II. Variation and Gorrelation in Lesser Celandiae lö3 



These coiiclusions show iis how cxtroinely cautious we rnust be in extending 

 conclusions deduced by intraraci.il calculatioiis to interracial relations. Intra- 

 racial and interracial variabilitics and correlations may bc (if a tofally diff'ereiit 

 Order. By the usual theory of regression wo havc the foliowing foniiiila' in find 

 tlie probable nuniber of staniens in a race given the inean niiniber of pistils, and 

 the probable luiiiibei- of pistils given the niean number of stamens: 



Probable nuan nuinber of .stamens in a] =-9939 x (observed niean iiinnhcr of 

 local race J pistils) + 7-185. 



Probable mean uumber of pistils in a| = '9100 x (observed mean number of 

 local race j stamens) -4-697. 



Comparing with actual values we have : 



Stamens from Pistils Pistils from Stamens 



The maximum error in the number of stamens is 30 and the mean error 13; 

 for the number of pistils it is 3-1 and the mean error again 1-3. Thus our 

 formulffi will give the number of stamens from pistils, or pistils fiom stamens 

 in any race with an average error of about one pistil or stamen, and a maximum 

 of about three. They are thus close enough for most practical purposes. The 

 diagram gives the lines to read stamens from pistils and pistils from stamens. 



(7) There are a number of other points which are suggested by our table, 

 but it seems better to defer their discussiou until more ample data are forth- 

 coming. There is hardly any district from which at present a series of 500 — 600 

 celandine heads would not be of value. But these series ought to be gathered on 

 a uniform plan, i.e. gathered as buds, each bud being wrapped in a separate soft 

 paper cover. Further since the labour of counting is very great, those who will 

 count as well as gather are the more valuable lielpers. In counting great care 

 rnust be taken to separate the individual pistils, and all heads with less than 

 three sepals ought to be carefully exaniined with a microscope or powerfui lens. 

 If possible two series should be taken from each locality, one early and one late in 

 the seasou, but the tiowers should not be taken otf the same plants, but off plants 

 having substantially the same euvironment. A record of date, environment and 



Biometrika ii 20 



