Isotonic lines and audition threshold 

Context in source publication

Context 1

... the vowels do not have a fixed pitch: they occupy a frequency band of 150 Hz in average. The stressed vowels are very often whistled higher in pitch within each band. As these bands sometimes overlap, they are used by the whistler to adapt to the phonetic contexts of the words: when there is a potential ambiguity, they distinguish clearly neighboring vowels by making the effort to place the vowels at opposed extremes of their own band (cf. examples below). Therefore, the overlapping, in whistles, of two bands of frequencies corresponding to two different vowels doesn’t mean that they are always interpreted in the same phonologic way. The five vowels in the modern Greek (i, y  , a, o) are whistled in the village of Antia in five bands of frequencies among which four might overlap with the band of another vowel. This creates statistically three major bands of frequencies: [i, ( y ,  ), (a, o)]. The accentuation in modern Greek has an intermediary degree of liberty, we have noticed that the whistlers reproduce the spoken accentuation nearly systematically (80%), with variations between the whistlers. The 8 vowels of Turkish (i, y, e,  a, o, u) are whistled in 8 bands of frequencies decreasing in pitch that might be grouped in three reductive bands as follows: [(i, y,  ), (e,  ), (a, o, u)]. Our results differ from the ones published by Leroy [5], only in the case of /u/, but we have the same results as Moles [11]. The influence of vowel harmony rules of Turkish help the whistlers to get rid of some ambiguities. However the ones which are not solved by the harmony system are sometimes overcome by the use of extremes of the bands of frequencies. For example for the common word “kolay” (/kolaj/), /o/ and /a/ are effectively distinguished because /a/ bears a higher pitch despite the fact that these two vowels are usually whistled in the same way. Silbo vocalic system is based on the Spanish spoken dialect of La Gomera island for which /o/ and /a/ are very near. The spoken vowels (i, e, a, o, u) are therefore whistled in five bands which can be grouped in three groups [i, e, (a, o, u)]. Similarly to Greek and Turkish, the three bands of frequencies of Silbo are around 2600 Hz for [i], 2100 Hz for [e] and 1600 Hz for [(a, o, u)]. Some of the best whistlers disagree with this classification. Their whistled description enables them to distinguish the five vowels. They can even differentiate /u/ from /o/ because its spectral envelope carries much more energy. This difference of point of view might be due to the survival of different whistled dialects in the island [3] and is always relative to the level of practice. The accentuation in Silbo uses the set of possibilities of each band of frequencies: for example, in the word “abajo”, the second /a/ is whistled higher than the first and the /o/ is whistled lower than the two former vowels. The main categories of whistled consonants described in Silbo and Turkish are still available in Greek. The occlusive stop consonants can be grouped in three types of loci movements: [P, T, K] equivalent to those of the formant 2 of spoken speech. /s/ and /dz/ behave as /t/ but with higher loci. The voicing is reproduced by a slight continuation in the amplitude. The shape of liquid and fricative continuants are displayed on Fig.2. together with [m]. The nasal [n] is usually whistled like [l] with sometimes a cut at the edge of the shape. From our study of Greek, we also concluded that from a cluster of consonants, only the most projecting consonant will remain in whistles. The whistling techniques vary in function of the distance of communication : intromission of a finger in the mouth (long distance), retro flexion of the tongue (middle distance) or bilabial (short distance). Whistling does not require the vibration of the vocal cords: it is produced by a shock effect of the compressed air stream inside the cavity of the mouth. When the jaws are fixed by the finger and/or the tighten lips (point 1 in Fig. 3) , the size of the hole is stable. The air stream expelled makes turbulences at the edge of the mouth. The faster the air stream is expelled, the higher is the noise inside the cavities. If the hole (mouth) and the cavity (intra oral volume) are well matched the resonance is tuned and the whistle is projected more loudly. The frequency of this bio-acoustical phenomenon is modulated by the variation of the volume of the resonating cavity which can be, to a certain extent, related to the articulation of the equivalent spoken form. The movements of the tongue and of the epiglottis play the main roles for tuning the vowels and consonants (Fig. 3). The whistled signal is shaped by a kind of whistled aperture. Such a behavior is aimed at reaching an optimal listener’s intelligibility. The pitches of the main band of frequencies of whistles are concentrated in a narrow bandwidth (1 kHz to 3 kHz) where the audition of human beings is more sensitive and selective as shown on Fig 4. The amplitude of whistled speech has reasonable limits in its dynamic range (less than 20 dB) whereas the range of spoken speech is more than 50dB. Long distance whistled speeches are higher in frequencies (approximately 100 to 250 Hz) than short distance ones. This aspect underlines that the range of frequency is relative to the distance of communication. Whistles are well carried in valleys which form a natural guide (the signal remains understandable at 8 km in La Gomera) and they exactly cover the central domain of frequencies for which the sounds resist to reverberation in forests [10]. Moreover, in natural conditions the background noise is weak in high frequencies (except in windy weather) so the signal to noise ratio is better than 6 dB at 1 km and is enough to be clearly heard. All these remarks show that whistled speech is particularly well adapted to human communication in noisy natural environments, and in long distance speaking in mountains or in forests. The comparative results obtained from several whistlers of various non tonal languages show how the properties of whistles are exploited to encode linguistic information. As a whistled signal is limited to three features: pitch, loudness and duration; the frequency and amplitude modulations are used in a complementary way. In this context, the analysis of the acoustic cues exploited by whistlers for vowels and consonants underline that the intelligibility of these sounds is relative to the lexical environment and to the structure of the concerned language. Such a conclusion is supported by the performances to psychoacoustic tests realized by Moles [11] in Turkey which showed that the intelligibility of words was eased when they contain the most frequent segmental features and that much more confusions where made for words extracted from their lexical context. These results combined with our analysis meet the interrogations of some studies on spoken languages which look for the phonetic basis of phonological representations [12] by analyzing the degree of reduction in different contextual conditions [13]. Our results taking into account the articulation, the perception, the acoustics and the use of the vowels in different cultural and lexical context suggest a good match of the whistled main band of frequency and of the perceptive formant (called F2') defined by Carlson, Fant and Granström [14]. Their approach of the mechanisms of data reduction in vowel perception, together with other psychoacoustic studies tackling this subject, provide similar results to the natural evolution displayed in the phonetic description of articulated whistles. Whistles extend this approach to consonants. Whistled languages are the result of the adaptation of the human perceptive and productive intelligence to a natural acoustic area (topography, vegetation, noise) and to a linguistic environment. They represent a strong model to investigate the perception of languages. I would like to thank the whistlers and the cultural leaders who took some time to work with me on the field, Laure Dentel who was volunteer to record during my fieldwork, Prof. René-Guy Busnel for his strong scientific support during the last 3 years, Bernard Gautheron for his advice, Ross Caughley for lending me some material in Chepang, the staff of the DDL Laboratory for providing technical material. This research was half financed by a BDI Ph.D grant of the CNRS and half by personal financial ...