Some thoughts on Jochem Sotschek’s report: On the possibilities of the identification of phonemes ANABELA CARDOSO It is a real pleasure to be able to offer to the readers of the IT C Journal the important paper by Jochem Sotschek, the renowned acoustics expert in telecommunications at the German Telekom, Research and Technology Centre in Berlin, Germany that we publish in this issue. Several experiments with Friedrich Jürgenson (FJ) devised and supervised by Professor Hans Bender (see ITC Journal Nº 40) yielded unexplainable utterances finally considered anomalo us by Bender. To check the acoustic characteristics of the latter and h elp determine their anomality or otherwise, Professor Bender cal led upon one of the most proficient Telecommunications technicia ns of that time in Germany, i.e. Jochem Sotschek. The present paper documents the work done with the acoustic samples containing the utterances recorded with Jürgenson, therein called ‘recordings’. The study of Instrumental Transcommunication results is a serious enterprise. The evidence provided by ITC communicatio ns can, for the first time in the history of the research into survi val of physical death, demonstrate that consciousness appears to surv ive physical death. The objective nature and permanence in physical form (audio and visual recordings) of the evidence enables its scrutiny by the use of technical and scientific tools. When Professor Bender worked with FJ the most updated methods available for speech recognition at that time were u sed by J. Sotschek to examine the ‘recordings’ (i.e., the anomalou s electronic voices). CUADERNOS DE TCI 11

Nowadays, other methods, in the main digital, are also availa ble for the same purpose. However, they are based exactly upon the same principles described by Sotschek in the work that we publi sh today. Having said this, things may seem easy to interested ITC partie s. However, I have a different opinion on this highly important t ool for ITC researchers and operators. I believe that proper evalua tion of allegedly anomalous recordings through such sophisticated ana lysis methods is not at the reach of a layperson. They are very useful and in most cases better than human listening panels as the paper reveals, but they demand expertise and state of the art softwa re to be reliable. I don’t think that the basic audio editing software , generally used by ITC operators, is of help for proper identification of t he meaning of anomalous recordings. In the hands of technically minded people it can serve as guidance but, even so, it should n ot be taken as definitive proof. Dedicated software for speech recogn ition is expensive and demands skilled users. It also demands sophisticated equipment such as professional, high quality soundboards, computers, amplifiers and loudspeakers. Thes e are the means used by the scientific police which cannot be adequate ly used by a normal person. They require years of study and practice and real expertise to yield valid results and serve as proof as happens with the police. We should keep in mind that speech recognition performed by the police can send to jail a presumed criminal. It is only natural that such work cannot be done by a lay person with Adobe Audition, just to give an example. We, ITC operators, should not forget this basic principle if we want to be respecte d something that, unfortunately, decreases daily because of our superficial behaviour, most likely well-motivated but unint entionally misleading. The fact that the guidelines above are generally not taken int o consideration is at the bottom of the highly objectionable pos ting in Internet of thousands of ‘voices’ and their contents, which, in the great majority of the cases, do not correspond to what the author , certainly convinced of its authenticity, publishes. This atti tude gives an extremely bad reputation to the discipline and jeopardizes the ITC JOURNAL12

admirable work of our outstanding pioneers – Jürgenson, Rau dive, Franz Seidl, among others, not to mention the most modern one s and certainly, also of our transcendental communicators. Publication of interpretations based upon listening tests can also be very dangerous for the following main reasons: 1. Listening is a subjective process to some degree (see Müller , 2011). 2. Correct listening requires listeners with good hearing s kills, as FJ already pointed out, and first-class playback equipment. 3. Anomalous electronic voices are normally mixed with noise. Even white noise disturbances can be dangerous, as Sotschek po ints out, but the so-called EVP-maker and other human-speech based acoustic supports are disastrous. As a distinguished Profes sor of Acoustics recently remarked about this ‘method’: “This is the most certain way to create pareidolia in the listener; it should not be used”. I could not agree more. 4. Listeners should be native-speakers of, or highly proficien t in, the language the anomalous voices are supposed to be in. Reading the full translation of Sotschek’s paper for the fir st time today, May 1, I was happy to verify that he recommends several steps that I had decided upon, based on my personal experience, years ago, namely that the listening should, in many cases, be done sou nd by sound and that the experimenter can pronounce and record the same words that he/she thinks comprise the anomalous recordin g and compare the graphic thus obtained with the supposed anomalous recording. For this rough inspection common sound editing software such as Soundforge, Adobe Audition, Wavep ad and so on are adequate because the listener only needs to compare the graphics of his/her own voice and of the anomalous recording o n the computer screen. In general, the latter will be corrupted wit h noise but it will still be a good way of roughly and easily checking the shapes of both sound waves on the computer screen. I had this purpose in mind when I decided to repeat the anomalous sentences I published in the CD Electronic Voices that complements my book with the same title. I wanted to allow the CUADERNOS DE TCI 13

listeners to digitally compare both graphics, a rather simple procedure. Up until now, I did not receive any negative remark about my interpretation of the content of the voices. On the othe r hand, the Portuguese speaking listeners said that they could v irtually understand every anomalous sentence and the majority of the nat ive speakers of other languages corroborated this evaluation al though, as it should be expected, they also stated that they could not understand the meaning. My purpose with these lines is, once more, to call the attention o f ITC operators and researchers to the need to be rigorous, preci se and objective in the evaluation of ITC results, either our ow n or other published ones, and to determinedly discard what does no t fulfil these requisites. Our marvellous communicators, who s pare no efforts to speak with us, will be thankful for this attitude a nd we will also render a great service to the study of this discipline that co ntains in it the power to change the prevailing human paradigm (see L aszlo 2008). But this most important goal can only be achieved if we are backed by the scientific establishment, something which has no t happened. However, it did happen in Jürgenson’s and Dr. Raudive’s time and the Austrian scientist, Franz Seidl, remarked in 1970: “ It is rare indeed for a special area of parapsychological study – as the phenomenon of the voices is – to awaken the interest of serious men of science and researchers to the extent that the source of thes e voices, which emanate from beings outside the Earth, is stud ied with the most modern of investigatory methods in universities and research laboratories, with the result that the reality of th eir existence is established.” (Appendix to Breakthrough 1971) I think we should rightly ask what has happened since then to “the interest of serious men of science and researchers” becaus e it is no longer visible. Many factors contributed to the present situ ation but one of the most important ones, if not the most important o ne, is the lack of rigour, critical sense and self-criticism of many ITC operators. Our communicators and our outstanding pioneers d o not deserve their work to be ignored the way it presently is. Let us endeavour to change the current scientific approach to ITC. The ITC JOURNAL14

issue at stake is too serious and vital for humanity and for pla net Earth to be treated this way. We absolutely need to understand t his and the decision is in our hands. References Bender, Hans. (2011). On the Analysis of Exceptional Voice Phenomena on Tapes . Pilot studies on the "recordings" of Friedrich Jürgenson. ITC Journal N o.40, pp. 61-78. Cardoso, A. (2010). Electronic Voices: Contact with Another Dimension? Ropley, Hants, UK: O Books - John Hunt Publishing, Ltd. Cardoso, A. (2010). CD: Electronic Voices (available from www.itcjournal.org). Laszlo, E. (2008). Quantum Shift in the Global Brain. Vermont, USA: Inner Traditions. Laszlo, E. (2008). An Unexplored Domain of Nonlocality: Toward a Scientific Explanation of Instrumental Transcommunicatio n. Explore, September/October, Vol. 4, Nº 5, 321-327. Müller, E. (2011). The Subjectivity of the Listening Process. ITC Journal 41, pp. 33-45. Raudive, K. (1971). Breakthrough: An Amazing Experiment in Electronic Communication with the Dead. Gerrards Cross, England: Colin Smythe. CUADERNOS DE TCI 15

Editorial Note On the possibilities of the identification of phonemes** JOCHEM SOTSCHEK The utilisation of visible-speech-techniques and other methods for the analysis of tape-recording ‘ playbacks’ A summary is given on the methods of determining speech components in the technique of communication. By means of these methods, analogous conclusions can be drawn for the analysis of ‘ play- back-phenomena’. Techniques and possibilities of visible-spe ech- techniques are explained by various examples. Auditory experiments in audiometry and telecommunications The usual and easiest way to recognise speech information is based on auditory experiments. The speech information to be te sted is presented for identification to one or several subjects, f or instance ITC JOURNAL16 Readers will recall that we published in issue 40 (pp 61-78) Professor Hans Bender’s account of experiments with Friedrich Jürgenson and the methods used to assess the paranormality of recordings where likely anomalous utterances were detected. Those were labelled ‘recordings’ («Einspielungen» in Germa n) and therefore, the same word used in the present report by Professor Bender’s invited telecommunications expert, Joche m Sotschek, should be interpreted in that sense i.e., the record ing of anomalous electronic voices. ITC - Historical Perspective

over headphones. This method is generally applicable. It is us ed, for instance, in audiometry for measuring and evaluating the hum an listening behaviour or in telecommunications to evaluate t he transmission quality during voice transmissions over unkno wn channels. In audiometry, groups of uniformly shaped words (multisyllabi c numbers or monosyllabic substantives) chosen from a greater repertoire are played to the subject to recognise the word samp les (Freiburg word test, DIN 45 621). For measuring the speech quality in telecommunications one can present language elements (sentences, words, syllables) to a testing group where the audibility is a measure of the transmission qu ality. Mostly in these auditory experiments sense-empty syllables, so called logotoms, are used, which auditory impressions have t o be noted by the subjects. They are performed from a fixed repertoir e of initial sounds, vowels and ablauts by randomly composing a gr eat quantity of different syllables. The syllable components are so und elements of at least one certain language; sound lists consis ting of sounds of all spoken languages are also used. However, in eve ry case the number of different sounds is limited: a list publis hed by the International Association of Telecommunications Administ rations contains 45 different initial sounds, 5 different vowels an d 49 different ablauts. The percentage of correctly understood syl lables in relation to the total number of transmitted syllables is calle d the syllable intelligibility of the transmission channel concerned . Auditory experiments for recognising ‘ recordings’ Conditions similar to those in the previously described metho d for recognising sense-empty syllables are present during recogn ition attempts when listening to ‘recordings’. But in contrast to th e aforementioned examples here the possible repertoire of vo ice elements is not taken only from one or several languages, which are usually known, but rather are completely unlimited. This appl ies not only to the number of different sounds but also to the rules aft er which the individual sounds are later put together into syllab les and words in the described experiments. In the Freiburg word test the listener only had to compare the sound heard with a rather limited vocabulary, and this just ta ken CUADERNOS DE TCI 17

from his mother tongue. He soon realised that the test words w ere composed of only two – or multisyllabic – numbers or represented a collection of monosyllabic substantives. Although the memb ers of testing groups in intelligibility measurements do not know a ll these different logotoms individually as syllables they can split the sound heard into sound parts and classify it into a repertoire of init ial sounds, vowels and ablauts they have learned over time. Also i n this case there is a strictly limited pool of possible language ele ments. In the case of trying to recognise ‘recordings’ these facilities o f a limited repertoire do not exist for the listener. He would ma ke it too easy for himself if he would assign to that which he heard only wor ds or parts of words (i.e., the sounds) from his own language or vocabulary. But the possibility of extending this to other lan guages that the subject also knows does not provide any major improvement. Proof of this is shown when a number of ‘recordin gs’ were differently evaluated by German subjects having good Eng lish language skills and American subjects with excellent German language skills. During listening attempts of ‘recordings’ the subjects shoul d care less about the ‘supposed’ semantic content of the speech sampl es but rather document the phonetic content. Because the two criteria cannot always be cleanly separated – a plausible sense is too easily interpreted into the heard soun d – the sound content should be analysed at least during the processing of the listening results sound by sound. Only after a statistical a nalysis of logged answers of an enormous number of subjects one can conclude from the accumulation of sound groups a possible wor d context. Analysing in accordance with these considerations the protocols of only a small number of subjects – for instance the results of 10 or 20 subjects – the sound analysis nevertheless will give only ran dom results, which are statistically not significant. Naturally, the question immediately arises of how many subjects are necessary to achieve statistically secured results for recognition of the sound cont ent of ‘recordings’. For this purpose experimental results from telecommunications can provide valuable guidelines. To prov ide an estimation of the required number of independent hearing te sts of ITC JOURNAL18

the same material, well-known connections of intelligibilit y measurement may be used. There are results of syllable recognition measurements on a wi de range of transmission systems. In this connection the results captured on transmission systems evoking a similar speech impression as in the ‘recordings’ are of special interest. T he most important parameters are the transmitted frequency band and the signal to noise ratio which results from the ratio of the signal value (speech – i.e. ‘recording’) to the value of all disturbances (i. e. noise and hidden speech). Due regard must be given to the spectral distribution (‘colouration’) of the disturbance signal. Al l these characteristics can be measured quite accurately. From the Vis ible Speech Spectrogram of the 12 ‘recordings’ of F. Jürgenson, in Mölnbo, Sweden, it can be seen that the frequency range of these ‘recordings’ extends between 200 Hz and 6...7 kHz, the range of perceiving ‘normal’ speech. The disturbance signals simulta neously occurring with the ‘recordings’ were in the same range. These spectrograms perform a three-dimensional visual presentat ion of a speech or sound event. Information about the signal noise ra tio of the sound event can be obtained by a level scriber. Here, the sig nal will be recorded over time without any frequency evaluation. Figure 1 CUADERNOS DE TCI 19

Based on the twelve aforementioned ‘recordings’ the signal t o noise ratio of the voltage levels between ‘recording’ and dis turbance signal was determined to be in an area between +10 and -10 dB. (Because of the large range between the lowest and the highes t possible value, amplitude specifications of sound events in telecommunications are made in the measurement decibel [dB] , the twenty times logarithm of a voltage or sound pressure ratio). Figure 1 shows such a level scriber writing. It belongs to the o utlined Visible Speech Analysis of the ‘recording’: ‘TAN(NER) TAN(NER)’ in figure 3. In the level scriber recording of these speech samples only the most powerful sound ‘A’ is shown. In th e above example the signal to noise ratio compared with the most powerful vowels is around 0 dB. From intelligibility measure ments, i.e. by Fletcher and Galt 1, the context between syllable understanding and signal to noise ratio is well known. These results apply to sense-empty syllables as speech material and white noise as disturbance signal (a noise signal is called ‘white noise’ i f it has constant power in equal frequency intervals). Desired and int erfering signals have been in bandwidths limited to 125 ...5700 Hz, wh ich approximately corresponded in the frequency range to the conditions during the ‘recordings’. However, the nature of these ‘artificial’ disturbance signals only approximately correspon ds to the actual combined speech and noise interferences. For this stil l quite favourable signal to noise ratio of +10 dB, which occurred in onl y two cases during the ‘recordings’, Fletcher and Galt reporte d a syllable intelligibility of 55%; in the case of a signal to nois e ratio of -10 dB a syllable intelligibility of 5%. If one assumes that the ‘recordings’ could be reasonable speech, i.e. single words, o ne can specify the result according to other authors 2 in word intelligibility. In this case the word intelligibility is about 90% (for +10 dB si gnal to noise ratio) and about 10% (for -10 dB signal to noise ratio). What does that mean for the practice of auditory experiments concerning ‘recordings’? Considering first the ‘recording s’ as sense- empty phoneme compositions, for the specified signal to voice r atio there are only 5...55% correct answers as an average of a large number of individual decisions. In logotom measurements the evaluation of 1000...2000 logotoms per transmission system is sufficient in this case. If, however, a limited stock of real w ords of a language is assigned to the ‘recordings’, the hit rates in aud itory ITC JOURNAL20

experiments under the same circumstances are between 10 ... 9 0%. Because it cannot always be presumed that the language is known to the subject, this hit rate is certainly too high. So, transferr ing the conditions and measuring methods of intelligibility tests t o recognition attempts in the above ‘recordings’, one would hav e to perform at least 1000 independent auditory experiments. In t he case of an adverse signal to voice ratio only about 5 ... 10% of the res ults will be correct. The evaluation method should allow recogniz ing this small number of hits. Due to an assumed limited word repertoir e of the subject the evaluations would not be sufficiently detaile d. As already explained, a better result can be expected only by an analys is performed sound by sound. However, in the case of the currently known ‘recordings’, which unfortunately are strongly disturb ed, this method seems less promising. Speech analysis from Visible Speech Representations A completely different method for recognizing speech sounds is offered by the analysis of Visible Speech Spectrograms . The Visible Speech Spectrograph, a modern device in linguistics, is an ana lysis device that allows the visible representation of a sound eve nt and its electric voltage in three coordinates: time, frequency and amp litude. The spectrograph manufactured by Voiceprint Laboratories and used for these studies has a 300 Hz wide fixed filter which glid es slowly as a ‘frequency window’ over the whole frequency band to be examined. It provides within 80 sec an analysis of a 2,3 sec lasti ng sound event stored on a tape. Analysing, for example, a voice signal, the spectrogram gives a spectral representation of the energy distribution over tim e (abscissa axis) of the speech sample. Thereby the frequency of the event is plotted on the ordinate axis (frequency marks at every full kHz) and the different size of every portion is presented by different s trong blackening. The device enables the representation of a dynami c range between 0 dB (no blackening) and about +50 dB (strongest blackening). Now we know that the different voiced speech sounds have energy concentrations in certain frequency ranges, the so-called formants. Every voiced speech sound has several characteristic formant areas with characteristic frequency positions. Thus , with the CUADERNOS DE TCI 21

recognition of formants in certain frequency positions in a spectrogram, which are not identifiable during listening t o the speech samples due to disturbances, it is possible to reconstr uct under certain circumstances the original sound from the physi cal analysis of the speech sample. Further support is offered by another property of voiced speech sounds generating particular speech-specific characteristics in a spectrogram. This special property results from the way in which the human voice formation system produces voiced sounds. Here, initially a constant air stream is pressed from the lungs as an energy source through the epiglottis of the larynx. The vocal folds fo rming the epiglottis are stimulated to perform periodical pulse-sh aped vibrations. Their fundamental frequency determines the pitch of the intonated speech sound. The vocal fold frequency may be modified arbitrarily within certain limits. It varies according to the w ay of speaking within a sentence. However, during normal speech it i s usually located within a given range. For women's and children 's voices this lies between 220 Hz and 330 Hz, for male voices betw een 120 Hz and 160 Hz. Then the periodical pulse-shaped air stream, which is very rich in harmonics, is passed through the three var iable cavities of the vocal tract (nasal, oral and pharyngeal cavitie s). According to the adjusted volume of the three cavities the des ired vowel is formed by their different resonance frequencies. Unvoiced sounds are produced by airflow or shutter sounds inside the vocal tract, on the teeth or lips without the invol vement of the vocal folds. The same mechanism of voice generation applie s to the ‘unvoiced’ soft speech where also no vocal fold oscillation s occur. A third group consists of voiced consonants, in which in additi on to non-periodic sound components occurs a periodic sound port ion due to vocal fold oscillations. These vocal fold oscillations, described above as a characteris tic of specific speech sounds, can also be seen during the analysis wi th the Visible Speech Spectrograph. If they occur in particular frequency ranges and can be assigned to particular formants, they perform an important criterion for differentiation betwee n useful signals and interfering signals, occurring in addition, dur ing voice recordings. These interfering signals, which mostly have a non - ITC JOURNAL22

periodic course (noise), are mapped accordingly, randomly, in t he Visible Speech presentation. However, if the disturbance al so consists of speech, a differentiation of the two speech sample s becomes very difficult. It would be conceivable at best only dur ing a simultaneous occurrence of different vocal fold frequencies ( i.e. different pitches of the two voiced speech samples). In this cas e unvoiced sounds from different speakers are not distinguish able at all. To illustrate this, the Visible Speech Spectrogram, in figure 2, of a clearly undisturbed voice sample is shown. It is the sentence ‘EINSPIELUNG NUMMER EINS’ pronounced by the male speaker B. In the figure the abscissa describes the course of time of the speech sample from 0 to 2.3 sec and the ordinate shows the frequencies from 1 to 7 kHz. Clearly the individual restricted formant regions of the voiced sounds can be recognised, in the darkest black areas. The broadband character of the sound ‘SCH ’ in ‘EINSPIELUNG’ ranges from 2 kHz to 7 kHz and beyond. The sound ‘S’ in ‘EINS’ also extends over a wide frequency range. Th e representation of the voice lip vibrations of vertical scannin g is clearly visible with all tones together with the voiceless ton es [the white areas in] SCH, P and S. The distance between the stripes indicates the period duration of the vocal fold oscillations. It can be seen that this period duration varies within a word or a whole sentence: The pitch of the speech changes with the sentence melod y. CUADERNOS DE TCI 23 Figure 2

The syllable ‘EIN’ is intoned higher than the ending -’IELUNG’ , the sound ‘N’ in ‘EINS’ is spoken very deep (long period duration o f the vocal fold oscillations). With the described analysis method the opportunity is created t o finally detect speech sounds, which are not clearly recognizable by the ear because of disturbances, or by optical decision criteria based on the position of the individual formant regions, consider ing the presence of vocal fold frequencies. The prerequisite is of cour se that the noise and the speech sound are distinguishable in the visu al representation. With regard to the frequency selection as well as to the reception of volume differences, the analysing skills of t he ear are very sensitive in a wide dynamic range and seldom are outdone by technical means. But for the exact technical measurement of sound events, often the ears are only accessories. If useful and interfering signals are exactly in the same frequ ency range and the disturbance has a higher energy than the useful si gnal it is very questionable whether a separation of the two signal s could be made naturally. However, there are also cases where a sound event is made inaudible to the ear by another sound event which h as not the same frequency as the first but a neighbouring one. In th e case of this ear-physiological phenomenon, the so-called occlusio n, the determining of this hidden and inaudible signal by metrol ogical analysis, i.e. as described above, is quite possible. A sound detection of ‘recordings’ using determinations of voiced/unvoiced sounds is of course only viable if the analysis o f the ‘recordings’ performed in the described manner provides representations of vocal fold frequencies. However, this ca n be assumed if the speech character of a ‘recording’ is clear by simpl e listening and most of normal spoken speech sounds have verif iable vocal fold oscillations which are also necessary for their gener ation. As shown in the next section, these premises are obviously applicable to ‘recordings’. Application of Visible Speech Analysis to ‘ recordings’ Figure 3 shows the Visible Speech Analysis of a ‘recording’. Next to the ‘recording’-signal it displays interferences in th e form of noise and speech. However, these interferences [made by the experimenters] are easily distinguished by listening to the r ecorded ITC JOURNAL24

material because, although both appear approximately equal ly loud, the ‘recording’ has a higher pitch than the language interfer ences. The ‘recording’ has the characteristics of a female voice. In the analysis the formant areas of the ‘recording’ are highli ghted by framings. The very quiet sounding ‘T’ is only visible in a smal l range of 3...4 kHz although it contains additional energy port ions particularly in higher frequencies. The formant of the sound ‘ A’ is visible. Mainly in the first syllable the fine structure caused b y the speech fundamental (vocal fold oscillations) is signified. H owever this is clearly visible in the first formant (about 1 kHz) of th e first ‘N’-sound. In comparison with the fundamental frequency representation in figure 2, one recognises a quite high speech fundamental frequency. This fits with the female voice pitch, subjectively perceived, of the ‘recording’. The other forman t structures belong to the speech interferences [by the experime nters]. The analysis of a further ‘recording’ is shown in figure 4. Her e, too, the suggestion of a fundamental speech frequency is visibl e particularly at the first ‘O’. In this Visible Speech Diagram the limitations of this method of sound analysis become evident. The sound ‘F’ (F corresponding t o CUADERNOS DE TCI 25 Figure 3 (3)

the V of the word VON) is clearly visible in the diagram. In accordance with the Visible Speech Analysis the vowels “O” [of VON WO] sound more like "A" than “O” and the sound "W" is only indistinctly visible. Thus, for "reading" the Visible Speech Diagram the phonetic listening impression is necessary as an aid. An example for the case in which the listening impression provides significantly more information than a Visible Spee ch Analysis is shown in figure 5. This ‘recording’ was done at the e nd of a conversation in which, from the beginning, it had been agr eed that it should be finished with a pause of about one minute. ITC JOURNAL26 Figure 4

CUADERNOS DE TCI 27 Figure 5

That is why the analysis shows on a first superficial observatio n only the (very highly amplified) image of an almost random noise interference with a predominance of certain frequency ranges, particularly at about 2.8 kHz. However, if we examine synchronously a listening attempt with the Visible Speech Diagr am, we can find some temporal and frequency bounded energy concentrations. They are in accordance with sounds in the hear d text ‘STOP – AS YOU LIKE ONE (?) A PAUSE’. The analysis also confirms the American pronunciation of the ‘O’ as an ‘A’ in the f irst word, already noticeable during a first listening attempt. If here in figure 5, in contrast to the view expressed in the first part o f this paper, the heard sounds are given a meaning, this is done sole ly for a better interpretation of the Visible Speech Diagram. The correctness of this interpretation should not be assumed at a ll. This also implies the sounds and sound sequences appearing in bra ckets, which can't be verified in the diagram itself. As a final example of an analysis of a ‘recording’ we will examin e a whispered ‘recording’ (figure 6). This ‘recording’ occurs wi thin a sentence spoken by B., during a break which is only disturbed by noise signals. Of course this is very suitable for sound detecti on. On the other hand, the whispered expression is very low, so that o nly the first syllable in the Visible Speech Diagram can be verified. In addition, the characteristic structure of the fundamental fr equency of the speech in the diagram cannot be used here to distinguish between speech and noise because soft speech arises only due to ITC JOURNAL28 Figure 6

noises in the voice formation channel without involvement of t he vocal folds. In figure 6 the actual ‘recording’ (denoted by ‘Einspielung’) and the utterances of speaker B. (‘Sprecher B. ’) are clearly distinguishable. For a better clarification of the sou nd content of the ‘recording’ in figure 6 there was another recording pro duced in which the speaker B. repeated the alleged meaning of this ‘recording’ and whispering occurred during his utterances. Fi gure 7 shows this repetition. The relationship between the whispe red voice sample and the first syllable ‘R (?) AS’ (the only part shown) in the ‘recording’ can be clearly recognised. In figure 6 the formant a rea of the ‘A’ and the area of the ‘S’ are relatively clearly identifiab le. CUADERNOS DE TCI 29 Figure 7

Because of the absence of the fundamental frequency of the speech in the figures nothing can be stated about the possible ident ity of the two speakers nor is the representation of the ‘recording’ so unds suited to assign the depicted energy concentration with exact certainty to certain speech sounds. The normally advanced argument that the ‘recording’ in figure 6 is produced by the spe aker B. himself, can be responded to in that it seems impossible to s witch from an unvoiced and very low whispered utterance to a voiced and loud spoken ‘I’ in the word ‘ICH’ that followed, in such a short period. This period can be seen directly in figure 6. Taking into account only the visible prefixes of the ‘recording’ it takes about 0. 05 sec. from the end of the ‘S’ until the beginning of the ‘I’ of ‘ich möch te Sie’. Taking into consideration the second syllable [MUS], w hich is not visible in figure 6 but audible, this period is certainly st ill shorter. Concluding remarks From these examples it can be concluded that the measuring and assessment methods of telecommunication undoubtedly can provi de assistance in identifying ‘recordings’. This applies to both the methodology of auditory experiments and the application of the Visible Speech Analysis to the sound detection of ‘recordings’ . Although there are limitations to the sound comparison betwe en ‘recordings’ and de facto spoken speech, in specific cases the evaluation of such an analysis can give some indication of what th e words/ speech might be in a recording. However, the Visible Speech Analysis does not give the opportunity to determine speci fic sounds with absolute certainty, but it only offers a help in combination with the listening impression of a ‘recording’ t o indicate its sound composition. In the case of voiced speech sounds, the sound determination becomes much simpler with this analysis. A n exact certainty, with the conclusion drawn on the basis of the Vis ible Speech Analysis, cannot be achieved. The method of the visible presentation of voice and noise sign als is most suitable for precise timing between the individual sp eech sounds, from which it is often possible to draw conclusions reg arding the origin of speech utterances. ITC JOURNAL30

Notes *Translation by Professor Uwe Hartmann of the German original article by J. Sotschek: Über Möglichkeiten der Erkenn ung von Sprachlauten - Zur Anwendbarkeit des Visible-Speech- Verfahrens und anderer Methoden auf die Analyse von Tonband- "Einspielungen"). *Published with the kind permission of Zeitschrift für Parapsychologie und Grenzgebiete der Psychologie, Jahrgang 12, Nr. 4, 1970, S. 239-254. (Journal of Parapsychology and Border Areas of Psychology). Grateful thanks to Professor Eberhard Ba uer for providing the original text for the present translation . 1. Fletcher, H. and R. H. Galt: The Perception of Speech and Its Relation to Telephony. Journ. Acoust. Soc. Amer. 22 (1950): 89-151 2. Flanagan, J. L.: Speech Analysis, Synthesis and Perception, i n: Kommunikation und Kybernetik in Einzeldarstellungen, Band 3, Berlin/Heidelberg/new York 1965, S. 234 ff. 3. The delineation in white of areas of sound in Figures 3,4,5 and 6 has been added to make them more visible on the page than the original black outlining. CUADERNOS DE TCI 31