from
Journal of the International Phonetic Association, vol. 27, pp. 27-34, 1997
Bulgarian
Speech Rhythm: Stress-Timed or Syllable-Timed?
Snezhina Dimitrova, University of Sofia
1. Introduction
Ever since the
terms “stress-timed rhythm” and “syllable-timed rhythm” were put forward by
Pike in 1945, scholars have tried to describe the rhythms of the world’s
languages in dichotomous terms. Thus, Abercrombie claims that “As far as is
known every language in the world is spoken with one kind of rhythm or the
other.” (1967:97) At the same time, experimental work has failed to produce
empirical evidence in substantiation of these theoretical claims (Roach 1982,
Dauer 1983). Therefore, it is not surprising that experiments designed to
determine which of the two rhythmic categories a given language belongs to
sometimes come to contradictory conclusions.
Thus, on the
basis of results obtained from a contrastive study of vowel reduction in
English and Bulgarian, Bulgarian has been claimed to have syllable-timed rhythm
(Yordanova 1987). However, data from a test on perception of Bulgarian rhythm
by native speakers seem to suggest that the language is characterised by
stress-timed rhythm (Kurlova and Krustev 1989).
On the other
hand, many scholars today prefer to view rhythmic differences between languages
not in strictly dichotomous but in scalar terms. But at the same time, despite
the lack of conclusive experimental evidence, the dichotomous distinction is
still considered both a phonetically and a phonologically useful concept.
Therefore, a
scalar model of speech rhythm, with two (hypothetical) languages at the two
extremes of the rhythm scale (Laver 1994), should be able to account better for
the observable facts than the traditional dichotomous distinction. Such a model
should make it possible to find the place of a given language on the rhythm
scale with reference to, for example, English or French.
One such model
has been put forward by Dauer (1987). Within the framework of this model, the
present paper attempts to establish the position of Bulgarian on the speech
rhythm continuum.
2.
Dauer’s
(1987) model
Dauer defines
rhythm as “the grouping of elements into larger units”, these elements in
language being syllables, whereas, in some languages at least, stresses (or
accents) also participate in order to mark off one group of syllables from
another. She hypothesises that all languages have rhythmic grouping, but that
not all necessarily have accent. Rhythm, then, is a total effect which involves
a number of components, namely, (i), syllable and vowel duration, syllable
structure and quantity as major factors responsible for length, (ii),
intonation and tone as means of achieving pitch distinctions, (iii), vowel and
consonant quality, and (iv), function of linguistic accent. Dauer also develops
a system of rating according to which each component is broken down into
“features’, and each feature is assigned a plus or a minus value (or sometimes
zero). In this way, a relative rhythm “score” for a given language is obtained:
the more pluses a language has, when assessed in terms of the above components,
the more likely it is, according to Dauer, that the language in question
has “strong stress” and that it is
“stress-timed”.
In the following table, English, Spanish and French are assessed on the
basis of information in Dauer
(1987):
English Spanish French
1. Length
1.1 Duration
+ O N
1.2 Syllable structure + -
-
1.3 Quantity
2. Pitch
2.1 Intonation +
N -
2.2 Tone
3. Quality
3.1 Vowels +
- N
3.2 Consonants + N -
4. Function of accent +
+ -
Table 1:
Components of language rhythm (N = not
assessed in Dauer's paper)
The languages
above are rather different in terms of rhythm, English being near to the
"stress-timing" end of the
rhythm scale and French being closer to the opposite
"syllable-timing" end; Spanish, on the other hand, would occupy an
intermediate position between these two extremes.
3.
Speech
rhythm in Bulgarian: an assessment with the help of the model
For an
assessment of Bulgarian along the same lines, two of the above components are irrelevant:
(i) Quantity: no
phonological vowel quantity distinctions exist in present-day standard
Bulgarian.
(ii) Tone: the
phonological use of pitch is restricted in contemporary standard Bulgarian to
the level of the phrase and the sentence; tone does not differentiate units at
the level of the word or the level of individual syllables.
For the other
components of language rhythm, the following scores can be given to Bulgarian:
3.1 Length
3.1.1Duration. Dauer proposes a + mark
for a language in which accented syllables, and accented vowels in particular,
are always longer than unaccented ones by at least 1.5. A 0 mark is proposed
for languages in which accented syllables are only slightly longer than unaccented
ones. Finally, a language is assigned a - if, in it, accent does not affect the
length of syllables, or the language has no accent.
According to one
author, stressed vowels in Bulgarian are about 1.5 times longer than unaccented
vowels, but in many cases they can be only slightly longer, equal to, or even
shorter than unstressed vowels (Misheva 1991). Tilkov and Boyadzhiev (1990)
report that in Bulgarian unstressed vowels are, on the average, 35% shorter
compared with the same vowels in the same environment under stress.
Since stressed
vowels in Bulgarian appear to be in many cases less than 1.5 times longer than
unstressd ones, a 0 marked can be assigned here.
3.1.2 Syllable structure. For a language
to be given a +, it must have a variety of syllable types - both heavy and
light syllables with many different possible syllable structures; in addition,
heavy syllables must tend to be accented, whereas light syllables must be
(predominantly) unaccented. If, on the other hand, a language has a very limited
number of syllable types, with CV or CVC syllables predominating, and if, in
this language, accent and syllable weight are independent, then it is given a
-. (The model does not provide for a 0 mark in this case.)
The maximum
phonological structure of the syllable in Bulgarian can be summarised as
(CCC)V(CCC). But here, the actual frequency of occurrence of the various syllable types must also be taken
into consideration.
In Bulgarian,
open syllables are, in general, more frequent than closed syllables, especially
in word-medial position. Moreover, diachronically, the appearance of closed
syllables is a fairly recent development: Old Bulgarian had open syllables
only. Today, only monosyllabic words can be characterised as having complex
syllabic structure, e.g., CCVC смес /smes/ (mixture), CCCVC страх /strax/ (fear), CCVCCC сфинкс /sfinks/ (sphinx), etc. In
disyllabic and polysyllabic words, on the other hand, open syllables
predominate. (Tilkov and Boyadzhiev 1990:142)
These
observations are supported by the results from an analysis of a text corpus
comprising 7056 syllables. Of these, 5986 or 84.8%, were open syllables, predominantly of the CV and the CCV type.
The closed syllables constituted 13.2%, or 1070 syllables, the majority of
which (12.66%) were of the CVC type (Misheva, ms).
Dauer also
mentions the importance of syllable weight - a concept relating the segmental
composition of a syllable to its potential for being made prominent in an
utterance. Some scholars define a “light” syllable as one which contains a
nonbranching rhyme, i.e., it has only one element in the peak and coda
positions taken together. A syllable is “heavy” if it has a branching rhyme,
i.e., at least two elements in the peak and the coda (Katamba 1989:176). Others
(Laver 1994:517-518) define a light syllable as one whose rhyme consists of a
short vowel followed by a maximum of one short consonant, and a heavy syllable
as one which can have in its rhyme either a long vowel, with or without a coda,
or a short vowel, followed in the coda by at least one long or two short
consonants.
Whichever of the
above definitions is adopted, it seems that the majority of syllables in
Bulgarian will be characterised as light. Only some monosyllabic lexical words
with more complex segmental make-up will be described as constituting a heavy
syllable, and in them this syllabic type will correlate with lexical stress. It
should also be noted that there will be cases, in polysyllabic lexical words,
when a light syllable will carry lexical stress in the immediate presence of a
heavy syllable, e.g. радост
/’ra.dost/
(joy), прелест
/’pre.lest/
(charm), etc.
Dauer’s model lacks a 0 mark for
this component. Therefore, Bulgarian must either be given a -, or else a new 0
mark must be introduced. The latter solution appears to be more appropriate
because in spite of the fact that in Bulgarian accent and syllable weight seem
to be independent and open syllables predominate, the language cannot be said
to have “only a limited inventory of syllabic structures”.
3.2 Pitch
3.2.1 Intonation. Here again Dauer's
system divides languages dichotomously. On the one hand, there are those
languages in which accent always correlates with pitch, accented syllables
being turning points in the intonation contour and having either high or moving
pitch; emphasis and contrast in such languages are realised primarily on the
accented syllables. Such languages receive a + mark. In languages which get a
-, intonation and accent are independent, and there may be a negative correlation
between the two.
In Bulgarian, in
some cases the beginning of the pitch change is associated with the syllable
immediately preceding the stress (Tilkov and Boyadzhiev 1990, Misheva1991). On
the other hand, since in the large majority of cases stress and pitch do
correlate in Bulgarian, it can be assigned a + mark.
3.3 Quality
3.3.1 Vowels. For a language to be given
a + here, its maximal vowel system should occur only in accented syllables, whereas
vowels in unaccented syllables should be reduced or centralised. A 0 mark is
assigned to languages in which, although the unaccented vowels are fewer in
number than the accented vowels, they are not necessarily centralised, and
there may be devoicing or raising typical of unaccented vowels only. Languages
with a - have the same vowels in both accented and unaccented syllables.
A characteristic
feature of the vocalic system of Bulgarian is that the phonetic realisation of
vowels is strongly dependent on accent. The maximum number of vowel phonemes
(6) is only met in stressed position, while in unstressed position this number
is smaller. In the pairs /a - з/
and /o - u/, the
distinction “close - open” is neutralised and the actual realisation depends on
the position of the respective vowel relative to the position of lexical
acccent in the word. In the pair / e - i / the close - open opposition, while
still functional in minimal pairs such as векове /veko’ve/ (ages) - викове /viko’ve/ (cries - noun, pl.),
also shows a tendency towards neutralisation (Tilkov and Boyadzhiev 1990:63-4).
Tilkov and
Boyadzhiev call this process reduction of unstressed open /a/ and /o/ in the
direction of the closer /з/
and /u/, respectively. But if reduction is defined as a (phonological) process
whereby a peripheral vowel is replaced with a more central vowel in unstressed
syllables (Lindblom 1963 cit. by Laver 1994:516), and since no centralisation
takes place in some of the above cases, then it would be probably more appropriate
to adopt Pettersson and Wood’s (1988) treatment of this process. They claim
that what is observed in unstressed position in Bulgarian is a process of
raising of unaccented /o/ in the direction of
/u/ and of /a/ in the direction of /з/.
In terms of Dauer’s moel, then,
Bulgarian should get a 0 mark here.
3.3.2 Consonants. Dauer proposes a + for
languages in which consonants are articulated with greater precision in
accented syllables, whereas in unaccented syllables they have special reduced
allophones, or are subject to neutralisation. If, in a given language, all
consonants have the same articulation regardless of their position in relation
to accent, then this language receives a - mark.
Consonantal
allophones in Bulgarian are positionally determined rather than being dependent
on accent. Although, for example, neutralisation of the phonological contrast
between voiced and voiceless consonants is typical of present-day standard
Bulgarian, such neutralisation takes place independent of accent, e.g.,
word-finally, before a pause, where only voiceless consonants can occur: хубав /’xubaf/ (nice) - лукав /lu’kaf/ (sly.
The major
consonantal allophones in Bulgarian are likewise independent of accent position
and are contextually determined. Thus, palatalization of /k/ and /g/ always
takes place before /e, i /, no matter whether they are accented or not, e.g. гибел [‘gJibel] (sacrifice), малки [‘malkJi] (small, pl.).
Again, Dauer's system lacks a 0 mark here, therefore
Bulgarian must be given a minus mark.
3.4 Function of accent. In Bulgarian
accent is free: it can occur in various positions in a word, and shifting it
from one syllable to another can result in a new word with a different meaning,
e.g. пара /’para/ (steam) - пара /pa’ra/ (a coin). Therefore,
Bulgarian is assigned a + mark here.
4. Conclusion
4.1 The place of Bulgarian on the rhythm
scale. A comparison of the assessment of Bulgarian rhythm presented here with that of English and
French as made by Dauer reveals that, on a scale of rhythm, Bulgarian occupies
an intermediate position between these two prototypical rhythm types:
|
|
French |
Bulgarian |
English |
|
1.1 Duration |
N |
0 |
+ |
|
1.2 Syllable
structure |
- |
- |
+ |
|
2.1 Intonation |
- |
+ |
+ |
|
3.1 Vowels |
N |
0 |
+ |
|
3.2 Consonants |
- |
- |
+ |
|
4 Function of
accent |
- |
+ |
+ |
Table 2:
Bulgarian rhythm compared to that of English and French
This result can
account for the contradictory conclusions reached in earlier studies of speech
rhythm in Bulgarian. Also, it appears that Dauer’s model can provide a useful
starting point for any investigation of speech rhythm in a given language.
4.2 Dauer’s model. Although Dauer's
model seems to account better for the rhythmical differences between languages
than the "classical" dichotomous stress-/ syllable-timing theory, it
could be further improved by, e.g., adding 0 marks for certain components. This
will make it possible for the model to capture finer rhythmic differences
between languages.
R e f e r e n c e s
1.
Abercrombie (1967)
2.
Dauer (1983)
3.
Dauer (1987)
4.
Katamba (1989)
5.
Kurlova and Krustev (1989)
6.
Laver (1994)
7.
Misheva (1991)
8.
Misheva (ms)
9.
Pettersson and Wood (1988)
10. Pike
(1945)
11. Roach
(1982)
12. Tilkov
and Boyadzhiev (1990)
13. Yordanova
(1987)