NATURE 



[October 26, 191 1 



We have been getting occasional pieces of a curious 

 material from the diamond mines, which may prove to 

 have a relation to the mineral described in Nature of 

 Septembei 7 (J. R. Sutton, "A New Mineral?"), and also 

 may throw some light ultimately upon the origin of the 

 diamond. To outside appearance, in extreme cases, the 

 material has a cindery look ; in less extreme cases its 

 diamond affinities are fairly evident. It can be readily 

 disintegrated witli a mineralogical file, but it has hard 

 corners which will scratch corundum. The specific gravity 

 j s 3-3 t0 3-5i <■«. -lightly lighter than diamond. It is 

 insoluble in acid, is feebly magnetic, and when suspended 

 by a light thread or floated on water (on a cork) shows 

 distinct polarity under the influence of an ordinary large 

 steel horse-shoe magnet. When it is crushed a small bar 

 magnet will readily take up small specks of it. (The 

 mineral previously described in Nature, by the way, shows 

 no polarity.) 



Somi months ago I casually examined some pieces of 

 this material, and concluded that they were diamond (bort) 

 with enclosed impurities. Some of the impurity is now 

 proved to be iron, which shows that the statement some- 

 times made that diamond is not found in association with 

 iron is not quite correct. 



Some pieces of this material which had been extracted by 

 the electromagnets at the pulsator were brought in by Mr. 

 Stewart (the manager of the pulsator) a few days ago. 

 They were very unlike the stuff readily recognisable as 

 diamond, but the chain of gradation from these to some- 

 thing more nearly approaching true bort is fairlv complete. 

 Whether a diamond buyer would put the same commercial 

 value upon them as he would upon bort is quite another 

 question. Up to the present time I have not come upon 

 any true bort which shows the same magnetic properties. 

 Like true bort, however, this material is a good conductor 

 of electricity. 



As a distinctive name for this variety of bort, or iron 

 bort— if bort it may strictly be called— Stewartite would be 

 suitable. J. R. Sutton. 



Kimberley, September 30. 



A Starling's Deception. 



Three weeks ago, or, to be quite correct, on Sep- 

 tember 22. I was considerably startled and surprised, on 

 going into the garden at 9.30 a.m., at hearing what I 

 thought was a wryneck's call in a tree not many yards 

 off. I listened, and in a few minutes the cry came again 

 clear and distinct as one hears it in the spring and early 

 summer. I was astonished, knowing it to be a rare thing 

 to hear the wryneck after the middle of July. I approached 

 the tree (in which two or three starlings" were chattering 

 and whistling) and tried to get a sight of the supposed 

 wryneck, but did not, although the call was repeated 

 several times. I put down my "failure to the thickness ol 

 the foliage and the ivy-grown trunk, somewhere in the 

 midst of which the bird was doubtless in hiding. 



Well, the next morning, and on several day's following, 

 the unseasonable, but otherwise very pleasant, note con- 

 tinued In be lend, and always from the same tree and. 

 apparently, in association with the starlings, for I noticed 

 that the cry invariably came after one of the starlings had 

 whistled. The whistle, in fact, seemed to he the signal 

 for the wryneck to sing. 



It struck me as being altogether verv curious, and I 



determined to find out, if possible, more about it. So one 



morning (September 27) I resolved to investigate the matter 



more closely. Standing under the tree, and after a little 



patieni waiting. I got a starling well into view and watched 



him carefully. Wagging his head from side to side he 



1 nattered and cackled for all he was worth ; then came the 



whistle, and immediately afterwards the wryneck's note, 



in uttering which I quite distinctly saw the quick movi 



merit of the beak. And so the mystery was solved ' 1 



Oping to see a repetition of the performance, but 



he bird, 1 fancy, caught sigh! of me and flew awav. On 



thn of Hi' following days I tried to catch him 



again, hut was not successful. In the early 



' ' "'" 1 thi 1 ry was not heard (at anv r: 



2191, vi >r . 87] 



myself), but n fell on my ear once more, and for the la~t 

 time, "ii October 6, and from the same tree. 



Starlings are great mimics, I believe, and 1 am wonder- 

 ing if this particular bird has Kern reared in thi 

 vicinitj ol a wryneck's nest, and so caughl the note from 

 the parent wryneck. However this may be, I thought the 

 incident would interest your readers, and perhaps 

 additional facts of a similar nature from some of them. 



I may add that in 1901, from August 19 to Sept 

 a friend and myself heard almost daily what we tirmh 

 believed to be a wryneck's cry. It surprised us, 

 but, other than being very interested in hearing the un- 

 seasonable note, we never investigated the matter properly) 

 The question now arises, were we and the neighbours 

 deceived by a starling in 1901 as I was so nearly deceived 

 by one this autumn? Basil T. Rowswell. 



" Les Blanches," St. Martin's, Guernsey, 

 October 18. 



Hot Days in 1911. 



Mr. MacDowall's dot diagram in Nature of October 12 

 certainly shows high correlation between the number of 

 hot days in a quinquennium and the difference between 

 this and the number of hot days in the next quinquennium, 

 and Mr. Corless in Nature of October 19 finds the value of 

 the correlation coefficient to be —0725 ; but the conclusion 

 is not that the number in one quinquennium is correlated 

 with the number in the next. 



If v, is the departure from mean value of the number 

 of hot days in one five-year period, and .v, that in the 

 next succeeding, then, if these are wholly independent 

 variables, sumx,x: 2 = o, the minus values neutralising the 

 plus, and the coefficient of correlation between x, and 

 x„—x lt which is 



sum x i (x 2 — x 1 )/ \/sum x, : X Vsum (x 2 — x,) 2 , 

 becomes 



— sum *,'/ v'sum iy X v^sum (x'+x*), 

 or — 1/\<2> since sum 3t/ = sura x* in a long series. 



The value —1/^2, or —0707, is within the limits 

 — o-7 2 5 + °'°S9 given, and the conclusion is that the correla- 

 tion between successive quinquennia is nil. 



This conclusion, based on the figures of Mr. Corless, 

 must render ineffectual Mr. MacDowall's endeavours to 

 make long-range forecasts of weather by correlations at 

 five years' distance, and will disappoint any hopes that the 

 new method may have raised in the minds of " official 

 meteorologists." H. E. Soper. 



University College, London, October 23. 



Mr. MacDowall, in dealing with the number of " hot " 

 days in a year (Nature, October 12. p. 485), com 

 two series of numbers which are not independent, and uses 

 the comparison in an attempt to make seasonal forecasts. 

 His method does not appear to be statistically legit in 



He obtains a series of numbers N + n,, N + » 



N+n,„_... representing the total number ot "hot" days 

 for periods of five years, 1, 2, 3, 4, 5:2, 3, 4, 5, 6, &c., 

 and plots a diagram showing the relation between N+n, 

 find «,._,.-, - 11,.. N being the mean of the five-year totals. 

 If the scales of ordinates and abscissa; were the same, and 

 the series of numbers N + n,, &c, represented a random 

 selection, we should expect to find in the diagram a 

 number of dots distributed more or less symmetrically 

 about a line bisecting externally the angle between the 

 axes. This is what Mr. MacDowall obtains in his 

 diagram on p. 485, allowance being made for his differ- 

 ence of scale. The diagram, as it stands, cannot ther 

 help the forecaster. 



We should expect also to find a large correlation 

 coefficient between N+n. and «,.+.,- «,-. For a long series 

 of numbers in which there was no correlation between 

 N + « r and N+n + „ the value of the coefficient between 

 N + n,., « r+ , n, would be —^2, or —0-71. say. Mr. 

 Corless finds from Mr. MacDowall's figures a value 

 —0-73. Clearly, therefore, this cannot he taken to prove 

 periodicity. 



The total number of " hot " days in the nine years 

 preceding 1911 is, according to Mr. MacDowall, ;86, com- 

 pared with an average of 9 ■ 77 693, 30 thai unless then 



