- 458

• 1. Introduction
• 2. Combinatorial Analysis of Bi-Vocalic Morpheme Structure in four Oceanic languages, in Proto-Polynesian and in Proto-Austronesian
• 2.1 Bi-Vocalic Morpheme Structure in Hawaiian (HAW)
• 2.2 Bi-Vocalic Morpheme Structure in Tuamotuan (TUA)
• 2.3 Bi-Vocalic Morpheme Structure in Maori (MAO)
• 2.4 Bi-Vocalic Morpheme Structure in Fijian (FIJ)
• 2.5 Bi-Vocalic Morpheme Structure in Proto-Polynesian (PPN)
• 2.6 Bi-Vocalic Morpheme Structure in Proto-Austronesian (PAN)
• 3. Comparison of Results
• 4. Summary
• 5. Tables 1 to 61

1. Introduction

This paper is an attempt to analyse an aspect of phonemic structure of bi-vocalic (disyllabic) morphemic forms in four living Oceanic languages—Hawaiian (HAW), Tuamotuan (TUA), Maori (MAO) and Fijian (FIJ)—and in two reconstructed Austronesian languages—Proto-Polynesian (PPN) and Proto-Austronesian (PAN). It is concerned specifically with two-member combinations of consonants (C) and vowels (V) functioning as constituents of morphemic forms.

The aim of the present work, as specified above, places it in the category of studies dealing with the distribution of phonemes within a higher level sequential unit such as the syllable, morpheme, or word. This sort of study is often termed phonotactics. 2

- 459

Both theoretical and practical aspects of phonotactics have been discussed by members of the Prague school, especially by Trubetzkoy, 3 Mathesius 4 and Trnka. 5 In present-day Czechoslovakia this tradition has been revived by Altmann 6 and Krámsky. 7 Copenhagen and other Scandinavian linguists ought to be mentioned in this connection as well, e.g. Vogt, 8 Diderichsen, 9 Spang-Hanssen 10 and Sigurd. 11 As far as American linguistics is concerned a theoretical publication by Harary and Paper 12 and a paper by Greenberg 13 are worth mentioning. An exhaustive bibliography of works dealing with phonotactics is given by Sigurd. 14

In the field of Austronesian languages this type of investigation has been carried out by Uhlenbeck, 15 Altmann 16 and Chrétien. 17 All of them have concentrated upon studying the structure of the disyllabic morpheme. Uhlenbeck's paper on the Javanese morpheme, in which he discovered various interesting structural regularities, has exerted a considerable influence upon Odendal who has analysed the structure of root morphemes in the Afrikaans language. 18 Altmann has applied strictly statistical techniques to the analysis of vocalic relations in several Indonesian languages, 19 and he, for the first time in the field of Austronesian languages, tests his hypotheses for statistical significance. Chrétien concentrates on one language, PAN, but his study is more inclusive. He takes into account not only V1—V2 combinations but all possible two-member combinations of both consonants and vowels, with the exception of C1—V2.

Basically this paper follows Altmann in its methodology and Chrétien in its scope. The aim of the present author is not to achieve methodological originality, but to take some elementary statistical techniques current either in linguistics or in other behavioural sciences, especially psychology, and to apply them to the analysis of the phonemic structure of bi-vocalic morphemic forms in Oceanic languages. All the languages included are genetically related. Thus a basis is obtained for a confrontation of the genetic classification of these languages with the typological one. Data needed for the typological evaluation and comparison of the languages involved are obtained from a quantitative anlysis of all two-member relations holding between the constituents of the bi-vocalic morphemic form, i.e. C1, V1, C2, V2 and, in PAN, C3.

The two-member relations mentioned above may be classified into three different types: (i) homogeneous relations (C1—C2, C1—C3, C2—C3, V1—V2); (ii) heterogeneous intrasyllabic relations (C1—V1, C2—V2, V2—C3); and (iii) heterogeneous intersyllabic relations (C1—V2, V1—C2, V1—C3).

The set of these relations is regarded as the structure of bi-vocalic morphemic forms of Oceanic languages for the purposes of this paper.

- 460

It should be added that no exhaustive description of all aspects of morpheme structure is given below. Lacunae are left, especially as far as the consonants are concerned. Only two rough classifications of consonants have been used, i.e. locational and modal classifications (ref. section 2). It is considered that Jakobson's classification of distinctive features 20 might have some bearing on problems discussed here. Besides, it would be interesting to examine heterogeneous relations under modal classification of consonants as well. Notwithstanding these and other gaps present in this work, it aims at making a contribution to several questions of a general nature, two of which deserve special mention.

Firstly there is the question of universal regularities of morpheme structure in various languages. Speakers of all languages communicate by means of the same articulatory apparatus. At the same time, all languages conform to the laws of communication. These two factors must certainly result in the existence of some structural features which are shared by morphemes of all languages. A similar view has been expressed by Sigurd, “A connection betwen distributional and articulatory properties seems very natural: the phonemes are produced by the articulatory organs and a priori it seems probable that restrictions in combinations are to some extent due to physical restrictions or tendencies of the movements of the vocal organs. Distributional features are often correlated to place of articulation, manner of articulation, and expiration.” 21 As an example of such a feature the so-called repulsion of like consonants can be adduced. Its existence has been demonstrated in all languages examined here. Besides, it has been reported from many other languages, e.g. from Afrikaans, Arabic, German, English, Swedish, Czech, Italian. 22

Secondly there is the question of the relevance of phonotactic research for historical and comparative purposes. According to a wide-spread a priori point of view there can be no parallels between the genetic and the typological classifications of languages. Comparison of Oceanic languages as to one aspect of their morpheme structure indicates, however, that a parallelism between typological and genetic classifications cannot be rejected in advance. The point is that most, if not all, current typological classifications of languages are highly episodic, unsystematic and subjectivistic. They simply do not bear any comparison with the highly systematised and fairly objectivistic classification of languages based on the degree of their genetic closeness. On the other hand, an exhaustive analysis of morpheme structure by means of some quantitative methods would furnish us with systematic material that would suit comparative purposes more than traditional typology. Sigurd is of the same opinion when he writes, “This model (i.e. a general model for the description of phonotactic patterns) may also be used for comparison between the Scandinavian languages (or between other languages) and seems to be suitable for the diachronic study of Swedish too.” 23

- 461

It was mentioned above that the subject of this paper is an analysis of phonemic structure of the bi-vocalic morphemic form in several Oceanic languages which can be schematised as C1V1C2V2 (C1V1C2V2C3 in PAN). This is the most frequently utilised canonic type of morpheme in all the languages involved. According to Malmberg's statement about the relation between the length of a word and its frequency, 24 the shortest morphemes are normally the most frequent ones. This does not hold for Oceanic languages either as to the speech frequency or as to the lexicon frequency. A more precise formulation is required here. There are monosyllabic, disyllabic, trisyllabic and possibly a few tetrasyllabic morphemes in the Oceanic languages involved. Since there is only one type of syllable (CV) in all these languages and since the number of phonemes is generally very small, the class of all theoretically possible monosyllabic morphemic forms never contains more than a few tens of items. For example, there are, in MAO, only 55 theoretically possible monosyllables, but there are 3025 theoretically possible disyllabic morphemic forms, 166,375 theoretically possible trisyllablic forms and 19,150,625 theoretically possible tetrasyllablic forms. Even if all 55 theoretically possible monosyllables were to be utilised in MAO, they would represent just a tiny group when compared to the number of disyllables actually used by the language. In MAO 1258 disyllabic morphemic forms have been identified by the present author. 25 The disyllables represent the strongest class of morphemic forms in all Oceanic languages investigated here. There are fewer than a hundred more than disyllabic morphemic forms in each of these languages. 26 Here the notion of utilisation should be introduced. The degree of utilisation of theoretical possibilities U is defined here as the ratio of all actually occurring morphemic forms of a given type Ob, and of all theoretically possible morphemic forms of that type T, i.e. U=100.0b/T. The ratio is multiplied by 100 in order to obtain a percentage. It was found in Oceanic languages that U decreases rapidly with the increase of the number of syllables. Thus in MAO the degree of utilisation of the monosyllabic forms equals 69.09%, that of the disyllabic forms 41.59%, while that of the trisyllabic forms is less than 0.001%. The above-mentioned statement by Malmberg may now be modified to the effect that the smaller the size of a type of morpheme in terms of syllables, the greater its utilisation.

It ought to be stressed once again that this is an analysis of the phonemic structure of morphemic forms. This means that the semantic aspect is put aside here. This operates two ways—two or more homonymous morphemes are counted as one item, while two or more free or conditioned variants of one morpheme are counted as two or more items. It should be said that there is a good deal of free variation in Oceanic languages as far as the phonemic shape of a morpheme is concerned. The variation h ˜ f is rather common in MAO, e.g. hea ˜ fea, hia ˜ fia, haro ˜ faro. The variation of n ˜ r is also found in MAO, e.g. - 462 naku ∽ raku, nehu ∽ rehu, wini ∽ wiri. A similar variation of n ∽ l occurs in HAW, e.g. ha-nana ∽ ha-lana, ma-nino ∽ ma-lino, nanu ∽ nalu. Vocalic alternations are even more common. Both assimilative and dissimilative processes are found to take place in all of the languages involved. As examples of an assimilative change the following MAO morphemes are given: hari→heri, kari→keri, kita→kite, taina→teina, tahi→tehi, ruahine→ruahini, inu→unu. Dissimilative changes can be illustrated by the MAO examples tupu→tipu, tumu→timu. Sometimes it is impossible to decide whether an assimilation or a dissimilation takes place, e.g. the MAO alternations umu ∽ imu, puru ∽ puri, humu ∽ himu, uho ∽ iho, uto ∽ ito and the HAW alternations hene ∽ hena, hili ∽ hilo, poho ∽ paho. Evaluation of these and similar variations from the historical point of view will not be possible until a reliable reconstruction of the pertinent proto-languages is available.

All two-member combinations of consonants and vowels as constituents of bi-vocalic morphemic forms will be subject to the chi-square (X2) test which is a goodness-of-fit test. The observed scores characteristic of the individual combinations of a relation type are cast into a contingency table (Table I).

V1\V2	a	e	i	o	u	Nj
a	72	50	47	49	48	266
	65.92	51.18	51.70	48.08	49.12
e	50	55	22	35	33	195
	48.32	37.52	37.90	35.25	36.00
i	43	19	55	42	25	184
	45.60	35.41	35.76	33.26	33.98
o	45	40	43	49	36	213
	52.78	40.99	41.40	38.50	39.33
u	45	34	33	11	48	171
	42.38	32.90	33.24	30.91	31.57
Ni	255	198	200	186	190	1029

It is assumed that the elements entering into combinations are independent. This is the null hypothesis. The X2 test is used in order to prove whether the null hypothesis should be accepted or not. It should be discarded if the computed X2 associated with the relation examined is equal to or greater than the proper tabulated X2 at the chosen level of probability (here 0.05) with (r-1). (c-1) degrees of freedom, where r=rows and c=columns. The X2 test is based on comparison of the observed values with some theoretical (expected) values. These are the - 463 values which would be found for combinations of independent elements.

They can be calculated as Eij=Ni.Nj/N, where E=expected value, N=the total of all morphemic forms involved and Ni, Nj are partial totals of rows and columns. When the expected values are computed (bottom figures in Table I), the X2 can be computed as X2=nΣ j=1 (Oij—Eij)2/Eij, where O=observed value. When the computed X2 is significant this means that the elements in relation are not independent. The investigator may be interested also in which individual combinations of the elements display significantly great deviations of the observed values from the expected values, i.e. which ones display positive or negative combinatorial tendencies (associations and dissociations respectively). The pertinent X2 will be calculated from the formula which was first applied to linguistic data by Altmann: 27

X2ij=(Oij—Eij)2/Eij+[(N—Oij)—(N—Eij)]2/N—Eij.

In section 2 of this paper the combinations whose X2ij are significant at least at the 0.05 level with 1 degree of freedom will be regarded as displaying combinatorial tendencies—associations if Oij > Eij; dissociations if Oij < Eij.

The diagonal test 28 requires that the observed values of the diagonal cells are summed up and the same is done with the observed values of the non-diagonal cells. After the pertinent expected values are summed up, the chi-square test is applied to evaluate the deviations of the observed values from the expected ones.

The position test indicates whether the frequency of occurrence of a phoneme is dependent upon the position (P) in which it occurs or not.

The expected values are calculated as E=P1+P2/2.

2. Combinatorial Analysis of Bi-Vocalic Morpheme Structure in four Oceanic languages, in Proto-Polynesian and in Proto-Austronesian

Throughout this section the locational classification of consonants is made in terms of their place of articulation, and the modal classification is made in terms of their manner of articulation. The zero consonant value is indicated by O.

2.1 Bi-Vocalic Morpheme Structure in Hawaiian (HAW)

The data on bi-vocalic morphemic form structure have been obtained from Pukui and Elbert's HAW dictionary. 29 1029 bi-vocalic morphemic forms have been found, which is 50.81% of the 2025 theoretically possible bi-vocalic forms in HAW.

- 464

There are 5 vowels a e i o u and 8 consonants in HAW. The consonants are classified locationally into front (F) m p w, middle (M) n l and back (B) k ? h, and they are classified modally into stops (S) p k ?, nasals (N) m n, fricative (F) w h and the liquid l.

As far as the V1—V2 combinations are concerned, the associated X2=64.89, which is significant at the 0.001 level with 16 degrees of freedom. This proves that combinatorial tendencies are present in vocalic combinations of HAW bi-vocalic morphemic forms. Three significant associative tendencies (e—e, i—i, u—u) and three dissociative tendencies (e—i, i—e, u—o) have been found. For observed values see Table 1.

The diagonal test proves that all associations occur significantly in the diagonal cells, while all dissociations occur outside the diagonal. To be more precise, combinations of like vowels are significantly preferred, while those of unlike vowels are significantly under-utilised. The associated X2=29.16, which is significant at least at the 0.001 level with 1 degree of freedom.

The position analysis (Table 48) brings no significant results, which means that all vowels occur with about the same frequency in both the first and the second syllables.

The X2 characteristic of the C1—C2 combinations under locational classification is equal to 26.25, which is significant at the 0.01 level with 9 degrees of freedom. The combination of F—F was found to be dissociative; that of B—F was found to be associative. For observed values see Table 2.

Combinations of like consonants are significantly under-utilised while those of unlike consonants are significantly over-utilised. This is proved by the diagonal test; the X2=14.19, which is significant at the 0.001 level with 1degree of freedom.

According to the position test (Table 49) O prefers the second position (X2=9.86, which is significant at the 0.01 level with 1 degree of freedom), F prefers the first position (X2=22.26, which is significant at least at the 0.001 level with 1 degree of freedom) and M prefers the second position (X2=4.84, significant at the 0.05 level with 1 degree of freedom).

Under modal classification of consonants the X2 characterising C1—C2 combinations equals 9.11, which is not significant at the 0.05 level with 16 degrees of freedom. No significant combinatorial tendencies exist in HAW bi-vocalic morphemic forms under modal classification of consonants. For observed values see Table 3.

As far as heterogeneous relations are concerned, the testing has brought no significant results. For C1—V1 (Table 4) X2=4.81, which is not significant at the 0.05 level with 12 degrees of freedom; for C2—V2 (Table 5) X2=14.63, which is not significant at the 0.05 level with 12 degrees of freedom; for C1—V2 (Table 6) X2=3.68, which is not significant at the 0.05 level with 12 degrees of freedom; and for V1—C2 (Table 7) X2=4.12, which is not significant at the 0.05 level with 12 degrees of freedom.

- 465

2.2 Bi-Vocalic Morpheme Structure in Tuamotuan (TUA)

The data needed for analysis have been extracted from Stimson and Marshall's TUA dictionary. 30 It is worth mentioning that Stimson was influenced by Williams' MAO dictionary, and this influence may have left its imprint on his dictionary of TUA. 1211 bi-vocalic morphemic forms have been found, which is 40.03% of the 3025 theoretically possible bi-vocalic morphemic forms in TUA.

There are 5 vowels a e i o u and 10 consonants in TUA. The consonants are classified locationally into front (F) m p v f, middle (M) n r t and back (B) ŋ k h, and they are classified modally into stops (S) p t k, nasal (N) m n ŋ, fricatives (F) f v h and the r.

As far as the V1—V2 combinations are concerned, the associated X2 equals 69.09, which is significant at least at the 0.001 level with 16 degrees of freedom. Four associative tendencies have been proved to exist (e—e, i—i, o—o, u—u) as well as four dissociative tendencies (e—i, i—e, i—u, u—o). For observed values see Table 8.

The diagonal test proves that combinations of like vowels are over-utilised while those of unlike vowels are under-utilised. The associated X2=22.82, which is significant at least at the 0.001 level with 1 degree of freedom.

The position analysis of frequency of the individual vowels (Table 50) indicates that all vowels occur with about the same frequency in both the first and the second syllable.

The strength of combinative tendencies of C1—C2 under locational classification of consonants (Table 9) is expressed by X2=39.51, which is significant at the 0.001 level with 9 degrees of freedom. Two significant dissociations, F—F and M—M, and one association, M—F, have been found.

According to the diagonal test (X2=28.74, which is significant at least at the 0.001 level with 1 degree of freedom) combinations of like consonants are under-utilised and combinations of unlike consonants are over-utilised.

The position test (Table 51) indicates that O prefers significantly the second position (X2=7.61, which is significant at the 0.001 level with 1 degree of freedom) and F prefers the first position (X2=27.56, which is significant at least at the 0.001 level with 1 degree of freedom).

Under modal classification of consonants no significant tendencies have ben found for C1—C2 combinations (Table 10). The X2=13.59, which is not significant at the 0.05 level with 16 degrees of freedom.

As far as heterogeneous relations are concerned, no significant chi-square values have been obtained, which indicates the lack of significant combinative tendencies. For C1—V1 (Table 11) X2=19.49, which is not significant at the 0.05 level with 12 degrees of freedom; for C2—V2 (Table 12) X2=13.69, which is not significant at the 0.05 level - 466 with 12 degrees of freedom; for C1—V2 (Table 13) X2=2.97, which is not significant at the 0.05 level with 12 degrees of freedom; and for V1—C2 (Table 14) X2=7.09, which is not significant at the 0.05 level with 12 degrees of freedom.

2.3 Bi-Vocalic Morpheme Structure in Maori (MAO)

Analysis of bi-vocalic morphemic forms in MAO is based on Williams' dictionary. 31 Here 1258 bi-vocalic morphemic forms have been extracted, which is 41.59% of the 3025 theoretically possible forms of this type in MAO.

There are 5 vowels a e i o u and 10 consonants in MAO. The consonants are classified locationally into front (F) m p w f, middle (M) n r t and back (B) ŋ k h, and they are classified modally into stops (S) p t k, nasals (N) m n ŋ, fricatives (F) f w h and the r.

Combinatorial tendencies operate in V1—V2 combinations, which is proved by the associated X2=83.20 (significant at least at the 0.001 level with 16 degrees of freedom). Four associative tendencies are found (e—e, i—i, u—u, i—o) and there are four dissociative tendencies (e—u, i—e, i—u, u—o). Observed values of all combinations are presented in Table 15.

According to the diagonal test combinations of like vowels are significantly over-utilised while those of unlike vowels are under-utilised (X2=18.70, which is significant at the 0.001 level with 1 degree of freedom).

The position analysis (Table 52) yields only one significant result: i prefers the second syllable (X2=9.57, which is significant at the 0.01 level with 1 degree of freedom).

For C1—C2 under locational classification X2=40.46, which is significant at least at the 0.001 level with 9 degrees of freedom. The combinations F—F and M—M are dissociative, while F—M and M—F are associative. Observed values are given in Table 16.

Combinations of like consonants proved dissociative and combinations of unlike consonants proved associative. The diagonal X2 is equal to 21.81, which is significant at least at the 0.001 level with 1 degree of freedom.

According to the position test, O and M prefer the second position (X2=5.38 and 6.80, which are significant at the 0.05 and 0.01 level respectively). F prefers the first syllable (X2=32.72, which is significant at a level higher than 0.001 with 1 degree of freedom). For observed values see Table 53.

Under modal classification of consonants for C1—C2 (Table 17) X2=15.82, which is not significant at the 0.05 level with 16 degrees of freedom. This indicates a lack of combinatorial tendencies.

For C1—V1 (Table 18) X2=27.65, which is significant at the 0.01 level with 12 degrees of freedom. Two tendencies found are the association of F—e and the dissociation of O—e. For C2—V2 (Table 19) X2=21.53, which is significant at the 0.05 level with 12 - 467 degrees of freedom. Here F—a is associative and F—o is dissociative. For C1—V2 (Table 20) X2=2.29, which is not significant at the 0.05 level with 12 degrees of freedom; and for V1—C2 (Table 21) X2=12.97, which is not significant at the 0.05 level with 12 degrees of freedom.

2.4 Bi-Vocalic Morpheme Structure in Fijian (FIJ)

FIJ is the only living non-Polynesian language discussed in this paper. It has been included in order to get a convenient background for comparing morpheme structures of the Polynesian languages. At the same time FIJ indicates to what extent regularities of morpheme structure occurring in the Polynesian languages can be expected in other Austronesian languages.

The data needed have been extracted from Capell's FIJ dictionary. 32 Altogether 1710 bi-vocalic morphemic forms have been identified, which is 23.67% of the 7225 theoretically possible bi-vocalic morphemic forms in FIJ. This percentage is considerably less than that for any of the three Polynesian languages examined above.

There are 5 vowels a e i o u and 16 consonants in FIJ. The consonants are classified locationally into front (F) m b v w, middle (M) c d dr l n r s t y and back (B) ŋ k q, and they are classified modally into voiceless stops (Sn) k t, voiced stops (Sv) b d q, fricatives (F) c s v, nasals (N) m n ŋ, vibrants (V) dr r, half-vowels (H) w y and the l.

For V1—V2 combinations (Table 22) X2=280.57, which is significant at a higher level than 0.001 with 16 degrees of freedom. There are four associative tendencies (e—e, i—i, o—o, u—u) and six dissociative tendencies (e—i, e—o, i—e, i—u, o—u, u—o).

The diagonal test proves that combinations of like vowels are over-utilised while those of unlike vowels are under-utilised (X2=160.56, which is significant at a level much higher than 0.001 with 1 degree of freedom).

According to the position analysis (Table 54) i prefers the second position (X2=26.40, which is significant at least at the 0.001 level with 1 degree of freedom), while o prefers the first position (X2=17.50, which is significant at least at the 0.001 level with 1 degree of freedom).

For C1—C2 combinations under locational classification (Table 23) three dissociations (F—F, M—M, B—B) and four associations (F—M, M—F, M—B, B—M) have been found. The associated X2=163.50, which is significant at a level much higher than 0.001 with 9 degrees of freedom.

Again combinations of like consonants are under-utilised and combinations of unlike consonants are over-utilised. The diagonal X2=114.04, which is significant at a higher level than 0.001 with 1 degree of freedom.

The position analysis (Table 55) gives only one significant result; O strongly prefers the second position (X2=51.37, which is significant at a higher level than 0.001 with 1 degree of freedom).

- 468

For C1—C2 under the modal classification of consonants (Table 24) X2=64.13, which is not significant at the 0.05 level with 49 degrees of freedom. A few significant combinatorial tendencies are found, however, mainly between members of the classes l and V. It is worth mentioning that there is a partial over-lapping of modal and locational classifications as far as l and V are concerned. If these are excluded (Table 25) X2=22.54, which is not significant at the 0.05 level with 25 degrees of freedom. The only tendency found is the association of N—N.

For C1—V1 (Table 26) X2=64.49, which is significant at the 0.001 level with 12 degrees of freedom; two associations (O—o, O—u) and two dissociations (O—a, O—e) have been found here. For C2—V2 (Table 27) X2=47.84, which is significant at the 0.001 level with 12 degrees of freedom; two associations (O—u, B—a) and four dissociations (O—o, F—i, M—a, B—u) have been found. For V1—C2 (Table 29) no significant tendencies have been found and the associated X2=17.76, which is not significant at the 0.05 level with 12 degrees of freedom. No significant tendencies have been found in the case of C1—V2 (Table 28) either (X2=6.99, which is not significant at the 0.05 level with 12 degrees of freedom).

2.5 Bi-Vocalic Morpheme Structure in Proto-Polynesian (PPN)

This analysis of PPN morpheme structure is based on the preliminary draft of the Proto-Polynesian Word List compiled by Walsh and Biggs of the Department of Anthropology at the University of Auckland. This preliminary draft was made available to me during my stay in Auckland. 33 Altogether 623 bi-vocalic morphemic forms have been identified in the list, which is about 44% of the estimated total of some 1400 bi-vocalic morphemic forms of PPN.

5 vowels a e i o u and 13 consonants f h k l m n ŋ p r s t w? have been reconstructed for PPN. The consonants are classified locationally into front (F) m p w f, middle (M) l n r s t and back (B) ŋ k h?, and they are classified modally into stops (S) p t k?, fricatives (F) f s h and sonants (Sn) m n ŋ l r w.

For V1—V2 combinations (Table 30) X2=113.68, which is significant at a higher level than 0.001 with 16 degrees of freedom. This indicates strong combinatorial tendencies. Four associations (e—e, i—i, o—o, u—u) and three dissociations (e—i, i—u, u—o) have been found.

The diagonal test proves that combinations of like vowels are significantly over-utilised while those of unlike vowels are significantly under-utilised (X2=60.55, which is significant at a higher level than 0.001 with 1 degree of freedom).

The position test (Table 56) indicates that a significantly prefers the first syllable (X2=8.58, which is significant at the 0.01 level with 1 degree of freedom), while i significantly prefers the second syllable (X2=6.84, which is significant at the 0.01 level with 1 degree of freedom).

- 469

For C1—C2 under locational classification (Table 31) X2=41.16, which is significant at least at the 0.001 level with 9 degrees of freedom. Two dissociations (F—F, M—M) and two associations (F—M, B—F) have been found.

According to the diagonal test combinations of unlike consonants are over-utilised while those of like consonants are under-utilised (X2=26.77, which is significant at least at the 0.001 level with 1 degree of freedom).

The position analysis (Table 57) indicates that O significantly prefers the second position (X2=12.13, which is significant at the 0.001 level with 1 degree of freedom), and F significantly prefers the first position (X2=17.90, which is significant at a higher level than 0.001 with 1 degree of freedom).

Under modal classification for C1—C2 combinations (Table 32) X2=14.59, which is not significant at the 0.05 level with 9 degrees of freedom.

The X2 is not significant in any of the heterogeneous relations at the 0.05 level with 12 degrees of freedom. For C1—V1 (Table 33) X2=17.97; for C2—V2 (Table 34) X2=9.40; for C1—V2 (Table 35) X2=3.85; and for V1—C2 (Table 36) X2=13.94.

2.6 Bi-Vocalic Morpheme Structure in Proto-Austronesian (PAN)

An exhaustive analysis of this problem has been undertaken by Chrétien in his paper submitted to the 1965 London Conference on Linguistic Problems of the Indo-Pacific Area. 34

The results of Chrétien's work are substantially the same as those obtained in this paper. Chrétien, however, has used a somewhat different technique, which makes a full confrontation of his results with ours impossible. Therefore the author of the present paper has applied techniques used for HAW, TUA, MAO, FIJ and PPN to the corpus of data contained in Chrétien's paper. Chrétien limits his attention to forms consisting of two syllables only. He disregards forms which are clearly reduplicated monosyllables. Besides, he has ignored the nasals, considering them to be of secondary origin; consequently all morphs containing nasal accretions have been treated as if they were CVCVC.

Consonants have been classified by Chrétien in terms of place of articulation (our locational classification) and in terms of type of articulation (our modal classification). Under the former classification he distinguishes the following six classes: labial m b p v, dental n d t l, retroflex d t l, palatal n′ d′ t′ g′ k′ j, velar ŋ g k γ and laryngeal e h. Under the latter classification he has found the following groupings: semivowel v j, aspirate e h, nasal m n n′ ŋ, stop voiced b d d d′ g′ g, stop voiceless p t t t′ k′ k and continuant l l γ;.

For the purposes of this paper Chrétien's dental, retroflex and palatal consonants are classed locationally as middle consonants (M), his velars and laryngeals as back consonants (B), and his labials as front consonants (F). His semivowels, nasals and continuants are classed modally as - 470 sonants (S), and his other classes remain unchanged, i.e. aspirates (A), voiced stops (Sv) and voiceless stops (Sn).

There are only four vowels in PAN—a i ∂ u (the PPN equivalent of ∂ is o).

For V1—V2 combinations (Table 37) X2 is as great as 140.63, which is significant at a higher level than 0.001 with 9 degrees of freedom. Three associations (i—i, ∂—∂, u—u) and five dissociations (a—u, i—u, ∂—i, u—a, u—∂) have been found.

The diagonal test proves clearly that combinations of like vowels are over-utilised and combinations of unlike vowels are under-utilised (X2=60.31, which is significant at a higher level than 0.001 with 1 degree of freedom).

According to the position analysis (Table 58) ∂ significantly prefers the first position (X2=54.01, which is significant at a higher level than 0.001 with 1 degree of freedom), while u prefers the second position (X2=8.43, which is significant at the 0.01 level with 1 degree of freedom).

For C1—C2 under locational classification X2=155.18, which is significant at a higher level than 0.001 with 4 degrees of freedom. All of the three like pairs have been found to be dissociative, while F—M, M—F, M—B, and B—M have been found to be associative. For observed values see Table 38.

The diagonal test proves that combinations of like consonants are significantly under-utilised and combinations of unlike consonants are significantly over-utilised (X2=124.72, which is significant at a much higher level than 0.001 with 1 degree of freedom).

According to the position analysis (Table 59) M prefers the second position (X2=21.21, which is significant at least at the 0.001 level with 1 degree of freedom) and B prefers the first position (X2=25.21, which is significant at least at the 0.001 level with 1 degree of freedom).

For C1—C3 under locational classification (Table 39) X2=11.02, which is significant at the 0.05 level with 4 degrees of freedom. The only significant tendency found is the dissociation of F—F.

The diagonal test turns out not to be significant (X2=2.38, which is not significant at the 0.05 level with 1 degree of freedom). In this case there is no rule that like consonants are avoided and unlike consonants are attracted.

According to the position analysis (Table 60) F strongly prefers the first position (X2=140.32, which is significant at a much higher level than 0.001 with 1 degree of freedom), M prefers the first position (X2=17.86, which is significant at a higher level than 0.001 with 1 degree of freedom) and B clearly prefers the third position (X2=121.54, which is significant at a much higher level than 0.001 with 1 degree of freedom).

For C2—C3 under locational classification (Table 40) X2=105.41, which is significant at a much higher level than 0.001 with 4 degrees of - 471 freedom. The three diagonal combinations F—F, M—M, B—B are dissociative, while F—M, M—B, B—F, B—M are associative.

The diagonal test proves that combinations of like consonants are under-utilised and those of unlike consonants are over-utilised (X2=79.30, which is significant at a much higher level than 0.001 with 1 degree of freedom).

According to the position test (Table 61) F clearly prefers the second position (X2=138.29, which is significant at a much higher level than 0.001 with 1 degree of freedom), M also prefers the second position (X2=71.41, which is significant at a much higher level than 0.001 with 1 degree of freedom) and B strongly prefers the third position (X2=258.06, which is significant at a much higher level than 0.001 with 1 degree of freedom).

For C1—C2 combinations under modal classification (Table 41) X2=34.00, which is significant at the 0.001 level with 9 degrees of freedom. There are three dissociations (A—A, Sv—Sv, Sn—S) and one association (Sv—S).

For C1—C2 combinations under modal classification (Table 42) X2=19.15, which is significant at the 0.05 level with 9 degrees of freedom. The only significant tendency found is the dissociation of S—S.

For C2—C3 combinations under modal classification (Table 43) X2=23.69, which is significant at the 0.01 level with 9 degrees of freedom. Two significant combinatorial tendencies have been found—the association of Sn—Sn and the dissociation of Sv—Sv.

For C1—V1 combinations (Table 44) X2=15.05, which is significant at the 0.05 level with 6 degrees of freedom. The only significant tendency is the dissociation of F—i. For C2—V2 (Table 45) X2=30.72, which is significant at the 0.001 level with 6 degrees of freedom. There are three associations (F—a, M—i, B—∂) and two dissociations (F—i, F—∂). For V1—C2 (Table 46) X2=17.85, which is significant at the 0.01 level with 6 degrees of freedom. The only significant tendency is the dissociation of u—F. For V2—C3 (Table 47) X2=130.66, which is significant at a much higher level than 0.001 with 6 degrees of freedom. Two associations (i—B, ∂—M) and four dissociations (i—F, ∂—F, u—F, ∂—B) have been found.

3. Comparison of Results

The aim of the previous chapter was to analyse an aspect of the phonemic structure of bi-vocalic morphemic forms in four living Oceanic languages (HAW, TUA, MAO, FIJ) and in two reconstructed languages (PPN, PAN). This aspect was the frequential relations of C1, C2, V1 and V2, in their two-member combinations. In addition, frequency of occurrence of individual phonemes was examined, as correlating with two different positions within the morphemic form.

All of the languages examined are genetically interrelated and, when viewed from the historical standpoint, represent three diachronic levels: (i) PAN, (ii) PPN and (iii) HAW, TUA, MAO and FIJ (see Fig. A).

- 472

It has been found that the frequency of occurrence of a phoneme (or class of phonemes) within the morphemic form may depend to some extent upon the nature of its homogeneous partner (V1 upon V2 C1 upon C2) as well as upon the position it takes up within the morphemic form (P1 or P2). In exceptional cases the frequency of a phoneme may be influenced by its heterogeneous partner.

Two types of tendency have been found to occur in the bi-vocalic morphemic forms of all the languages examined here—associations (positive combinatorial tendencies, where the observed value attested for a combination is significantly greater than its expected value) and dissociations (negative combinatorial tendencies, where the observed value is significantly smaller than the associated expected value). As far as homogeneous relations are concerned (i.e. V1—V2 and C1—C2) combinations of like phonemes proved in general to be different from those of unlike phonemes. In the case of V1—V2, associations are found in combinations of like vowels while dissociations are always displayed by combinations of unlike vowels. The reverse is true of C1—C2 combinations; associations occur with combinations of unlike consonants while dissociations occur with combinations of like consonants. The latter phenomenon, which may be termed repulsion of like consonants, seems to be rather widespread. It has been reported for Swedish, German, Afrikaans, Italian, English and Arabic. 35 This repulsion is considerably stronger when the consonants are classified as to the place of articulation than it is when they are classified according to the manner of articulation. The explanation of this phenomenon should be looked for in the field of the physiology of speech.

A survey of all associations and dissociations, for both V1—V2 and C1—C2 relations, is given in Figures B and C. As far as vocalic relation is concerned, most dissociations are found within the same vertical level (i—e, e—i, u—o, o—u) and also within the upper horizontal level (i—u, but not u—i). Fewest dissociations are found when vowels of different levels combine (i—o, o—i, e—u, u—e and, especially, a—V and V—a). As far as consonantal relation is - 473 concerned, the most associations are between F and M and between M and B.

In Figures B and C the individual phonemes and phoneme classes are circled, and arrows denote tendencies holding between them. Associations are marked by —> and dissociations by - - ->.

As the position analysis indicates (Tables II and III), i prefers the second position in MAO, FIJ and PPN, and o prefers the first position in FIJ and PAN. More interesting results are obtained for consonants. O clearly prefers the second position in HAW, TUA, MAO, FIJ and PPN, while F prefers the first position in all instances except FIJ and PAN C1—C2.

- 474

	HAW	TUA	MAO	FIJ	PPN	PAN
a	0	0	0	0	1	0
e	0	0	0	0	0	—
i	0	0	2	2	2	0
o	0	0	0	1	0	1
u	0	0	0	0	0	2

	HAW	TUA	MAO	FIJ	PPN		PAN C1—C2 C1—C3 C2—C3
O	2	2	2	2	2	—	—	—
F	1	1	1	0	1	0	1	1
M	2	0	2	0	0	2	1	1
B	0	0	0	0	0	1	2	2

Thus far, the individual languages have been compared for the occurrence and distribution of positive and negative combinatorial tendencies. Now a comparison as to the total strength of combinatorial tendencies in all types of structural relations will be carried out. The strength of combinatorial tendencies of a given relation is expressed by the value of the associated X2. Since the value of X2 is affected by the number of morphemic forms included (N) as well as by the number of combinations (r.c), these two factors have to be eliminated in order to obtain comparable figures. Tschuprow's T Coefficient of contingency will be used for this purpose. It is given by the formula

T=□ø2/□(r—c)·(c—1)

where øØ2=X2/N and r and c are the numbers of cells in rows and columns. The greater T is, the higher is the degree of contingency. The values of T for all types of binary relations of all languages examined are presented in Table IV.

Table IV reveals that the strongest tendencies are found with the homogeneous relations (V1—V2, C1—C2). Considerably weaker tendencies are obtained for heterogeneous intrasyllabic relations (C1—V1, C2—V2), and the weakest tendencies are observed for heterogeneous intersyllabic relations (C1—V2, V1—C2). It should be added that this is the case when the consonants are classified as to the place of articulation. In Table V, the average values of T () for the three types of relations are presented.

- 475

	HAW	TUA	MAO	FIJ	PPN	PAN
V1—V2	0.13	0.12	0.13	0.20	0.21	0.16
C1—C2	0.09	0.10	0.11	0.18	0.15	0.21
C1—C3	—	—	—	—	—	0.06
C2—C3	—	—	—	—	—	0.17
C1—V1	0.04	0.07	0.08	0.10	0.09	0.06
C2—V2	0.06	0.06	0.07	0.09	0.07	0.08
C1—V2	0.03	0.03	0.02	0.03	0.04	—
V1—C2	0.03	0.04	0.05	0.05	0.08	0.06
V2—C3	—	—	—	—	—	0.17

Homogeneous Combinations	0.15
Heterogeneous Intrasyllabic Combinations	0.07
Heterogeneous Intersyllabic Combinations	0.04

The values of the coefficient of contingency T will be used for computation of distance of any two languages as d (Lx, Ly)=□Σ(Tx — Ty)2, where Lx and Ly are two different languages in which T displays the values x and y respectively. The above formula may be recognised easily as a formula for computing distance of two points in n-dimensional Euclidean space. 36 As far as the present author knows, it was first applied to linguistic problems by Altmann. 37 Distances computed for Polynesian languages (including PPN) are presented in Table VI. Table VII gives values of distance for the living Oceanic languages, and Table VIII gives distances between the living Oceanic languages and PAN. An inspection of the results indicates that the distance between any two Polynesian languages is smaller than the distance between any Polynesian language on one hand and FIJ on the other hand. The average dT (PNx, PNy)=0.04 while the average dT (FIJ, PN)=0.12. In this respect a full agreement with the historical and genetic classification of Oceanic languages exists. Interestingly enough, the Polynesian languages were found to - 476 be more similar to PAN than to FIJ. The average distance dT (PN, PAN)=0.09. FIJ is somewhat closer to PAN—the associated dT (FIJ, PAN)=0.07. Another fact confirmed by genetic classification is that MAO is closer to TUA than to HAW, and the distance between MAO and TUA is smaller than that between HAW and TUA. Strangely enough, the distances between PPN and the living Polynesian languages are greater than those between the latter and PAN. This might be explained perhaps by the fact that the corpus of data available for PPN is considerably smaller than that available for PAN.

Distances Between Polynesian Languages

TUA	MAO	PPN
0.04	0.05	0.13	HAW
	0.02	0.12	TUA
		0.10	MAO

TUA	MAO	FIJ
0.04	0.05	0.14	HAW
	0.02	0.12	TUA
		0.11	MAO

TUA	MAO	PAN	FIJ
0.04	0.05	0.10	0.14	HAW
	0.02	0.09	0.12	TUA
		0.08	0.11	MAO
			0.07	PAN

- 477

Because of the simplicity of morpheme structure in Oceanic languages, the number of all theoretically possible bi-vocalic morphemic forms can be computed as C2. V2, where C is the number of consonants plus zero and V is the number of vowels. Various distributive and combinative restrictions, to which phonemes are constituents of morpheme structures are subjected, account for the fact that the number of all theoretically possible morphemic forms is much greater than the number of actually occurring morphemic forms. As the number of theoretically possible morphemic forms P increases, the number of actually occurring morphemic forms Ob increases as well, but Ob increases at a much lower rate. This is connected closely with the essence of language, and specifically with the fact that a language with an infinite number of morphemes cannot exist. The dependence of Ob upon P is expressed here as a non-linear regression. 38 The independent variable P acquires the values 2.0 for HAW (the number of P in thousands), 3.0 for TUA and MAO, and 7.2 for FIJ. The dependent variable Ob acquires the values 1.0 for HAW, 1.2 for TUA and MAO, and 1.7 for FIJ. The relation between Ob and P is given as Ob=0.723+0.148 P—0.002 P2. The computed values coincide entirely with the observed values, Ob1 being equal to 1.0, Ob2 to 1.2, and Ob3 to 1.7. Now we are able to compute the approximate value of Ob on the basis of P only if 2.0 <P> 7.2. This is the case for PPN as reconstructed by Walsh and Biggs. 39 The number of theoretically possible bi-vocalic morphemic forms of PPN equals 4900. Thus 4.9 substitutes P in the above equation, and the estimated number of observed bi-vocalic morphemic forms in PPN is computed as ObPPN=0.723+0.148.4.9—0.002.4.92=1.4. This means that about 1,400 bi-vocalic morphemic forms are to be postulated for PPN. This figure does not differ substantially from that arrived at by means of linear interpolation. 40 Now we are able to evaluate the exhaustiveness of the reconstruction. Since some 620 bi-vocalic forms have been reconstructed from the total of 1400 estimated actually occurring forms, the exhaustiveness of the reconstruction would be as high as some 44 to 45%.

The corpus of data employed in this paper did not include borrowed morphemes. This is why it can be said that the results achieved here hold for Oceanic languages of the pre-European cultural period. An attempt will be made here to show that borrowings from European languages conform to the patterns of the recipient languages. The main source of borrowing for the four Oceanic languages used here is English (and, in the case of TUA, French). The phonemic structure of the English morpheme deviates considerably from that of Oceanic languages. This means that phonetic transformation of the borrowed elements is inevitable. The degree of phonetic modification of an adopted element is ruled by two factors. The first factor requires a more or less precise formal reproduction of the borrowed element in the recipient language. - 478 The second factor, which often interferes with the first one, requires that the borrowed elements conform to the structural pattern of the recipient language. When the adoption takes place spontaneously, which is typical of Oceanic languages, the latter factor is more important. Here the phonetic modification is far-reaching, and, especially with phonetically complex and long elements, only rough features of the model element are preserved in the recipient language, e.g. English handkerchief becomes Maori aikiha.

There are restrictions of two types: (i) inventory restrictions—a given set of phonemes that may be used in adoption is not greater than that of the phonemic system of the recipient language; (ii) distributive restrictions—the choice of the particular phonemes ought to conform to the extant combinatorial tendencies characteristic of the morpheme structure of the recipient language.

The latter restrictions are more interesting from the linguistic point of view. Let us take as an example the association of like vowels in MAO. When the English monosyllables of the type CVC are adopted in MAO, the final consonant must be avoided. This takes place through adding a vowel after the final consonant. In most cases this added vowel is identical with that of the preceding syllable, e.g. jug→ haaka, supper→ hapa, bell→ pere, fig→ piki, meat→ miiti, dish→ riihi, hall→ hooro, bull→ puru, foot→ putu, etc. This assimilative vocalic accretion takes place not only when the final consonant is concerned but also when consonantal clusters are involved as the following modifications show: umbrella→ (a) marara, candle→ kaanara, apostle→ aapotoro, clerk→ karaka, etc. This does not mean that the “auxiliary” vowel is always chosen according to the principle of assimilation. Sometimes different formations appear. Thus, e.g., i may be employed as an auxiliary vowel if the final consonant of the model element is voiceless. This is because the final -i is often devoiced in MAO speech. For example, match→ mati, map→ mapi, coat→ koti. The factor requiring the precise reproduction no doubt works here. The same holds for adoptions like steamer→ tima, where the syllabic final -∂ is substituted by -a.

The examples listed above indicate that borrowings conform to at least some of the structural imperatives which are characteristic of morpheme structure of Oceanic languages. However, further research is needed before any definite conclusions can be made concerning this problem.

4. Summary

This paper has aimed at an analysis of certain aspects of phonemic structure of bi-vocalic morphemic forms in four living Oceanic languages as well as in PPN and PAN. Use of original methodology was not the ambition of the present author. Rather he intended to apply some elementary statistical techniques that had been previously used either in linguistics or in other social sciences to the linguistic data of several Austronesian languages.

- 479

Bi-vocalic morphemic form was regarded as a system consisting of inventory and structure. Inventory was defined as a set of elements C1, V1, C2, V2 (plus C3 in PAN). Structure was interpreted as a set of the following two-member relations: (i) homogeneous relations (V1—V2, C1—C2, C1—C3, C2—C3); (ii) heterogeneous intrasyllabic relations (C1—V1, C2—V2, V2—C3); and (iii) heterogeneous intersyllabic relations (C1—V2, V1—C2, V1—C3).

One of the major tasks was to find out whether any combinatorial tendencies operate in the phonemic structure of the bi-vocalic morphemic form. It was found that the strength of combinatorial tendencies is greatest with the homogeneous relations. It is smaller with the heterogeneous intrasyllabic relations and it is smallest with the heterogenous intersyllabic relations. Significant combinatorial tendencies (either positive or negative) occur mainly with the homogeneous relations.

As far as the distribution of the significant tendencies is concerned, all the languages investigated here are very similar in this respect. It happens that a combination characterised in one language as associative or dissociative is neutral in another language, but there is no case of a combination that is associative in one language and dissociative in another or vice versa.

The following combinatorial tendencies have been discovered for homogeneous vocalic relations: (i) e—e, i—i, u—u are associative in all languages where they are present; (ii) o—o is neutral in HAW and MAO and associative elsewhere (=∂—∂ in PAN); (iii) i—o is associative in MAO and neutral elsewhere; (iv) u—o is dissociative in all languages (=u— ∂ in PAN); (v) i—u is dissociative in all languages except HAW, where it is neutral, i—e is dissociative in all languages except PPN, where it is neutral, e—i is dissociative in all languages except MAO, where it is neutral; (vi) e—u is dissociative in MAO and neutral elsewhere, e—o is neutral everywhere except in FIJ, where it is dissociative, o—u is dissociative in FIJ, a—u, e—i and u—a are dissociative in PAN and neutral elsewhere; (vii) a—a, a—e, a—i, a—o, e—a, i—a, o—a, o—e, u—e and u—i are neutral in all languages.

Consonantal homogeneous relations have been examined after the consonants had been classified in terms of the place of articulation (locational criterion) and the manner of articulation (modal criterion). Much stronger combinatorial tendencies have been found under the former classification than under the latter, e.g. in HAW Tloc=0.13, while Tmod=0.05. The following combinatorial tendencies have been found for homogeneous consonantal relations under locational classification of consonants: (i) F—F is dissociative in all languages; (ii) M—M is dissociative in all languages except HAW, where it is neutral; (iii) B—B is dissociative in FIJ and PAN and is neutral elsewhere; (iv) M—F is associative in all languages except HAW, where it is neutral, F—M is associative in MAO, FIJ, PAN and PPN and is neutral in HAW and TUA, M—B is associative in FIJ and PAN and is neutral elsewhere, B—F is associative in HAW and B—M is associative in FIJ and PAN and is neutral elsewhere; and (v) O—O, O—F, - 480 O—M, O—B, F—O, F—B, M—O and B—O are neutral in all languages in which they are present.

Under modal classification of consonants only a few significant tendencies have been found. A few occasional tendencies have been found for heterogeneous relations as well.

Values of the coefficient T associated with the particular types of structural relations have been used for computing distance between languages. It was noticed that the distances computed on the basis of the phonemic structure of the bi-vocalic morphemic form (which is the most important morphemic type in the languages involved) correspond exactly to results arrived at by means of traditional comparative linguistics. The question of possible parallels between typological and genetic classification of languages has been touched on here. When the languages have been ordered as to the greatness of the distances it has been found that there are three groupings—FIJ, PAN and Polynesian. Polynesian languages turned out to be closer to each other than any of them is to either FIJ or PAN. The distance between FIJ and Polynesian languages is greater than that between these two and PAN. It is rather interesting that MAO is closer to TUA than to HAW and that HAW is closer to TUA than to MAO.

Combinations of like phonemes proved to differ significantly from those of unlike phonemes. Combinations of like vowels are over-utilised while those of unlike vowels are under-utilised. This is another important typological feature shared obviously by most if not all Austronesian languages. It has been proved to exist in PAN, Malay, Hanuno'o, Bare'e, Angkola-Batak and Sundanese 41 as well as in HAW, TUA, MAO, FIJ and PPN. This ‘vocalic harmony’ 42 has been weakened in FIJ and in the Polynesian languages, which is in perfect agreement with the overall simplification that has taken place east of Indonesia.

The situation is reversed with the consonantal combinations. Combinations of like consonants are under-utilised while those of unlike consonants are over-utilised. This is true of all languages investigated here. Unfortunately, data from Indonesian languages are not available yet. Interestingly enough, a similar situation exists in many other languages. 43 It may be assumed therefore to be a fairly widespread phenomenon. Additional evidence is needed, however, to corroborate the hypothesis of the universality of this phenomenon.

It has been found that there is a correlation between the position in which a phoneme occurs within the morpheme and its frequency. Thus, e.g., front consonants significantly prefer the first position while zero prefers the second position.

Finally, it was pointed out that structural regularities found to hold for genetically inherited morphemic forms are, to some extent at least, obligatory for borrowed morphemes as well.

- 481

V1\V2	a	e	i	o	u	Nj
a	72	50	47	49	48	266
e	50	55	22	35	33	195
i	43	19	55	42	25	184
o	45	40	43	49	36	213
u	45	34	33	11	48	171
Ni	255	198	200	186	190	1029

C1\C2	O	F	M	B	Nj
O	22	23	35	51	131
F	49	17	88	101	255
M	44	40	58	101	243
B	72	79	113	136	400
Ni	187	159	294	389	1029

C1\C2	O	S	N	F	I	Nj
O	22	37	28	27	17	131
S	68	126	69	57	59	379
N	41	70	44	31	32	218
F	35	54	33	24	35	181
I	21	44	19	19	17	120
Ni	187	331	193	158	160	1029

C1\V1	a	e	i	o	u	Nj
O	36	19	24	27	25	131
F	71	53	45	50	36	255
M	61	48	44	51	39	243
B	98	75	71	85	71	400
Ni	266	195	184	213	171	1029

- 482

C2\V2	a	e	i	o	u	Nj
O	39	35	39	36	38	187
F	55	29	21	26	28	159
M	67	58	55	59	55	294
B	94	76	85	65	69	389
Ni	255	198	200	186	190	1029

C1 V2	a	e	i	o	u	Nj
O	33	27	23	26	22	131
F	62	44	57	45	47	255
M	60	47	43	44	49	243
B	100	80	77	71	72	400
Ni	255	198	200	186	190	1029

V1\C2	O	F	M	B	Nj
a	45	45	78	98	266
e	40	30	50	75	195
i	33	27	52	72	184
o	38	28	68	79	213
u	31	29	46	65	171
Ni	187	159	294	389	1029

V1\V2	a	e	i	o	u	Nj
a	92	62	72	62	67	355
e	54	55	31	37	35	212
i	59	22	62	53	25	221
o	51	44	50	60	30	235
u	54	36	37	12	49	188
Ni	310	219	252	224	206	1211

- 483

C1\C2	O	F	M	B	Nj
O	19	37	55	47	158
F	65	33	128	105	331
M	65	80	91	138	374
B	62	59	134	93	348
Ni	211	209	408	383	1211

C1\C2	O	S	N	F	r	Nj
O	19	50	37	30	22	158
S	73	137	92	69	59	430
N	50	66	64	37	34	251
F	46	67	49	30	41	233
r	23	43	29	24	20	139
Ni	211	363	271	190	176	1211

C1\V1	a	e	i	o	u	Nj
O	44	18	31	33	32	158
F	116	66	59	52	38	331
M	98	66	70	79	61	374
B	97	62	61	71	57	348
Ni	355	212	221	235	188	1211

C2\V2	a	e	i	o	u	Nj
O	45	40	44	38	44	211
F	70	34	38	33	34	209
M	96	76	86	77	73	408
B	99	69	84	76	55	383
Ni	310	219	252	224	206	1211

- 484

C1\V2	a	e	i	o	u	Nj
O	38	30	37	29	24	158
F	85	57	69	62	58	331
M	104	66	75	67	62	374
B	83	66	71	66	62	348
Ni	310	219	252	224	206	1211

V1\C2	O	F	M	B	Nj
a	54	65	115	121	355
e	41	31	71	69	212
i	40	39	77	65	221
o	43	37	77	78	235
u	33	37	68	50	188
Ni	211	209	408	383	1211

V1\V2	a	e	i	o	u	Nj
a	93	72	75	60	70	370
e	64	66	45	51	30	256
i	54	21	56	48	22	201
o	56	51	54	54	35	250
u	55	31	38	4	53	181
Ni	322	241	268	217	210	1258

C1 C2	O	F	M	B	Nj
O	19	36	54	49	158
F	60	24	152	110	346
M	60	84	108	124	376
B	63	67	137	111	378
Ni	202	211	451	394	1258

- 485

C1\C2	O	S	N	F	r	Nj
O	19	53	33	33	20	158
S	66	133	98	72	60	429
N	51	83	62	50	42	288
F	44	67	65	27	40	243
r	22	48	26	25	19	140
Ni	202	384	284	207	181	1258

C1 V1	a	e	i	o	u	Nj
O	44	21	30	33	30	158
F	114	90	50	54	38	346
M	109	70	69	73	55	376
B	103	75	52	90	58	378
Ni	370	256	201	250	181	1258

C2 V2	a	e	i	o	u	Nj
O	46	34	47	35	40	202
F	73	42	44	21	31	211
M	103	93	91	84	80	451
B	100	72	86	77	59	394
Ni	322	241	268	217	210	1258

C1\V2	a	e	i	o	u	Nj
O	40	28	33	29	28	158
F	85	64	76	64	57	346
M	101	71	81	63	60	376
B	96	78	78	61	65	378
N1	322	241	268	217	210	1258

- 486

V1\C2	O	F	M	B	Nj
a	54	80	120	116	370
e	43	32	98	83	256
i	34	34	74	59	201
o	42	34	91	83	250
u	29	31	68	53	181
Ni	202	211	451	394	1258

V1\V2	a	e	i	o	u	Nj
a	168	68	118	78	93	525
e	67	90	38	16	50	261
i	69	14	90	51	18	242
o	95	58	73	123	45	394
u	80	41	50	17	100	288
Ni	479	271	369	285	306	1710

C1\C2	O	F	M	B	Nj
O	7	18	51	13	89
F	49	34	278	56	417
M	116	261	334	199	910
B	32	51	190	21	294
Ni	204	364	853	289	1710

- 487

C1\C2	O	Sn	Sv	F	N	I	V	H	Nj
O	7	13	11	23	12	10	12	1	89
Sn	30	41	52	57	50	30	43	7	310
Sv	38	44	29	54	44	43	33	9	294
F	51	71	57	81	53	45	49	10	417
N	27	33	25	36	45	26	28	5	225
I	13	23	23	32	20	10	3	6	130
V	27	26	21	28	19	6	9	5	141
H	11	17	13	21	11	13	15	3	104
Ni	204	268	231	332	254	183	192	46	1710

C1\C2	O	Sn	Sv	F	N	H	Nj
O	7	13	11	23	12	1	67
Sn	30	41	52	57	50	7	237
Sv	38	44	29	54	44	9	218
F	51	71	57	81	53	10	323
N	27	33	25	36	45	5	171
H	11	17	13	21	11	3	76
Ni	164	219	187	272	215	35	1092

C1\V1	a	e	i	o	u	Nj
O	4	5	15	34	31	89
F	133	73	50	85	76	417
M	281	144	133	216	136	910
B	107	39	44	59	45	294
Ni	525	261	242	394	288	1710

- 488

C2\V2	a	e	i	o	u	Nj
O	53	23	54	21	53	204
F	117	64	60	51	72	364
M	210	136	206	157	144	853
B	99	48	49	56	37	289
Ni	479	271	369	285	306	1710

C1\V2	a	e	i	o	u	Nj
O	22	14	20	20	13	89
F	116	68	97	62	74	417
M	268	139	187	152	164	910
B	73	50	65	51	55	294
Ni	479	271	369	285	306	1710

V1\C2	O	F	M	B	Nj
a	49	129	263	84	525
e	36	64	118	43	261
i	34	41	130	37	242
o	48	80	190	76	394
u	37	50	152	49	288
Ni	204	364	853	289	1710

V1\V2	a	e	i	o	u	Nj
a	74	32	41	35	43	225
e	19	30	8	11	10	78
i	21	8	35	24	9	97
o	25	16	29	35	16	121
u	28	11	24	0	39	102
Ni	167	97	137	105	117	623

- 489

C1\C2	O	F	M	B	Nj
O	5	12	25	19	61
F	38	12	91	49	190
M	43	56	69	67	235
B	20	36	55	26	137
Ni	106	116	240	161	623

C1\C2	O	S	F	Sn	Nj
O	5	19	14	23	61
S	43	96	29	94	262
F	19	30	12	48	109
Sn	39	54	25	73	191
N1	106	199	80	238	623

C1 V1	a	e	i	o	u	Nj
O	23	3	16	12	7	61
F	80	21	25	35	29	190
M	73	36	41	46	39	235
B	49	18	15	28	27	137
N1	225	78	97	121	102	623

C2 V2	a	e	i	o	u	Nj
O	31	15	24	17	19	106
F	28	21	21	18	28	116
M	57	35	56	44	48	240
B	51	26	36	26	22	161
N1	167	97	137	105	117	623

- 490

C1\V2	a	e	i	o	u	Nj
O	16	11	14	10	10	61
F	54	30	45	32	29	190
M	61	36	49	38	51	235
B	36	20	29	25	27	137
Ni	167	97	137	105	117	623

V1\C2	O	F	M	B	Nj
a	29	49	87	60	225
e	16	17	23	22	78
i	21	20	35	21	97
o	24	14	51	32	121
u	16	16	44	26	102
Ni	106	116	240	161	623

V1\V2	a	i	∂	u	Nj
a	385	152	57	217	811
i	155	75	24	77	331
∂	128	41	70	92	331
u	172	65	16	205	458
Ni	840	333	167	591	1931

C1\C2	F	M	B	Nj
F	31	291	113	435
M	240	267	199	706
B	135	332	87	554
Ni	406	890	399	1695

- 491

C1\C3	F	M	B	Nj
F	26	133	278	437
M	61	226	409	696
B	63	188	313	564
Ni	150	547	1000	1697

C2\C3	F	M	B	Nj
F	10	176	240	426
M	86	208	593	887
B	49	181	173	403
Ni	145	565	1006	1716

C1\C2	A	Sv	Sn	S	Nj
A	3	75	94	105	277
Sv	20	76	135	206	437
Sn	30	180	258	232	700
S	14	65	91	111	281
Ni	67	396	578	654	1695

C1\C3	A	Sv	Sn	S	Nj
A	68	11	84	115	278
Sv	130	5	118	176	429
Sn	186	20	213	292	711
S	92	8	93	86	279
Ni	476	44	508	669	1697

- 492

C2\C3	A	Sv	Sn	S	Nj
A	14	4	17	33	68
Sv	119	4	102	172	397
Sn	153	16	205	212	586
S	193	23	192	258	666
Ni	479	47	516	675	1717

C1\V1	a	i	∂	u	Nj
F	209	57	83	124	473
M	309	139	131	153	732
B	237	108	99	152	596
N1	755	304	313	429	1801

C2\V2	a	i	∂	u	Nj
F	225	60	23	141	449
M	387	191	85	266	929
B	178	63	51	142	434
N1	790	314	159	549	1812

V1\C2	F	M	B	Nj
a	209	371	195	775
i	70	164	67	301
∂	90	151	65	306
u	80	243	107	430
Ni	449	929	434	1812

- 493

V2\C3	F	M	B	Nj
a	86	239	461	786
i	9	88	211	308
∂	41	74	39	154
u	27	180	358	565
Ni	163	581	1069	1813

	a	e	i	o	u
Position 1	266	195	184	213	171
Position 2	255	198	200	186	190
Totals	521	393	384	399	361

	O	F	M	B
Position 1	131	255	243	400
Position 2	187	159	294	389
Totals	318	414	537	789

	a	e	i	o	u
Position 1	355	212	221	235	188
Position 2	310	219	252	224	206
Totals	665	431	473	459	394

- 494

	O	F	M	B
Position 1	158	331	374	348
Position 2	211	209	408	383
Totals	369	540	782	731

	a	e	i	o	u
Position 1	370	256	201	250	181
Position 2	322	241	268	217	210
Totals	692	497	469	467	391

	O	F	M	B
Position 1	158	346	376	378
Position 2	202	211	451	394
Totals	360	557	827	772

	a	e	i	o	u
Position 1	525	261	242	394	288
Position 2	479	271	369	285	306
Totals	1004	532	611	679	594

- 495

	O	F	M	B
Position 1	89	417	910	294
Position 2	204	364	853	289
Totals	293	781	1763	583

	a	e	i	o	u
Position 1	225	78	97	121	102
Position 2	167	97	137	105	117
Totals	392	175	234	226	219

	O	F	M	B
Position 1	61	190	235	137
Position 2	106	116	240	161
Totals	167	306	475	298

	a	i	∂	u
Position 1	811	331	331	458
Position 2	840	333	167	591
Totals	1651	664	498	1049

- 496

	F	M	B
Position 1	435	706	554
Position 2	406	890	399
Totals	841	1596	953

	F	M	B
Position 1	437	696	564
Position 3	150	547	1000
Totals	587	1243	1564

	F	M	B
Position 2	426	887	403
Position 3	145	565	1006
Totals	571	1452	1409

• ALTMANN, G., 1964. Kvantitativne stôdie z indonézistiky. Dissertation, Bratislava. (Typescript).
• — —, 1966. Review of Some Aspects of the Relative Efficiencies of Indian Languages by B. S. Ramakrishna, K. K. Nair, V. N. Chiplunkar, B. S. Atal, V. Ramachandran and R. Subramanian. Jazykovedny casopis, 17:109-111.

• BUSACKER, Robert G. and Thomas L. SAATY, 1965. Finite Graphs and Networks. New York, MCGraw-Hill. (International Series in Pure and Applied Mathematics.)

• CAPELL, A., 1941. A New Fijian Dictionary. Sydney, Australasian Medical Publishing Co.
• CHRETIEN, C. Douglas, 1965. “The Statistical Structure of the Proto-Austronesian Morph.” pp. 243-270 in Indo-Pacific Linguistic Studies, Part I, G. B. Milner and Eugénie J. A. Henderson (eds.), Amsterdam, North-Holland Publishing Co.

- 497

• DIDERICHSEN, P., 1953. “Bidrag til en analyse af det danske skriftsprogs struktur.” Selskab for Nordisk Filologi, Arsberetning 1951-1952, pp. 7-22.

• GREENBERG, Joseph H., 1950. “The Patterning of Root Morphemes in Semitic.” Word, 6 : 162-181.

• HARARY, Frank and Herbert H. PAPER, 1957. “Toward a General Calculus of Phoneme Distribution.” Language, 33 : 143-169.
• HILL, Archibald A., 1958. Introduction to linguistic structures; from sound to sentence in English. New York, Harcourt, Brace.

• JAKOBSON, Roman, C. Gunnar M. FANT and Morris HALLE, 1952. Preliminaries to speech analysis; the distinctive features and their correlates. Massachusetts Institute of Technology, Acoustics Laboratory, Technical Report No. 13.

• KRAMSKY, J., 1964. “A Quantitative Phonemic Analysis of Italian Mono-, Di-and Trisyllabic Words.” Travaux linguistiques de Prague, 1 : 129-143.
• KRUPA, V., 1967. “Dissociations of Like Consonants.” Asian and African Studies, 3.

• MALMBERG, Bertil, 1963. Structural Linguistics and Human Communication. Berlin, Springer.
• MATHESIUS, V., 1929. “La structure phonologique du lexique de tchèque moderne. TCLP, 4 1: 67-84.
• — —, 1931. “Zum Problem der Belastungs- und Kombinationsfähigkeit der Phoneme.” TCLP, 4 : 148-152.

• NORKINA, S. B. (ed.), 1963. Elementy vycislitelnoj matematiki. Moscow, Vysshaya shkola.

• ODENDAL, F. F., 1962. Die struktuur van die afrikaanse wortelmorfeem. Cape Town, H.A.U.M.

• PUKUI, Mary Kawena and Samuel H. ELBERT, 1957. Hawaiian-English Dictionary. Honolulu, University of Hawaii Press.

• SIEGEL, S., 1956. Nonparametric Statistics for the Behavioral Sciences. New York, McGraw-Hill.
• SIGURD, B., 1958. “Tendencies in the Combination of Prevocal and Postvocal Consonants in Swedish Monosyllables.” Studia Linguistica, 12 : 27-51.
• — —, 1965. Phonotactic Structures in Swedish. Lund,—.
• SPANG-HANSSEN, H., 1953. “Analyse ved hjœlp af kombinations-diagrammer.” Selskab for Nordisk Filologi, Arsberetning 1951-1952, pp. 22-26.
• STEEL, R. G. D. and J. H. TORRIE, 1960. Principles and Procedures of Statistics. New York, McGraw-Hill.
• STIMSON, J. Frank with Donald Stanley MARSHALL, 1964. A Dictionary of Some Tuamotuan Dialects of the Polynesian Language. Massachusetts, Peabody Museum of Salem, and The Hague, Hot Koninklijk Instituut voor Taal-, Land- en Volkenkunde.

• TRNKA, B., 1935. A Phonological Analysis of Present-Day Standard English. Prague, Charles University.
• — —, 1936. “General Laws of Phonemic Combinations.” TCLP, 6 : 57-62.
• TRUBETZKOY, N. S., 1939. “Grundzüge der Phonologie.” TCLP, 7.
• TWADDELL, W. F., 1939-1941. “Combinations of Consonants in Stressed Syllables in German.” Acta Linguistica, 1 : 189-199; 2 : 31-50.

• UHLENBECK, E. M., 1950. “The Structure of the Javanese Morpheme.” Lingua, 2 : 239-270.

• VOGT, H., 1942. “The Structure of the Norwegian Monosyllables.” Norsk Tidsskrift før Sprogvidenskap, 12 : 5-29.

• WALSH, D. S. and Bruce BIGGS, 1966. Preliminary draft of [Proto-Polynesian Word List I. Te Reo Monographs. Auckland, Linguistic Society of New Zealand, 1966].

• WILLIAMS, Herbert W., 1957. A Dictionary of the Maori Language. 6th ed., Wellington, Government Printer.