4aSC14

1
4aSC14
Formants of Japanese function particles Setsuko
Shirai University of Washington Department of
Linguistics, Phonetics Laboratory
May 27th 2004ssetsuko_at_u.washington.edu
Hypothesis 2
DISCUSSION
RESULTS
ABSTRACT
There is much debate about whether vowel
centralization results from reduction or
undershoot. This study investigates this
relationship in Japanese. The results of research
conducted by Keating and Huffman (1984) indicated
that Japanese vowels in prose were formed closer
to the center of vowel space than in words.
However, they did not verify their results using
statistics. Moreover, they provided little
information about the preceding consonants in the
prose. Thus, there was a possibility that the
consonantal context led to the reported vowel
centralization through undershoot. To address
this issue, I conducted research in which
Japanese vowels a, e, o in function
words were compared to a matched set of content
vowels. Key indicators of increased
centralization were F1 values for /a/ lower
centralization, F2 values for /e/ lower
centralization, and F2 values for /o/ higher
centralization. There were mixed results in
comparing content vowels to matched function
vowels. For /a/ F1 there was a significant
effect, but there were no significant for/e/ F2
or for /o/ F2. Furthermore, I found a lexical
effect in the absence of a durational difference
. The results of factorial ANOVA with F1
categorical durations indicate that there is a
significant lexical effect in addition to
undershoot.
If durational reduction is the source of spectral
reduction, which is undershoot (Lindblom 1963),
then differences in F1 will be linked to
differences in duration, If lexical difference is
the source of spectral reduction, which is
independent from durational reduction, then
differences in F1 will be observed in the absence
of differences in duration.
The results of F1 show that function /a/ is
reduced. In addition, overall results of distance
show that function vowels are more central than
content vowels.
Hypothesis 1 Function /a/ shows statistically
reliable reduction, but /e/ and /o/ do not.
Possible reason for lack of lexical effect on F2
for /e/ and /o/ The preceding consonants /d/
and /t/ have little coarticulation effect on F2.
Locus target distances between these consonants
and vowels are short. The magnitude of lexical
effect is small consequently, no significant
lexical effects are observed. Hypothesis 2
The results of a factorial ANOVA show that there
are significant effects for Lexical status and
for duration on F1 of /a/ and duration effect but
no significant interaction between the two.
Observed centralization is independent of the
observed differences in duration. The
trajectories indicate that the onset F1 of
content /a/ does not agree with the onset F1 of
function /a/ replicating van Bergems findings
for Dutch. Thus, increased contextual
assimilation is involved with spectral reduction.
The results of a factorial ANOVA indicate that
the target of function /a/ is different from the
target of content /a/. The categorical durations
indicate that shorter vowels have lower F1
implicating undershoot. However, content /a/ has
a systematically higher F1 than function /a/ at
all durations, indicating lexically based
reduction that is independent from undershoot.
Formant plot illustrating the average F1 and F2
for /a/, /e/ and /o/. Note the relative
centralization of function /a/ on the F1
dimension.
TrajectoriesTo test whether or not the duration
is sole determinant of formant displacement,
formants are measured at five places at the
beginning, ¼, ½, ¾ and at the end of a vowel.
Categorical durationTo test whether or not
there is lexical difference when the duration is
the same, each vowel duration is classified into
3 durational categories (short, medium, long).
Figure 1. Formant plot showing content
vowels (red) and function vowels (blue).
RESULTS Duration is not the sole determinant of
spectral vowel reduction. The trajectories of F1
/a/ shows that target of content F1 seems
different from that of function F1. The results
of factorial ANOVA indicate that there are
significant effects for LEXICAL status and for
DURATION on F1 and D F1 but no significant
INTERACTION. In addition, the box plot graph in
Figure 5 shows that F1s of content /a/ are lower
than those of function /a/. Figure 4.
Scatter graphs illustrating F1 of /a/
trajectories Categorical duration and F1 (left),
or D F1 (right) are plotted in Figure 5. The
graph show that F1 of function /a/ is lower than
F1 of content /a/ when the duration is the same.
D F1 of content /a/ is greater than D F1 of
function /a/. It indicates that the target of F1
for function /a/ is different than that for
content /a/. Figure 5. Box Plot
graphs illustrating F1s (left) and DF1s (right)
for each categorical duration. The results of a
factorial ANOVA with F1 of /a/ as dependent
variable, LEXICAL and CATEGORICAL DURATION as
independent variables Main effect for
LEXICAL on F1 of a F(1, 126) 12.639, p
0.001 for DURATION on F1 of a F(2, 126)
4.232, p 0.017 for INTERACTION on F1 of
a F(2, 126) 0.889, p 0.413 The results
of a factorial ANOVA with D F1 of /a/ as
dependent variable, LEXICAL and
CATEGORICAL DURATION as independent
variables Main effect for LEXICAL on D F1 of
a F(1, 126) 13.236, p for DURATION on D F1 of a F(2, 126)
16.568, p INTERACTION on D F1 of a F(2, 126) 0.368,
p 0.693
PREVIOUS RESEARCH

Table 1. Averages of F1 and F2
CONCLUSION
Hypothesis 1 Japanese function vowels are
spectrally reduced although Japanese does not
have a stress contrast. Hypothesis 2 Spectral
vowel reduction in Japanese is caused by lexical
reduction in addition to undershoot.

• Box plots illustrating F1 and F2 of content vs.
  function. Note the relatively large effect on F1
  for /a/.
• Figure 2. Box plot showing that formants of
  content vowels and function vowels. Left F1
  Right F2
• The results of a repeated measure ANOVA
• Main effect LEXICAL on F1 of /a/ F 1, 65
  73.401, p
• F2 of /e/ F 1, 68 3.549, p 0.067,
• F2 of /o/ F 1, 71 1.150, p 0.287

ACKNOWLEDGEMENTS
Hypothesis 1
I would like to thank subjects who participated
in recordings. I appreciate the advice of
Richard Wright and Alicia Beckford Wassink,
Sharon Hargus and Ellen Kaisse at the University
of Washington. I would like to thank the
phonetic lab members who provided for their
useful comments.
The Japanese function vowels, /a/, /e/, and /o/
in function words will be spectrally reduced
compared to content vowels.

• PHONETIC ENVIRONMENTS
• Following phonemes the same in content and
  function
• Location of the target syllables - almost same
  position in function and content
• Length of sentences almost same in function and
  content

STATISTICAL ANALYSIS Results were submitted to
a repeated measure ANOVA with F1, F2, and
DISTANCE split by vowel (a, e, o) as the
dependent variable, and LEXICAL (content,
function) as independent variable.
REFERENCES
Keating, P and Huffman, M. 1984. Vowel
Variation in Japanese. Phonetica, 41,
191-207. Lindblom, B. 1963. Spectrographic
Study of Vowel Reduction. Journal of Acoustical
Society of America Vol. 35. No. 11. pp
1773-1781. Lindblom, B. 1996. Speech A Special
Code. Cambridge the MIT press. Moon, S.J, and
Lindblom, B. 1994. Interaction between
duration, context, and speaking style in English
stressed vowels. Journal of Acoustical Society
of America Vol 96. No. 1 pp 40-55. Nearey, T. M.
1978. Phonetic feature systems of vowels. PhD
thesis, Indiana University Linguistics Club Nord,
L. 1986. Acoustic Studies of Vowel Reduction in
Swedish. Quarterly Progress and status report.
Vol. 4 19-36. Shirai, S. 2003. The Duration of
Function Words in Japanese. UW Working Papers in
Linguistics, Vol. 21, 70-92. Vance, T. J. 1986.
An Introduction to Japanese Phonology. New
York State University of New York. Van Bergem,
D. R. 1993. Acoustic vowel reduction as a
function of sentence accent, word stress, and
word class. Speech Communication, 12,
1-23. Wright, R. 2003. Lexical Competition and
Reduction, Papers in Laboratory Phonology VI,
75-87.
TOKENS /ga/ C bokushi, kiga
kikitsukeruF katoo-san-wa ki-ga kikuC
nakamura-san-wa jinbutsuga toku-ni osukide
irrashaimashita F ano eiga-wa jinbutsu-ga
toku-ni yoku byooshasareteimasu /to/ C hato
kubi-o mawashita F ha-to kuki dake shika-nai C
kookoosei kogoto karuku kikinagusu F kookoo-no
kogo-to kanbun-wa muzukashii /de/ C kotowaza-ni
tade kuu mushi-mo sukizuki-to F taue-no ato-de
ta-de kuu meshi-wa umakatta C suizoku-kan-ni
iku-made hitode konna-ni kirei-da-nante
shiranakatta F katoo-san-wa Toukyoo-no hito-de
korokoro-to yoku waratte aisoo-ga ii
SUBJECTS Six female and six male speakers of the
Tokyo dialect MEASURING POINTS Formants are
measured at the center of vowels

• DISTANCE
• Finding the center of vowel space using Neareys
  method of normalizing formants (Neary 1978 ).
• Based on the center of vowel space, each formant
  is normalized, which are called G1and G2
• Distance is calculated as follows