Intonation and Sentence Type:
The Emergence of Conventions for Attitudinal Meanings
Sunwoo Jeong
Department of Linguistics, Stanford University
Introduction
Experimental Results: Attitudinal Meanings (Polar-Qs)
Statistical Analysis
Intonational meanings and context dependency: different perspectives
Consistent effects of intonation on participants’ interpretations of attitudinal meanings, across sentences with
different interpretational biases and across different illocutionary inferences
A series of linear mixed effects models was fitted to
the data, with:
authority
0
inv
req
amb
inf
inv
Legend: falling: x level: x rising: x
req
Polar-Qs
There are qualitatively different types of context
independent intonational convention signaled by
0
inf
inv
Ordering: annoyance:
accuse
20
10
Methods
info−s
0
accuse request
info−s invitation request
A series of perception experiments with stimuli:
No consistent intonational effects on speech acts:
• Representing
• Pre-coded
biases as the main predictor
• Significant intonational effects only for sentences
with ambiguous biases
Stimuli
Each of the 31 sentences manipulated to yield:
•3
types of terminal contours (rising: L* H-H%,
level: !H* H-L%, falling: !H* L-L%)
Pitch (semitones re 100 Hz)
24
12
+10 st.
0
-12
-10 st.
-20.84
Do you have a problem?
0
1.224
Time (s)
info−s invitation request
• Each
of the attitudinal ratings as the main
dependent variables
• Intonation, participants’ illocutionary
inferences, and sentence-type as
independent variables (+ all possible interactions)
• Speaker and participant as random effects
Coef. S.E. p <
intercept
41.57 3.47 .001
level
12.31 2.64 .001
rise
−11.87 2.67 .001
info-s
−8.50 2.42 .001
invitation
−7.09 2.77 .05
Wh
7.60 3.63 .05
rise * inv
−11.16 4.01 .01
level * Decl
−18.97 4.97 .001
invitation * Wh −24.32 4.92 .001
Model fragment: Annoyance rating
Imprs
80
80
60
60
40
10
0
0
stance
30
20
20
Polar-Qs
'Do you want to do the laundry?'
illocution count
illocution count
30
stance:
40
info−s invitation request
info−s
Polar-Qs
'Do you have a problem?'
amb
60
annoyance
info−g
req
Wh-Qs
0
0
info−s accuse request
inv
politeness:
@
@
@
60
20
20
0
authority:
40
40
50
A
A
A
inf
amb
Polar-Qs
60
100
req
Other sentence types
info−giving bias
illocution count
illocution count
150
They signal stable attitudinal meanings
(e.g. speaker authority, politeness, stance, etc.)
across diverse contents and contexts.
Sentence
Where do armadillos live?
How do manatees swim?
Do you want to go to the movies?
invitation
Do you want to grab a bite?
Can you close the window?
request
Can you carry this box?
Do you have a problem?
ambiguous
Do you want to the laundry?
0
Decls
info−seeking bias
terminal contour (e.g. L* H-H%)
+
sentence type (e.g. Declarative).
Bias
info
seeking
20
Speech Acts
Main Argument
diverse sentence types (16 polar-Qs
& wh-Qs, 7 declaratives, 8 imperatives), contents
(4 types of biases), and contexts
• Produced by diverse speakers (6 in total), and
acoustically manipulated in terminal contours
amb
20
Discussion & Conclusion
stance
inf
40
40
20
0
60
60
40
20
80
80
60
40
stance
Rise > {Fall Level}
annoyance
60
polite
Rise > {Fall Level}
politeness
What other types of intonational contours have such
conventionalized effects? What kinds of meaning
contributions do they make? How do they operate?
authority
Fall > Level > Rise
annoyance
contours always have
context-dependent effects on meanings [1]
• At least some types of intonational contours have
context independent effects on meanings [3]
e.g. L*+H in English: scalar values, speaker
uncertainty → conventional implicature
stance
annoyance
Level > Fall > Rise
• Intonational
40
20
20
0
0
info−s invitation request
Significant effects of intonation + sentence-type
on attitudinal meanings; for instance:
request
wish
info−g
Questions & Procedure
240 participants (Amazon MTurk) listened to all 31
sentences (randomly presented in 1 of the 3 intonations), and answered the following questions:
• Verification
Q1 : What did the speaker say?
• Speech acts (forced choice)
Q2 : What is the most likely interpretation of the
utterance?: (a) info-seeking, (b) request, (c)
invitation, (d) info-giving, (e) accusation, etc.
• Attitudinal meanings (ratings from 0–100)
Q3 : How annoyed does the speaker sound?
Q4 : How authoritative does the speaker sound?
Q5 : How polite does the speaker sound?
Q6 : What kind of attitude does the speaker have about
the listener? (from negative (0) to positive (100))
• Falling
+ Impr: positive stance, authority
• Rising + Polar-Q: politeness, positive stance
Potential sources for the conventions? cf. [2]
Full data: https://github.com/sunwooj/labph
Related paper (with Chris Potts):
https://github.com/sunwooj/perlocution
Acknowledgement
I thank Chris Potts, Meghan Sumner, Cleo Condoravdi, Sven Lauer, and Rob Podesva for their valuable comments.
References
[1] Cutler, A. (1977). The context-dependence of intonational meanings. Proceedings of the 13th regional
meeting: Chicago Linguistic Society, 104-115.
[2] Scherer, K. R., Ladd, D. R., & Silverman, K. E. A. (1984). Vocal cues to speaker affect: Testing two
models. The Journal of the Acoustical Society of America, 76(5), 1346?1356.
[3] Ward, G., & Hirschberg, J. (1985). Implicating uncertainty: The pragmatics of fall-rise intonation.
Language, 747?776.