Intraspeaker variability in vowel production: An investigation of motherese, hyperspeech, and Lombar

1
Intraspeaker variability in vowel production An
investigation of motherese, hyperspeech, and
Lombard speech in Jamaican speakers

• Alicia Beckford Wassink
• Richard Wright
• Lisa Galvin
• Amber Franklin
• University of Washington, Seattle, USA
• wassink_at_u.washington.edu

UW Linguistics Colloquium 17 October, 2003
Department of Linguistics Speech and Hearing
Sciences
funded by a grant from the Center for Mind, Brain
and Learning, University of Washington
2
0.0 Introduction

• Purpose
• Characterize and compare several acoustic
  phonetic parameters associated with so-called
  perturbed speech
• Extend knowledge regarding the sources of
  systematic variability in speech
• --The Lombard effect and sociolinguistics
• studies of audience design and attention paid to
  speech (Mahl 1975 Labov, 1984 Bell, 1984) have
  been influenced by investigations into auditor
  effects and speakers ability to self-monitor
• 1960s (Labov) the notion of formality lies on a
  simple dimension of attention paid to speech
• formality or styleas a discrete variable,
  rather than a continuous one WORD LIST gt
  MINIMAL PAIRS gt READING gt INTERVIEW gt
  CASUAL
• Bell argues that we need a construct that does
  not confuse the code with factors influencing the
  code...Language doesnt covary with style. Style
  is an axis of its own. (1984)

3
1.0 Background Types of Exaggerated Speech

• Child-directed speech (CDS, Kuhl et al. 1997)
• Speech directed to infant auditors is produced
  with a higher fundamental frequency, exaggerated
  pitch range, and slower rate of speech
• more expanded F1 x F2 vowel space (92 larger
  than ADS)
• Question Do adult speakers of a duration
  contrasting language show stretching of cues
  crucial to phonemic contrast? Do hyperarticulated
  vowels facilitate the childs category
  development?
• The Lombard reflex (Lane Tranel, 1971)
• Speech in which intensity is adjusted to
  compensate for changes in background noise. An
  informational response.
• Question Does the Lombard reflex affect
  parameters other than intensity?
• Hyperspeech (The Hyperspace Effect, Johnson et
  al. 1993)
• Talkers modify F1 and F2 of phonetic targets in
  an effort to alleviate perceived difficulties on
  the part of the listener in recovering
  information in the signal. An informational
  response.
• e.g., corrective behavior, clarification of the
  content of an intended utterance (to another
  native-speaking adult addressee)

4
1.1 Goals of this talk

• Goal 1 Characterize the effects of four task
  types on key phonetic parameters associated with
  vowel contrast.
• Research Question 1 Do these four task types
  differ in the pressures they exert on key
  phonetic parameters?
• a.) F1 (Hz) c.) duration (sec) e.) intensity
  (dB)
• b.) F2 (Hz) d.) f0 (Hz)
• Goal 2 Characterize cross-lectal similarities
  and differences in exaggeration strategies.
• Research Question 2 Are the phonetic parameters
  exaggerated differently in a linguistic variety
  that employs spectral differences for vowel
  contrast than in another which relies upon
  temporal differences for signaling contrast?
• a.) Jamaican acrolect (spectral) c.) future
  American English (spectral)
• b.) Jamaican basilect (temporal)

5
1.2 Acrolectal and Basilectal Segmental
Inventories

• Vowels
• --Vowels are sociolinguistic markers in Jamaican
• American English 1.21

6
2.0 Methods 1

• A. Materials and Equipment
• 1.) 6 vowels in 3 tense-lax pairs
• acrolect basilect
• HeedieHiddie /i? / /I/ /i?/ / i/
• HaughdieHaddie /ç? / /a/ /a?/ / a/
• WhoodieHoodie /u? / /U/ /u?/ /u/
• 2.) Critters and critter cards
• 3.) Tascam DA-P1
• 4.) 2 Shure WL184 supercardioid lavalier
  microphones
• 5.) Shure UT4 wireless bodypack
  transmitter/receiver
• 6.) Digital tape of white noise - 40dB noise
  (Lombard task only) delivered over closed
  circumaural headphones (Sennheiser HD280)

7
2.1 Methods 2

• B. Tasks and Amounts of Data
• Four randomized tasks (semi-crossed design)
• unscripted/ child/ noise/
• scripted adult no noise
• (Spontaneous-Adult) (unscripted) (adult) (no
  noise)
• (Familiarization)
• Spontaneous-Playtime (CDS) unscripted child both
  (12)
• Wordlist in carrier (CDS) scripted child both
  (24)
• Map (Hyperspeech)-corrective scripted adult both
  (24)
• Map (Hyperspeech)-noncorrective scripted adult bot
  h (24)
• Subjects
• Two groups of 10 mothers/group (of infants 6-9
  months of age). Data for 5 mothers/group are
  considered here.
• -- basilect 5 mothers x 6 vowels x /-84 tokens
  2520
• -- acrolect 5 mothers x 6 vowels x /- 84
  tokens 2520

8
2.2 Methods 3

• C. Analysis
• 1.) digitized at an 44kHz sampling rate
• 2.) downsampled to 11.025kHz for analysis in
  Praat 4.04
• 3.) overall measures duration
• 4.) midpoint measures f0, F1, F2, F3 (in Hz),
  intensity (dB)
• 5.) f0 range for each speaker
• 6.) Inter-measurer reliability (10) 91
• 7.) Inferential Statistics Factorial MANOVA (4
  x 2 x 2 x 2 design)
• --Appropriate post-hoc comparisons
• --Independent variables (discrete) Task (4
  levels), Group (2 levels),
• Noise (2 levels) and Auditor (2 levels)
• --Dependent variables (continuous) f0, F1, F2,
  intensity, duration
• --Speaker (within subject factor)

9
3.0 Summary of Results

• Research Question 1 (restated) Do these four
  task types differ in the pressures they exert on
  key phonetic parameters?
• By-parameter results
• 1.) Intensity the Lombard task shows the
  highest dB values. CDS means not significantly
  different from Lombard.
• 2.) f0 mothers show the widest ranges in CDS
  and Lombard tasks
• Research Question 2 (restated) Are the phonetic
  parameters exaggerated differently in a
  linguistic variety that employs spectral
  differences for vowel contrast than in another
  which relies upon temporal differences for
  signaling contrast?
• By-group results
• 1.) Region Main effect of region on vowel
  duration (plt0.01)
• 2.) F1 Main effect of region on F1 (plt0.01)
• 3.) F2 Main effect of region on F2 (plt0.03)

10
3.1 Results by Phonetic Task
Table 1 Mean values for 5 phonetic parameters,
by task ( indicates a statistically significant
difference at plt0.05)
11
3.1 Results by Phonetic Task
Table 1 Mean values for 5 phonetic parameters,
by task ( indicates a statistically significant
difference at plt0.05)
12
3.1 Results by Phonetic Task
Table 1 Mean values for 5 phonetic parameters,
by task ( indicates a statistically significant
difference at plt0.05)
13
3.1 Results by Phonetic Task
Table 1 Mean values for 5 phonetic parameters,
by task ( indicates a statistically significant
difference at plt0.05)
14
3.2 Results by Group (Acrolect vs. Basilect)

• Table 2 Between-group differences for 5 phonetic
  parameters. ( indicates a statistically
  significant difference at plt0.05). Means for F1,
  F2 are computed using distance scores for tense
  vs. lax vowels (e.g.,( F1Bas/i/-F1Bas/I/)-(F1Acr/
  i/-F1Acr/I/)).

15
3.2 Results by Group (Acrolect vs. Basilect)

• Table 2 Between-group differences for 5 phonetic
  parameters. ( indicates a statistically
  significant difference at plt0.05). Means for F1,
  F2 are computed using distance scores for tense
  vs. lax vowels (e.g.,( F1Bas/i/-F1Bas/I/)-(F1Acr/
  i/-F1Acr/I/)).

16
4.0 Discussion

• Register Theory, a more integrated view
• Variation on the style dimension within the
  speech of a single speaker derives from and
  echoes the variation which exists between
  speakers on the social dimension. (Bell,
  1984151)
• Present study suggests that it is most profitable
  to consider style as an axis. Linguistic
  features are used as resources available to the
  speaker from the set available in social
  variation. The speaker chooses among available
  features (which comprise variables) as they
  design speech for a particular audience, based
  upon their perception of that speakers
• linguistic experience (competence in target
  language)
• environmental experience (noise)
• social features (age)
• discourse context (attitude toward the topic,
  conversational factors)

17
4.0 Discussion

• Audience Design Theory, Bell 1984
• Variation on the style dimension within the
  speech of a single speaker derives from and
  echoes the variation which exists between
  speakers on the social dimension. (Bell,
  1984151)

18
4.0 Discussion

• Audience Design Theory, Bell 1984
• Variation on the style dimension within the
  speech of a single speaker derives from and
  echoes the variation which exists between
  speakers on the social dimension. (Bell,
  1984151)

19
4.0 Discussion

• Examination of a single set of phonetic
• parameters across several exaggerated
• speech tasks has yielded
• A rich database for inquiries into intra-speaker
  variability
• A more sophisticated understanding of phonetic
  differences that obtain between different types
  of exaggerated speech.
• A larger pool of languages for which CDS has been
  examined (i.e., new data from a creole language),
  and further clarified differences between two
  varieties of this language.
• Results with implications for language
  acquisition Do speakers emphasize those
  dimensions that will be crucial for the childs
  phonemic category development? If between-group
  durational differences disappear under CDS, but
  spectral differences remain, are spectral
  differences of particular importance in category
  development? Not necessarily. Both emphasize
  durational differences in tenselax contrast.

20
References

• Bell, A. (1984) Language style as audience
  design. Language and Society (13). Cambridge
  Cambridge.
• Crothers, J. (1978) "Typology and universals in
  vowel systems." In Universals of Human Language
  (J. H. Greenberg, C. A. Ferguson and E. A.
  Moravcsik, eds.). Stanford Stanford UP,
  93-152.
• Johnson, K., Flemming, E., Wright, R. (1993)
  The hyperspace effect Phonetic targets are
  hyperarticulated. Language, 69 (3), (505-528).
• Kuhl, P. K., Andruski, J. E., Chistovich, I. A.,
  Chistovich, L. A., Kozhevnikova, E. V., Ryskina,
  V. L., Stolyarova, E. I., Sundberg, U.
  Lacerda, F. (1997) Cross-Language Analysis of
  Phonetic Units in Language Addressed to infants,
  Science (277), 684-686, Aug 1.
• Labov, W. (1984) Field Methods of the Project on
  Linguistic Change and Variation. In, Language in
  Use, (J. Baugh and J. Sherzer, editors).
  Englewood Cliffs Prentice-Hall.
• Lane, H., Tranel, B. Sisson, C. (1970)
  Regulation of Voice Communication by Sensory
  Dynamics, Journal of the Acoustical Society of
  America, 47(2), 618-624.
• Lane, H., Tranel, B. (1971). The Lombard sign
  and the role of hearing in speech. Journal of
  Speech and Hearing Research, 14, 677-709.
• Lehiste, I. (1970) Suprasegmentals. Cambridge,
  Mass. MIT, 18-33.
• Mahl, (1975)
• Snow, C. E. Ferguson, C. A. (1997, eds.)
  Talking to Children Language Input and
  Acquisition. Cambridge Cambridge UP, 31-49.
• Wassink, A. B. (2002) Theme and Variation in
  Jamaican Vowels, Language Variation and Change
  13(2).