Phonetica 33 (1976), p. 81-92

=====================================

Systems of Ultimate Phonological Units

V. Y. Plotkin

Department of English, University of Kishinev

Abstract. The paper upholds the existence of a phonological unit smaller than the phoneme, serving as the component of the latter. This is the linguistic correlate of articulatory movements together with their auditory reflexes, i.e. Baudouin de Courtenay’s kinakeme. The acceptance of the kinakeme as the ultimate phonological unit makes the notion of distinctive phonemic features redundant in linguistic theory. The kinakemes of any given language form a paradigmatic system and combine syntagmatically in phonemes.

The emergence of phonology as a branch of linguistics naturally resulted from the discovery of the phoneme as a linguistic entity, and the latter has always been the pivotal figure in phonological studies. It is traditional to treat it as the indivisible, minimal phonological unit of language. However, there is also considerable support for the thesis that the phoneme is not the smallest linguistic unit, that it is composed of still smaller phonological entities. These are commonly described as phonemic distinctive features. The thesis was explicitly propounded in 1938 by two prominent scholars from Prague in their papers prepared for the 3rd International Congress of Phonetic Sciences: ‘Les composants phonologiques les plus petits ne sont pas les phonèmes, mais les traits distinctifs de ceux-ci’ [Trnka, 1938, p. 270; quoted from Vachek, 1960, p. 32]. ‘Nous identifions les phonèmes d’une langue donnée en les décomposant en leurs caractères phonologiques constitutifs’ [Jakobson, 1962, p. 272]. In his paper ‘On the identification of phonemic entities’, written in 1949, Jakobson [1962] took a further step in the elaboration of the thesis: ‘In dissociating the phoneme into distinctive features we isolate the ultimate linguistic constituents charged with semiotic value’ [p. 422]. ‘Only in resolving the phonemes into their constituents and in identifying the ultimate entities obtained does phonemics arrive at its basic concept’ [p. 425].

And yet today, more than a quarter of a century later, the status of the ultimate phonological unit remains dubious, for it has failed to gain universal recognition among phonologists, although, on the other hand, the notion does not seem to have met any significant resistance. We cannot but agree with Jakobson’s [1962, p. 635] remark in his ‘Retrospect’ that the idea of the phoneme as the ultimate, indivisible unit ‘obstinately survives even to the present’. This hesitancy may perhaps be explained by the fact that essential corollaries have not been deduced from the thesis about the divisibility of the phoneme into ultimate phonological entities. And it was none other than Jakobson, the chief protagonist of this thesis, who played a significant part in blocking the way to two indispensable steps in this direction.

The first corollary may appear to be a technicality, since it concerns the term for the designation of the ultimate entity. While forcefully defending the existence of units smaller than the phoneme, Jakobson never doubted the wisdom of describing them as features of the latter. But it is logically inconsistent to describe a linguistic unit as a mere feature of another unit. Things consist of components, not features. A stone consists of its molecules, into which it can be divided, but it does not consist of and cannot be divided into its size, weight or colour [e.g. Solntsev, 1971, p. 197]. If the ultimate phonological entity is to be firmly established in linguistic theory, it needs a better term than ‘distinctive feature’. ‘A distinctive feature is ... a unit of the message ensemble rather than a property of the signal ensemble. The term “distinction” or “minimal category” would have been more appropriate and might have led to less confusion concerning their nature and use’ [Fant, 1967, p. 51].

But discarding the word ‘feature’ is not enough. It has recently been shown that the system of language does not and cannot contain any special distinctive unit, that units at every level of the system serve to distinguish the larger units whose components they are. ‘Es gibt ... keine besondere Kategorie von sprachlichen Unterscheidungseinheiten ... Die diakritische Funktion kann grundsätzlich durch jeden Teil der betreffenden Einheit ausgeführt werden’ [Grucza, 1970, p. 106]. It follows that the required new term should not imply any reference to distinction as the specific function of the unit in question. Therefore the term ‘mérisme’ meaning ‘delimitation’ in Greek [Benveniste, 1964, p. 267; 1966, p. 121] is to be declined. Neither are the terms ‘subphoneme’ [Panov, 1967, p. 113] and ‘phononeme’ [Grucza, 1970, pp. 77 ff.] acceptable, since the ultimate phonological unit is not in itself a unit of sound, and this makes the root ‘phon’ undesirable.

The phoneme retains its status of the minimal unit of sound in the language system. Its indivisibility should be qualified as inability to be broken up into smaller units of sound. As for the ultimate phonological unit, it is an instrument for the linguistic structuring of extralinguistic substance which might be called prephonic rather than phonic. This is the mechanism of articulatory and auditory innervations underlying speech and its perception [e.g. Jakobson, 1962, p. 420]. It seems highly appropriate to revive a term coined 60 years ago by Baudouin de Courtenay, who used the terms ‘kinema’ and ‘akousma’ for the psychic images of articulatory movements and their auditory counterparts; both terms were then blended into ‘kinakeme’ to designate the bilateral psychophonic unit [Baudouin de Courtenay, 1963, pp. 199, 290].

Jakobson et al. [1955] and Jakobson and Halle [1956] have led the way in the quest for a universal inventory of phonemic distinctive features. But if the latter are recognized as ultimate phonological units, the very idea of such an inventory must be rejected. All systems of language units known in linguistics are language-specific, and we have no reason to make an exception for kinakemes. Thus, the other corollary to the thesis about the divisibility of phonemes into ultimate phonological units is the necessity to discover in each language its own peculiar system of kinakemes. It is undeniable that the pioneering research by jakobson and his colleagues into the distinctive features of a great variety of languages has been extremely useful and stimulating, bringing to light a vast amount of valuable data on how universal anthropophonic substance is organized and utilized in languages. But its proclaimed aim of establishing a universal list of distinctive features leads phonology away from what is feasible and necessary. It must be stressed that language units at all levels are paradigmatically organized in systems, not mere inventories. An inventory of phonemic features might be useful in describing the systems of phonemes, but it undermines the status of its items as linguistic entities with a value of their own.

Let us now examine the phonologies of several languages with the aim of describing their specific kinakemic systems. In all languages such systems consist of two parts, vocalic and consonantal. The two have much in common both in their structures and substance, they are closely interconnected, but they are undoubtedly autonomous subsystems. Since our present investigation is confined to consonantal subsystems, the rank-indicating prefix ‘sub’ will henceforth be omitted.

The simplest consonantal kinakemic systems are found in the Polynesian languages. The ten consonant phonemes of Maori are satisfactorily described by four pairs of distinctive features (table I). This will satisfy the phonologist whose aim is to present the system of phonemes; the inventory of features is for him only a tool in the identification of individual phonemes. But the kinakemic analysis only begins where the phonemic analysis ends. The object of the former is to discover the paradigmatic structure of the kinakemic system, the substance and syntagmatic properties of the kinakemes.

Table I. The matrix of Maori consonant phonemes

Features	Phonemes
	p	t	k	m	n	ŋ	wh	h	w	r
Sonority	–	–	–	+	+	+	–	–	+	+
Occlusion	+	+	+	+	+	+	–	–	–	–
Labiality	+	–	–	+	–	–	+	–	+	–
Velarity	–	–	+	–	–	+	0	0	0	0

Two kinakemes correspond to the feature of sonority in the Maori matrix, and they determine the type of sound produced: one produces tone, the other results in noise. The second pair of kinakemes is concerned with the type of obstruction: one provides an oral occlusion, the other makes a constriction. The third pair of kinakemes operates the articulators – the lips or the tongue. The last pair is responsible for the point of articulation in the roof of the mouth – front or back. The four pairs of kinakemes thus cover all the vital choices in the production of consonants – the choice between periodic or non-periodic vibration (tone or noise, in other terms sonority or discordance), between complete or incomplete obstruction (occlusion or constriction), the choice between the lips or the tongue as the articulator, the choice of the point for the obstruction (front or back). The first two pairs of kinakemes determine the manner or mode of articulation and may be classified as modal kinakemes; the two other pairs deal with the place of articulation and may be described as locational kinakemes. Modal kinakemes are subdivided into phonal and obstructional; locational kinakemes are articulatoral or pointal.

Table II. The subsystem of consonantal kinakemes in Maori

Consonantal kinakemes
Modal				Locational
Obstructional		Phonal		Articulatoral		Pointal
Occlusion	Constriction	Sonority	Discordance	Precentral (labiality)	Central (non-labiality)	Central (dentality)	Postcentral (post-dentality)

The Maori system of consonantal kinakemes displays four hierarchic tiers (table II). On all the tiers the organization is strictly binary with predominantly equipollent relations. But certain facts indicate privative relations on the lowest tier between paired kinakemes.

Of the two phonal kinakemes the one of sonority is based on the positive articulatory action of the vocal cords, whereas discordance (noise) does not result from any specified action. It is plausible that the sonority kinakeme is positive and the discordance kinakeme is its negative counterpart. Both occlusion and constriction imply positive articulatory movements. But of the two obstructional kinakemes that of constriction is more limited in its syntagmatic capabilities, for it does not combine with any pointal kinakemes in the same phoneme – the Maori constrictives /h r/ do not distinguish front and back articulation (this is why the English alveolar /s/ becomes /h/ in Maori, e.g. hooiho < horse, Tiihema < December). As syntagmatic limitations are more typical of positive members in correlated oppositive pairs, the constriction kinakeme may be regarded as positive and that of occlusion as negative.

Among the locational kinakemes, the assignment of a positive or negative value seems to be governed by the rule that the more usual central articulations are the domain of negative kinakemes, whereas their positive counterparts are based on the rarer peripheral articulations. Then the precentral labial kinakeme is positive, while the lingual kinakeme is negative and is more properly described as non-labial, since it also embraces the faucal articulation of /h/. Articulation points in the front of the mouth are obviously more easily and frequently employed than those in the back; hence the positive interpretation of the peripheral postdental articulation, palatal and velar, while the negative pointal kinakeme is realized in the dental or alveolar area. It is clear that the positive articulatoral kinakeme of labiality precludes any pointal kinakeme in the consonant phoneme.

A matrix can now be built for the Maori consonants, which reflects the structure of the kinakemic system as well as the kinakemic composition of phonemes (table III).

Table III. The kinakemic structure of Maori consonant phonemes

Kinakeme groupings		Phonemes
		p	t	k	m	n	ŋ	wh	h	w	r
Modal	Obstructional	–	–	–	–	–	–	+	+	+	+
	Phonal	–	–	–	+	+	+	–	–	+	+
Locational	Articulatoral	+	–	–	+	–	–	+	–	+	–
	Pointal	0	–	+	0	–	+	0	0	0	0

We turn now to the consonantal kinakemes of English. It is evident that their system cannot be as simple as in Maori, since four pairs of kinakemes can produce no more than 16 different combinations, and English has 24 consonant phonemes. But it is not only a matter of numbers. Certain kinds of consonants cannot be generated by a kinakemic system like that of Maori. For instance, the Maori system is incapable of generating affricates or voiced noise consonants. Affricates combine two types of obstruction – occlusion and constriction, and voiced noise is a combination of sonority with discordance. Such phonemes must therefore include two obstructional or two phonal kinakemes. This is, however, out of the question in Maori, as two paired kinakemes are in privative relation to one another and incompatible within the same phoneme.

The existence of affricates in English shows that the obstructional kinakemes of occlusion and constriction do not make a privative pair. In English they are separated and enter into two different pairs: occlusion is paired with its true privative correlate non-occlusion, constriction is likewise paired with non-constriction. Being thus both positive kinakemes in different privative pairs, occlusion and constriction are able to combine in the affricates /tš dž/. Here the phonic realizations of both kinakemes are located in the same area and are consecutive in time. There is a third English phoneme with the same combination of two positive obstructional kinakemes: this is /l/, where the two kinakemes are realized concurrently in time, but at different locations, occlusion being alveolar and constriction lateral (the phonetic proximity of the lateral sonant to affricates has been noticed [e.g. Jakobson, 1962, p. 496]; it appears to result from the identity of their obstructional kinakemes). A combination of two negative obstructional kinakemes is logically possible, but hardly feasible in consonants, since it is tantamount to the absence of any obstruction.

Thus the transformation of one pair of kinakemes (as represented in Maori) into two related pairs results in three instead of two phoneme classes. A similar transformation is also manifest in the phonal kinakemes of English. While Maori consonants are either pure sonants or pure discordants, English possesses an intermediate class of consonants with less sonority than in the sonants and less noise than in the pure discordants. This testifies to the separation of phonal kinakemes into two privative pairs: sonority versus non-sonority, discordance versus non-discordance. The intermediate class presents a combination of two phonal kinakemes, which may be both positive or both negative. In the former case, the resulting consonants are voiced, combining distinct tone with considerable noise; otherwise they are non-sonorous and non-discordant, phonetically lax, lenes. English obviously represents the latter case, in which the discordants are usually tense, fortes and contain in their phonic realization special means of increased noise production, such as aspiration.

The English locational kinakemes display similar departures from the Maori pattern. Instead of one positive articulatoral kinakeme of precentrality (i. e. labiality) versus negative centrality (non-labiality), English has developed another positive kinakeme of postcentrality, phonically realized in the faucal articulation. This leads to the formation of three articulatoral classes: precentral labial, central lingual, postcentral faucal, of which the second combines two negative articulatoral kinakemes of non-precentrality and non-postcentrality. Phonemes of this class only may take pointal kinakemes, which have the same structural organization as the articulatoral ones. Instead of one positive pointal kinakeme in Maori, which is that of postcentrality and is opposed to the negative kinakeme of centrality, there are two positive pointal kinakemes in English: one is the kinakeme of postcentrality, the other of precentrality. Lingual consonants are divided accordingly into three pointal classes: precentral prealveolar (i.e. dental), central alveolar, postcentral palatal and velar. As in the case of articulatoral classes, the central class is formed by a combination of two negative kinakemes.

Table IV presents the kinakemic structure of English consonant phonemes. Certain rules can be observed in it, which regulate the intraphonemic combination of kinakemes. There is a marked difference in this respect between modal and locational kinakemes. Every phoneme contains four modal kinakemes, among which from one to three are positive. But a phoneme never contains more than one positive locational kinakeme and may contain only negative locational kinakemes. All locational kinakemes are excluded from phonemes with two positive obstructional kinakemes; all pointal kinakemes are excluded by positive articulatoral kinakemes.

Table IV. The kinakemic structure of English consonant phonemes

Phonemes	Kinakemic groupings
	Modal				Locational
	Obstructional		Phonal		Articulatoral		Pointal
	Occlus.	Constric.	Sonor.	Discord.	Preling.	Postling.	Prealveol.	Postalveol.
p	+	–	–	+	+	–	0	0
t	+	–	–	+	–	–	–	–
k	+	–	–	+	–	–	–	+
b	+	–	–	–	+	–	0	0
d	+	–	–	–	–	–	–	–
g	+	–	–	–	–	–	–	+
m	+	–	+	–	+	–	0	0
n	+	–	+	–	–	–	–	–
ŋ	+	–	+	–	–	–	–	+
f	–	+	–	+	+	–	0	0
θ	–	+	–	+	–	–	+	–
s	–	+	–	+	–	–	–	–
š	–	+	–	+	–	–	–	+
h	–	+	–	+	–	+	0	0
v	–	+	–	–	+	–	0	0
đ	–	+	–	–	–	–	+	–
z	–	+	–	–	–	–	–	–
ž	–	+	–	–	–	–	–	+
w	–	+	+	–	+	–	0	0
r	–	+	+	–	–	–	–	–
j	–	+	+	–	–	–	–	+
tš	+	+	–	+	0	0	0	0
dž	+	+	–	–	0	0	0	0
l	+	+	+	–	0	0	0	0

Just like phonemes may show variation in their phonic realization due to the influence of neighbouring phonemes, kinakemes may receive varying realizations when combined with certain other kinakemes in the same phoneme. One instance of this has already been mentioned above: the combination of occlusion and constriction is realized differently in the sonant /l/ and the non-sonants /tš dž/. The pointal kinakeme of postcentrality, i.e. postalveolarity, is realized differently in the occlusives /k g ŋ/, which are velar, and in the constric-tives /š ž j/, which are palatal [e.g. Jakobson, 1962, p. 428].

The kinakemic structure of phonemes affects interphonemic combination as well. In English initial consonant clusters, two neighbouring phonemes will often have no common positive kinakemes – e.g. occlusives with constrictive sonants /pr tr kr br dr gr tw dw kw/, constrictive /s/ with occlusive sonants /m n/. In /sp st sk/ both components have the positive kinakeme of discordance. The clusters /pl kl bl gl/ combine two occlusions of different location (note the absence of /*tl *dl/ where the occlusions would coincide in location). Clusters may also combine two constrictions of different location, e.g. /fl sl fr θr šr θw sw/ (the constrictions would coincide in location in *sr, which is accordingly impossible). Final clusters display greater freedom of kinakeme combination; thus, two occlusions of the same location occur in /mp nt nd ŋk lt ld/.

The kinakemic system of English consonants, presented in table V, shows the same structural pattern as that of Maori, with a hierarchy of tiers and binary relations on every tier. But there are five tiers in English against four in Maori. The enlargement results from the fact that each kinakeme of the simpler system is replaced by a privative pair of kinakemes in the richer system. This involves a significant change in the relations between kinakemes. In Maori, non-sonority, for instance, is equivalent to discordance; but this is not the case in English. Occlusion and discordance, which in Maori serve as negative kinakemes, are positive in English, and new negative kinakemes emerge on a strictly privative basis. Thus, identical phonic substance is incorporated differently into different systemic structures.

The two highest tiers are the same in both languages. The vocalic and consonantal subsystems are obviously universal. The same can be said about the next tier, which we shall designate as the tier of categories – modal and locational. The units of the lowest tier are the kinakemes themselves. Above them is the tier whose units are best described as oppositions; the latter are thus treated not as pure relations between phonemes, but as units in kinakemic systems, the smallest paradigmatic groupings of two kinakemes with different signs [e.g. Jakobson, 1962, p. 421]. The tier of oppositions is the only one between categories and kinakemes in Maori; but English has two tiers here. Let us describe the higher of them as the subcategorial tier. Then the two tiers of subcategories and oppositions, kept apart in English, merge into one in Maori. This is probably a mechanism which regulates the numerical composition of the phoneme inventory – the two tiers are separated for richer inventories and coalesce in the case of poor inventories.

Besides the four-tier structure of Maori and the five-tier structure of English, which demonstrate complete coalescence and complete separation of subcategories and oppositions, many languages have systems of an intermediate structure. In them some subcategories evolve two oppositions, while the others contain only one opposition each.

An instance is found in French. Only one subcategory is fully developed here with two oppositions: the phonal subcategory with the kinakemes of sonority and discordance as the positive members of two oppositions. Accordingly, there are three phonal classes of consonants, the intermediate class combining two positive kinakemes in the voiced /b d g v z ž/. The absence of affricates shows that the obstructional subcategory is undeveloped, its two kinakemes make one opposition. Both locational subcategories are also undeveloped; the one positive articulatoral kinakeme is that of precentrality (labiality), the one positive pointal kinakeme marks postcentrality (palatality or velarity). It is noteworthy that the French sonants /l r/ differ from their English ‘homographs’ in their kinakemic composition. The former does not join the non-existent affricates, it is the central constrictive sonant. In English, this constitutes the kinakemic composition of /r/; the ‘homographic’ French sonant /r/ finds another place in the system, acquiring the positive pointal kinakeme of postcentrality with its uvular realization.

Kinakemic systems of the three structural types analysed so far generate comparatively small phoneme inventories. The simplest system with totally undeveloped subcategories in Maori produces 10 consonants, partial development creates 17 phonemes in French, while full development of all subcategories in English results in 24 consonants. Richer inventories of more than 30 phonemes require further reinforcement of kinakemic systems. But new tiers are no longer possible, for all the logical resources are used to the full in the fifth tier. The only way to increase the potential of the kinakemic system is to bring into action new substance resources. This is achieved by introducing supplementary oppositions into the system. Instances are the modal opposition of protensity (gemination) in Old English, the locational opposition of palatalization in Russian. The positive kinakemes of supplementary oppositions usually combine easily with other kinakemes, and this enables them to almost double the phoneme inventory. Thus 16 geminates were added to the 18 simple consonants in Old English, raising the total to 34 phonemes (only /w j/ had no correlated geminates). In Russian 18 palatalized consonants are added to the 19 non-palatalized consonants (only /j/ has no correlate), creating an inventory of 37 phonemes.

In languages with the richest consonant inventories, with more than 50 phonemes kinakemic systems may contain not one, but several supplementary oppositions. An instance is Abkhazian with its 58 consonants, where there are three supplementary oppositions – the modal opposition of glottalization and the locational oppositions of labialization and palatalization.

Zusammenfassung

Systeme der elementaren phonologischen Einheiten

Das Phonem ist nicht die kleinste phonologische Einheit; es besteht aus elementaren phonologischen Einheiten, die als Kinakeme bezeichnet werden und den Begriff der ‘distinktiven Merkmale’, überflüssig machen. Kinakeme bilden paradigmatische Systeme mit hierarchisch geordneten Stufen von streng binarer Struktur. Es werden einige konsonantische Kinakemsysteme mit einfacher, völlig oder teilweise entwickelter, komplexer Struktur beschrieben.

Résumé

Systèmes des unités phonologiques élémentaires

Le phonème n’est pas la plus petite unité phonologique; il consiste en unités phonologiques élémentaires: les kinakèmes qui rendent superflue la notion de «trait distinctif». Les kinakèmes forment des systèmes paradigmatiques avec une hiérarchie de niveau et une structure strictement binaire. Des sous-systèmes de kinakèmes consonantiques de structure simple, complètement ou partiellement développée et de structure complexe sont decrits.

References

Baudouin de Courtenay, I. A.: Izbrannyje trudy po obščemu jazykoznaniju, vol. 2

(Izdatelstvo Akademii Nauk SSSR, Moskva 1963).

Benveniste, É.: Les niveaux de l’analyse linguistique. Proc. 9th Int. Congr. Linguistics

(Mouton, The Hague 1964).

Benveniste, É.: Problèmes de linguistique générale (Gallimard, Paris 1966).

Fant, G.: Sound, features and perception. Proc. 6th Int. Congr. Phonetic Sciences

(Prague 1967).

Grucza, F.: Sprachliche Diakrise im Bereich der Ausdrucksebene des Deutschen.

Beiträge zur allgemeinen Sprachtheorie (Państwowe wydawnictwo

naukowe, Poznań 1970).

Jakobson, R.: Selected writings. I. Phonological studies (Mouton, The Hague 1962).

Jakobson, R.; Fant, C. G. M., and Halle, M.: Preliminaries to speech analysis.

The distinctive features and their correlates (MIT, Cambridge 1955).

Jakobson, R. and Halle, M.: Fundamentals of language (Mouton, The Hague 1956).

Panov, M. V.: Russkaja fonetika (Prosveščenije, Moskva 1967).

Solntsev, V. M.: Jazyk kak sistemno-strukturnoje obrazovanije (Nauka, Moskva 1971).

Trnka, B.: Poznamky ke kombinatorickým variantám a k neutralisací.

Čas. mod. Filol. 24: 261-270 (1938).

Vachek, J.: Dictionnaire de linguistique de l’École de Prague (Spectrum, Utrecht 1960).