In the paper Zur allgemeinen Theorie der phonologischen Vokalsysteme published in the first volume of Travaux du Cercle Linguistique de Prague (pp. 39, etc.), Professor N. Trubetzkoy formulated phonological laws which apply to all linguistic systems. According to these general laws the phonological polytony necessarily involves the presence of the phonological quantity of vowels, which presupposes the existence of the timbre oppositions of vowel phonemes, based on the fundamental vocalic opposition of sonority. These laws stating the hierarchy of phonological factors in languages may be reduced, as Professor Jakobson1 shows in one of his treatises, to either of these logical formulas: If there is a, b also exists. If a exists, b is missing. All linguistic laws that cannot stand this test, e. g. M. Grammont’s laws of dissimilation, are therefore only of a limited validity.
Trubetzkoy’s laws refer to phonological relations which are abstracted from the given linguistic systems. Now the question arises if the combinations of phonemes in the same morpheme are also regulated by rules of general applicability or not. This question may be answered in the affirmative. In examining the combination and iteration of English phonemes in monomorphemic words consisting of not more than two syllables,2 I have tried to give, according to the method suggested by Trubetzkoy, the special rules restricting phonemic combinations in English. One of those rules, by which is stated that phonemes differentiated by a mark of correlation never combine in the same morpheme, applies to all languages of which I have some phonological knowledge, and may be regarded — at least provisionally — as a general law admitting of no exceptions. Thus we do not find such consonantal combinations as [pf, fp, tϑ, ϑt] and [pb, td, kg], etc. in those languages where the correlations of plosion or voice exist, e. g. in Anglo-Saxon and Modern English. Similarly the combinations [ph - p, p3 - ph, th,- t, t - th], etc. do not occur in Sanscrit and Old Greek,3 where aspiration is a correlative mark. In German, Czech, Magyar and other languages in which the quantity of vowels is phonological, the combinations of two vowels differentiated only by the correlative mark of quantity (e. g. [ii:], [uu:]i etc.) are non-existent. The same is true in regard of the combinations of stressed and unstressed vowels in Russian. In Modern French the nasalized vowels are never combined with the corresponding oral ones. Nor can palatalization, which is an important correlative mark of consonants in Russian, be the only mark of differentiation between two neighbouring phonemes and consequently such combinations as [tt’, dd’, t’t, nň, ňn], etc. do not occur in this language. The assumption of this and another complementary correlation, which may be called post-alveolar, for Czech (cf. [t : t’, d : d’, n : ň] and [s : š, z : ž, c : č, r : ř) would also account for the non-occurrence of the monomorphemic combinations tt’, t’t, ňn, nň sš, šs, cč, čc, čc, cč etc.
This rule (which may be expressed by the logical formula: if there is £, it must be neither followed nor preceded by p1 in the same morpheme) may be called the law of the minimal phonological contrast. The combinations of the marked phoneme with the corresponding unmarked one are below the limit of the minimal contrast unless they occur at the suture of two morphemes where they are tolerated (cf. bathed [beiδd] in English). If however the morphological suture is shifted from its original position or entirely disappears (cf. OE -9æt Se >Me flatte) we see that the law of the minimal contrast comes into force automatically through the assimilation (or dissimilation) of both sounds. On the other hand, the existence of the combinations of two phonemes differentiated only by one mark of correlation, signalizes them as members4 of two different morphemes (cf. let -te, vifi-te in Czech).
If this is true, the law of the minimal phonological contrast is an important negative test of the correlative character of phonemes. According to Trubetzkoy’s definition given in Travaux, IV, p. 97, the phonological correlation is constituted by at least two pairs of phonemes not connected by another correlative mark. Thus the plosion correlation in English is guaranteed by the fact that there are at least two pairs of phonemes [p/f] and [?/?] which are not otherwise correlated to each other. On the same degree of relationship, therefore, we could place the nasal opposition of m : £ (or b), n : _t (d), and էլ : k (g) in English, because none of these pairs has another correlative mark in common. If we put however both series of phonemes to our test of the minimal phonological contrast, we may see that the existence of such combinations as mp(mb), nt(nd), r|k(T]g) shows a different degree of relationship in the latter series, the phonological difference between m and £<b), n and_t(d), rj and k(g) being so wide as to allow the members of the same pair of phonological opposition to combine with each other. It is necessary, therefore, to distinguish this kind of phonemic relationship, which may be called parallelism, from correlation which represents such a close affinity that it deprives the members of the same pair of the capacity of being contrasted, as individual phonemes, in one monomorphemic combination. Phonemic groups consisting of parallel phonemes constitute therefore the minimal contrast occurring in languages, whereas the maximal degree is represented by the combinations of vowel + consonant. In Teutonic, Slavonic and other languages intermediary degrees exist in addition to the minimal and maximal contrast (cf. such combinations as տյջ, st, £k, ks, sl, £r, jai, tr, U, £t, db, kt in Czech), which are totally missing from others, e. g. the Bantu languages.
The observation of the combinative capacity of phonemes suggests the classification of so-called disjunct phoneme s-which have hitherto defied phonological analysis. At the present stage of phonological investigation, disjunct phonemes are defined to be all those that do not take part in correlations, although it is not doubted that the phonological relationship between phonemes [p, _t, k] is different from e. g. [r ] : [j]. It is true that even the former phonemes are neither parallel nor correlative to each other, but the process of their articulation is analogical and takes place symmetrically on the common axis, so that it results in an acoustic similarity. With regard to this fact we are entitled to speak of symmetrical phonemes on the one hand, and of purely disjunct on the other. The phonological symmetry is constituted by the existence of at least three disjunct (i. e. uncorrelated and unparalleled) phonemes of analogical articulation which may be brought into the proportion p : ? = ? : ?. Thus the phonemes p, ? and ? are symmetrical both in English and Russian, whereas the nasals, which form the series of phonemes m, n, ij in English and m/m, n/n in Russian, are symmetrical only in the former language, m and n being correlated to m and n, respectively, in the latter. It goes without saying that symmetrical phonemes may take part either in parallelisms or correlations, so that symmetry represents a basic relationship underlying those of parallelism and correlation. E. g. in Czech two of the series of symmetrical phonemes £, v,5 _t and ? are the members of the nasal parallelism (cf. p//m, _t//n) and voice correlation (cf. p/b, _t/d), while the correlation of palatalization refers to the phoneme ? and its corresopnding d and n (cf. է/է՛, d/d1, n/n). The phonemes v and ? are outside any correlation or parallelism in native words, [f] and [g] occur ing as phonemes only in words of foreign, or onomatopoeic, origin. It is the symmetry of phonemes which helps to account for the fact that the Czech language found it easy enough to adopt these two sounds as phonemes, whereas the phonemic function of rj could not be transferred into Czech on account of the non-correlative character of nasals in this linguistic system.6
Let us return to the main subject of this paper, the combination of phonemes. It may be assumed as a working hypothesis that the combinations of phonemes7 as well as their position in words, are regulated by strict special rules, the cogency of which is no less than that of other phonological factors. Unlike some phonological laws (e. g. that of voice neutralization in Czech) these rules, however, do not extend to foreign words which may contain monomorphemic combinations not occurring in native ones, at least not in the same position. E. g. in English the monomorphemic combinations of symmetrical phonemes pt and kt are found only in words of foreign, especially Latin, origin, but their existence in the language of adoption is undoubtedly due to the fact that the dimorphemic combinations gt, kt, bd, gd, etc. (e. g. in dipped [dipt], licked [likt], sobbed [sobd], begged [begd]) occur in the same position as do the monomorphemic £t and kt in foreign words (e. g. sept, sect). If there were not the native dimorphemic combinations of this kind in English, we should probably expect that English speakers would likely pronounce sept, sect, etc. always without_t, simplifying the consonantal combinations in a similar way as they actually do in the positions which are without any analogy in native words (cf. ptarmigan [ta :migan], mnemonic [ni’monik]). The adoption of unusual combinations (or even foreign simple phonemes) is made possible, as we may see, by the actualization of the potential functional elements of the language, which might not otherwise be utilized. Thus the monomorphemic character of the combinations £t, kt and mn, which is the foreign feature of the words mentioned, is within the range of the possibilities of the English linguistic system where new monomorphemic combinations may always come into existence by the loss of morphological suture. Similarly the functional value of the Czech variant [j* ] in words of foreign origin and character is in conformity with the voice correlation of other symmetrical phonemes, whereas the neutralization of this correlation is necessarily without exception in both native and foreign words.
The comparison of monomorphemic combinations in various languages entitles us to risk formulating the general tendencies of consonantal combinations as follows:
1. The existence of phonologically parallel combinations (e. g. mb, nd, r|g in English) presupposes the presence of the maximal degree of phonemic contrast, e. e. that of vowel + consonant, or consonant + vowel.
2. The combinations of symmetrical phonemes (e. g._£t, kt, mn in Czech) presuppose the existence of the combinations of disjunct phonemes (e. g. ££, pl, £s, pl, br, bl, etc. in Czech).
3. The occurrence of consonantal combinations at the beginning, or end, of words involves the possibility of their use between vowels, but not vice verse (cf. pl, kl, kw as against mp, mb, ljp, in English).
A few words may be said of the question relative to the number of phonemes of which the word consists. Although languages differ a good deal in this respect and interesting differences may be also stated in regard to the various grammatical classes of words, one feature probably common to all languages may be stated, viz. that only vowel phonemes can constitute independent words by themselves, without any combination with other phonemes (cf. E. eye [ai], owe [ou], Czech a, i). This rule is so general that it seems valueless for phonological analysis, but it is not entirely without importance, because it reveals the specific character of vowels as against consonants which can form independent words only in combination with vowels or other consonants (cf. trn, pin, prst in Czech).
1. R. Jakobson, Remarques sur l’évolution phonologique du russe. Travaux II, p. 17.
2. _A Phonological Analysis of Present-day Standard English (p. 30). Prague Studies in English, V (1935). As to phonological terminology, s. pp. 176, etc.
3. P—ph in such Greek words as Sappho may be accounted for as the geminated, or long, phoneme ph.
4. Cf. R. Jakobson’s article Zur Struktur des russischen Verbums in Charisteria G. Mathesio oblata (Prague, 1932), pp. 74, etc.
5. [v] may be regarded as a voiced labio-dental explosive in Czech before voiced consonants excepting [b] and [m],
6. The assumption of parallelism and symmetry as special cases of phonological opposition leads us to the conclusion that the sonority oppositions of vowels (Offnungsgradgegensätze in Trubetzkoy’s terminology) may be also interpreted as a series of symmetrical phonemes (e. g. _i — e — æ and u — ? — ?>
This interpretation, based on Trubetzkoy’s theory of consonantal phonemes (cf. Travaux, IV, pp. 96, etc.), tries to remove an inconsistency in his conception of phonological correlation. Whereas he assumes in that article that the correlative in English). The timbre oppositions (Ëigentogegensätze) are, on the other hand, either parallelisms (cf. _i//u, e//o, æ//t> in English; _i//u, e//o, in Czech) or form the groups ofphonemes interlinked Бу two correlations, if there are at least three members of these oppositions. E. g. the German system of short vowel phonemes (i — u — u, e — ö — ?, a) involves the labial and velar correlations~(i : u7 էլ : <է, u — u, o’ : o) forming two triangular groups : series consists of the pairs of phonemes differentiated by a mark of correlation, his theory of vowel phonemes (cf. Travaux I, pp. 33, etc.) involves the hypothetical assumption that the triads of vocalic phonemes of the same correlation may exist in addition to correlative pairs (cf. i : u : _u, e : ö : _o in German). What would then be the basic phonemes TArchiphonemen) according to his theory? — The analysis of the vowel oppositions given here is in agreement with the fact that the phonemic combinations iu, iii, ùu, uii, eö, öe, etc. do not exist, whereas ui, iu, eo, etc. may~õccur, in the same language.
7. The importance of phonemic combinations for the structural analysis and classification of languages was emphasized by V. Mathesius in his paper Ia structure phonologique du lexique du tchèque moderne, published in Travaux, I, pp] 57, etc.
*From Travaux du Cercle Linguistique de Prague, VI: 57-62 (1936).