An Introduction to Indonesian Linguistics/Essay 4

From Wikisource
Jump to navigation Jump to search

translated by Charles Otto Blagden (1916)




(The original was published in 1915.)


1-38. Section I : Fundamental Considerations.

39-66. Section II : Enumeration and Description of the Indonesian Sounds. 39-40. The Original Indonesian Phonetic System. 41-3. The Phonetic Systems of the Living Languages, compared with that of Original Indonesian. 44-6. Fixed and Varying Pronunciation. 47-50. Full and Reduced Pronunciation. 51-66. Preciser Description of the several Indonesian Sounds.

67-88. Section III: Quantity and Quality of the Vowels, Doubling of the Consonants. 67. Quantity in General. 68-72. Quantity of the Accentuated Syllable in Words of more than One Syllable. 73-5. Quantity of Vowels in Monosyllabic Words. 76-9. Quantity of Unaccentuated Syllables. 80. Quantity in Old Javanese. 81. Quantity in Original Indonesian. 82. Quality of the Vowels. 83-8. Doubling of the Consonants.

89-118. Section IV : Phonetic Laws of the Simple Sounds, summarily stated. 89-90. Preliminary Observations. 91-6. Laws of the Vowels. 97-8. Laws of the Semivowels. 99-101. Laws of the Liquids. 102. Laws of the Laryngal q. 103-5. Laws of the Velars. 106-8. Laws of the Palatals. 109-111. Laws of the Dentals. 112-4. Laws of the Labials. 115. Laws of the Spirant s. 116. Laws of the Aspirate h. 117-8. Laws of the Simple Sounds in Indo-European and in Indonesian. 119-56. Section V : The most important Indonesian Phonetic Laws, set forth in detail. 119-20. Preliminary Observations. 121-8. The Pěpět-Law. 129-39. The RGH-Law. 140-53. The Hamzah-Law. 154-5. The Law of the Mediæ. 156. Comparisons with Indo- European.

157-79. Section VI : The two most important Phonetic Combinations and their Laws. 157-9. Preliminary Observations. 160-76. Tlie Diphthongs and their- Laws. 177-9. The Aspirates and their Laws.

180-224. Section VII : Special Phenomena of Initial, Interior, and Final Sounds. 180. Preliminary Observations. 181-6. Initial, Medial, and Final Enunciation. 187-92. The Initial. 193-9. The Medial. 200-11. The Final. 212-23. The Final in Rottinese. 224. Comparison with Indo-European.

225-60. Section VIII : Certain Special Classes of Phonetic Phenomena.

261-65. Section IX : Phenomena connected with the Aggregation of Sounds into Syllables.

266-73. Section X : Phonetic Phenomena connected with the Combination of Word-bases with Formatives.

274-80. Section XI : Abbreviation of Words.

281-90. Section XII : Phonetic Phenomena in Loan- words.

291-306. Section XIII : Phonetic Phenomena in the Sentence.

307-378 Section XIV: Accent. 307-8. In General. 309-15.Accentuation of the Word-base. 316-9. Accentuation of Derivatives from the Word-base. 320-2. Accentuation of Doubled Words and Compounds. 323-5. Accentuation of the "Complex": i.e., Word of Substance + Word of Weak Stress. 326. Accentuation of Loan-words. 327-8. Quality of the Accent. 329. The Unaccentuated Syllables. 330. Original Indonesian Accentuation. 331-2. Comparison with the Accent of the Indo-European Word. 333-7. Sentence Stress.

338-41. Section XV : Lagu.

342-51. Section XVI : As to the Invariability of Phonetic Laws.



1. The present monograph is a dehneation of the phonetic phenomena of the IN languages.

Note. — As to the method of transcription, see § 39, as to the abbreviations, § 38.

2. Up to the present no comprehensive work on this subject has appeared, but a sufficient quantity of material for such a work has been published in the shape of IN grammars and vocabularies and a number of treatises. I shall not enumerate these sources and prelitninary works here, because I intend to refer to them in detail in my “Geschichte der IN Sprachforschung” which is to appear shortly. — The works of my predecessors have furnished me with a relatively small part of the materials, either rough hewn or more or less worked up; the greater part has been collected by myself. In its whole plan, as well as in the execution of the individual sections dealing with the subject from various points of view, my monograph takes its own independent line.
3. I have to delineate the IN phonetic phenomena of the past as well as those of the present time. The past history of IN sounds can be gathered from the written documents handed down to us, or it can be deduced by the usual methods of linguistic science, above all by the method of comparison. On account of its heritage of written documents dating from former periods, Javanese is of special importance for the study of IN phonetics; Bugis, Sundanese, Malagasy, and some other tongues, are of much less moment.
4. For our deductions we often require a basis to start from; and that basis is Original IN. In this matter I follow the same procedure as Brugmann in his “Kurze vergleichende Grammatik der indogermanischen Sprachen” . Just as IE comparative research has inferred from the Sanskrit dhūmás, Latin fumus, etc., an Original IE dhūmós, “smoke”, and as Brugmann (KvG, § 85), in deahng with the vowel ū, proceeds from this dhumos and other such Original IE words as by the same method have been shown to contain the vowel u ; so from Hova telu, Toba tolu, etc., there follows an Original IN tĕlu, “three”. This tĕlu, together with the other words in which an Original IN ĕ has been inferred, serves us as a point of departure for the discussion of the sound ě and its derivatives.
Note. — By far the greater part of the IN words occurring in this monograph have the accent on the penultimate syllable. In that case I do not mark it, and accordingly write telu, tolu; on the other hand, in § 5 I write talǒ, because this Pangasinan word is accentuated on the final syllable. For the reasons given in § 330 I cannot indicate the accent in the reconstructed Original IN words. — As regards quantity, see §§ 67 seqq.
5. I will now demonstrate by an individual example the nature of the method by which I reconstruct the Original IN forms.


“Original IN possessed a neutral, colourless vowel, styled in Javanese, and accordingly also in IN comparative linguistics, the pĕpĕt, which is represented (not very aptly) by the symbol ě, and occurs for example in the Original IN word tĕlu, ' three” .


I. As “three” in Pangasinan is taló, in Hova telu, in Sundanese tilu, in Toba tolu, in Tinggian tulu, the variegated character of the vowel of the first syllable can be most satisfactorily explained as a case of differentiation from a neutral original, just such as the pĕpĕt.
II. The pĕpĕt still actually exists, even though in a minority of the IN languages, yet in the most diverse local areas of the family. Thus in Karo in Sumatra, in Balinese next to Java, in Tontemboan in Celebes, etc., the word for “three” is tĕlú.

III. Old Javanese likewise has tĕlú; and how important Old Javanese is will be shown in § 6.

IV. Nias has no pĕpĕt; where other tongues have ĕ, Nias has an o. But this o has a peculiar pronunciation, it is articulated further back in the mouth than the o of a different origin. If I represent the front o by o1 and the back one by o2, I get (for example) the equations: Nias bo2li, “price” = Original IN, and likewise Gayo, Malay, etc., bĕli, but Nias o1no1, “child” = Original IN, and likewise Old Javanese, Tagalog, etc., anak. The peculiarity in the articulation of the o2 accordingly points to an originally peculiar sound, in fact to the pĕpĕt.

V. Iloko knows no pĕpĕt; where other languages have a pĕpĕt, Iloko puts an e. But the consonant, which immediately follows this e, is doubled ; thus the equivalent of Original IN, and likewise Old Javanese, Malay, etc., lĕpas, “free”, is Iloko leppás. This doubling of the consonant does not occur after an e of any other origin. Now Madurese says lĕppas; it also doubles the consonant, but leaves the ĕ unchanged. If we compare the Mad. procedure with the Iloko, it follows that the Iloko e, after which the consonant is doubled, points back to an original pĕpĕt.

VI. Talautese lacks the pĕpĕt; an a occurs where other languages have ĕ. But after this a the liquid l is articulated differently than it is after an a which descends from an Original IN a. So in Tal., too, we have an indication of the existence of the pĕpĕt in Original IN.

VII. Hova possesses no pĕpĕt; for an ĕ of other languages it puts in an accentuated syllable an e, in an unaccentuated one an i. Original IN, and likewise Karo, etc., tĕlĕn, “to swallow”, has therefore in Hova the equivalent télina. Now before this i < ĕ Hova preserves Original IN l unchanged, whereas before a Hova i < Original IN i it becomes d; thus tĕlina < Original IN tĕlĕn, but dimi, “five” < Original IN lima. Here again, then, we find in a language whicli itself has no pepet an argument for the existence of the pĕpĕt in Original IN.


The evidence of I.-VII. supra, to which many other testimonies could be added, shows conclusively that the phonetic system of Original IN must be credited with the vowel called pĕpĕt.

6. The phonetic conditions of Old Javanese coincide in most cases with the phonetic system of Original IN as inferred by the comparative method. Hence we get, from objective documents, a confirmation of what has been attained merely by inference. — To this harmonious agreement there are two exceptions :

I. Original IN r2 (= uvular r) disappears in Old Jav.; hence Old Jav. atus, " hundred ", from Original IN r2atus.

II. Original IN successive vowels are often contracted in Old Jav. Original IN, and also Malay, etc., disyllabic lain, "other", becomes len in Old Jav.

7. It is not possible in IN linguistic research, any more than in IE, to discover the corresponding original values of all the phonetic phenomena of the living languages. Some IN languages possess the sound called hamzah; but, as stated in § 40, I am not at present in a position to decide with absolute certainty whether it should be ascribed to Original IN.

8. Between any of the phonetic types that exist to-day and its corresponding archetype in the original mother- tongue, there may have been intermediate stages. IE hnguistic research possesses the means of determining such intermediate stages in many cases. Thus Kluge, in his "Etymologisches Worterbuch der deutschen Sprache", s.v. Met, shows that between Original IE mĕdhus and the Modern German Met (= mēt), " mead ", we have to intercalate Original Germanic mëdus. Old High German mëto, and Middle High German mët as intermediate forms. IN research possesses such means to a much more limited extent, seeing that it has only got one single language with really important written records of some antiquity, namely, Javanese; besides which, Old Jav. mostly exhibits the same phonetic state as Original IN. Nevertheless, IN research is in some instances able to detect such intermediate forms, and the cases where it is possible may be taken to fall pretty much under the following heads:

I. The intermediate form is found in Old Jav.:

Original form Intermediate stage Final result
Original IN Old Jav. Modern Jav.
dir2yus dyus adus, "to bathe"

II. The intermediate form is represented by the native spelling:

Original form Intermediate stage Final result
Original IN Written Minangkabau Spoken Minangkabau
sĕlsĕl sasal sasa, "to regret"

III. The intermediate form exists in a cognate dialect:

Original form Intermediate stage Final result
Original IN Tunong-Achinese Achinese Proper
batu batew batee

IV. The intermediate form can be ascertained by inference. If Original IN bar2a, "glowing embers", results in Bungku wea, we must assume a form waya as an intermediate stage: see § 136.

9. We often have reports that the older living generation adheres to an older phonetic stage, while a newer phonetic type has developed in the speech of the younger people. In Kamberese Original IN s becomes h, e.g. Kamb. ahu, "dog", < Original IN asu, but "one often hears old people pronouncing the s" (Wielenga).

10. Phonetic changes either take place unconditionally or are dependent on definite conditions. Original IN pĕpĕt unconditionally, in all cases where it occurs, becomes e in Dayak; thus Original IN tĕkĕn, "staff", results in Day. teken. In Hova, which, betrays a somewhat near relationship with Day., the pĕpĕt only becomes e when it bears the accent; hence Original IN tĕkĕn produces Hova tĕhina. Accentuation, therefore, is the condition for the change of ĕ into e in Hova.

11. The condition under which a phonetic change takes place is one thing, and the cause which calls it into existence is another. The conditions are very often recognizable in IN, but as regards the causes the same observation applies to IN as Hirt, in his “Handbuch der griechischen Laut- und Formenlehre” ", § 71, made about Greek: “We are often unable to detect the causes of phonetic change”. Nevertheless IN linguistic students have set up many a theory on this subject, and I here repeat some of them, without commenting thereon: “A peculiarity of certain of the Toraja languages is the change of s into h. It appears to us that the custom of filing the teeth quite short or partially knocking them out, may be the cause of this phonetic change” (Adriani). — In Karo, Original IN a remains a, but alongside of jah, “yonder”, a form joh has appeared, “in consequence of a movement of the lips, with which one indicates the direction ‘yonder’ ” (Joustra). — “The custom of chewing betel explains why the Javanese often pronounce a velar instead of a labial, e.g., kĕstul for pĕstul, ‘pistol’ ” (Roorda).

12. In the evolution of IN sounds a number of other forces bear sway, which operate in the way of influencing,furthering, hindering, crossing, etc., though they cannot be called “causes” or “conditions” in the strict sense. These are analogy, popular etymology,* the tendency towards differentiation, phonetic symbolism, onomatopoeia, euphem- ism, and the tendency towards disyllabism.

13. Analogy plays as great a part in the phonetic evolution of the IN languages as it does in the IE family. Thus- among the IN, as among the IE languages, there is hardly one in which the numerals have not been affected by its influence;

* [The tendency which produces forms like “Hobson-Jobson”, “sparrow-grass”, etc.] cf. Paul, “Prinzipien der Sprachgeschichte”, under the heading “Kontamination”.

In Original IN, “hundred” is r2atus and “thousand”, r1ibu, but for “thousand” Bajo says ribus, having transferred, to it the s of r2atus.

14. Popular etymology also has the same importance in IN as in IE. Persian lāzuwerdi, “sky blue”, becomes in Javanese rojowěrdi, in imitation of rojo, “king”, as if it meant the royal colour. —Particularly frequent is the occurrence in IN of a species of popular etymology which I will style grammatical popular etymology. From Sanskrit yoga comes Karo iyoga, “yoke” ". But as i- in Karo is a prefix, it appears to the people who speak Karo as if iyoga were made up of the prefix i + the intermediaiy sound y + oga, and hence they have abstracted out of iyoga a WB oga, which is now employed alongside of iyoga. Or, since ka- is a very common prefix in Old Javanese, the Sanskrit kawi, “poet”, makes the impression of being a derived word, and from it is extracted a WB awi, “to compose (poetry)”, from which in its turn various derivatives are formed, e.g., aiviawian, “poetry”.

15. Tendency towards differentiation. Where the originally single meaning of a word is differentiated, a phonetic differentiation may also be induced, in IN as in IE. Just as in the dialect of Lucerne the Middle High German mësse has evolved into Mäss, “the religious ceremony styled the mass”, and Määs, “an annual feast and fair”, so too the Original IN ulu, “head”, appears in Bimanese as ulu, “formerly”, and uru, “beginning”.

16. Phonetic symbolism. In many IN languages we find phonetic symbolism at work in the duplication of words, as in the Sundanese uncal-ancul, “to hop hither and thither”, alongside of ancul, “to hop”. So too the substitution of a sonant for a surd in the Nias aizo2-aizoo2, “somewhat sour”, beside aiso2, “sour”, and other cases, may be due to phonetic symbolism. On the other hand, I do not share the view that in durative formations such as the Old Javanese mamanah from the WB panah , “to shoot”, the m replacing the p, that is a continuous sound instead of a momentary one, indicates duration; in a former monograph I have given a purely mechanical phonetic explanation of this phenomenon, and I am convinced that all IE scholars will approve of my view.

Madurese exhibits a peculiar phenomenon in connexion with words where the WB is partially reduplicated, after the fashion of los-alos, “ very fine ”, te-pote, “ quite white ”; alos, “ fine ”, is evolved phonetically from the Original IN, and likewise Malay, etc., halus, and similarly pote from putih. But alongside of los-alos, te-pote, we also find forms which have preserved the Original IN vowels, viz., lus-alus, ti-puti, and these denote a still higher or more superlative degree than los-alos, te-pote. In these cases, then, it is the more archaic phonetic type that denotes the higher degree of quality.

17. Onomatopœia exercises its influence on the evolution of the IN sounds mainly in the way of impeding the consistent operation of phonetic laws. It manifests itself in the first place in interjections that mimic a sound. In Minangkabau an Original IN liquid at the end of a WB disappears; Original IN lapar, “hunger”, is also written lapar in Mkb., but pronounced lapa. Further, Original IN final explosives turn into hamzah in spoken Mkb.: thus Original IN atĕp, “ roof ”, > spoken Mkb. atoq. So no r and no p can occur as finals in spoken Mkb. But interjections like gar, “crack !”, dapap, “ plop ! ”, etc., are exceptions to this rule. — Besides these, the operation of onomatopœia is seen in words of substance, mostly in names of animals, which have been formed in imitation of natural sounds. In Tontemboan, in the case of WB's consisting of a doubled root, the final consonant of the first half must as a rule become q: thus Original IN korkor, “ to scratch ”, becomes Tontb. koqkor. But the onomatopceically formed kerker, the name of a species of bird, retains the r in the first half.

18. Euphemism. For reasons of euphemism certain words, especially such as are connected with sexual matters, have been deliberately deformed in the IN languages. A number of these are given in Van der Tuuk's Toba vocabulary, e.g., ilat, disfigured from pilat, "membrum virile". Such deformations usually occur on the analogy of some other, more or less comiected, word: thus Hat, on the analogy oiila, "shame".

19. Tendency towards disyllabism. Whereas the forces thus far mentioned, analogy, popular etymology, etc., operate in IE as well as in IN, the tendency towards disyllabism is exclusively peculiar to IN. Its significance was already recognized by Humboldt in his "Kawisprache", pp. ccccii seqq. The WB's of the IN languages are as a rule disyllabic, and the genius of the IN languages is often impelled to squeeze into this mould such words as are not really disyllables at all or have lost that form in the course of linguistic evolution. Thus the Dutch word lijst, "list", appears in several IN languages under the form eles, with a prothetic formative ĕ which has no meaning or significance; and "Rome", i.e., Constantinople, is called Ruhum in Minangkabau, not Rum.

20. Between the written language and the colloquial in IN there are often phonetic differences. The one of most frequent occurrence is that the colloquial allows abbreviations which are avoided in the written language. Thus spoken Javanese says dulur, "brother or sister", for the written sĕdulur.

21. The phonetic phenomena hitherto described occur in ordinary, normal speech. Besides this we find in IN certain special modes of speech. These are the language of children, the language of animals in the beast fables, poetic language, and various artificial languages.

22. The language of children in IN has the four following characteristics :

I. Substitutioti of one sound for another. "As long as a Bareqe child is unable to pronounce the velars, it regularly replaces them by the dentals; thus it says atu for Original IN, and also Bareqe, aku, 'I'. Small children often pronounce c for s, and accordingly say cucu for Original IN, and likewise Bareqe, susu, 'breast'" (Adriani).

II. Infantile repetition. By the change of a consonant, words are adapted to this type. Thus Bareqe children say jeje for keje, “membrum virile”; Tontemboan children, kiqkiq for kiqciq, “to bite”. In Tontb. titiq for hiliq, “to sleep”, both changes have occurred, viz., substitution of the dental for the velar and also adaptation to the infantile habit of repetition.

III. Transformation of combinations of sounds which are difficult for children to pronounce. Thus Karo children say a-pe for laṅ-pe, “not at all” .

IV. Besides the above, the language of children exhibits other isolated phenomena, which cannot be classed under any general category. Thus Tontemboan children say lileq instead of lěleq, “to bathe”.

23. When parents speak with children, they use either the normal form of speech or the children's language; but they also sometimes make a compromise between the two. In the preceding paragraph, under subsection I, we saw that the Bareqe children use cucu for susu, “breast”. But in normal Bareqe the palatal tenuis only occurs after the nasal, so that forms like cucu do not exist in the speech of adults. On the other hand, the palatal media is not subject to the same restrictions as the tenuis, and so it comes about that parents, when they speak to children, say neither susu nor cucu, but juju.

24. It is not uncommon for childish words to make their way into the language of adults, particularly the forms involving infantile repetition. In Original IN and in most of the living IN languages, “father” is ama, “mother”, ina; but several languages employ the infantile forms mama and nina. In Tontemboan, “grandfather” is apoq, and “uncle”, itoq; but the vocatives of these words are papoq and titoq. — In Bugis the word for “little girl” is běsseq or běcceq, the first form being used only of princesses. According to subsection I of § 22, the form with s is the normal one, while the one with c was originally the infantile form.

25. The phenomena of childish speech recur to a great extent in IE. In certain of the Swiss dialects the word for “father” is Ätti, but other dialects replace it by Tätti, using therefore the form that involves infantile repetition: see “Schweizerisches Idiotikon” , I, 585.

26. The language of animals employs (inter alia) the method of infantile repetition, like the language of children. In the sixteenth tale in Adriani's “Leesboek in de Bareqe taal” , p. 17, 1. 10, the old mouse says kuko, for duṅko “crust of the rice-pap in the pan”.

27. Poetic language. The requirements of rhythm and rhyme produce all sorts of phonetic changes. Certain literatures, it is true, e.g. the Bareqe, do not tolerate such disfigurements, but others put up with a great deal in this respect. Such poetical deformations may be divided into two classes, viz., those which exhibit changes that are still within the limits of linguistic possibility, and on the other hand such as exemphfy deliberately artificial modification.

I. To the first category belongs the poetic licence in Bisaya, whereby i before a vowel may be treated as a consonant, e.g., motya, for the trisyllabic motia, “pearl”. The change of i in this position into a consonant is found in the normal form of many IN languages: the Old Javanese WB ipi, “to dream”, has a conditional aṅipya.

II. To the second category belong the most varied forms of licence, which for the most part are based on no principle. Sometimes they result from metric difficulties. Thus in the Balinese Epic Megantaka, strophe 318, verse 7, we find tos, for totos, “descendant”, because if totos had been used the verse would have had one syllable too many. In the second place, they may be due to difficulties connected with the rhyme. In the Minangkabau Epic “Kaba Sabay nan Aluyh”, verses 446, 447, read: "That we say yes, yes, that we say no, no" = maq kami bario-io, maq kami batido-tido. Here the form tido is a deformation of the normal tidaq, “no”, made to suit the rhyme, which consists in a similarity of both

vowels of the WB's. Thirdly, these changes may be produced by the requirements of the lagu, i.e., the current mode and fashion of reciting. Achinese has (inter alia) a special lagu for the recitation of solemn or tragic poems. In this lagu the several syllables are pronounced very long, and here and there extended into two syllables by pronouncing the vowel twice over with the intercalation of an between the two: for instance, puṅucoq instead of the normal pucoq, “tip”.

28. In IE we also find both kinds of poetic licence, as depicted in the preceding paragraph. If in the Aeneid we have to scan conubjo, that corresponds to motya in subsection I, while the mutilated form navyasā vacas, cited in Wackernagel, “Altindische Grammatik” , I, p. xvii, is parallel to the deliberately artificial deformations of subsection II.

29. In reading aloud, certain phonetic peculiarities also occur. "It is customary at the Javanese Court, in reading out official documents, but only in that case, to aspirate initial vowels, e.g. to say hadalěm for adalěm, 'to dwell'" (Poensen).

30. Artificial languages. In IN there are quite a considerable number of artificial languages: e.g., priestly languages, languages of ceremonious politeness, languages specially used when hunting, thieves' languages, etc. The peculiarities of these artificial forms of speech are lexicographical and morphological, but also phonetic. . From the phonetic point of view two principles in particular are operative:

I. Metathesis. The Toba thieves' language, for example, interchanges the two syllables of the WB, saying therefore tema for mate, “dead”.

II. Analogical transformation. The Dayak priestly language says rohoṅ, “sword”, for the dohoṅ of normal speech, by analogy with rohes, “to slay”. The Javanese language of ceremonious politeness changes kuraṅ, “too few”, into kiraṅ, by analogy with liraṅ, “half”.

31. One of the methods of formation of the Javanese language of ceremonious politeness consists in replacing

various word- endings by -jiṅ or -jěṅ: thus from esuq, “morrow”, it makes enjiṅ and from buru, “to hunt”, bujěṅ. I shall style this mode of formation the jěṅ-type. Now we find isolated representatives of this jěṅ-type in other languages also. Malay has a word anjiṅ, “dog”, Makassar a word tojeṅ, “true”, with e instead of ě. These words do not belong to an artificial stratum of these two languages, but to their normal form of speech. But inasmuch as anjiṅ coexists with the Original IN, and likewise Old Javanese, etc., asu, and tojeṅ with the Dayak, etc., toto, one must assume that anjiṅ and tojeṅ were originally artificial forms, transformations of asu and toto in accordance with the jěṅ-type, and that they subsequently found their way into normal speech and displaced asu and toto. This is an interesting case of the influence of the artificial type of language upon the normal type.

32. The word anjiṅ is genuine Malay, the word tojeṅ is genuine Makassar, they are not borrowed from Javanese, for the simple reason that Javanese does not possess these words. Thus we find the jěṅ-type of word formation as an estabhshed institution in several widely separated languages. Hence we may perhaps be entitled to ascribe this particular mode of artificial word formation even to Original IN.
33. Influence of foreign languages. This influence, it must be admitted, shows itself most strongly in the vocabulary, and only shghtly in phonetic evolution.
I. Phonetic influence of other IN languages. Kulawi changes s into h, and accordingly says tahi, “lake” , for Original IN tasik. “But many of the men, who nearly all know Palu, which has preserved the s, often pronounce the s even now, whereas the women, who for the most part only know Kulawi, regularly use h” (Adriani). In Ruso-Talautese the normal Talautese k of a final syllable is pronounced s, e.g., ápuka, “lime”, becomes ápusa; " but this peculiarity has been steadily disappearing since the settlement in Ruso of a number of people from Niampak, who mock at this idiosyncrasy of the Ruso population" (Steller). The Tojo-Bareqe has partially adopted the accentuation of its neighbour, the Bugis. "“His dwelling-place”" in Bareqe is banúa-ña, in Bugis wanuwá-na; but Tojo-Bareqe under Bugis influence says banuá-ña.
II. Influence of non-IN languages. Madurese had originally no ƒ, but the Madurese have no difficulty in pronouncing the sound and therefore mostly preserve it unchanged in loan-words from Arabic or from European languages, so that we must now include the sound ƒ in the Mad. phonetic system. Bimanese rejects all original final consonants, and treats loanwords in the same way, thus saying asa for the Arabic aṣal, “origin”. "But educated Bimanese often pronounce the final consonant" (Jonker).
34. Influence of school teaching. Tontemboan has changed the Original IN, and likewise Old Javanese, Malay, etc., media g into the spirant ɤ.* “Under the influence of school education, which is given in Malay, the younger generation now uses the media instead of the spirant” (Adriani).
35. The native systems of writing and spelling are of importance for linguistic research in two sets of cases:
I. The spelling of certain languages, particularly in Sumatra, exhibits a more archaic phonetic stage than the pronunciation. IN research establishes that the word for “free” in its original form was lḝpas. Minangkabau says lapeh, but writes lapas; the written language, therefore, has preserved the original final of the word. Such spellings accordingly confirm the conclusions of linguistic comparison.
II. Words that lean proclitically or enclitically on a principal word are in several languages written continuously with it. Thus in the Makassar tale I Kukang, p. 5, 1. 15: “He was always presented (with) money” = nanitanrotanrówimo doweq. Here na, “he”, and mo, an emphatic particle, are written together continuously with the principal word nitanrotanrówi, “to be always presented (with)”. From the point of view of linguistic science this habit must be regarded as correct.

*[See §41, IV, footnote.]

36. For the understanding of IN phonetic phenomena it is absolutely necessary to study texts. Naturally those texts are most satisfactory which mark accent, quantity, sandhi, and the like. One can often get more Hght from the texts than from the explanations of the manuals. For instance, Seidenadel, in his grammar of the Bontok language, gives no theory of quantity, but out of his most conscientiously edited texts we can construct the theory for ourselves. Not infrequently the texts even correct the data of the grammars. Matthes, in his Bugis grammar, § 193, says that the pronoun of the first person ku is abbreviated proclitically, but never enclitically, to u, but in the Budi Isětiharatě, edited by himself, p. 294, 1. 8, we find: “My husband loves me” = He loves me, husband my = na-elóriy-aq woroiwané-u. Moreover for several languages we possess carefully edited texts, indicating accent, quantity, sandhi, etc., but as yet no grammars or vocabularies.

37. Comparison of IN with IE. In this monograph I compare, where it seems to me feasible, the phonetic conditions of IN with those of IE. The idea of comparing IE linguistic phenomena with IN is nothing new. Humboldt and Bopp did it, though with an inadequate comprehension of the IN material. Kern does it with a true insight into both the IN and the IE material, and the critical student is grateful to Kern for his work. But recently certain voices have made themselves heard, denying the desirability of such comparisons. Therefore I must adduce some considerations in support of my point of view.
I. IE research has advanced further than IN, its subtle and highly developed methods can, indeed must, serve as a guide to IN research. For example, many IN scholars classify the IN languages according to the sounds they admit as finals; others have classified them on the basis of their genitive construction, particularly as regards the position of the genitive before or after the principal word. Both systems depend upon a single linguistic phenomenon. In the IE sphere we find (inter alia) a classification of the Germanic languages into East Germanic and West Germanic. But Kluge, “Urgermanisch”, § 146, bases this division not upon a single criterion, but upon a whole series of them, and yet the classification is not accej)ted by all scholars. That sort of thing ought to make IN scholars wake up; either they must discover additional criteria or abandon their classification of the IN languages.

Note. — The classification of the IN languages on the basis of one single linguistic phenomenon would only be reasonable if it were proved that it was the most important, significant, and characteristic, of all linguistic phenomena. But no such proof has been given, either in support of the phenomena of final sounds or of the position of the genitive. For my own part, I do not see why the phenomena of final sounds should be deemed more important than those that affect sounds in the interior of words (see §§ 193 seqq.), or the position of the genitive in relation to the principal word more important than (e.g.) that of the predicate in relation to the subject. In the last few years IN research has devoted an undue amount of attention to the genitive.
II. Conversely, the results of IN linguistic research may also be applied with profit to IE study. For example, in Meyer-Lübke's “Historische Grammatik der französichen Sprache” , I, § 43, the word tante , “aunt” , is explained as having been formed under the influence of the principles of infantile repetition from an older form ante ᐸ Latin amita. This explanation finds its parallel and confirmation in the IN phenomena of our § 22.
III. Students of linguistic psychology make use of IN material, often in fact they seem to prefer it, as a basis for their inferences. But as their own training has been IE, they will be enabled to feel their way with greater certainty into the sphere of IN linguistic phenomena, if these are presented to them accompanied by IE parallels. For I have shown clearly enough in a former monograph* how" even the most

*[See " Prodromus ", § 28.]

eminent students of linguistic psychology may err, when they venture into IN without definite guidance.
IV. Many scholars who compare the vital phenomena of different families of speech, inter alia IE and IN, have it as their aim, either principal or subsidiary, to ascertain what linguistic phenomena should be esteemed as expressions of the higher intellectuality. Without exception, they arrive at the conclusion that the IN languages, as compared with the IE, bear the mark of inferiority. Now if the deductions which led them to that conclusion were unassailable, one would have to submit to them ; but so far as the IN languages are concerned, I will undertake to show that these arguments, also without exception, betray inadequate knowledge, partiality, etc. As regards two scholars, Durand and Taffanel, I demonstrated that in a former monograph.* Let us now take a more recent case. Finck, in his work "" Die Haupttypen des Sprachbaues", p. 94, deals with the structure of the sentence in Samoan, and in connexion therewith, rightly enough, dis- cusses the part which is played by the numerous particles — i.e., prepositions, conjunctions, words of emphasis, etc. — in knitting together the several portions of the sentence. He then arrives at the conclusion that these particles have not the power to weld the Samoan sentence into a unity, and his final verdict is that Samoan does not possess the complete, definite sentence-structure that IE has. This implies a judgment that convicts a language of the IN-Polynesian family of inferiority in an important manifestation of its linguistic vitality. But Finck overlooks the fact that Samoan, like all the languages of the IN-Pol. family, has other means of attaining the completeness, and in particular the definite rounding off, of the sentence, means which can be employed in addition to, or in lieu of, the particles. One such means, for example, is the tonal accentuation of the sentence (see § 335), whereof Finck says not a syllable. And how inadequately Finck — and his authorities — grasped the real nature of these very particles, is drastically illustrated by the way in which he translates the title of the Samoan text selected by him as an
* [See " Tagalen und Madagassen ", §§ 12, 56.]
example. This reads: ’o le tala i fuṅafuṅa, and Finck (p. 86. 1. 12 from the bottom) translates it: "O! (or "indeed") the tale in the sea-cucumber". In reality 'o (in my spelling qo. § 39), which is derived from an older form ko, is a preposition accompanying the nominative (see Kern FI, p. 30, 1. I); and i is a preposition with an extensive (i.e., vague and general) sphere of mneaning which in several IN-Pol. languages happens also to serve for the genitive relation. So Finck has rendered a preposition by an interjection, and has arbitrarily and wrongly translated a preposition of vague and general import by a locative one. — Now if as against this sort of faulty comparison of IN and IE another method of comparison is propounded, which avoids the mistakes of the former and may therefore be termed the objective method, it would follow that the former method could no longer maintain itself. And if our objective method had no other aim or purpose than to cut away the ground from under the feet of that unscientific, unjust mental attitude, which is so offensive to our common sentiment of humanity as well, would not that be a sufficient justification for its existence?

38. In this monograph the following abbreviations (besides such as are obvious) have been used :

IN = Indonesian.

IE = Indo-European.

WB = Word-base.

Brugmann KvG = K. Brugmann, "Kurze vergleichende Grammatik der indogermanischen Sprachen."

Meillet GvP = A. Meillet, "Grammaire du vieux Perse."

Kern FI = Kern, "De Fidjitaal."

Bijdr. = Bijdragen tot de Taal-, Land- en Volkenkunde van Nederlandsch-Indië.

Schwarz- Texts = Tontemboan texts, edited by J. Alb. T. Schwarz. Steller-Texts = The texts in K. G. F. Steller, "Nadere Bijdrage tot de kennis van het Tala- oetsch."

Seidenadel-Texts = The texts in C. W. Seidenadel, " The first Grammar of the language spoken by the Bontoc Igorot."

Tuuk Lb = H. N. van der Tuuk, " Bataksch Lees- boek."

Hain-Tenj = Jean Paulhan, " Les Hain-Teny Meri- nas."*

* [See also Essay II, § 15.]



The Original Indonesian Phonetic System.

39. Original Indonesian must be credited with the following phonetic system :

Vowels a i u e o ĕ

Semi-vowels y w

Liquids r1 r2 l

Laryngal q*

Velars k g *

Palatals c* j ñ*

Dentals t d n

Labials p b m

Sibilant s

Aspirate h

40. Observations on this table of sounds :
I. The two vowels e and o in the hving IN languages are mostly of secondary origin. In a former monograph † I was able to prove their existence as Original IN sounds only in two words, viz., bela," companion, avenger, to share the same fate", and sor, "below".
II. The liquid r1 is a lingual r, while r2 is a uvular r.
III. The laryngal q, also called hamzah, is almost always secondary in the living IN languages. Only in a single case (see § 181) can it with some probability be ascribed to Original IN.
IV. The palatals are regarded by some scholars as not being original; in their opinion they have been evolved from
  • [See also Essay I, § 11, I, footnotes.]

† [See Essay II, § 26.] dentals. But no valid arguments have been advanced against my view, which I supported in a former monograph.*

V. Precisely the same applies to the labial media, which some scholars likewise refuse to attribute to Original IN.
VI. We must not overlook the fact that the picture which we are at present able to draw of the Original IN sounds is very much in the rough. For example, it is certain that Original IN possessed the dental series, but we are not in a position to form any precise view as to whether they were postdental, or supradental, etc.
VII. The symbolization of the pĕpĕt by ĕ is clumsy and misleading, but in general use. It is quite a mistake to represent the hamzah by an apostrophe, since the latter has also to serve entirely different purposes, e.g. to indicate the omission of a sound. The objectionable ambiguity caused by using the apostrophe for the hamzah is plainly shown by such a book-title as "De Bare'e-sprekende Toradja's": here the first apostrophe stands for the hamzah, while the second one serves to separate the sign of the plural from a noun. — For my part, I denote the hamzah by q.

The Phonetic Systems of the Living Languages,

compared with that of Original Indonesian.

41. The modern IN languages exhibit the following peculiarities in phonetics as compared with Original IN.
I. Some languages have lost certain of the original sounds; some more, some less. In Old Javanese, r2 has disappeared. Rottinese has lost the pepet, the palatals, and r, and has got y and w only in interjections.
II. Some languages have created new sounds; thus Hova has created the spirants f and z.
III. Some languages have lost certain of the Original IN sounds, but have formed them again out of other sounds. Original IN h has disappeared in Hova, hence Hova, fulu <

  • [See Essay II, §§ 37-40.] Original IN puluh, "ten", but h has again been evolved

from k, hence Hova hazu < Original IN kayu, "tree".

IV. The sounds found in living IN languages, which cannot however be ascribed to Original IN, are :

The modified ("Umlaut") vowels ä ö ü.*

The nasalized vowels.†

The cerebrals.‡

The spirants y, χ; š, z; f. §

42. Some of the IN languages possess some sound or other in two distinct shades; thus Nias has two o’s (see § 5), Talautese two l’s (see § 5); Original IN had two r’s (see § 129).
43. Sounds with unusual articulation, i.e., such as rarely occurs in human speech in general, are scarce in IN. Busang has a labio-dental b, formed by the contact of the lower lip with the upper teeth. But has an h formed by expelling the breath through the nose.

Fixed and Varying Pronunciation.

44. Some of the IN languages have a constant pronunciation of their sounds, others exhibit variations in some sound or other. In the Philippine languages "i is often not to be distinguished from e" (Scheerer). In Dayak "the sound of o varies between o and u, indeed the same person in uttering the same word will pronounce the sound sometimes more like an o, and at other times more like a u" (Hardeland). Probably Bontok exhibits the extreme of arbitrariness in this respect; thus (inter alia) in the short story entitled Rolling in Seidenadel-Texts, pp. 555 seqq., one and the same narrator pronounces the word for "then" sometimes isaed and sometimes išaed (see Rolling 1 and Rolling 10).
45. Such varying pronunciation may be a preparatory step towards certain phonetic changes. Dayak is somewhat
  • [Pronounced as in German, or nearly so.]

† [As in French.]

‡ [As in Sanskrit, and some other Indian languages.]

§ [See § 65 ; y is the voiced sound corresponding to the unvoiced χ.] closely related to Hova, and it is to be observed that in Hova the sound o no longer varies between o and u, but has become completely identified with the latter, so that Hova no longer possesses any o at all.

46. The varying pronunciation of sounds also occurs in certain IE languages. Thus Finck in his "Lehrbuch des Dialekts der deutschen Zigeuner", § 1, note 4, notices a case of variation between w and b.

Full and Reduced Pronunciation.

47. In some IN languages certain sounds are pronounced not in their normal, full form, but in a weak, reduced form. In Bontok, final g, d, b "are often scarcely audible" (Seidenadel). In Gayo "in the combinations ng, ñj, nd, mb the media is so very much weakened in pronunciation that in many cases it is impossible to make out whether it is present at all" (Hazeu). In Hova "final vowels are on the point of disappearing altogether" (Rousselot).
48. This weak pronunciation is displayed particularly by such furtive vowels as the Minangkabau ǎ in such a word as púluǎh (disyllabic) < Original IN puluh, "ten"; by vowels that owe their existence to the principle of the repetition of sounds mentioned in § 232, like the y in the Hova phrase ari gyaga, "and is surprised ", for ari + gaga; by such sounds as merely serve to separate or link together two vowels, like the w in Bugis wanuwa, "land", for which some other languages say wanua. The weak pronunciation of the last-named class of sounds is reflected in the varying spelling of the manuscripts, which sometimes write and sometimes omit the corresponding letter. In the Bugis tale Paupau Rikadong the phrase "to the child" = ri + anaq is written riyanaq (p. 4, 1. 4) and rianaq (p. 10, 1. 18).
49. Weak pronunciation is the preparatory step towards complete disappearance. Thus the media after the nasal, which as mentioned in the preceding paragraph is weakly pronounced in Gayo, has disappeared altogether in certain other languages, e.g. in Rottinese; hence Rot. tana, "mark" < Original IN tanda.
50. Reduced pronunciation of certain sounds is also found in IE. In Latin n was weakly pronounced before s, e.g. in mensa (see Sommer, "Handbuch der lateinischen Laut- und Formenlehre", § 136). Here too reduced pronunciation is a preparatory step towards complete disappearance, hence the Romansch form mesa.

Preciser Description of the Several Indonesian Sounds.

51. In the following I give a somewhat more precise description of the several IN sounds, so far as seems to me necessary and sufficient for the purposes and aims of the present monograph.
52. Vowels. These will be described in greater detail in the following Section, with reference both to their quantity and quality. Only the pěpět will be discussed here.
53. I. The pure pěpět. "The Javanese pěpět is the indeterminate vowel, the sound of the voice when the mouth is not put into any particular position so as to form a definite vowel like a, i, etc." (Roorda). The shape of the mouth-cavity in pronouncing the Madurese pěpět is "the same as in ordinary breathing" (Kiliaan).
II. The modified pěpět. In this the articulation inclines somewhat towards the position of a, or i, or u. "The pronunciation of the Bugis ě partakes somewhat of the sound of a" (Matthes). In Old Javanese the articulation of the pěpět must have approximated somewhat towards the position of u, for it changes into u when, after the loss of a consonant, it happens to stand before a vowel; hence Old Jav. bwat for běat < Original IN běr2at.
This shade of a, i, or u is the transitional stage to the perfect a, i, or u. In Bugis the pěpět has the shade of a, while in Makassar, which is very closely related to Bugis, it appears as a perfect a. III. The fleeting pěpět. In some languages the pěpět shares the characteristics of the other vowels: it can occur in long and short form, accentuated and unaccentuated. In other languages, e.g. in Tontemboan, it only appears as a short vowel. Or else, as in Gayo, it cannot carry the accent: hence Gayo túluk, "to verify", but "tělúk", "bay".

This fleeting character of the pěpět is causally connected with various IN phonetic phenomena. So far as I am aware, the pěpět does not become a diphthong in any of the IN languages. In Old Jav., u before a vowel turns into a consonant, hence the conjunctive of těmu, "to meet with", is atěmwa; but before the pěpět the u persists, and the pěpět is simply absorbed, without any lengthening of the u; hence the gerund těmun < těmu + en.

54. The modified (Umlaut) vowels are described and discussed in another connexion (§§ 251 seqq.).

55. The nasalized vowels are not largely represented in IN. The nasalization is caused either by a preceding or a following nasal consonant.

I. The nasal consonant precedes. "In Achinese the nasals impart their strongly nasal sound to the following vowel" (Snouck Hurgronje).

II. The nasal consonant follows. "In Hova, as in French, the nasalization is coincident with the commencement of the vowel" (Rousselot). "In Sakalava, in the case of nasal vowels, one also hears the nasal, e.g. in the first a of the word mandea, "to go", the n sound" (Fahrner).

56. The semi-voivels y and w. " Javanese y is a semi-vowel like the French y in il y a" (Roorda). "Dayak y is to be pronounced as in the English you" (Hardeland). " Bontok w is as in (the English) winter; a consonantal u" (Seidenadel). "Makassar w is to be pronounced like the ou in the French ouate" (Matthes).

With this articulation of the two semi-vowels all sorts of IN phonetic phenomena are connected. " When speaking slowly the Dayak pronounces y as a short i, thus yaku, " I ", as a trisyllable, iaku" (Hardeland). In several languages initial w receives a prothetic u; thus Original IN walu, "eight", is pronounced walu and uwalu in Tontemboan.

There are however also other ways of pronouncing the semi-vowels in IN. "Bungku w is dentilabial" (Adriani). As w is represented in Rottinese by f, e.g. in falu < Original IN walu, "eight", and y in Hova by z, e.g. in hazu, "tree" < Original IN kayu, we must assume as transitional stages semi-vowels accompanied by fricative sounds.
57. The liquids r and l.
I. The liquid r. "In certain regions the Malay r is formed by the tongue and teeth, in others by the tongue and palate, in others again it is uvular" (Ophuijsen). "In the north (of the Peninsula, the Malay r) is guttural" (= uvular) (Winstedt). "Madurese r is coronal-cacuminal" (Kiliaan). "The northern dialects of Sangirese have a labial r" (Talens).
A few IN languages have two differently articulated r's; thus Běsěmah possesses a lingual one and a uvular one. That was also the case in Original IN (see § 40).
II. The liquid l. "The Gayo l is formed by the articulation of the tip of the tongue against the roots of the upper teeth" (Hazeu). "Madurese l is pronounced by the articulation of the edges of the tip of the tongue against the foremost part of the hard palate, the tip of the tongue being bent upwards and backwards" (Kiliaan)." Bada has a pre-palatal l as well as a supradental one" (Adriani).
58. The laryngal q. "The hamzah is the explosive formed by the glottis" (Adriani). "The hamzah is formed by the sudden opening of the closed vocal chords" (Snouck Hurgronje). "In Ampana the hamzah is as a rule weakly pronounced" (Adriani).
59. Velars. As regards these there is nothing further to be said.
60. Palatals. "In the Madurese palatals the back of the tongue, more precisely the middle part of it, articulates against the back part of the hard palate" (Kihaan). "The Javanese c is supradental (alveolar), the Malay one palatal, but not purely explosive like the Tontemboan one, but to some extent fricative" (Adriani). "Bontok j and c" (which Seidenadel writes dj and tj) "are dentals, not palatals; frequently they are near ds and ts (d and t ' mouillé ') " (Seidenadel).
61. From these and other descriptions of the palatals, it appears that their articulation varies very considerably in the several languages, so that the name "palatal" is often inappropriate, but more particularly that in several languages they are not purely explosive but accompanied by a fricative sound; in that case they do not represent a single consonant but rather two. From this circumstance many IN linguistic phenomena can be explained:
I. Just as no IN word may have more than one consonant at the end, so too a palatal is not permissible in that position.
II. In Dayak two consonants coming together make the preceding vowel short, as in sǎnda, "pawn, pledge", a simple media makes it long, as in lādin, "knife", but before the palatal media the vowel is always short, as in mǎja, "to visit". Thus j operates like two consonants together.
III. In Sundanese the accent falls on the last syllable when the penultimate contains a pěpět; thus for example in tělúk, "bay"; save that if two consonants follow immediately after the pěpět, as in děnki, "envious", the accent can remain upon the ě, and similarly if a palatal follows, as in sěja, "plan".
Note.—After the descriptions in § 60 we can understand why the native alphabets sometimes write the palatal nasal and sometimes the dental one before the palatals, thus tuñjuṅ or tunjuṅ, "water-lily".
62. Cerebrals or cucuminals. "In Madurese the cacuminals are produced by the articulation of the tip of the tongue against the front part of the hard palate, the tip of the tongue being bent upwards and backwards" (Kiliaan).
63. Dentals. "The Achinese d is formed by the articulation of the tip of the tongue against the gums close to the roots of the upper teeth" (Snouck Hurgronje). "Malay d and t are supradental" (Fokker). "In Lebonese d and t are supradental" (Adriani).
64. Labials. Here there is nothing more to be said.
65. Spirants. " The Dayak s is hard, like the hard s in German"* (Hardeland). "The Tontemboan s is supradental" (Adriani). "The Gayo s is pronounced a little between the teeth, somewhat lisping" (Hazeu).
"The Nias x sounds like the German ch in the word wachen" (Sundermann). "Tontemboan has no velar media; in place of it there is a spirant which is pronounced at the back part of the hard palate" (Adriani). Bontok š is like "sh as in (the English) shield" (Seidenadel). Bontok š is "as in (the English) fine" (Seidenadel). "Buli f is bilabial" (Adriani).
66. The aspirate h. "Gayo h, as in Dutch, distinctly audible even at the end of a syllable" (Hazeu). "Javanese h is mute when it is the initial of a word, and very weak as the final of a word, likewise between two different vowels, while between two similar vowels it is hke the Dutch h" (De Hollander).

* [Or in English.]



Quantity in General.

67. In most of the IN languages there are two gradations of quantity: long and short. In Bontok "lengthened syllables are but little longer than short syllables" (Seidenadel). "In figures the quantity of the German long a might be estimated to be 2, the Malay long ā 1½" (Fokker). Sangirese has three gradations, the long vowels resulting from contraction being longer than the rest. Madurese has no gradations of quantity.

Quantity of the Accentuated Syllable in Words of More than One Syllable.

68. In a considerable number of IN languages there prevails a law of quantity which we may call "the IN law of quantity" and which in its two branches is as follows:
I. The law of length: The vowel is long when followed by only one consonant, e.g. in wālu, "eight".
II. The law of brevity: The vowel is short when followed by more than one consonant, e.g. in gǎntuṅ, "to hang".
69. In several languages the IN law of quantity is modified by the interference of special laws, e.g.:
I. In Dayak the IN law of length is restricted by the tact that before unvoiced sounds the vowels are mostly short, e.g. the a in ǎso, "dog"; and even before the voiced palatal the vowels are short, always (see § 61).
II. In Modern Javanese the IN law of brevity is restricted by the fact that before a nasal + a cognate explosive the vowels are mostly long, e.g. in dīntěn, "day".
III. Seidenadel, in his Bontok grammar, gives no theory of quantity, but an examination of his texts leads to the following results: The law of brevity exhibits hardly any exceptions; in Lumawig 69 we find the word ākyu, "sun", which is contrary to that law. The law of length exhibits more exceptions, esjjecially the one whereby a short vowel is frequently found before a nasal, as in Lumawig 1 ǎnak, "child", 13 tǎnub, "reed, hollow stalk", Kolling 10 wǎnis, "trousers". Before f all the texts exhibit no exceptions; thus we find only forms like tukfīfi, "star", etc.
70. There are however also IN languages that display a law differing entirely from the IN law of quantity, e.g. Dairi. In Dairi the vowel of every accentuated syllable is long; thus for example in pōstěp, "to begin".
71. When the accent is shifted from the penultimate syllable to the final one, as happens in the case of contractions and in many languages in the vocative, two separate tendencies assert themselves:
I. The vowel is long. So in Old Javanese in contractions, where the fact is indicated by the symbol of length in the manuscripts; e.g., Ramayana, VII, 40, 2: "In order to enter into the interior" = tumamā riṅ abhyantara. Here tumamā = the aorist tumama + the sign of the conjunctive a, the WB being tama. — Or in Gorontalese in the vocative, of which Breukink says: "Suku akhir itu boleh měnjadi panjaṅ, jikalaw kata itu ditilik sapěrti kata sěruhan ataw suruhan" = "The vowel of the final syllable becomes long when the word is used as a vocative or a command".
II. The vowel is short. So in Hova in contractions, according to Ferrand. Thus it appears that the Hova imperative milazá, "tell!" < indicative milazá + imperative sign a, has a short final vowel.
72. Many parallels can be drawn between the IN and IE phenomena of quantity. The IN law of quantity coincides with the German one; cf. Siebs, "Deutsche Bühnenaussprache", in the section entitled "Vokale". Madurese knows no differences in quantity, precisely like Rumanian; cf. Tiktin, "Rumänisches Elementarbuch", § 15.

Quantity of Vowels in Monosyllabic Words.

73. In some languages the monosyllabic words of substance are long, e.g. in Karo. Thus even the Karo word pět, "to seek", wherein ě is — inaptly — used for the pěpět, is pronounced long. In other languages such words are short; so in Hova, e.g. in , "negation".
74. The monosyllabic affirmative a or o is long in most languages, as is shown especially by the spelling of the texts; thus in the Kamberese Story of the Top, Bijdr. 1913, p. 83, l. 28, we find: "Yes, yes, said they" = ā ā hiwada.
75. Monosyllabic ;;words of form are mostly short, because (for one thing) they have but a weak stress in the sentence. But they may be long; thus according to Meerwaldt the Toba , "every", and , "even", are long. When a shortly pronounced word of form is formed by composition into a word of substance, length of vowel may ensue. In the Bareqe Tale "The Monkey and the Pig", Bareqe Leesboek, p. 15, 1. 4, we find: "To dig up roots" = maṅkae toraa. Adriani's spelling with aa indicates the length of the final vowel. "Root" = torā, with the accent on the ā, really stands for "that (which is) in (the earth)", the word for "in" being .

Quantity of Unaccentuated Syllables.

76. The syllables which precede the accentuated one are almost always short. Bugis has some long ones, but a search through the dictionary only reveals about half a dozen cases, and these are mostly unexplained etymologically, as' mēñcána, "shallow".
77. Syllables which come after the accentuated one are not infrequently long, especially when they end in a vowel. In Dayak all vowels at the end of words are long; thus hūmā, "house", has both vowels long, the penultimate one being accentuated. Bugis in certain cases has vowel length in the unaccentuated final syllable, even when it ends in a consonant, e.g. in dīmēṅ, "longing".
78. The phenomenon that syllables which precede the accent are hardly ever long, while those that follow it are often long, is parallelled by the fact that the former hardly ever contain diphthongs, whereas the latter often do (see § 171).
79. In Bugis, when a word accentuated on the final syllable becomes the first member of a compound, the accent may be thrown back; in that case, if the final syllable was long, e.g. on account of contraction, it loses its length. From WB táppa is derived tappāṅ, "creation, model" < tappa + āṅwith the accent on the final syllable; in the compound táppāṅ-matuwa, "model for a father-in-law" = "future father-in- law", the accent has been thrown back and the vowel has become short.

Quantity in Old Javanese.

80. The Old Javanese manuscripts indicate the length of the vowels. But it is noticeable how seldom the marks of length occur in them. Judging from the quantity of Modern Javanese they ought to be much more frequent. In the Ramayana the long vowels are found, apart from Sanskrit loan-words, only in interjections, in certain monosyllabic words of substance, as in kūṅ, "longing", but not in sih, "pity", in contractions like matī, "to let die" < mati + i, and in compensatory lengthenings, as in ikū, "tail" < Original IN ikur2. Hence we meet with whole verses without a single long vowel, e.g. Ramayana, V, 68, 2 : "She then, quite alone, entered unafraid" = sira juga tuṅga-tuṅgal anusup tamatar matakut. Were there perhaps in Old Javanese three gradations of quantity, as in Sangirese (§ 67), and is only extreme length marked in the manuscripts?

Quantity in Original Indonesian.

81. As we have got an "IN law of quantity", but it is counteracted by all sorts of special laws, as further there are difficulties about quantity in Old Javanese, and lastly as in not a few IN languages the available data about quantity are insufficient, we are not at present in a position to form a definite and trustworthy picture of quantity in Original IN.

Quality of the Vowels.

82. As regards the quality of vowels in IN we chiefly meet with two tendencies:
I. The quality depends upon the quantity. Long vowels are close, short ones are open. This law holds good for several languages.
II. The quality depends upon the sounds that follow. Thus in Minangkabau accentuated e before s, as in leseq, "zealous", is close, while before r, as lereṅ, "slope, descent", it is open.

Doubling of the Consonants.

83. What is called gemination, doubling of consonants, and the like, may represent several different phonetic values: see Sievers, "Phonetik", in his chapter entitled "Silbentrennung". As regards the nature of the IN double consonants, the following definitions (inter alia) give us some information. "In all these (i.e., certain Philippine) languages the gemination is real, that is, the two consonants are distinctly pronounced" (Conant). "In Bugis the consonants that are written double are pronounced so that the consonant both closes the preceding syllable and begins the following one" (Matthes). "The dividing line of syllabic stress* lies in the
* [The author gives the following illustration of what is meant by "the dividing line of syllabic stress" (Druckgrenze): "In the Italian word anno, "year", the an- is spoken decrescendo, and the -no crescendo. After the n of an- the voice is feeblest, weakest, and this is the 'Druckgrenze'."]
geminate consonant itself" (Kiliaan).—In Bontok the two consonants may also be separated by a hamzah as well; thus in Seidenadel-Texts, Headhunters' Ceremonies, 4, we find: "The old people" = nan amamqma.
84. The rarest cases of doubhng are those of h and q. Madurese has a few instances, e.g., ěhham, "ham", and leqqer, "neck".
85. As a rule doubling only occurs between vowels; before a consonant it is rarer, e.g., as in the Madurese lommra, "accustomed", in accordance with the law given in § 86, III.
86. Consonantal doubling in the living IN languages owes its origin to several distinct factors; these are:
I. Doubling of the root, which is one of the methods of WB-formation, when the root begins and ends with a similar consonant, as in the Kangeanese tottot, "tame". This case also occurs especially in the language of children, e.g. in the Achinese childish word mammam, "cakes".
II. Derivation from the WB. Here it may be simply a case of mere addition, as when in Toba from the prefix mar + AVB rara there results the adjective marrara, "red". Or it may involve phonetic processes, as when in Madurese from ṅator + the suffix aghi we get ṅatorraghi, "to offer".
III. Various phonetic laws. Before r or l Madurese doubles every consonant except , n, and w; hence the above-cited lommra, as compared with lumrah in other languages. In Talautese r is pronounced double when it follows immediately after the accentuated vowel.
As for consonantal doubling after the pepet, see § 5.
IV. Assimilation. In spoken Toba, in the combination nasal + cognate tenuis, the nasal is assimilated to the tenuis: thus Original IN and ṅwritten Toba gantuṅ ^ spoken Toba gattuṅ. Certain cases of assimilation also especially occur in sentence-sandhi ; thus in Tuuk Lb, I, p. 1, 1. 11, we find written: "Red because of their ripeness" = Red now because r. their = marrara do dibahen lamun-na; but the spoken language says dibahel lamun-na.
V. Haplology, as for example when in Iloko apó-apó becomes appó, "grandfathers", from the singular apó.
VI. Sandhi phenomena not dependent upon assimilation. These occur for example in Timorese, as instanced in the text "Atonjes Nok", Bijdr. 1904, pp. 271 seqq. There we find, e.g., p. 271, 1. 7: "To marry a woman" = M. w. a = sao bifel-l-es, from sao + bifel + es.
VII. Analogical transference. In Makassar, final is assimilated to the immediately following possessive na, hence "His king" = karaenṅ-na > karaénna; through transference this nna is also added to words ending in a vowel, hence matánna, "his eye", from mata, "eye".
VIII. Some interjections, e.g. Madurese awwa.
87. Of all these cases of consonantal doubling only the one mentioned under I. supra can be positively ascribed to Original IN.
88. The phenomena connected with the doubhng of consonants have many parallels in IE. Thus, for example, the Madurese doubling mentioned in III. supra may be compared with the West Germanic consonantal lengthening (Kluge, "Urgermanisch", §§ 157 seqq.). The IE doubling of consonants in personal names (Brugmann, KvG, § 366, 6) has nothing corresponding to it in IN.


Preliminary Observations.

89. I have prepared for my own use a list of all the phonetic laws of all the hitherto known IN languages. From that list I here give a selection of the more important phenomena, being guided in my choice by the interests of IN research on the one hand and those of IE study on the other.
90. Phonetic changes are either unconditional or conditional (see § 10); in the latter case I add the condition. But it may happen that the condition is composed of several different factors, which it would take too long to go into; or alongside of the cases that follow the law there may be a serious number of exceptions; or the material at my disposal may be incomplete: in such cases I employ the neutral formula: "the phonetic change occurs in certain cases".

Laws of the Vowels.

91. Original IN a. I. It persists for the most part unchanged in the living languages. Original IN anak, "child", appears as anak in Old Javanese, Dayak, etc., as anaq in Bugis, etc.
II. Original IN a becomes o in several languages; thus in Tontemboan before w, hence Original IN awak > Tontb. owak, "body".—It becomes e in several languages; thus in Sumbanese by Umlaut,* hence Original IN tasik > Sumb. tesi, "lake".—It becomes i in several languages; thus in Taimuruna by complete assimilation, hence Original IN lima > Taim. limi, " five ".—It becomes o in certain cases in Gayo,
* [" Umlaut " is a particular case of partial assimilation: see § 251.] hence Original IN ina > Gayo inö, "mother". — It becomes ě in Besemah when final, hence Original IN mata > Běs. matě, "eye". — It becomes aw in Sěraway when final, hence Original IN mata > Sěr. mataw.
III. Original IN a rarely disappears altogether; it does so in certain cases in Hova after Original IN y, hence Original IN laya2 > Hova lay, "to sail".
92. Original IN i. I. It persists for the most part unchanged in the living languages. Original IN lintah, "leech", appears as lintah in Old and Modern Javanese, Malay, etc., as dinta in Hova, etc.
II. Original IN i becomes e in a few languages, thus in certain cases in Madurese, hence Original IN lintah > Mad. lenta. — It becomes ey in several languages when final; thus in Tiruray, hence Original IN tali > Tir. taley "rope". — It becomes oy in certain cases in Achinese when final; hence Original IN běli > Ach. bloy, "to buy".
III. Original IN i rarely disappears altogether; it does so in Tontemboan under the exigencies of metre. In Schwarz-Texts, p. 317, Songs of Martina Romoas, 13, we find: "Do you mean?" = cua-mu. That cua < icua = prefix i + WB kua has really lost an i, is proved by the presence of the c, which can only occur after an i (see § 103).
93. Original IN u. I. It persists for the most part unchanged in the living languages. Original IN tunu, "to burn", appears as tunu in Old Javanese, Hova, etc.
II. Original IN u becomes o in a few languages; thus in certain cases in Madurese, hence Original IN putih ^ Mad. pote, "white". — It become ü in Bontok by Umlaut, hence Original IN babuy > Bont. fafüy, "boar". — It becomes i in Loindang by complete assimilation, hence Original IN kulit > Loi. kilit, "skin". — It becomes ew in several languages when final; thus in Tiruray, hence Original IN pitu ^ Til. fitéw, "seven". — It becomes ee in Achinese in certain cases when final, hence Original IN palu > Ach. palée, "to strike".
III. Original IN u rarely disappears altogether; it does so in Kupangese owing to sentence-sandhi. In Kup. the word for "to draw (liquor)", when pronounced by itself or in a pause, is sulu. But in the Story of the Fool, Bijdr. 1904, p. 259, 1. 13, we find: "And drew in order to pour (into another vessel)" = ti sul le doan.
94. The vowel e. I observed in § 40 that bela, "companion, avenger, to share the same fate", is the only word I have hitherto positively ascertained as possessing an original e. This bela remains unchanged in the several languages where it occurs, such as Gayo, Bimanese, etc., only the Achinese has bila.
95. The vowel o. I mentioned in § 40 that sor, "below", is the only word I have positively ascertained as possessing an Original IN o. This o persists unchanged everywhere, thus in the Old Javanese sor, Tontemboan sosor, etc.
96. The vowel ě: see §§ 121 seqq.

Laws of the Semi-Vowels.

97. Original IN y. I. It persists unchanged in many living languages. Original IN layar2, "to sail", appears as layar in Malay, Sundanese, etc., as layag in Tagalog, etc.
II. Original IN y becomes j in several languages, thus in Bugis between a, o, or u, and an immediately following vowel, hence Original IN layar2 > Old Bugis lajaq, "to sail". — It becomes z in certain cases in Hova, hence Original IN kayu > Hova hazu, " tree". — It becomes l in Sangirese between vowels, hence Original IN kayu > Sang. kalu.
III. Original IN y disappears altogether in several languages; thus in Toba, hence Original IN kayu > Toba hau.
98. Original IN w. I. It persists unchanged in many living languages. Original IN walu, "eight", appears as walu in Tettum, as waluh in Gayo, etc.
II. Original IN w becomes u in Toba when initial, hence Original IN walu > Toba ualu, "eight", a word of three syllables, also pronounced uwalu. — It becomes h in Mentaway, as in halu, "eight". — It becomes f in Rottinese, as in falu, "eight". — It becomes ww in Old Javanese, hence Original IN wara > Old Jav. wwara, "to be" (the substantive verb). — It becomes gu in Inibaloi, as in gualo, "eight". — It becomes h in Mamuju, hence Original IN tawa > Mam. taha, "to laugh".
III. Original IN w disappears altogether in a few languages, thus in Modern Javanese between a consonant and a vowel, hence Modern Jav. lir. "manner", for Old Jav. lwir.

Laws of the Liquids.

99. Original IN r1, the lingual r. I. It is preserved in many living languages, but pronounced in various ways. Original IN pira, "how much", is also pira in Old Javanese, Kamberese, etc., firi in Hova, etc.
II. Original IN r1 becomes l in several languages; thus in certain cases in Bisaya, as in pila, "how much". — It becomes d in several languages; thus in certain cases in Balinese, as in pidan, "how much". — It rarely becomes g; thus in certain cases in Toba, hence Original IN ir1uṅ > Toba iguṅ, "nose". — It becomes χ in Nias in the cases where Toba has g, as in iχu, "nose".
III. Original IN r1 disappears altogether in several languages when final; thus in Hova, hence Original IN butir1 > Hova wutsi, "bud".
100. As for Original IN r1, see §§ 129 seqq.
101. Original IN l. I. It mostly persists unchanged in the living languages. Original IN laṅit, "sky", is also laṅit in Old Javanese, lanitra in Hova, etc.
II. Original IN l becomes r in several languages; thus in Toba by a regular assimilation whenever the word contains an r, hence Original IN lapar > Toba rapar, "hunger". — It becomes y in several languages; thus in Bareqe between vowels, hence Original IN jalan > Bar. jaya, "path". — It becomes w in several languages; thus in certain cases in Tagalog, hence Original IN puluh, "ten" > Tag. powo. — It becomes n in certain cases in Timorese, hence Original IN kali > Tim. hani, "to dig". — It becomes d in several languages; thus in Hova before an original i, not an i derived from ě, hence Original IN lima > dimi. — It becomes g in several languages; thus in certain cases in Batanese, hence Original IN ulu > Bat. ogo. — It becomes h in some Formosan dialects in certain cases, hence Original IN ulu > Form, uho, "head",
III. Original IN l disappears altogether in several languages; thus in Boano, hence Original IN balay, "house" > Boa. bae.

Laws of the Laryngal q.

102. For the laws of the laryngal q, see §§ 140 seqq.

Laws of the Velars.

103. Original IN k. I. It persists for the most part unchanged in the living languages. Original IN kuraṅ, "deficiency", appears also as kuraṅ in Old Javanese, Makassar, etc., as 'koraṅ in Tarakan, etc.
II. Original IN k becomes g in several languages; thus in Tiruray between vowels, hence Original IN laki > Tir. lagey, "man" (as opposed to "woman"). — It becomes h in several languages; thus in Hova when initial or between vowels, hence Original IN kuku > Hova huhu, "claw". — It becomes q in several languages; thus in Bugis when final, hence Original IN anak > Bug. anaq, "child". — It becomes c in Tontemboan when an i precedes, hence Original IN tasik > Tontb. taqasic, "lake". — It becomes t in Hova in sentence- sandhi before s, e.g., Hain-Teny, p. 264, 1. 4: "Young lark" = zanat surúhitra < zanak < zánaka, "young" + surúhitra, "lark". — It becomes s in Kawangkowan Tontemboan in the cases where the standard Tontemboan has c < k, thus Kaw. taqasis for the above taqasic.
III. Original IN h disappears altogether in several languages; thus in certain cases in Bugis, hence Original IN kulit > Bug. uliq, "skin".
104. Original IN g. I. It mostly persists unchanged in the living languages. Original IN gantuṅ, "to hang", appears in Old Javanese, Sundanese, etc., as gantuṅ, in Bugis as gattuh, etc.
II. Original IN g becomes k in Bugis after w, hence Original IN tuṅgal ^ Bug. tunke, "alone". — It becomes gh in Madurese, hence ghantoṅ, "to han ". — It becomes a velar spirant in Tontemboan (see § 65). — It becomes h in Hova when initial, hence Hova hantuna, "to hang".
III. Original IN g rarely disappears altogether; it does so in Bottinese after n, hence Makassar, etc., geṅgo appears in Rot. as ṅgeṅo, "to rock to and fro".
105. Original IN n. I. It mostly persists unchanged in the living languages. Original IN aṅin, "wind", is also aṅin in Old Javanese, Malay, etc., haṅin in Tagalog, etc.
II. Original IN becomes n in several languages; thus in Hova, save before a velar, hence in ánina, "wind". — It becomes ñ in several languages; thus in certain dialects of Tontemboan after i, hence Original IN and Tontb. liṅa > dialectic Tontb. liña, "to hear". — It becomes k in several languages by assimilation; thus in spoken Toba, hence Original IN baṅkay > written Toba baṅké > spoken Toba bakke, "corpse".
III. Original IN disappears altogether in several languages when final; thus in Nias, hence Original IN ar1ěṅ > Nias aχo, "charcoal".

Laws of the Palatals.

106. Original IN c. I. It is preserved in some of the languages. Original IN r1acun, "poison", appears in Old Javanese and Malay as racun, in Bimanese as racu, etc.
II. Original IN c becomes s in many languages; thus in Tagalog, hence lason, "poison".
107. Original IN j. ;. I. It persists unchanged in some of the languages. Original IN jalan, "path", appears also in Bontok, Besemah, etc., as jalan, in Bareqe as jaya, etc.
II. Original IN j becomes c in Bugis after n, hence Original IN janji > Bug. janci, "promise". — It becomes jh in Madurese, as in jhalan, "path". — It becomes d in several languages; thus in certain cases in Old Javanese, as in dalan, "path". — It becomes z in certain cases in Hova, hence Original IN tuju > Hova tuzu, "direction". — It becomes s in Lalaki, as in sala, "path".
108. Original IN ñ. I. It persists unchanged in some of the languages. Original IN pěñu, "turtle", appears also in Old Javanese as pěñu, in Madurese as pěñño, with doubling of the ñ in accordance with the law in § 5, V, etc.
II. Original IN ñ becomes n in some languages; thus in Toba, as in ponu, "turtle".

Laws of the Dentals.

109. Original IN t. I. It mostly persists unchanged in the living languages. Original IN tali, "rope", appears also in Old and Modern Javanese, etc., as tali, in Tettum as talin, etc.
II. Original IN t becomes d in several languages; thus in certain cases in Sawunese, hence Original IN mata > Saw. mada, "eye". — It becomes is in Hova before i, hence Original IN tilik, "to peep at" > Hova tsidika. — It becomes k in several languages; thus, according to Aymonier and Cabaton, in Cham before l, hence Original IN telu > tlu > Cham klaw, "three". — It becomes χ in certain cases in some of the Formosan dialects, hence Original IN tai, "dung" > Form. χe. — It becomes h in several languages; thus in certain cases in Kamberese, hence Original IN pitu > Kamb. pihu, "seven". — It is cerebralized in several languages; thus in certain cases in Madurese. — It becomes s in Bolaang-Mongondou when in contact with i, hence Original IN kulit > Bol.-Mong. kulis, "skin".
III. Original IN t disappears altogether in several languages; thus in Nias when final, hence Original IN kulit, "skin" > Nias uli.
110. Original IN d. I. It persists unchanged in many languages. Original IN dagaṅ, "stranger", appears also in Old Javanese, Toba, etc., as dagaṅ, in Bimanese as daga, etc.
II. Original IN d becomes t in several languages when final; thus in Malay, hence Original IN añud, "drift" > Mal. hañut. — It becomes dh in certain cases in Madurese, hence Original IN damar > Mad. dhamar, "resin". — It is cerebralized in several languages. — It becomes r in several languages; thus in Bugis after n, hence Original IN linduṅ > Bug. linruṅ, "shade".
II. Original IN d disappears altogether in some languages; thus in Kulawi after n, hence Original IN tanduk > Kulawi tonu, "horn".
111. Original IN n. I. It is preserved in the living languages in a great majority of the cases. Original IN anak, "child", appears also in Old Javanese, etc., as anak, in Nias as o1no1, etc.
II. Original IN n becomes in several languages when final; thus in Bugis, hence Original IN aṅin > Bug. aṅiṅ, "wind". — It becomes l in several languages; thus in certain cases in some of the Formosan dialects, as in alak, "child ". — It becomes t by assimilation; thus in spoken Toba, hence Original IN guntuṅ > spoken Toba gattuṅ, "to hang".
III. Original IN n is lost in several languages; thus in Nias before t, hence Original IN lintah > Nias lita, "leech".

Laws of the Labials.

112. Original IN f. I. It mostly persists unchanged in the living languages. Original IN pitu, "seven ", is also pitu in Old Javanese, Masaretese, etc., opitu in Gorontalese, etc.
II. Original IN p becomes b in Achinese when final, hence Original IN idup > Ach. ndeb, with metathesis of the vowels and change of i into e. — It becomes f in many languages; thus in Hova when initial or between vowels, as in fitu, "seven". — It becomes w in Nias in sentence-sandhi in accordance with the law changing surds into sonants (§ 302). Original IN par1ay, "rice", appears in Nias as faχe; but in the Dancing Hymn in Bijdr. 1905, p. 12, 1. 4 from the bottom, we find: "I winnow rice" = u siχi waχe. — It becomes k. "Some of the tribes of the Eastern Toba cannot pronounce

f and make a k of it, thus kiso for the standard Toba piso, 'kiso'" (Van der Tuuk). — It becomes h in Rottinese, as in hitu, "seven".

III. Original IN p disappears entirely in several languages; thus in certain cases in Kissarese, hence Original IN pira > Kis. ira, "how much".
113. Original IN b. I. It persists unchanged in many of the living languages. Original IN baṅaw, "heron", appears also in Malay, Dayak, etc., as baṅaw, in Old Javanese as baṅo.
II. Original IN b becomes bh in certain cases in Madurese, hence Original IN buru ^ Mad. bhuru, "to hunt". — It becomes p in Buli in certain cases, hence Original IN bulu > Buli plu, "hair". — It becomes w in several languages; thus in Hova when initial or between vowels, as in wanu, "heron". — It becomes f in Rottinese, hence Original IN r1ibu > Rot. lifu, "thousand". — It becomes h in the Silayarese dialect of Makassar in certain cases, hence Original IN běli, "price" > Mak. balli > Sil. halli.
III. Original IN b disappears altogether in several languages; thus in Gayo in certain cases when initial, hence Original IN batu > Gayo atu, "stone".
114. Original IN m. I. It persists for the most part unchanged in the living languages. Original IN mata, "eye", is also mata in Old Javanese, Bagobo, etc., matan in Tettum. etc.
II. Original IN m becomes n in several languages: thus in Hova when final, hence Original IN inum, "to drink" > Hova inuna, "to drink poison". — It becomes in a few languages when final; thus in Bugis, as in inuṅ, "to drink".

— It becomes p by assimilation; thus in Toba, hence Original IN lumpat > spoken Toba luppat, "to jump".

III. Original IN m disappears altogether in several languages when final; thus in Bareqe, as in inu, "to drink".

Laws of the Spirant s.

115. Original IN s. I. It mostly persists in the living languages. Original IN susu, "breast", is also susu in Old Javanese, Malay, etc.
II. Original IN s becomes š in several languages; thus in Mentaway when initial, hence Original IN siwa > Ment. šiba, "nine". — It becomes h in several languages; thus in Kamberese, as in hiwa, "nine". — It becomes t in several languages; thus in Buol, hence Original IN si, the article > Buol ti.
III. Original IN s disappears altogether in several languages; thus in Hova in certain cases, hence Original IN běsi > Hova wi, "iron".

Laws of the Aspirate h.

116. Original IN h. I. It persists unchanged in a minority of the IN languages. Original IN pěnuh > "full", appears also as pěnuh in Old Javanese, as panuh in Tarakan, etc.
II. Original IN h becomes q in a few languages; thus in certain cases in Tontemboan, hence Original IN lintah > Tontb. lintaq.
III. Original IN disappears altogether in the majority of the living languages; thus in Bugis, hence Original IN pěnuh > Bug. pěnno, "full", and Original IN ilih ^ Bug. ile, "to choose". — In such cases a u or i preceding the h becomes o or e in Bug., whereas final Original IN u and i remain unchanged, hence Original IN and likewise Bug. tunu, "to burn", kali, "to dig".

Laws of the Simple Sounds in Indo-European and in Indonesian.

117. A large majority of the IN phonetic changes also occur in IE, partly under similar conditions, and partly under different ones. I give here a selection of parallels between IE and IN:
Sanskrit and Toba: s + s > ts. — Sansk. vatsyāmi, "I shall dwell" < vas + syāmi; Toba latsoada < las + soada, "not yet".
Old Persian and Kamberese: s > h. — Old Pers. hainā, as compared with Sanskrit senā, "army", Meillet GvP, § 130; Kamb. hiwa < Original IN siwa, "nine".
Armenian and Rottinese: p > h. — Arm. hing, "five", as compared with Sanskrit pañca, Greek pente; Rot. hitu, "seven" < Original IN pitu.
Greek and Modern Javanese: w > nil. — Gr. oikos, as compared with Sanskrit veśa; Modern Jav. lir, "manner" < Old Jav. lwir.
Latin and Toba: y between vowels > nil. — Lat. tres < treyes; Toba hau, "tree" < Original IN kayu.
Old Bulgarian and Makassar: All original diphthongs become simple vowels, cf. Leskien, "Grammatik der altbulgarischen Sprache", §§ 43 seqq.
Old Prussian and Cham: tl > kl. — Old Prus. stacle, "support" < statle (Trautmann, "Die altpreussischen Sprachdenkmaler", § 67); Cham klaw < tlu < Original IN telu, "three".
Germanic and Hova: k > h. — Gothic hian, "to conceal", as compared with Latin celare; Hova hazu, "tree" < Original IN kayu.
Old Irish and Rottinese: w > f. — Old Ir. fer, "man", as compared with Latin vir; Rot. falu, "eight" < Original IN walu.
Sicilian dialect and Bugis: media after nasal > tenuis. — Sic. ancilu, "angel" < Latin angelus; Bug. jañci, "promise" < Original IN jañji. German dialect of Lucerne and Mori: nlnd. — Luc. dialect Määndig, "Monday" (High German Montag); Mori mondasu, "sharp", as compared with Petasia montaso, from the Original IN WB tajĕm.

118. There are two IN phonetic laws for which I know of no parallels in IE; both are pecuhar to Achinese and both appear in the word lhee, " three " ᐸ Original IN tělu : Initial Original IN těl ᐳ Ach. lh, and final Original IN u ᐳ Ach. ee. See also Section V, ad fin.





Preliminary Observations.

119.The most important IN phonetic laws are four in number: the pepet-law, the RGH-law, the hamzah-law, and the law of the mediæ.

120. Now the course of our enquiry is as follows:

I. In the case of the pepet-law we have to ascertain what are the representatives of the Original IN pepet in the living IN languages.

II. In the case of the RGH-law our enquiry has to pursue the same course as with the pepet-law, we have to find out by what sounds the Original IN uvular r (r2) is represented in the living IN tongues. This law is also called, after its discoverer, by the name of "Van der Tuuk's first law". I have preferred to designate it by the more convenient and significative name of "RGH-law", a name based on the fact that Original IN r2 is represented in many of the living languages by g, in some by h; and I observe that this designation is gradually gaining ground.

III. In the case of the hamzah-law our business is to set forth from what Original IN sounds or by what linguistic processes the sound hamzah has originated in the living IN languages. Here, therefore, the procedure differs from that which is to be followed in the case of the pěpět-law and RGH-law; we start from an Original IN multiplicity and arrive at one uniform result in the living languages, viz., the hamzah. For the hamzah is a secondary sound in IN ; we cannot therefore proceed from an Original IN hamzah. The hamzah-law is peculiarly interesting for the following reason: in IN the hamzah is very widely distributed, in IE on the other hand it is very little known; the phenomena connected with the hamzah therefore mark an important difference between IE and IN.

IV. The law of the mediæ. In a number of IN languages media and continuant interchange in one and the same word. Thus in Bugis the WB for "to cut off" appears sometimes as bětta and sometimes as wětta, thus at times with the momentary media h, and at other times with the continuant w. Probably there was nothing corresponding to this in Original IN; so in this case (in contrast to the other three laws) we cannot have recourse to Original IN at all.

The Pepet-Law.

121. Original IN possessed the vowel ě, described in § 53, produced with the mouth-cavity in a position of indifference, and often called pěpět.
122. This pěpět has been preserved in a minority of the IN languages, e.g. in Old Javanese, Karo, Bugis, etc. Thus Old Jav. has preserved the original phonetic condition in the case of the pěpět, while abandoning it in the case of the EGH-law.
123. The pěpět can change into all the other vowels: a, i, u, e, o. Original IN teken, "staff", results in Makassar takkaṅ, Dayak teken, Tagalog tikin; Original IN enem, "six", becomes in Toba onom, in certain Formosan dialects unum.
124. In several languages Original IN e has a multifarious resultant.
I. The accent is the determining factor. In Hova ě in an accentuated syllable becomes e, in an unaccentuated one i, hence Original IN ěněm > Hova ěnina. In Kolo in the same way the resultants are o and u, hence Original IN ěněm > Kolo onu. Notice the parallel between the more sonorous e and o and the less sonorous i and u of Hova and Kolo respectively.
II. The consonant following the pěpět is the determining factor. In Pabian-Lampong, ě before r becomes a, while before m it becomes u, etc. Hence Original IN sěmbah, "respectful salutation" > Pab.-Lamp. sumbah.
III. The determining factors cannot be ascertained; thus in Bimanese, where the pěpět can be replaced by all the other vowels.
125. In Old Javanese, in consequence of the loss of an r2, the pěpět may be left standing before a vowel, in which case it changes into w and forms with the vowel a rising diphthong, as in bwat < Original IN běr2at, "heavy". In its further evolution the diphthong becomes a simple vowel, hence Modern Jav. abot < Old Jav. bwat.
126. In a few languages the pěpět disappears altogether.
I. Disappearance before the accentuated syllable, in several languages, when the pěpět stands between a mute and a liquid, e.g., Original IN běli, "to buy" > Gayo bli, also however pronounced běli. — In Tagalog this phenomenon only occurs when the word is also extended by a prefix, so that even after the loss of the e the word remains disyllabic, as in itlóg, "egg" < Original IN tělur2.
II. Disappearance after the accentuated syllable, in several languages, when the pěpět stands between a semi-vowel and a consonant. Thus Original IN dawěn, "leaf", results in Dayak in dawen, which is quite in conformity with the rule (§ 123), but in Malay it is not represented, as one might have expected, by dawan, but by daun (a disyllable).
III. In Old Javanese inscriptions the pěpět is often omitted. Thus we find in Kawi Oorkonden, II, 10, b: "Shall be seized by tigers" = dmakěn iṅ macan. As Modern Javanese pronounces the pěpět in these cases, thus saying děmaq, "to seize", I do not quite know what to think of this omission of the pěpět.
127. Languages that are closely related to one another often display a similar treatment of the pěpět. But that also happens in the case of tongues that are widely apart; thus both in Toba and in Bisaya, ě changes into o, hence tělu, "three" > Toba tolu, Bis. tolo. Finally, it also happens that languages which are very closely related to one another differ just in their treatment of the pěpět; thus in the two principal dialects of Minangkabau it is precisely the difference in the representation of the Original IN e that forms the chief differentia between them: the Agam dialect has a, hence Original IN běr2as, "rice" > Agam bareh; the Tanah Datar dialect has o, hence boreh.
128. The pěpět and prosody. In some languages the pěpět is replaced by another vowel when the verse accent falls upon it. In Tontemboan it is changed into e. "God" in Tontb. is ěmpun, "friend", rěṅan, "friends", rěṅa-rěṅan; but in a poem in Schwarz-Texts, p. 139, 1. 16, in an iambic verse, we find: "Now, gods, friends, ho!" = ja empuṅ rěṅa-rěṅan e. — In Talautese, ě has become a, but under the influence of the verse accent even this a is replaced by another vowel, either e or o; thus in Steller-Texts, p. 66, 1. 2, we find sasobaṅ for sasabbaṅ, "to appear", and elo for allo, "sun".
Note. — For other phenomena occurring in connexion with the pěpět, see §§ 5 and 148.

The RGH-Law.

129. Original IN had two r sounds, a lingual r (=r1) and a uvular r (=r2). "Thousand" in Original IN was r1ibu, but "hundred" was r2atus
130. The Original IN condition has only been preserved in very few of the living languages, and even there not quite undisturbed, for example in Běsěmah. Běs. r1ibu, "thousand", r1acon, "poison", sur1oṅ, "to push", contain r1, like the corresponding Original IN words r1ibu, r1acun, sur1uṅ; Běs. dar2at, "mainland", jar2om, "needle", nior2 "coconut palm", are pronounced with r2, like Original IN dar2at, jar2um, niur2.
131. In several languages r1 and r2 have coalesced: thus Madurese pronounces the r in soroṅ < Original IN sur1uṅ just like the one in jharum < Original IN jar2um, both being cacuminal.
132. Original IN r1 becomes l in some languages, as in Tagalog libo < Original IN r1ibu, in others it becomes d, in others again some other sound (see § 99). I have included these phenomena of the r1 sound under the designation of "the RLD-law", a name which is analogous to that of the EGH-law.
133. Original IN r2, that is to say, the r of the RGH-law, in some languages remains r, which need not however be uvular; in some others it turns into g, or into h; in a few it also becomes t or y or q. Example: Original IN ur2at, "vein" = Malay urat = Tagalog ugát = Dayak uhat = Pangasinan ulát = Lampong uyak = Tontemboan oqat. See also §§ 135 and 139.
134. A comparison of the two preceding paragraphs shows that the further developments of r1 and r2 are in part identical, e.g. both r1 and r2 can result in l. But they never have an identical evolution in one and the same language. Thus, for example, in Tagalog, r1 does in fact become l, as in libo from r1ibu, "thousand", but r2 turns into g, as in ugát < ur1at, "vein".
135. In several languages Original IN r2 is represented by more than one sound.
I. Its position in the word is the determining factor. In Talautese, r2 becomes k when final, otherwise r; hence Original IN bar2at, "west" > Tal. bárata, but Original IN niur2, "coconut" > Tal. níuka.
II. The contiguous sounds are the determining factor. In Sangirese, r2 results in h, but after o < ě it appears as ɣ; hence Original IN r2atus, "hundred" > by metathesis r2asut > Sang, hasuq, but Original IN běr2as, "rice" > Sang. boɣaseq.
III The determining factors cannot be ascertained. In Hova, Original IN r2, sometimes produces r as in awáratra < Original IN bar2at, "west", sometimes s as in wesatra < Original IN betoat, "heavy", sometimes z as in zatu < Original IN r2atus, "hundred", and sometimes disappears altogether as in wau, "new" <^ bar2u.
136. When Original IN r2, becomes y, further developments may occur. The semi-vowel y may unite with the preceding vowel to form a falling diphthong, as in Lampong ikuy, "tail" < Original IN ikur2. By a further phonetic process such diphthongs may become simple vowels, as in Pampanga iki < ikuy < ikur2.
137. In some languages Original IN r2 disappears altogether, particularly in Old Javanese, as in dyus < Original IN dir2us, "to bathe". When the r2 was final in Original IN, then in Old Javanese the preceding vowel is lengthened by way of compensation for the disappearance of the to, as in Old Jav. ikū, "tail" < Original IN ikur2.
138. It is to be supposed that this disappearance was not a single, momentary change. In Old Javanese, r2 probably first turned into h, as in Dayak; such an h is still preserved in wahu, "new" < bar2u. — In other languages where r2 has likewise disappeared the hamzah may have been the transitional sound. In Tontemboan, Original IN r2 becomes q, but in several cases this q has disappeared: Original IN ular2, "snake" > Tontb. ulaq, but Original IN timur2, "south" > Tontb. timu.
139. Languages that are closely related to one another often exhibit a similar treatment of the r2- But this also occurs in the case of languages that are widely distant from one another ; thus both in Lampong and in Pampanga, r2 > y. Finally, it may happen that languages, which difier so little from each other that one can only call them dialects of one another, nevertheless diverge in this matter of the treatment of the To; thus in the various dialects of Talautese it is just the divergent representation of the r2 that forms the differentia between them. The chief dialect turns final Original

IN r2 into k, hence Original IN niur2, "coconut" > niuka, with the supporting vowel a; for this oiiuka other dialects have niuca, niuha, niuta.

The Hamzah-Law.

140. In the living IN languages the hamzah is found as an initial sound before vowels, as a medial sound between vowels or between a vowel and a subsequent consonant, and as a final sound after vowels; in such positions, in fact, as in the Achinese qancó, "to melt", Madurese leqer, "neck", Bugis biriqta, "report", Makassar anaq, "child". — Other positions are rare, such as in the Bontok allqo, "pestle", Tontemboan ĕlaqb, "torch"; and the words in question nearly always offer etymological difficulties.
141. Hamzah is found occasionally as initial, medial, or final of the WB, in which case it is not derived from another sound.
142. In many IN languages words that "begin with a vowel" are pronounced with an emphatic enunciation, where by in fact a hamzah is sounded as initial before the vowel; this rule holds good for Achinese, Tontemboan, etc. It is true that neither the native script nor the transliteration made by scholars is in the habit of representing this hamzah; thus the word for "child" in Tontb. is written anak, but in reahty pronounced qanak, with an initial hamzah.
143. In several languages we find hamzah as a medial between the two vowels of the WB, when there is no other consonant there. In Nias, as the dictionary shows, this is often the case, though it is of course true that a percentage of these q’s result from k, as in ataqu, "to fear" < Original IN takut, and so fall under § 147. In Madurese such a hamzah is found in cases where other languages in the respective words have h or w or no sound at all, as in Mad. poqon, "tree" = Malay pohon = Bugis pōṅ, or in soqon, "to carry on the head" = Javanese suwun.
144. In many languages hamzah is found as a final, abruptly closing the final vowel.
I. In very many interjections, thus in Bugis, Tontemboan, etc.; the interjection "fie!" in particular very often has a q as final: Makassar ceq, Bugis caq, Sangirese siq, etc. The frequent occurrence of q in interjections is connected with the emphatic, abrupt way in which they are uttered.
II. In names of relationship in the vocative. Original IN ama, "father", results in the Tontemboan amaṅ, with a particle welded on to it; but the vocative is amaq. Here the abrupt utterance has created the q.
III. In names of relationship generally. Thus alongside of Original IN pu, "grandfather", we find the Tontemboan apoq, beside Old Javanese bi, Modern Javanese bibi, "woman, aunt", the Madurese bhibbhiq, etc. These forms with q were originally vocatives.
IV. In numerals. In Madurese the numerals which in Original IN ended in a vowel are pronounced with a final q, whenever they are used by themselves, thus Original IN tělu > Mad. tělloq, "three", but těllo ratos, "three hundred". The occurrence of the q is either due to the force of analogy, on the pattern of empaq, "four" < Original IN pat, where the q is in accordance with rule, or else it is connected with the abrupt enunciation which sometimes occurs in counting.
V. In negatives, very frequently. Thus alongside of Original IN, and likewise Old Javanese, aja, a vetative negative, we find the Bugis ajaq, beside the Malay tiada a form tidaq, etc.
VI. In words of any category of meaning. In Busang an Original IN final vowel usually has a hamzah added to it; in particular, Original IN i becomes eq and u becomes oq, hence Original IN beli, "to buy" > Bus. běléq and Original IN batu, "stone" > Bus. batóq.
145. Just as in several languages the interjections often end in q, so in other languages they are particularly frequently found with an h as final, thus in Madurese.
146. Just as vocatives and negatives may have a final q added to them, so in certain languages they may receive a final a. Toba has a vocative aṅgiá corresponding to aṅgi, "younger brother"; alongside of the Tagalog negative di there is a Dayak dia.
147. Hamzah proceeds from certain Original IN sounds in conformity with phonetic laws:
I. From Original IN k in several languages, thus in Talautese; Original IN aku, "I" > Tal. iaqu = article i + aqu.
II. From r2 in several languages, thus in Makelaqi; Original IN jar2um, "needle" > Make. raqum. From h (see § 116).
148. While some languages double the consonant that follows upon an accentuated pěpět (see § 86), others develop a hamzah between the pěpět and certain consonants that follow it. In Makassar the pěpět becomes a, but the hamzah persists; thus Original IN kěděm, "to close the eyes", becomes kaqdaṅ in Mak.
149. Hamzah arises in several languages from special laws of the interior sounds of words, when the interior is of the taptap type (see § 198), thus in Tontemboan; hence Original IN pukpuk, "to break in pieces" > Tontb. puqpuk.
150. Hamzah results in many languages from special laws of the final sounds:
I. In some languages Original IN final k > q, thus in Malay, hence Original IN anak ^ Mai. anaq.
II. In Minangkabau all Original IN final explosives result in q, hence Original IN atep, "roof" ^ Mkb. atoq.
III. In Bugis, q results from all final consonants save the nasals and h, hence Original IN nipis, "thin" > Bug. nipiq.
151. Many IN languages add to the Original IN final consonant a supporting vowel, thus Hova, hence Hova ánaka, "child" < Original IN anak; several languages employ supporting vowels + hamzah, thus Makassar, hence Original IN nipis, "thin" > Mak. nipisiq.
152. Hamzah arises in sentence-sandhi from the abbreviation of words that are weak in stress. Thus the Sangirese preposition su can be pronounced q. We find in the "Children's Games", Bijdr. 1894, p. 520, 1. 2: "Yonder in the inland country" = dala q ulune for dala su ulune.
153. Initial and final hamzah may be lost in sentence-sandhi, either regularly or arbitrarily. Tontemboan aṅe, "hither", is pronounced qaṅe in conformity with § 142, but in the text Weweletěn (Sacrificial Prayers), Schwarz-Texts, p. 309, 1. 7, we find: "Come eat here!" = mai cuman aṅe, the q being lost. In Busang the word for "house" is umaq. But in the poem Boq Uyah Batang, p. 285, 1. 2, we find: "The house (named) Lang Děhaq" = umaq Laṅ Děhaq, and on p. 284, 1. 2, uma Laṅ Děhaq

The Law of the Mediae.

154. We meet with the law of the mediæ especially in Celebes and the neighbouring smaller islands, thus in Sangirese, Talautese, and Tontemboan, three languages that are closely related to one another, in Cenrana, and in Bugis; also apart from Celebes in Ibanag, Nias, Mentaway, and Hova.
155. I. The Sangirese law. In Sang. the media comes after a consonant; after vowels, the media g turns into the spirant ɣ, the media d into the liquid r, the media h into the semi-vowel w; thus the instantaneous mediae become continuants; and this holds good both of a single word and of words in a sentence. As initial of a single word pronounced by itself or at the beginning of the sentence, the media persists. Hence bera, "to speak", měqbera, the future active of the same, but iwera, the future passive. In the story in Bijdr. 1893, p. 354, 1. 1, we find: "I will tell of the ape" = iaq měqbio n baha, but 1. 4: "Said the ape" = aṅkún i waha.
II. The Talautese law agrees with the Sangirese, Thus the word for "house" is bale, as in Sang., and the word for "edge" is biṅgi, but in the Story of Parere, Steller-Texts, p. 89, 1. 2, we find: "At the edge of the river" = su wiṅgi it sáluka.
III. The Tontemboan law. The media g becomes ɣ in all cases; d and h interchange with r and w as in Sangirese. But as initial of a single word pronounced by itself or at the beginning of a sentence, the continuant is used, in contrast with the Sangirese usage. Original IN balay, "house", therefore, is Sang. bale but Tontb. wale; "to remain at home" in Tontb. is maqmbale. Within the sentence the law only operates in certain cases. Thus in the Story of the Newly Wed, Schwarz-Texts, p. 82, 1. 3 from the bottom, we find: "In the house" = am bale < an wale; but 1. 20: "Yet corals" = taqan wiwin, without alteration of the w.
IV. The Cenrana law. The mediæ d and h become r and w respectively after a vowel, e.g., dami, "only", but mesa rami, "one only".
V. The Ibanag law. Initial d becomes r, when an a is put before it: dakay, "badness", but marakay, "bad".
VI. The Bugis law. In Bug., initial w and r turn into b and d respectively, when a prefix is put before these sounds, no matter whether the prefix ends in a vowel or a consonant. Thus from wěnni, "night", are formed maqběnni, "to spend the night (somewhere)", and paběnni, " to cause (somebody) to spend the night (somewhere)", and from rěmme, "soft", maqděmme, "to soften", and paděmme, "to cause to soften". But the rule is not consistently carried out: from wětta, "to cut", comes maqbětta, "to cut off", but also pawětta-wětta, "headhunter". Evidently compromise has been at work here, and probably the regular rule is the one exemplified in pawětta in relation to maqbětta.
VII. The Nias law. When a WB begins with d or b and a prefix is put before these sounds, b becomes w and d becomes r; in similar circumstances x becomes g, thus conversely the continuant turns into a media. Thus bua, "fruit", but mowua, "to bear fruit"; dua, " two ", but darua, " to be a pair"; χaru, "dig", as a WB, but mogaru, the verb "to dig". In Nias also the rule is not strictly observed.
VIII. The Mentaway law. We always find the media b, never w instead of it. As initial the media g always appears, but in the interior of words g and ɣ interchange pretty irregularly. An examination of the whole of Morris' texts shows that the word for "banana (plantain)" occurs four times under the form of bago and twice under that of baɣo. "The media d is mostly a variant in pronunciation for r" (Morris).
IX. The Hova law. In Hova, Original IN initial g > h, hence hántuna < Original IN gantuṅ, "to hang", and húruna < Original IN guluṅ, "to roll". But Original IN k also becomes h, hence Hova húditra < Original IN kulit, "skin". When the prefix ma + nasal is put before h < k, the h < k disappears in conformity with § 16,* hence manuditra, "to peel". But if this same prefix appears before h < g, the g reappears, hence maṅgúruna, "to roll". But here too there have been changes based on analogy, for from hántuna is formed manántuna, instead of the maṅgántuna which one would have expected. From húdina < Original IN guliṅ, "to turn", is formed the verb manúdina, but the substantive saṅgúdina, "top".

Comparisons with Indo-European.

156. We do not find many parallels in IE to the four principal IN laws.
I. The IE indeterminate vowel turns into i or a (Brugmann KvG, § 127), just as in IN a and i (inter alia) result from é, but after the description in § 53 we cannot absolutely identify the pěpět with the IE indeterminate vowel.
II. In contrast with the IN r, the IE r is a very constant sound.
III. The hamzah plays but a very small part in IE. Just

* [See also Essay Til, § 30.]

as IN possesses many interjections ending in hamzah, so in the Lucerne dialect the word for “yes” is in certain cases pronounced yŏq instead of .
IV. With the law of the medise may be compared certain instances in Italian dialects, such as donna and la ronna (Gröber).




Preliminary Observations.

157. The two most interesting phonetic combinations in IN are the combination of a vowel with a semi- vowel and the combination of an explosive with the aspirate h. The first are called diphthongs, the second aspirates. In the diphthongs the semi- vowel may precede, as in the Dayak yaku, "I", or follow, as in the Tagalog patáy, "to kill". Only the second kind can lay claim to special interest, and we shall therefore deal only with it.
158. In many writings on IN subjects — to some extent, I regret to say, in my own former monographs also — the semi-vocalic components of the diphthongs have not been distinctly indicated. Thus in Malay textbooks one meets with such spelhngs as, for example, bau, "smell", and rantau, "coast", the end of the word being spelt in each case in the same way, though it is only in the second case that the end is a diphthong, while in the first word the a and the u belong to two separate syllables; I now write bau, and, on the other hand, rantaw.
159. Madurese spelhng has no means of denoting aspiration, and accordingly spells ghuluṅ, "to roll", and gulun, "a delicacy made from glutinous rice", in the same way.

The Diphthongs and their Laws.

160. The IN diphthongs are mostly combinations of vowels with one or other of the two semi-vowels y and w. Other possible combinations are rare; Original IN final i becomes in certain Achinese dialects the diphthong oy, thus Ach. bloy, "to buy" < Original IN bĕli, but the standard dialect

says bloe, with a semi-vocalic e.

161. The IN diphthongs only appear exceptionally in the accentuated syllable of the WB. That is a contrast with IE, if we call to mind such cases as the Greek kairios, "fitting", or the Gothic skaidan, "to sever". Hova has certain cases like táwlana, "bone", where contraction has produced the diphthong, as there is an Old Javanese tahulan corresponding to táwlana. Mentaway has also a few cases, e.g., räwru, "to journey downstream"; they are mostly words of which the etymology is obscure.
162. In by far the greater number of the cases the IN diphthong appears in the final syllable of the WB, and constitutes the end of it. Though that is usually the syllable that does not bear the principal accent, its sonority is, nevertheless, not very much weaker than that of the accentuated syllable (see §§ 329 seqq.).
163. The diphthongs most commonly met with in the IN languages are aw, ay, and uy, and these we must ascribe to Original IN. The words paraw, "hoarse", baṅaw, "heron", patay, "to kill", balay, "house", apuy, "fire", babuy, "pig", which are found in many IN languages, must also be regarded as Original IN, the first of them in the form par2aw.
164. The Original IN diphthongs aw, ay, and uy, have undergone various vicissitudes in the living IN languages :
165. In many languages the diphthongs, as mentioned in § 163, have been preserved. Iloko has, for example, the words patáy, apuy, etc.
166. The a of ay and the u of uy may, under the influence of the y, be modified in the manner technically termed "Umlaut", hence Dayak atäy, "heart" < Original IN atay, Bontok fafüy, "pig" < Original IN babuy.
167. The first component of the diphthong may change into another vowel, thus in certain dialects of Borneo both ay and uy appear as oy, hence patoy as well as baboy. — As Original IN patay results in Bontok in padöy, we must there assume patoy as an intermediate stage.
168. The diphthongs become simple sounds:
I. The first component of the diphthong disappears, as in Malay api < Original IN apuy.
II. The second component disappears, as in Hova afu < Original IN apuy.
III. The two components unite into a simple vowel, the sound of which lies intermediately between the two components, as in Toba pate < patay and poro < paraw.
169. The Original IN vowel sequences au and ai, where the vowels belong to two distinct syllables, as in tau, "man", lain, "other", are in several languages contracted to o and e respectively, hence e.g., Old Javanese len < lain. Here we must assume their pronunciation as diphthongs as an intermediate stage, thus: taw and layn.
170. This contraction to simple vowels takes place:
I. Without limitations, in several languages.
II. Only when the word is burdened with an enclitic, in Karo. Thus, "water" = Karo lau, "his water" = lo-na, "distance" = dauh, "his distance" = doh-na.
171. In several IN languages new diphthongs have arisen, which do not therefore represent Original IN diphthongs:
I. Several languages turn the i and u of the final syllable of the Original IN WB into ey and ew respectively, for example, Tiruray; hence Tir. taley, "rope" < Original IN tali, Tir. fitew, "seven" < Original IN pitu.
II. Other languages turn i into ay or oy, u into iw or aw. Thus Original IN bĕli, "to buy", becomes in Daya-Achinese blay, in Tunong-Achinese bloy; Original IN batu, "stone", apppears as batiw in Lanniga-Achinese, and as bataw in Miri, a Borneo dialect.
III. Sĕraway turns Original IN final a into aw; thus Sĕr. mataw", "eye" < Original IN mata.
172. All the cases of diphthongization mentioned in the preceding paragraph occur as a rule only when the original vowel, from which the diphthong has proceeded, was a final. This phenomenon runs parallel with the fact that the Original IN diphthongs aw, ay, and uy also appear only as finals. Only in certain languages of Borneo are vowels closed by consonants also sometimes turned into diphthongs, thus in Dali and in Long Kiput, hence Dali laiiayt, "sky" < Original IN laṅit and Long Kiput pulawt, "gum" < Original IN pulut.
173. Diphthongization may also be the result of various other phonetic processes. Such processes are:
I. Vocalization of consonants, in Lampong. e.g., ikuy, "tail" < Original IN ikur2.
II. Reduction of vowels that originally belonged to two distinct words, as in Bangkalan saybu, "a thousand" < sa + ebu.
III. Reduction of vowels after the droj)ping out of a consonant, as in Hova fay, "ray (fish)" < Original IN par2i.
174. When originally simple vowels are turned into diphthongs, as in Tiruray taley, "rope" < tali, the process of diphthongization must have been preceded by length of the vowel; from § 77 we know that final vowels are often long.
175. As stated in § 76, syllables coming before the accentuated syllable are very seldom long, and similarly diphthongs very rarely occur before the accentuated syllable. In a few Sanskrit loan-words beginning with s, Lampong has the diphthong ay in that position, as in saygara, "sea" < Sansk. sāgara.
176. I know of no case of the pĕpĕt turning into a diphthong. After what has been said in § 40, I, there can be no question of the diphthongization of Original IN e or o.

The Aspirates and their Laws.

177. As the majority of the IN languages only tolerate combinations of consonants to a limited extent, some of them merely the combination of a nasal with a cognate explosive, the aspirated consonants are not widely distributed in IN.
178. Aspirated consonants have arisen in the hving IN languages in the following ways :
I. They are found in WB's formed by the doubling of roots having h as their initial and an explosive as their final, as in Old Javanese hathat, "to take care", Bisaya hagliag, "texture". These cases are not numerous.
II. A few languages, like Tagalog, allow the combination of most of the consonants with a subsequent h, and consequently also that kind of combination which we call the aspirates. Hence in Tagalog a word like bugháw, " blue ",is just as permissible as a word like panhik, "to ascend".
III. In Madurese, aspirates arise in conformity with phonetic law by the change of Original IN mediæ into aspirated mediæ; hence Original IN gantuṅ, "to hang" > Mad. ghantoṅ; Original IN jalan, "path" > Mad. jhalan; Original IN dagaṅ, "stranger, trader" > Mad. dhaghah; Original IN kĕmbaṅ, "bud" > Mad. kĕmbhaṅ, "flower".
IV. In a few languages, as in Cham and Achinese, aspirates arise through the elision of vowels. Hence Original IN pohon, "tree" > Cham phun; Original IN tahu, "to know" ^ Ach. thee.
V. Aspirates are found in loan-words from the Sanskrit, as in Tagalog katha, "speech".
VI. Achinese renders the Arabic f by ph, as in kaphé, "the infidel".
179. Only the few isolated cases of aspiration mentioned under § 178, I, can be ascribed to Original IN. There is therefore a great difference in the relative importance of the aspirates as between Original IN and Original IE.




Preliminary Observations.

180. We have often been able to notice in Section IV that the condition under which a phonetic change takes place consists in the fact that the sound in question is an initial, interior, or final sound. Thus Original IN a changes in Bĕsĕmah into ĕ when it ends a word, but otherwise remains unchanged. We will not discuss these phenomena again in this place, but will deal with a series of phonetic facts which are particularly characteristic of the nature of the initial, interior, and final sounds and which we have, on that account, reserved for this Section. Hereto belong also the phenomena connected with the enunciation of vowels, whether as initial, medial, or final, in words.

Initial, Medial, and Final Enunciation.

181. The enunciation of IN words, that begin with a vowel, may be soft, hard, or aspirated. Hard enunciation, that is to say the sounding of a hamzah before the vowel, is evidenced for a considerable number of IN languages (see § 142), and we may therefore with very fair certainty ascribe this phenomenon to Original IN. Moreover, hard and aspirated enunciation may interchange. "At the beginning of Achinese words h and q occasionally interchange, either because one dialect uses q and another uses h, or because the choice between them is left to the fancy of the individual speaker " (Snouck Hurgronje). The Minangkabau vocabulary contains a great number of words beginning either with or without h such as hindu and indu, "mother".
182. Initial hamzah is replaced in certain languages by other sounds : ỿ, y, w.
I. In Muna we find ỿ, as in ỿate, "heart" < Original IN atay, or more precisely: qatay.
II. In Buli we get y, as in yataf, "roof" < Original IN atĕp, or more precisely: qatĕp.
III. In Bulanga-Uki we meet with w, as in wina, "mother" < Original IN ina, or more precisely: qina.
183. The occurrence of these sounds, ỿ, y, and w, instead of q, is explained by the phenomena of sandhi. In Malay the word ĕmpat, "four", when it stands alone or at the beginning of a sentence, is pronounced qĕmpat. In the phrase "four pieces of sugar-cane" = s. f. p. = tĕbu wĕmpat buku, the initial q is replaced by w", under the influence of the preceding u (Fokker). Now in Bulanga forms such as wina this w has simply become permanently affixed, and the y of Buh and ỿ of Muna are the result of similar processes.
184. Medial enunciation. In some IN languages the most various vowels may succeed one another. The sequence of vowel + pĕpĕt or pĕpĕt + vowel is rare; it is found in a few instances in Madurese, as in taĕn, "rope". In some languages intermediary or separating sounds arise between two vowels. And here two cases in particular are to be observed:
I. Between u + vowel or i + vowel the appropriate semi-vowel steps in. Some of the IN languages say buah, "fruit", others buwah; some say ia, "he", others iya.
II. In some languages q or h appears between vowels, especially when they are similar, as in Malay leher, Madurese leqer, "neck".
185. Final enunciation of a word ending in a vowel may, like initial, be soft, aspirated, or hard (with a hamzah). In Madurese every word ending originally with a vowel receives a final h, Mad. matah, "eye" < Original IN mata. — In Busang in the like case a hamzah is used, and i turns into e, u into o. Table of examples:
Original IN lima Busang limáq, "five"
děfa depáq, "span"
buta hutáq, "blind"
běli beléq, "to buy"
laki lakéq, "man"
tali taléq, "rope"
asu asóq, "dog"
batu batóq, "stone"
kayu kayóq, "tree".
186. In Madurese all three grades of enunciation may occur in the same word. If a word in Original IN ends in a vowel or diphthong, as mata, "eye", laju, "to proceed", patay, "death", the word is pronounced in Mad. with aspirated enunciation, hence matah, pateh, lajhuh, "thereupon". In the interior of a sentence the aspiration is lost, and thus in the texts appended to Kiliaan's Grammar, I, p. 124, 1. 12, we find: "Thereupon (he) died" = lajhu mateh. Before a pause, due to the speaker being at a loss how to proceed, the word is pronounced with a hamzah, thus lajhuq ... mateh.

The Initial.

187. IN words can, as a rule, begin with a vowel, a semi-vowel, or a simple consonant; and this state of affairs is to be regarded as Original IN. In connexion therewith the following points are also to be noted:
I. Before the initial vowel some languages sound a hamzah (see § 181).
II. Few words in the IN languages begin with the semi-vowel y, and none of them can be shown to be Original IN. Initial w is more frequent, but it mostly originates from b. Of Original IN w there are only three cases: walu, "eight", wara, "to be, to exist", and way, "water".
III. In contrast with the IE languages, is not a rare phenomenon among the initial consonants.
188. Some IN languages also admit an initial formed of two consonants. The commonest cases are mute + liquid and nasal + cognate explosive. In connexion with initial double consonants the following points are also to be noted:
I. Initial consonantal sequences are the same as those that occur in the interior of words; thus in Nias, where we have, for example, mb both initially and medially, as in mbawa-mbawa, "spotted", alongside of mambu, "to forge (as a smith) ".
II. The initial allows fewer consonantal sequences than the interior of words, as in Hova. The sequence n + t + s does in fact occur medially, e.g. in untsi, "banana (plantain)", but not initially.
III. The initial admits of more consonantal sequences than the interior of words, as in Rottinese. The sequence n + d does in fact occur initially, as in ndala, "horse", but not medially.
189. Three consonants, mostly nasal + cognate explosive + liquid or semi-vowel, rarely occur initially. Nias, for example, has n + d + r, as in ndrundru, "hut". Old Javanese has n + d + y as in ndya, "where, what".
190. When words begin with two or three consonants, they usually remain unchanged in all positions in the sentence. In the Gayo text, "The Blue Princess", p. 46, 1. 4, a sentence begins with a word having the initial nt: "That I may not marry" = N. I m. = nti aku kĕrjön. In Rāmāyana, VIII, 171, the above-mentioned ndya occurs after the word toh, "well", which ends with a consonant.
191. Initials of more than one consonant are not Original IN; they have arisen through various phonetic processes:
I. In conformity with phonetic law, e.g. in Hova. Original IN d changes in certain cases in Hova into tr, as in trúzuna < Original IN duyuṅ, "sea-cow".
II. Through loss of a vowel, as in Gayo bli < Original IN bĕli, "to buy", or in Makianese into, "eye" < Original IN mata.
III. Through word-formation. Alongside of the Old Javanese ndya we find an Old Jav. ndi and a Toba dia, with the same meaning; accordingly ndya is analysable into the three elements n + di + a. I have dealt with such combinations of words of form in previous monographs.
192. Through the process of word-abbreviation (§§ 274 seqq.) initials arise which would not otherwise be possible in the languages in question. Examples:
I. In Tontemboan k is pronounced c when an i comes before it, either in a word or in a sentence, but in no other case. Thus from the two elements raqi + ka comes the negative raqica. This is often abbreviated into ca, and then the initial c is allowed to remain unchanged even if there is no i before it, as several passages prove. Thus in Schwarz-Texts, p. 61, 1. 17, we find: " He said: He does not catch " = kuanao : ca maindo.
II. In conformity with the law of the mediæ (§ 155, III) an initial media must become a continuant in Tontemboan. But in proper names abbreviated as explained in § 276 the media persists, as in Biraq, abbreviated from Imbiraṅ, a personal name.

The Medial.

193. In the interior of the IN WB, which is mostly disyllabic, thus between the two vowels of it, we meet either with no consonant, or one, or two, but rarely three.
194. Of the cases where there is no consonant, or only one, between the two vowels, there is nothing more to be said.
195. Among the cases where two consonants occur between the two vowels, two types in particular are frequent, the Lintah-type and the Taptap-type. Both are to be ascribed to Original IN.
I. The Lintah-type. In almost all the IN languages the combination of nasal + cognate explosive is permitted medially. Thus, for example, the word lintah, "leech", recurs, with n + t, in nearly all the IN languages.
II. The Taptap-type has originated from the doubling of the root, as in Old Javanese, etc., taptap, "to strike".
195. Now some of the living IN languages have preserved both the Original IN types, others have modified them.
197. The Lintah-type has remained unchanged in by far the greater number of IN languages; only a small percentage of the languages has altered it, entirely or partially, and in the follomng ways:
I. Some languages, such as Toba, assimilate the nasal in the combination of nasal + tenuis to the tenuis; thus spoken Toba gattuṅ, "to hang" < written Toba, and likewise Original IN, gantuṅ.
II. Some few languages allow the nasal to disappear entirely; thus Nias, as in lita < Original IN lintah. But mb and ndr < nd persist, as in tandru, "horn" < Original IN tanduk, tandra, "mark" < Original IN tanda.
III. Conversely, other languages allow the explosive to disappear; thus Rottinese, as in tana, "mark" < Original IN tanda.
198. The Taptap-type has remained unchanged in Old Javanese, Karo, Tagalog, etc., but yet in fewer languages than the Lintah-type. The modifications it has undergone are of the following kind:
I. Assimilation has taken place, as in Makassar; Original IN, and likewise Old Javanese, paspas, "to cut off", appears in Makassar as páppasaq = pappas + the supporting syllable aq.
II. The first of the two medial consonants becomes a hamzah, as in Tontemboan, which accordingly says taqtap instead of taptap.
III. The first of the two consonants disappears entirely,, as in Bĕsemah, hence Bĕs. tatap.
Note. — The rules mentioned in this paragraph do not hold good of all possible cases of the Taptap-type, but they always apply to certain classes of cases, determined by regular laws.
199. Sequences of three consonants in the interior of words are rare and cannot be ascribed to Original IN. They arise chiefly in two ways:
I. By mere operation of phonetic law. The sequence nd in Nias becomes ndr, hence Nias tandra < Original IN tanda, "mark".
II. By the springing up of intermediary sounds. From the Old Javanese WB prih is derived the verb amrih, "to strive", for which Madurese has ambri, the b having arisen as an intermediary sound forming the transition from the m to the r. Of like origin is Modern Javanese ambral < amral < admiral.

The Final.

200. In Original IN a word could end in a vowel, a diphthong, or a single consonant, other than a palatal (see § 61, 1). A final palatal is quite rare in the living languages, though found in Tontemboan as the resultant of k (see § 103).
201. The developments in the living languages of the Original IN final vowels have been dealt with in §§ 91 seqq., those of the diphthongs in §§ 160 seqq.; those of the consonants, which can lay claim to very special interest, will now be discussed.
202. The Original IN condition as regards final consonants persists unchanged in Old Javanese, and also, with very few exceptions, the loss of h for example, in several of the Philippine languages.
203. In the remaining IN languages we can discern three tendencies in the treatment of consonantal finals: unification, loss, and addition of a supporting vowel.
204. Unification. This is applied in various degree in the several languages, as will be shown here by reference to the explosives:
I. Malay unifies the mediæ with the tenues. Original IN bukid, "hill" > Mal. bukit. Thus among the explosives only the three tenues k, t, and p, are in this case capable of serving as finals.
II. Masaretese behaves like Malay and further unifies p with t, hence Original IN atěp = Malay atap = Masaretese atet, "roof." Here, therefore, among the explosives only two tenues, k and t, are capable of serving as finals.
III. Primitive Bugis, as I showed in a former monograph, unified all the explosives into k, hence Primitive Bug. laṅik "sky" < Original LN laṅit, and atěk, "roof" < Original IN atěp. Here, therefore, out of all the explosives only the one tenuis k is capable of serving as a final.
205. Loss. The disappearance of the final consonants takes place in the several IN languages in various degrees:
I. Makassar allows only one consonant, namely h, to disappear, as in panno < Original IN pěnuh, "full".
II. In Hova, s, h, and the liquids disappear, hence Hova manifi < Original IN nipis, "thin", fenu, "full" < Original IN pěnuh.
III. In Bimanese, Nias, and some other languages, all final consonants disappear.
206. Addition of a supporting vowel. In this connexion two tendencies may be discerned among the several IN languages :
I. The same supporting vowel is added in all cases; in Talautese and Hova a, in Ampana i, in Kaidipan o, or exceptionally u, etc. Hence Original IN inum, "to drink" > Tal. ínuma = Hova ínuna = Kaid. ínumu; Original IN putih, "white" > Kaid. pútiho.
II. The supporting vowel imitates the vowel that immediately precedes the final consonant, as in Mentaway, hence Ment. túkulu, "to push", alongside of Karo tukul, but rápiri, "wall", bóbolo, "a species of lily", etc.
III. A few languages, such as Makassar, further add a hamzah after the supporting vowel, hence Original IN nipis, "thin" > Mak. nipisiq, lěpas, "free" > Mak. láppasaq, atur, "to put in order" > Mak. átoroq.
IV. The consonant saved by the supporting vowel may nevertheless disappear owing to further phonetic processes, while the supporting vowel may be preserved, as in Ambon; hence Original IN tuwak, "palm-wine" > Amb. túwao, Original IN atěp, "roof" > áteo.
207. Some of the IN languages only recognize one of the tendencies dehneated in §§ 204 seqq., others two, others again all three of them.
I. Nias only has loss of the final: all Origmal IN final consonants disappear.
II. Minangkabau has both unification and loss. The explosives are unified to q, the liquids disappear. The nasals and h persist, s becomes h.
III. Makassar employs all the methods. The aspirate disappears, the nasals are unified into , the explosives into q, the liquids and s receive a supporting vowel + hamzah.
208. In all the IN languages we meet with the phenomenon that final consonants are interchanged. In Malay alongside of butir, "grain", which is in conformity with phonetic law, there is also a form butil. Hova has as a pendant to the Malay burut, a phonetically regular word wúrutra, but alongside of it it also has a form wúruka, "broken, torn, rags", etc. This phenomenon occurs everywhere in isolated cases, mostly only in a few cases. Probably these are due to prehistoric formative processes, or to the working of the principle of analogy, and the like.
209. We very often meet with the phenomenon that in some language a word ends in a consonant, while the Original IN and some of the other living languages have a vowel as final. Here we are dealing with words of form that have become annexed to the original word. "How much" in Original IN is pira but in Makassar piraṅ; "this" in Old Javanese is ika, but in Modern Jav. kaṅ. The is an article welded on, originating from such formulas as the Old Jav. for "this child " = Greek tuto to teknon = Old Jav. ika ṅ anak. Such annexed articles are also not uncommon in the IE sphere; we find one for example in the French lierre, "ivy", etc,
210. Now, when a suffix is added to a final that has been modified through the influence of the laws affecting finals, we observe the following phenomena:
I. The original state of the final, as it was in Original IN, again appears. When from the Bugis nipiq, "thin" < Original IN nipis a verb "to thin" is formed by means of the suffix -i, it does not take the form nipiqi but nipisi. More correctly expressed, the formation nipisi has been handed down from a period when people still said nipis.
II. The derivative exhibits the modern state of the final. Original IN baṅun, "to stand up," appears in Makassar as baṅuṅ, and from it is derived the verb baṅuṅaṅ, "to raise". This formation dates from a period when n had already turned into .
III. The derivative displays an intermediate state of the final, a stage of development lying between the Original IN and the modern form. As shown in a previous monograph, Original IN sělsěl, "to regret", turned in Primitive Bugis into sěssěr, whereof the Modern Bug. has made sěssěq. The derivative "reproof" in Modern Bug. is pasěssěrrěṅ; it dates from a period when people no longer said sělsěl, and had not yet begun to say 'sěssěq.
IV. The derivative has both the original and the modified state of the final, side by side. Original IN lěpas, "free", appears in Minangkabau as lapeh. The derivative "to free" is both malapasi and malapehi. To this there is an exact parallel in Hova. Original IN lěpas appears in Hova as lefa. But the passive imperative is both alefasu and alefau.
V. The derivative shows none of the forms we should be disposed to expect; for just in this sphere there have been many cases of analogical transference. Original IN ěpat, "four", appears in Primitive Bugis as ěppaq, in Modern Bug. as ěppaq, but the derivative "to divide into four" is ěppáti. This formation is based on the analogy of words like appaq: appári, "to display", where the r is in conformity with phonetic law, for Malay and other languages have the form hampar.
211. Behaviour of the supporting vowel in derivatives and with enclitics:
I. On the addition of a suffix the supporting vowel is dispensed with. From Makassar sássalaq < sělsěl comes the derived verb sassáli, "to refuse".
II. Before enclitics we find both persistence and disappearance of the supporting vowel. In the Makassar romance Jayalangkara, p. 72, 1. 9, is the expression: "The people of Egypt" = P. E. the = tu-Máserek-a, < Meṣir has the supporting syllable -eq, and this persists before the enclitic article, the hamzah turning into a k. In the Hova Fables of Rahidy, V, 1. 3, we find: "Killed by him" = nuwunúini, for nuwunúina + ni. Here the supporting vowel has disappeared, and the two n<nowki>'</nowiki>s coming together have coalesced into one.

The Final in Rottinese.

212. Rottinese displays peculiar phenomena in the final, which require special consideration.
213. In Rot. three of the consonants are capable of serving as finals, viz. k, n, and s, Original IN awak, "body" > Rot. aok; Original IN ur1an, "rain" > Rot. udan; Original IN nipis, "thin" > Rot. niis. But Original IN běr2at, "heavy" > Rot. belak; Malay, etc., ruaṅ > Rot. loak, "room".
214. One might endeavour to explain these facts by the principle of unification (§ 204).
215. But we are faced by a circumstance which excludes any explanation based on the principle of unification; the circumstance is this, that in an extraordinary number of cases the final is different from what one would expect on general linguistic principles or from IN usage. Examples: Original IN jalan, "path" > Rot. dalak; r2atus, "hundred" > natun; matay, "to die" > mates, "dead"; lidi, "nerve of a leaf" > lidek. And alongside of the above-mentioned niis > nipis there is a form niik.

216. To explain this state of affairs, one might then have recourse to the principle of the interchange of finals, as in § 208. But this is countered by the fact already mentioned, that the phenomena illustrated in the preceding paragraph are exceedingly common in Rot., whereas interchange of finals in other languages only occurs in isolated cases.

217. Accordingly we must look around for another explanation, namely the following one:

I. Negative part of the explanation. The cases in question are not really instances of a law affecting finals. The three final consonants, k, n, and s, are not the representatives of Original IN finals; even the n in udan is not a survival of the n of Original IN ur1an.

II. Positive part. In a former period of its existence Rot. cast off all its final consonants; there was, therefore, a time when it said dala, "path", uda, "rai", nii, "thin". This condition is the same as the closely related Bimanese has preserved to this day, e.g. in ura, "rain", nipi, "thin", etc. The finals which occur nowadays in Rot. are articles, which have become annexed and have lost their force, like the ones mentioned in § 209.

218. Articles and demonstratives beginning with k, s, or n, are found very frequently in the IN languages. It is also no uncommon phenomenon for such words of form to occur without a vowel. The Old Javanese article for things is aṅ and . The Bontok article for persons si is often abbreviated into s, and so is the Inibaloi si. Thus in the text "Kalinas", in Scheerer, "The Nabaloi Dialect", p. 149, 1, 5, we find: "I have met the captain" = H. + m. I the c. = inaspol ko s kapitan.

219. The article is put after the principal word in many IN languages, and particularly in those that are near neighbours. and relatively close connexions of Rot., for example in Sawimese. Thus in the Sawunese Story of Pepeka, Bijdr. 1904, p, 283, 1. 10 from the bottom, we find: "In the cave" = In c. the = la roa ne. The Modern Rottinese article a also follows the principal word.

220. The fact that we therefore have to credit Rottinese with four articles, k, n, s, and a, makes no difficulty, for the number of articles in Bugis is even larger, as I showed in a previous monograph.
221. We therefore assume that in a former period of its evolution the final in Rottinese had become exclusively vocalic, and that in the modern period it has again to a large extent resumed the consonantal form owing to the welding on of articles containing no vowel. The assumption of such a divergency in development involves no impossibility. As was remarked above, the Bimanese is a language with purely vocalic finals, yet it employs certain enclitic pronouns even in forms devoid of vowels. "Child" in Bim. is ána < Original IN anak, "my child" is aná-ku or aná-k. And such forms even occur at the end of a sentence; thus in Mpama Saṅaji Ali in Jonker's "Bimaneesche Texten", p. 55, 1. 15 from the bottom, a sentence ends with the words: "At the house of our prince" = At p. o. = labo rumá-t. In Bim. such pronouns are still mobile, they have not been welded together with the principal word to form a new WB as in Rot.
222. The crucial test of the correctness of these conclusions consists in the following: If the finals k, s, and n, are articles that have been annexed and have lost their original force, they must not occur in verbal words, vocatives, or the like. And that is really the case. The word taek, "young man", is tae in the vocative. "To rain" is uda, "(the) rain" udan. Accordingly the Original IN ur1an has undergone the following development in Rot.:
Original IN ur1an.
Rot., older period uda.
Modern Rot., verb uda, "to rain".
Modern Rot., substantive udan, "rain".
It is true that there have also been compromises and cases of transference. In niik or niis (§ 215) we should not have expected to find an annexed article but a vocalic final.
223. Phenomena similar to those of Rot. are also exhibited by some other languages spoken on islands in the same part of the sea, for example by Timorese.

Comparison with Indo-European.

224. The IN phenomena of initials, medials, and finals have a very large number of parallels in IE. Only a few of them will be selected for mention here :
I. Old Slavonic and Buli. Before an originally initial vowel an i-sound appears. Buli yataf < Original IN atĕp; for the Old Slavonic see Leskien, "Grammatik der altbulgarischen Sprache", § 57.
II. Greek and Madurese. Between medial m and r the intermediary sound b appears. Mad. ambri < amrih (§ 199); Gr. mesēmbria, "noon", alongside of hēmera, "day".
III. Portuguese dialect of Alta Beira and Talautese. Final consonants receive a supporting vowel. Alta Beira deuze, "God" ; Tal. inuma < inum, "to drink".




225. In this Section will be discussed certain phonetic phenomena which occupy a somewhat special position and are also usually dealt with separately in the IE textbooks. These phenomena are: prothesis, anaptyxis, repetition of sounds, metathesis, haplology, assimilation, "Umlaut", dissimilation, and fracture.

226. The most frequently occurring kind of prothesis is the affixing of a pĕpĕt before words that were originally monosyllabic or had become monosyllabic through some process of phonetic change. The cause of this phenomenon is the tendency towards disyllabism (§ 19). Old Javanese goṅ, "gong", appears in Modern Jav. as ĕgoṅ as well as goṅ. Original IN dur2i, "thorn", passes through rur2i into Old Jav. rwi, in accordance with the principle mentioned in § 137, and then in Modern Jav. undergoes a further evolution into ri, by the side of which now appears a form ĕri. Dutch lijst, "list", appears in IN languages as les and ĕles.

227. This prothetic e is also subject to the phonetic laws of change; hence "gong" in Toba is oguṅ, as Toba alters the pĕpĕt into o.

228. Instead of the pĕpĕt, i may also appear before y and u before w. The Old Javanese conjunction ya, "that", is likewise ya in Tontemboan, but people also use both the forms ĕya and iya, in accordance with what has been said above. Original IN buwah, "fruit", becomes Old Jav. wwah; from this is regularly derived the Modern Jav. woh, but alongside of the latter there is also a form uwoh. 229. Besides this prothetic ĕ, i, or u, we meet in various languages with yet other vowels affixed before words that originally had a consonantal initial. Original IN, and likewise Old Javanese, Malay, etc., lintah, "leech", appears in several languages as alintah; Original IN tĕlur2, "egg", is represented in Tagalog by itlóg, etc. I am not in a position to decide whether these are cases of a purely phonetic process or whether we are here dealing with creations of a formative import. Parallel phenomena offering similar difficulties of explanation are also found in IE : see (inter alia) Hirt, "Handbuch der griechischen Laut- und Formenlehre", §§ 193 seqq.

230. Anaptyxis is found in Pabian-Lampong, where an ĕ appears between r and the immediately following consonant. To Malay, Karo, etc., sĕrdaṅ, corresponds a Lamp, form sarĕedaṅ, " a species of palm". This sort of anaptyxis may be compared in the IE sphere with phenomena hke the Oscan aragetud = Latin argento. Of another kind is the appearance of vowels between consonants in loan-words, where facilitation of pronunciation is the cause of the phenomenon (see § 284).

231. Repetition of sounds may affect vowels or consonants, may be progressive or regressive, and may occur merely in isolated cases or in series of cases.

232. When in Hova an i is put before a velar, it is without exception repeated after the velar, in consonantal form and at the same time very softly pronounced. " Surprised " in Hova is gaga, and "to be surprised " is not migaga but migyaga.

233. Bajo "fractures"* an a of the second syllable of a WB into ea, when the word ends in ; Original IN bĕnaṅ, "thread", therefore appears in Bajo as heneaii. In the one case of geantéaṅ < Original IN gantan, "a particular measure of capacity", the e has been repeated in the first syllable.

234. In very many cases in IN languages there appears before a consonant a nasal, which is wanting in other IN languages, and which we cannot ascribe to Original IN either. The word for " brain " in some of the languages is utĕk, in others untĕk; "to pursue" is usir and uṅsir in Old Javanese. Now there are in the IN languages a great number of very commonly used prefixes and suffixes containing nasals, and it is from these that the nasal has forced its way into the interior of the WB by means of the principle of the repetition of sounds. From the Old Jav. WB usir, for example, comes the active aṅusir or maṅusir, and from this aṅ- or maṅ- the has been projected into the variant form uṅsir of the WB.

235. Repetition of sounds is also found in IE, and IE research avails itself of the same explanations that have been applied above to IN: see (inter alia) Zauner, " Altspanisches Elementarbuch ", § 78.

236. Metathesis is a phenomenon of very frequent occurrence in IN ; it turns up in all sorts of forms, either sporadically or in regular series.

237. The following are the sorts that occur most frequently :

I. The vowels of the two syllables of the WB change places. Original IN, and likewise Malay, etc., ikur, "tail", is pronounced ukir in some other languages; thus in some dialects of Madagascar: uhi < ukir.

II. The consonants of the first half of the word change places. Original IN, and likewise Old Jav., etc., waluh, "pumpkin", appears in Bugis as lawo.

III. The consonants of the second half of the word change places. Original IN r2atus, "hundred", appears in several languages as rasut.

IV. Two interior consonants change places. Toba purti < Sanskrit putrī, "daughter".

238. Tontemboan possesses a peculiar, optional kind of metathesis, which will be illustrated by the following example. In the Story of the Poor Woman and her Grandchild, Schwarz-Texts, pp. 107 seqq., we find on p. 110, 1. 5: "Why should we respect?" = What the cause-for-respect = sapa ĕn ipĕsiriq. On p. 109, 1. 1, we find as an equivalent for the same phrase sapa im pĕsiriq. From ĕn ipĕsiriq there has been metathesis to in ĕpĕsiriq, then the initial e has disappeared, making in pĕsiriq, and finally by assimilation im pĕsiriq has resulted.

239. Metathesis may either be definitive or optional, permitting both forms, the original and the modified one, to exist side by side. Original IN par2i, "ray (fish)", appears in Tontemboan as pair, and in no other form ; but in Sundanese ayud and aduy, "soft", exist side by side.

240. We notice in various languages a certain preference for particular kinds of metathesis.

I. The preference is connected with the position of the sounds in the word. The Mantangay dialect of Dayak favours metathesis in the case of the first syllable of trisyllabic words; it has dahaṅan for the hadaṅan, "buffalo", of the standard Dayak dialect.

II. The preference is related to a particular result. In Sawunese metathesis mostly operates so that an a of the second syllable comes into the first one, the a also changing into ĕ; hence Original IN pira, "how much" > Saw. pĕri, r2umah, "house" > ĕmú, etc.

241. In certain languages we meet with metatheses occurring in regular series. When in Original IN an l immediately precedes the second vowel of a word and an r immediately follows it, then these consonants invariably change places in Gayo; hence Gayo tĕrul, "egg" < Original IN tĕlur2, arul, "brook" < alur2, etc.

242. Haplology. This occurs in IN in the first place sporadically in various languages, thus in Tsimihety. In "Chansons Tsimihety", Bulletin de I'Académie Malgache, 1913, p. 100, V. 10, we find mañi-reboṅo, "growing densely", for maniri-rehono, from the WB tsiri, "to grow".

243. Haplology also occurs in most regular series in connexion with the doubling of words. Here either the first or the second term of the duphcation may be abbreviated by the method of haplology. The first kind, the abbreviation of the first term, appears in every imaginable form, whereof we now give a selection :

I. The last sound is omitted: Dayak luyu-luyut, "somewhat soft", from luyut, "soft"; similarly aki-akir, "to push slightly", etc.
II. The last two sounds are omitted : Bull lis-lisan, "broom", from lisan, "to sweep".
III. All sounds except the last are omitted : Tontemboan u-anu, "such and such a person", "Mr. What's his name", from anu, "some one".
IV. All sounds except the first are omitted ; Mentaway o-ogdog, "a tool for opening coconuts", from the WB ogdog.
244. The second term is abbreviated by the method of haplology. This case is rare. Padoe laqika-ika, "hut", from laqika, "house". Javanese Roso-so, familiar mode of addressing a person whose name is Roso.
245. As all the species of haplology instanced in §§ 243 seqq. are cases of the eUmination of sounds that were not in direct contact wdth the corresponding similar ones, they are analogous to the haplological phenomena of IE described in Brugmann KvGr, § 338, A, 2, like the Latin latrocinium < latronicinium.
246. Assimilation displays in IN all the possibilities that also occur in IE: compare the cases in Brugmann KvG, §§ 319 seqq., with the following scheme. In the IN languages assimilation may be:

Ia. Vocalic: Original IN tau > Tontemboan tow, "man".

Ib. Consonantal: Original IN gantuṅ > Toba gattuṅ, "to hang".

IIa. Progressive: Original IN garuk > Bimanese garo, "to scratch".

IIb. Regressive: Original IN tau > Tontemboan tow.

IIIa. Unilateral: Original IN tau > Tontemboan tow.

IIIb. Mutual: Original IN aur > Bimanese oo, "bamboo".
IVa. With contact: Original IN gantuṅ > Toba gattuṅ.
IVb. Without contact: Original IN kulit > Loinan kilit, "skin".
Va. Partial: Original IN babuy > Bontok fafüy, "pig".
Vb. Complete: Original IN kulit > Loinan kilit.
247. Assimilation occurs in IN not only within the WB, but also, though not often, between the WB and the formative:
I. The formative affects the WB: Dayak tuli, "to moor", talian, "mooring place".
II. The WB affects the formative: Tontemboan sěraq, "to eat ", sěraqan, gerundive of the same < sěraq + formative ěn; siriq, "to honour", siriqin, gerundive of the same < siriq + formative ěn; and so with all the vowels, when the WB ends in a vowel + hamzah.
248. Assimilation of one part of a compound word by the other is rare. In Busang do, "day" + halěm, "past", becomes dahalěm, "yesterday".
249. Certain assimilations occur in IN in regular series. If in an Original IN word there is an l together with an r, then in Toba the l is in all cases assimilated to the r; hence Original IN lapar2 "to hunger" > Toba rapar. The assimilations between WB and formative mentioned in § 247 also occur invariably.
250. We find in several IN languages the change from Original IN final aya and ayu to ay. We may believe that in these cases there was first assimilation into ayi and then simplification of the y + i. Examples: Original IN kayu > Sigi kay; Original IN layar2, "to sail" > Hova lay.
251. The term Umlaut is really a superfluity in linguistic terminology, for it only denotes a species of partial assimilation. But the term is used in IN as well as in IE, and usually denotes the partial assimilation of the three vowels a, o, and u, under the influence of i.
252. Umlaut is fairly widely distributed in IN. Examples :

Umlaut of a > ä : Original IN lima > Dayak limä, "five".

Umlaut of a > e : Original IN hatay > Sumbanese eti, "heart".

Umlaut of a > ö : Original IN patay > Bontok padöy, "to kill".

Umlaut of o > e : Toba subdialect oyo > Toba eo, "urine".

Umlaut of u > ü : Original IN babuy > Bontok fafüy, "pig".

Note. — The symbol ä is used by Hardeland for an a modified by Umlaut in Dayak, and I have retained it.
253. Gayo has a sound similar to the German ö, but not originating through the modifying influence of an i, as in dödö, "breast" < Original IN dada'.
254. Umlaut may be a preliminary step towards more complete assimilation. Original IN lima appears in Dayak as limä; and Hova, which is very closely related to Dayak, has dimi.
255. Dissimilation is rarer in IN than assimilation.
256. Dissimilation occurs:
I. If two similar sounds would otherwise follow one another. The duplication of Modern Javanese ro, "two", takes the form of loro. Original IN babuy, "pig", and babah, "to carry", appear in Mandarese as bagi and baga respectively.
II. If three similar sounds would otherwise follow one another. Original IN aṅin, "wind", appears in Bugis as aṅiṅ, but maṅinaṅ, "to dry in the air", has preserved the n through the operation of dissimilation. As in waṅuṅĕṅ, "to raise", from waṅuṅ, "to stand up" < Original IN baṅun, the same principle has not been at work, we may assume that in maṅinaṅ the vowel i has been a contributing factor. — Or is this case simply an application of the rule in § 210, I ?
257. Dissimilation operates either with or without contact :
I. Dissimilation with contact: e.g. in Toba in the case of s + s, when these two sounds would otherwise come together, e.g. in lat-soada, "not yet" < las, "yet" + soada, "not". II. Dissimilation without contact: thus in Dayak in the case of s : s, e.g. in tuso, "breast" < Original IN susu, tisa, "remainder" < Sanskrit śeša.

258. Dissimilation operates between WB and formative. In Sangirese the suffix -aṅ is replaced by -eṅ when the last syllable of the WB contains an a.

259. A special case of dissimilation is the one that goes so far as to cause one of the two sounds to disappear entirely. Whereas Dayak says tisa for sisa < śeša (§ 257), in Minangkabau the word takes the form iso, alongside of siso. This is a proceeding similar to the elision of the r in the dialectic Greek phatria, "brotherhood" < phratria (see Brugmann KvG, § 336).

260. Fracture* is a term used in IE research to denote various processes; I use the term for the change of a into ea. Bajo changes the a in final aṅ into ea, at the same time transferring the accent on to the e; hence Bajo padéaṅ, "grass" < Original IN .pádaṅ The cases are numerous.

* [In the original, " Brechung ".]



261. Each individual syllable has a summit. In IN this is almost always a vowel, and only quite exceptionally a voiced sound of some other kind. It is true that IN possesses words of form having no vowel, such as n, "of", m, "thy", but these appear almost invariably only after a vowel, with which they then combine to form a syllable. "Thy gain", in Toba, is labám < laba + m, but "thy house" is bagasmu.—An exception is formed by Gayo, where n, "of", can stand between two consonants, as in bět n se, "(after the) fashion of this". Here the nasal n is placed between two sounds that are deficient in sonority; it must therefore be a nasalis sonans, and is accordingly the summit of a syllable. The same is the case in Dayak phenomena like blióṅ-m, "thy chopper", where m for mu also comes after a consonant and is a nasalis sonans. An illustration in support of this is to be found in the Story of Sangumang, Bijdr. 1906, p. 201, 1. 10: "How many choppers hast thou?" = How + many are choppers thine = pirä aton blióṅ-m.—Seidenadel says in § 17 of his Bontok grammar: "Final l often becomes a sonant liquid, similar to l in our (English) word bottle"; but he gives no instance, and in all the texts I have found nothing to correspond with this assertion.

262. Certain phonetic processes may cause a shifting of the summit of a syllable. Most of the IN languages (as stated in § 4) accentuate the penultimate, thus áwak, "body", báyar, "to pay", and the like. The summit of the first syllable of each of these words is the a before the semi-vowel. Toba changes awak into aoak, Tawaelia turns bayar into baeari. It must not, however, be imagined that the o and e here merely fulfil the consonantal function of the w and y : they become full vowels, and hence take on the accent and become the summits of syllables; accordingly the results are the trisyllabic aóak' and the quadrisyllable baéari.

263. When a word consists of several syllables, there arises the question where the limits of the syllables lie. In Bontok "two intervocalic consonants are divided and distributed among two syllables", but "ds (and) ts are considered as one sound " (Seidenadel). According to § 60, ds and ts represent Original IN palatals. "In Achinese, when there is a combination of nasal with explosive or even of nasal + explosive + liquid, as in cintra, 'wheel', the first syllable ends with the vowel and the second begins with the combination" (Snouck Hurgronje). This rule must also hold good for some other IN languages; various phenomena point in that direction. In several languages, as in Nias (§ 188), a WB can begin with nasal + explosive ; in others, as in Modern Javanese (§ 69), such a combination does not make the preceding vowel short. And is perhaps also the bĕt n se of § 261 to be regarded as bĕt + nse ?

264. Variahilitij in the division into syllables also occurs. "In Madurese a hamzah between vowels may be pronounced as the end of the first syllable or the beginning of the second" (Kiliaan): accordingly poqon, "tree", is either poq-on or /poqon, or even poq-qon.

265. In Bontok we find a few cases where the limit between syllables is further marked by the shutting of the vocal chords, i.e. by hamzah; thus in the Headhunters' Ceremonies, Seidenadel-Texts, p. 512, 1. 3: totokqkoṅan, "to watch".



266. In the IN languages the WB's occur either unchanged or else combined with formatives, i.e. prefixes, infixes, or suffixes. In Ophuijsen's “ Bataksche Texten, Mandailingsch Dialect ”, p. 16, 1. 14: “ The story of the old ox ” = S. of ox which old = hobaran ni lombu na tobaṅ, the words lombu and tobaṅ are unchanged WB's, but hobaran consists of the verbal WB hobar, “ to tell a story ” and the suffix -an, which is used to form substantives.

267.   The addition of formatives may, or may not, involve phonetic changes. In the “ Pantun Mĕlayu ”, edited by Wilkinson and Winstedt, Pantun 4, 1. 1, we find: “ Whence flies the dove?” = W. d. f. = dari-mana punay mĕlayaṅ, and 5, 1. 1: “ How can one catch a porcupine ?” = How catch p. = bagay-mana mĕnaṅkap landaq. The WB's are layaṅ and taṅkap; in the derived form mĕlayaṅ we observe no phonetic change, but in mĕnaṅkap the t has turned into n.

268.   The phonetic phenomena occurring in connexion with the extension of the WB by means of formatives are either the same as those which we also observe in the interior of the WB itself, or they are different. Old Javanese contracts the Original IN sequence a + i in a WB into e; but it similarly contracts if this sequence should happen to occur in a derivative word. Original IN lain, “ other ” > Old Jav. len, but likewise ma + iṅĕt > meṅĕt, “ to take care ”. — Toba assimilates the r of a prefix to an immediately following l, and accordingly pronounces par + lanja as pallanja, “ carrier of burdens ”. But in a WB, such as torluk, “ bay ”, the r persists unchanged. 269. The phonetic phenomena which we observe in the various IN languages in connexion with the addition of suffixes, are chiefly and more particularly the following; and it is to be noted incidentally that the suffixes nearly all begin with a vowel:

I. Intermediary sounds appear; after i-sounds naturally y, after u-sounds w. Hence Bugis tunuwaṅ, “ to set on fire ” < tunu + aṅ from WB tunu, “ to burn ”. Or the intermediary sounds are h or q; thus in Southern Mandailing parkalahan, “ prophetic tables ” < par + WB kala + an; or in Madurese mateqe, “ to kill ” < pate, “ death ” + e.

Such intermediary sounds may interchange. After e in Makassar the intermediary sound is y, after o it is w; but in the “ Journal of the Princes of Gowa and Tello ” we always meet with Bontoya, “ the (country of) Bonto ” < Bonto + the article a, instead of the modern form Bontowa, e.g. on p. 8, 1. 15.

II. The final vowel of the WB becomes consonantal' before the suffix: thus in Old Javanese i > y,u > w, as in katunwan, “ to be burnt ” < ka + WB tunu + an.

III. The vowel of the WB and that of the formative are in many languages contracted together. Old Javanese WB kĕla or kla + the gerundial termination -ĕn results in klān, as in Kawi Oorkonden, I, 3, 20: “ Shall be cooked in (the) cauldron of Yania ” = klān i kawah san Yama.

IV. The consonantal final of the WB is doubled in Madurese and a few other languages. From the WB ator Mad. forms the verb ṅatorraghi, “ to offer ” < + ator + aghi.

V. In Gayo a final nasal of the WB changes into cognate media + nasal, e.g. in kuödnön, “ more to the right ” < WB kuön, “ right ” + ön. Mentaway inserts the tenuis instead of the media, as in mämäräpman, “ to want to sleep ”, from WB märäm. Illustration: Ghost Stories, in Morris' texts, p. 82, 1. 8: “ I want to sleep there ” = Sleep + want + to there I = mämäräpman lä aku

VI. In Bontok a media becomes a tenuis. From WB kaeb is formed the verb kapen, “ to make ”, from faeg, “ to whip ”, fayeken. VII. In Madurese a tenuis becomes a media. The WB for “ to suck ” in Old Javanese and Mad. is sĕpsĕp, and “ to suckle a child ” in Mad. is ñĕpsĕbbhi.

VIII. In the combination nasal + cognate media between the two vowels of the WB the media is lost in Maañanese. From WB endäy, “ to take ”, there is a derivative form enäyan.

270. Quantity in contraction.

I. In most cases vowel-length results; thus in Makassar, e.g. in kasaláṅ, “ compensation (for a wrong done) ”, which has a long a in the final syllable < ka + sala, “ to err ” + aṅ.

II. In other languages length of vowel does not result; thus in Toba, as in parhután, “ locality of a settlement ”, with short a' in the final syllable < par + huta, “ settlement ” + an; see also § 71.

271. The phonetic phenomena to be observed upon the addition of prefixes to the WB are less multifarious than in connexion with the addition of suffixes. They are the following:

I. Elision. In Bugis from WB onro, “ to dwell ”, are formed both paonro and ponro, “ to cause to dwell ”.

II. Contraction. In Old Javanese, from ma + WB iṅĕt is formed meṅĕt, “ to take care ”.

III. Appearance of intermediary sounds. Thus in Daïri pĕhuwap < + uwap, “ steam ”.

IV. Change of the explosive, with which the WB begins, into the cognate nasal, as in Malay mĕnaṅkap, “ to catch ”, from WB taṅkap, see § 16.

272. Upon the addition of formatives vowel-harmony may also supervene (see § 247).

273. We noticed in § 168 that the Original IN final diphthongs, as in punay, “ dove ”, patay, “ to kill ”, become reduced to simple vowels in several languages, as for instance in Bugis, which says pune. — But in connexion with this appearance of simple vowels instead of diphthongs we meet in several languages with phenomena that cannot be simply-explained as contraction, weakening, and the like. For the consideration of these cases we will use the following table as a basis:

Original IN: gaway, “ to make ” patay, “ to kill ” punay, “ dove ”.
Tagalog: gaway, “ to bewitch ” patáy punay.
Old Javanese: gaway pati
Later Old Jav.: gawe
Malay: mati punay.
Dayak: gawi patäy punäy.

Here we are struck by two sets of facts. Why do Malay and Dayak in some of these cases have the diphthong, as in punay and punäy, respectively, while in others they have a simple vowel, as in mati and gawi? Why does Original IN ay appear in the Old Jav. pati as i, while in gawe we observe that it is regularly contracted into e?

The answer to these questions is given by certain phenomena of the Philippine languages. In these a word ending in a diphthong appears in different forms according to whether it stands alone or has a suffix or enclitic attached to it. Original IN balay is also in Ibanag baláy, “ house ”, but “ their house ” is balé-ra. In Tagalog the word for “ to give ” is bigáy, but its passive is bigyán. Now I assume that a similar change used to take place in Original IN, so that one and the same word was e.g. sometimes pronounced gaway and sometimes gawi; Old Javanese, Malay, Dayak, etc., then compromised the matter, so that in some cases the form with the diphthong and in others the form with the vowel came to be used exclusively, and hence e.g. Malay mati side by side with punay.


274. Abbreviation of words occurs in very various kinds of cases; certain languages use it in words of every category, others in certain sorts of words, e.g. proper names. Most commonly a word is abbreviated at the beginning, less often at the end, and least frequently in its interior, as for example in the case of Napu au < Original IN anu, “such and such a one, that which”. Very rare indeed are such cases of irregular compression as the Karo ĕrbubai, “to announce a marriage formally” < ĕrdĕmu bayu. Abbreviation always occurs in isolated cases, here and there pretty commonly, but never in definite series determined by phonetic laws, apart from the haplological abbreviations in the doubhng of words (§ 234), which are, however, a special phenomenon. The full form may give rise to several short forms: thus of the above mentioned anu there are in Napu the two short forms au and u. All three forms figure side by side in the Napu text, “The Creation of the World” ; p. 393, 1. 6: “that which we see” = anu ta-ita; p. 394, 1. 11 : “that which is wild” = au maila; p. 394, 1. 11 : “that which lives” = u tuwo.

275. Abbreviation in WB’s, irrespective of the category to which they belong.

I. In Achinese, in consequence of the accentuation of the last syllable, the first syllable of many WB’s is dropped. In the Story of the Clever Blind Man, appended to Van Langen’s grammar, p. 109, 1. 12, we find in the second sentence two abbreviated WB’s next to one another: “To climb a coconut palm” = ik ur. Here ik - < Original IN naik and ur < niur2.

II. In Cham we meet with similar abbreviations, e.g. in lan, “month” ", “aphseresis from Original IN bulan” (Cabaton). Cham has a very large number of loan-words from neighbouring languages having monosyllabic WB's, and these loan-words have had their influence on a part of the disyllabic WB's of IN stock.

276. Abbreviation in certain categories or functions of words.

I. In exclamations. In various languages a disyllabic WB will lose a -syllable when it is used as an exclamation, thus imitating the interjections, which are very often monosyllabic. — In Tontemboan they say deq, “oh, horror !”, from indeq, “horror”. Thus we find in “The Burning of Kinilow”, Schwarz-Texts, p. 156, 1. 2 from the bottom: “Horror, oh horror for me !” = deq e deq aku.

II. In vocatives. In many languages abbreviations are used in the vocative. Vocatives, like exclamations, imitate interjections, and besides that, many languages accentuate the last syllable of a word when used in the vocative.[1] Such vocative abbreviations are found, in the first place, in words of relationship and friendship, as in the Madurese coṅ, “lad !” < kacoṅ, “the lad”. Thus in the Story of Kandhulok in Kiliaan's Texts, II, p. 153, 1. 9 from the bottom, we find: “Well, lad !” = kĕmma coṅ! Longer formations are reduced to disyllables or trisyllables, as in the Toba vocative maén from parumaen, “daughter-in-law” . — In the second place, they occur in personal names, as in the Modern Rĕmbang- Javanese vocative “Wir !” < Wiryadimejo. — In some languages such abbreviations are also used when the word is not employed vocatively, e.g. Tontemboan itow, “the little boy” < maṅalitow. — The Rottinese feo < feto, “sister”, is primarily a vocative but is also used in other ways.

III. In the imperative, which is very similar in its nature to the vocative, e.g. Toba botson, “give (it) here !” < boan tuson.

IV. In proper names, especially those of persons. In some languages, as for example in Dayak, personal names are formed from descriptive words by omitting the initial consonant. Such Dayak names are, e.g., Agap < tagap, “strong”, Adus < radus, “stout”, Ilak < kilak, “love”, Inaw < ginaw, “to shine” . These abbreviations imitate the words of relationship “father”, “mother” , and “child” , which in most of the IN languages begin with a vowel, the Original IN types being, of course, ama, ina, and anak.

V. In technical terms. Here we often meet with very drastic abbreviations, just as in similar cases in IE : of. English pops < popular concerts (Brugmann KvG, § 366, 5). Thus the Bugis ida-ida, which denotes a certain poison, is an abbreviation of “quickly- working poison” = p. q.-w. = racuṅ maqpacidacida.

VI. In compounds. Here, in the first place, the abbreviation may occur at the point of junction of the two words. The first member of the compound loses its final vowel, more rarely a consonant and in that case mostly the aspirate h. Examples: Ampana torarue, “water-spirit” < spirit + water = torara + ue; Minangkabau tigari, “a festival” < tiga, “three” + hari, “day” . Or else the first member of the compound, which usually has the weaker stress, may be abbreviated; as in Busang bĕtaóq, “right side” < beh, “side” + taóq, “right”.

Particular notice is due to certain Bugis abbreviations wherein the first member of the compound loses a final ṅ, as in po-lila, “back part of the tongue” < pon, “stem” + lila; similarly po-lima, “back part of the hand” . The abbreviation is explicable by the fact that a sequence like + l is not permissible in the interior of a WB ; but why should there be abbreviations like po-kanuku, “back part of the nail” seeing that the sequence + k occurs very often in WB's ? The reason is that these are analogical formations; po-kanuku has imitated po-lima.

VII. In groups of words that denote a single idea and therefore approximate to compounds. In the first place, such groups as have for their first element a title. In the story Ja Bayur, in Ophuijsen, “Bataksche Texten, Mandailingsch Dialect”, p. 74, 1. 4 from the bottom, we find: “His name became Ja Bayur” = N. h. the b. J. B. = gorar nia i manjadi Ja Bayur, where Ja is an abbreviation for raja, “prince”. — Other abbreviations falng under the present category are, e.g., Sawunese dupamu, “wife” < “person in (the) house” = dou pa ĕmu; Napu anaṅkoi, “little child” < ana, “child” + anu, “which” + koi, “little” . — We particularly often meet with such cases of compression in words of form, as in Dayak ranen, “and so on” < ara, “name” + enen, “whatever”.

VIII. In numerals. In Javanese, in counting (according to Poensen), or in counting rapidly (according to J. N. Smith), the disyllabic digits 1-10 are usually docked of the first syllable, e.g. people say tu for pitu, “seven” . Here the abbreviated forms imitate the forms which are really monosyllabic, like pat, “four”.

IX. In auxiliary verbs. In verbs which are usually followed by another, dependent, verb that contains the leading idea, abbreviation may occur in several languages, the medial sounds of the word being reduced. Thus, for example, in Karo, dapĕt, “to be able” , is abbreviated to dat. In Minangkabau the full form of the word which would correspond to the Malay, etc., pĕrgi, “to go”, no longer occurs, but only the short form pai or pi. But an analysis of the Manjau Ari shows that pai or pi mostly occurs only in the above-mentioned kind of context; thus p. 8, 1. 1: “We go to fetch (him)” = written: kita pi japut = spoken: kito pi japuyq.

X. In enclitics and proclitics: see § 302.

XI. In euphemisms: see § 18.

XII. loan-words: e.g. Modern Javanese deler < Dutch edele heer. Here the tendency towards disyllabism very often asserts itself.

XIII. In colloquial language: see § 20.

XIV. In poetry: see § 27.

277. Abbreviation, particularly in the case of compounds, may go so far that the significative nucleus is lost altogether. This is especially the case with compound negatives in several languages. The Tontemboan negative raqi is often strengthened by the particle ka, thus forming raqica (in accordance with the rule in § 103, II), which again by abbreviation becomes ca. As appears from an examination of all the texts, this ca is especially found in dialogues.

278. The full form and the short form may serve side by side in the language. In Cham " a certain " is holey, or, abbreviated, ley. Now, in the story Mu Gajaung, p. 22, 1. 29, we find: " On a certain day " = harey holey; but in 1. 11: harey ley.

279. In abbreviations, phonetic phenomena may occur which are not otherwise possible in the language in question. Rottinese has a word bindae, " a sort of vessel " ᐸ bina " shell " + dae, " earth "; but in the interior of Rot. WB's the sequence n + d does not occur.

280. A considerable proportion of the IN abbreviations of words have parallels in IE. In IE, as in IN, the species of abbreviation whereby sounds in the interior of words are eliminated, is the rarest: see Brugmann KvG, § 366, 5. — Elimination of the final vowel of the first part of a compound is found in Gothic, e.g. in hauhhairts, " proud ", as compared with armahairts, " merciful " : see Wilmann, " Deutsche Grammatik ", the section entitled " Der Vokal in der Kompositions- fuge". Abbreviations of titles are found, e.g. Middle High German ver ᐸ vrouve, Italian wadonna. An instance of abbreviation in exclamations is the Swiss-German mäntSacrament. A case of the abbreviation of the negative going so far as to deprive it of its significative nucleus is the Swiss-German at, “nothing”.


281.   When a loan-word is taken up into an IN language, its sounds must accommodate themselves to the phonetic capacities of the language that accepts it.   Exceptions are rare and are found mainly among educated persons; but here and there a foreign sound has persisted even in popular pronunciation. Madurese has no f, but loan-words containing that sound preserve it even in popular pronunciation “ pretty generally ” (Kiliaan).

282.   The change of sounds takes place in certain cases because the recipient language absolutely does not possess the sound that occurs in the loan-word.

I.   Loan-words from IN languages. The commonest case is that of the palatals, which are wanting in certain IN languages; they are replaced by a velar, a dental + i, or by the semi-vowel y.   Table:

Malay jambatan > Napu gambata, “ bridge ”.
Malay janji         > Sangirese diandi, “ to promise ”.
Malay jaga         > Tontemboan yaga, “ watch, guard ”.

II.   Loan-words from non-IN languages. The commonest case is that of the various sibilants, as most of the IN languages possess only one, viz. s.   Thus the Dutch sjaal, pronounced šāl, " shawl ", appears in Madurese as sal or cal.

283.   The sound may occur in the loan-word in a position which it is not allowed to occupy in the recipient language. In Busang no word ends in s, Original IN r2atus, “ hundred ”, becomes atu, and hence Bugis > Bugit and English > Iṅgĕlit. The s has only persisted in kĕrtas, “ paper ”. — A particularly common case is that an IN language having only vocalic finals adds a vowel to loan-words that end in a con

sonant.   In the Tsimihety Poem on the Telegraph, p. 116, we find telegrafi, Parisi, and Madagasikara.

284.   The phonetic combination may be alien to the recipient language. Here it is mostly a case of combination of consonants. The linguistic methods then applied by the IN languages are the following:

I.   Elimination. In Jonker's Rottinese Texts, p. 44, 1. 1, we find: “ Service letter ” = L. s. = susula dis; dis being from the Dutch dienst.

II.   Metathesis, as in the Old Malagasy Serafelo, the name of a certain angel < Arabic Asrafil. This form is found in Ferrand's text Niontsy, p. 24, 1. 1 from the bottom: “ Where art thou, O Asrafil?” = aiza hanaw ra Serafelo?

III.   Insertion of a sound, as in the Bugis porogolo < Dutch verguld, “ gilt ”.

285.   The selection of the inserted vowel is determined:

I.   By the nature of the neighbouring vowel, as in Makassar parasero < Portuguese parceiro, “ partner ”.

II.   By the nature of the neighbouring consonant. Between the s and the χ of a Dutch initial sch Makassar inserts an i, as in sikau < schout, “ mayor ”.

286.   Special consideration is due to the developments of the phonetic combinations of explosives + h, i.e. the aspirates of loan-words in languages which themselves have no aspirates.

I.   The aspiration disappears altogether, as in Malay bumi < Sanskrit bhūmi, “ earth ”.

II.   A vowel appears between the explosive and the aspiration: Makassar pahala, “ utility ” < Sanskrit phala.   Madurese, in accordance with the principle in § 184, II, has paqalah. Daïri dĕhupa < Sanskrit dhūpa, “ incense ”.

III.   Owing to some secondary process the aspiration disappears, but the inserted vowel persists; hence Toba daupa and budá < budaha < buddha.

287.   The phonetic phenomena hitherto delineated are either sporadic or else form regular series.   Of the latter kind

is the Bugis rendering of the Dutch initial sch-; the dictionary contains half a dozen cases, and in all of them sch- is rendered by sik-, e.g., sikemboro < Dutch schenkhord, “ tray ”. The Hova dictionary has eight loan-words which in their original languages began with br. In five cases br- becomes bur-, as in burákitra < English bracket. In the three other cases the inserted vowel is determined by the neighbouring one, as in biriki < English brick.

288.   Loan-words may either, submit to the laws of phonetic change governing the several IN languages or they may struggle against them. In Saqdanese w is omitted in loan-words as it is in native words. That appears from the text “ Tunaq Pano Bulaan ”, where (inter alia) on p. 225, 1. 6 from the bottom, we find saa, “ snake ” < Original IN sawa, and on p. 228, 1. 8, deata, “ God ” < Sanskrit dewatā. — In Minangkabau an Original IN final at becomes eq, but loan-words preserve the pronunciation at unchanged even in colloquial, hence Mkb. adat, “ customary law ”.

289.   In connexion with the reception of loan-words the forces of analogy and popular etymology are particularly operative. The word for “ veil ” in Bugis is bowoṅ, or in its contracted form bōṅ, and in imitation of it the Dutch bom, “ bomb ”, appears not only as bōṅ but also as bowoṅ. In Hova it chances that no words begin with l + a + b, but several with l + a + m + b, hence the French la bride appears in Hova as lamburidi. In the Old Sundanese legend Purnawijaya, verse 154, the hound of hell is called Sirabala; that is a deformation of the Sanskrit śabala made under the influence of the article si, which in Sund. is used with names of animals.

290.   In the reception of loan-words IE displays much the same sort of phenomena as IN. To mention only a single case, we observe in Italian as in Makassar the insertion of vowels into awkward consonantal combinations, hence Ital. lanzichenecco < German Lanzenknecht, “ spearman ”, like Mak. parasero < parceiro (§ 285).



291. In the interior of the sentence we may either meet with the same phonetic phenomena as in the interior of words, or with different ones.
I. In the standard dialect of Tontemboan a k after an i changes into a c, both in the sentence and in the individual word. Hence in the story told by S. Pandey, Schwarz-Texts, pp. 12 seqq., we not only get on p. 13, l. 25, lalic < lalik, "to go to law (about something or other)", but also in l. 23 si cayu < si kayu, "the tree".
II. In the Kawangkoqan dialect the change of k into c takes place only within the word, not in the sentence. Hence in the story told by A. W. Rompas in the Kawangkoqan dialect, p. 156, 1. 5, we read pasicolaan, "school-house", from WB sicola < sikola, but on p. 155, 1. 11, we find si kayu, "the tree".
292. A sentence may either be a perfect unit, or may contain within it certain parts which combine into a more closely connected group. Such groups may either be knit together more intimately by the sense, thus to the Hnguistic consciousness of the people of Nias the combination of "principal word + subjective genitive" is more intimate than that of "principal word + objective genitive". Or the closer relation between certain parts of a sentence may be constituted by the fact that they are subordinated to a single accentor stress). This is the case with the group of "proclitic or enclitic + word of substance". — Now in these groups of more intimate relation phonetic phenomena may occur which other-wise do not appear in the body of the sentence (see § 302).
293. The phonetic phenomena that appear in the sentence, as such, are especially the following: assimilation, metathesis appearance of intermediary sounds, doubhng of final consonants, turning of vowels into consonants, contraction, loss of vowels, and loss of consonants. These phenomena are to a great extent similar to those that have been noticed in connexion with the combination of the WB with formatives (§§ 266 seqq.).
294. Assimilation, in many languages, e.g. in Toba. In the story Nan-Jomba-Ilik, Tuuk Lb, p. 1, 1. 4 from the bottom, we find written: “Why comest thou?” = di-bahen ro hamú, but it is pronounced di-baher ro hamú.
295. Metathesis, in Kupangese. According to the text Bihata Mesa, Bijdr. 1904, metathesis occurs in certain cases in the second syllable of a WB when used in a sentence. Original IN aku, “I” , appears also in Kup. as aku, and laku, “to go”, as lako; hence on p. 253, l. 1, we find: “Then (he) went and reported (it)” = Then w., then r. = ti lako, ti tek. But on p. 253, 1. 2, we find: “I went to hang him up” = I w. hang + up = auk laok tai.
298. Appearance of vowels or consonants as intermediary sounds. In the Tontemboan story “Kariso and his Children”, Schwarz-Texts, p. 129, 1. 8 from the bottom, we find: “A relation of his” " = ěsa taranak-ě-na. The intercalated pĕpĕt is the intermediary sound; na = “of him”. Hain-Teny, p. 186, verse 5, has: “To be able to keep back the stream” = nahatan-d-riaka. Here Hova employs the consonant d as an intermediary sound between nahatan (a) and riaka.
297. Doubling official consonants, in Ibanag. “I am big” = B. I = dakall ak < dakal and ak.
298. Change of vowels into consonants , in several languages, as in Old Javanese, Timorese, etc. Old Javanese, from Mpu Tanakung's Prosody, str. 41, v. 1: “A bird likewise” = pakšy adulur < pakši and adulur.
299. Contraction, in Old Javanese and other languages. Rāmāyana, II, str. 43, v. 1 : “His big bow” = Bow his big = laras nirāgöṅ < nira and agöṅ.
300. Loss of vowels.
I. ATien tlie vowel ends the word and the next one begins with a vowel, e.g. in Hova. Hain-Teny, p, 136, v. 6: “To be disquieted” = Have disquiet = manan eritreritra < manana eritreritra.
II. When the vowel ends the word and the next one begins with a consonant, in Kupangese. From the story Bihata Mesa, Bijdr. 1904, p. 257, 1. 3: “(They) sit together” = dad buan < dada and buan.
III. lTien the vowel is closed by a consonant, in Timorese. From the story Atonjes, Bijdr. 1904, p. 271, 1. 17: “This mother” = M. t. = ainf i < ainaf and i.
301. Loss of consonants, in Kamberese. From the Story of the Top, Bijdr. 1913, p. 82, 1. 7: “Pasture (for) horses” = pada njara < padaṅ and njara.
302. Special phenomena of the groups of intimate relation mentioned in § 292.
I. In Old Javanese certain pronouns when in a proclitic position may lose a final vowel, even before a word that begins with a consonant. Thus Eamayana, XXII, str. 17, v. 1: “Then shall I recognize thy love” = ṅke k tona asih ta. The k is an abbreviation of ku, the proclitic pronoun of the first person, which appears in that form and with that function in many IN languages; tona is the future of ton, “to see”. Apart from these cases Old Javanese does not employ elision but only contraction or the change of a vowel into a consonant.
II. In Nias after a final vowel in certain groups of intimate relation the voiceless initial consonant of the next word is changed into a voiced (or sonant) one; thus in the combination “principal word + subjective genitive” , or the combination “preposition + principal word” . The word for “heart” in Nias is to2do2, but in the story Siwa Ndrofa, Bijdr. 1905, p. 34, 1. 7, we find: “In (the) heart” " = ba do2do2.
303. The phonetic phenomena of the sentence sometimes take place with strict regularity, sometimes less regularly.
I. Voicing in Nias takes place with strict regularity.
II. Elision in Hova in the group “predicate + object” is left to the discretion of the speaker, at least so it appears from Hain-Teny. On p. 188, v. 2, we find: “To want to swallow stones” = W. to sw. st. = hitelim batu < hitelin (a) watu, but p. 80, V. 2: “To smell (of) lemons” = manitra wuasari. Elision of the vowel would produce mani buasari.
304. Interjections often dechne to conform to the laws affecting the sentence. In Toba the final a of a word invariably disappears before an initial a of the next one, as in the Riddle Stories, III, Tuuk Lb, I, p. 50, 1. 1 from the bottom: “If it is not permitted” = molo soada adoṅ, which is pronounced molo soad adoṅ. But if the word with final a is an interjection, the a persists, as in Riddle Stories, I, Tuuk Lb, I, p. 49, 1, 11 : " No, father " = indadoṅ ha amáṅ.
305. Within a group of intimate relation the O'peration of the phonetic laws of change is often suspended.
I. In words of substance. In Makassar, final k- changes into q, as in Mak. anaq < Original IN anak, “child”, but before the article added enclitically this change does not take place, e.g. in anak-a, “the child”.
II. In proclitics and enclitics. In Minangkabau, final a changes into o, as in niato, “eye” , <C Original IN mata; but proclitic words hke the j)reposition ka keep their a unchanged.
306. Finally, here are some parallels between IN and IE :
I. Assimilation in Greek and in Toba. Greek dialect in Thumb, “Handbuch der griechischen Dialekte” , § 203: tōl Labyadān < tōn Labyadān. Toba sal lappis, “one layer” < san lampis.
II. Voicing in Sardinian and in Nias. Sardinian tempus, “time”, but su dempus, “the time”. Nias to2do2, “heart” but ba do2do2, “in (the) heart”.
III. Loss of vowel. The word of substance loses a vowel before an enchtic: Rumanian and Balinese. Rumanian: casa, “the house” < casā and the article a. Balinese, from the texts appended to Eck's grammar, p. 62, 1. 2 : “To be contained in the letter” = muṅgw iṅ surat < muṅguh iṅ surat. — The enclitic loses a vowel: Latin and Karo. Latin viden , “seest thou ?” ᐸ vides + ne . Karo, from the story Dunda Katekutan, p. 34, 1. 17: " (It is) done " = ĕňgom ᐸ ĕňgo + the emphatic particle .

IV. Resistance of inferjections to the phonetic laws affecting the sentence: Sanskrit and Toba. Sanskrit, in Wackernagel, “Altindische Grammatik” ", I, § 270. Toba: ba amáṅ(see I 304).


In General.

307. Accent in the IN languages is either determinate or free. It is determinate, when its place in the word is determined by definite rules; it is free, when such rules are wanting.

308. The positition of the accent in the WB is either on the penultimate or the final syllable. Other modes of accentuation are rarer phenomena.

Accentuation of the Word-Base.

309. There are four systems of accentuation applicable to the IN word-base:

I. All WB's accentuate the penultimate. This is the penultimate type.

II. All WB's accentuate the final. This is the final type.

III. The WB's accentuate either the penultimate or the final, according to definite rules. This is the Toba type.

IV. The WB's accentuate either the penultimate or the final, but without definite rules. This is the Philippine type.

310. The penultimate type is the most widely distributed one. — Moreover in the Toba type also, and in many repre- sentatives of the Philippine type, accentuation of the penultimate syllable preponderates. This appears clearly from an examination of accentuated texts, e.g. the Mandailing texts in Van der Tuuk's Toba grammar, p. 31 (which exempHfy the Toba type) or the text Lumawig in Seidenadel-Texts, pp. 485 seqq. (Philippine type). Hence by far the greater number of the IN WB's have the accent on the penultimate 311. The languages which accentuate the penultimate, however, admit exceptions to the general rule.

I. In several languages of the penultimate type the pěpět cannot take the accent. Therefore if the penultimate contains a pěpět, the accent falls on the final syllable, as in Gayo sělúk, “tortuous”. If both syllables have a pěpět, some of the languages accentuate the penultimate, others the final.

II. Some languages of the penultimate type have a small number of words of substance that are accentuated on the final ; thus Mentaway, e.g. , arát, “to go in”. These are mostly words for which no cognates are to be found in the other IN languages.

III. Several languages of the penultimate type possess a few words of form, especially demonstratives, that are accentuated on the final. Examples: Mentaway, otó , “so” , Bugis manrá, “yonder”, Hova iti, “this”. Mas accentuates most of its demonstratives on the final syllable.

In various languages of the penultimate type we find words of form that are accentuated either on the penultimate or the final, but with variations in meaning; e.g. Sangirese táṅu , “on that account”, táṅu , “thereupon”.

IV. In interjections too we not infrequently meet with accentuation of the final, as in Bugis aui , “indeed !” (implying surprise).

312. The final type comprises but few languages. It includes, for instance, Busang, which accordingly pronounces anál , for “child”.

313. The Toba type comprises Toba and cognate languages, such as Mandailing. Here too the accent mostly falls on the penultimate. But in certain definite cases, determined by rules which are given in the grammars, accentuation of the final occurs. One such rule is : Verbal WB's denoting a condition that has been caused (by some external agency), accentuate the final ; hence the accentuation of tanóm, “to be huried”, as against húndul, “to sit”.
314. In the languages of the Philippine type some WB's accentuate the penultimate and others the final, without there being any rules on the subject. We can form no idea why Bontok says piló, “seven”, but wálo, “eight”, the more so as there is no certain etymological explanation of these words.
315. Unusual modes of accentuation: accentuation of the antepenultimate results from the addition of a supporting vowel in all languages that add it. Hence Hova ánaka, “child” < Original IN anak, Makassar nipisiq, “thin” < Original IN nipis.
Equal accentuation of both syllables of the WB is found in some languages in onomatopoeic formations, as in Toba búmbám, “to beat”.

Accentuation of Derivatives from the Word-Base.

316. When a disyllabic — or polysyllabic — WB is extended by means of prefixes, the accentuation is not affected thereby; Bugis pésĕq, “to feel”, and papésĕq, “feeling”, are accentuated alike.
317. When suffixes are added, we observe the following phenomena :
I. In languages of the penultimate type the accent is shifted, so that it always falls again upon the penultimate. From the Bugis tiwiq, “to bring” < Primitive Bug. tiwir, are derived: tiwiri, “to bring to somebody” , and patiwiriyaṅ, “to give something to somebody to take with him and bring it”. Only a few languages of the penultimate type fail to shift the accent, e.g. Gayo, which accordingly accentuates kĕbáyakan, “riches” < bdyak, “rich”.
II. The other types also shift the accent, hence Toba isian, “vessel” < isi, “contents”. But alongside of this they also possess suffixes which attract the accent to themselves. In Toba the suffix -an of the comparative takes the accent, thus: biroṅán, “blacker”, from biroṅ, “black”, as. against the above-cited isian.
III. When contraction results upon the addition of suf fixes, accentuation of the final syllable is also produced, as in Toba haduwán, “the day after to-morrow” " < formative ha + duwa, “two” + formative an. — If the feeling, that the word is a derivative, becomes lost, the accent may be shifted back again; hence Mandailing hadúwan, “the day after to-morrow” .
318. When suffixes are added to monosyllabic WB's —which in all the IN languages amount to only a very small percentage of the vocabulary — there is nothing new to be remarked as to the accent. From the Bugis noq, “down- wards” << nor < sor (§§ 40, 1, 150, III), is derived: nóri, “to bring down” , which gives rise to no observations.
319. When prefixes are added to monosyllabic WB's, the general rule is that the accent does not shift away from the WB, e.g. Bugis panóq, “to let down” < noq, “downwards” . Here, therefore, even the languages of the penultimate type have the accent on the final. — But if the feehng of derivation becomes obscured, the accent may shift back. Bungku has opá, “four”, from ěpat = prothetic ě (§ 226) + Original IN pat, but Nias has o2fa, with o < ě in conformity with § 227.

Accentuation of Doubled Words and Compounds.

320. When a word is doubled, the first element preserves its accent in some of the languages, but loses it in others. In Dayak both alternatives occur side by side, with differentiation in meaning: gila-gila, “all stupid” , gila-gila, “somewhat stupid”.
321. Here too the Toba type has all sorts of peculiarities, e.g., jalák-jalák, “to seek everywhere” , alongside of manjálak:, “to seek”, from the WB jálak.
322. In Bugis a certain number of words that have a long and accentuated final syllable, such as apĕllán, “cooking utensils”, atinrón, “sleeping apartment”, arúṅ, “king”, shift back the accent, and thereby also lose the length of the final syllable, whenever they serve as the first element of a compound; e.g., árum-póne, "king of Bone" < aruṅ and Bone. In a Bug. sentence it hardly ever happens that two accentuated syllables follow one another, for almost every word is accompanied by enclitics ; hence in a compound an accentuation like árum-póne makes a disagreeable impression, and is therefore altered. As to mp < mb < + b, see § 117.

Accentuation of the " Complex ": i.e. Word of Substance + Word of Weak Stress.

323. The complex (or conglomerate) may consist of a word of substance + a monosyllabic enclitic. In that case we sometimes find the accent shifted, and sometimes not, in accordance with definite rules:
I. In Makassar, for example, when the article a is affixed, the accent shifts if the principal word ends in a vowel, but not if it ends in a consonant: hence úlu, "head", ulúw-a, "the head", járaṅ, " horse ", járaṅ-a, "the horse".
II. If the enclitic loses its vowel, that does not prevent the shifting: Bimanese aná-t, "our child" < aná + ta.
III. The Toba particle tu, " too", attracts the accent to itself: madae-tu, "too bad" < madde + tu. This is in imitation of the accentuation of the comparative (§ 317, II).
324. Again, when disyllabic or several ynonosyllabic enclitics are added, shifting of the accent may result, or it may not, or the complex may even have two accents. An instance with two accents is found in Paupau Rikadong, p. 19, 1. 4 from the bottom, in Matthes' Bugis grammar: "They also reported it" = T. r. a. it = na-lěttúri-tó-n-i . Here n < na is an emphatic particle, homonymous with na, "they".
325. There is little to remark in connexion with the addition of proclitics. When a monosyllabic proclitic combines with a monosyllabic WB, some of the languages accentuate the WB, others the proclitic. Toba says si-gák, "the crow" < the article si + gak, Sundanese, on the other hand, si-pus, "the cat".

Accentuation of Loan-words.

326. Loan-woids mostly acccmmcdate themselves to the native laws of accentuation; thus the Dutch gezaghebber, “ruler”, becomes sahébar in Dayak. Exceptions are rare, e.g. the Bugis sikelewá < Dutch schiláwacht. This cannot have been a case of imitation, for Bug. native words never end in an accentuated a.

Quality of the Accent.

327. All our previous researches have been concerned with the position of the accent. Let us now enquire as to its quality. In IN the accentuated syllable may differ from the unaccentuated ones thus: by greater loudness, by a higher pitch, or by increased length.
328. Select descriptions of the quality of the IN accent: “Accent in the IN languages is of a different kind from what it is in the IE. In Dutch, and particularly also in English, the principally accentuated syllable is pronounced loudly, the other syllables softly. That is not the case in the IN languages. There the unaccentuated syllables receive fairer treatment, but in consequence the accent is of course less distinct. In some languages the accent is nothing more than a lengthening or extension of the accentuated syllable. But Sangirese has not gone to such lengths as that ; its accent is distinctly audible” (Adriani). — “In the IE languages accent is stress, but in many IN languages it is a rise in pitch of the voice. It is true that this rise in pitch is accompanied by an increase in loudness, but that does not cause the un- accentuated syllables to be pronounced in a more cursory manner. As in Tontemboan the accent is produced by a rise in pitch and the unaccentuated syllables are all distinctly and perfectly pronounced, the Tontb. accent gives one the impression of being weak. It is, however, distinctly audible that it falls upon the penultimate” (Adriani). — “In Rottinese the accent is distinctly audible and falls upon the penultimate” (Jonker). — “In Minangkabau all the syllables have the same loudness but the penultimate one sounds somewhat longer or more extended and thus has the accent” (Van der Toorn). — “When pronounced alone, in fact simply mentioned, all Achinese words are sounded so that both syllables have an equal stress, but the second syllable is pronounced in the higher tone” (Snouck Hurgronje).

The Unaccentuated Syllables.

329. It appears from § 328, that so far as the loudness of the tone is concerned the unaccentuated syllables do not differ very considerably from the accentuated. At the same time the syllable preceding the accentuated one is somewhat weaker than the one following the accentuated syllable. On this fact are based all sorts of phenomena that we have noticed in the preceding parts of this monograph, e.g. that the syllable after the accentuated one is pronounced long in several languages, and that it is capable of becoming a diphthong. On the other hand, length is of very rare occurrence in syllables preceding the accentuated one, and diphthongization is still rarer. Ampana sounds the syllables before the accentuated one so softly, “that it is only when a person speaks slowly, that one can hear what vowel they have” (Adriani). In several languages a syllable preceding the accentuated one may lose its vowel: in Dayak they say blaku as well as balaku, “to ask”. Loss of a vowel following after the accentuated syllable is very rare; it is found in Makianese, which has lim, “five” < Original IN lima.

Original Indonesian Accentuation.

330. In former monographs I assumed that the determinate system of accentuation, and in particular the penultimate type, was the modern representative of the Original IN law of accentuation. Since then, doubts have arisen in my mind. In IE there are languages with determinate, and others with free, accentuation, and Original IE is credited with the free system, the IE languages with the determinate accent being taken to represent a secondary development. Might not something of the kind be possible in IN also? In that case the free Philippine type would be the primitive original, and not the determinate penultimate type. This supposition arose in my mind when I observed that in the languages of the penultimate type, there occur, though very sporadically, cases of accentuation which are at variance with the law of penultimate accentuation and coincide with corresponding Philippine cases. The Philippine languages often accentuate personal pronouns on the final syllable, and Nias (which in other respects follows the penultimate type) also has ami, “ you ”.

Comparison with the Accent of the Indo-European

331. In this sphere also a large number of parallels between the two families of language may be found. For instance, Latin and Makassar have quite similar systems of accentuation:

I. Principal rule. The accent falls either on the penultimate or on the antepenultimate, as in Lat. cadáver, “ corpse ”, Mak. kandáwo, “ hollow ”, Lat. cádere, “ to fall ”, Mak. káttereq, “ to cut ”.

II. Subsidiary rule. In a minority of cases the accent is on the final syllable, viz. as a result of contraction, as in the Lat. perfect audit < audivit (Sommer, “ Handbuch der lateinischen Laut- und Formenlehre ” , § 71, I, e), Mak. kodi with a long i, “ to make bad ” < WB kodi + the suffix i.

332. It is true that the principal rule in Latin has a different linguistic basis from the one in Makassar. In Latin the quantity of the penultimate is the determining factor, while in Makassar the question turns upon the origin of the final, viz. whether it is an original syllable or merely a syllable added as a support.

Sentence Stress.

333. Under this head we must consider the relative accentuation of the several parts of the sentence, and especially the phenomena connected with the accentuation of the end of the sentence, for these are of great importance as characteristics of the IN languages.
334. Relative accentuation of the several parts of the sentence. " Toba only has accentuation of a syllable of a word. It does not employ word-stress, which we use in order to throw emphasis upon a particular word in the sentence " (Van der Tuuk). — " Busang accentuates the final syllable of the last word of the sentence ; but one can also accentuate any word in the sentence, if it contains a leading idea " (Barth). — " Accent in Javanese consists merely in this, that the last two syllables of each subdivision of a sentence are pronounced somewhat long and slowly, but both in an equally high tone. All the other syllables of a subdivision of a sentence are pronounced in a similar tone. If it is desired to throw special emphasis upon a word, it is given a position just before a break in the sentence, so that its last two syllables are as a matter of course pronounced more slowly, with the accent as defined above " (Roorda). — " In the Achinese entence it is not the several words that are the units for the purpose of accentuation, but rather groups of two or three words, linked together as one whole. In the phrase, ' a new-born child '= c. n. b. = anöq baro na, the na deprives the other two words altogether of any distinctive accent; they become, if one lies to put it that way, unaccentuated " (Snouck Hurgronje).
335. Relative accentuation of the end of the sentence. "The end of a Sundanese sentence is always pronounced long and in a singing (zangerig) tone, and the penultimate syllable of the sentence mostly receives a special emphasis " (Coolsma). — " The pronunciation of the Mantangay dialect of Dayak is much like that of the Pulopetak dialect, only the last word of each sentence is pronounced longer and louder " (Hardeland). — “In Minangkabau the last word of a sentence, or its final syllable, bears the principal accent; thus they say, with a stronger intonation : ' He sleeps ' = inyo lalóq ” (Van der Toorn). “In Bada the last syllable of a sentence is spoken with a rising (opgang) of the voice, i.e. with a rising accent” (Adriani). — “As regards rise and fall of tone, or the musical

accent, Malay pronounces the phrase: ‘Is that a stone ?’ = That stone = ini batu, in a rising tone, but : ‘ That is a stone ’ = ini batu, in a falhng tone” " (Van Ophuijsen).

336. The interrogative sentence. In the Bontok interrogative sentence the intonation rises and reaches “its highest tone at the final vowel of the sentence” (Seidenadel). — “The assertive and the interrogative sentence in Dayak may be illustrated by the following examples. Assertive : ' He is sick ' = iä hábăn. Interrogative: ‘Is he sick?’ = id häbán, in an interrogative tone which somewhat accentuates, and makes half long, even the last syllable of haban” (Hardeland).
337. In many IN languages the vocative, whether standing by itself or forming part of a sentence, throws the accent on to the last syllable of the word or group of words. Hence in many languages: iná, “O mother”, the vocative of ina, “mother”. Karo, from the story Kaja Kĕtĕnahĕn, in Joustra, “Karo-Bataksche Vertellingen” , p. 92, 1. 19: “Weep not, father !” = Not thou w., f. = ola ham taṅis bapá. Ibid., p. 91, 1. 18: “Let us go home to eat, my prince” = Eat we to house, prince mine = mati hita ku rumah, raja-ṅkú. — This fashion of accentuating the vocative must be regarded as Original IN.


338. The word lagu in IN signifies “modulation of the voice, melody, tempo, and style, in speaking or reciting” .
339. We may distinguish between three kinds of lagu, viz. those characteristic of particular languages, particular individuals, and particular circumstances, or the emotions arising therefrom, respectively. Of the second sort there is nothing to be said here.
I. The lagu of particular languages. “The Sundanese are in the habit of speaking slowly and quietly, in a peculiar tone, lagu, which sounds singing and prolonged” " (Coolsma). — “The Achinese speak rapidly” (Snouck Hurgronje). — “The Puqu-m-Boto dialect is spoken in a tone that sounds more cheerful and is more prolonged, than the average Bareqe. The tone of the To-Lage dialect sounds somewhat proud and mocking, even in the mouths of slaves and children” (Adriani).
II. The lagu of particular circumstances. Here the excessive lengthening of vowels and even of consonants is a phenomenon of particularly frequent occurrence. “In Minangkabau, if one wants to express pity for the person addressed, one says : tuaaan, ‘lord !’ ; if a man sees a runaway horse, he shouts out: kudooo!” (Van der Toorn). — “In Madurese, they say kab . . . bhi, ' all !', instead of kabbhi, if they want to express astonishment” .
340. From the tem fo there result certain phonetic phenomena, viz. the lento an d alltgro forms. In Dayak the article i coalesces with the pronoun aku to form yaku, “I” . When speaking slowly the Dayak says iaku. This iaku is the lento form, and at the same time the exceptional one, the normal form being yaku. According to Ophuijsen the Malay duwabĕlas, "twelve", when spoken rapidly, is sounded dobĕlas. This dobĕlas is the allegro form, and likewise the exceptional one. In Achinese bah + le, " let be !", has become bale, but when people are speaking quite slowly the h reappears,

bahle being therefore the lentissimo form.

341. On the differentiation into lento and allegro forms depend such double forms in Latin as nihil and nil (Sommer, "Handbuch der lateinischen Laut- und Formenlehre ", § 80); the case of the Karo negative lahaṅ, beside lan, is exactly similar.


342. When one studies certain descriptions of IN phonetic conditions, it appears not infrequently as if the IN languages were less consistent in their phonetic phenomena than the IE ones. But the trouble is not always in the language, it may be due to the writer:

I. Something may be given as a striking instance of a phonetic phenomenon, though in fact it is not a phonetic phenomenon at all. On the assumption that the phrase “come here !” is mari in Malay and some other languages, but mai in Bugis, it has been asserted that in Bugis the r has disappeared. This would be the only case of the loss of r in Bugis. However, mari < ma + ri is a verbal derivative from the locative preposition ri, while mai is a derivative from the locative preposition i, and does not mean "come here !" but "to go yonder". So the Bug. mai is not a case of phonetic change at all.

II. False etymologies are propounded. Thus in the Old Javanese dictionary (Kawi-Balineesch-Nederlandsch Woordenboek), vol. IV, p. 226, the Old Jav. pula, "to plant", is connected with the Dayak pambulan, "garden". But Dayak pambulan < prefix p(a) + imbul, "to plant" + suffix on. In conformity with a strict law of Dayak phonetics (§ 247), the i of imbul has had to assimilate itself to the a of the suffix.

III. The phonetic phenomena are wrongly explained. Original IN tunu, "to burn", appears in Pampanga as tun. Now according to Conant, in his article entitled “Monosyllabic Roots in Pampanga”, Journal of the American Oriental Society, 1911, p. 392, tun < tunu has lost the final u by apocope. That would, however, be the only case of apocope in a WB in Pampanga ; and accordingly every representative of IE scholarship will regard this explanation as unacceptable, because it leaves the case standing as an isolated phenomenon. — In reality, Original IN tunu was changed by meta- thesis (an extremely common phenomenon in IN, as we saw in § 236) into tuun, which was then contracted to tun.

IV. Too httle consideration is given to the meaning of words. Conant ibid., p. 392) adduces yet another instance of apocope: sut as compared with the Bisaya suta. But according to Bergaiio sut means “humillarse, rendirse yendo a la presencia de aquel a quien se humilla” ; while suta, according to Encarnacion, signifies: “descubrirse, hacerse patente, publico” . The meanings of sut and suta are therefore very divergent, so that it is impossible to connect these two words together.
V. But there is, above all, yet another thing which makes it appear as if the phonetic evolution of the IN languages were less subject to the rule of law than is the case with the IE ones. And that is a certain practice, widely spread in IN research, and not exactly wrong per se, but defective and apt to give rise to confusion. It is this: many lexicographers are in the habit of adding etymologies to their key- words; but in doing so they omit to indicate whether the words adduced for comparison from other languages are to be considered identical with the particular key-word in conformity with some phonetic law, or are merely in some way or other related to it. Example: in the “Kawi-Balineesch-Nederlandsch Glossarium”, p. 313, we find: “Panas; Malay, Sundanese, Madurese idem, Bimanese pana, Malagasy fana, ‘ warmth’ ”. Here the words panas, pana, and fana coincide with one another, in perfect conformity with phonetic law. — But on p. 302 we find:. “Pakan; Sundanese hakan, Madurese kakan, Malay makan, ‘to eat’ ”. Here the words do not coincide according to phonetic law, for in no case does, e.g., an Old Jav. p correspond with a Sund. h; we have here several variant formations from the WB kan, which in its monosyllabic shape exists in many IN languages. In Old Jav. the WB kan has been combined with the formative syllable pa-, in Sund. with ha-. — The IN lexicography of the future must perform this part of its work more precisely than has been the case hitherto.
343. In reality the occurrence of phonetic phenomena in IN is certainly not attended by any greater irregularity than it is in IE. We observe in IN the strictest regularity in a very great nmnber of cases. Brugmann KvG, § 19, 7, remarks: “That certain phonetic changes take place in regular conformity with some law, is often enough an obvious fact, e.g. the change of Original IE -m at the end of a sentence into -n in Greek” . We can say precisely the same of the treatment of Original IN -m in Hova, for it appears there invariably as -na = n + the supporting vowel, e.g. in inuna < Original IN inum, “to drink” .
344. Though on the one hand we observe in many cases the strictest consistency in IN, yet on the other we also sometimes observe the contrary, but such instances are not more frequent than in IE nor do they dift'er in kind from IE cases.
345. There is a series of phonetic phenomena, in IN as in IE, in which science neither can, nor does, expect absolute invariability. Such are metathesis, assimilation, dissimilation, and the like. But even here IN not infrequently displays a thorough-going consistency (see § 241).
346. A strikingly large percentage of the IN vocabulary is of onomatopoeic origin ; and it has already been observed in § 17 that onomatopoeic formations may evade the operation of phonetic laws. For the actions of beating, tapping, and pounding, there are in the various IN languages the interjections tuk or duk or puk or bug. Now from these interjections are derived a large number of WB's, whose meaning preserves the fundamental idea of beating or the like, or has diverged from it by transference. Examples: Karo tuktuk, “to knock”; Gayo tumbuk, “to beat” ; Malay tumbuq, “to pound”; Old Javanese gebug, “to beat”; Karo batuk, “cough” ; Malagasy dialects tútuka, “beak”; Old Javanese tutuk, “mouth”; in several languages tuktuk, “woodpecker” ; Karo pukpuk, “to labour hard” ; Tontemboan sinduk, “pounded rice”  ; Javanese pupugan, “fragment”.
Here, for example, there is no question of any phonetic connexion between duk in Tontb. sinduk and buq in Mal. tumbuq, for a Tontb. d never corresponds with a Mal. b.
347. In IN and IE the operation of phonetic laws is very often countered by the powerful influence of analogy and popular etymology. The power of popular etymology is very aptly and generally noticeable in names of animals, especially trisyllabic ones, as exemplified by the “Schweizerisches Idiotikon” " on the one hand and the Old Javanese vocabulary on the other: see e.g. in the former s.v. Ameise, “ant” , and in the latter s.v. alifan, “centipede” .
348. In IE research difficulties arise in connexion with certain phenomena that have been termed root-variation, root-determination, and the like: see Brugmann KvG, § 367. I refer to such cases as the existence alongside of one another of forms like the IE trep: trem: tres, in the Latin trepidus, “timid” , tremere, “to tremble”, and the Sanskrit trasati < treseti, “to tremble” . We find precisely similar phenomena in IN also: thus in Tontemboan there are the forms rěp, rěm, and rěs, in urĕp, “to cover”, rĕrĕp, “to overlap” ; urěm, “to clasp round”, tirěm, “to enclose”; kĕrĕs, “to clasp round”, kurĕs, “to cross one's arms” . And in IN such phenomena are even less easy to tackle than they are in IE. As a rule we are not even in a position to form any certain idea that they really involve phonetic questions.
349. But there still remain in IN as in IE some phonetic phenomena, of which one can only say either: “Here law- less chance holds sway”, or: “Research has failed to dis- cover the principle of the occurrence” . To such alternatives I feel that I have to resign myself after considering the representation of Original IN t in Bimanese and Original IN k in Nias. I. Original IN t in Bimanese:
Original IN Bimanese
tanda tanda, "mark".
tanah dana, "earth".
r2atus ratu, "hundred".
batu wadu, "stone".

II. Original IN k in Nias:

Original IN Nias
karaṅ kara, "coral".
kandan kandra, "stable".
kima gima, "shell".
kasaw gaso, "rafters".
kěn χo2
kait χai, "hook"'
kayu eu, "wood".
kulit uli, "skin".
350. Regularity in the occurrence of phonetic phenomena is greater in some languages than in others; it is greater, for instance, in Minangkabau than in Bimanese. It is also greater in some sounds than in others: in the nasals very much more than in the liquids, so that Bopp in his "Ueber die Verwandtschaft der malayisch-polynesischen Sprachen mit den indisch-europaischen", p. 66, 1. 15, rightly speaks of a "fluctuation of the liquids".
351. When we observe the phonetic processes of the IN languages we often get the impression that the movement is still going on and tending towards some end, which it has not yet attained. Such a presumable end, for example, is that "in Bugis initial tenues are endeavouring to disappear".
I. Initial k has to a great extent disappeared already, as in uliq, "skin" < Original IN kulit.
II. Initial p has disappeared in two words, viz. uso, heart" < Original IN pusu and uro, quail" < Original IN puruh
III. Of the disappearance of initial c and t there are no certain instances.

  1. [See Essay II, § 79.]


1. Die Beziehungen des Malagasy zum Malaiischen.*
2. Die Geschichte von König Indjilai. Eine bugische Erzählung ins Deutsche iibersetzt.†
3. Tagalen und Madagassen.‡
4. Ein Prodromus zu einem vergleichenden Wörterbuch der malaiopoly nesischen Sprachen.
5. Mata-Hari.
6. Wurzel und Wort in den indonesischen Sprachen.§
7. Sprachvergleichendes Charakterbild eines indonesischen Idiomes.
8. Gemeinindonesisch und Urindonesisch.§
9. Das Verbum.§
10. Der Artikel des Indonesischen verglichen mit dem des Indog..r manischen.
11. Indonesisch und Indogermanisch im Satzbau.
12. Die Lauterscheinungen in den indonesischen Sprachen. §


1. Der Natursinn in den älteren Litteraturwerken der Malaien.
2. Die Geschichte von Hang Tuwah. Translated from Malay.
3. Die Geschichte von König Indjilai. Translated from Bugis.
(Already mentioned above on account of its detailed linguistic commentary).
4. Die Grindung von Wadjo. Translated from Bugis.
5. Die Geschichte von Djajalankara. Translated from Makassar.


1. (a) Das Lehnwort in der bugischen Sprache. (b) Die Lehnwörter, welche der Luzerner Mundart und der bugischen Sprache gemeinsam angehören. In Drei Abhandlungen über das Lehnwort.
2. Anlaut und Auslaut im Indogermanin und Malaio polynesischen. In Album Kern.
3. Die Sprache der Liebe in der makassarischen Lyrik. In Mélanges de Saussure.
4. Die Stellug der minahassischen Idiome zu den übrigen Sprachen von Celebes einerseits und zu den Sprachen der Philippinen anderseits. In F. Sarasin, Versuch einer Anthropologie der Insel Celebes.

*Translated into English under the title: " The relationship between the Malagasy and the Malayan languages", by R. Baron, in the Antananarivo Annual and Madagascar Magazine, 1894-95, Antananarivo.
† Translated into Dutch under the title: " Geschiedenis van Koning Indjilai ", by M. C. Poensen, in the Indische Gids, 1900.
‡ Translated into Spanish under the title: "Tagalog y Malgache ", by P. L. Stangl, the first part in the Revista Historica de Filipinas, año 1, the second in the Biblioteca Nacional Filipina, año I, Manila.

§ Translated in the present volume.







(1-5) REHATSEK (E.). Mir Khwānd's “Rauzat-iis-Safa or” “Garden of Purity.” 1891 to 1894. Price 10s. a volume.

(6) TAWNEY (C. H.). The Kathā Kosa. 1895. Price 10s.

(7) RIDDING (Miss C. M.). Bana's Kadambari. 1896. Price 10s.

(8) CoWELL (E. B.) and THOMAS (F. W.). Bāna's Harsa Carita. 1897. Price 10s.

(9) CHENERY (T.). The first twenty-four Makāmāts of al Harīrī. 1898. Price 15s.

(10) STEINGASS (F.). The last twenty-four Makāmāts of al Hariri. 1898. Price 15s.

(11) GASTER (M.). The Chronicles of Jerahmeel. 1899. Price 10s.

(12) DAVIDS (Mrs. Rhys). The Dhamma Sangani. 1900. Price 10s.

(13) BEVERIDGE (Mrs. H.). Life and Memoirs of Gulbadan Begum. 1902. Price 10s.

(14, 15) WATTERS (T.). On Yuan Chwang's Travels. Edited by T. W. Rhys Davids and S. W. Bushell. 1904-5. Price 10s. a volume.

(16) WHINFIELD (E. H.) and MIRZA MUHAMMAD KAZWINI. The Lawā'ih of Jāmī. Facsimile of an old MS., with a translation and a preface on the influence of Greek philosophy upon Sufism. Second edition, 1914. Price 5s.

(17) BARNETT (L. D.). Antagada-dasāo and Anuttarova-viya-dasāo. From the Prakrit. 1907. Price 5s.

(18) KEITH(A. Berriedale). The Śāṅkhāyana Āranyaka. 1908. Price 5s. (19, 22) ROGERS (A.). Memoirs of Jahāngīr. Edited by H. Beveridge. Vol. I, 1909. Price 10s. Vol. II, 1914. Price lOs. a volume.

(20) NICHOLSON(R. A.). The Tarjumān al-Ashwaq of Ibn al-' Arabī. Text and Translation. 1911. Price 7s. Qd.

(21) WARDOP(Miss M.). The Man in the Panther's Skin. By Shot'ha Rusfhaveli. 1912. Price 10s.

(23) WARDOP (0.). Visramiani. The Story of the Loves of Vis and Ramin. A romance of Ancient Persia. Translated from the Georgian Version. Price 10s. Margoliouth (D. S.). The Hesht Bihisht. (In preparation.)



(1) GERINI (Colonel G. E.). Researches on Ptolemy's Geography (Further India and the Indo-Malav Peninsula) 8vo. 1909. Price 15s.

(2) WINTERNITZ (Dr. M.). Catalogue of Sanskrit MSS. in the R.A.S., with an Appendix bv E. W. Thomas. 8vo. 1902. Price 5s.

(3) HiRSCHFELD (Dr. H.). New Researches into the Composition and Exegesis of the Qoran. 4to. 1902. Price 5s.

(4) DAMES (M. Longworth). The Baloch Race. 8vo. 1904. Price 5s.

(5) LE STRANGE (Guy). Mesopotamia and Persia in the Fourteenth Century a.d., from the Nuzhat-al-Kulūb of HamdAllah Mustawfī. 8vo. 1903. Price 5s.

(6) BROWNE (Professor E. G.). Chahár Maqála of Niḏẖá-mí-i-'Arúdí-i-Samarqandi. 8vo. 1899. Price 3si

(7) CODRINGTON (0.), M.D., F.S.A. A Manual of Musal-man Numismatics. 8vo. 1904. Price 7s. 6d.

(8) GRIERSON (G. a.), C.I.E. The Piśāca Languages of North- West India. 8vo. 1906. Price 5s.

(9, 10) DAMES (M. Longworth). Popular Poetry of thenBaloches. Text and translation. Two vols. 8vo. 1907. Price 15s.

(11) SAYCE (Professor A. H.) and PINCHES (T. G.). The Tablet from Yuzgat in the Liverpool Institute of Archaeology. 8vo. 1907. Price 5s.
(12) BAILEY (Rev. T. Grahame). The Languages of the Northern Himalayas, being studies in the Grammar of Twenty six Himalayan Dialects. 8vo. 1908. Price 5s.
(13) BAILEY (Rev. T. Grahame). Kanauri Vocabulary. 8vo. 1911. Price 3s.
(14) lE STRANGE(Guy). Description of the Province of Fārs, in Persia, from the MS. of Ibn-al-Baḵẖĩ. 8vo. 1912. Price 5s.
(15) BRANSTETTER (R.). An Introduction to Indonesian Linguistics, translated by C. 0. Blagden. Svo. 1916. Price 7s. 6d.
(16) PINCHES (T. G.). Tablets from Lagaš and other Babylonian Sites in the possession of Randolph Berens, Esq. Svo. 1915. Price 5s.
(17) BAILEY (Rev. T. Grahame). Himalayan Dialects. (In the press.)

GASTER (M.). El-Asatir, or the Samaritan Apocalypse of Moses. (In preparation.)



(1) Hultzsoh (Professor E.). Prākritarupāvatāra of Simhāraja. The Text in Nagari characters, with Notes, Introduction, and Index. 8vo. 1909. Price 7s.
(2) BODE (Mabel Haynes), Ph.D. The Pali Literature of Burma. 8vo. 1909. Price 5s.
(3) Hultzsoh (Professor E.). The Mēghadūta with Vallabha's Commentary. 8vo. 1911. Price 7s. 6d.
(4) Bray (Denys de S.). The Life-History of a Brāhūī.

Svo. 1913. Price 5s.


A special discount is allowed to members of the Society