October 28, 1909] 



NA TURE 



533 



down on the British Admiralty Chart for 1907 are in error 

 by amounts reaching 16 units in the third decimal C.G.S., 

 and those of the Deutsche Seewarte for 1902 (the most 

 recent chart not being available to us just now) require 

 corrections running up to 12 units iii the third decimal 

 on the values obtained by the Carnegie. A part of this 

 error is, of course, due to secular variation, but the major 

 portion appears to be due to defective data. Voir both 

 charts there are five negative corrections, amounting in the 

 maximum to 4 un-its in the third decimal, four zero values 

 and thirteen -positive ones for the " B.A." chart, and two 

 zero and fifteen positive corrections for the " Seewarte " 

 chart. On the average for the Atlantic, from Long Island 

 to Falmouth, the chart values are too low by about i/45th 

 part of the value of the force, even at times amount- 

 ing to 10 per cent, to 20 per cent. In the Pacific Ocean 

 the intensity charts gave, in general, too high values by 

 about i/25eh part. 



Since the above was written the Carnegie has been 

 swung in Falmouth Bay, observations of the three 

 magnetic elements (declination, dip, and horizontal in- 

 tensity) being made on eight equidistant headings. The 

 mean results of the entire swing are in excellent agreement 

 with w'hat would be deduced from Riicker and Thorpe's 

 magnetic charts of the British Isles applying secular varia- 

 tion corrections as deduced froin the records of the 

 Falmouth Magnetic Observatory. The declinations and 

 dips a^iree within two or three minutes, and the horizontal 

 intensity within i /2000th part. This again proves that 

 there are no deviation corrections of whatever nature to 

 be applied to the Carnegie results ; also that the instru- 

 mental constants have been well determined. We have 

 here also a satisf.':ictorv proof that if the distribution of 

 magnetism is uniform, as appears to be the case here, the 

 sea values, upon careful ineasurement, will bo found in 

 agreement with the shore values. 



\.. .\. Bauer. 



W. J. Peters. 



THE ANALYSIS OF SOUNDS USED IN 

 SPEECH. 

 'T'HE characteristics of the simpler sounds which form 

 the elements of speech have been studied by many 

 physicists. At first, attempts were made to reproduce the 

 vowel sounds synthetically ; Helmhollz achieved noticeable 

 successes in this endeavour, but by themselves synthetical 

 methods can never be quite satisfactory, since it is neces- 

 sary to prove that the ear does not possess properties 

 similar to those possessed by the eye ; it is well known 

 that two mixtures of light may produce identical effects 

 on the eye, although the component waves may be quite 

 different in the two cases. 



Many attempts have been made to analyse complex 

 sounds, but the results obtained up to the present have not 

 been quite conclusive. Dr. Erskine-Murray used a tliin 

 membrane stretched slackly over the small end of a conical 

 tube ; a light mirror was adjusted so that it was rocked 

 to and fro by the motion of the membrane, and a beam 

 of light was reflected from this on to a revolving mirror 

 and thence on to a screen. Such an arrangement is very 

 sensitive for sounds of low pitch, but it is mechanically 

 unsuitable for analysing sounds of high frequency. 

 Hermann, McKendrick, and Bcvicr have attacked the 

 same problems by analysing phonograph records, and have 

 obtained much useful information ; but there is small 

 reason to suppose that a phonograph record, however good 

 It may be, affords ait exact equivalent to human soecch. 



For some time past Mr. G. Bowron, of 57 Edgware 

 Road, has devoted attention to the construction of an 

 oscillograph which nroiects on a screen a curve represent- 

 ing the .sounds produced by a gramophone ; a brief notice 

 of this oscillograph was given in. N.^iTL'RE of May 21, iqoS 

 (vol. Ixxviii., p. 6q). The curves obtained are very in- 

 structive, and thev appear to possess as much detail as 

 those obtained by the elaborate method of Prof. 

 McKendrick ; but the problem of the analysis of the sounds 

 used in speech can scarcely be solved in this way. 



The best curves representing vowel sounds have been 



NO. 2087, VOL. Sll 



obtained by Mr. Duddell ; some of these are published in 

 the Journal of the Institution of Electrical Engineers (vol. 

 .xxxix., No. 186, 1907, pp. 545-6). They were obtained 

 by speaking into a " solid-back " telephone transmitter, 

 and transmitting the periodic current obtained thereby 

 through the high-frequency electrical oscillograph invented 

 by Mr. Duddell. The regularity of these curves is very 

 striking ; a perfectly definite curve corresponds to each 

 vowel sound, and with some practice it would be possible 

 to decipher a telephone message by inspecting the curve 

 which corresponds to the telephone current. It may be 

 mentioned, in passing, that the analysis of the sounds 

 which constitute speech is now acquiring commercial 

 importance in connection with the problem of telephone 

 transmission. 



Mr. Bowron has constructed an acoustic oscillograph 

 intended to show the vibrations comprised in ordinary 

 speech ; but although this gives .some interesting informa- 

 tion, the curves obtained are not nearly so good as those 

 obtained by Mr. Duddell. The arrangement used is some- 

 what like that due to Dr. Erskine-Murray, but the slack 

 membrane is replaced by a ferrotype telephone diaphragm. 

 In practice it is found that Mr. Bowron 's oscillograph does 

 not give good curves for sounds of low pitch : this is dite 

 to the fairly high frequency of the natural vibrations of 

 the diaphragm. It has been pointed out by Mr. A. 

 Campbell (Jour. Inst. Elec. Engineers, vol. xxxix.. No. 

 iS6, p. 533) that even a microphone transmitter has certain 

 definite free periods, and that sounds of the corresponding 

 pitches are greatly reinforced. Mr. Campbell's experi- 

 ments in proof of this are worthy of mention here. A 

 solid-back microohone is put in circuit with a battery of 

 6 volts or 8 volts and the primary circuit of a fair-sized 

 spark coil. The secondary of this coil is connected to a 

 rellcctinqf electrostatic voltmeter reading up to 10 volts. 

 The deflections of the voltmeter afford a sensitive indica- 

 tion of sounds received by the microphone. If the nearly 

 pure note of a stopped organ-pipe be sounded, a deflection 

 of the voltmeter is produced, and if the nitch of the note 

 be gradually raised the spot of light rushes off the scale 

 when the note attains certain definite frequencies. This 

 arrangement might be used with advantage in cases where 

 measurements of the intensities of sounds are required. 



There can be little doubt that much remains to be done 

 in connection with the analysis of the sounds used in 

 soeech. Thus the curve obtained by Mr. Duddell for the 

 00 sound in " coo " scarcely differs from a sine-wave 

 curve; a slight alteration of the mouth introduced the 

 oct.ave of the fundamental vibration. According to the 

 accented theory of vowel sounds, each vowel is distinguished 

 by the reinforcement of those partials which lie near to 

 certain definite pitches. The question could be .answered 

 dePnitely if we possessed a diaphragn which would vibrate 

 indifferently for all pitches ; and since the drum of the ear 

 does this, it may be hoped that we may be able to realise 

 the conditions necessary to the solution of the problem. 



Edwin Edser. 



A 



HEAT TRANSMISSION. 

 P.APER on heat transmission was read by Prof. 

 W. E. Dalby before the Institution of Mechanical 

 Engineers on Friday, October 15. The object of the re- 

 search was to place before the members a general view 

 of the work which has been done relating to the trans- 

 mission of heat across boiler-heating surfaces, and in 

 carrying this out more than 500 papers have been read 

 and abstracts of the more important prepared. 



In a furnace heat is transmitted by three methods 

 simultaneously, viz. radiation, convection, and conductiort. 

 It is extremely difficult to analyse the results of experi- 

 ments, so that the heat transmitted by each of these 

 methods may be stated separately. Formula: of simple 

 type have been devised to express the results of definite 

 sets of experiments ; the application of such formul.-c should 

 be strictly limited to cases in which similar conditions, 

 prevail and fall within the range of the original experi- 

 ments. 



In discussing radiation, the author gives Stefan s law 



