Hirst, Daniel and Albert Di Cristo (eds.)(1998):
Intonation Systems. A Survey of Twenty Languages. Cambridge:
Cambridge University Press. 499 pages
reviewed by Johannes Reese, University of Muenster, Germany
Twenty-six researchers, among them famous names like Dwight D.
Bolinger and Dafydd Gibbon, contributed to this collection. They
made up a volume describing the intonation systems of twenty
languages, as can be seen already by the title.
Two of them are described in two chapters and varieties, sharing
almost everything but (a few features in phonology and) intonation,
i.e. American and British English and European and Brazilian
Except two of them, all chapters have a similar construction:
1.1 General prosodic characteristics
1.2 Theoretical background and approach
2. Description of the intonation patterns
2.1 Description of a basic non-emphatic pattern
2.2 Mode and expressivity
2.3 Focalisation and contextual effects
2.4 Phrasing and textual organisation
2.5 Other patterns
3. Comparison with other intonation systems
3.1 Comparison with other dialects
3.2 Comparison with other languages
4. Implications and conclusions
This makes the overview very consistent and comparable, besides,
it improves readability.
The chapters deal with (authors in brackets):
American English (Dwight Bolinger)
British English (Daniel Hirst)
German (Dafydd Gibbon)
Dutch (Johan 't Hart)
Swedish (Eva Gaarding)
Danish (Nina Gr/onnum)
Spanish (Santiago Alcoba and Julio Murillo)
European Portuguese (Madalena Cruz-Ferreira)
Brazilian Portuguese (Joao Antonio de Moraes)
French (Albert Di Cristo)
Italian (Mario Rossi)
Romanian (Laurentia Dascalu-Jinga)
Russian (Natalia Svetozarova)
Bulgarian (Anastasia Misheva and Michel Nikov)
Greek (Antonis Botinis)
Finnish (Antti Iivonen)
Hungarian (Ivan Fonagy)
Western Arabic (Morocco) (Thami Benkirane)
Japanese (Isamu Abe)
Thai (Sudaporn Luksaneeyanawin)
Vietnamese (Do The Dung, Tran Thien Huong, and Georges Boulakia)
Beijing Chinese (Paul Kratochvil)
The first chapter is written by the editors, describing the way to
go. At first, they confront the situation in the field of
intonation with that of other linguistic fields, the poor research
situation in this respect makes the collection a milestone in
They foresee a great future for this science, as in automatic
speech production the lack of intonation hinders progress, i.e.
understandability of technically produced spoken language, a
point often repeated in the conclusions of the chapters. It is
especially difficult, as they further argue, to do research in
this area without begin native or near-native speaker of the
language in question. This is an argument for choosing
the publication form of a collection. As intonation research is
still in its toddler years, their project was not a typological
relevant overview of the world's intonation systems, but rather a
first step towards comparison of intonation systems.
Facing this modest pretension, their choice of languages is
remarkable, including two tone languages (Chinese, Vietnamese),
one pitch accent language (Japanese), one language that
knows syllables without vowels (Moroccan Arabic),
near-tone languages like Swedish and Danish, various stress
systems and various ways of rhythm in speech production.
The field of research is intonation proper, that means principally
non-lexical pattern of fundamental frequency (F0) movement.
The editors want to emphasize on the differences between
the intonation system in order to finally find the common base,
which they leave to future research.
As a practical step, they developed a system of transcribing
intonation (INTSINT), which is used by the majority of the
Instead of using the "intonation orthography" provided by
the editors, Bolinger prints the ordinary letters of English with
rise of pitch represented by higher printed letters and vice versa.
His parameters for detecting intonation patterns are interest
("new information"), power, and closures ("pause").
A concatenation of interest highlighting makes high "power".
Questions tend to have higher overall
level, exclamations favour high level followed by fall.
The bearers of intonational profiles are stressed
syllables, unless the urge to emphasize makes the speaker put more
power into an utterance than there are stressed syllables.
In longer words, word stress patterns may deviate from the
"official" rules, especially the first syllable tends to receive
The description is not based upon empirical investigation.
For Hirst, what we know as non-syntactically marked
questions are no questions, but "requests for information".
Hirst rejects the notion of focus inside intonation research, he
prefers the notion of emphasis, because they are pragmatically rather
than syntactically motivated. Intonation Units tend to be not more
than eight words long, causing longer syntactic units to break into
up to eight words long Intonation Units, only in WH-questions breaking
is restricted. syntactic units are a kind of target breaks for the
latter. Hirst creates a tree model for English intonation. Therefore,
he needs a bit more effort to build
it up for the downstepping RP pattern, as compared to the American
(and Scottish) recurrent pattern.
Gibbon states that the overall functional load of
intonation is smaller than in languages like English due to the
extensive use of discourse particles.
He claims there to be a number of regional standards.
Gibbon detected a stylised hesitation pattern.
Southern dialects usually have right-displaced prominence peak,
whereas in Standard German the peak is on the prominent syllable.
In his point of view, the intonation patterns of German are coined
by its word order, the verb-final structure hindering final nuclei
(=most prominent syllable of the utterance).
Pitch modulation in German is generally smaller than in English,
due to the less functional load.
't Hart only describes one out of six possible Dutch intonation
patterns, the "hat pattern". Longer utterances follow the same
patterns as shorter ones, but they start at a higher pitch level,
whereas any utterance ends at approximately the same level.
As for the functional load of intonation, 't Hart describes
Dutch as having a lot of free variation within a pattern
type, though there are certain restrictions to non-final falls.
Focalisation can be achieved by reducing the number of rises.
Gaarding observes the textual level of Swedish intonation as well.
Sequences of sentences can be attached to each other by similar
principles as within sentences. The main textual features
thus expressed are "additiveness", "equivalence", and
"adversativeness". The arguable tone structure
of Swedish does not prevent it from being subject to focus,
attitude/emotion, and interrogation being expressed by
intonational means. Focal accent is achieved by leveling intonation
after the focus accent, whereas in focusless sentences the last
accented syllable receives most weight. The same is true for
question, where rising begins at the focus point or, in lack
of such, from the start. She gives some hints to the effect of
emotions, in her examples, angryness widens the range at the
beginning, happyness the one at the end of an utterance.
Also the notion is not used, in Danish we find right-displaced
prominence peak again. What is more, the st/od seems to have an
impact on intonation, in that it inserts an additional fall into
the pattern. The striking observation of the article is the
syllable structure in Danish, which puts an intervocalic consonant to
the preceding rather than to the following syllable, at least as
far as the intonation movement is concerned. Again, focalisation is
achieved by reducing the pitch of non-focal accents. Questions
show a horizontal contour as opposed to the declining declarative
one. Declination (=pitch level drops during the utterance)
is also found beyond sentence level. A
sequence of sentence shows ever more declining.
Spanish is said to be a trailer-timed language, i.e. the stressed
and the preceding syllables form a unit. The authors presume tonic
groups (TG), which correspond roughly to grammatical phrases;
these are the basic units for intonation.
In declarative sentences, there is usually falling terminal juncture.
Yes/no questions are opposed to WH-questions, as the latter have
the normal declarative pattern, the former coincide with the
non-terminal pattern, marked by a rising terminal juncture.
Focalisation is mostly expressed by syntactic structure, the focus
is put to the end. In opposition to that, the T stress (=high
tone) appears more often outside their ordinary initial position,
which is interpreted as a signal of emphasis.
The most common European Portuguese
contour is a low-falling tone. It is to be found both in
statements and in WH-questions. In the latter, the high level
pre-nuclear pattern marks the asking character of the utterance.
Low-rise occurs both in yes/no-questions and expressive
utterances. Principally, there is end
focus like in Spanish (making use of the freedom of word order in
Portuguese), unless grammatical constraints forbid this, in which
case the nucleus can be non-final. In this case, however,
intonation groups usually break into two, resulting in an end focus
in one of them again. Low-rise marks incompleteness, too.
In Brazilian Portuguese, too, there is a end focus tendency. Its
patterns are similar to the European Portuguese one, except that
European speakers prefer a fall-rise in yes/no questions, where
Brazilians use rise on the tonic syllable, falling down in
Unlike many other languages, French knows a Received
Pronunciation. Besides the Tonal Units and
the Intonational Units, Di Cristo introduces an intermediate
level, which he calls the Prosodic Word; this seems not always to
occur; it contains exactly one primary stress and is motivated
mostly by the lack of account so far of the "temporal factors",
i.e. syllable (vowel) length, which may be more in stressed
syllables. Ordinary declarative intonation is rising-falling or
"circumflex", to be found in Intonation Units.
There is a useful section on regional variation of French.
Italian intonation standard is less well defined, so Rossi in his
paper on Italian intonation takes some regional intonation
patterns, which comes along with some empirical problems.
He introduces a new term, intonation morpheme or
intoneme, describing the syllable that carries the intonation
relevant pitch. Italian intonation is marked by its penultimate
word stress; whereas in French a syncretism of lexical stress (AC) can be
found, coining the intonation system, in Standard (i.e. Tuscan)
Italian both are kept apart. That is why in dialects with ultimate
lexical stress intonation patterns are found that are close to the
French ones; the same is valid for the speakers of those dialects
if they speak the properly penultimate Standard Italian.
Romanian focuses by means of high level in statements
and WH-questions and low level in yes/no questions.
The default rising-falling pattern marks the unmarked
declarative sentence and is independent of
information structure. Declination can be observed. The rising
intonation is the only marker for yes/no questions, other
question-marking elements are redundant. This final rise occurs on
the last stressed syllable.
Svetozarova has observed that Russian intonation
adds a lot of information to sentences. Information structure is
as it were the least of them, it distinguishes various
communicational types of utterances; as was expected,
yes/no questions are merely marked by intonation, i.e.
"a sharply rising high tone at the beginning of the stressed
syllable of the word receiving main stress", followed by
deep lowering. Part of the information is revealed by
divergences from the neutral sentence stress, which
is rule-based, though barely investigated. Some of these
divergences can be observed in separating the prosodic
subdivisions, which play such a great role in Russian syntax.
Following the Russian
tradition, they call prosodic units syntagma and investigate
average syntagma length. Syntagmata represent semantic/syntactic
units and can be separated by their perceived pauses, which may be
real or subjective, i.e. based on abrupt changes in intonation
flow. The authors divide two different accents, the focal one,
which they call semantic accent, and the dichotomy of continuation
vs. finality. Among syntagmata, only a quarter contain more than
one semantic accents, in that case there are normally two. Accents
vary in degree, though there is no difference in quality, as to
different kinds of information that could be revealed by the
accents. The most remarkable feature in the article is the
observation that there are certain words that attract the accent
to a neighbouring word, either preceding or following. One of
those is the negation particle, which is almost invariably
followed by an accent.
Like Danish, Greek knows no neutral intonation; intonation has an
semantic impact. Botinis emphasizes on the stress pattern of
Greek; stress patterns coin the prosodic manifestation of an
utterance. WH-questions show no difference in pattern to
declarative utterances. He even doubts yes/no question, which
receive no syntactic marking, being necessarily marked by
intonation; this is understandable only in terms of his discourse
point of view. The "famous" continuative intonation , e.g., has
the function of keeping the turn in his interpretation. So the
basic and autonomous units of description are turn-units
and their subdivisions, the STU (sub-turn-units).
In Finnish, dialects play a vivid role, yet they are influenced by
the standard. Compounds may contain
morpheme boundaries within them, marked by intonational means and
laryngealisation. Creaky voice can be found as final marker of
utterances as well, as the end point of a declination. Information
structure is part of the intonation encoding, Iivonen
differentiates between accent for rheme (AR), accent for contrast
(AC), and accent for emphasis (AE); the latter extends to more
than one syllable. Striking is the feature of delayed F0 peak,
which has a function of hinting, assertion plus inherent doubt; it
means that the intonation peak is about 100ms delayed, yielding
the second syllable or the end of a long vowel as peak. Particular
questions (=WH-questions) are marked be a high start and then
constant falling, general questions (=yes/no questions) be high
level until the focus, with flattening onwards.
Fonagy bounds traditional intonation research to the modern one.
He cites the 5 tone levels introduced by Varga for Hungarian and
connects them to empirical investigation. He uses whispering for
control. Whereas simple focalisation is recognizable in whispered
speech, thus outside the domain of intonation, focalisation with
implication (i.e. contrast) is only encoded by "melody", i.e.
intonation. Attitudes are "lexicalized" and follow
special patterns, for which he gives one example (implication of
hurry), emotions are expressed by "expiratory, laryngeal, and
articulatory strategies" which are partly paralinguistic.
Western Arabic (Morocco)
Surprisingly, Benkirane has the very unusual hypothesis that
releasing of plosives leads to the conclusion that there are
syllable-initial. A final schwa is said to be optionally
added to any (otherwise defined as) final consonant.
In terms of duration, closed syllables are as long as two open ones
(of course, in Benkiranes point of view, closed syllables are two
open ones, too). The theory allows to say that both word and sentence
stress are on the penultimate syllable. Besides, Moroccan Arabic
shows all the characteristics that have been so common while
reading the volume: focalisation pitch peak, focalised WH-words,
intonational theme-rheme-marking, splitting up of longer
utterances. As a footnote, there is a focalisation particle in
Moroccan Arabic, too.
The study of Japanese intonation faces an empirical problem: As a
pitch-accent language (Cruttenden 1986) its word stress
interferes with the physical appearance of intonation, rather than
"working together", as in the "intonation languages" (Cruttenden
1986). The results, though, do not differ too much from those of
languages discussed in the earlier chapters. Questions are marked
by rising intonation on the final element ka, making the latter
redundant. What is more, Abe points out some interesting
\methods to go beyond the borders of present-day intonation
research. By examining one-word-utterances,
he shows a range of meaning changes imposed to segmentally the
same utterance by different pitch levels and contours. There are
several constraints stemming from the word accents, though, that
point to the fact that pitch serves for two purposes in Japanese.
Although it is a tone language, Thai has
stress and information which is encoded in intonation. Along with
the five lexical tones of Thai, five different pitch behaviours
occur for each of the presented intonation patterns. Some of
the patterns cause tones to become less salient. Intensity proved
to be a cue in the physical recognition of tone or intonation,
especially for the low ones and particularly in whispering.
Luksaneeyanawin is the only one who asks the question what stylised
intonation is good for; his answer is summarized as
"predictability". The basic intonation group and focalisation
pattern applies equally to Thai as to non-tone languages. In
comparison to statements, other forms of intonation, which are due
to semantic constrasts or to attitudes, are
accompanied by a narrower pitch range, except "Tune 4" (which widens
pitch range), conveying "emphatic, agreeable, interested,
or believing attitudes".
Vietnamese has a wide range of "pragmatic" particles, good for
expressing the same as intonation normally does.
Yet intonation is used the same way as in other languages.
The results concerning the behaviour of tones are very
detailed, in an impressionistic attempt to summarize:
Different tones behave differently in different environments.
They are somehow "attracted" by the intonation movement. Register
tones are leveled, low tones widen their range, high tones are more
strongly marked in interrogatives. Yet intonation contours seem
to be a bit more "cautiously used" in Vietnamese than in other languages,
including Thai: In questions, e.g.,
there is normal declination until the proximity of the
sentence final question marker, where the rise begins; the overall
register is yet higher than in declaratives.
Chinese is on its way towards a foot structure, partly due to
the development of the lexicon. Accent is normally marked by
relatively higher pitch, but with the third (falling-rising)
tone, it is marked by lower pitch. Kratochvil states that
for describing tone, pitch, duration, and amplitude must
be taken into consideration. The perception of Chinese
tone needs 50 to 100 msec. to be distinguishable. Slower
speech allows more sticking to the intonation pattern.
Focusing enlarges the channel. As they bear no
tone, Intonation carriers (final sentence particles)
allow to sort of "save" the whole intonational information
or protect the tones before.
Nearly all of the authors describe stylised patterns for the
language in question. Dialectal variation varies across the languages.
Nearly all intonation patterns convey thematic relations.
The range of information borne by intonation is only slightly
different among the described languages.
It takes some time to get used to the INTSINT system, fortunately,
most researchers add some F0 curves, whereas the Bolinger
system fits intuition, but the problem is that "mute" letters get
a pitch mark as well. Romance languages and Arabic
partly adapt their syntactic structure to intonation
patterns, a typologically very relevant observation.
Especially the description of the tone languages is very
revealing. The method of whispering occurs in a number of
contributions, especially in the chapters describing tone languages;
it seems that power or amplitude can substitute F0
movement, wherever intonation gets into trouble by lack of voice
(whispering) or interference with tone. In cooperation with tone,
tone allots any device of intonation its F0 movement. Fortunately,
the Vietnamese and Thai chapters clearly show how the can be
specified. Kratochvil walks into another direction.
His results seem to be more theoretically relevant than empirical.
He embeds Chinese tone into a non-tonal pattern, showing how tones
settle down in the framework set by intonation, without losing
their distinctiveness. I have tried to represent the merits of the
other authors in the summary of their chapters
above. They provide us with a clearer picture of the similarities
and little differences of the languages and dialects. Especially
for French, English, Italian, a useful grid of dialectal diversion is
built. Abe shows for Japanese that the assumption that
pitch-accent languages have less freedom for using intonation
(Cruttenden 1986) can not be supported. The Russian chapter and the
Dutch and German in comparison to the "Englishes" seem
to show that there are languages that stick out for their
excessive use of intonation or at least their intonation
use is more overt.
The only inconvenience for the reader is provided by the fact
that the authors use different terminologies.
The interaction of the contributions draws a clear and
comprehensive picture of how intonation works, what seems to be
universal and what liberties a specific dialect has.
Certain interactions have been dealt with:
tone and stress, tone and intonation, intonation
and syntax, stress and intonation, pitch and power.
Cruttenden, Alan (1986): Intonation (Cambridge textbooks in
linguistics). Cambridge: Cambridge University Press
Johannes Reese has a M.A. degree in linguistics obtained at the
University of Osnabrueck, Germany. After a short time of working
in an Internet company, he got a position at the English
department of the University of Muenster, Germany,
where he has just started to write his Ph.D.-thesis,
which will be in the field of translation
studies. His interests lie in all fields of
linguistics that are relevant for facilitating language learning,
especially syntax, phonetics, and typology.