Linguistic Stress in Language and Speech

1
Linguistic Stress in Language and Speech

• Kenneth de Jong
• Indiana University

2
Chapter N1. Suprasegmentals

• The last chapter in phonetics descriptions
• Consists of Tone, Stress, Quantity, may-be
  juncture
• Or fundamental frequency, loudness, duration,
  may-be syllable stuff
• What do these have in common?

3
Chapter N1. Suprasegmentals

• Scary
• When doing basic transcriptions, we can sorta
  skip them -- e.g. no tonal minimal pairs.
• in English (most languages have tone
  contrasts, most languages have quantity
  contrasts)
• Tend not to fit well with a segmental model of
  phonetic structure.
• Vary with spoken context. Intonation is a
  property of the sentence duration varies overall
  by tempo
• Tend not to be as well understood
  (linguistically) as things like aspiration,
  point of articulation, vowel quality

4
Stress

• What is it?
• Why is it?
• What does it tell us?

5
What stress is phonetic observations

• OK, we do need it in transcriptions deepened
  vs. depend
• D.B. Fry (1955, 1958, 1965) perception
• F0 pattern (some complicated stuff about pitch)
• Duration (longer)
• Intensity (more intense)
• Other stuff (vowel quality more extreme)
• Stress vs. Accent making sense of context
• Accent F0 pattern varies qualitatively by
  context, e.g. statement vs. question
• Other stuff more attached to the word itself

6
What stress is Phonological observations

• Many languages have something similar to English
  stress
• Cross-language studies, such as de Jong
  Zawaydeh (1998) Arabic is surprisingly like
  English
• Various patterns appear in a number of languages
• Keeping track of them all creates things like
  metrical phonology

7
What stress isMetrical observations

• Reduced Contrast Unstressed items can have
  fewer contrasts.
• Domain Stress is expressed over a syllable.
• Alternation Stressed and unstressed material
  tends to be collated.
• Spacing Stresses tend to be distributed evenly.
  
• Accent Location Stressed items often are the
  site for accents.
• Culminativity Stresses may bear a one-to-one
  relationship with
• a higher-level unit, such as a phrase.
• Weight Sensitivity Stresses tend to fall on
  heavy syllables
• heavy syllables are ones with long vowels and
  sometimes
• consonantal codas.
• Boundedness Stress location is often fixed in
  relation to a
• location within a word.
• Boundedness Variation Stress locations may
  either be
• determined by position in morpheme or by weight
  sensitivity.

8
What stress isCharacterizing the other stuff

• Loudness vs. Clarity
• Loud people
• Brits Sweet (1892), Jones (1960) pulmonic
  force -gt heard as loudness
• Americanists Bloomfield (1933), Trager Smith
  (1951)
• More sophisticated Lehiste (1970), Beckman
  (1986)
• Clear people
• Brits Walker (1781) Jones (1960) prominence
  distinctness
• Swedish research Ohman (1967), Engstrand (1988)
• American speech Kent Netsell (1971) Harris,
  1978)

9
What stress is Characterizing the other stuff

• Loudness vs. Clarity
• Speech production work
• Similarities
• Open up vowels
• Close down consonants for contrast
• Make it longer
• Loudness is a way of being clear
• Differences
• Care with respect to targeting
• Being clear is harder than being loud

10
What stress is Characterizing the other stuff

• de Jong (1995)
• Compare production of words with /o/ in context
  of coronal consonants
• Use X-ray microbeam facility to see whats going
  on inside
• Found vowels with further tongue body retraction
• Degree of retraction was not predicted by
  duration increases, so it cant be due to
  undershoot mechanisms
• de Jong et al (1993)
• Consonant coarticulation
• de Jong (1998)
• Looked at articulation of post-vocalic /t/ /d/
  with stress variation
• Find variation due to something like degree of
  effort

11
Illustrative Results
Tongue tip movement patterns for phrase Put
the t__ . Solid unstressed Put, Dashed
stressed Put
12
What stress isHyperarticulation

• Clarity, sweet clarity
• Connected to Hyperarticulation (Lindblom, 1990)
• Speech production happens in a sea of variability
• Some of this is due to mode of production
• Hypoarticulation maximize production system
  considerations
• Hyperarticulation maximize likelihood the other
  guy will understand you
• Hyparticulation local to the syllable (de Jong,
  1995)
• More attention in production (de Jong, 1998)

13
What stress isTesting Hyperarticulation

• Lindblom hyperarticulation premium on getting
  information in signal
• Hyperarticulation happens in corrective focus
• I said bed, not chair.
• IF corrective focus gt hyperarticulation stress
  gt hyperarticulation,
• THEN stress corrective focus should have same
  effect as corrective focus.
• de Jong Zawaydeh (2000) de Jong (2004) test
  this with vowel duration and quality effects

14
de Jong (2004) results for voicing X vowel
duration

• Words like flowerbed
• bed longer than bet
• Focus makes difference bigger

15
de Jong (2004) results for voicing X vowel
duration

• Add words like bed - primary stress
• bed longer than flowerbed
• Stress focus have similar effect
• Stress focus get even larger effect

16
de Jong (2004) results for voicing X vowel
duration

• Add words like rabid and rabbit
• Much shorter
• No voicing difference
• Get a tiny effect with focus

17
Results, de Jong (2004)

• To Summarize
• Both stress and focus increase duration
• Both stress and focus increase duration contrast
  - specified differences get bigger
• Stress and focus interact so that contrasts get
  much larger in focused stress material
• Side note on stress shift

18
What stress isGeneral Attentional Model

• Other work on auditory attention in time (Jones,
  Kidd)
• Various properties
• Attentional selectivity some parts of a stimulus
  are more readily acted upon than others
• Attentional capture parts which change in
  salient ways tend to garner selective advantages
• Attentional integration aspects which work
  together to define an event get attended to as a
  unit
• Temporal expectancy events forming regular
  temporal patterns will focus attention on
  particular up-coming times

19
What stress isGeneral Attentional Model

• Stress some syllables are attentionally
  selected
• The attentional selectivity arises from
  attentional capture by acoustic events with
  sudden changes
• And may exhibit attentional integration where
  bits of speech which cohere and are regular form
  units
• Attention modulation can be governed by temporal
  expectancy, wherein high attention areas can come
  at regular intervals
• Attention modulation characterizes both hearer
  and speaker
• Speakers put important stuff in high-attention
  areas
• Hearers look for high-attention areas
• The match between speaker and hearer is A Good
  Thing

20
What stress isPhonological properties

• Reduced contrast unstressed low attention area
  a bad place for information
• Domain syllable onsets places of sudden change
  gt attentional capture syllables tend to be
  unitary acoustic objects gt attentional
  integration
• Alternation since attention is relative,
  attending to one event detracts attention from
  neighboring events

21
What stress isPhonological properties

• Spacing temporal patterning, especially regular
  spacing in time, tends to make high attention
  areas occur at regular intervals
• Accent location accents help direct attention to
  syllables which are hyperarticulated by the
  speaker
• Culminativity if stresses are attentional
  objects, having one stress per meaningful unit
  would make a mechanism which allows speakers to
  present speech a series of meaningful tasks
• . Good so far

22
What stress isPhonological properties

• Weight Sensitivity so why DO syllables with a
  final consonant tend to get stressed?
• Boundedness and why do stresses come in
  particular places in the word?
• Boundedness variation oh yeah? if there are
  good places in the word for stress, why are
  different languages so different with respect to
  WHERE?
• Actually if stress is so functional, Mr. Stress
  Man, why DO languages stress different syllables?
• Better take a good look at language differences
  

23
Korean Case Study Korean Stress Rule

• Korean stress rules
• Polianov (1936) if at end of utterance, stress
  the first syllable, otherwise stress the last one
• Huh (1985) Lee (1992) stress the first
  syllable always
• Lee (1974, 1985, 1989) Weight sensitive -gt
  stress a long vowel or an initial syllable with a
  coda, otherwise stress the second syllable
• Yu (1989) Weight sensitive -gt stress the first
  heavy syllable you come to, otherwise stress the
  last one
• Lee (1990), Kim (1998) Weight sensitive -gt
  stress a heavy first syllable, otherwise stress
  the second syllable
• Zong (1965), Cho (1967) Unbounded -gt its
  unpredictable so you just memorize it.

24
Korean Case Study Lim (2000)

• Lims Expectations

25
Korean Case Study Lim (2000)

• Lims Expectations
• Stress heavy first syllable

26
Korean Case Study Lim (2000)

• Lims Expectations
• Stress heavy first syllable
• Stress light second syllable

27
Korean Case Study Lim (2000)

• Lims Results
• No systematic differences by position
• No effect of weight on position

28
Korean Case Study Lim (2000)

• Production Results
• No systematic differences in vowel durations,
  except that vowels at the end are longer
• Syllable weight makes no difference
• Compares with Balinese production studies
  (Barber, 1977 Herman, 1998)
• Barber is a very reliable and experienced field
  worker who relied on impressionistic
  transcriptions
• Herman ran acoustic measurement studies
• No systematic differences in vowel durations,
  except that vowels at the end are longer
• Syllable weight makes no difference

29
Balinese Case Study Herman (1998)

• Barber (1977), first
• "There is no strong word-stress in Balinese in
  ordinary speech, there is only a slight variation
  in loudness and energy between the syllables of a
  sentence.
• Barber (1977), then (same page later on)
• "In words of more than two syllables (not
  counting suffixes), the penultimate syllable is
  stressed unless the vowel is e."

30
Balinese Case Study Herman (1998)

• Herman (1998), her comment
• "It is theoretically impossible to prove that
  some entity does not exist. Therefore, it is
  impossible to prove that word-level accentuation
  does not exist in Balinese. However, if
  word-level accentuation in some form did exist,
  one might expect to find certain indications of
  it."

31
Korean Case Study Lim (2000)

• Production Results
• No systematic differences in vowel durations,
• Summary
• KOREAN DOESNT HAVE STRESS
• Korean Intonation Tutorial

32
Korean Case Study Lim (2000)

• Point Though a stress system might be
  functional, languages work perfectly well without
  them.
• One more question so what do we hear as stress
  when listening to non-stress languages?

33
Korean Case Study Lim (2000)

• Perception Results - ask them to locate stress
• Korean listeners tend to say stress occurs in the
  vicinity of a pitch peak
• Pitch rises and falls in Korean are used to mark
  the edges of words
• Perception Results - suggests weight sensitivity
• The presence of consonants determines where,
  exactly, pitch peaks show up
• If stresses grow out of locations for pitch
  peaks, then consonants can indirectly determine
  where stresses get located
• This can explain weight sensitivity
• This explanation doesnt directly use attentional
  selectivity to consonants.

34
Why is stress?

• The functional nature of attention modulation.
  It has to do with the dynamics of speakers
  production systems and/or the dynamics of
  hearers perception systems and their linkage.
• The not particularly functional nature of
  language history. It has to do with the (much
  slower) dynamics of language groups.

35
What does it tell us

• The functional nature of stress
• Plasticity in production people are more skilled
  then they are given credit for.
• Acquisition patterns not all segmental material
  is created equal.
• Fluency complexity speech takes place in a sea
  of variability.
• The not particularly functional nature of
  language history.
• Cross language differences and bio-physical
  explanations
• Second language acquisition

36
de Jong (2004) results