Beyond Lab Phonology The Phonetics of Speech Communication

1
Beyond Lab PhonologyThe Phonetics of Speech
Communication

• Klaus J. Kohler
• IPDS, Kiel, Germany

Paper at the Conference on Methods in
Phonology Berkeley, 20 - 23 May 2004
2
1 Introduction

• Every science develops paradigms
• sets of theoretical and methodological
  principles,
• which are only partly determined by scentific
  phenomena, far more by the sociology of science
• they are passed on through the teaching of an
  influential, often missionary discipleship,
• and they are finally codified in textbooks.

3

• Change means revolution
• Historical Linguistics of the 19th c.
• Structuralism in 1/20th c.
• Generative Grammar since the 2/20th c.
• Applies to the analysis of the spoken medium too.

4

• experimental, signal-oriented phonetics gt
  science discipline Rousselot, Scripture,
  Panconcelli-Calzia
• descriptive, symbol-oriented phonetics gt
  humanities discipline Jespersen, Passy, Sievers,
  Sweet, Viëtor
• phonology in Prague Circle and American
  Structuralism gt new discipline within
  linguistics and humanities
• conceptualization of a science-humanities dualism

5

• Linguistic concepts, e.g. the phoneme, imported
  into psychology and engineering labs
• to be filled with phonetic substance in
  production and perception experiments
• adequacy of linguistic concepts for new questions
  taken for granted
• gt categorical speech perception, the speech
  code, the motor theory of speech perception

6

• analysis of minute detail in word-phonology frame
• e.g. array of phonetic parameters for voiced/
  voiceless plosives in word or logatome contrasts
  in isolation or in metalinguistic phrases
• even rejection of established phonological rules,
  e.g. neutralization of word-final voicing in
  German

7

• poor methodology in subject selection, word
  material, experimental design and blind
  application of statistics
• inferential significance interpreted as category
  difference
• results of limited value for explanation of
  speech communication
• This is the paradigm of phonology-going-into-the-
  lab.

8

• It reached its climax with the Lab Phonology
  series.
• Lab Phonology gt natural science
• filling known phonological categories with
  phonetic substance under lab conditions
• thus alleviating the modularization into
  phonetics and phonology
• but new dilemma categoricalness vs. gradience of
  phonological categories

9

• Neither the phonological categories nor the
  phonetic measurements of Lab Phonology
• need represent language structures in
  communication
• they may even reflect incongruous metalinguistic
  domains
• thus extrapolation to real speakers and listeners
  problematic, but standard practice in Lab
  Phonology
• return to the philosophy of science approach of
  early 20th c. in spite of sophisticated
  theorizing and analysis

10

• Pierrehumbert, J., Beckman, E. M., Ladd, D. R.
  Conceptual foundations of phonology as a
  laboratory science. In N. Burton-Roberts, P.
  Carr, G. Docherty (eds.), Phonological Knowledge.
  Oxford OUP, 273-3003 (2000).

11

• Part of the Lab Phonology paradigm is the
  prosodic framework of Autosegmental-Metrical
  Phonology/ToBI
• none of the many prosodic paradigms
• British School, Halliday, Dutch Model, Swedish
  Model, Danish Model, Fujisaki-Model,
  AM-Phonology/ToBI, KIM (The Kiel Intonation
  Model)
• have been carried round the globe with greater
  zeal than AM-Phonology and its tool ToBI.

12

• It has little concern for the communicative
  categories Time, Listener and Function
• it lacks concern for Time, because it defines
  intonation contours independently of time
• it also lacks concern for the Listener, because
  it focuses on production
• and it lacks concern for Function in a wide
  sense, because it concentrates on linguistic
  function, if it considers function at all.

13

• But these categories are corner-stones in the
  paradigm of phonology-coming-out-of-the-lab.
• I will now look at the phonology of f0 peaks
  under two perspectives
• Lab Phonology
  with phonology-going-into-the-
  lab
• L-H categorization independent of time
• phonetic alignment independent of the listener
• detached from function

14

• Communicative Phonetics
  with phonology-coming-out-of-the-lab
• time, listener, function define
  intonational categories
• phonetic substance determines phonological form
• and provides a direct link to function The
  Frequency Code

15
2 The Phonology of f0 Peaks
2.1 Pitch accents and alignment in AM/ToBI

• Timeless phonological categorization of
  intonation peaks in AM-Phonology/ToBI
  HL vs (L)H vs LH
• post hoc introduction of time as phonetic
  alignment in a considerable number of lab speech
  studies in English, German, Dutch, Greek, e.g.
  R.D. Ladd

16

• Esther Grabes Comparative Intonational Phonology
  of English and German (1998)
• lab data on the production of HL were collected
  for both languages in parallel contexts
• it is postulated, but not explicated that both
  languages contain the same intonational category
  in their phonological inventories
• this category is filled with phonetic substance
  through measuring f0-peak alignment
• result later position in German than in English

17

• but the contexts in the data acquistion were not
  identical for the two languages
• Anna and Peter are watching TV. A photograph of
  this week's National Lottery winner appears. Anna
  says Look, Peter! It's ...! Our new neighbour!
• Anna und Peter sehen fern Ein Lottogewinner
  wird vorgestellt. Anna sagt Na sowas! Das ist
  doch Herr ...! Unser neuer Nachbar!

18

• surprise in Germ. Na sowas! (Well I never!),
  reinforced by doch, absent from Look, Peter!
• in such a context German uses a semantically
  contrastive late peak position
• this shows that
• function is already important at data
  collection
• different phonological synchronizations of f0
  contours with articulation need to be
  distinguished from variable phonetic alignment to
  avoid misinterpretation

19
2.2 Time, Listener, Function in f0 contours
2.2.1 Synchronization of pitch patterns

• backed by long-standing research at IPDS Kiel
  KIM The Kiel Intonation Model cf Lab
  Phonology I
• global contours (peaks, valleys)
• new experimental paradigm
• whole F0 peak contour shifted in equal steps
• through segmentally constant utterance
• for perceptual pitch changes
• and associated semantic features

20

• synchronization of pitch and articulatory time
  courses
• 3 peak contour positions to be differentiated in
  relation to articulatory timing of accented
  syllables
• early
• medial
• late
• listener has a central role

21
Germ. Sie hat ja gelogen. Shes been lying.
l
où
22

• pragmatic function of peak contour
  synchronization
• early - finality
• knowing
• summarizing
• coming to the end of an argument
• resignation

23

• medial - openness
• observing
• realising
• starting a new argument
• late - unexpectedness
• observing, realising in contrast to ones
  expectation
• surprise
• disbelief

24
2.2.2 Internal pitch timing in peak contours

• recent research at IPDS Kiel
• Oliver Niebuhr MA dissertation 2003 Perzeptorisc
  he Untersuchungen zu Zeit-variablen in
  Grundfrequenzgipfeln in German
• Tamara Khromovskikh MA dissertation 2003
  Perzeptionsuntersuchungen zur Intonation der
  Frage im Russischen
• the rise and the fall of a peak contour
• slow
• fast

25

• independent changes of rise and fall speeds
• softening of finality of early peak by slow
  fall
• further increase of openness by fast rise
• perceptual interaction between synchronization
  and internal timing

26
Germ. Sie hat ja gelogen. Shes been lying.
27
2.2.3 AM Phonology and KIM compared

• alignment in ToBI and peak position in KIM are
  fundamentally different concepts
• in KIM the time dimension is anchored in the
  phonological categories themselves
• in ToBI it is a phonetic addition post festum
• Time has the same conceptual value at the
  prosodic level in KIM as it has at the segmental
  level in Articulatory Phonology

28
2.3 Findings from other languages

• Russian
• yes-no questions lack syntactic markers
• synchronization and internal timing effects in F0
  coding of statements vs. yes-no questions
• early vs. late peak positions
• combined with slow rise fast fall vs. fast rise
  slow fall
• and by additional lower vs. higher peak value

29

• Bulgarian
• narrow-focus statement vs. question show the same
  differences in
• synchronization
• and internal timing
• as in Russian
• Bistra Andreeva, Saarbrücken production data

30

• Pisa Italian
• broad focus vs. narrow contrast
• in former, F0 maximum of peak contour later and
  trailing off more slowly
• Barbara Gili Fivela, Tonal alignment in two Pisa
  Italian peak accents, Speech Prosody 2002,
  production data

31

• Neapolitan Italian
• statement vs. question
• later synchronization of F0 peaks
• and strengthening of high F0 in the descent
• for questions
• Mariapaola d'Imperio, The Role of Perception in
  Defining Tonal Targets and their Alignment, PhD
  thesis, OSU, 2000, perception data

32

• Bari Italian
• commands vs. questions
• based on the sentence "lo mandi a Massimiliano
• later peak position as well as a faster rise in
  questions
• Martine Grice Michelina Savino, Low tone
  versus 'sag' in Bari Italian intonation a
  perceptual experiment, ICPhS Stockholm 1995,
  perception data

33
2.4 Explaining the data

• reference to two theoretical principles
• auditory contrast in contours at specific
  syllable points (auditory enhancement, cf. Diehl
  Kluender)
• and J. Ohalas Frequency Code
• contrastive high-low vs low-high pitch change in
  consonant - vowel transition of the accented
  syllable for early vs medial peak

34

• consonant - vowel transition crucial because of
  increase in intensity, heightening pitch change
• low-high change later in vowel late peak
• focus on change to low or high pitch in the
  accented vowel linked to semantics of finality
  vs. opennesss in the German data

35

• J. Ohalas Frequency Code
• an attempt to relate phonetic substance
• high vs. low F0
• to social behaviour
• subordination vs. dominance
• subsequent explanation of linguistic form
• use of high or rising F0, e.g. in questions in
  the languages of the world

36

• may also be applied to the high/low contrast for
  the semantics of opennesss vs. finality,
  which includes subordination vs. dominance
• all peak alignment data and functions they serve
  in the different languages can be subsumed under
  the same two principles of auditory enhancement
  and Frequency Code
• later, faster rising and higher F0 peak
  configuration contains all the ingredients for a
  low-high pitch contrast in an accented vowel to
  mark the question function vs statement/command

37
3 Developing the new paradigm

• The goal of phonetics is the elucidation of
  speech communication
• of the relationship between phonetic substance
  and communicative function
• with linguistic form being derived from this
  relationship.

38

• Corollaries
• neither substance nor function can be analysed
  without the other
• measurement must take place within communicative
  domains
• go beyond lab speech
• take spontaneous speech into the lab

39

• functional categories must be established in
  relation to substantive parameters in production
  and perception
• go beyond systemic linguistic contrasts
• include the whole spectrum of the behavioural
  sound - meaning relationship

40

• metalinguistically derived phonological form has
  no more than a heuristic value in this
  elucidation
• word phonology must be supplemented by the
  phonetic manifestation in utterances
• prosodic categories of isolated sentences by the
  prosodic structures of speech interaction

41

• The supplement is provided by systematic analysis
  of large corpora of speech interaction
• segmental and prosodic annotation on the basis of
  provisional phonological categories of lab
  speech, e.g. Kiel Corpus of Spontaneous Speech
• context-sensitive search operations
• measurements for sound classes and pitch patterns
  in search files
• statistics applied to symbol and signal data

42

• return to lab speech experiments on the basis of
  results of corpus analysis
• revision of the initial heuristic categories to
  bring them in line with the phonetics of speech
  communication
• phonology-coming-out-of-the-lab
• This progression of steps has largely been
  carried out in the analysis of f0 peaks in German.

43
4 Outlook

• Speech analysis is not just a metalinguistic
  academic pursuit
• but aims at describing and explaining language
  and speech behaviour
• in realistic communicative situations
• with reference to such central concepts as
  function, time and the listener
• and general principles in production and
  perception.

44

• There is growing unease with mainstream prosodic
  theory and practice, e.g. ToBI
• fair amount of rumbling at Speech Prosody 2002/4
• Yi Xu went as far as giving priority to function
  over lingistic form.
• When we combine this with Björn Lindbloms
  priority of substance over linguistic form, we
  capture the future of phonetics, which I have
  attempted to sketch in this paper
• the relation between function and substance
• linguistic form as derivative from it.

45

• This movement will gather momentum in years to
  come
• and the categories I have been talking about
  today will no doubt play a central role
• in the development of a comprehensive theory of
  speech communication
• and in the description of speech behaviour in the
  languages of the world.
• We will then have a new paradigm, the Paradigm of
  Function-Oriented Experimental Phonetics.