VOICES FROM WORLD WAR ONE TIMES AS MEDIA-ARCHAEOLOGICAL EVENT

Wolfgang Ernst
Institut für Musik und Medienwissenschaft an der Humboldt Universität zu Berlin

The Berlin Lautarchiv, as its very name expresses, is not just an audio archive of human voices and ethnic songs from the past, but as well an archive of Laute, which in German refers to phonetic and sonic, even noisy articulation - that is, all kind of acoustic enunciations.¹ Listening to the records with media-archaeological ears, one detects not only the human speech but the expression of the recording apparatus und storage media themselves - the scratches and the revolving rhythms of the Edison cylinders. In the online-inventory of the Lautarchiv, among page-long enumeration of recorded ethnic songs, two artefactual devices are listed which embody the media-archivological condition for listening to such voices from the World War One past at all: items no. (ID) 9311 (type "Plastisches Objekt") Zwei Tonabnehmer (electro-magnetic pick-ups).

The traditional archival approach to recorded voices has been to provide them with separate metadata on formula sheets of paper. In times of "digital humanities" this can be alternatively accessed by sonic analytics, which is: the algorithmic analysis of that enormous mass of voice recordings once they have been digitized ("big data"). Cultural Analysis which historically contextualizes the Lautarchiv recordings and concentrates on their ambiguous cultural meaning is now being matched by cultural analytics which is expressed by spectographs for audio content - a dramatic hift?? of emphasis from the symbolical textual field to the processing of the real audio signals. All of the sudden, voices may be identified by their very spectral individuality, not exclusively subjected to alphabetic registration in written metadata any more. Next to the well-known symbolic order of the archive (which goes with the symbolic order of administration, bureaucracy and the state governmentality) a signal memory arises: the phonographic record.

Almost immediately after its invention, the Edison phonograph was announced in the journal Scientific American. It obviously triggered phono-archival phantasms in the Romantic tradition of the historian of the French Revolution Jules Michelet, who in early Nineteenth century believed to hear the murmurs of the dead in the archives. A true Lautarchiv is being declared:

"That the voices of those who departed before the invention of the wonderful apparatus [...] are for ever stilled is too obvious a truth; but whoever has spoken or whoever may speak into the mouthpiece of the phonograph, and whose words are recorded by it, has the assurance that his speech may be reproduced audibly in his own tones long after he himself has turned to dust. [...] A strip of indented paper travels through a little machine, the sounds of the latter are magnified, and our great grandchildren or posterity centuries hence hear us as plainly as if we were present."²

Natural sound is evasive, liquid, in itself unrecordable beyond the bodily range, but technical media (different from alphabetic phonetic writing which "freezes" the human voice into a range of a very limited symbolic code) are able to de-freeze recorded voices in almost all frequencies (that is, the Lacanean "real" of the voice) by re-play. After two millennia of the phonetic alphabet there is a new kind of cultural technology as sound recording.

The target of sonic analytics is not individual speech in terms of meaningful content, but first of all subsemantic insights which can be derived from the very materiality of sono-cultural articulation: phoné (German "Laut").³ Very literally, the phonographic collection of early voice recordings (Lautarchiv) based at Humboldt University, Berlin is an ideal subject for such a sonic archaeology. The Lautarchiv encompasses three groups: a) Famous voices (which for political reasons were partly neutralized or even destroyed after 1945); b) truly archival recordings of local speech dialects, based on a set of artificial word sequences in order to achieve formal comparability (so-called Wenker-sentences) with the speed of the recording beeing controlled by a supplementary oscillographic time code, and c) recordings for musical ethnology (mostly Africans and Indians from the French and British Army in the World War One Halbmond prisoner camp at Wünsdorf south of Berlin).⁴ The phonological target was inscribed into the Lautarchiv by its promoter Wilhelm Doegen from the beginning - notwithstanding the circumstances of its coming-into-being with recordings in a prisoner camp. While cultural analysis concentrates on this ambivalent historical and discursive context, with a different epistemological vantage point media archaeology lends its ears to knowledge which can be derived from the actual media articulation contained in the technical archive itself.

When these recordings since April 1920 became integrated as Department of Phonetics (Lautabteilung) into the Prussian State Library in Berlin to be reproduced on schellack discs and as transcription for educational distribution⁵, the original relation between spoken orality and its grama-phonic derivative (the phonetic alphabet⁶) was reversed again by the intrusion of real audio signals into the symbolical order of the librarians' Gutenberg world of letters, resulting in a kind of animated phonetic library.⁷ Printed text as it were start to speak from a gramophonic storage medium which (different from the alphabet) does not discriminate between signal and noise any more.⁸ Therefore the Lautabteilung consequently accumulates natural and artificial noise („Geräusche natürlicher und künstlicher Art und andere“) such as the sound of tree leaves in the wind. What had started as interlinear auditory hallucinations in romantic literature becomes real in sub-symbolic recording media. The gramophonic recording method for waveforms in the so-called glyphic system on wax discs inscribes even sonic warfare into the new cultural memory as écriture automatique.⁹

When it came to detect minute variances and to eliminate subjective inexactitudes in listening to the recordings of foreign dialects and voices, the limits of hand-written phonetic transcription became obvious, leading instead to the application of visual oscillograms and Fourier Analysis of the phonetic wave forms.¹⁰ When explicit listening gets replaced by technographical measuring of sonicity, the gap between cognitive musical understanding and physical recording (the material, tonally integrative engraving of a musical event in the phonographic groove) opens. Just like the point of the gramophone needle can make only one movement at one time, "the illuminated disk of the oscilloscope shows only one line, no matter how many tones are sung into the microphone simultaneously. [...] what the apparatus registers as one wave, we hear as multiplicity of tones - and as a organized multiplicity. [...] mathematical analysis of the shape of the line permits us to deduce the individual waves that are combined in it. Yet [...] our ear accomplishes, effortlessly, continuously, and instantaneously, what costs the skilled mathematician a considerable expenditure of time and energy"¹¹ - until the Fast Fourier Transform algorithm arrived in real-time digital computing of sound. Even the much more detailled spectral voice analysis which had just been developed in Zuckerkandl's generation subjected the complex dynamics of sonic events once more to the visual knowledge regime since sonagrams, though expressing delicate micro-temporal variations, tend deciphered analog to alphabetic writing.¹² But the tempor(e)ality of sonicity can never be caught in a frozen state but always points beyond the moving still - as has been discussed by Bergson's critique of chronophotography and the cinematographic illusion of "movement".

Such ancient phonetic oscillograms today represent the truest media-historiography of that time - while at the same time challenging the historical narrative of their recording context. The real archive of sonic articulation emanating from such recordings is no longer literary stories but numerical analysis - finally resulting in digital sampling of the analogue records which is the transduction of ghostly voices into computability.

It was on the linguistic field that effective algorithms for recognition have first been developed - as transformation of physically measurable wave forms of speech signals into electric impulses. The operation is based first on electronic transduction and then the transformation of the time-signal to its frequency number.¹³ Thus, sonicity can not be reduced to the dynamics of waveforms, but encompasses mathematical operations and subsequently their machinic computing as well. Once a series of digits can represent waveforms, sound is liberated from its acoustic phenomenology. The statistic tools from corpus-based linguistics have been adopted for music analysis:

"While the basic elements and features (or tokens) over which statistics are computed naturally differ between linguistics and musicology, the statistical concepts that allow us to infer regularities within the specific domain are quite similar or nearly identical. Among the chief statistical concepts that can be derived from frequency counts of tokens / features, and that are employed in both fields, are Markov models, entropy and

mutual information, association measures, unsupervised clustering techniques, and supervised classifiers such as decision trees."¹⁴

Sonic analysis in a Lautarchiv focuses on the materiality of sound equally valuable in its acoustic and its technological sense. In modern Greek radio broadcasting is called radiophonia. Analog to telephony, not speech or music as semantic content is named here, but the phonetic materiality (ancient Greek phoné / German Laut) of any kind which is transmitted by a neutral medium called radio. In terms of a (media) archaeology of acoustics, the nature of sound is spectral, thus undermining the symbolical (Pythagorean) order of harmonic tonal relations in integer numbers - just as the letters in an alphabet only symbolically relate to the phyiscality of actual speech phonems which are as "differential" (Arseny Avraamov) as the glissandi of the Theremin Vox contructed as the first mass-reproduced electronic music instrument by Leon Thermen in revolutionary Soviet Union.¹⁵ With sound production which is subliminal to human perception, sonicity (different from sonority) starts.

Sonic analytics has been provided by Nikita Braguinski for such an archival recording: the Russian Volk Song Vo kuznice, recorded 1916 with a chorus of Russian war prisoners during World War One. Lautarchiv, inventory no. PK135-Mersbach)

Instead of traditional alphabetical transcription, open source software like Praat allows for (and incites) new kinds of "archive" mobilization: signal-based speech analysis, active archaeology of past sounds. Under such observation, audio recordings are not just archival objects any more, but become items in an experimental laboratory of presence. This presence is a distorted one, though. Trendelenburg describes the distortions of sound fidelity which are essential features of phonographic and grammophonic records.¹⁶ This is the bandwidth limit of mechanical sound records from the past as compared to electro-magnetic and finally digitally processed recording.

1 The following arguments are being elaborated in the "Lautarchiv" chapter of: Wolfgang Ernst, Sonic Time Machines. Explicit Sound, Sirenic Voices and Implicit Sonicity in Terms of Media Knowledge (forthcoming)

2 Anon. (The Editor), A Wonderful Invention - Speech Capable of Indefinite Repetition from Automatic Records, in: Scientific American, 17. November 1877, 304; see chap. 6 "A Resonant Tomb", in: Jonathan Sterne, The Audible Past. Cultural Origins of Sound Reproduction, Durham / London (Duke University Press) 2003, 287-334 (297f)

3 For several socio-linguistic and computer-based analyses in the techno-culturally variant coding of human voice frequencies see Zakharine / Meise (eds.) 2013

4 See Britta Lange, Ein Archiv von Stimmen. Kriegsgefangene unter ethnografischer Beobachtung, in: Nikolaus Wegmann / Harun Maye / Cornelius Reiber (eds.), Original / Ton. Zur Mediengeschichte des O-Tons, Konstanz (Universitätsverlag) 2006, 317-341 (esp. 335f). An almost complete list of the both phonographically and symbolically registered recordings is provided online: http://www.sammlungen.hu-berlin.de/sammlungen/78.

5 Lautbibliothek: Phonetische Platten und Umschriften, ed. by the Lautabteilung der Preußischen Staatsbibliothek, 1920 onwards

6 The architectural front of the German Library at Leipzig (Deutsche Bücherei), founded in 1913, still displays a monumental quote from a Schiller poem: "Körper und Stimme leiht die Schrift dem stummen Gedanken [...]."

7 „Die toten Buchstaben und Büchertexte werden hier durch die Ergänzung der Lautplatte lebendig und verkörpern eine wirkliche Lautbücherei." Wilhelm Doegen, Die Lautabteilung, in: Fünfzehn Jahre Königliche und Staatsbibliothek 1921, Berlin (Preußische Staatsbibliothek) 1921, 253-258 (253)

8 „In Graphie und/oder Phonie des Titelworts `Sprache´ steckt die Lautverbindung `ach´“: Friedrich A. Kittler, Aufschreibesysteme 1800 / 1900, München (Fink) 1985, 48

9 „Gewehrfeuer (gun fire) for a theory of sonic explosion, and the sound of air planes ("Fliegergeräusche"): Doegen, op. cit.

10 Alois Brandl, Lebendige Sprache: Beobachtungen an Lautplatten englischer Dialektsätze, mit einem Anhang von Wilhelm Doegen, Zur Lautanalyse aus dem Klangbild des englisches Dialektwortes "man", aus der Lautplatte gewonnen nach dem elektro-oszillographischen Verfahren, in: Sitzungsberichte der Königlich Preußischen Akademie der Wissenschaften, Phil.-hist. Klasse (1928), 72-84

11 Victor Zuckerkandl, Sound and Symbol. Music and the External World, New York (Pantheon) 1956, 333f

12 See Ralph K. Potter / George A. Kopp / Harriet C. Green, Visible Speech, New York (Van Nostrand) 1947, and Boris Yankovsky's sound spectrography (as mentioned above).

13 H. Schnelle, Automatische Sprachlauterkennung, in: Kybernetische Maschinen. Prinzip und Anwendugn der automatischen Nachrichtenverarbeitung, Frankfurt/M. (S. Fischer) 1964, 208-219 (211)

14 Daniel Müllensiefen / Geraint Wiggins / David Lewis, High-level feature descriptors and corpus-based musicology: Techniques for modelling music cognition, in: Systematic and Comparative Musicology: Concepts, Methods, Findings, hg. v. Albrecht Schneider, Frankfurt am Main u. a. (Peter Lang) 2008, 133-153 (140)

15 See Andrey Smirnov, Sound in Z. Experiments in Sound and Electronic Music in early 20th Century Russia, London (Koenig Books) 2013, 44

16 Ferdinand Trendelenburg, Klänge und Geräusche. Methoden und Ergebnisse der Klangforschung, Schallwahrnehmung, grundlegende Fragen der Klangübertragung, Berlin (Julius Springer) 1935, 51