oue in Middlesex-and spreading the observations over some time, so as to experiment with different bees. and under varied circumstances. Adding the numbers together, it of course follows that the greater the preference shown for each colour the lower will be the number standing against it.

The following table gives the first day's observations in extenso:

In the next series of experiments the bees had been trained for three weeks to come to a particular spot on a large lawn, by placing from time to time honey on a piece of plain glass. This naturally gave the plain glass an advantage; nevertheless, as will be seen, the blue still retained its pre-eminence. It seems hardly necessary to give the observations in detail. The following table shows the general result

The precautions taken seem to me to have placed the colours on an equal footing; while the number of experiments appears sufficient to give a fair average. It will be observed also that the different series agree well among themselves. The difference between the numbers is certainly striking. Adding together 1, 2, 3, 4, 5, 6, and 7, we get 28 as the total number given by each journey; 100 journeys therefore give, as the table shows, a total of 2,800, which divided by 7 would of course, if no preference were shown, give 400 for each colour. The numbers given, however, are-for the blue only 275, for the white 349, yellow 405, red 413, green 427, orange 440, and plain glass as many as 491.

Another mode of testing the result is to take the per-centage in which the bees went respectively to each colour first, second, third, and so on. It will be observed, for instance, that out of a hundred rounds the bees took blue as one of the first three in 74 cases,

and one of the last four only in 26 cases; while, on the contrary, they selected the plain as one of the first three only in 25 cases, and one of the last four in 75

I

100 100 100 100

may add that I was by no means prepared for this result. Müller, in his remarkable volume on Alpine Flowers, states that bees are much more attracted by yellow than by white.' In the same work he gives the following table :

This table does not indeed show any absolute pre

ference for one colour rather than another. In the first place, the number of species compared is very different in the case of the different colours; and in

'Alpenblumen, p. 487.

the second place, the results may of course be due to the taste, quantity, or accessibility of the honey (all of which we know exercise a great influence), rather than by the colour of the flower. Still the table rather seemed to indicate that bees preferred red, white, and yellow, to blue.

I may very likely be asked, if blue is the favourite colour of bees, and if bees have had so much to do with the origin of flowers, how is it that there are so few blue ones? I believe the explanation to be that all blue flowers have descended from ancestors in which the flowers were green; or, to speak more precisely, in which the leaves immediately surrounding the stamens and pistil were green; and that they have passed through stages of white or yellow, and generally red, before becoming blue. That all flowers were originally green and inconspicuous, as those of so many plants are still, has, I think, been shown by recent researches, especially those of Darwin, Müller, and Hildebrand.

But what are the considerations which seem to justify us in concluding that blue flowers were formerly yellow or white? Let us consider some of the orders in which blue flowers occur with others of different colours.

For instance, in the Ranunculaceae,' those with simple open flowers, such as the buttercups and Thalic

'I take most of the following facts from Müller's admirable work on Alpine Flowers.

trums, are generally yellow or white. The blue delphiniums and aconites are highly specialised, abnormal forms, and doubtless, therefore, of more recent origin. Among the Caryophyllacea the red and purplish species are amongst those with highly specialised flowers, such as Dianthus and Saponaria, while the simple open flowers, which more nearly represent the ancestral type, such as Stellaria, Cerastium, &c., are yellow and white.

Take, again, the Primulacea. The open-flowered, honeyless species, such as Lysimachia and Trientalis, are generally white or yellow; while red, purple, and blue occur principally in the highly specialised species with tubular flowers. The genus Anagallis here, however, certainly forms an exception.

Among the violets we find some yellow, some blue species, and Müller considers that the yellow is the original colour. Viola biflora, a small, comparatively little specialised fly-flower, is yellow; while the large, long-spurred V. calcarata, specially adapted to humblebees, is blue. In V. tricolor, again, the smaller varieties are whitish-yellow; the larger and more highly developed, blue. Myosotis versicolor we know is first yellow and then blue; and, according to Müller, one variety of V. tricolor alpestris is yellow when it first opens, and gradually becoines more and more blue In this case the individual flower repeats the phases which in past times the ancestors have passed through.

The only other family I will mention is that of the