IL Variation and Correlatioii in Lesser Celandine löl 



The .slight cxccption t(i this rulo — Suiroy corrolation of pctals and stainons — is 

 well withiii Ihü probable error of thc deteruiinations. 



SLiiiiining up thcsc rcsults in a singio Statement, we conelude froni llie data 

 that were gathcred and eouiited on one uniform plan that : 



The lücal races of the lesser celandine which have flowers with rnore nmaerous 

 purts, have those jmrts laore variable and nwre highli/ correlated. 



Lot US compare thesc results with the data obtained by othcr observers. 

 Oermaiiy contributes nothing, for stamens and pistils have not yet bcen couiited. 

 The Swiss data, if in(li\idually a little anomalous, form, if clubbcd together, a 

 striking contirmation of our argument, — they have fewer pistils and stamens thaii 

 our four scries, the variations are less, and thc correlations for stamens and pistils 

 are low — Gais has thc lowest result of all, and Trogen oidy about the Guernsey 

 value. Unfortunatcly we can only compare one corrolation and this has been 

 workod out for only small numbers, 285 in one case and SO in the other. With 

 rcgard to Belgium the comparative results are not so satisfactory. Thc early 

 Belgium flowers have about the same number of stamens as the Italian and fewer 

 pistils, the variability of both stamens and pistils is Icss. Thc lato Belgium 

 flowers have the least number of stamens and pistils of any series, their variability 

 in both stamens and pistils is less than all four of our series, and less than the 

 mean of the two Swiss scries. Here again wc have unfortunatcly only one corro- 

 lation availablc for eomparison, that of stamens and pistils. It is for the early 

 flowers somewhat greatcr than that for thc Italian sorios, when it ought to be 

 about equal to it, while for the late flowers, it is thc highest of all when it should 

 be the Icast from thc above rule. Bat just in this material a new factor conies 

 into play, which, if not the source of the result, deserves to be taken into account. 

 The late flowers are said to be from the same plants as the early flowers. But 

 pistils and stamens cannot be counted unless the flower be gathered. Hence the 

 plants from which the later flowers are gathered may be heterogeneous in character, 

 namely plants naturally floweriug late, and plants which have been more or less 

 injured by one or more flowers being rcmovcd and which have reflowered in 

 a more or less exhausted condition. Such heterogeneity would tend to increase 

 correlation. Clearly in this as in so many easos the conditions are very complex, 

 and we cannot without further investigation club together flowers from plants 

 (a) which have had earlier flowers gathered, (b) which naturally flower late, with 

 the late flowers of plants which have been flowering throughout the season 

 undisturbed, and tenn thc group ' late flowers ' without danger of introducing 

 heterogeneity. Till therefore the whole subject of the corrolation of late and 

 early flowers has been goue into carefully with aniple data, we may take, we hold, 

 the above geueral principle — that increase in the number of parts denotes increase 

 in the variability and correlation of those parts — as a working hypothesis to 

 be demonstrated or modifled by further researches. We, of course, do not 

 extend thc principle beyond the species for which the data are deduced, even as a 

 " w'orkiug hypothesis," and wc state it oidy for pctals, stamens and pistils. 



