Can pitch accent type convey information status in yes-no 
questions? 
Martine Grice* and Michelina Savino*'i" 
*Institute of Phonetics, FR 8.7, University of the Saarland, 
Postfach 1150, 66041 Saarbrticken, Germany 
1-Politecnico di Bari, Italy 
mgrice@coli.uni-sb.de, esavino@poliba.it 
Abstract 
This paper analyses the intonation of polar 
questions extracted from a corpus of task- 
oriented dialogues in the Bari variety of 
Italian. They are classified using a system 
developed for similar dialogues in English 
where each question is regarded as an init- 
iating move in a conversational game 
(Carletta et al 1995). It was found that there 
was no one-to-one correspondence between 
move-type and intonation pattern. An alter- 
native classification was carried out taking 
into account information status, that is, 
whether or not the information requested by 
the speaker is recoverable from the previous 
dialogue context. It is found that the degree 
of confidence with which the speaker believ- 
es the information to be shared with the inter- 
locutor is reflected in the choice of pitch 
accent and postfocal accentual pattern. Low 
confidence polar questions contain a L+H* 
focal pitch accent and allow for accents to 
follow it, whereas high confidence ones con- 
tain a H*+L focal pitch accent, followed by 
deaccenting or suppression of accents. 
1 Introduction 
In this paper we examine the intonation of 
polar questions. Such questions may elicit an 
affirmative or negative reply concerning in- 
formation which is totally new, i.e. where the 
speaker believes the information is not re- 
coverable from the dialogue context. These 
are often referred to as information-seeking 
questions. They may also; however, refer to 
old information, i.e. that which the speaker 
believes has already been conveyed. These are 
often referred to as confirmation-seeking 
questions. Using a corpus of task-oriented 
dialogues, we investigate how far it is possible 
in such questions to make a clear-cut binary 
distinction between old and new information 
and whether this distinction is reflected in the 
choice of intonation pattern used. 
A distinction has been made in the analysis 
of a dialogue corpus in English (Carletta et al, 
1995) where information-seeking questions 
are referred to as QUERIES and confirmation- 
seeking questions as CHECKS. In the canon- 
ical examples given in English, QUERIES and 
CHECKS are syntactically distinct (Kowtko et 
al, 1992; Carletta et al, 1995); "Do you have a 
rockfall?" and "So you want me to go down 
two inches?", respectively. The former uses 
interrogative and the latter declarative syntax. 
In Italian, on the other hand, where all yes-no 
question types have the same syntactic form 
as declaratives, the intonation pattern plays a 
greater role. "Vado a destra" can be translated 
as "I go to the right" (statement), "Do I go to 
the fight?" (QUERY) or "(So) I go to the 
right?" (CHECK). We therefore investigate 
dialogues in Italian where it is possible to 
isolate the role of intonation patterns. 
Intonation contours can be analysed as 
having two types of tonal specification: tones 
which have a prominence-lending function, 
referred to here as pitch accents, and those 
which delimit intonational phrases, referred to 
here as boundary tones (Pierrehumbert,1980; 
Beckman and Ayers 1994). In Standard Italian 
(SI), it is argued that the boundary tone or a 
combination of pitch accent and boundary 
tone play a role in distinguishing questions 
from statements (Avesani, 1990; Chapallaz, 
1979; D'Eugenio, 1982; Canepari, 1980; 
Agard and di Pietro, 1965). In all cited cases 
of SI yes-no questions have a final rising 
contour or high boundary tone. In a number of 
other varieties of Italian, those spoken in Bari 
(Grice and Savino, 1995) and Palermo (Grice 
29 
1995) for instance, it is solely the pitch accent 
which has the distinguishing function. Our 
analysis here is based on a corpus of Bari 
Italian dialogues, allowing us to concentrate 
on the relation between pitch accent type and 
question type. 
2 Dialogue Corpus 
The corpus analysed consists of task-oriented 
dialogues between six pairs of Bari Italian 
speakers. The task, based on the HCRC Map 
Task (Anderson et al, 1991), involves verbal 
co-operation (via auditory channel only) be- 
tween two participants, each having a map, 
with the aim of transferring as accurately as 
possible a given route from one map to the 
other. There are a number of discrepancies in 
placement and positioning of the landmarks 
on the maps. Since our aim was to examine 
intonation contours, the landmark names con- 
tained mainly sonorants and were controlled 
for word stress pattern. 
3 Intonation analysis 
The intonation analysis employs a modified 
version of the ToBI transcription system 
(Beckman and Ayers 1994) using two tones, 
H (high) and L (low). When they occur in 
pitch accents, one tone is starred, indicating 
association with a metrically strong syllable 
(Pierrehumbert 1980). They may also function 
as boundary markers for one of two phrase 
types: intermediate (or minor), indicated by 
"-" after H or L, and intonation (or major) 
phrase, marked with "%". The pitch accents 
referred to in this paper are L+H*, which 
involves a low pitch target just before a high 
accented syllable, H+L*, which involves a 
high pitch target immediately preceding a low 
accented syllable, and H*+L, a high target 
early in the accented syllable followed by a 
rapid fall (see Grice and Savino 1995 for a 
discussion of peak placement). In addition, H* 
and L* involve a high or low target, 
respectively, on the accented syllable, with no 
specification as to the pitch contour flanking 
it. The boundary tones referred to are L- and 
the combination L-L%, both of which give a 
low pitch value at the end of the phrase, and 
the combination L-H% which gives a slight 
rise up to the end of the phrase, with a low 
starting point at some distance before the 
endpoint. 
4 Question types 
The questions occurring in our corpus are 
described with the coding scheme for 
conversational games used to describe the 
English HCRC Map Task corpus (Kowtko et 
al 1992, Carletta et al 1995). Conversational 
games are sequences of acts, referred to as 
moves, such as the possible sequence of 
QUERY-REPLY-ACKNOWLEDGE moves 
within a QUERY game. Since each move 
within a given game may have a distinct 
intonation pattern, we confine our analysis to 
individual moves. 
The analysis here concentrates on 
QUERIES and CHECKS as described in 1 
above, both of which are initiating moves 
within games of the same name. The Map 
Task coding scheme has another question-type 
where the speaker is attempting to get 
evidence that the transfer was successful, so 
that s/he can move onto the next game. This is 
called an ALIGN. Examples of ALIGNS 
might be "Have you drawn it?" or "OK?". 
They may seek information about new or old 
material within the discourse and may thus 
pattern with either of the above-mentioned 
question moves. Another move-type found in 
our corpus is not categorised in the current 
Map Task coding scheme, but may be fitted 
into the framework as a responding move 
which indicates that the communication has 
been unsuccessful. We refer to this move as 
OBJECT. It is used to point out that there has 
been a break-down in communication, such 
that the game cannot continue until common 
ground is re-established. OBJECT moves 
contrast with ACKNOWLEDGE moves. 
These latter indicate that communication has 
been successful, and encourage the 
interlocutor to proceed with the game. 
Examples of OBJECT moves discussed below 
are of the type that are categorised elsewhere 
as 'echo questions' (inter alia Cruttenden 
1986), because they echo, or repeat, all or part 
of what has just been said by the interlocutor. 
Because these types of OBJECT move are 
considered to be a category of question in the 
intonation literature, they are analysed here 
alongside moves of questioning force. How- 
ever, since they could be responding within 
30 
one game as well as initiating another (sub)- 
game, they cannot be classified as simple 
questions, which have only an initiating funct- 
ion (Carletta et al, 1995). 
5 Intonation and moves 
The intonation contours used for each type of 
move will be discussed and exemplified 
below, along with the dialogue context from 
which the example has been excised (where G 
is the route giver and F the follower). A 
nulnber of examples are accompanied by F0 
traces. The basenames of the associated 
speech and F0 files are given in angled 
brackets. 
QUERY Moves - QUERIES have a rising- 
falling intonation pattern, rising up to the 
accented syllable and down from it, tran- 
scribed as L+H* L-L%. The final boundary 
tone may be H% instead of L%. An initial H* 
pitch accent (PA) is optional; it is not present 
in all examples of QUERY move. 
1. Hai aGNELlo? 
H* L+H* L-L% 
context: 
G: hal agnello? 
(do you have lamb?) 
F: no 
If the accented syllable is final in the phrase, 
the fall on that syllable is curtailed 
substantially, as illustrated in Grice et al 
(1995). If the focus (see e.g. Gussenhoven 
1983 for a definition of focus) is not on the 
final lexical item in the phrase (henceforth 
referred to as non-final), the PA on the focus- 
sed item is followed by an intermediate phrase 
(ip) boundary. The rest of the question con- 
stitutes another ip which has a reduced range 
and has an identical PA and boundary tone. 
That second ip is similar to what has been de- 
scribed by Crystal (1969) as a subordinate 
tone group. For a theoretical account see 
Grice (1995) which refers to analogous cases 
in Palermo Italian ~ 
2. <un-giro> fa un Glro (foc) intorno all'alBERgo? 
L+H* L- L+H* L-L% 
1 The interrogative PA in Palermo Italian is L*+H, not 
L+H* as in Bari Italian. 
context: 
F: ... no, aspetta, scusa, la la strada passa da sotto, hai 
detto 
(...no, wait a minute, sorry, the the path goes 
underneath, you said) 
G: si °, da sotto l'albergo Malaga 
(yes, under hotel Malaga) 
\[pause\] 
F: ecco, e poi la la il devo questa strada devo farla ehm 
tocc.., devo devo 
ehm dove dove la faccio terminate? 
Fa un GIRO intorno all'albergo? 
(right, and then the the I have to this - path - I have to 
erm touch... I have to erm where where do I make it 
finish ? Does it go ROUND the hotel?) 
G: ecco, 
(right,) 
The above case involves what appears to be 
reaccenting, as discussed by Cruttenden 
(1993), on the word albergo (hotel), which is 
given in the dialogue context. In cases where 
the subject occupies sentence-final position, it 
also constitutes a subordinate intermediate 
phrase with its own accent, the focal PA being 
in the first ip, as in 3. 
3. era numeRAto questo perCORso? 
L+H* L- L+H* L-L% 
context: 
G oh adesso segna la croce alia sinistra del lago 
Anomalo e siamo arrivati 
(Oh now put the cross on the left of lake Anomalo and 
we're there) 
F era numerato questo pereorso? 
(was it numbered, this path?) 
G no 
CHECK Moves - The intonation of 
CHECKS is influenced by the degree of 
speaker confidence as to the correctness of the 
inferred material it contains, that is, the degree 
of speaker confidence that the material is old. 
CHECKS which are tentative may have an 
intonation contour which is indistinguishable 
from that of QUERIES; in terms of 
information structure, neither contain 
information which is confidently deemed by 
the current speaker to be old. An example of a 
tentative CHECK is in 4 below. 
4. <curvR> 
QUINdi anche QUI (foc) c'e' una leggera CURva? 
H* L- L+H*L- L+H* L-L% 
(So there's a gentle curve even HERE?) 
31 
Confident CHECKS have a different pitch 
accent: H*+L. Moreover, there is no 
reaccenting of post-focal material. Where the 
focus is non-final, the focussed pitch accent is 
followed by a word receiving a strong 
prominence without a pitch excursion (in 
British School terms, the appropriate syllable 
in the word is stressed but not accented; 
Crystal 1969). We refer to such a prominence 
as a suppressed accent. In example 5 below, 
there is an accent on "devo" and a suppressed 
accent on "scritta". 
5. <non-devo> non DEvo andare verso la scritta 
H*+L L% 
context: 
G: ... continua sempre come grossa curva intorno tutto 
al lago Anomalo fino a un certo punto. Ci sei con la 
mente? 
(...keep going in a big curve around lake Anomalo until 
a certain point. Are you with me ?) 
F: No. Altora, il lago Anomalo devo ehm praticamente 
non DEVO andare verso la scritta 
(No. So, lake Anomalo, I have to erm practically... 
I don "t HA VE to go towards the wriang) 
An equivalent syntactic structure to that in 
example 3, where the subject is sentence-final, 
is in 6 below. In both examples 5 and 6, the 
choice of H*+L pitch accent is motivated by 
the perceptual impression of a sharp fall on 
the accented syllable. In fact, the F0 peak 
occurs very early in the accented syllable in 
contrast to the F0 peak on L+H* pitch accents 
which generally occurs in the second half of 
the syllable. 
It is unclear whether the analysis should 
include a L- boundary after "devo", with the 
implication that what follows is an ip with no 
pitch accent, a combination so far only 
postulated for tags. Since without further 
analysis, for example of duration data, there is 
no obvious way of distinguishing between the 
two transcriptions, we opt provisionally for 
the analysis without an ip boundary. 
6. <tutto-sommato> 
Quindi sono tutto sommato viCIni (foc) 
H*+L 
banane e meloni e lago anomalo. 
L-L% 
(So, all in all, they are close, bananas and melons 
and lake anomalo. ) 
ALIGN Moves - ALIGNS of the type "Va 
bene?" or "Ci sei?" (Are you with me?) are 
realised with the same intonation pattern as 
for QUERIES. Although example 7 below has 
a L% boundary tone, they may have either L% 
or H%. 
7. <va-bene> va BEne? 
L+H* L-L% 
context: 
G: allora quindi il percorso lo facciamo al contrario. 
Va bene? 
(well so we'll do the route backwards. 
Alright?) 
These constitute the majority of ALIGNS. 
However, they may also involve seeking con- 
firmation of inferred material. The intonation 
pattern used in these cases depends on the 
speaker's degree of confidence as to the cor- 
rectness of his/her inferral, as in the CHECK 
category above. 
OBJECT moves In Bari Italian, the 
OBJECT moves we have examined have the 
same tonal analysis as QUERIES of the yes- 
no kind. It is not the tonal analysis but rather 
other parameters which appear to distinguish 
OBJECTS from QUERIES, viz. "breathy" 
voice quality and/or expanded range, both of 
which can signal incredulity. An example 
from the corpus is in 8. 
8. <anima-mia> Anima rnla 
(L+)H* L+H* L-L% 
context: 
G: C'e l'hai il ristorante Anima mia? 
(Do you have it, restaurant "Anima Mia"?) 
F: Anima Mia?! 
It might be argued that although the actual 
lexical material in OBJECT moves is old, the 
degree of speaker confidence that the 
information is shared is low. In example 8, the 
speaker is, amongst other things, conveying 
the fact that the interlocutor's assumption that 
the information is shared is incorrect. 
A summary of the moves and their pitch 
accents is in table 1. It is clear that in two 
types of move either L+H* or H*+L can be 
used, thus precluding a one-to-one corres- 
pondence between move-type and pitch 
accent. 
32 
Move L+H* H*+L 
QUERY 
CHECK 
+ 
+ + 
ALIGN + + 
OBJECT i + 
Table 1 Nuclear Pitch Accents in question moves 
6 Non-finality 
For completeness, example 9 is given below 
of a command, referred to as an INSTRUCT 
move, which is non-final in a list of 
commands. 
9. <legger> 
aDESso PIEga verso il BASso leggerMENte 
H* L- H* !H* L* L-H% 
(Now turn slightly downwards,) 
This pattern is similar to that given for 
'questions' in Standard Italian, but is distinct 
from any of the patterns found for the 
questions examined here for Bail Italian. An- 
other non-final pattern (H* H-H%) is 
discussed in Grice et al (1995). 
7 Conclusion 
In this dialogue corpus study, we have ob- 
served that conversational game move types 
do not necessarily have a specific intonation 
contour; CHECKS and ALIGNS may have 
either of two distinct types of pitch accent. 
QUERIES and OBJECTS have the same pitch 
accent, but differ in other parameters. How- 
ever, we have found that pitch accent type re- 
tlects the degree to which the speaker believes 
whether the current question contains shared 
material. Where the speaker is lacking in con- 
fidence as to whether the material is old or 
not, the same pitch accent, L+H*, accompan- 
ied by reaccenting of post-focal material, is 
used as when it is totally new. The fact that 
OBJECTS contain repeated lexical material, 
although they pattern with utterances referring 
to new information, is explained by the fact 
that this repeated material is considered to be 
erroneously regarded as shared with the inter- 
locutor. 
It therefore appears that speakers do not use 
pitch accent type to distinguish between 
totally new and possibly new material, but 
that they do distinguish between those cases 
and ones where they are reasonably confident 
that material is old. Further research using 
perceptual testing is planned to confirm this 
hypohesis. 

References 
Agard, F.B. and R.J. di Pietro. 1965. The 
sounds of English and Italian. University 
of Chicago Press. 
Anderson, A.H., M. Bader, E.G. Bard, E. 
Boyle, G Doherty, S Garrod, S. Isard, J. 
Kowtko, J. McAllister, J. Miller, C. 
Sottillo, H Thompson, and R. Weinert. 
1991. The HCRC Map Task Corpus, 
Language andSpeech, 34, 4, pp. 351-366. 
Avesani, C. 1990. A contribution to the 
synthesis of Italian intonation. Proc. ICSPL 
90, vol 1, pp. 833-836. 
Beckman M. and G. Ayers. 1994. Guidelines 
for ToBI transcription, version 2. Ohio 
State University. 
Canepari, L. 1980. Italiano standard e 
pronunce regionali.. Padova:Cleup. 
Carletta, J. A. Isard, S. Isard, J Kowtko, G. 
Doherty-Sneddon, and A. Anderson. 1995. 
The coding of dialogue structure in a 
corpus. Twente Workshop on Language 
Technology on Corpus-Based Approaches 
to Dialogue Modelling, Twente, The 
Netherlands. 
Chapallaz, M. 1979. The Pronunciation of 
Italian: A practical introduction. London: 
Bell and Hyman. 
Cruttenden, A. 1986. Intonation. Cambridge: 
CUP. 
Cruttenden, A. 1993. The Deaccenting and 
Reaccenting of Repeated Lexical Items. 
ESCA Workshop on Prosody 1993, 
Working Papers 41, Dept. Linguistics, 
University of Lund, Sweden. 
Crystal, D. 1969. Prosodic Systems and 
Intonation in English, Cambridge: CUP. 
D'Eugenio, A. 1982. Major problems in 
English phonology with special reference 
to Italian-speaking learners, Foggia: 
Atlantica. 
Grice, M. 1995. The intonation of interr- 
ogation in Palermo Italian; implication for 
intonation theory. Niemeyer, L.A. series. 
Grice, M. and M. Savino. 1995. Low tone 
versus "sag" in Bari Italian intonation; a 
perceptual experiment. Proc. ICPhS 95 , 
Stockholm, Sweden. 
Grice, M., R. Benzmtiller, M. Savino, B. 
Andreeva. 1995. The intonation of queries 
and checks across languages: data from 
Map Task dialogues. Proc. ICPhS 95,, 
Stockholm, Sweden. 
Gussenhoven, C. 1983. Focus, mode and the 
nucleus. Journal of Linguistics, vol 19, 
377-417. 
Kowtko, J., S. Nard, and G. Doherty-Sneddon. 
1992. Conversational Games within Dia- 
logues. HCRC research paper, Edinburgh. 
Pierrehumbert, J.B. 1980. The phonetics and 
phonology of English Intonation. Doctoral 
dissertation. MIT. 
