tutorial on experimental work - maria...
TRANSCRIPT
![Page 1: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/1.jpg)
Tutorial on Experimental workMaria Polinsky & Stavros Skopeteas
Part II. experimental studies on Georgianby Stavros Skopeteas
Paris, September 23, 2016
South Caucasian Workshop The U Chicago Center Paris
![Page 2: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/2.jpg)
experimental methods in grammatical research
linguistic research based on two sources of evidence:
- observational research: possible linguistic entities in corpora
- introspective research: grammaticality judgments
recent focus on repeated-observation designs:
- quantitative corpus studies or production experiments
- questionnaire studies on speakers‘ intuitions
tutorial, part 1: offline methods, not informative for cognitive processes, crucial for testinggrammatical hypotheses
motivation for these developments: increase reliability, estimate gradience
![Page 3: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/3.jpg)
contents of this part
experimental data in language comparison
- increase reliability
illustrative study: focus in OV languages (Georgian, Armenian, Turkish)
(questionnaire study)
speaking in the lab vs. speaking in the world
- same contrasts, different sources of variation
illustrative study: prosodic effects of order and focus in Georgian
(scripted speech vs. corpus data)
![Page 4: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/4.jpg)
experimental data in language comparison
![Page 5: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/5.jpg)
grammatical background
Georgian, Armenian and Turkish
are OV languages, but differ in details:
local p-expression of focus(independent of order)
Georgian Armenian Turkish
optional V-fronting(creates word order flexibility)
![Page 6: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/6.jpg)
grammatical background
Georgian, Armenian and Turkish
OV languages, differing in details:
local p-expression of focuspossible: XPF YP V
Georgian Armenian Turkish
optional V-frontingpossible: V XPF V
![Page 7: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/7.jpg)
illustrationQA vin ipova burTi?
who found ball?A kiram burTi ipova.
Kira ball found
(does not fit) 1 2 3 4 5 6 7 (fits)
factorial designCONTEXT (question):
Sentence-focus|Subject-focus|Object-focus|Verb-focus
WORD ORDER (answer):
SVO|SOV|OVS|OSV|VSO|VOS
random factors48 lexicalizations, 48 speakers
method
![Page 8: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/8.jpg)
result
SF XFV XFYV XFV
OF XFV XFV
SVO SOV OVS OSV
Georgian
1
3
5
7
SVO SOV OVS OSV VSO VOSorder
acce
ptab
ility
object-focus subject-focus
1
3
5
7
SVO SOV OVS OSV VSO VOSorder
acce
ptab
ility
object-focus subject-focus
1
3
5
7
SVO SOV OVS OSV VSO VOSorder
acce
ptab
ility
object-focus subject-focus
SF XFV XFYV VXF XFV
OF VXF XFV XFV
SVO SOV OVS OSV
SF XFV VXF XFV
OF VXF XFV XFV
SVO SOV OVS OSV
Armenian Turkish
![Page 9: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/9.jpg)
on the reliability of controlling sources of variance in a Between Languages design
tutorial, part 1: acceptability vs corpus: what do we learn about modularity from the mappingbetween different data types?
discussion
![Page 10: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/10.jpg)
speaking in the lab vs. speaking in the world
![Page 11: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/11.jpg)
observation
L H L H L
ni no ma mas e lo li a ve ba
100
350
150
200
250
300
Pitc
h (H
z)
Time (s)0.4 2.1
L H L H L
ni no e lo li a ve ba ma mas
100
350
150
200
250
300
Pitc
h (H
z)
Time (s)0 1.6
Nino father cared Nino fathercared
H-phrase tones are integrated within a single domain of downstep in SOV but not in SVO.
![Page 12: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/12.jpg)
method
Context * Order
(all and only felicitous permutations)
Items (4)nino mama-s e-loliav-eb-a.Nino(NOM) father-DAT O3-care-THM-DEP.S3.SG
Nino cares about the father.
Context questionsALL: What happens?
VP: What do we hear about Nino?
S: Who cares about the father?
eight speakers
two repetitions per item
Total= 832 tokens
SOV SVO OSV OVS
allF
VPF
SF
VF
OF
![Page 13: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/13.jpg)
results
150
200
250
300
subject object verb
mea
n F0
(Hz)
focus ALL VP O V
150
200
250
300
subject verb object
mea
n F0
(Hz)
focus ALL VP S O V
two sources of non-downsteped H-phrase tones:
• right-edge of a non-final V
• preceding a focused XP
![Page 14: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/14.jpg)
theoretical relevance
reported phrase-languages
ALIGN(focus-i;φ-i)
Align the i-edge of the focus with the i-edge ofa prosodic phrase.
(Büring 2010, Féry 2013)
hypothesisIn a subset of these languages (among elseGeorgian), the prosodic edges are mapped on nucleus (not on focus)
ALIGN(nucleus-i;φ-i)
Align the i-edge of the nucleus with the i-edgeof a prosodic phrase.
![Page 15: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/15.jpg)
syntax-phonology mapping
stress-based languagemaximally embedded: stressed
phrase-based languagemaximally embedded: separated
*
* *
* * *
[S [V [O]]
*
* *
* * *
[S [[O] V]
( )
( ) ( ) ( )
[S [V [O]]
( )
( ) ( ) ( )
[S [[O] V]
![Page 16: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/16.jpg)
speaking in the lab vs. speaking in the world
external validity?
• is the inference from the sample to the world valid
• or is it an artefact of the lab situation?
background:
• current discussion in phonetics about scripted data. (Xu 2010; Wagner, Trouvain, & Zimmerer, 2014, among else)
![Page 17: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/17.jpg)
Georgian corpus
NarrativesActivity description
Ancestor story
Path description
Event description
Comparative description
Stylenon-scripted data; not spontaneous, but less attention to speech; naturalistic behavior (laughter, spontaneous speech planning)
24 speakers
- residents of Tbilisi;
- speaking Georgian as a first language;
- generally high education level;
- age range: 20-58 (average 29.3).
material speakers
![Page 18: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/18.jpg)
corpus annotation
speaker XP V YP XP YP V dur (sec)
01 8 5 207
02 5 3 497
03 17 5 268
04 11 8 309
05 6 4 228
06 9 4 179
09 10 7 320
10 14 5 365
11 8 5 218
12 9 2 173
14 5 3 196
15 7 4 290
16 4 6 296
17 10 10 389
18 12 8 361
20 10 2 197
21 6 3 215
22 11 3 483
23 13 2 167
24 7 4 99
total 182 93 91 min.
phenomena of interest- order: XP V YP vs. XP YP V
- focus: approximated through givenness
phenomena outside the scope of the intended generalization => excluded- non-declaratives, embedded clauses
- laughter, disfluencies, etc.
phenomena for which the generalization is intended to apply (random factors)- speakers
- syllable structures
- syntactic category of the XP
![Page 19: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/19.jpg)
resuls
measurementsvoiced part of the first and last syllable
of the target constituent
F0-mean of three equal intervals per target σ
givennessas predictor of focus
given = referent mentioned in the pre-text
word order effect
final vs. non-final V
![Page 20: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/20.jpg)
0.0
0.1
0.2
0.3
0.4
-10 0 10rise in the medial word (semitones)
dens
ity
order SVO (n=64) SOV (n=64)
0.0
0.1
0.2
0.3
0.4
-10 0 10rise in the medial word (semitones)
dens
ity
order XP V YP (n=187) XP YP V (n=97)
scripted speech unscripted speech
comparison between data types
rise in the medial word: F0-max (last syllable) – F0-min (first syllable) of the medial constituent
![Page 21: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/21.jpg)
- on the external validity of experimental data
- controlling variation vs outbalancing variation
- are there lab grammars?
Continuing the discussion in tutorial, part 1: limits of experimental designs, when should I come back to the traditional methods?
final discussion
![Page 22: Tutorial on Experimental work - Maria Polinskyparis2016.mariapolinsky.com/.../11/Tutorial-Georgian.pdfGeorgian, Armenian and Turkish are OV languages, but differ in details: local](https://reader035.vdocuments.mx/reader035/viewer/2022071502/6121f58ae21f3d13414a0452/html5/thumbnails/22.jpg)
Tutorial on Experimental workMaria Polinsky & Stavros Skopeteas
Part II. Georgianby Stavros Skopeteas
Paris, September 23, 2016
South Caucasian Workshop The U Chicago Center Paris