Contribution of each , Ancestor to the Heritage 401 tlian the left .
The difference between the means of the two sides would appear to vary in different races ; also it is distinctly greater in woman than in man .
( 11 ) The absolute correlation between the lengths of the long bones is high , while , on the other hand , the correlation between length and breadth Avould appear to be weak .
( 12 ) On comparing with the Aino it would seem that serially homologous bones are more strongly correlated with one another than are non-boraologous bones .
Also the femur and tibia are more closely correlated than the humerus and radius .
Distal bones would appear to be somewhat less correlated with one another than are proximal bones , c.g. } there is slightly less correlation between the tibia and radius than between the femur and humerus .
In the New Race woman was distinctly less correlated than man .
( 13 ) Sometimes it is desirable to refer the variations to some standard , such as stature .
The index-correlations tend to be lower than the absolute correlations .
From the results we have obtained it appears doubtful as to wrhat meaning we can attach to the difference between the observed correlation of ratios and Professor Pearson 's " spurious correlation .
" ( 14 ) The New Race approached the Negro in the proportional lengths of the limb-bones to one another , but the sacral and scapular indices were nearly identical with those of Europeans .
" The average Contribution of each several Ancestor to the total Heritage of the Offspring .
" By Francis G ALTON , D.C.L. , Sc. D. , F.R.S. Received and Read June 3 , 1897 .
In the following memoir the truth will be verified in a particular instance , of a statistical law of heredity that appears to be universally applicable to bisexual descent .
1 stated it briefly and with hesitation in my book 'Natural Inheritance ' ( Macmillan , 1889 ; page 134 ) , because it was then unsupported by sufficient evidence .
Its existence was originally suggested by general considerations , and it might , as w'ill be shown , have been inferred from them with considerable assurance .
Consequently , as it is now found to hold good in a special case , there are strong grounds for believing it to be a general law of heredity .
I have had great difficulty in obtaining a sufficient amount of suitable evidence for the purpose of verification .
A somewhat extensive series of experiments with moths were carried on , in order to supply it , but they unfortunately failed , partly owing to the diminishing fertility of successive broods and partly to the large disturbing effects of differences in food and environment on different 402 ' Dr. F. Galton .
The average Contribution of each broods and in different places and years .
No statistical results of any consistence or value could be obtained from them .
Latterly , while engaged in planning another extensive experiment with small , fast-breeding mammals , I became acquainted with the existence of a long series of records , preserved by Sir Everett Millais , of the colours during many successive generations of a large pedigree stock of Basset hounds , that he originated some twenty years ago , having purchased ninety-three of them on the Continent , for the purpose .
These records afford the foundation upon which this memoir rests .
The law to be verified may seem at first sight too artificial to be tru6 , but a closer examination shows that prejudice arising from the cursory impression is unfounded .
This subject will be alluded to again , in the meantime the law shall be stated .
It is that the two parents contribute between them on the average one-half , or ( Q-,5 ) of the total heritage of the offspring ; the four grandparents , one-quarter , or ( 05)2 ; the eight great-grandparents , one-eighth , or ( 0'5)3 , and so on .
Thus the sum of the ancestral contributions is expressed by the series { ( 0'5)-|-(0'5)24-(0'5)3 , \amp ; c. } , which , being equal to 1 , accounts for the whole heritage .
The same statement may be put into a different form , in which a parent , grandparent , \amp ; e. , is spoken of without reference to sex , by saying that each parent contributes on an average one-quarter , or ( 0'5)2 , each grandparent one-sixteenth , or ( 0*5)4 , and so on , and that generally the occupier of each ancestral place in the \#187 ; th degree , whatever be the value of n , contributes ( 0'5)2 " of the heritage .
In interbred stock there are always fewer , and usually far fewer , different individuals among the ancestry than ancestral places for them to fill .
A pedigree stock descended from a single couple , m generations back , will have 2m ancestral places of the mth order , but only two individuals to fill them ; therefore if 10 there are 1024 such places ; if m \#151 ; 20 there are more than a million .
Whenever the same individual occupies many places he will be separately rated for each of them .
The neglect of individual prepotencies is justified in a law that avowedly relates to average results ; they must of course be taken into account when applying the general law to individual cases .
No difficulty arises in dealing with characters that are limited by sex , when their equivalents in the opposite sex are known , for instance in the statures of men and women .
The law may be applied either to total values or to deviations , as will be gathered from the following equation .
Let M be the mean value from which all deviations are reckoned , and let Dls D2 , \amp ; c. , be tlLe means of all the deviations , including their signs , of the ancestors in the 1st , 2nd , \amp ; c. , degrees respectively ; then \#163 ; ( M + D , )+KM+D2)+\amp ; c. = M+(|D1-f-fD2+\amp ; c. ) several Ancestor to the total Heritage of the Offspring .
403 It should noted that nothing in this statistical law contradicts the generally accepted view that the chief , if not the sole , line of descent runs from germ to germ and not from person to person .
The person may be accepted on the whole as a fair representative of the germ , and , being so , the statistical laws which apply to the persons would apply to the germs also , though with less precision in individual cases , Now this law is strictly consonant with the observed binary subdivisions of the germ cells , and the concomitant extrusion and loss of one-half of the several contributions from each of the two parents to the germ-cell of the offspring .
The apparent artificiality of the law ceases on these grounds to afford cause for doubt ; its close agreement with physiological phenomena ought to give a prejudice in favour of its truth rather than the contrary .
Again , a wide though limited range of observation assures us that the occupier of each ancestral place may contribute something of his own personal peculiarity , apart from all others , to the heritage of the offspring .
Therefore there is such a thing as au average contribution appropriate to each ancestral place , which admits of statistical valuation , however minute it may be .
It is also well known that the more remote stages of ancestry contribute considerably less than the nearer ones .
Further , it is reasonable to believe that the contributions of parents to children are in the same proportion as those of the grandparents to the parents , of the great-grandparents to the grandparents , and so on ; in short , that their total amount is to be expressed by the sum of the terms in an infinite geometric series diminishing to zero .
Lastly , it is an essential condition that their total amount should be equal to 1 , in order to account for the whole of the heritage .
All these conditions are fulfilled by the series of l -f -|2 + -|3 + \amp ; c. , and by no other .
These and the foregoing considerations were referred to when saying that the law might be inferred with considerable assurance priori ; consequently , being found true in the particular case about to be stated , there is good reason to accept the law in a general sense .
The Bassets are dwarf blood-hounds , of two , and only two , recognised varieties of colour .
Excluding , as I have done , a solitary exception of black and tan , they are either white , with large blotches ranging between red and yellow , or they may in addition be marked with more or less black .
In the former case they are technically known and registered as " lemon and white , " in the latter case as " tricolour .
" Tricolour is , in fact , the introduction of melanism , so 1 shall treat the colours simply as being " tricolour " or " non-tricolour ; " more briefly , as T. or N. I am assured that transitional cases between T. and N. are very rare , and that experts would hardly ever disagree about the class to which any particular hound should be assigned .
A stud-book is published from time to time 404 Dr. F. Gallon .
The average Contribution of each containing the pedigrees , dates of birth , and the names of the breeders of these valuable animals .
The one I have used bears the title ' The Basset Hound Club Rules and Stud-Book , ' compiled by Everett Millais , 1874-1896 .
It contains the names of nearly 100*0 hounds , to which Sir Everett Millais has very obligingly , at my request , appended their colours so far as they have been registered , which during later years has almost invariably been done .
The upshot is that I have had the good fortune to discuss a total of 817 hounds of known colour , all descended from parents of known colour .
In 567 out of these 817 , the colours of all four grandparents were also known .
These two sets are summarised in Table I and discussed in Table V , and they afford the data for Tables II , III , and IV .
In 188 of the above cases the colours of all the eight great-grandparents were known as well ; this third set is discussed in Table VI .
Partly owing to inequality in the numbers of the tricolours and non-tricolours , and partly owing to a selective mating in favour of the former , the different possible combinations of T. and R " .
ancestry are by no means equally common .
The effect of this is conspicuous in Table I , where the entries are huddled together in some parts and absent in others .
Still , though the data are not distributed as evenly aS could be wished , they will serve our purpose if we are justified in grouping them without regard to sex ; or , more generally , if we treat the 2 " components of each several A " , whatever be the value of \#187 ; t , as equally efficient contributors .
Our first inquiry then must be , " Is or is not one sex so markedly prepotent over the other , in transmitting colour , that a disregard of eex would introduce statistical error ?
" In answering this , we should bear in mind a common experience , that statistical questions relating to sex are very difficult to deal with .
Large and unknown disturbing .causes appear commonly to exist , that make data which are seemingly homogeneous , very heterogeneous in reality .
Some of these are undoubtedly present here , especially such as may be due to individual prepotencies combined with close interbreeding .
For although this pedigree stock originated in as many as ninety-three different hounds , presumably more or less distant relations to one another , some of them proved of so much greater value than the rest that very close interbreeding has subsequently been resorted to in numerous instances .
In order to show the danger of trusting blindly to averages of sex , even when the numbers are large , I have compared the results derived from different sets of data , namely from those contained in the last two columns of Table I , where they are distinguished by the letters A and B , and have treated them both separately and together in Table II .
They will be seen to disagree widely , concurring only in showing that the dam is prepotent over the sire in transmitting colour .
According to the A data , their several Ancestor to the total Heritage of the Offspring .
405 relative efficacy in this respect is as 58 to 51 , say 114 to 100 ; according to the B data , it is as 47 to 32 , say 147 to 100 .
Talcing all the data together , it is as 54 to 45 , say 120 to 100 , or as 6 to 5 .
It does not seem to me that this ratio of efficacy of 6 to 5 is sufficient to overhear the statistical advantages of grouping the sexes as if they were equally efficient , the error in one case being more or less balanced by an opposite error in the other .
It is true that the reciprocal forms of mating are by no means equally numerous , the prevailing tendency to use tricolours as sires being conspicuous .
Still , as will he found later , on the application of a general test , the error feared is too insignificant to be observed .
Should , however , a much larger collection of these data be obtained hereafter , minutiae ought to be taken into account which may now be disregarded , and the neglect of female prepotency would cease to be justified .
The law to be verified supposes all the ancestors to be known , or to be known for so many generations back that the effects of the unknown residue are too small for consideration .
The amount of the residual effect , beyond any given generation , is easily determined by the fact that in the series | + \amp ; c. , each term is equal to the sum of all its successors .
Now in the two sets of cases to be dealt with the larger refers to only two generations , therefore as the effect of the second generation is that of the unknown residue is j also .
The smaller set refers to three generations , leaving an unknown residual effect of S\#166 ; .
These large residues cannot be ignored , amounting , as they do , to 25 and 12'5 per cent , respectively .
We have , therefore , to determine fixed and reasonable rules by which they should be apportioned .
The requisite data for doing this are given in Table 111 , which shows that 79 per cent , of the parents of tricolour hounds are tricolour also , and that 56 per cent , of the parents of non-tricolour hounds are tricolour .
It is not to be supposed that the trustworthiness of these results reaches to 1 per cent. , but they are the best available data , so I adopt them .
It will be convenient to use the following nomenclature in calculation :\#151 ; a0 stands for a single member of the offspring .
cii for a single parent ; a2 for a single grandparent , and so on , the suffix denoting the number of the generation .
A parallel nomenclature , using capital letters , is :\#151 ; A0 stands for all the offspring of the same ancestry .
Ajfor the two parents ; A2 for all the four grandparents , and so on .
Consequently Art contains 2 " individuals , each of the form atl , and A " contributes ( 0'5)H to the heritage of each a0 ; while each a " contributes ( O'S)2'1 to it .
In the upper part of Table IV the ratios are entered of the average VOL. lxi .
2 F 406 Dr. F. Gallon .
The average contributions of T. supplied by known ancestors .
Nothing further need be said about these , except that they are styled coefficients because they must be multiplied into the total number of offspring , in order to calculate the number of them that will , on their separate and independent accounts , be probably tricolours .
We have next to explain how the coefficients for the unknown ancestry have been calculated , namely , those which are entered in the lower part of Table IV .
Suppose all the four grandparents , A2 , to be tricolour , then only 079 of A3will be tricolour also , ( 079)2 of A* , and so on .
These several orders of ancestry will respectively contribute an average of tricolour to each cio of the amounts of ( 0-5)3X ( 079 ) , ( 0'5)4 x ( 079)2 , \amp ; c. Consequently the sum of their tricolour contributions is ( 0-5)3x(079 ) { l + ( 0-5 ) x(079 ) + ( 0-5)2x ( 079)2+\amp ; c. } , which equals O'1632 .
The average tricolour contribution from the ancestry of each of the four tricolour grandparents must be reckoned as the quarter of this , namely , 0-0403 .
By a similar process , the average tricolour contribution from the ancestry of each non-tricolour grandparent is found to be 0-0243 .
When the furthest known generation is that of the great-grandparents , the formula differs from the foregoing only by substituting ( 0'5)4X ( 079 ) for ( 0-5)3X ( 0-79 ) .
This makes the average tricolour contribution from the ancestry of the whole eight tricolour great-grandparents equal to 0'08160 , and that from the ancestry of each of them to be one-eighth of this , or 0'0102 .
In a similar way the tricolour contribution from the ancestry of each non-tricolour great-grandparent is found to be 0-0061 .
The following example shows how the coefficients in Table IY were utilised in calculating the general coefficients entered in Table Y. 2 Parents , Ti ( personal ) ... ... ... . .
0-5000 3 Grandparents , T2 ( personal ) ... ... 0-1875 1 Grandparent , N2 ( personal ) ... ... .
3 Grandparents , T2 ( ancestral ) ... ... .
0-1224 1 Grandparent , N2 ( ancestral ) ... ... .
0-0243 Total tricolour contribution . .
0'8342 The coefficient 0'83 will consequently be found under the appropriate head in Table Y , where the total number of offspring ( " all cases " ) is recorded as 119 .
By multiplying these together , viz. , 0-83 x 119 , the " calculated " number of 99 is obtained .
It will be seen that the observed number was 101 , a difference of only 2 per cent. , The extraordinarily close epincidence throughout the two tables , several Ancestor to the total Heritage of the Offspring .
407 V and VI , between calculation and observation , proves that the law is correct in the present instance , and that the principle by which the unknown ancestry was apportioned , is practically exact also .
It is not so strictly exact as it might have been , because the whole of the available knowledge has not been utilised .
The 0'79 applied to A4 , \amp ; c. , requires some suiall correction according to the known colours of the offspring of A3 .
If they had been alt tricolour the 0'79 would have to be increased ; if all non-tricolonr , it would have to be diminished .
Having insufficient data to check a theoretical emendation , I note its omission , but shall not discuss the matter further .
It will be easily understood from these remarks how collateral data are to be brought into calculation , for if the collaterals were more tricolour than the average of hounds , the 079 would have to be somewhat increased ( but not beyond the limiting value of 1'00 ) ; if less tricolour than the average , the 0'79 would have to be diminished .
The knowledge of collaterals would be superfluous , if that of the direct ancestry were complete , but this important prolongation of the present subject must not be considered further on this occasion .
There are three stages in Tables V and VI at which comparisons may be made between calculated results and observed facts .
( 1 ) The Grand Totals.\#151 ; In Table V the sum of all the calculated values amounts to 391 ; that of all the observed ones to 387 , which are closely alike .
In Table VI they amount to 180 and 181 respectively , which is a still closer resemblance .
Consequently the calculations are practically exact on the ivhole , and the error occasioned by neglect of sex , \amp ; c. , is insignificant .
( 2 ) The Subordinate Pairs of Totals.\#151 ; These are entered at the sides of the tables , and are nine in number , namely , 236 , 239 ; 149 , 139 ; 6 , 9 ; 53 , 56 ; 52 , 56 : 9 , 9 ; 8 , 6 ; 49 , 46 ; 9 , 8 .
The coincidences are striking , in comparison with such results as statisticians have usually to be contented with ; the second pair , 149 , 139 , is the least good , and will be considered in the next paragraph .
( 3 ) Individual Pairs of Entries.\#151 ; There are 32 of these ; here also calculation compares excellently well with observation , excepting in the line that furnishes the " subordinate totals " of 149 , 139 , where the " all cases " of 37 , 158 , 60 yield the tricolour contingents of 20 , 79 , 36 .
Dividing each tricolour by the corresponding " all cases , " we obtain what may be called " Coefficients from Observation , " to compare with the calculated coefficients .
They are as follows:\#151 ; Diff. Diff. Coefficients from observation ... .
54 I ( \#151 ; 4 ) I 50 !
( + 10 ) 60 " , , calculation ... .
66 I ( \#151 ; 8 ) | 58 I ( \#151 ; 7 ) 51 The great irregularity of the entries in the upper line shows that the observed values cannot be accepted as true representa2 f 2 408 Dr. F. Galton .
The average Contribution of each tives of the normal condition .
I have not unravelled the causes of this error , and it is not urgent to do so , since its ill effects are swamped by the large number of successes elsewhere .
In order to satisfy myself that the correspondence between calculated and observed values was a sharp test of the correctness of the coefficients , I made many experiments by altering them slightly , and recalculating .
In every case there was a notable diminution in the accuracy of the results .
The test that the theory has successfully undergone appeared on that account , to be even more searching and severe than 1 had anticipated .
It is hardly necessary to insist on the importance of possessing a correct law of heredity .
Yast sums of money are spent in rearing pedigree stock of the most varied kinds , as horses , cattle , sheep , pigs , dogs , and other animals , besides flowers and fruits .
The current views of breeders and horticulturists on heredity are contradictory in important respects , and therefore must be more or less erroneous .
Certainly no popular view at all resembles that which is justified by the present memoir .
A correct law of heredity would also be of service in discussing actuarial problems relating to hereditary longevity and disease , and it might throw light on many questions connected with the theory of evolution .
-Pedigrees of Parental and Grand Parental Colours ( Tricolour ( T ) or .Non-Tiicolour The Sex of the Ancestors is taken into Account , but not that of the Offspring .
several Ancestor to the total Heritage of the Offspring .
409 Others of which all a. p. 's are not known .
B j 87 20 15 17 30 64 11 6 s CM | Totals . ?
J 05 00 LOO0 Tft 05 05 lO CO ^5 54 ri i\#151 ; i O rH 1 CM rH rH jg ^ 1 : | o : : : : : : : : j : |Zh^ | S j . .
riH . .
rH* CO J S . .
HH H H . .
HP i*\gt ; i | T ?
CO .
rf\#171 ; CO rH rH .
| 2 ; 00 t \#163 ; \#163 ; HH CO .
j \#163 ; Eh h !
2 ; . . . .
CM 00 LO CO JJ BB j .\lt ; ?\#187 ; | CO .
CO CM . . . .
00 j * !
4 1 ! . .
4 05 E-t !
25 | | ^ 1 CM tH H CO HP rH . .
a HH j 's I 1\gt ; rH . . . . .
a j E-i H | \lt ; L\gt ; 12 2 11 5 S | EH\amp ; | EH ^ Cl CM CO rH O 05 Cl X \amp ; | HH 4^ CM O CO t ?
CO . .
Tfl rH rH rH . .
101 EH H I I H \#163 ; \#171 ; o | g\#171 ; : : $S : : QO CO rH HB EH EH e 10G 13 20 17 CO lO rH Offspring .
Try Non-T Try IS'on-T Try Non-T Try Non-T j Totals .
i : EH H i i t i \#151 ; i | Sire .
4 4 |1 410 Dr. F. Galton .
The average Contribution of each Table II \#151 ; Offspring of one parent Tricolour ( T ) and of one Non-tricolour ( N ) .
The sex of the parents is not regarded .
| From Tabled Observed .
Per cents .
Tricolour .
Non- tricolour .
Tricolour .
Non- tricolour .
Sire T , dam N A 114 109 223 51 49 B 30 64 94 32 68 Sum 144 173 317 45 55 Dam T , sire If A 25 18 43 58 42 B 15 17 32 47 53 Suin 40 35- \#166 ; 75 - | 54 1 : j 46 .
Table III.\#151 ; Distribution of T and N colour in Parents , when the Offspring are T and N respectively .
From Table I. ^ * C I _ TT No. of T offspring .
J Parents* of T offspring .
Sires .
Dams .
A ' J B. Total .
T. l N. T T 239 87 326 652 0 T N 25 15 40 40 40 JT T 114 30 144 144 144 N N 9 11 20 0 40 Total s t , r ... ... ... .
530 836 224 Per cent , of parents* ... .
... .
50 79 21 From Table I. No. of N offspring .
Parents* of N offspring .
Sires .
, Dams .
A. : \#166 ; .
* , " * .
B. Total .
T. N. T i i T 38 20 58 116 0 T , N 18 17 35 35 35 St : T 109 64 173 173 173 N N 15 6 .
21 0 42 Totals \#171 ; mi 287 1 324 250 Per cent , of Parents* 50 56 44 * More properly " Parental Places " ; the number of these , though not that of the Individual parents , being always double the number of any group of offspring .
several Ancestor to the total Heritage of the .
411 Table IV.\#151 ; Tricolour coefficients .
Personal allowance of T for each Tricolour parent .
" grandparent * " great-grandparent ( No allowance for Non-tricolours .
) Ancestry known i Grandparents .
ip to and inclusive of Great-grandparents .
0*2500 0*0625 0-2500 0 0625 0-0156 Ancestral allowance of T for each Tricolour grandparent .
0-0408 \#151 ; Non-tricolour " 0 0243 !
| Tricolour great-grandparent \#151 ; 0 -0102 Non-tricolour " " 0 0061 Table V.\#151 ; Calculation and Observation Compared .
The pedigrees are utilised up to the second ascending generation .
Sex not taken into account .
No. of tricolours in parents .
Number of tricolours in grand- I parents .
Total tricolour offspring .
4 1 1 3 2 1 Calculated .
Observed .
a beei dfffjkm hlno All cases 119 119 28 n f Coefficient ... ... 0-91 0 83 0-76 0-68 2 \lt ; t Tricolour calc 'd .
10$ 99 21 s 236 L " observed 106 101 24 8 t \ 239 r All cases 37 158 60 6 f Coefficient 0-66 0-58 0-51 0-43 1 \lt ; 1 Tricolour calc 'd .
n 92 30 3 149 '\#166 ; L " observed 20 79 36 4 139 r All cases 18 6 o i Coefficient . . . .
0-26 0-18 1 Tricolour calc 'd .
5 1 6 * " observed ** i 7 2 9 Grand totals . . .
r r .nroV-r\lt ; Lj i 391 ^0 t\ 387 ; 412 Contribution of each Ancestor to the Heritage of Offspring .
Table VI.\#151 ; Calculation and Observation Compared .
The pedigrees are utilised up to the third ascending generation .
Sex not taken into account .
No. of tricolours Number of tricolours in great-grandparents , Total tricolour offspring .
in parents .
in grand- parents .
8 7 6 5 4 Calcu- lated .
Observed All cases 2 25 14 16 Coefficient 0-96 0*94 0-92 0-90 r 4 J Tricolours calc 'd .
2 H 13 u ... 53 L , , observed 2 25 14 .
15 .* 56 All cases ... ... 18 21 16 6 \ Coefficient 0*87 0-85 0-83 0-81 O J 3\lt ; Tricolours calc 'd .
16 18 13 5 52 " observed 17 19 14 6 56 r All cases ... ... 3 2 3 3 f Coefficient ... ... 0*81 0*79 0-77 0-75 2 J 1 Tricolours ealc 'd .
2 3 2 2 9 " observed 2 2 3 2 9 r All cases 2 i 9 f Coefficient 0-69 0-67 0-65 4 J _____ 1 Tricolours calc 'd .
1 1 6 S J L " observed 1 . .
5 . .
6 r All cases ... ... 1 28 14 31 9 " | 3J Coefficient 0-64 0*62 0-60 0-58 0 *56 _ 1 Tricolours calc 'd .
1 n 6 ?
18 5 49 L " observed .
1 16 12 \lt ; 8 9 ' 46 All cases . .
4 13 2 f Coefficient . .
0-54 0-52 1 ' 1 Tricolours calc 'd .
2 7 9 " observed 1 !
* .
1 7 8 Grand totals . .
180 1181 1 1 The summed data derived from Table IY , form the coefficients entered in Tables Y and VI .
These are multiplied into the corresponding number of w all cases , " and the result gives the " calculated number of tricolour hounds among them .
Influence of ci Magnetic Field on Frequency .
413 The entries of " all cases " and of " tricolours observed " in Table V are deduced from Table I , by combining the appropriate columns .
The letters at the top show which columns are combined .
Seven other observed cases , disposed in three groups , are scattered beyond the limits of Table VI ; two of these seven cases are tricolour .
\#171 ; Further Note on the Influence of a Magnetic Field on Radiation Frequency .
" By Oliver Lodge , F.R.S. , assisted by Mr. Benjamin Davies .
Received May 19 , \#151 ; Read June 3 , 1897 .
Referring to a former communication of mine , on the subject of Zeeman 's discovery , printed on page 513 of the ' Proceedings of the Royal Society ' for February 11 this jear , vol. 60 , No. 367 , I wish to add an observation to those previously recorded , as I have recently acquii*ed a concave Rowland grating ( 3|- X H-inch ruled surface , 14,438 lines to inch , 10 feet radius of curvature , being the one used by Mr. George Higgs ) , of which the spectra of the first and third orders on one side are very satisfactory .
It is said on page 513 , " If the focussing is sharp enough to show a narrow , dark reversal line down the middle of each sodium line , that dark line completely disappears when the magnet is excited .
" ' With the greater optical power now available the dark reversal line is often by no means narrow , and though in some positions of the flame it does still tend to disappear or become less manifest when the flame is subjected to a concentrated magnetic field , the reason of its partial disappearance is that it is partially reversed again\#151 ; i.e. , that a third bright line , as it were , makes its appearance in the midst of the dark line , giving a triple appearance to each sodium line .
More completely stated the phenomena are as follows:\#151 ; After* obtaining each sodium line with a prominently double aspect by manipulating the flame , the magnet is excited , and the dark band in the midst of each sodium line is then seen to widen out considerably in the region of most intense magnetisation , while a bright intrusion line makes its appearance .
On closer examination this new line is seen to be double , by reason of a dark division down its middle ; and I apprehend that with still more magnetic power this dark band might itself open out into two ; but this last phenomenon I have not yet observed .
The whole sodium group is thus seen as if it were octuple .
The effect is not due to a mere mechanical disturbance or rearrangement of the gases of the flame by the agency of magnetism ; because a

