Proceedings
of the Eleventh International Congress of Phonetic Sciences (ICPhS)
August 1-7, 1987, Tallinn,
Estonia
Volume
2, p. 415-418
The
Ultimate Phonological Unit
as the
Smallest Morpheme Shape
Dept. of English Philology, Pedagogical Institute, Tula,
Russia
The
phoneme is divisible not only because it consists of ultimate constituents
traditionally known as distinctive features, here termed kinakemes, but also
because morpheme boundaries can run through phonemes. This is made possible by
the ability of a kinakeme not only to participate in distinguishing morpheme
shapes as a phoneme constituent, but also to provide such a shape by itself.
Instances of inflexional and derivational affixes with shapes consisting of a
single kinakeme are found in various languages, e.g. Estonian, Gaelic, Latvian,
Nivkh (Gilyak), Romanian, Russian. A morpheme boundary can also run through a
phoneme when an affix shape consists of a kinakeme cluster smaller or larger
than a phoneme; the boundary then dissects the phoneme in question or its
neighbour.
Ever since the notion of the phoneme
as the basic unit in the sound system of language came into being, the problem
of its (in)divisibility has always been present, though not always explicit in
phonological theory. The insistence on the unquestionable absolute
indivisibility of the phoneme, so characteristic of phonology’s early days,
soon gave way to the recognition of the existence within the phoneme of smaller
truly ultimate constituents, named distinctive features [11, 272, 422-5; 12,
25], merisms, phononemes, subphonemes etc. For reasons explained elsewhere [14,
82-3] the best term is Baudouin’s coinage ‘kinakeme’ [1, 199, 290].
Recent research has demonstrated that these entities possess all the fundamental
properties of language units:
(1) They are language-specific
and cannot therefore be items in a universal inventory, Jakobsonian [11, 484-6]
or Chomskyan [8, 335], any more than phonemes, syllables or words could be
listed so [9, 152].
(2) In each language they are
paradigmatically united in a kinakemic system whose structure follows universal
principles, but provides, like any other language system, a unique way of segmenting
and organizing extralinguistic substance, which is not sound, linguistically
organized by the phonemic system, but the speaker’s cerebral activity in
initiating sound and the listener’s subsequent perceptive cerebration [14, 83
ff.; 15, 277-83].
(3) Each language has its
specific syntagmatic patterns for kinakemic combination in phonemes, which is
basically non-linear simultaneous [15, 277-83].
(4) Kinakemic systems play a
leading role in the phonological evolution of languages and determine the
direction of phonemic change [7].
The establishment
of the kinakeme as the ultimate language unit does not, however, take the
question of phoneme (in)divisibility off the phonological agenda, for the
problem has more than just one facet. An analogy may be appropriate here with
the atom, whose very name reflects its indivisibility: despite its decomposition
into a host of particles it remains the ultimate quantum of a chemical element
and is indivisible on that level. Likewise, the phoneme is segmentable in
certain aspects and indivisible in others.
The discovery of
the phoneme in 20th-century phonology was in a sense a rediscovery, for the
original discovery dates back to the invention of alphabetic writing, when
letters were created as symbols for phonemes. As long as the sound substance
behind them was not analysed, they were treated as representing indivisible
units of sound. The advent of phonetics in the 19th century put an end to the
notion of integral sound units symbolized by letters and led to a two-pronged
attack against them, pointing out the wide range of their variation and the
complexity of their production and perception. The emergence of phonology was
stimulated above all by the urgent necessity to uphold the notion of sound
quanta and to protect them from being disintegrated in a continuum of variable
phonic realizations. Hence the firmness with which the founders of phonology
rejected every infringement on the principle of phoneme indivisibility.
The two questions
concerning the unity of the phoneme with its variability, its articulatory and
auditory complexity, have been answered differently by modern phonology.
Allophonic variation has found its place in phonemic theory and no longer
threatens phonemic unity. As for the inner complexity of phoneme structure, the
discovery of the kinakeme as its constituent has of course shown the phoneme to
be divisible into units of a lower level. But there is another aspect of
phoneme structure which in its time attracted attention in connection with the
problem of phoneme divisibility – the question of monophonemicity for sounds
with temporally varying articulation, i.e. diphthongs and affricates. It must
be stressed that kinakemic divisibility of phonemes does not affect their
monophonemicity if it is established by the well-known criteria of classical
phonology, which remain fully valid. Since kinakemic combination in a phoneme
is non-linear and the kinakemes, clustered to form a phoneme, are activated
more or less simultaneously, none of them can occupy a temporal segment of its
own.
The kinakemic level is
ultimately responsible for the conversion of sense into sound and the
reconversion of sound into sense. Sound as the physical vehicle for the
externalization of the speech signal is obviously so different from the
cerebral activities with which the kinakemic system is concerned, that it has
to be represented in the language system by a separate level. Deprived of its
classical status of ultimate phonological (indeed, linguistic) unit, the
phoneme retains its ultimateness on that level. Its properties including
(in)divisibility are determined by the needs of the level it belongs to. Its
kinakemic divisibility, far from being an obstacle to its function in
organizing sound, is absolutely indispensable for the purpose. The phoneme is
divisible in two other respects, and in both cases the divisibility is
determined by the needs of other language levels without affecting the
functioning of the phoneme. First, a phoneme can be crossed by a syllabic
boundary. That is not possible for vowels and many languages do not permit it
in consonants either. But languages that make use of the kinakeme of vowel
checking may put the syllabic boundary after checked vowels inside a single
consonant, between its onset and release, as in English body, supper.
Secondly, a phoneme can be crossed by a morpheme boundary, when one of the
bordering morphemes uses a kinakeme and not a whole phoneme as its sole
exponent.
The employment of kinakemes
as exponents for morphological categories is observed in several languages. The
analysis of protensity in Estonian provides abundant instances of the
phenomenon. Estonian protensity with its traditionally recognized three levels
(short, long, over-long) is now treated as resulting from a combination of two
separate oppositions: long vs. short, over-long vs. long [cf. 16, 17]. The
difficult problem of the exact phonic natures of both must be put aside in the
present paper. What concerns us is the functional difference between the two:
the opposition of long vs. short is mostly active in distinguishing lexical
items (aasta ‘year’ – aste ‘degree’, mure ‘grief’ – murre
‘dialect’), while the predominant function of the opposition of over-long vs.
long lies in the grammatical sphere, where it is widely used to distinguish
noun cases: part.sg. saAli ‘hall’ – gen.sg. saali, gen.sg. hoOne
‘building’ – nom.sg. hoone, part.sg. liNna ‘town’ – gen.sg. linna,
elat.sg. kaLlist ‘dear’ – part.sg. kallist. It is obvious that
the kinakemes of the former opposition, which may be described as lexical
protensity, do not possess semantic values of their own, whereas the kinakemes
of the latter opposition of grammatical protensity serve as sole exponents of
the categorial meanings of cases. That shows that the kinakeme of grammatical
protensity or over-length is in itself a morpheme shape – a kind of case infix.
As such it must occupy a certain fixed position in the stem shape and therefore
must be able to join any phoneme in that position. The phoneme that
incorporates the infix always follows the syllabic peak and may be a vowel; it
is either different from the syllabic vowel and forms a diphthongal cluster
with it (part.sg. laUlu ‘song’ – gen.sg. laulu, gen.sg. laIne
‘wave’ – nom. sg. laine), or identical with it and forms a bimoric
monophthong with it, which becomes trimoric when the infix is added to it (see saAli,
hoOne above). The infix can also enter consonants, single (part.sg. sePpa
‘smith’ – gen.sg. sepa, elat.sg. riKkast ‘rich’ – part.sg. rikast),
in a cluster (part.sg. oKsa ‘branch’ – gen.sg. oksa) or geminated
(see liNna, kaLlist above).
In Irish Gaelic [2, 81]
categorial distinction of number and case in noun paradigms is regularly
achieved by the kinakeme of palatalization: bád – báid ‘boat’, béal –
béil ‘mouth’, bonn – boinn ‘coin’, cnoc – cnoic ‘hill’.
In some German dialects [3,
384-5] noun number is distinguished by certain kinakemes alone, e.g. voicing
(pl. barγ – sg. barç
‘mountain’, pl. brev – sg. bref ‘letter’), vowel checking (pl. fiš
– sg. fīš ‘fish’) or vowel fronting (pl. hünd – sg. hund
‘dog’).
Before their dat.sg.
inflexions -m, -i Latvian nouns display a stem-final vowel /a/,
/u/, /e/, /i/ (galdam ‘table’, tirgum ‘market’, zemei
‘land’, sirdij ‘heart’). The sole exponent for the locative case is the
kinakeme of protensity included into that vowel (galdā, tirgū, zemē, sirdī);
for the accusative the sole exponent is the kinakeme
of tongue-raising added to the vowel (galdu, tirgu, zemi, sirdi) [10; 5, 232].
The use of kinakemes as
categorial exponents is widespread in Romanian, where it is found in the
paradigms of nouns and verbs. The best-known example is the use of the
palatalization kinakeme to mark the plural of nouns (pl. lupi ‘wolf’ –
sg. lup) and the 2nd person in verbs (dormi ‘sleepest’ – 1st sg. dorm).
But there are other instances as well. The four canonical forms expressing
number and gender, e.g. sg.m. vecin ‘neighbour’, mîndru ‘proud’,
pl.m. vecini, mîndri, sg.f. vecină, mîndră, pl.f. vecine,
mîndre, demonstrate that kinakemic distinctions also exist between the
two genders: both masculine forms end in high vowels /u/, /i/, regularly reduced
to zero representations after single consonants (as in vecin, vecini),
whereas both feminine forms end in non-high vowels /ă/, /e/; the masculine
gender is thus marked by the positive kinakeme of tongue-raising, the feminine
remains unmarked and usually includes its negative counterpart into its final
vowel. But the feminine gender uses the positive kinakeme of tongue-lowering
for its definite article: def.sg. casa ‘house’, cartea ‘book, ziua
‘day’ with low vowels – /a/, front /ea/, back /ua~oa/, absent from the
indefinite forms (casă, carte, zi). In the 3rd person of
many Romanian verbs indicative and conjunctive forms are distinguished only by
the presence or absence of the positive kinakeme of fronting in the inflexional
vowel (ind. bate – conj. bată ‘beat’, conj. poarte – ind. poartă
‘carry’). Thus, the positive kinakemes of consonant palatalization, of
tongue-raising and tongue-lowering can alone act as exponents for the
grammatical categories of number, gender, definiteness in Romanian nouns, of
person and mood in verbs.
The ability of single
kinakemes to serve as grammatical exponents is not confined to morphological
categories. They may also function as indicators of syntactic relations. For
instance, in the Nivkh (Gilyak) language [4; 6] syntactic subordination is
marked by changing a modal kinakeme in the initial consonant of the headword.
When a verb takes an object or a noun takes an attribute, it reflects
its syntactic domination by a kinakemic restructuring of its initial consonant:
the kinakeme of occlusion is replaced by that of constriction and vice versa,
voicing is interchangeable with devoicing (bod' ~ vod' ~ pod' ‘to
hold’), the kinakeme of aspiration with that of constriction (khu
~ xu ‘arrow’). The role of kinakemes as markers of syntactic dominance in
Nivkh is functionally analogous to the Persian e-z`āfe,
i.e. the suffix attached to the noun when it takes an attribute (pesar-e
mard ‘the man’s son’, āb-e garm ‘hot water’).
The same pattern of
kinakeme interchange is widely used in Nivkh lexico-grammatical derivation.
Replacement of initial occlusion by constriction may form a causative verb (kukud'
‘to fall’ – γukud'
‘to drop’), a change in the opposite direction creates a verbal noun (fuvd'
‘to saw’ – phuf ‘a saw’). The kinakeme of voicing in the
initial consonant of a qualitative word indicates intensity (tuzla
‘cold’ – duzla ‘very cold’) [2, 102; 6-I, 72; 6-II, 41]. The use of a
single kinakeme as an exponent of lexico-grammatical derivation is found in
Russian, where the inclusion of the palatalization kinakeme into the stem-final
consonant is highly productive in building verbal and adjectival nouns: подписать – подпись,
связать – связь, обувать
– обувь, бездарный
– бездарь, удалой
– удаль, новый
– новь.
The conclusion can safely
be reached that a wide range of kinakemes, both vocalic and consonantal, among
them vowel checking and fronting, tongue raising and lowering, protensity,
palatalization, occlusion, constriction, aspiration, voicing and devoicing, are
quite capable of serving as the sole exponents of various derivational meanings
– syntactic dominance in a phrase, morphological in the categories of case,
number, gender, definiteness, person, mood, lexico-grammatical in shifting a
word from one class to another.
In general linguistic terms
the capability of kinakemes to function as morpheme exponents refutes the idea
of the morphemic indivisibility of the phoneme. When a kinakeme, which is
naturally incapable of externalization outside a phoneme, constitutes a
morpheme shape in itself and is thus from the morpho-semantic viewpoint
independent of the phoneme it enters, it is separated from the other kinakemes
in the phoneme structure by a morpheme boundary.
Kinakemes are capable of
building affix shapes not only singly, but also in clusters. Strictly speaking,
every affix shape can be described as made of a kinakeme cluster, since a
phoneme is always such a cluster. But there is no need to do so if the affix
shape consists of entire phonemes. However, the identity between kinakeme
clusters in their two constitutive functions – making up phonemes and morpheme
shapes – is not obligatory, and a cluster as an affix shape may not equal any
phoneme or phoneme combination.
The affix shape is often
larger than a phoneme and contains an extra kinakeme which is incorporated into
the adjacent phoneme of the stem. Linguistic tradition has treated such
instances as sound alternations, phoneme replacements, which entails separate
treatment for each case of replacement. Relegating the phenomenon to the
kinakemic level brings more uniformity to linguistic description.
Of the languages discussed
above Gaelic and Russian abound in affix shapes that
expand beyond the inflexional phonemes and penetrate into the phonemes of the
stem. For instance, the Irish possessive noun prefix of the 3rd person
contains, besides the entire phoneme /«/, the kinakeme of constriction for the masculine sg. (port – a phort
‘his port’, cota – a chota ‘his coat’), the kinakeme of voicing for the
plural (a bport ‘their port’, a gcota ‘their coat’). In contrast
the corresponding prefix for the fem.sg. is equal to the phoneme /∂/ (a port
‘her port’, a cota ‘her coat’) [2, 79-83].
In Russian the vowel /e/ is
unable to begin a suffix shape alone and is therefore always accompanied in it
by the kinakeme of palatalization implanted into the final consonant of the
stem: loc.sg. столе, dat. sg. траве, inf. твердеть,
where the palatalization kinakemes in /l'/, /v'/, /d'/ do not belong to the
stem shapes, but to the affix shapes together with the vowel /e/. As a morpheme
boundary separates the kinakeme of palatalization from the rest of the phoneme
it joins, the stem shapes by themselves do not undergo any changes on the
kinakemic level despite the changes in the kinakeme structures of their final
consonants.
The Russian vowel /i/ is
not always accompanied by the kinakeme of palatalization in suffix shapes. It
equals the suffix shape in some noun inflexions (nom.pl. пилы) ,
but in verb inflexions beginning with the same vowel phoneme it is accompanied
by the kinakeme of palatalization placed in the last consonant of the stem: пилит.
Affix shapes may also be
smaller than a phoneme, which then has to fill the resulting gap in its
structure by admitting a certain kinakeme from the stem shape. This is the
essence of synharmonism. For instance, in Finnish the kinakeme of vowel
fronting or its negative counterpart is carried over from the stem vowels into
the vowel of the suffix: inf. puhumaan ‘speak’ – leikkimään
‘play’.
In English the suffix shape
in hopes, moves contains only the kinakemes common to both
phonemes /s/, /z/, and the suffix shape in hoped, moved likewise contains only the kinakemes common
to /t/, /d/. In other words, the suffix shapes do not show any variation
determined by the phonemic context. The kinakemes of voicing and devoicing
which enter the suffixal consonants belong to the stem shapes and not to the
suffix shapes [13].
Affix shapes larger or
smaller than phonemes have a special role to play in strengthening the unity of
the derived word, as the penetration of one morpheme shape into the phonemes
that otherwise belong to the other morpheme, the resulting participation of a
phoneme in two morpheme shapes at once are factors which help to cement the
ties between the morphemes. But each direction of penetration has a typological
significance of its own. When the affix shape is larger than the affixal
phoneme and spills over into the stem, it serves to emphasize the constitutive
role of the affix in the structure of the word and accordingly reduces the
discernibility of the stem; this is a characteristic trend in syntheticized
languages. On the other hand, analyticized languages show a typological
propensity to emphasize the pivotal role of the stem by its easy separation
from the affixes, and that requires the stability of the stem, well-defined
morpheme boundaries; the unity of the word is also enhanced by morpheme
boundaries running through phonemes, but the phonemes affected in such
languages belong to affixes, whose shapes are smaller than the phonemes. It can
be said that in the former type of languages the unity of the word is based on
the power of its affixes, whereas in the latter type it uses the stem as its
bulwark.
References
[1] И. А. Бодуэн де Куртенэ. Избранные труды по общему
языкознанию, т.2.
М., 1963.
[2] Л. Г. Герценберг.
Морфологическая структура слова в ирландском языке.
Морфологическая
структура слова в индоевропейских языках. М., 1970.
[3] В. М. Жирмунский. Немецкая
диалектология. М.-Л., 1956.
[4] Е. А. Крейнович. Фонетика
нивхского (гиляцкого) языка. М.-Л., 1937.
[5] Общее языкознание.
Внутренняя структура языка. М., 1972.
[6] В. З. Панфилов. Грамматика
нивхского языка, ч. I,
1962; ч. II, 1965. М.-Л.
[7] В. Я. Плоткин. Эволюция
фонологических систем. М., 1982.
[8] N. Chomsky, M. Halle. The
Sound Pattern of English. Harper & Row, 1968.
[9] E. Fischer-Jørgensen.
Trends in Phonological Theory: A Historical Introduction.
Akademisk
Forlag. Copenhagen,1975.
[10] M. Halle. Discussion of
[13], ibid.
[11] R. Jakobson. Selected
Writings, I. Phonological Studies. Mouton, 1962.
[12] R. Jakobson, L. R. Waugh.
The Sound Shape of Language. Harvester, 1979.
[13] F. Palmer. Grammatical
Categories and their Phonetic Exponents.
Proc.
9th Int. Congr. Linguists. Mouton, 1964.
[14] V. Y. Plotkin. Systems of
Ultimate Phonological Units. Phonetica, 33 (1976), 2.
[15] V. Y. Plotkin. The Kinakeme
as the Ultimate Unit of Language.
Kwartalnik
Neofilologiczny, XXV (1978), 3.
[16] V. Tauli. Standard
Estonian Grammar, I. Uppsala, 1973.