48x36 Poster Template

1
Acoustic voice correlate variation by dialect
Data from Venezuelan Spanish Stephanie LainThe
University of Texas at Austin
Abstract
Research Questions

• Fixed effect dialect showed significance (p
  .05) for RMS in initial and medial contexts (p
  .004 p .042).
• Fixed effect voicing category was significant
  for consonant duration, VF, RMS, F1 onset and
  burst in initial and medial contexts.
• Voicing category showed significant values for
  preceding vowel duration and CV ratio in medial
  position.
• Voicing category values for VOT were significant
  in initial position (p .000), but not in medial
  position (p .142).
• F0 contour failed to achieve significance in
  either initial (p .996) or medial position (p
  .552).
• The fixed effect condition showed significant
  values for consonant duration in both initial and
  medial contexts (p .000).
• VOT and RMS were also significant for condition
  in initial position only (p .000).
• In initial position, the interaction of voicing
  category dialect was significant for consonant
  duration and VOT (p .000 p .026).
• Figure 2 shows the interaction of voicing
  category dialect for consonant duration
  (initial position).

• Figure 4 shows the interaction of voicing
  category dialect for RMS (initial position).
• FIG. 4. Initial RMS interaction voicing category
  dialect.
• For the interaction dialect condition
  (speeded/ unspeeded), the only significant
  variable was preceding vowel duration (p .005).
• RMS was shown to uphold the difference between
  MAR and MER. Difference greater in initial than
  in medial contexts.
• All correlates (with the exception of F0
  contour) showed significance for the fixed effect
  voicing category.

• What robust acoustic correlates to voice
  emerge?
• Are the acoustic correlate inventories
  substantially different in the two dialects?
• Assuming differences are found, what
  generalizations can be made?
• Are there observable differences across prosody
  and rate of speech?

• The present study is an investigation of
  acoustic correlates corresponding to the category
  voice in two dialects of Venezuelan Spanish.
  The Andean mountain dialect Mérida (MER) and
  Caribbean coastal dialect Margarita (MAR) are
  thought to differ systematically in the phonetic
  implementation of the Spanish phonological stop
  series along the lines of lowland and highland
  divides commonly reported for Latin American
  Spanish. Specifically, MER has been characterized
  by a greater percentage of occlusive
  pronunciations, MAR by more fricative and/or
  approximant realizations of phonological stops.
  To test what repercussions these differences in
  consonant articulation have on the acoustic
  correlates that encode voice, a production
  experiment was run.
• Informants were 25 adult monolingual speakers of
  Venezuelan Spanish from the areas of El Tirano
  (Margarita Island) and San Rafael de Mucuchíes
  (Mérida state). The materials were 44 CV syllable
  prompts. Target syllables were analyzed with
  respect to the following consonant closure
  duration, VOT, VF, RMS, preceding vowel
  duration, CV ratio, F1 onset frequency, F0
  contour, and burst. Statistical analysis using a
  linear mixed model ANOVA tested for fixed effects
  of voicing category, dialect and condition
  (speeded/unspeeded) and interactions of voicing
  category dialect and dialect
  condition. Results showed that the dialects MER
  and MAR vary significantly in RMS. In addition,
  the following correlates were significant for the
  interaction of voicing category dialect
  consonant duration, VOT, VF, RMS, CV ratio and
  burst. Generally, the nature of the differences
  indicates a greater separation between voice
  values in MER than in MAR (notably divergent are
  VOT and RMS). These results imply that while the
  same acoustic correlates of voice are operative
  in both fortis and lenis dialects of Spanish,
  voice categories relate differently.
  Furthermore, with regard to prosody and rate of
  speech, most significant differences in condition
  occurred in initial position while most
  significant differences in the interaction of
  voicing category dialect were linked to medial
  position. The results of this study are relevant
  to current research on the specifics of dialectal
  variation in consonant systems. They also have
  wider implications for the general mapping of
  phonetics to phonology in speech.
• Articulatory gestures condition the contrastive
  use of voice in speech across languages
  (Pickett, 1999).
• However, many fine laryngeal adjustments
  contribute to perceived quality of voicing these
  vary cross-linguistically.
• Substantial language variation, several ways to
  do a 2-way voicing contrast across languages
  (Hoole, Gobl, Ni Chasaide, 1999 Ni Chasaide
  Gobl, 1993).
• Is this true of dialects as well?
• Lewis 2000- Found systematic differences between
  Bilbao, Caracas and Medellín dialects.
• Hualde 2005- Reports variation between dialects
  with respect to stop consonant voicing.
• Martínez Celdrán, 2006- Murcia data show
  phonetic voicing in both
• voice.
• Two dialects of Venezuelan Spanish (Margarita
  MAR and Mérida
• MER) commonly reported to differ in stop
  consonant pronunciation (Lipski 1994).
•  

Materials and Procedure

• Informants were 25 adult monolingual speakers of
  Spanish- 14 from MAR (4 males, 10 females) and 11
  from MER (4 males, 7 females)- between 20-35 yrs.
  of age.
• Materials were 44 CV syllable prompts, preceded
  by the word son.
• During the session, subjects were asked to
  create a nonsense word based on the syllable that
  appeared
• (e.g. Son TO -gt Son toróto)
• Contexts? Word-initial (n CVC) and word-medial
  (n CVC).
• Speeded and unspeeded trials.
• Measurements taken from a Wavesurfer (Sjölander
  Beskow, 2006) spectrogram display viewed in
  conjunction with waveform and pitch tracks.
• Measured 9 acoustic correlates.
• 7 in initial consonant duration, VOT, VF, RMS,
  F1 onset frequency, F0 contour following closure,
  presence/ absence of release burst.
• 9 in medial (see above) preceding vowel
  duration and CV ratio.
• A linear mixed model statistical analysis was
  performed using SPSS software. Word-initial and
  word-medial were analyzed separately.
• The linear mixed model used in this study tested
  for fixed effects of dialect, voicing category,
  and condition (speeded/ unspeeded).
• It also tested for interactions between voicing
  category dialect and dialect condition.
• Figure 1 shows an example of a VCV (/ibi/)
  waveform spectrogram slice for MAR (on the left)
  and MER (on the right).

Discussion
Background
Results
Selected References

• Hoole, P., Gobl, C., Chasaide, A. (1999).
  Laryngeal coarticulation. In W. Hardcastle N.
  Hewlett (Eds.). Coarticulation Theory, Data and
  Techniques (pps. 105-143). Cambridge Cambridge
  University Press.
• Hualde, J. (2005). The Sounds of Spanish.
  Cambridge Cambridge University Press.
• Lewis, A. (2000). Acoustic variability of
  intervocalic voiceless stop consonants in three
  Spanish dialects. In H. Campos, E. Herberger, A.
  Morales-Front, T. Walsh (Eds.),
  Hispanic Linguistics at the Turn of the
  Millennium Papers from the 3rd Hispanic
  Linguistics Symposium (pp. 101-114).
  Sommerville Cascadilla Press.
• Lipski, J. (1994). Latin American Spanish. New
  York Longman.
• Martínez Celdrán, E. (2006). Remarks delivered
  during a keynote address. 3rd Conference
  on Laboratory Approaches to Spanish Phonology,
  Sept. 2006
• Ni Chasaide, A. Gobl, C. (1993). Contextual
  variation of the vowel voice source as a function
  of adjacent consonants. Language and Speech,
  36(2-3), 303-330.
• Pickett, J. (1999). The Acoustics of speech
  communication Fundamentals, speech perception,
  theory, and technology. Boston Allyn and Bacon.
• Sjölander, K. Beskow, J. Wavesurfer (2006). An
  open source speech tool, Centre for
  Speech Technology, KTH, Drottning Kristinas väg
  31, SE-100 44, Sweden.
•  
•  
•  
•