Plosive-related glottalization phenomena in read and spontaneous speech - A stød in German?

This document is based on a poster presented at ICPhS 99 Satellite Meeting on Non-modal Vocal-Fold Vibration and Voice Quality, San Francisco, 31 July, 1999.

K.J. Kohler, IPDS Kiel, Germany

1. Definitions
2. Physiological explanation
3. Plosive-related glottalization phenomena in German
• 3.1 Conditions of occurrence in production
• 3.2 The Kiel Corpus of Read/Spontaneous Speech
• 3.3 Results of corpus analysis
• 3.4 Discussion of results
4. Glottalization phenomena in other languages
5. Explaining the data with reference to general principles of speech production
6. References

• glottalization phenomena = glottal stop and any deviation from canonical modal voice
  • glottal stop or low frequency irregular glottal pulsing (variable in frequency, amplitude and waveform) = glottalization
  • breathiness
  • breathy voice


• linguistic function of glottalization phenomena
• vowel-related glottalization phenomena
  boundary signal of a word or morpheme beginning with a vowel, typically in German, but also in other languages, e.g. English: variation between glottal stop and any other glottalization phenomenon along a scale of phonatory weakening
   
• plosive-related glottalization phenomena
  reinforcement of plosive by glottal stop, or replacement of plosive along a scale of phonatory weakening from glottal stop to any other glottalization phenomenon, e.g. in English or German
   
• syllable-related glottalization phenomena
  along a scale of phonatory weakening from glottal stop to glottalization, e.g. in the Danish stød
   

• paralinguistic function of glottalization phenomena
• laryngealization: prosodic phrase final relaxation of phonation = glottalization alternating with breathy voice and breathiness, but not with glottal stop
   
• truncation glottalization: tensing of phonation in utterance truncation = glottal stop alternating with glottalization (Nakatani and Hirschberg 1994, Local and Kelly 1986)
   

2. Physiological explanation

• glottal stop associated with feature of reinforcement, compared with weakening in other glottalization phenomena: extrapolation from physiological data on Danish stød
• different degrees of medial vocal fold compression
• due to increased/decreased vocalis and lateral crico-arytenoid activities
   
• phrase-final laryngealization (`creak' or `creaky voice') and its alternation with breathy voice or breathiness fits in with low-F0 utterance-final relaxation in preparation of glottal opening for non-speech function; this relaxation excludes glottal stop
   
• utterance-internal speech truncation before correction is most effectively achieved by cutting off the air stream at the glottal valve; thus tensing of the vocal folds for a glottal stop is the most natural process
   

• glottal stop and glottalization
  general conditions
  • simple glottal valve action to cut off air stream for stop articulation, added to, or instead of, more complex combination of supraglottal oral/velic closures
     
  • irregular glottal pulsing to reduce, rather than stop, air stream; different frequency from environment - higher/lower for tensing/relaxation, latter typical
     
  • stop not released into a vowel but in most cases followed by another complete or partial oral occlusion - nasal, plosive or lateral.
     
  specific contexts
  1. `sonorant - plosive - sonorant' (especially nasal)
      
  • for fortis and lenis stops at all places of articulation
  • word-internally or across word boundaries
  • word-internal sequence results from []-elision before nasals or laterals of canonical forms
  • oral closure in nasal - plosive and plosive - nasal adjusted to plosive place of articulation throughout
  • accompanied by velic opening as the complete or partial interruption of the air stream is transferred to the glottal valve
  • e.g. könnten [] or halten [] or Stunden [] or sind noch [], instead of the more elaborated canonical pronunciations [][][][]
  • timing of glottalization
    four possible temporal alignments of glottalization with the sonorant, e.g. /n/ in "könnten"
    1. medial [], most common in all contexts
    2. final [], next frequent for lenis stops
    3. initial [], next frequent for fortis stops
    4. complete []
    glottalization may also extend into preceding vowel
     
  2. `vowel - (fricative) fortis plosive - consonant' (esp. nasal)
      
  • higher probability of glottally reinforced plosives (with velic raising) than in (1)
  • in plosive - nasal same place adjustment as in (1)
  • in plosive - nasal velic opening may occur very early after oral occlusion, accompanied by glottal stop or glottalization
  • e.g. zweiten [], Leipzig [], hat nicht [].
     
• breathy voice and breathiness
  • in complete nasal context of (1) breathy-voiced or voiceless nasals instead of plosives also possible, breathy-voiced especially for lenis
  • must be due to glottal (interarytenoidal) opening: preservation of plosive phonation features, again combined with velic lowering, as required for the environment
  • modal-voice nasal context still interrupted by different type of phonation, reflecting more complex plosive articulations
     
• modal voice with(out) F0/amplitude modulation
  • for lenis in nasal context of (1) also further progression to modal voice, e.g. reduction to [nn] in "einverstanden"
  • for fortis only possible in unstressed function words, e.g. "könnten", and elements of compounds, e.g. "-zehnten" in numerals
  • may be complete, or there may be a weak trace of the plosive in the form of a medial amplitude and/or F0 dip in the nasal stretch
  • so speaker can still signal break, albeit towards the low effort end of a reduction scale ranging from plosive to complete nasalization

• two large, phonetically labelled corpora of read and spontaneous speech for German of approximately 31,500 and 37,500 running words, respectively: `The Kiel Corpus of Read/Spontaneous Speech'
   
• basically linear segmental phonemic transcription with componential additions (marking glottalization phenomena amongst others), using the SAMPA alphabet
   
• the read corpus and about one half of the spontaneous corpus have also been labelled prosodically using PROLAB, based on the Kiel Intonation Model (KIM)
   
• signal analysis and labelling environment xassp
   
• distribution of signal and label files on CD-ROM: one for read speech and 3 for spontaneous speech so far
   
• lexicon-oriented databank with awk retrieval tools
   
• library of awk search scripts for large number of phonetic questions, including glottalization phenomena
   
• wide array of phonation features related to plosive articulations are examined in these two data bases
• frequency distributions of the various phenomena, also in relation to unreduced plosive productions
• comparison for possible speaking style effect
• analysis is both symbol and signal oriented
   

Table 1

Frequencies of glottalization phenomena in the canonical pattern `sonorant - plosive - // - sonorant' in read (R) and spontaneous (S) speech, set against other realizations; (a) one // syllable (b) two // syllables in succession

(a)

1 syll	abs      R      %		abs      S      %
total	479	100.0	874	100.0
-schwa	425	88.7	866	99.1
glott	125	26.1	461	52.7
nas breath	4	0.8	76	8.7
nas	48	10.0	53	6.1
nas/lat plos	224	46.8	272	31.1
asp	24	5.0	4	0.5
+schwa	2	0.4	0	0
asp +schwa	52	10.9	8	0.9

(b)

2 syll	abs      R      %		abs      S      %
total	16	100	25	100
-schwa +schwa	12	60	10	40
-schwa -schwa	0	0	15	60
+schwa +schwa	4	40	0	0

Table 2

Frequencies of glottalization phenomena in the pattern `sonorant - plosive -// - sonorant' of monosyllables in read (R) and spontaneous (S) speech, for /ptk/ (1) and /bdg/ (2), set against other realizations

Table 3

Frequencies of prosodic boundaries occuring after items in phonetic classes of Table 1a; nas/lat plos, asp (+schwa), and +schwa have been combined to rest. Numbers in brackets are the frequencies of phrase-internal occurrences of the 4 classes; both numbers in each abs field add up to the sums of phrase-internal + phrase-final occurrences listed in corresponding classes of Table 1a

Table 4

Frequencies of glottalization phenomena in the pattern `vowel - (fricative) fortis plosive - // - sonorant' in read (R) and spontaneous (S) speech, set against other realizations

Table 5

Frequencies of glottalization (1) and of its absence (2) across word boundaries before initial nasals and after sonorants, vowels or other segments preceding the word-final plosive in read (R) and spontaneous (S) speech

3.4.1. Pattern (1): `sonorant - plosive - // - sonorant'

• // realization is rare in either corpus, but is even less frequent in spontaneous than in read speech (0.9% vs. 11.3%)
   
• decisive difference between the two speaking styles in proportion of plosives vs. glottalization phenomena (glottalization + breathy nasalization) + modal nasalization

  R	63.1% plosives	26.1% glottaliz	36.9% glottaliz phen + nasaliz
  S	32.5% plosives	52.7% glottaliz	67.5% glottaliz phen + nasaliz

  
  
• still holds for fortis and lenis plosives separately
  
  fortis

  R	14.3% //	75.5% plosives	23.1% glottaliz	24.5% glottaliz phen +nasaliz
  S	1.1% //	35.0% plosives	53.1% glottaliz	64.9% glottaliz phen + nasaliz

  
  lenis

  R	7.3% //	46.6% plosives	30.1% glottaliz	53.4% glottaliz phen + nasaliz
  S	0.0% //	19.1% plosives	51.1% glottaliz	80.8% glottaliz phen + nasaliz

  
  
• comparison of data for fortis and lenis in each speaking style shows increase of // deletion and substantial increase of glottalization phenomena for lenis, due to stop nasalization
   
• 2 // syllables in sequence:
  read sp 2nd // always, 1st 40% realized
  spont sp 1st // never, 2nd 60% not realized
   
• // syllables before prosodic boundaries
  comparable distributions across 4 reduction categories phrase-final and -internal:
  independence of plosive-related glottalization phenomena from position in prosodic phrase; but in R 2/3 plosives, in S 2/3 glottalization phenomena in both positions: greater degree of plosive reduction in S vs. R.

3.4.2. Pattern (2): `vowel - (fricative) fortis plosive - // - sonorant'

• // realization again rare in either corpus, but again more deletions in spont sp
   
• in both corpora nasal/lateral plosion predominant pattern with ca. 80%
   
• glottalization rare, but more frequent in spont sp
  i.e. frequency of glottalization completely different in the two segmental patterns in either speaking style

3.4.3. glottalization across word boundaries

• occurs in both patterns
   
• most frequent before nasal at beginning of next word

• paralinguistic function of glottalization phenomena
  (laryngealization, truncation glottalization) wide-spread in human speech communication; language universal on physiological basis?
   
• vowel and plosive related glottalization phenomena also found across languages, but with different distributions:
  British English
  • in stressed word-initial vowels
  • reinforcement of fortis plosives
  • replacement of intervocalic /t/
  American English
  • in stressed word-initial vowels
  • for fortis stops, also in nasal environment, e.g. Clinton
     
• syllable-related glottalization phenomena in Danish
  • prosodic phenomenon of certain syllable types, requiring sufficiently long stretch of voicing (long vowels or short vowels + sonorants
  • bound to primary or secondary stress
  • prosodic nature strengthened by relation to accent 1 words in Norwegian and Swedish
  • vennen (definite form of ven "friend")
  • [], besides emphatic []
  • coincides phonetically with possible realization of German wenden and American pronunciation of the name Fenton, but has different phonological function

  From the phonatory point of view German has a stød.

• in both patterns for plosive-related glottalization in German, 2 steps to simplify articulatory program
   
  1. elimination of central oral opening - closing gesture nasal or lateral plosion
      
  2. elimination of velic raising during oral occlusion and transfer of air stream interference to the glottis glottalization phenomena
  
  
• in bilateral nasal context of pattern (1), b. eliminates need for synchronization of velic control, velum can remain lowered in entire sequence, but listener still gets signal break for canonical plosive through a glottal stop, glottalization or some other change in phonation: articulatory reorganization with same function
   
• in unilateral nasal contexts of patterns (1) and (2), b. makes timing of velic control more flexible
   
• reduction in extent and synchronization of velic movement accords with a sluggish articulator
   
• in short closure lenis stops, variable velic timing can result in complete assimilation to modal-voice nasality
   
• temporal indeterminacy of velic action further heightened by flexible timing of phonation changes
   
• reduce demands on articulator coordination and help to ease production whenever called for by context of situation
   
• reading style, as against unscripted dialogue speech, sets a frame for greater articulatory precision; under this condition there are fewer articulatory reductions of the types described
   
• the same considerations apply to the plosive-related glottalization phenomena in English and any other language
   
• it is a great task for future research to investigate glottalization phenomena with a linguistic function in a large number of phonetically and phonologically diverse languages to arrive at utterance-based typologies and universals of phonation and its coordination with supraglottal articulation under different conditions of communication

(a) g086a004 KAK []	(b) g077a002 TIS []
(c) g074a000 HAH []	(d) g075a008 TIS []
(e) g091a014 FRS []	(f) g072a014 HAH []

Figure 1
Different manifestations of glottalization in "könnten" from the Kiel Corpus of Spontaneous Speech, 1 female (FRS) and 3 male speakers

Figure 2
Breathiness in "neunten", Kiel Corpus of Spontaneous Speech; g106a014 NAR, female speaker; speech wave, spectrogram, and SAMPA labels

Figure 3
Breathy voice in "-hängenden", Kiel Corpus of Spontaneous Speech, g105a009 NAR, female speaker; speech wave, spectrogram, and SAMPA labels

Figure 4
Amplitude and F0 dips in "-zehnten", Kiel Corpus of Spontaneous Speech, g115a002 REK, male speaker; speech wave, spectrogram, SAMPA labels, fundamental frequency, and energy

Figure 5
No plosive reflex in "-zehnten", Kiel Corpus of Spontaneous Speech, g146a000 BAC, male speaker; speech wave, spectrogram, and SAMPA labels

Figure 6
Shift of plosive and of vowel-related glottalization to the left and right into the vowel, respectively, in "achtzehnten elften", Kiel Corpus of Spontaneous Speech, g411a004 HEE, female speaker; speech wave, spectrogram, and SAMPA labels

Figure 7
Plosive and vowel-related glottalization and phrase-final laryngealization in "achtzehnten Oktober", Kiel Corpus of Spontaneous Speech, g083a003 KAK, male speaker; speech wave, spectrogram, and SAMPA labels

Figure 8a
Truncation glottalization in "(in einer) Messe/+ a=/+ in (Hannover auf der Messe)", Kiel Corpus of Spontaneous Speech, g274a008 SIH, female speaker; speech wave, spectrogram, SAMPA labels, and fundamental frequency

Figure 8b
Final breathiness in 2nd occurrence of "Messe" in utterance of Figure 8a

Figure 9
Truncation glottalization in "am O=/+ nach (Ostermontag)", Kiel Corpus of Spontaneous Speech, g315a009 SVA, male speaker; speech wave, spectrogram, SAMPA labels, and fundamental frequency

Figure 10a
Speech waves and F0 traces in 3 repetitions of "dem könnten wir uns (anschließen)", scripted lab speech, illustrating high-frequency glottalization for tensing; male speaker

Figure 10b
Corresponding spectrograms to Figure 10a

Fischer-Jørgensen, E. (1989): Phonetic analysis of the stød in Standard Danish. Phonetica 46, 1-59.

Kohler, K.J. (1996): Phonetic realization of German //-syllables. AIPUK 30, 159-194.

Local, J. & Kelly, J. (1986): Projection and "silences". Notes on phonetic and conversational structure. Human Studies 9, 185-204.

Nakatani, C.H. & Hirschberg, J. (1991): A corpus-based study of repair cues in spontaneous speech. JASA 95(3), 1603-1616.

© Klaus J. Kohler, ipds Kiel, 1999