;o2 



NA TURE 



[OCTOI 



9-^3 



Letters to the Editor. 



[ The hiiitor does not hold himself responsible for 

 opinions expressed by liis correspondents. Neither 

 can he undertake to return^ nor to correspond "with 

 the writers of, rejected manuscripts intended for 

 this or any other part of Na'IURK. No notiie is 

 taken of anonymous communications.] 



Correlation of Upper Air Variables. 



1 DO not see that Prof. Mahnlanobis (Nature, 

 September i, p. 323) has given any goo<l reason for 

 the statement that the correlation coefTicients that 

 1 have obtained from the EngHsh balloon ascents are 

 to be taken as the upper limit of what is p)ossible, 

 excepting that Capt. I)ouglas working on a different 

 system in one specific instance has obtained a lower 

 value. I freely admit they may be wrong ; unless 

 one luis a very large sample one always has to reckon 

 with the casual error of a correlation coefficient, but 

 there seems no rea.son why I should not equally well 

 accept Dr. Chapman's conclusion that they are too 

 low. 



Taking Prof. Mahalanobis's equation (i) (Memoirs 

 of the Indian Meteorological Department, vol. xxiv., 

 pt. ii., p. 12), transposing it somewhat, and re- 

 arranging, we get the following expression for the 

 correction for the observational errors : 



•'■'■ I 



I ' liv' f ,v, 



\ ^o« *'j:,Vi 



~bx I 



.S. / 



-\r ir '"-r '*^* 



where x and v denote the true departures from the 

 mean, ;r, and Vi the observed departures, and a and b 

 the errors. 



Let us take -the special case of the correlation 

 between pressure and temperature at a fixed height 

 between 4 and 8 kilometres. Here r^ ^ is equal to 

 0-85 and the ratios 5„/Sx and s^ls, are known to have 

 a value of about 1/5. 



Substituting approximate numerical values the 

 correction is 



o-2o(o-85r„, 4 o-85r„, - >-„, - r^^) +o-04(o-85 - r,j) 



-o-o2(»'„-ytJ2. 



Owing to its comparatively high numerical co- 

 efficient the first bracket is the important one, and 

 a negative correction requires that r^^ and r^, should 

 be negative and r^y and r^, positive. I can see no 

 reason why the correlation values should be anything 

 but casual ; they will certainly be small. Moreover. 

 X and y are positively and highly correlated and 

 therefore r^x and r„, are likely to have the same 

 sign ; so are r^t and r^y, hence it does not seem 

 likely that the term can supply a large correction 

 either positive or negative. 



In the second bracket the coefficient Ta^ is certainly 

 positive for the special case where a and b refer to 

 the errors of temperature and pressure at about 6 

 kilometres height. This is apparent because r-, is 

 calculated by Lagrange's formula and a positive value 

 of (a) increases the value of yi and therefore increases 

 (6), but the casual error of Vi due to faulty cahbration 

 or incorrect working up will prevent the correlation 

 between a and b being as high as 085 and the term 

 will be positive. The third bracket is the square of 

 a small quantity multiplied by 0-02 and is insignificant. 

 Thus it appears probable that on the whole the com- 

 puted correlation coefficients are somewhat too low. i 



NO. 2814, VOL. I 12] 



There can be no reasonable doubt that the corr« : 

 tion I>etwecn certain variables in the upper air 

 very high, and any theory of the genesis of cvf In; 

 and anticyclones to be satisfactory must ace 

 such correlation. 



I should like to add that I have never thuu, 

 that the seat of atmospheric disturbances wan in • 

 :'■• * : 'icro, but, since upi>er air obscr\ 

 lablf, have held tnat the winds <' 

 ci.. .M.ii.oii in the upper part of the tn>' 

 responsible for the formation and mrii 

 cyclones. This fits in sati^f ■'"'•• 'v with i.i. ivi>... 



variations of temperature. 

 Hen.son, Wallincfonl, I3erk- 



\V. H. Dines 



Greek Orthography In Scientific Nami'^ 



It is difficult, as corresp)ondents in Nati hi. ii.im.- 

 noted, to preserve orthography in scientific names 

 derived from the Greek. A good example of the con- 

 fusion which has been allowetl to become inevitable 

 occurs in the similarity of the generic title of two very 

 dissimilar shrubs. Chionanthus Virginica has been 

 named from X"^*" — snow — because of the masses of 

 white blossom it bears at midsummer ; while Chimi 

 anthus fragrans, flowering in midwinter, ought to 1>«- 

 written Cheitnonanthus, from x"A'w*'. winter. To 

 each of these Greek generic names a Latin adjective 

 has been tacked, which serves to distingin 

 species, but may offend the .scholar. 



Hi.Kiu Ri Maxwell. 



Monreith, Whauphill, 



Wigtownshire, N.li. 



X-Rays and Crystal Symmetry. 



It has long been recognised that angular measure- 

 ments do not always carr>' one beyond a determination 

 of the system, and that other methotls of investigation 

 are needed if the crystal is to be assigned to its class 

 of symmetry. But different methods do not always 

 give the same result, so that some principle of dis- 

 crimination has to be adopted. In the past the 

 principle universally applied has been that of greatest 

 common measure, the crystal being correspondingly 

 relegated to the highest class, the symmetry of which 

 is common to the various symmetries observed (in 

 most cases this leads to the lower of two observed 

 symmetries, since the symmetry of one is generally 

 wholly contained in that of the other). It must be 

 noted that all class assignments are provisional and 

 liable to modification (necessarily in the direction of 

 lower symmetry) as new evidence is forthcoming. 



The above symmetry has hitherto always been 

 regarded as the true symmetn>- of the internal struc- 

 ture. This \dew has been somewhat questioned by 

 E. T. Wherry- (Atner. J. Sci., 1922, vol. 4, p. 237) and 

 repudiated by R. W. G. Wyckoff {ibtd. vol. 3, 

 p. 177 ; vol. 4. p. 469). It is much to be regretted 

 that considerations of space prevent any discussion 

 of Wherr\-'s paper, for it is in many ways suggestive. 

 The issue raised by Wyckoff is, however, more clearly 

 defined. As a result of a renewed X-ray examination 

 of sal-ammoniac he finds that there is no possible 

 model which will simultaneously satisfy Tschermak's 

 symmetry, deduced from surface studies, and the 

 X-ray data (a model can be found to agree with either 

 of two higher symmetries, the ambiguity arising from 

 an impossibility of placing the hydrogen atoms on 

 account of their small scattering p)ower). This leads 

 to an entirely new definition of symmetry, as being 

 that of the constituent parts (the atoms) as revealed 

 by X-rays. The evidence of such surface phenomena 



