Abbildungen der Seite

The Mechanical Recording of Speech, by G. OSCAR RUSSELL

(abstract only)

The principal reasons for making mechanical acoustic records of speech in an investigation such as that to be discussed by the Conference on a Linguistic Atlas of the United States and Canada are:

1. Mechanical records will remain unchanged after the dialect studied has changed. If the change is slight, a record in phonetic notation may not enable an observer to detect the change, and it can never enable him to determine the precise extent of the change. The acoustic record will.

2. In a scientific study of speech, the acoustic phenomenon is always much more serviceable than a necessarily inadequate written representation or a description of it.

3. The acoustic record will serve to set matters straight (a) whenever doubt arises with regard to any written notation and (b) when problems arise that were not foreseen at the beginning of the investigation.

4. Any phonetic symbol involves an interpretation of sound heard, and it can be understood only by means of an interpretation back into sound. Both processes are subject to error. Where we have the original, a check-up can be made by any number of persons who care to listen. If we relied entirely on phonetic notation, we should have to take one man's word.

5. Mechanical records can be listened to again and again by many observers until disputed observations are settled. A dozen observations, made independently, are of necessity more reliable than the observation of a single individual.

Four types of mechanical recording are now in use:

1. The wax cylinder, vertical cut record made by electrical process. That of Mr. Johnson and the one constructed in the Ohio State University Phonetics Laboratories are of this type.

Advantages: The wax cylinder has a ball-point needle for reproduction which fits the groove closely on all sides, thereby picking up all the speech vibrations most faithfully. The oscillating distortion caused by uneven and varying pressure of the needle against the groove is eliminated. Since the record is cylindrical, speech waves of the same frequency keep the same relative length, which makes for less mechanical distortion. All of these factors tend to favor especially those high frequencies which are represented in very fine vibrations lying along the superficial layer of the slower and longer length curves. They are less

likely to be ground down by reason of having to pull the needle along, or to fail of effective registration or of reproduction in consequence of jamming them together, which commonly results on disk records as the needle moves toward the center.

Disadvantages: Wax cylinders are very perishable. They represent a fire hazard. They crack easily and hence cannot be readily shipped about or stored as permanent records. They can be played only about 100 times without becoming dim. Master records in metal can be made, but for that purpose unshaved records of fixed diameter are required, and the process is expensive.

2. The wax matrix, composition stamped, lateral cut record now most commonly made by electrical processes of various types, which reproduces speech so well as to be almost humanly normal. The Victor Orthophonic and the Columbia records, made by the Western Electric Company's process, are of this type. The Present-day English Section of the Modern Language Association and Columbia University have already made a number of these through Professors Ayres and Greet. Advantages: Too well known to be in need of any statement here. Disadvantages: They can be made only in a well equipped laboratory by highly skilled experts. However, Mr. Steinberg states that the Victor Company makes a portable recorder which could probably be purchased, or else leased, for the duration of the investigation. Other disadvantages are mentioned in connection with the advantages enu merated for the cylinder records.

3. Light ray, or valve, film records, best when electrically recorded. These range from the Western Electric, De Forest, and Bristol types to the Metfessel type.

Advantages: Long speech records can be made without interrupting the subject, whereas the average phonograph record is limited to from 2 to 8 minutes. Filing space for a large number of records can be very much reduced, since the film can be done up in a very small bundle. Scraping and frictional noises are eliminated. High frequencies could be recorded better than by any other means.

Disadvantages: Films are perishable and much more expensive. The recording equipment is much harder to handle.

4. Metal, directly recorded, lateral cut discs, made by electrical process, and efficient enough to reproduce the various sounds very reliably. Advantages: Records inexpensive (perhaps 35 cents each), permanent, easily stored. Recording set simple to operate and easily portable.


If a recording machine is used in the study of American speech it is as essential to put it in charge of a specialist in experimental phonetics who knows its every idiosyncrasy, as it is to place the direction of such an investigation in the hands of a specialist trained in the processes of linguistic research.

The developments that have taken place in the mechanical recording of speech during the last decades give us an overwhelming advantage over those who had to construct the linguistic atlases of Europe. It would be a shame if we did not profit by this recent progress. If we do, this investigation will put at the disposal of linguists of the future an acoustic record of speech without parallel in history.

Discussion of Professor Russell's paper:

STURTEVANT: (1) Note that the Europeans have employed no mechani-
cal methods in preparing their speech atlases. Phonographic records
of speech will be of great value 20 years hence for the study of now
moribund dialects. (2) Suggested that remarks be restricted to the
place of such records in the project now on hand.

KURATH: Does the Speak-O-Phone distort any of the vowels?
RUSSELL: No. It records all the frequencies from 250 up to 4000.

Some of the voiceless consonants are, however, imperfectly recorded.
KURATH: Does the Speak-O-Phone record [s] and [f] as distinct sounds?
RUSSELL: Yes, though untrained subjects fail to distinguish them when
listening to the record.

KURATH: If the fundamental of the voice is below 250 vibrations per second, it is not recorded by the Speak-O-Phone. Can such records be used in a study of American intonation?

RUSSELL: Yes. The rise and fall of the voice is heard by means of the difference tones as if fully recorded.

KURATH: It is sometimes hard to tell by ear, when listening to orthophonic records such as those made by the Victor Company under the direction of Professors Ayres and Greet, whether a Southerner says [hat] or [hrt] (hurt). Would a physical analysis of the recorded sound help in deciding the question?

RUSSELL: Yes, if the high frequencies of [r] are actually recorded. STURTEVANT: There is a difference of opinion possible as to whether mechanical methods should be used or not.

MOORE: I am in favor of mechanical methods, to avoid the inevitable differences in interpretation of the same sound by different listeners. Russell, in giving his data on efficiency, stated that there was greater

efficiency in interpreting the living voice than in interpreting the record, but with training the interpreter could get approximately as good results from the record as from the living voice. But will not differences in training, that is, in perception habits, constitute a source of error?

RUSSELL: The actual facts were given in my paper without any glossing over. But note that practically for the vowels the same efficiency results from interpreting the living and the recorded voice. The great difference is in the final consonants, of which there is a very small number relatively. Training is of great importance in the interpretation of mechanical records. Dictaphone operators were tested and showed 18 % efficiency the first day, but after several days of training they showed an average of 92 %.-In answer to two questions, Mr. Russell stated that syllabic nonsense words, and his own lists, not those of the Bell laboratories, were used in getting the figures which he cited.

STURTEVANT: Mechanical records can be heard many times and by more people than the original utterance of the subject. These are important advantages.

BLOOMFIELD: The records can be carefully analyzed later!

RUSSELL: Yes. The ear will miss subtle differences which analysis will show.

HANLEY: The ordinary dictaphone has proved very useful for rough work. Why not use it for collecting material, and for permanent records pick out the best subjects. The Speak-O-Phone seems to be the best device mentioned by Mr. Russell for this purpose. SALESKI: What part would Professor Metfessel's device described in his Phonophotography in Folk Music play if the Speak-O-Phone were


METFESSEL: It is used after the records are made, and need not be considered at this point. Would it be possible to make records of telephone conversations from the exchange?

KURATH: Only part of the material can be phonographically recorded. Other features must be noted in long-hand, e.g. such groups as [fɔg, lǝg, frǝg] but [kag, klag] in Chicago. Do Romanic scholars think Edmont heard correctly in collecting material for L'atlas linguistique de la France?

MÜLLER: Yes, in all but one or two per cent of the cases.

Such unity is lost if mechanical devices are used in addition to recording by ear the living speech of the subject.

MOORE: If one man does the job there is unity, of course; but no more so than if mechanical methods are used to check up on the field workers' hearing.

MÜLLER: reaffirmed his stand.

STURTEVANT: Do you recommend that one man collect all the material in the United States and Canada?

MÜLLER: One man might work in the South, one in the East, and one in the West.

ROEDDER: A wrong impression has been given as to the lack of mechanical records of dialect material in Europe: (1) The PhonogrammArchiv at Vienna has a large collection; (2) there are plans to supplement the Sprachatlas des Deutschen Reiches by records of the 44000 communities represented on the maps of the atlas.

HANLEY: Mention of such records in France was made at the last meeting of the MLA. However the present plan differs from the European projects in starting with records. As for unity in Müller's sense, we cannot hope for it in this vast country. The field work must be done by several investigators. This fact makes mechanical recording all the more important.

THE SECOND SESSION was called to order by Professor Sturtevant at 9.00 o'clock on Saturday morning, August 3, and Professor Karl Young was elected Chairman. The discussion of the problems proposed in the preliminary circular was at once begun.

I. What action, if any, shall be taken with regard to the effect of
movements of population on dialect?

KRAPP: This is a part of the interpretation of the material, and need
not concern us at first, when our object is merely the collection of
detail. Moreover, it is impractical at present, since practically
nothing is known about the subject. Moved that no action be taken
at present on the question of the influence of movements of population
on the English language in America. Seconded by Long.
KNOTT: Endorsement of Krapp's attitude. (1) The more we spread
out, the less feasible our plan. (2) We are not in a position to con-
sider this question, as we know nothing of American dialects and little
of population movements.

KURATH: granted that the historical interpretation must come later. But in order to choose our 500 communities wisely, we must be guided in part by historical information.

« ZurückWeiter »