pat r4ba.eftir bv[i sem var.at eigi ero] sva vitni vmb.bau Qll er t0cilig ero. 'and (one) can not talk about that-accord