Riede, Tobias : Vocal changes in animals during disorders

18

Chapter 3. Nonlinear phenomena - common components of mammalian vocalization or indicators for disorders: three case studies

Harmonic phonation is characterized by periodic vibration of the vocal folds. This oscillation is due to a repetitive sequence of the same vibration pattern, the duration of which is called the period. According to van den Berg's myoelastic-aerodynamic theory (1958) vocal fold vibration is based on a dynamic equilibrium between viscoelastic forces depending on mass, damping, length, and tension of the vocal folds, and aerodynamic forces related to the Bernoulli effect. The effective length, mass, and tension of the vocal folds are determined by muscle action, which allows the fundamental frequency and the waveform of the pulses to be controlled. The vocal tract then acts as a filter which transforms the primary signals (Fant 1960; Fitch & Hauser 1995).

As explained in chapter 2.2, the vocal folds can be considered as two coupled oscillators and constitute a highly nonlinear self-oscillating system (Herzel et al. 1995). Nonlinearity means that the factors (vocal fold amplitudes, glottal air flow, intraglottal pressure) vary in ways that are not linearly proportional to each other. This results in a complex relationship between pressure and flow. Nonlinear systems display a number of typical phenomena which are briefly described here. Aperiodic (or chaotic) oscillations are characterized by irregularity, and in extreme cases there are no repeating periods at all. Period doubling (subharmonic regime) is another characteristic of nonlinear dynamical systems. It is characterized by a sudden change in the frequency of the oscillations, such that the spacing between spectral components is halved.

In special cases such as vocal fold paralysis or anatomical asymmetry of the larynx, these frequencies can be detuned, causing desynchronized vocal fold vibrations. Subharmonics in the oscillation spectrum often correspond to integer ratios of the frequencies of the left and right vocal fold (e.g. 1:2 or 2:3) (Steinecke, Herzel 1995). This phenomenon is termed frequency locking or entrainment and results in a vibration pattern characterized by more than one oscillation maximum. If the ratio of two frequencies is not a rational number, the dynamics corresponds to a torus - a superposition of two independent frequencies. The coexistence of two audible frequencies has been termed biphonation. Subharmonic regimes and biphonation show often sudden transitions to irregularity. Such chaotic oscillations show no repeating pattern over the duration of the vocal segment. By the phrase "nonlinear phenomena" hereafter it is refered collectively to subharmonics, biphonation, deterministic chaos, or any subset of these.


19

Nonlinearities are found in normal phonation of humans (e.g. Fields, 1973; Stark, Rose, McLagen, 1975; Buhr, Keating, 1977; Kent, Murray, 1982; Robb, Saxman, 1988; Titze et al. 1993) and they have been characterized as an integral part of mammalian vocalization (Wilden, Tembrock, 1994; Tembrock, 1996 a, b; Wilden et al. 1998; Brown, Cannito, 1995). Moreover, they are relevant as indicators of pathologies (e.g. in humans: Sirviö, Michelsson, 1976; Herzel et al. 1994; Omori et al. 1997). That is, the occurrence of nonlinear phenomena increases during disorders of the vocal apparatus or some kinds of systemic diseases with impacts on phonation. The voice, as an often used instrument is also in animal communication is suspected of showing variability due to the vocalisers state of health.

The conspicuous rough vocalization of three cases have been observed. Spectral analyses allowed us to relate the conspicuous audible roughness to the occurrence of biphonation, subharmonics and chaos.

3.1 The Japanese macaque infant

This chapter is a revised version of: T. Riede, I. Wilden, G. Tembrock (1997): Subharmonics, biphonations, and frequency jumps - common components of mammalian vocalization or indicators for disorders? Zeitschrift für Säugetierkunde 62 (Suppl. II): 198-203. (Riede et al. 1997)

The starting-point of this study was the detection of the three acoustic phenomena - subharmonics (SH), biphonations (BP), and frequency jumps (FJ) - in the repertoire of infant Japanese macaques (Macaca fuscata) and the comparable higher amount of these phenomena in the repertoire of one infant with a metabolic disease. Unfortunately, the disease could not be diagnosed in more detail.

The following question was the basis of the study: Are there differences in the frequency of irregularities such as SH, BP, FJ in the repertoire of healthy and the ill individual?


20

3.1.1 Material and Methods

Recordings for this investigation were originally made within a different study (Riede, 1997). Vocalizations from 10 individually known infants were recorded several times during September 1994 to January 1995, in the Zoologischer Garten Berlin and the Tierpark Berlin-Friedrichsfelde. Recordings were done with a NAGRA SN tape recorder and a NAGRA directional microphone. For spectral analysis we used the HYPERSIGNALTM-MACRO software package. 200 calls per animal undergone a 'Fast Fourier Transformation' (FFT), with 40 kHz sampling frequency and 512 points FFT order, i.e. narrow band analysis. Narrow band analysis is essential to recognize the phenomena in the frequency domain display.

The context was always the same: the infant tried to come in contact with its mother by crying and staying in one place. As soon as the mother came in ventroventral contact with the infant, the infant stopped crying.

The quantitative analysis was done by judging each call on the existence of the expected phenomena. The relative frequency (in %) of the occurrence of the phenomena was calculated.

3.1.2 Results

We found SH, BP, FJ. All three phenomena occur in various expressions (Fig. 3.1). FJ appear in the spectrogram as sudden changes in the fundamental frequency (f0) (Fig. 3.1a). This phenomenon is similar to the jump between two voice registers in humans (e.g. between modal and falsetto register). The most characteristic feature is the break between the two f0 regimes. This break can be represented by a pause or by a noisy segment (Fig. 3.1a).

SH are frequencies that lie between or below the harmonic frequencies and are rational divisions of the f0 or their integer multiples of, for instance, one half, one third, refering to period doubling or period tripling (Fig. 3.1b).

When two independent and audible pitches are simultaneously produced we speak of BP. In the case of the Japanese macaques the lower f0 is entrained to the higher one, resulting in a very typical picture in the spectrogram - 'side bands', i.e. parallel bands around the f0 and the harmonics (Fig. 3.1c) and in the time series we find the typical 'beats' (Fig. 3.1c).

The quantitative analysis was done with the first 200 recorded calls from each of the 10


21

infants. Each call was judged on the existence of the expected phenomena. The amount of irregularities ranges between 3.5 % and 45 %, i.e. 45 % of the calls of one infant contained FJ, BP or SH (Tab. 3.1). The ratio of the three phenomena shows a very individual specific pattern.

Figure 3.1a: 3.1a shows the time series and the spectrogram of two single calls each with a frequency jump (FJ). The FJ appears as a sudden change of the fundamental frequency. It is indicated by the arrows.


22

Figure 3.1b: 3.1b shows the time serie, the spectrogram, the averaged power spectrum and a zoomed segment of the time series of a call containing subharmonics (SH), indicated by the arrow in the spectrogram. The SH appear as parallel lines between the overtones (=’harmonics‘) of the fundamental frequency. The power spectrum gives further information. The energy peaks of the SH lie in a determinated distance (indicated by the numbers in the power spectrum) of the overtones of the fundamental frequency (f0 = 1.0 kHz), here the determinated distance is about 0.5 to 0.55 kHz. The subharmonic segment gives also a charcteristic picture in the time series if zoomed.

Figure 3.1c: In 3.1c the biphonic call shows two characteristic features, firstly, some elements can be found where the lines between the overtones of the original fundamental frequency are not parallel, and secondly, the distance of the lines is not necessarely related to the original f0. The distance between the parallel lines (here: 0.45 kHz) represent the f0 of the second pitch, which is only represented by its overtones in the spectrogram. The energy of the f0-line of the second pitch is too low that it can not be represented in the spectrogram. The zoomed segment of the time series gives a characteristic picture.


23

Figure 3.1: The three phenomena are presented. 3.1a shows the time series and the spectrogram of two single calls each with a frequency jump (FJ). The FJ appears as a sudden change of the fundamental frequency. It is indicated by the arrows. 3.1b shows the time serie, the spectrogram, the averaged power spectrum and a zoomed segment of the time series of a call containing subharmonics (SH), indicated by the arrow in the spectrogram. The SH appear as parallel lines between the overtones (=’harmonics‘) of the fundamental frequency. The power spectrum gives further information. The energy peaks of the SH lie in a determinated distance (indicated by the numbers in the power spectrum) of the overtones of the fundamental frequency (f0 = 1.0 kHz), here the determinated distance is about 0.5 to 0.55 kHz. The subharmonic segment gives also a charcteristic picture in the time series if zoomed. In 3.1c the biphonic call shows two characteristic features, firstly, some elements can be found where the lines between the overtones of the original fundamental frequency are not parallel, and secondly, the distance of the lines is not necessarely related to the original f0. The distance between the parallel lines (here: 0.45 kHz) represent the f0 of the second pitch, which is only represented by its overtones in the spectrogram. The energy of the f0-line of the second pitch is too low that it can not be represented in the spectrogram. The zoomed segment of the time series gives a characteristic picture.

Table 3.1: The table shows the actual distribution of subharmonics, biphonations and frequency jumps in the vocalization of ten individuals. The relative frequency (rel. fr. [%]) of all three phenomena was calculated from 200 calls per individual.

Infant

1

2

3

4

5

6

7

8

9

10

sex

M

m

m

m

m

f

f

f

f

f

SH

3

0

2

22

6

2

0

7

10

2

BP

2

4

1

4

2

1

4

11

8

22

FJ

4

6

9

30

76

4

9

9

5

66

rel. fr. [%]

4.5

5.0

6.0

28.0

41.9

3.5

6.5

13.5

16.5

45.0


24

3.1.3 Discussion

FJ, SH, BP appeared in the vocalizations of 10 Japanese macaque infants. They occured in sufficient numbers to merit consideration as likely features of occurrence in the infant's vocalization. Like in the vocalization of the African Wild Dog (Wilden, Tembrock, 1994), as well as in the vocalization of many other mammals (Tembrock, 1996 a, b; Wilden et al., 1998), in the vocalization of Japanese macaques nonlinear phenomena can be found.

The quantitative analysis should give us an idea about the distribution of these phenomena in one age class of Japanese macaques. The amount of irregularities produced by infants range between 3.5 % and 45 %. The frequency of SH, BP, FJ in the human voice shows also a relatively high variability. That was shown by quantitative voice analyses of healthy humans: Keating, Buhr (1977): 6 individuals, age 33-169 weeks: SH, BP, FJ in 0 - 36 % (range) of the calls per person; Keating (1980): 4 individuals, age 16-69 weeks: in 13,8 % (mean) of the calls; Robb, Saxman (1988): 14 individuals, age 11-25 months: in 6 % (mean) of the calls. In these investigations newborns and infants were considered. Anatomically, the laryngeal structures undergo substantial growth during this time (Hirano et al., 1981) as do the neurological systems involved in speech and language (Netsell, 1981). The higher probability of occurrence of SH, BP, FJ in the vocalization of newborns and infants may be caused by lower control abilities. Adult humans learn to control their voice, but are still able (to varying extent) to produce these phenomena intentionally. For that reason it would be interesting to observe the amount of irregularities in the whole repertoire of the macaque species, to see if the amount of irregularities can be a measure for the controll over vocal utterances. Assuming infant vocalization under a very low level of control, in contrast, we would expect for instance the vocalization during aggressive encounters as more controlled, since it shows in some cases some very ritualized characters.

One of the female infants produced the highest amount of irregularities - 45 %. This infant also showed clear indications of a disease - underweight, a sudden and total loss of hair and black pigmented skin. This disease could not be diagnosed in more detail. The rest of their behaviour was not conspicuous in comparison to infants of the same age. The contact time with the mother was not significantly increased. We assume a metabolic disease due to the lack of any other conspicuous behaviour than the acoustic behaviour and the visible symptoms. Other infants showed no symptoms.

From the simultaneous appearence of the acoustic phenomena and the external symptoms we can not automatically conclude a causal relationship. Furthermore we could only


25

investigate one ill individual. However, there are several investigations of diseases in human infants which mention the simultaneous occurrence with irregularities in the voice:

1) severe diseases of the central nerval system in children: Sirviö, Michelsson (1976); Michelsson et al. (1977); Wermke (1986)

2) premature and asphyxiated newborn infants: Michelsson (1971)

3) severe malnourished children: Juntunen et al. (1978)

4) hearing impared children: Monson (1979)

5) acute laryngitis or papilloma in adults: Herzel, Reuter (1996)

The ratio between the described acoustic phenomena seems to change in a characteristic way for each disease. Michelsson et al. (1977) examined the vocalization of children suffering from meningitis (until the age of 6 months) compared to healthy infants. They found an increase in the occurrence of BP, a decrease in the occurrence of SH and no significant change in the occurrence of FJ. A change in vocalization during illness as well as a "normalization" after beginning the therapy could be shown.

In the Japanese macaque infant with the assumed metabolic disease the high amount of FJ was most significant. However, in Japanese macaque infants the FJ are much more common than the BP and SH.

3.2 The domestic cat infant

This chapter is a revised version of: T. Riede, A. Stolle-Malorny (1998): Spektrale Analyse der Stimmveränderung bei einem Kater mit Schädel-Hirn-Trauma. Kleintierpraxis 43: 773-780, and T. Riede, A. Stolle-Malorny (1999): The vocal change of a kitten with cranio-cerebellar trauma - a case study. Bioacoustics 10: 131-141. (Riede, Stolle-Malorny 1998, 1999)

The case involved a 3 month old male cat with craniocerebellar trauma (CCT). The animal suffered an accident in which a heavy board fell on its head, causing the CCT. During the 8 day clinic stay, the vocalisation of this kitten seemed markedly different from that of similarly-aged kittens. Also conspicuous was a vocal change on the seventh day in the clinic, which was correlated with a general clinical impression of improvement. Here, we define a 'vocal change' as short term reversible change as distinct from an ontogenetically caused, irreversible change of the animal's voice.


26

The purposes of this paper is to give a case report of the vocal changes during the course of recovery in this cat.

3.2.1 Case history and course of the disease

Admission: A three month old male Maine Coon kitten (1.3 kg body weight), with a CCT caused by a heavy board which fell on its head from the left was admitted to the Clinic for Small Animals, Free University of Berlin. The cat was in lateral recumbency, disoriented, showed hemorrhage from both ear canals and shock symptoms, e.g. shallow breathing (12 per min.) with an increased bronchial breathing sound, fast and shallow pulse (200 per min.), rectal temperature 34.6°C, white mucosa. Additionally a left sided soft tissue swelling and skull deformation observed above the eye. Shock therapy was immediately initiated with 25 mg Prednisolon i.v., 5 mg Furosemid i.v., and 200ml Sterofundin i.v.. 30mg Amoxicillin i.v. and 3mg Diazepam i.v. for mild sedation was also applied. For an improved respiration and oxygen flow to the brain, the cat was transferred to an oxygen cage. The blood investigation showed no other abnormalities than a leucocytosis (34000/µl). After stabilization a neurological examination was performed. The cat was disorientated and unable to stand. An increased muscle tone of all extremities and from time to time involuntary paddling movement of all four limbs could be observed. It showed a torticollis to the right side, but no positional or spontaneous nystagmus. Physiological nystagmus could not be observed. Furthermore the pupils were miotic and direct and indirect pupillary reflex was absent. There was no menace response on both sides. There was a decreased facial sensitivity. Both ear canals showed bleeding. Damage to the ear drums was suspected but could not be demonstrated. By taking radiographs no skull fracture could be demonstrated. Radiographic pictures of the thorax and abdomen showed no abnormalities. The medication during the clinical stay was 30mg Amoxicillin, initially i.v., later oral administration, twice a day, and Dexamethasone starting 0.1 mg per day and 0.05 mg per day from the 5th day onwards. In addition eye treatment with tears substitute (Vidisic) and antibiotics (Refobacin) was applied. Up to the sixth day a fluid substitution was given.

2nd day: The cat was responsive. It was able to stand with support but still circling and falling to the right. It showed a head tilt to the right. The cat showed an anisocoria with the narrower pupil on the left side. The facial sensitivity was still absent and facial nerve deficits were significant on the left. The animal appeared to be blind because the 'following movement test' and the 'menace test' were negative. The cat did not react to acoustic stimuli. It was unable to consume food independently.

3rd day: The cat was awake but it seemed still disorientated. It tried to stand up but always fell down again, but could succeed given manual support. Movement with a tendency to the right and a mild head tilt to the right still could be observed. Anisocoria was still present as


27

well as an absent palpebral reflex and absent corneal reflex on the left. The left cornea was dry with a diffuse edema and a mild anterior uveitis was present. During the fourth and fifth day no significant improvements were observed.

6th day: An improvement of the general clinical impression was recognized. The kitten was bright, alert and responsive, no longer leaned to the right and had no more head tilt. It showed moderate ataxia. On the left eye a mild anterior uveitis with a narrower pupil was still present. The cat still seemed to be blind, i.e. the 'menace test' and the 'following movement test' were still negative.

7th day (one day before clinic discharge): The cat was active, it showed minor ataxia, i.e. the animal was able to walk with assurance and to feed independently. A partial failure of the facial nerve still remained, it could not close the left eye. The visual tests were still negative, but the animal avoided obstacles and followed an object visually if the object moved slowly 20 cm in front of the cat. The cat reacted to acoustic stimuli. The clinical improvement was so good that the cat was discharged the following day with continuing cornea treatment (tears substitute).

60 days after clinic discharge: The cat walked with stiff hind legs. Its visual abilities were restored. It was still unable to close the left eye.The skull deformation on the left side could still be palpated. The cat reacted when its owner called to it.

3.2.2 Acoustic analysis

The kitten was housed in a wire cage (50 x 50 x50 cm) coated with a 5 cm thick foam layer. Spontaneously uttered calls were recorded from the third day on in the clinic using a cassette recorder (SONY Professional) and a directional microphone (Sennheiser ME 80) on chrome-II-oxide cassettes (BASF). Recording conditions were as follows: distance between microphone and cage was 10 cm, but the animal could move freely in the cage, so the distance between microphone and the kitten effectively varied from 10 to 40 cm. Spectrographic analysis was completed using Fast Fourier Transformation, 512 points, 75% frame overlapping, 30 kHz sampling frequency, Hamming window. Acoustic analysis was performed using the software package HYPERSIGNALTM-MACRO.

10 frequency and time parameters were extracted from the spectrogram and the spectrum.

All recorded calls can be categorized as 'isolation calls' according to Roman, Ehret (1984), Carterette et al. (1979).

Additionally we characterised the amount of nonlinear phenomena. Certain types of acoustic


28

events are referred to as nonlinear phenomena (i.e. subharmonics, biphonation, deterministic chaos). These phenomena are normally audible and are detectable in the spectrogram and the frequency spectrum. There are different hypotheses considering the generation of nonlinear phenomena, but all agree that they are generated by particular oscillation patterns of the vocal folds (Herzel et al. 1994). In this paper we use the vocabulary suggested for mammal vocalisation by Wilden et al. (1998).

Figure 3.2: Spectrographic representation of the nonlinear phenomena, subharmonics, chaos, biphonation, in the calls of the cat. Harmonic windows appear within chaotic segments.

The nonlinear phenomena appear in the cat as follows (fig. 3.2):

Subharmonic frequencies: frequencies that lie between or below the harmonic frequencies and are rational divisions of the preceding fundamental frequency (e.g. 1/2, 1/3) or their integer multiples.

Biphonation: Phonation with two independent pitches, acoustically observed as two non-commensurate fundamental frequencies which can appear as nonparallel harmonic lines in a spectrogram as either or both pitches change.

Deterministic Chaos: Chaotic segments often follow abrupt changes between harmonic and aperiodic call segments. They are preceded by subharmonics. Within chaotic segments there are often harmonic windows.

3.2.3 Results

A total of 158 calls were analysed. 143 calls were recorded during the stay in the clinic and


29

15 calls after clinical discharge at the animal's home. For technical reasons, all 10 parameters could not be measured in all calls. In 141 calls, all 10 parameters were measured. All 158 calls were analysed to determine the prevalence of nonlinear phenomena.

Table 3.2: Means and standard deviations of 10 parameters extracted from the calls: f0 A-fundamental frequency at the beginning of the call; f0 E- fundamental frequency at the end of the call; f0 max - maximal fundamental frequency; t max. - distance between the beginning of the call and the point of maximal fundamental frequency; t gesamt - total length of the call; 1st quart, 2nd quart, 3rd quart - point of first, second and third energy quartil; f1/f0 - ratio of the relative amplitudes of the second harmonic and the fundamental frequency; f-peak - frequency with the highest peak in the spectrum; Hz - Hertz, ms - Milliseconds; N - number of calls investigated. On the 8th day the animal was discharged from the clinic. On two days thereafter the animal was acoustically recorded.

 

3rd day in the clinic

N= 39

4th day in the clinic

N= 44

5th day in the clinic

N= 1

7th day in the clinic

N= 46

1st day after Entlg

N= 6

55th day after Entl. N= 5

f0 A (Hz)

873±178

857±127

719

1064±198

1310±179

1190±176

f0 E (Hz)

902±73

788±59

883

789±88

1003±156

1179±192

f0 Max (Hz)

1201±83

1076±87

1048

1143±146

1403±161

1398±83

t max (ms)

234±105

189±149

123

78±94

145±83

322±73

t gesamt (ms)

884±279

1108±212

1696

797±194

1302±185

868±122

1st.quart (Hz)

2931±655

3288±625

3093

2870±860

4035±1001

3534±630

2nd quart (Hz)

4125±740

4659±730

4144

4399±1033

4444±390

3631±920

3rd quart (Hz)

6210±1280

6440±730

6128

6763±1473

7137±430

6531±1102

F1/f0

5.5±5.3

5.1±5.4

4.3

3.3±3.7

7.2±6,1

1.8±3.2

f-peak

3536±1089

3776±883

3109

3349±1314

3023±1715

1799±1112

a) parametric call description

third and fourth day in the clinic: Calls on these days show a harmonic structure, i.e. in the spectrogram the fundamental frequency (= first harmonic) is clearly visible and a number of overtones (= second up to n-th harmonics) are to be found in a distance of an integer multiples of the fundamental frequency as parallels (fig. 3.2). The frequency modulation (i.e. the change of the fundamental frequency over time, here represented by the difference of


30

'start fundamental frequency' and 'maximal fundamental frequency' related to 't-max') of circa 200 to 250 Hz is restricted to a very short call segment at the beginning of the call. This results in an audible impression similar to a long sung tone, but not typical of a normal cat 'meow' (Buchwald, Shipley 1985). The calls were very loud compared to the calls of other hospitalised cats. Sound pressure level measurements are unfortunately lacking.

seventh day in the clinic: On the fifth and sixth day it was impossible to record a sufficient number of calls since the animal did not call spontaneously. On the seventh day the calls were again very numerous. The maximal fundamental frequency is very near to the beginning of the call, the call often starts with the highest fundamental frequency. This results in a very "pressed" auditory impression. The vocalisation was not as loud as on the previous two days.

Table 3.2 gives the mean values for the acoustic parameters. The differences comparing the fourth and seventh day were significant for the start frequency (fundamental frequency at the beginning of the call) (t-test, N1=44, N2=46, T=5.99, P<0.0001), the maximal fundamental frequency (t-test, N1=44, N2=46, T=2.63, P<0.02), the total length of the call (t-test, N1=44, N2=46, T=7.23, P<0.0001), the location of the maximal fundamental frequency (t-test, N1=44, N2=46, T=4.17, P<0.001), the first energy quartile (t-test, N1=44, N2=46, T=2.91, P<0.01), the ratio of the relative amplitudes of the second harmonic and the fundamental frequency (f1/f0) (t-test, N1=44, N2=46, T=2.42, P<0.02), the peak frequency (t-test, N1=44, N2=46, T=2.0, P<0.05), but not for the fundamental frequency at the end of the call (t-test, N1=44, N2=46, T=0.08, P=0.94), the second energy quartile (t-test, N1=44, N2=46, T=1.46, P=0.14), the third energy quartile (t-test, N1=44, N2=46, T= -1.37, P=0.17).

b) audible impression and nonlinear phenomena

A high amount of nonlinear phenomena were found on the seventh day and one day after clinical discharge as well.

All 157 calls were individually inspected for the occurrence of nonlinear phenomena, results are given in Table 3.3.


31

Table 3.3: Number of calls which contained nonlinear phenomena. SH - Calls containing Subharmonics, BP - Calls containing Biphonationen, CH - Calls containing Deterministic Chaos, SH+CH - Calls containing Subharmonics and Deterministic Chaos, N - number of calls investigated

 

3rd day in the clinic

N= 39

4th day in the clinic

N= 44

5th day in the clinic

N= 1

7th day in the clinic

N= 59

1st day after discharge

N= 7

55th day after discharge N= 8

SH

0

3

0

21

3

0

BP

0

0

0

2

1

0

SH+CH

0

0

0

16

1

0

CH

0

0

0

2

2

0

gesamt (%)

0

6.8

0

69.5

100

0

Figure 3.3a: Three calls from the third day in the clinic. The calls show a harmonic structure. The middle call starts with a noisy (chaotic) call segment. Some few calls from the fourth day show non linear phenomena. The time series are represented above each spectrogram to give an impression of the amplitude modulation.


32

Figure 3.3b: Three calls from the seventh day in the clinic. These calls show harmonical segments with clear fundamental frequency and further harmonics. We also found a high amount of calls containing segments with non linear phenomena. In the left there is a chaotic segment passing over in a short subharmonic regime and ending with a harmonic structure. In the middle call biphonation can be seen. In the right call we found chaotic and subharmonic segments as well as harmonic windows.

Figure 3.3: 3.3a: Three calls from the third day in the clinic. The calls show a harmonic structure. The middle call starts with a noisy (chaotic) call segment. Some few calls from the fourth day show non linear phenomena. The time series are represented above each spectrogram to give an impression of the amplitude modulation. 3.3b: Three calls from the seventh day in the clinic. These calls show harmonical segments with clear fundamental frequency and further harmonics. We also found a high amount of calls containing segments with non linear phenomena. In the left there is a chaotic segment passing over in a short subharmonic regime and ending with a harmonic structure. In the middle call biphonation can be seen. In the right call we found chaotic and subharmonic segments as well as harmonic windows.

Nonlinear phenomena were present during the fourth day (fig. 3.3) in the clinic but not on the third or fifth day. The amount increased on the seventh day and decreased after clinic discharge. Two months after clinic discharge nonlinear phenomena were not found.

3.2.4 Discussion

Vocal changes in a three month old kitten with craniocerebellar trauma were described by using a parametric call description method and a description of nonlinear phenomena. The


33

animal showed a vocal change on the seventh day of hospitalisation which coincided with clinical improvement.

In the CCT cat an increase in fundamental frequency parameters and a shorter total call length was observed. In human neonates the majority of disorders with a vocal change result in abnormally high fundamental frequencies (Furlow 1997). No uniform behaviour could be observed in the time parameters (e.g. total call length) during the several disorders (Furlow 1997).

The amount of nonlinear phenomena give a clear representation of the vocal change on the seventh day. Nonlinear phenomena have also been described in the non verbal vocalisation of human infants (Michelsson 1980; Robb, Saxman 1988). These phenomena occur in the voice of normal human infants but more during several diseases (Sirviö, Michelsson 1976; Herzel et al. 1994). In the vocalisation of normal animals they are often quite common as well (Tembrock 1996a, b; Wilden et al. 1998) and they may increase during diseases (Riede et al. 1997 and this study).

What is the reason for the vocal change?

The cat did not react to acoustic stimuli suggesting the presence at least of a temporary deafness. The deafness may be the reason for an increased sound pressure level. Auditory feedback is essential for a normal vocalisation (Shipley et al. 1988). The calls on the third and fourth day were very loud suggesting that the hearing ability during these days was reduced. The bleeding from both ears supports this assumption as well, which may have been caused by damage of the base of the skull and/or by the destruction of the outer ear (ear drum). Shipley et al. (1988) described an increase in call length and sound pressure level in experimentally deafened cats compared to normal litter mates. They observed no compensation of this parameter change in deaf cats over a period of 3 years. Thus, our cat showed a temporary auditory disorder, since the loudness decreased and the total length of the 'meow' calls decreased, and the cat reacted to acoustic noise after the seventh day in the clinic. To summarise: The cat may have produced highly harmonic calls with increased loudness and call length, reduced frequency modulation and only few nonlinear phenomena due to a decreased motor control over the voice production apparatus and a restricted hearing ability. The motor control and the hearing ability recovered during the course of observations, resulting in a vocal change.


34

The higher amount of nonlinear phenomena after the vocal change (after the 6th day) may be the result of a high charge of the voice generating apparatus. During the third and fourth day in the clinic we observed a high rate of vocalisation. This could have caused a hoarseness due to exhaustion of the vocal folds of this 3 month old kitten which was recognisable on the seventh day. Hoarseness can occur following extensive use of the voice production apparatus in humans, i.e. hoarseness can be the result of exhaustion in vocally untrained humans (Wendler et al. 1996). Hyperphonation have also shown in dogs as a source for vocal fold damages (Gray et al. 1987; Gray, Titze 1988). Spectrally hoarseness might be represented by the occurrence of nonlinear phenomena (Omori et al. 1997).

The high correlation between the vocal change and the clinical general impression suggests that further systematic studies of vocalisation in ill animals are warranted.

3.3 The dog-wolf hybrid

This chapter is a revised version of: T. Riede, H. Herzel, D. Mehwald, W. Seidner, G. Böhme, E. Trumler, G. Tembrock: Nonlinear phenomena in the natural howling of a dog-wolf mix. J. Acoust. Soc. Am. in press (Riede et al. in press)

The observed nonlinearities occurred in the female's vocalization during chorus howling of the pack. The howl is the best studied form of acoustic communication in wolves. It is a frequency modulated harmonic vocalization and plays a major role in territory maintenance, pack integration and individual recognition (Harrington, Mech 1979; Harrington 1989). It is a form of communication that is effective over long distances (e.g. 1-2 km). A single wolf usually begins howling and is followed after some time by other pack members (Joslin 1967). Three or more animals can be involved in the chorus (Klinghammer, Laidlaw 1979) which can start spontaneously, i.e. without an obvious release, or may follow a special trigger (Theberge, Falls 1967; Klinghammer, Laidlaw 1979).

Our aim is to understand the underlying physiological mechanisms of the nonlinear phenomena observed in this animal. It is known from human vocalization studies, that a variety of mechanisms can induce nonlinear phenomena. For example left-right asymmetries and strong source-tract coupling can desynchronize vocal fold vibrations (Mergell et al. 1997).

Unlike in humans, no direct investigation of the vocal fold vibrations with stroboscopy or high-speed glottography was possible. Consequently, we have to rely on indirect data to discuss


35

the mechanisms of the instabilities. First, we compare the animal vocalizations with a surprisingly similar voice of a young woman, in whose case the underlying mechanism of the voice instabilities is well understood. Second, we perform a post mortem investigation of the animal's larynx to look for anatomical peculiarities.

3.3.1 Material and Methods

Studied animals

The observed animals lived in a group. All were derivatives from a male hybrid (Persian wolf, New Guinea singing dog, Australian dingo) and a female mixed-breed (Golden shakal, Elk dog, Siberian husky). Since 1981 this pack (in 1997 it consisted of 8 animals: 5 males and 3 females) had lived without any further genetic influence on a 5 000 m² area of the Eberhard-Trumler-Station (Birken-Honigsessen, Germany).

The female with the conspicuous voice (named 'Schaka') died in December 1997 in the age of 7.9 years and with 18 kg body weight after a severe undetermined disease. A male (4 years) and an additional female (5 years) provided larynges for anatomical comparison. These two animals died after aggressive interactions with another dog group. All three animals were deep frozen until dissection. Unfortunately a detailed pathological investigation of the carcasses was not possible for technical reasons.

Acoustical analysis

Video and audio recordings of the animals were made within a different study (Mehwald 1998). Video and audio recording was done with two cameras (Video-Hi 8, Blaupunkt, CR 8700H and CCD-V200E Video 8 PRO). Recordings were made in two observation periods, first period: June 1994, 15 choruses recorded, second period: April to June 1997, 64 choruses recorded. Three (from 1994) and twelve (from 1997) choruses respectively were available in which one, two or three of the anatomically investigated animals were clearly recognized. Additional recordings served for comparative purposes.

Spectral analyses were made with HYPERSIGNALTM-MACRO software. The calls were analyzed using the Fast Fourier Transformation (FFT), with 8 kHz sampling frequency and 512 points FFT order, i.e. narrow band analysis. Hanning windows and a 75% overlapping of the successive windows were applied.


36

For the safe identification of an individual's calls during the howling chorus, the spectral analysis was compared with the simultaneous video recordings. This procedure was repeated by a second investigator and only those calls identically identified by both investigators were used for analysis.

For the description of the nonlinear phenomena in the calls we used the nomenclature as suggested by Wilden et al. (1998) for the mammal vocal repertoire. Subharmonics are characterized by parallel bands between pre-existing harmonics or at multiples of one third of the original pitch. Chaos is associated with abrupt transitions to noise-like segments. Periodic windows often appear within these chaotic segments. Biphonation is characterized by a series of non-parallel bands related to two independent pitches.

For quantification we used a method used in humans (Robb, Saxman 1988) and once in macaques (Riede et al. 1997) and cats (Riede, Stolle-Malorny 1998). For each individual a number of clearly identified calls were extracted from the choruses. We obtained a sample of 291 calls from 5 animals. The spectrogram of each call was then categorized according to the occurrence of nonlinear phenomena, and the relative duration of the nonlinear phenomena to the total call duration was calculated. For that purpose the duration of all 291 calls was measured and summed for each individual. Then the duration of all nonlinear phenomena was measured, summed up for each individual and divided by total call duration.

To ensure that the selected spectrograms were free of artifacts such as aliasing, clipping or reverberation, the samples were subjected to perceptual review by a second investigator (D.M.).

Anatomical investigation

The larynges (fix in formaline, 7%) were dissected in the dorso-ventral midline and macroscopically inspected. For microscopic investigation the vocal folds were excised in toto and embedded in paraffin. 6 µm- sections were done at three levels (ventral, middle part, dorsal) and colored with Haemalaun-Erythrosine for a global inspection.

3.3.2 Results

Acoustical analysis


37

Figure 3.4 shows the spectrogram of a howling chorus (60 s total duration). Five animals are involved in this chorus. For reason of clearity only the calls of 'Schaka' are marked with arrows. The first call of Schaka exhibits a harmonic structure whereas her subsequent calls show a variety of complicated patterns.

1. Nonlinear phenomena in Schaka's vocalization

Figures 3.5 to 3.7 show spectrograms of Schaka's calls displaying a variety of transitions between harmonic vocalization, subharmonics, biphonation and chaos. The calls were extracted from choruses, therefore they are sometimes overlaid by calls of other animals. Relevant points in the spectrogram are indicated by arrows.

The transitions from harmonics to chaos (e.g. after arrow 3 in Fig. 3.5) or chaos to biphonation (e.g. after arrow 1 in Fig. 3.6) are abrupt. Chaotic elements always transition to periodic, i.e. harmonic (Fig. 3.7) or biphonic (Fig. 3.5 and 3.6) elements. Chaos can occur at the beginning of a call (Fig. 3.6 and 3.7) but calls never ended with chaos.


38

Table 3.4: Total number and percentage of calls containing nonlinear phenomena and relative duration of nonlinear phenomena in five animals. The calls come from different chorusus.

animal‘s name

age (years)

sex

Number of calls

(number of choruses)

calls containing nonlinear phenomena

calls containing nonlinear phenomena (%)

relative duration of nonlinear phenomena (%)

Schaka

7.3

f

72 (10)

23

32

17.9

Chinuk

4.2

m

21 (4)

5

24

4.9

Weisspitz

6.2

m

107 (9)

11

10

2.7

Graue

6.2

f

29 (5)

1

3

1.4

Koja

6.2

f

62 (5)

0

0

0


39

Figure 3.4: Time series and the spectrogram of a chorus. The upper graphs represent the first 20s, the middle ones the next 20s and below, the final 20s of the chorus are shown. The calls uttered by ’Schaka‘ are marked over the total call duration by arrows.


40

Figure 3.5: Time series and spectrograms of a 10 s - chorus cut-out with calls uttered by ’Schaka‘: The first call (between arrow 1 and 2) was harmonical (f0 at the beginning of the call 390 Hz, at the end 410 Hz, maximum 470 Hz, call duration 3.7 s). This call was overlaid by the call of another animal with very similar fundamental frequency. Schaka‘s second call started harmonically with increasing fundamental frequency (arrow 3). There was an abrupt change to a nonperiodic element with chaos at the beginning which continues to biphonic structures. The biphonation ended suddenly (arrow 4) and passed on to a harmonic element with decreasing fundamental frequency from 430 Hz to 380 Hz (arrow 5). This call element was overlaid by the howling call of another animal with increasing fundamental frequency. Note, that the time series displays a high amplitude during the chaotic and biphonic episode.


41

Figure 3.6: The call starts (arrow 1) with a chaotic element which passed on to a harmonic part with side bands (65 Hz distance). This part shows an abrupt transition (arrow 2) to a harmonic part with a fundamental frequency of 430 Hz, which decreases to 390 Hz (arrow 3). The total call duration is 6.4 s.


42

Figure 3.7: Two calls of ’Schaka‘ from the middle of a howling ceremony are shown. Both calls are overlaid by calls of other animals. Both calls start with a chaotic element (arrows 1 and 4 respectively). It follows a harmonic window and a second and much longer chaotic element (ending at arrows 2 and 5, respectively). Both calls end with a harmonic element with a fundamental frequency decreasing from 410 Hz to 390 Hz and 430 Hz to 380 Hz.

The percentage of calls containing nonlinear phenomena is presented in Table 3.4. Nonlinear phenomena can be found in the calls of four out of five animals. Schaka shows the highest amount with 32%, the other animals range between 0 and 24%.

Table 3.4 also presents the relative duration of the nonlinear phenomena. Schaka again showed the highest value of 18%, the other animals range between 0 and 5%.


43

2. A human is able to produce similar acoustic features

The mammalian larynx is quite similar in gross anatomy between species (Negus 1949, Schneider 1964, Harrison 1995) suggesting similarities in the physiology, for instance in the vibration pattern of the vocal folds. For that reason we compared Schaka's vocalization with that of a 24 year old woman who was able to produce biphonation intentionally and whose production mechanism is well understood (see Mergell, Herzel 1997 for details regarding high speed glottography and modeling).

Figure 3.8: Time series and spectrogram of two vocalizations. The left one is from a 24 years old woman who was asked to imitate Schaka‘s vocalization which is on the right side. The woman is able to produce biphonation intentionally.

This woman is able to phonate simultaneously at two different fundamental frequencies (i.e. biphonation) during forceful expiration. The woman was asked to imitate Schaka's howl. From a subjective viewpoint, she was able to simulate Schaka's vocalization very closely. Figure 3.8 shows spectrograms of the woman's voice (left) and Schaka's original howling call


44

(right).

Spectral analysis of the woman's voice signal showed that during biphonic vocalization the fundamental frequency is of the same order as the first formant. When the fundamental frequency falls towards considerably lower values or reaches sufficiently higher values than the first formant frequency, the biphonic spectral pattern disappears.

3. 'Source tract interaction' in Schaka's vocalization - comparison with the woman's voice

Now we discuss in detail a representative call of Schaka with biphonation. In the power-spectrum in Fig. 3.9 the peaks of two fundamental frequencies (termed f0 and g0) are visible at f0= 515 Hz and g0= 875 Hz. The major ratio of the two fundamental frequencies (f0/g0 ) in the call from figure 3.9 is about 0.59 which is significantly different from 1/2 or 2/3 as expected for subharmonic vocalization. Consequently we term this occurrence of two independent frequencies biphonation.

The major peaks in the power-spectrum are harmonics or linear combinations of these two fundamental frequencies. For example (Fig. 3.9), the spectrum demonstrates that all significant peaks can be expressed as linear combinations of two frequencies f0 and g0. As explained by Mergell, Herzel (1997), Reuter, Herzel (1999) for the human voice, and by Nowicki, Capranica (1986) and Fletcher(1992) for bird vocalization, the appearance of linear combinations indicates a nonlinear interaction of two frequencies. In humans direct observation of the vibrating vocal folds revealed that these biphonations represent a glottal or source-generated mechanism caused by asynchronous vibration pattern of the left and right vocal fold (Kiritani et al. 1991, 1993; Mergell, Herzel 1997).

As discussed in Mergell and Herzel (1997), slight asymmetry in the laryngeal frame-work is not sufficient to generate biphonation. Only together with source-tract interactions was biphonation sustained. This interaction is enhanced if the fundamental frequency coincides with the first formant. Moreover, high intensities due to large subglottal pressure support biphonation.


45

Figure 3.9: Time series, spectrogram and spectrum of the biphonic call from Figure 2. The spectrum represents only a short term segment (50 ms) around 0.9 s in the spectrogram. Indications of vocal tract resonances are found with LPC analysis around 550 Hz and 1800 Hz.

In Schaka's case we observed that during biphonic vocalization the fundamental frequency seems to be near to the first formant. The overlaid LPC curve showed a first formant peak at


46

about 550 Hz. From post mortem measurements we know that Schaka's vocal tract length (from glottis to lips) is 16 cm, predicting a first formant frequency of 550 Hz, according to the equation for formant frequency calculation from the vocal tract length (VTL) and speed of sound in warm moist air (c=350 m/s): F(1)= c/4*(VTL) (for humans: Titze 1994; for animals: Fitch 1997; Owren, Bernacki 1998). Thus the fundamental frequency of the source signal prior to the biphonation comes into resonance with the first resonance frequency of the vocal tract, enhancing the interaction between the glottal source and vocal tract. The biphonic pattern does not appear if the fundamental frequency falls towards lower values. For example, Schaka's first call in Figure 3.4 reaches a maximum fundamental frequency of 470 Hz which is below the first formant frequency of about 550 Hz. This call is purely harmonic.

The access to subglottal pressure information is possible indirectly via examination of the time series and the amplitude envelope estimation. Subglottal pressure is the air pressure caudal from the glottis maintained by contraction of the intercostal muscles and the abdominal muscle. The time series in Figure 3.8 shows an increase in amplitude (indicating an increase of the subglottal pressure) during the woman's biphonic phonation. In Schaka's biphonic calls, the amplitude is about one third higher than during harmonic phonation (Fig. 3.5).

Anatomical investigation

Macroscopic investigation of Schaka's larynx showed no obvious peculiarities of the vocal folds. The lengths of the left and right vocal folds were 16 mm and 18 mm, respectively. The Processus cuneiformes on the left side was larger than on the right side. The other wolve's larynges showed no macroscopic peculiarities. The vocal folds of the male were 18 mm on the left and 17 mm long on the right, and those of the female 16 mm on the left and right side.

The main result of the microscopic inspection was an upward extension of both vocal folds of 'Schaka' but not of those of the two other investigated individuals (Fig. 3.10). These structures were about 400 µm high, 200 µm broad and 8 mm long, situated on the edge of the vocal folds. This feature reminiscent of 'vocal lips' in some primates (Brown, Cannito 1995; Mergell et al. 1999) and 'vocal membranes' in bats (Grosser 1900; Elias 1908; Pye 1967). It also resembles oedema and Sulcus vocalis in pathological human voices.


47

Figure 3.10: Horizontal sections of the middle part of the vocal fold of Schaka (with vocal lip) and of a normal sounding animal (without vocal lip). Compare to the section of the larynx in figure 2.1.


48

3.3.3 Discussion

A female dog-wolf mix displayed an unusually high amount of nonlinear phenomena in its howling vocalization. Four other animals also showed nonlinear phenomena in their calls, but at a much lower rate. Further, they never produced biphonation, but only subharmonics and deterministic chaos.

The generation mechanism of nonlinear phenomena in Schaka's howling

The anatomical investigation of 'Schaka's' larynx showed a slight asymmetry in the Processus cuneiformes (a part of the arytenoid cartilage). Hirano et al. (1989) showed in humans that the laryngeal framework is typically asymmetric, and that the degree of asymmetry did not differ among age groups or between sexes. They concluded that there must be some mechanism compensating for the asymmetric framework to keep the vocal fold edges relatively symmetric. Assuming, that a similar compensatory mechanism exists in wolves, this slight asymmetry in Schaka's larynx is probably not solely responsible for the high amount of nonlinear phenomena in her voice.

Investigations of human voices have shown that asymmetries in vocal fold anatomy, as well as asymmetries in the vibration pattern, can be the basis for nonlinear phenomena (e.g. Ptok et al. 1993; Moore et al. 1987). These findings received further support from computer models of the vibrating vocal folds (Steinecke, Herzel 1995; Tigges et al. 1997; Mergell, Herzel 1997). These models showed that slight asymmetries are able to produce nonlinear phenomena (Herzel, Reuter 1997; Mergell, Herzel 1997) if parameters such as the subglottal pressure or vocal tract shape are appropriately adjusted. The coincidence of vocal tract resonance frequencies and fundamental frequency of the vocal fold vibration support interactions and reciprocal reinforcement.

Studies with anaesthetized and centrally electrical stimulated dogs (Solomon et al. 1994) showed that macroscopically normal looking vocal folds produced normal sounding dog vocalizations, but a subject with nodules produced rough sounding utterances (not analyzed in more detail).

In Schaka, we found a slight asymmetry of the laryngeal framework. Second, in biphonic calls we observed a coincidence of the fundamental frequency of the source with the first formant frequency of the vocal tract, suggesting the possibility of source-tract interaction. Third, the relative amplitude in biphonic call parts was about one third higher than in harmonic call parts. All three points together may be responsible for biphonic, chaotic and subharmonic


49

regimes.

The microscopic investigation delivered a further aspect to be considered - an upward extensions of the mucous on the edges of the vocal folds. In two other animals we did not find such structures. These structures are well known as 'vocal lips' (syn. vocal membranes) in bats and some primate species (Harrison 1995). In canids there is no consensus about the frequency of vocal lip occurrence. Harrison (1995) did not describe vocal lips in the wolf. Jiang et al. (1994) described no vocal lip like structures in the vocal folds of 13 mongrel dogs. Two other references showed histologic pictures of dog vocal folds which obviously had vocal lip like structures but the authors did not explicitly mention them in the text (Duckworth 1912; Negus 1929).

There are two main suggestions about the function of vocal lips in voice production. First, they might be responsible for the production of very high fundamental frequencies (Schön-Ybarra 1995) as seen in some small primate species. Second, a computer simulation showed that vocal lips can lower the subglottal pressure at which phonation is supported, thus increasing vocal efficiency (Mergell et al., 1999). Moreover, vocal lips induce vocal instabilities in the model, suggesting that the 'upward extension' in Schaka's vocal folds may support the production of nonlinear phenomena. Also Brown, Cannito (1995) discussed vocal lips as being responsible for biphonation (they called it 'polyphonic vocalization'), hypothesizing that vocal lips represent a separately vibrating structure inducing a second fundamental frequency.

Communicative relevance of nonlinear phenomena

From video recordings we know, that Schaka's conspicuous voice was present for at least the last 3 years. In the social rank order Schaka was in the alpha-position during the last 4 years and she reproduced 3 times during that period. Whether or not there is a relationship/coincidence between the occurrence of a high amount of nonlinear phenomena and any event in Schaka's life (for instance a disease) is unknown. The increased amount of nonlinear phenomena can be explained in two ways. On the one hand, Schaka might have produced the high amount of nonlinear phenomena involuntarily without communicative effect due to her special laryngeal anatomy (vocal lips, slight asymmetry), just displaying an idiosyncratic voice disorder. On the other hand, Schaka might have produced the high amount of nonlinear phenomena voluntarily, aided by her special laryngeal anatomy, and reinforced by auditory feedback, with unknown communicative effect.

Generally, the problem whether or not nonlinear phenomena are pathological or integral part


50

of the repertoire remains an open question. There are examples pointing in both directions. Riede et al. (1997) showed in a Japanese macaque (Macaca fuscata) infant with an assumed metabolic disease an increased amount of nonlinear phenomena in the vocal repertoire, but unfortunately sufficient behavioural observations were lacking. East and Hofer (1991) showed spectrograms of spotted hyenas (Crocuta crocuta) whoops which indicated that nonlinear phenomena seem to be rather common in this type of vocalization. In the painted hunting dog (Lycaon pictus) Wilden (1997) showed nonlinear phenomena in the vocalization of all age-classes. Wilden et al. (1998) screened the literature on mammalian acoustic signals and found nonlinear phenomena in many published spectrograms. The widespread occurrence of such phenomena suggests that they are integral parts of the vocal repertoire in many mammals.

In humans, rough sounding voices are primarily studied in connection with voice pathologies. There are, however, examples of nonlinear phenomena in normal voices which are well-known from modern pop singers and infant cries (Mende et al. 1990). In these cases, rough sounding voices elicit strong attention from listeners. Hecker, Kreul (1970) showed that slightly rough sounding voices (i.e. voices with a higher amount of nonlinear phenomena) do not indicate an unhealthy voice to human listeners suggesting that nonlinear phenomena are common to a certain degree in the normal human voice. Hecker and Kreul (1970) found also that a higher degree of roughness affects the listener's ability to guess the speaker's age: listeners evaluated the speaker as older than he/she was.

Together, these observations suggest that nonlinear phenomena in the otherwise primarily harmonic vocalization exhibit communicative relevance in humans. Further studies, including behavioural observations, are necessary to substantiate the communicative role of nonlinear phenomena in nonhuman mammal vocalization.


[Front page] [Acknowledgements] [Attached Audiofiles] [1] [2] [3] [4] [5] [6] [7] [Bibliography] [Declaration]

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.

DiML DTD Version 2.0
Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML - Version erstellt am:
Thu Sep 28 14:01:51 2000