Upload
claire-matthews
View
220
Download
0
Tags:
Embed Size (px)
Citation preview
Eye Movements in Reading Syntactically Ambiguous Sentences in Russian Language
Victor N. Anisimov, Anna S. Jondot, Olga V. Fedorova, Alexander V. Latanov
M.V. Lomonosov State University, Moscow
Moscow, 2015
30 min recording, sampling rate of 8 fps
First study of eye movements in reading syntactically ambiguous sentences
fixation
Syntactic ambiguity in English
(a) They gave her dog candies (b) They gave her dog candies They told her cat stories They told her cat stories
Two interpretations of the sentence:
Number of fixations normalized for 10 participants and reading every sentence with definite interpretation - (a) or (b).
Syntactic ambiguity in English
dog candies (b)
Num
ber
of fi
xatio
n
Number of fixations normalized for 10 participants in reading every sentence with definite interpretation - (a) or (b).
Num
ber
of fi
xatio
n
her dog (a)
The criminal shot
the servant of the actress who was on the balcony
the servant the actress
early closure late closure
Who was on the balcony?
Syntactic ambiguity
(early/late closure)
?dominates in English dominates in Russian
A priory interpretation complexity during reading structurally (syntactically) ambi-guous sentences slows reading in comparison with reading sentences without any syntactic ambiguity.
MethodMonitor for stimuli presentation
IR - mirror
IR - light
Hight-speed digital camera
Top view
45 deg.
Eye image, reflected from IR mirror, is projected to camera
matrix
Eye
Monitor
IR mirrors
• 29 subjects, 18-24 years old;• 20 sentences Global (G) and Control (C);• originally developed eye tracker based on fast
digital camera FV300;• sampling rate of 250 fps;• matrix resolution 640х480 pixels;• angular size of the monitor for text presentation
47 deg.;• average angular size of the character 0,6 deg.
Method (1st part)
Example of reading sentence with ambiguity
regression
1 s
2nd str
3id str
1st str
Example of stimuli (overall 40)
Eye movements parameters in reading 2nd string (test and control). Data averaged over all sentences and 30 subjects. Spread presented by standard error of mean, sample size in parenthesis).
Parameter Global (G) Control (C)Significant
level, р
Reading time (ms)1393±20 (1169)
1195±14 (1169)
<0,00001
Number of fixations (in one string)
5,29±0,05 (1169)
4,72±0,04 (1169)
<0,00001
Regression frequency (in one string)
0,662±0,029 (1169)
0,336±0,019 (1169)
<0,00001
Fixation duration (ms)207±1(5964)
200±1(5326)
<0,00001
Results. 1st experiment.
Anisimov et. al., 2014
Parameters of eye movements in reading 2nd string in G and C
Regressive saccadesComparison of regressive saccades in 1st and 2nd strings in reading sentence with global ambiguity and control
•F2(G1/G2)=47,61;
p<0,01, n=16
•F2(G1/C1)=0,80;
p<0,37, n=16
•F2(G2/C2)=62,9;
p<0,01, n=16
p<0,01
1st string 2nd string
G1 C1 G2 C2
Types of ambiguity
• G – global
• LE – local, early closure
• LL – local, late closure
• C – control
Gardener met daughter of the cook who went to school.Садовник встретил дочку кухарки, которая пошла в школу.
Girl caught snowflake that melt on the palm. Девочка поймала снежинку, которая растаяла на руке.
Sentences with and without ambiguity
· 31 subjects, 18-24 years old;
· 12 sentences LE, LL, G and C;
· originally developed eye tracker based on fast digital
camera FV300;
· sampling rate of 250 Hz;
· matrix resolution 640х480 pixels;
· angular size of the monitor for text presentation
47 deg.;
· average angular size of the character 0,6 deg.
Method (2nd part)
Example of reading sentence with ambiguity
Results. 2nd experiment. Eye movement parameters (M±SEM, sample size in parentheses) averaged over
all subjects in reading LE, LL, G and С.
In reading LL subjects demonstrate more fixations than in reading LE. Significant differences revealed by Student t-test criterion.
Number of fixations in reading sentences with local ambiguity.
In reading LL regression frequency significantly higher than in reading G. Number of fixations is quasi significantly higher than in reading G. Significant differences revealed by Student t-test criterion.
Regression frequency in reading sentences with ambiguity.
1. Syntactic ambiguity leads to slowing reading because of language processing complication.
2. LL is more complicated in terms of syntactic interpretation than LE.
3. Higher regression frequency and greater number of fixations in reading LL in comparison with reading G indicate that interpretation of LL is more complicated than interpretation of G.
4. Disambiguation of G in Russian (as well as in English) apparently doesn't provoke additional complexity of syntactic analysis and its interpretation is predetermined by domination of early closure (high attachment) principle in Russian to a considerable extent. The absence of any significant differences between eye movement parameters in reading G and LE proves this point.
Thanks for you attention!
[email protected]@ya.ru,