- 89

Illustration

BY comparing these with the numerals found in other dialects or languages of Oceania, we may make an endeavour to ascertain their original forms, that is, the forms of these words in the Oceanic mother-tongue.

One.—In Malagasy isaka, iraika, are formed from isa, iray, by attaching to them the well-known Malagasy (and Oceanic) formative suffix ka (see “The Malagasy Language,” by G. W. Parker). This explains Epi saka, Mallicolo soka, Efate dialect sikítika, New Guinea dik, Mangarei isaku. Sikitika is a reduplication, in which the numeral is pronounced both si and ti—without the suffix ka, in the - 90 believe that ira(i) is the same word as isa. The i is found suffixed same dialect, the numeral is ite. Efate sikai suffixes i to the suffix ka, and this suffix occurs also in Malagasy iray (i.e. irai); and I also in Epi tai, Eromanga sai, Cayagan tadday (tadai). This last tadai has the prefix ta or t' which is seen in MallicoIo tes, and in Samoan tasi, which Bopp, I believe rightly, thought a demonstrative particle. Another prefix is seen in Vaturanga kesa, Tanna kadi (dialects keri, kwati, kirik, kilik, d to t, r, and e), Anudha kedha (the dh like th in that). The Mallicolo san has the suffix n, seen also in Ceram san, Pelew tong, Marshall Islands thuon (thong-ul, 10; i.e. one-ten). Compare Malagasy singa, and with Mallicolo soka, New Guinea dik, Efate (siki-) tika; compare also Malagasy toka. With tadday compare Malagasy tsirairay, as to the prefix. It would thus appear that the original word in the Oceanic mother-tongue was a dissyllable (Malay asa, kiss ista, ida, Malagasy isa, ira(i), Timor aida, Savu aisa; while in Epi the word occurs in one dialect both as taga and rai, and in another both as tai and ri), of which the final syllable began with a consonant, now found as s, t, r (and l), th, and d. Of these consonants, d is probably the original in this word.

Two.—In this word the initial consonant in the above table is prevailingly r, which is of necessity l in Samoan. The r was, as is in the highest degree probable, the original form of this consonant in the mother-tongue. This r has become l as in Samoan, and d as in Malay dua, or duwa. In the Maori, rua is the common form; but rie, and rienga also occur dialectically; rie may answer to rua, Celebes dia, Tambora (ka)läe, Mangarei (lo)lai; but rienga has the suffix nga (for na) already noticed under the word “one.” Vaturanga ruka, Malagasy, (Sir Joseph Banks) rica, have the suffix ka above noticed under “one.”

Three.—As the above table indicates, telo, &c., t and l were the original consonants of this word in the mother-tongue. The Maori toru has of necessity r for the original l, and the Maori tengi seems to compare with the Gilolo sangi: it is worthy of remark that in Ambrym, Laman (off Epi), Eromanga, Tanna, and Aneityum, as in sangi, the initial consonant of this word occurs as s. Malay tiga has ga for ka, the suffix already noted under the words “two,” and “one,” and compares with Aneityum seik (dialectically seij, k to j as usual in Aneityum). In both seik and tiga the l has been elided before the suffix (ka), so in tengi and sangi before the suffix (n) ngi.

Four.—It is beyond doubt (1.) that t was the final consonant of this Oceanic numeral in the mother-tongue (Malagasy dialects effat, effats, efar), (2.) that its penultimate consonant was a labial, and (3.) that there was a syllable before this penultimate consonant, Tagala apat, Malay ampat, dialect enpa, Savu uppah, Batta opat. In Samoan fa, Bouru pa, ha, the final t is elided.

- 91

Five.—The final consonant of this word in the mother-tongue must have been as the table indicates m. As to what the initial consonant was originally the question is more difficult, and by no means settled in favour of l as the table suggests. In Bakian (Epi) it is jimo, jumo, or limo, and the Pelew has im, and Ceram has nima. But this throws little light on the matter. In Aneityum the word occurs as ikma or ijma (for kima or jima) and in that language k is changed to j, as already noted above under the word “three.” The question then is whether in this word for “five,” which means also in many dialects “hand,” that is the five fingers, the k or the l (d, j) is the more original. Evidence will be adduced below to show that this question must be decided in favour of k. The word in Vaturanga is kima, and jehe, in Gao kame and lima.

Six.—The Java nânâm is a reduplication. This word in Malay is anâm, Tagala anim, Bissayan unum. Like the final m in these words the final na of Malagasy enina is a suffix. Samoan ono, like Malagasy ene (Sir Joseph Banks) does not show this suffix. See further as to this suffix below under the word “nine.” The Mallicolo won suggests that an initial consonant in this word has been lost in many of the modern dialects. Bopp regarded Tahiti fene as from a form hene for sene. But fene compares rather with Mallicolo won, Ceram wonen, Gilolo butanga, Tembora bata-in, Pelew malong (=butanga). The Pelew lollom (=nânâm), Malagasy ene, &c., have not this initial consonant, while Sula gane, Guham gurum have the prefix g' (i.e., k') instead of it: the f therefore in fene is a prefix like the ma or m' in malong of which m' is, I believe, the original form. This is a well-known prefix to Oceanic numerals. Is the n in ene &c., the original consonant of this word, or a variant of a consonant more original? Gilolo butanga, Tidore rora, Mangarei daho, Guham gurum, Caroline hol and others, show that the answer to this question can only be uncertain till other evidence be adduced to show that the original consonant was t and not n.

Seven.—The table shows pitu, fito the prevailing form of this word in Oceanic, and the Mallicolo ntit or tit (Malagasy, Sir Joseph Banks, titou, i.e., titu, Mysol tit, Malay tuju) is the same word, the labial having been changed into a dental as it often is in that dialect. The same form of the word is seen in Sirang titura the ra being a suffix as in takura, 1). In tuju the final t of the word is changed to j. There can be no doubt that pitu or fitu is the more original form, the labial in this case having the priority.

Eight.—In the table valu is the most incorrupt form, but Savu panu compared with Atshin (Sumatra) lappan (lapan), Malay de-lapan, or se-lapan show that n was probably the original final consonant, and not l (as in valu, balu, Caroline ual, and uan), and that lapan is the fuller form from which by elision of the initial syllable, both panu and valu come.

- 92

Nine.—The forms in the table are all from one original which occurs in a more incorrupt form in Tagala siyam, Bissayan siam. The initial s is elided in Samoan, and has become hh in Mallicolo (2). The final m in siam, which we have already seen under the word “six,” as a suffixed particle, has been changed in this word in Malagasy and Samoan to v, Mallicolo (2) to p. Santo to w, and merged into the vowel in su. In sanga it has become ng: compare under the word “six,” above.

Ten.—Stripping off the prefixes from this word in the above table we have it as fulu, puluh, wul, vil, and til. In the Mallicolo til we notice the same interchange of v, or b, and t as appeared in tit: see above under the word “seven.” In Bouru the word occurs as boto, Amblaw buro; Santo bulu, fulu, fura, ula; Mallicolo pur, ful, bur, b'; Savu bo, uru; Ambrym hul, ul, pi; Oba fulu, buka (hanga-fulu, 10, nga-buka, 20); Pelew ok; Ternate yagi; Caroline sik or sig; Maori huru, wiri; Timbora, sarene, or sarone; Mangarei turu; Malagasy (Sir Joseph Banks) tourou, i.e., turu (see under “seven,” titu, which he writes titou); Mysol yah, lafu; Tagala polo and pobo, pulu and puvu; Bugotu lage; Nifilole kolu; Savo tale or sale; Ceram vuta, hutu, husa (this last is probably for hutusa, in which sa= 1, as in husa lesa 11, i.e., husa one ten, le and sa one, compare hutu=10, in the words for 20 and 30: see Wallace, “Malay Archipelago,” Appendix); Matabello sow (probably for ser, compare ter=10, in the words terwahei 11, ternorua 12, teranrua 20, terantola 30). What was the original form of the word thus singularly changed phonetically, not only in different dialects but sometimes in one and the same dialect? Manifestly the word in the Oceanic mother-tongue had two consonants, the last of which was probably r or l, and the first of which is represented in the above as p, f, v, b, w, h, y, spiritus lenis (h not pronounced), t, and s. There can be no doubt of the actual occurrence of each of these consonants as the first consonant of the Oceanic numeral “ten.” Thus though the Mallicolo til should be a local substitution of t for v this leaves the Malagasy turu, and most of all the Mangarei turu (Gabelentz & Latham), and Matabello ter, tera (Wallace) to be accounted for, and the latter two cannot be accounted for in this way. Then saro-ni, sare-ni sow, tale, and sale (Gao sale), along with hulu, huru would seem to indicate that probably s was the original letter which became t and h; h (see above under the word “six”) then probably became f (v, p, b, w). See the next word “eleven” for a confirmation of this view as to the priority of s as the first consonant of this numeral “ten.”

Eleven.—This in Java is suwalas; 12, rolas. Malay sablas, duwablus; in Java 14 and 16 are patbâlas, nâmbálas. No doubt therefore ro-las is, by contraction, for ro-bâlas; compare Malagasy roa amby ny folo, 12; iraika amby ny folo, 11; for the same construction, that is the construction in which, in the compound numerals - 93 from 11 to 19, the digit is put first, the “ten” last, as in the English “thirteen,” “fourteen.” In the above table it is seen that in the Samoan, Santo, and Mallicolo the “ten,” on the contrary; is put first, and the digit last, the compound expression for 11 being “ten and one;” whereas in Malagasy and Java it is “one and ten;” and so with the others up to nineteen. Now in these compound numerals, as the table shows, there is no difficuly, save in the Javanese, in pointing out the three parts of the expression, namely: that denoting the “ten,” that denoting the digit, and that equivalent to “and.” To take the last first, in Santo it is rav, in Mallicolo dromon, in Samoan ma le, and in Malagasy amby ny; and, I believe, in Java it is lâla or la. On looking down Wallace's list, under the words “eleven,” and “twelve,” it will be seen that this same “and” is expressed in Cajeli (Bourn) by le, and in Amboyna by ala or ela, and in Ceram by la or le; and that in these we have the same la as appears in Java suwâlas or suâlas, 11; rolas, 12; &c., there can be little if any doubt. Now this brings us to the conclusion that in Java the part of these compound numerals which denotes “ten” is s, thus rolas, 12, is ro 2, la “and,” s' 10. Probably, instead of s' we should say as, in that case rolas being ro, la, and as', for thus the elision of the final consonant of the word “ten” would be more easily accounted for. This s (or as), it need scarcely be said, is held to be the s in the above sow, saro-ni, &c. It can be shown beyond all doubt that the second consonant of the Oceanic word “ten” is found elided or lost in widely separated dialects, for instance in Sava bo-aisa (aisa=1), Mallicolo singāb—probably for singa-ab (singa=1), dialect singeap (singe-ap)—Ceram husa (sa=1), &c. That the same word (originally) should now occur in two such different forms in the same dialect, as (in Java) puluh and s, is not without parallel in other dialects as to this very word: compare Matabello sow, ter; Savo tale (sale), bolo; Gao boto, 10; but sale (kaheni) 11, sale (paluni) 12.

Hundred.—Java atus, Malagasy zato, Sula ota, Bouru ūtun and botha, Mallicolo ngut, Malay ratus, Bouton säatu, appear all to spring from one original; the s in atus, z in zato, n in utun, r and s in ratus, and sa in saatu, being non-radical. If bot is the same then from the others, an original labial has been elided; and if ngut is the same this original labial was probably m.

Thousand.—Malagasy arivo, Malay ribu, Java ewu, Samoan afe, Santo rowuna, Tagala libu, Bissayan livu, manifestly represent the same original of which the initial consonant was r or l, and the final consonant a labial.

Quinary System.—What Gabelentz in his work on the Melanesian languages first called by this name need not detain us long. A good example of it occurs in Enganho, a small island off the south-east coast of Sumatra; and a comparison of the Enganho numerals with - 94 those of Santo, Efate, and Ambrym will sufficiently show the nature of this system:—

In the Santo lina-rav-e, literally “five and one,” lina-rabi-rua, “five and two,” we have the same rav=and, already seen in another Santo dialect: see above under the word “eleven.” In Efate la-tesa (for lima-tesa), and Ambryn limsi (lim 5, si 1) the “and” is left out. In the above four languages the common Oceanic word for “ten” is seen in three, only the fourth (Efate) having instead rualima literally “two (of) five.” It should be observed that the ma or mo prefixed to the Santo numerals is the “Verbal Pronoun” of the third person, hence matea=it is one, mo-rua= they are two. Taking such as these Santo and Ambrym numerals we can only conclude that the ancestors of those now using them had either forgotten the separate words for 6, 7, 8, 9, or preferred to use these very easy and natural substitutes, while the Efatise added to them the word for 10 also. The Efatise is therefore the most completely “quinary.” Nevertheless, as is well known, the Efatise system of numeration is as decimal as is the English. That is to say, “quinary” can only be applied to the words denoting the numbers etymologically considered. And as these words are simply compounds of the common Oceanic numerals above discussed, they throw no additional light upon our present enquiry.

In the following table the forms of the separate numerals on the one side are forms actually occurring all from one original, that of the Oceanic mother-tongue; and on the other side are placed the consonants, with apostrophes for the omitted vowels, of the probable original forms:—

Turning now to the Continent of Asia, and to the same quarter as before in the case of Personal Pronouns—see Journal of the Polynesian - 95 Society, Vol. I., page 259, on “The Asiatic Origin of the Oceanic Personal Pronouns”—we find that

• 1—is in Arabic, ahad', fem. ihda(y); Ethiopic, ahadu, f. ahati; Syriac, had, f. hdo; Modern Syrica (common, i.e., mas., or fem.), ha, hda.
• 2—Arabic (C. de Percival), etnein, t'nein, fem-etnetein, tintein; Syriac, treyn, f. tarteyn; Modern Syrica (common), tray, or trai; Mahri (c.), tharo; Sokotra (c.), tarawah.
• 3—Arabic (C. de Percival), t'laté, f. t'lat (ancient), thalath', f. thalathat'; Ethiopic, salastu, f. salas; Amharic (c.), sost: Modern Syriac (c.), t'laa; Mahri, shathet; Sokotra, tataah; Syriac, f. tolto.
• 4—Arabic, arba', f. arbaat'; Ethiopic, arbā'tu, f. arbā', reb'e; Amharic (c.), arut (ubah, at Arkeeko); Modern Syriac (c.), arbā'; Mahri (c.), robot; Sokotra (c.), arba'ah.
• 5—Arabic (C. de Percival), khamsé, f. khams; Modern Syriac (c.), hamsha; Mahri (c.), khomo; Sokotra (c.), khemah; Amharic (c.), amst.
• 6—Arabic (C. de Percival), sitté, f. sitt; Mahri (c.), iteet; Sokotra (c.), yitah; Modern Syriac (c.), ishta; Assyrian, sudu.
• 7—Aaabic, sab', f. sab'at'; Ethiopic, sab'atu, f. sab'u; Amharic (c.), sabāt; Assyrian, sibit; Modern Syriac (c.), shoa; Mahri, ibet; Sakotra, yibi'ah.
• 8—Arabic (C. de Percival), t'manié, t'mani, f. t'man; Ethiopic, samantu, f. samani; Amharic (c.), sment; Modern Syriac (c.), tmania; Sokotra (c.), tamani (Mahri, thimaneet).
• 9—Arabic (C. de Percival), tis-à, f. tis'; Ethiopic, ts'atu, f. ts'u; Mahri (c.), iset; Sokotra (c.), sa'ah; Modern Syriac (c.), itsha; Amharic (c.), zatang.
• 10—Arabic (C. de Percival), achra (i.e., ashra), f. acher (i.e., asher); Ethiopic, asartu f. asaru; Hebrew, asarah, fem. eser; Syriac, esro, f. 'sar; Modern Syriac (c.), usra; Amharic (c.), eser and eseru; Mahri (c.), ashareet; Sokotra, ashri.
• Teen—That is 10 in the words 11-19 inclusive is found in Modern Syriac, thus—(c.) 11, hadusar; 12, trusar; 13, tiltasar—in Amharic (c.), 11, eseru ande; 13, eseru sost—and in vulgar Arabic (c.), 11, hidash; 12, tnash; 13, tlittash; 14, artatash, &c. The Syriac and Arabic put the 10 last, as do the Malay and Javanese; the Amharic puts it first, as do the Mallicolan, Samoan, &c.
• 100—Arabic, miāt'; Ethiopic, mĕĕtĕ; Amharic, mato.
• 1000—Arabic, alf'; Hebrew, elef; Assyrian, alapu; Ehkili (dialect of South Arabic, sister dialect to Mahri) of (the l elided).

Let it be observed (1) that these Semitic numerals are the well known common words peculiar to the Semitic family, and that, separately considered, they are sprung from one original; (2) that in the Ancient Semitic languages the numerals have a masculine and a feminine form, while generally in their modern representations this distinction of gender is lost, as may be seen in the above numerals marked (c.), that is, common gender; (3) that the feminine form of the numerals was distinguished by a suffixed t, which often became h, or was elided; (4) that this feminine or abstract form was used with masculine nouns, and hence became the most commonly used form, and almost the sole form used in modern dialects, in the words denoting the numbers 3–10 inclusive, as may be seen by noting the words marked (c.) above; (5) that some grammarians call the feminine-form numerals, as Dillmann (see the Ethiopic above), masculine, because they are used with masculine nouns, and some feminine (see the Arabic above), as Caussin de Percival, because they are of the - 96 feminine or abstract form; (6) that the numerals 1 and 2 were used, those of the masculine form with masculine nouns, and those of the feminine form with feminine nouns. It has to be observed also that Himyaritic sometimes suffixed m and sometimes n to these numerals (Halèvy and Prideaux), that is to the numerals 1–10, exclusive of the word for 2. It will be observed from the above that Ethiopic suffixed u, as in ahadu 1, and to the t in the numerals 3–10 (the t being the feminine or abstract ending). These three suffixed particles m, n, and u, are demonstrative particles of pronominal origin (Dillmann and Helèvy),

“No words,” says Professor Sayce (“Assyrian Grammar,” page 134), “are more used than those which denote the numerals; and consequently no words are more liable to be contracted, changed, and, in short, to undergo all the phenomena of phonetic decay.” As an example of this he gives Assyrian edu 1 as a contraction of ekhadu (=Ethiopic ahadu 1), “Assyrian Grammar,” page 135. Whether this be a correct derivation of edu, or not, it is undeniable that the Amharic, andĕ 1; Tigre, ade; modern Syriac, ha and hda, are at least equally striking corruptions of the original word. Then as to the word for 2, Professor Sayce remarks that the “Aramaic tĕrēn shows how an often-repeated word could change its primitive form.” The construct or short form of trēn was trē, hence the modern Syriac (sole) form given above trai. Gesenius, under the Hebrew shnaim, construct shnē 2, remarks of trēn that it “is very different from the primary form,” the original n having become r, as it has also in Mahri tharo, Sokotra tarawah, given above: he adds, “the primary form of the numeral appears to be tnē (or thnei), from which have been softened Sanscrit devi, dual dwáu . . . whence English, German, two, zwo: Greek, Latin, δôo, duo.” Gesenius, under the Hebrew word shalosh, 3, remarks that “amongst the Indo-Germanic languages the primary form appears to be retained in the Zendic teshro, whence with the letters transposed are both the Aramaic telát, and Greek, Latin, τpε˜ῖς, tres. The Sanscrit has the abbreviated tri.” He also compares the Semitic word for 5, Arabic khams, and Sanscrit pantshan; Hebrew shesh, 6, and Sanscrit shash; Hebrew sheba', and Sanscript sapta, English seven; and Hebrew ehad, 1, and Sanscrit eka; but the Semitic words (see above) for 4, 8, 9, 10, 11–19, 100, and 1,000, he finds no Indo-European words to compare with. Before passing from this comparison by Gesenius of these Semitic and Indo-European numerals, it may be remarked that the above comparison as to the words for 5 may be left out of account, and that in fact Renan (“Histoire des Langues Semitiques,” page 464) does leave it out of the account, while he marks the comparison as to 1 (ehad, Sans:eka) as doubtful, and acknowledges the resemblances between the Semitic and Indo-European words for - 97 2, 3, 6, and 7, as also that between certain of the Semitic and Indo-European personal pronouns, but denies that the resemblance arises from the identity of their origin, holding it sufficiently accounted for by the fact that both the Semitic and Indo-European speakers were alike human beings. Happily, it is not necessary to the discussion of the subject of this paper to express any opinion in this controversy. It may be noted, however, that according to Professor Sayce, who endeavours to trace the numeral words, 1–10, to their sources or radical ideas, that is according to his derivation of these words, the whole of the Semitic and Indo-European numerals, notwithstanding the above striking resemblances, are radically different from one another. It is not alien to our purpose to note that, according to Professor Sayce, the Semitic word for 5, khams, which has lost its kh in Amharic, and “has changed it into s in the Berber summus,” radically means “hand with its five fingers,” being connected with “qometz (Hebrew) the first;” and that the Semetic word for 10 (above) is connected with asar, to bind together, “referring to the combination of the two hands,” or, as Gesenius expresses it, “the conjunction of the ten fingers.”

In the above given Semitic numerals it will be noticed that in the word for 3, Amharic and Mahri elide the l retaining the final radical and also the feminine ending t, and that Modern Syriac retains the l while eliding both the third radical and the ending t; that in the word for 4, Amharic retains the ending t eliding the radical b, while at Arkeeko b is retained and the radical r and ending t elided; that in the word for 5, the ending t is elided in Modern Syriac, the radical sh (s) being retained, while in Mahri and Sokotra both the radical s and ending t are lost; that in the word for 6, Mahri elides the initial radical s, as also does Sokotran, but Sokotran changes the ending t (as usual) to h, while Mahri retains it; that in the word for 7, Mahri and Sokotran elide the initial radical s and Modern Syriac the radical b, and as before Mahri retains the ending t, which Sokotran changes to h; that in the word for 9, Mahri and Sokotran elide the initial radical t, Mahri retaining the ending t, Sokotran changing it to h, and Modern Syriac eliding it; and that in the word for 10, Hebrew changes the ending t to h, Syriac and some others elide it.

In accordance with the foregoing, we expect, in comparing the Oceanic numerals with those of the Asiatic group indicated, to find that those from 3–10 inclusive should be, while of common gender like the Modern Syriac and Amharic, also, like these, representations of the ancient feminine-form numerals, and that the numerals “one” and “two” should represent the ancient masculine forms of these words rather than the feminine (for 1, the Modern Syriac uses both ha and hda indiscriminately, for 2, the representative of the ancient masculine form). We may now compare the following actual forms, - 98 which, if they are of one origin, settle the question as to the Asiatic origin of the Oceanic numerals in all their forms:—

It may be noted that the ancient feminine ending t is still unmistakably seen in the words for 4 and 7 in the Oceanic, and that as the final u in the word for 7 manifestly compares with the same in Ethiopic, so we may regard also the final u in the words for 3 and 8. On both sides in this table, in the words for 6, 7 and 9, the original initial consonants s, t, are elided, while in Oceanic the initial t or s has also been elided in the word for 2, and in panu, one of the forms of the word for 8. It is easy to see how tman became pan, and trai, rai or lai, or tharo, ro.

To discuss this fully is not necessary here. Suffice it to say that the t in Samoan tasi, 1, Maori tahi, is a demonstrative particle found similarly used in other dialects; and the same may be said of the Maori ko in kotahi, 1. The ka prefixed to the numerals in Malay and Efate, forming the ordinals, is the same k' as is found in other dialects prefixed to the cardinals, as e.g. in Tanna kadi, 1, and has the same k' as the Maori ko. The suffixed ka in isaka, 1, tiga, 3, is found suffixed in the Harari (Semitic) forming the ordinals, and to nouns and adjectives in Mahri forming abstract nouns and adjectives, as in Malagasy also, and other Oceanic languages (vide “South Sea Studies”). The suffix m in anam, 6, siam, 9, is found also as a suffix to all the numerals 1–10, except 2, in Himyaritic. The suffixes 'n' (ng) and i are also found both in the Island and Asiatic groups.

Illustration