9/16/2010

1

Amalia Arvaniti

University of California, San Diego

TIE4, Stockholm, 9-11 September 2010

Research funded by CNRS-Lacito and WUN (through UIUC)

There is no “the current model” of intonation

Researchers do not seem to agree on basics:

o What are the primitives of intonational structure?

o Is intonation a phonetic or a phonological phenomenon?

o Do we need a phonology of intonation, and if so, of what form?

o What is the function of intonation?

2

Explore these issues as they play out in the analysis of Romani

Introduce the language and briefly sketch its prosodic system Concentrate on aspects of

o phonetic realization and phonological representationo the function of intonation in marking focus and expressing

information structure

Discuss how Romani challenges received ideas but can also answer some of the outstanding questions

Discuss how we may achieve a better understanding of the nature of intonation cross-linguistically

3 4

Romani is Indo-Aryan language with extensive dialectal variation

The Komotini variety is a Vlach Romani variety

The speakers from Xanthihave more diverse backgrounds

The prosodic patterns described here appear valid for both areas

Speakers are trilingual in Romani, Turkish and Greek

Romani is the language of the home and the community

Turkish and Greek are used mainly for trade and other professional activities

Most speakers have little or no formal schooling in any of their languages; they are certainly not literate in Romani

5

Photo: E. Adamou

6

Ten speakers of both genders and various ages Approximately 45 minutes of speech

Photo: Roma children for Camera Project

Genres▫ Spontaneous conversations

between two or more speakers▫ Conversations with Adamou▫ Story telling

http://lacito.vjf.cnrs.fr/archivage/languages/Romani_fr.htm

▫ Elicitation with Adamou (QUIs) Broad transcription, AM

analysis and glossing in PRAAT

9/16/2010

2

Romani has word-final stress in its native vocabulary

Stress may appear on the penult or antepenult in loan words asteˈnava “hospital” korˈkodilo “crocodile”ˈapo “pill” ˈapora “pills”

7 8

σp.a.

Accent before “continuation rise”

Default (nuclear and prenuclear) accent

Nuclear accent in polar questions

Declaratives, wh-questions

Continuation rises, wh-questions Polar questions

Wh-questions

L-L%

H-H%

L-H%

H-L%

H*

L*

L*+H

9

Few studies are available on this issue: o Ohala & Ewan (1973)o Sundberg (1979)o Caspers & van Heuven (1993)o Xu & Sun (2002)

Xu & Sun (2002) o claim that their results have established the physiological

limits of pitch speed and excursion sizeo argue that given these physiological limitations, AM

assumptions about the coordination of segments and tones are untenable

10

All based on laboratory speech

ka dikhav tut

will look.after you

H* L-L

Time (s)0 1.242

Pitc

h (

Hz)

100

500

11

13 ST or 45 ST/s

X&S,2002: 5 ST excursion; 40.1 ST/s speed

bas-ter i koptʃa viiiii

laughterpress.CAUS.IMP.2SG the button (imitates sound) laughter

H* H*

Time (s)0 2.428

Pitch

(H

z)

100

600

12

17 ST or 50 ST/sX&S,2002: 5 ST excursion; 40.1 ST/s speed

9/16/2010

3

ni daras tʃe sabia

NEG scared INTERJ. Sabiha

L*+H L*+H H-L

Time (s)0 1.539

Pitc

h (

Hz)

50

450

13

9.7 ST or 68.3 ST/s7.1 ST or 78.8 ST/s14 ST or 70.4 ST/s

X&S,2002: 5 ST excursion; 40.1 ST/s speed

Female Romani speakers are often above published physiological averages (and limits)

There are no physiological reasons for this We know that

o pitch range shows variation depending on language, culture, and gender (e.g. Bezooijen 1995)

o pitch speed can be quite variable (e.g. Arvaniti & Ladd, 2009)

Limits established in the laboratory may not be representative and cross-linguistically valid

It is premature to conclude that we have established the physiological limits of pitch production

It is even more premature to base our understanding of the phonetics of intonation on said limits

14

Studies have shown that the use of plateaux is extensive:

o British English (Knight & Nolan 2006)

o Neapolitan Italian (D’Imperio 2000, D’ Imperio et al. 2000)

o Italian and German (D’Imperio et al. 2010)

Turning points (TPs) have been seen as a cornerstone of AM

The importance of TPs is questioned by plateaux: intonationalevents may not be as localized as TPs suggest

Questioning TPs can be interpreted as evidence against AM principles of phonetic organization of intonation

15

ama ka dʒas kaj pazari P kin mange P jek momia

if will go at market P buy to-me P a scarf

H* H* L* H-H H* L* H- H* H* L-L

Time (s)0 3.698

Pitc

h (

Hz)

50

450

16

o adla(n) senerdoktor si

Adnan psychiatrist is

H* H* L-L

17

ama phagindol P kazanamadum xasardom

if breaks I win.NEG I lost

H* H* H-H H* H* L-L

i phabaj

the apple

H* H-H

ka sarɯ majmuna

to yellow monkey

H* H* H-H

i kafe pasta P kaste te dav

the brown cake to,who to give.1SG

H* H* H-H H* L-L

18

9/16/2010

4

The alternation in the Romani data suggests that plateaux may be simply a matter of phonetic realization:o plateaux are frequent but they alternate with TPs in structurally

and pragmatically similar contours

Tones are realized either as points or as levels Languages may prefer one or the other realization Why such differences exist and how are they produced? Conventions for how to measure “flat” tones are required and

how to understand their coordination with the segmental string But this does not denigrate the relevance of TPs as possible

reflexes of underlying tones or more general principles of phonetic organization of intonation

19

It is often assumed that tonal patterns should have

o transparent representations

o similar (if not identical) representations across linguistic

systems (Ladd 2008, chap. 3)

Although transparency is certainly desirable, it may not always be possible, for various reasons:

o economy in the representation of a system (cf. Arvaniti et al. 2006)

o the system does not allow for cross-linguistic similarity in representation (cf. Arvaniti & Baltazani 2005)

20

dui tʃhiriklja lole

two birds red.PL

H* H* L-L

amen laʃti amen dara(t)sa

we cannot we scared

H* H*L- H* H* L-L

kaj solako

on left

H* L-L

ama tu(t) but ʃukar san

but you very pretty you-are

H* H* L-L

21

The H* pitch accent of Thrace Romani shows several variantso peako (short) plateauo riseo fallo rise-fall

Preferable to keep the representation simple and look for context-related reasons for the variation, such as o the prosodic position of the accent in the phraseo the location and nature of adjacent toneso the segmental context in which a tone is realized

22

The phenomena of Romani examined here suggest that there is more variation in intonational realization than is often expected

This is often seen as a problem with intonation However, intonational data

o are not more variable than “segmental” aspects of the speech signalo variation is equally lawful and probably due to the same overall

organization

Our expectations, however, appear to be different

23 24

Focus and Romani intonation

9/16/2010

5

For narrow focus, Thrace Romani uses combinations of prosodic, syntactic and morphological marking (contra Büring, 2009):o Prosodic and syntactic marking

o Prosodic, syntactic and morphological marking

o Prosodic marking

Prosodic focus marking (on its own) is not necessarily preferred to more “costly” strategies such as syntactic marking (contra Skopeteas & Fanselow, in press)

25from Arvaniti & Adamou, in press

WCCFL 2010

from Arvaniti & Adamou, in press

ka sarɯ majmuna de

to yellow monkey give.IMP.2SG

H* L-H

Time (s)0 1.407

Pitc

h (H

z)

75

250

ka loli majmuna de

to red monkey give.IMP.2SG

H* H* L-L

Time (s)0 1.558

Pitc

h (

Hz)

75

250

26

von da kelen ipolojistes

they FOC play.3PL computers (Q)

L*+H H-L

Time (s)0 1.852

Pitc

h (

Hz)

75

300

mande da si bakre

to,me FOC is sheep

H* L-L

Time (s)0 1.524

Pitc

h (

Hz)

75

300

27

ko rom ka marel tut

your man will beat.3SG you

H* H- L*+H H-L

Time (s)0 1.333

Pitc

h (

Hz)

75

500

mo rom buki ni kerel

my husband work NEG does

H* H* H* L-L

Time (s)0 1.244

Pitc

h (

Hz)

75

500

28

iklan kaj lolo kher da

got,out.2SG at red house FOC

L*+H H-L

Time (s)0 0.9689

Pitc

h (

Hz)

0

250

lazɯm me da te dikhav me romes

must I FOC to look after.1SG my.OBL husband.OBL

H* H* L*+H H-L

Time (s)0 1.865

Pitc

h (

Hz)

0

600

29

si late dui piperke P jek loli ek da sarɯ

is her.LOC two peppers one red oneand yellow

H* H* L-L H* L* H- H* L-L

Time (s)0 3.518

Pitc

h (

Hz)

75

300

30

9/16/2010

6

i m kaj maria si P em loli P em sarɯ P piperja

the M (disfl)on Maria is and red and yellow pepper

H* L* H- H* L* H- H* L-L H* L-L

ka lefter-esko ʃoro opre si em pismara em tawʃano

on Lefteri-GEN head above is and cat and rabbit

H* H* H* H* L* H- H* L-L

el va em kokini em sarɯeh yes and red and yellow

H* H* L* H- H* L-L

31

but traʃano traʃano but

very coward coward very

H* L- H* H* L-L

Time (s)0 1.365

Pitc

h (

Hz)

0

300

but dʒuvlen dikhlom but murʃen dikhlom but dʒenen dikhlom

many women saw.1SG many men saw.1SG many people saw.1SG

H* L-L H* L-L H* L-H

Time (s)0 3.832

Pitc

h (

Hz)

50

450

32

The data from Thrace Romani show that the relationship between information structure (IS) and intonation is indirect and language specific

This relationship may in part depend on the interaction of intonation with other components of the grammar

Listeners, therefore, do not expect direct highlighting of new information by means of an accent

Rather, listeners derive IS by comparing the prosody of an utterance with the system they know

33

Romani provides us with evidence that our understanding of intonational form and function is still limited

Many aspects that we take for granted are questioned by these data

Some obvious answerso Study more languageso Expand research beyond laboratory speecho Study different varieties of discourse, as pattern frequency may

vary with style

34

The phonological nature of intonation is often not fully acknowledged:o the use of ToBI as a transcription systemo Ladd’s call for phonetically transparent representations (Ladd

2008, chap. 3) o Gussenhoven’s “half-tamed savage”

If intonation is seen as a straightforward component of phonologyo context–related variation will be expected and toleratedo a degree of representational opacity will be tolerated both

within and across linguistic systemso function will not be seen as the sole determiner of contrasto sociolinguistic variation will be expected and studied

35

Intonational structure is not dictated by IS

Intonation has its own rules and requirements which on many occasions will go against “common sense”

This point has been made before, but

o it is not always fully acknowledged

o this Romani corpus is among the first extensive natural spoken corpora demonstrating such phenomena

36

9/16/2010

7

These Romani data show that we have a long way to go before we can say we truly understand intonation

It is imperative that o we open up the range of languages and speaking styles examinedo we question even the most common sense ideaso we seriously entertain the possibility that intonation may appear as a

“half-tamed savage” only because we do not yet know its culture

Starting with the assumption that intonation is a phonological phenomenon that is not radically different from all other phonological phenomena should help resolve many of the current puzzles and disagreements and allow us to concentrate on figuring out the rituals of intonational culture

37 38

Thank you!

my collaborator, Evangelia Adamou (CNRS-Lacito)

our speakers,including Sabiha

CNRS-Langues et Civilisations à Tradition Orale (LACITO) and a Worldwide Universities Network (WUN) grant from the University of Illinois at Urbana-

Champaign (with M. Terkourafi as PI) that made the fieldwork and our collaboration possible

Thanks also to