ExLing 2020 Disambiguating cues of disjunctive questions

Abstract

This paper aims to find out the similarities and differences between Jordanian (JA), Egyptian (EA), and Kuwaiti (KA) Arabic in which cues disambiguate alternative questions (altqs) and disjunctive yes-no questions (dynqs): intonation contour and choice of disjunctive element (DE). A perception study was run in the three dialects, replicating Pruitt & Roelofsen's (2013) perception study on English. Mixed-effects logistic regression was used to explore the results which revealed all dialects show a main effect of both intonation and DE choice; a rise contour and use of aw significantly increased the likelihood of dynq responses. The effect of intonation was larger than that of DE choice in all dialects. The differences between the dialects lay in the relative strength of the DE coefficients.

Full text

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/344663816
ExLing 2020 Disambiguating cues of disjunctive questions
Conference Paper · October 2020
CITATIONS
0
2 authors:
Some of the authors of this publication are also working on these related projects:
prosody; intonation; phonology; Arabic linguistics View project
Intonational Variation in Arabic View project
Mohammad Bani Younes
The University of York
2 PUBLICATIONS0 CITATIONS
SEE PROFILE
Sam Hellmuth
The University of York
33 PUBLICATIONS147 CITATIONS
SEE PROFILE
All content following this page was uploaded by Mohammad Bani Younes on 15 October 2020.
The user has requested enhancement of the downloaded file.

International Society of Experimental Linguistics
ExLing 2020
Proceedings of 11th International Conference of
Experimental Linguistics
12-14 October 2020
Athens, Greece
Edited by Antonis Botinis

ExLing 2020: Proceedings of 11th International Conference of Experimental
Linguistics, 12-14 October 2020, Athens, Greece
Disambiguating cues of disjunctive questions
Mohammad Ali. S. Bani Younes, Sam Hellmuth
Language and Linguistic Science, University of York, UK
Abstract
This paper aims to find out the similarities and differences between Jordanian (JA),
Egyptian (EA), and Kuwaiti (KA) Arabic in which cues disambiguate alternative
questions (altqs) and disjunctive yes-no questions (dynqs): intonation contour and
choice of disjunctive element (DE). A perception study was run in the three dialects,
replicating Pruitt & Roelofsen’s (2013) perception study on English. Mixed-effects
logistic regression was used to explore the results which revealed all dialects show a
main effect of both intonation and DE choice; a rise contour and use of aw
significantly increased the likelihood of dynq responses. The effect of intonation was
larger than that of DE choice in all dialects. The differences between the dialects lay in
the relative strength of the DE coefficients.
Key words: Alternative questions, disjunctive yes-no questions, intonation, prosody.
Introduction
A question like Do you want coffee or tea? can be perceived differently in English:
either as an altq (possible answer: Tea) or as a dynq (possible answer: No). The
two readings of the string-identical utterance are perceptually disambiguated by
prosodic cues: with a fall, the utterance is perceived as an altq, but with a rise,
as a dynq (Pruitt, Roelofsen 2013, Dayal 2016). They also claim that placing
accents on both disjuncts ‘X’ and ‘Y’ (in the ‘X or Y’) favours an altq reading.
One reason why these questions may be disambiguated by prosody in English
is that there is only one DE or which is used in both question types. Modern
Standard Arabic (MSA) has two DEs: am in altqs and aw in dynqs (Fakih,
2012). Colloquial Arabic dialects also display two DEs - willa and aw - which
vary in their mapping to English or. To replicate Pruitt & Roelofsen’s study in
Arabic, therefore, one must first decide which DE should be used in such a
perception study.
A review of corpus data and prior literature suggests that some Arabic
dialects might be English-like in employing one DE in both altqs and dynqs
(e.g. EA) while other dialects might be MSA-like in using two DEs, one for
altqs and one for dynqs (e.g. KA). Previous research (Bani Younes, Hellmuth
2020) has found that intonation (rise) and DE choice ( aw) both favour a dynq
reading in JA though both DEs are accepted in both question types, placing JA
in the middle of the two preferences between English and MSA. Having
established the disambiguating cues, how similar or different are the three
dialects in these cues? And what is the relative contribution of each of the two
cues in the disambiguation? The present study, thus, adapts Pruitt &
You can cite as follows:
Bani Younes, M. A. S., & Hellmuth, S. (2020). Disambiguating Cues of Disjunctive
Questions. In A. Botinis (Ed.), Proceedings of 11th International Conference of
Experimental Linguistics. Athens: International Society of Experimental Linguistics

M.A.S. Bani Younes, S. Hellmuth
2
Roelofsen’s design for JA, EA, and KA, using the two cues: intonation
(rise/fall) vs. DE choice (willa/aw). Figure 1 below shows the same utterance
produced with both contour shapes and DEs.
Figure 1. Contour shape vs. DE choice of lljom azmatak æia alift u:r aw/willa
al ada? ‘Did Aya invite you to breakfast or lunch (to have breakfast or lunch)?’
Methodology
There were 74 JA, 52 EA, and 70 KA listeners (mean age 36; range 18-53). 24
lexically-distinct target sentences and 36 fillers were recorded. The task was a
two-alternative forced choice between an altq and a dynq paraphrase of the
question they heard. Each target utterance was recorded in four conditions (see
Figure 1 above), yielding 96 target tokens; 24 unique tokens were presented in
each of four blocks in Qualtrics, in a Latin-Square design. The four conditions
in each block are: 6 aw+rise (2r), 6 aw+fall (2f), 6 willa+rise (wr), and 6
willa+fall (wf) tokens. Participants were sent the experiment link and were
asked to listen to the recordings via their own laptop or mobile phone in a quiet
room, free from noise, using ear/headphones. After they clicked on the link,
participants were randomly assigned to one of the four blocks containing 60
tokens (24 unique target utterances and 36 fillers). Two paraphrases appeared
below each recording, so participants ticked one of them. Paraphrases were
randomly ordered for each participant within each trial, and trials were also
randomly ordered. After listening to each token, they selected the paraphrase
that they thought was the best equivalent to what they had heard.
Results and discussion
Table 1 presents the results of the three separate GLMM models (one per
dialect: glmer(resp_numeric~intonation*DE + gender + age +Education + (1
+ intonation | listener) + (1 | stimulus), data = data2, family = binomial,
control = glmerControl(optimizer = "bobyqa"))). Dynqs were coded as 1 while
altqs as 0. Figure 2 shows the spread of values across participants for counts of
tokens interpreted as dynqs, by dialect.

Disambiguating cues of disjunctive questions
3
Figure 2. Median and interquartile range for count of dynq responses across
participants for each dialect.
Fixed effects Estimate SE z value p-value
JA
Intercept -0.53423 0.52680 -1.014 0.311
intonation1 0.95554 0.11590 8.244 2e-16 ***
DE1 0.34167 0.08524 4.009 6.11e-05 ***
intonation1:DE1 0.12901 0.08517 1.515 0.130
EA
Intercept -0.2955836 0.4127871 -0.716 0.4739
intonation1
0.4677226
0.1001999
4.668
3.04e
-
06 ***
DE1 0.3723088 0.0872576 4.267 1.98e-05 ***
intonation1:DE1 0.1470266 0.0871239 1.688 0.0915 .
KA
Intercept -0.225396 0.436883 -0.516 0.6059
intonation1 0.602045 0.089545 6.723 1.78e-11 ***
DE1 0.147364 0.066593 2.213 0.0269 *
intonation1:DE1 0.065525 0.066591 0.984 0.3251
Table 1. Estimates of coefficients in JA, EA, and KA.
The control variables (gender, age, education) shown in the model syntax
above were not included in the table as they did not reach significance. There
are main effects of intonation and of DE choice in each dialect, suggesting that
both cues are important in all three dialects. However, the effect size of DE
choice compared with intonation is subtly different across the three dialects. JA,
EA, and KA are thus similar in that both intonation and DE choice
significantly influence the likelihood of dynq responses. Likewise, it seems that
intonation was more important than DE choice in increasing the likelihood of
dynq responses in the three dialects. This matches the results of prior studies
on the role of intonation in characterising the realisation of altqs and yes-no
questions (ynqs) in JA, EA, and KA (e.g., Al-Khalifa 1984, Rammuny 1989, Al
Amayreh 1991, Alharbi 1991, Hellmuth 2018, Winans, 2019).
The findings also corroborate the fact that the contour shape of dynqs is the
same as that of ynqs. In other words, the fact that tokens with a rise

M.A.S. Bani Younes, S. Hellmuth
4
significantly favoured dynq responses, over altq responses, is also in line with
Winans’ (2019) claim that aw-dynqs have the same contour as ynqs in EA; she
explained this similarity by the fact that both are ynqs. This similarity in the
contour of dynqs and ynqs might explain listeners’ tendency to interpret rise-
tokens more as dynqs in JA, EA, and KA. For this reason, interpreting tokens
with a rise as dynqs is unsurprising.
The results show slight differences between the dialects. The coefficient
estimates for intonation and DE in EA are similar in size, suggesting that the
relative importance of DE choice in EA was the strongest across the three
dialects, although intonation still plays the main role in the disambiguation. JA
and KA were similar in having DE coefficient values which are much lower
than those of intonation, but the effect of DE choice in KA was still smaller
than that of DE choice in JA, making KA different from JA in this respect, too.
In conclusion, the findings are both similar to and different from Pruitt and
Roelofsen’s (2013) study on English, taking into account the idiosyncratic cues
of each language. Pruitt and Roelofsen’s findings suggest that two cues
(intonation vs. accent distribution), both prosodic, were significant in
disambiguating altqs and dynqs. They also show that intonation played the
primary role while accent distribution played a supporting role. This paper adds
a lexical cue (DE choice) and finds that both cues in Arabic, which are prosodic
and lexical, are also needed, though the primary role is again associated with
intonation and the secondary role with DE choice.
References
Al Amayreh, M.M.K. 1991. A study of the basic intonational patterns in Standard and
Jordanian Arabic (Master’s thesis). University of Jordan. Amman, Jordan.
Alharbi, L.M. 1991. Formal analysis of intonation: The case of the Kuwaiti Dialect of
Arabic. Ph.D. Dissertation. Heriot-Watt University, UK.
Al-Khalifa, F.E. 1984. An intonational study of Kuwaiti Bedouin Arabic. Ph.D.
Dissertation. University of London, UK.
Bani Younes, M., Hellmuth, S. 2020. The role of prosody in disambiguation of
disjunctive questions in Jordanian Arabic. Paper presented at Arabic Linguistics
Forum Conference, University of Leeds.
Dayal, V. 2016. Questions. Oxford University Press.
Fakih, A. H. 2012. The syntax of questions in Modern Standard Arabic: A minimalist
perspective. Germany: LAMBERT Academic Publishing.
Hellmuth, S. 2018. Variation in polar interrogative contours within and between Arabic
dialects. oland, 13-16 June.
Pruitt, K., Roelofsen, F. 2013. The interpretation of prosody in disjunctive questions.
Linguistic Inquiry, 44(4), 632-650.
Rammuny, R. 1989. Instrumental and auditory analysis of Colloquial Jordanian Arabic,
an update. Journal of Arabic Linguistics (20) 23-42.
Winans, L. 2019. Disjunction and alternatives in Egyptian Arabic In K. von Heusinger
et al. (Eds.), Questions in Discourse (pp. 237–285). Leiden, The Netherlands: Brill.
View publication statsView publication stats