METHODS OF RESEARCH IN FLOWER POLLINATION 195 
species of flowers by biting them through. Bombus terrester also frequently plays 
the part of a nectar-thief (cf. pp. 116-17). 
As already mentioned (see p. 66) Loew has proposed a classification of 
flowers corresponding to the allotropous, hemitropous, and eutropous groups 
of insects. Under A//otropous Flowers he places the classes An, Po, E, and EC— 
under Hemiztropous Flowers the classes C and S—and under Lusropous Flowers 
the bee flowers (Hb), humble-bee flowers (Hh), and lepidopterid flowers (L). 
The statistical investigations in flower pollination made by Loew, as well as by 
MacLeod, Heinsius, and myself, show that there actually is a marked agreement 
between the corresponding groups of flowers and insects. But there is still a need 
of more numerous and more thorough special investigations to make quite clear the 
relations between the two sets of groups. 
The four groups of flower guests established by Loew are connected with one 
another by intermediate stages and transitional forms. This classification leaves 
untouched those theoretical speculations which have reference to the genetic develop- 
ment of the various insects. 
X. Methods of Research in Flower Pollination. 
The statistical method introduced by Hermann Miller to determine the 
reciprocal dependence between flowers and insects, and more especially employed 
by him in his ‘Alpenblumen’ (pp. 477-525), does not count the individual visits of 
insects to a species of flower, but only the number of insect species which seek out 
a particular kind of plant. At first sight this method seems unreliable, if not quite 
worthless. As a matter of fact it cannot be denied that a disadvantage is involved, 
though it is one that can hardly be avoided, for it is almost impossible to count all 
the individual visits that a conspicuous flower receives during a considerable period 
of observation when the weather is fine, or to establish how many flowers are 
pollinated by a species of insect in a given time’. 
The value of a method must, however, be judged by its results. Those 
attained by this method, as employed first by Hermann Miiller and afterwards by 
E. Loew, J. MacLeod, and myself, show that (to quote from Loew, ‘ Blumenbesuch,’ 
II, p. 147) it has greater possibilities than any one possessing only a superficial 
knowledge of it would be inclined to believe. In particular the reproach that it must 
afford an inaccurate idea of the number of pollinators, because it counts the visits of 
species and not of individuals, is of no importance. This appears from the agree- 
ment which exists between the determinations of Miiller and those of all subsequent 
observers regarding the numerical relation of visits—not the absolute values, of 
course, but their proportions in a series—and many of the sets of observations were 
made quite independently of one another. 
1 Professor F. Dahl, of Kiel, informs me that he has constructed an apparatus that automatically 
catches all the visitors of a flower or an inflorescence. This would make it possible not only to 
ascertain the number of insect species that visit flowers, but also the number of individuals visiting a 
flower within a given time. Meanwhile Prof. Dahl will make a communication with regard to this 
apparatus at the meeting of the Deutsche Zoologische Gesellschaft (1898) at Heidelberg. The trap 
described and figured by him in his memoir (‘ Vergleichende Untersuchungen tiber die Lebensweise 
wirbelloser Aasfresser,’ SitzBer. Ak. Wiss., Berlin, 1896) is made on a similar principle. 
02 
