January 19, 1906.] 



SCIENCE. 



87 



sample of what is permissible might obvi- 

 ously be carried on forever. In the great 

 majority of cases this type of classification 

 is the work of authors who have quite lost 

 sight of the fact that it is not the magni- 

 tude but the constancy of the differences 

 which is of real significance in classifica- 

 tion. There is, for instance, no more dan- 

 gerous theory than that because a particu- 

 lar trait, let us say the presence of a corona, 

 forms an excellent generic distinction in 

 one family it must necessarily be of generic 

 significance in another. 



The frequent occurrence in botanical lit- 

 erature of such expressions as 'a generic 

 difference,' 'distinctions of generic rank,' 

 etc., seems to indicate a more or less wide- 

 spread feeling that differences of a par- 

 ticular kind or magnitude, or relating to 

 special plant-structures, are in some way 

 recognizable as diagnostic characteristics 

 of generic value as opposed to those which 

 can be used merely for the separation of 

 species or varieties. There is a common 

 idea, for instance, that a difference of floral 

 structure, or especially one of fruit or seed, 

 is almost infallibly trustworthy in the sepa- 

 ration of genera. This belief has certainly 

 the justification that distinctions in the 

 essential parts of the flower or in the fruit 

 are much more apt to be fixed, are at least 

 far more constant, than those of habit, 

 foliage or pubescence, since the latter ap- 

 pear much more subject to modification 

 with varying environment. Nevertheless, 

 this theory, as a rule of classification, may 

 easily be carried too far. When carefully 

 examined, very few differences in the struc- 

 ture of flower or fruit will be found to 

 be invariably of generic value. Nearly 

 all appear at some point in the plant 

 system to be in a fluctuating state. For 

 example, the position of the cotyledons, 

 although possessing a high value in sepa- 

 rating many cruciferous genera, com'pletely 

 breaks down in the genus Lepidium, where 



in plants of the closest habital similarity 

 aecumbent and incumbent cotyledons are 

 found. In the Rosacece the unlike adnation 

 and connation of the floral parts furnish in 

 some portions of the great family excellent 

 generic and even tribal distinctions. In 

 other rosaceous genera, however, the same 

 diversity of adnation and connation occurs 

 in species which are obviously of such close 

 relationship and so connected by interme- 

 diates that their generic separation is by no 

 means convincing. Many other instances 

 of this kind might be cited to show that a 

 difference having great classificatory sig- 

 nificance in one place may be almost value- 

 less in another. 



It is fair to inquire how we know this. 

 If, for example, genera of the Cruciferas 

 can be readily separated by the position 

 of the cotyledons, why is it not after all 

 the most logical course to treat Lepidium 

 virginicum, with its aecumbent cotyledons, 

 as constituting a genus distinct from the 

 other species of Lepidium in which the 

 cotyledons are incumbent? There are two 

 pretty obvious reasons against this. One 

 is that in Lepidium there are other species 

 which have transitional cotyledons, exhibit- 

 ing various degrees of obliquity. In the 

 second place, the aecumbent cotyledons of 

 Lepidium virginicum, although a striking 

 character, are unaccompanied by any other 

 difference of moment. From this fact we 

 may reasonably infer that in this particular 

 group the difference in the position of the 

 cotyledons is of relatively recent origin, 

 for it has not had time to become correlated 

 with any other trait of significance or con- 

 stancy. This brings us to a" matter of great 

 practical as well as theoretical importance 

 in classification, namely, that few, if any, 

 genera carry conviction as natural groups, 

 or, to be more precise, naturally delimit- 

 able groups, unless they can be separated 

 by more than one feature. The ideal genus 

 is certainly oiie in whicV several' di^lh- 



