arrangement. It may indeed be objected that the evil would have been equally avoided by all four anthers being placed in one horizontal line. Of two perfectly equal arrangements, nature, however, can but select one. But, in fact, the horizontal position would not answer all the requirements. Let any one watch a bee visiting a woodsage. He will see that the anthers strike the top of its head just between the eyes, and smudge the whole intervening space with pollen. Had the two lower anthers been placed in a horizontal line with the upper ones, their pollen would have been shed on the eyes of the insect, and its visits discouraged. The pollen, moreover-and this is the point of chief importance-would have adhered to parts which are quite external to the range of a centrally placed stigma.

If the explanation I have just given be the true one, it should apply not merely to woodsage, but to all those numerous other flowers-labiates, figworts, and the rest-which have a similar arrangement of their stamens. Thus all such didynamous plants must be supposed to require the agency of insects for their due fertilisation. Now that this is really the case is, I think, highly probable. Why otherwise are they, as a rule, furnished with sweet secretions and aromatic odours which serve to attract insects? Why are their four stamens so arranged against the walls of the corolla as to leave a free and open access to the nectary? Why are all the four anthers and the style placed on the same side of the corolla, and in the median line? Why is the dehiscent surface turned towards the path which leads to the nectary and away from the stigma? Why does this latter so often become mature at a later period than the anthers? Why are there only four stamens, even when the corolla has five divisions, and why is the missing stamen invariably the posterior one? Why, lastly, do we so frequently find some or other arrangement in these didynamous flowers, which seems directly calculated to render self-fertilisation a matter, to say the least, of difficulty? All these questions admit of easy answers, on the hypothesis that didynamia are fertilised by insects, and are unanswerable on any other hypothesis.

Let any one, for instance, examine a spike of foxglove, beginning with the upper or less mature flowers, and going gradually downwards to the more mature. He will find, first, flowers in which neither anthers nor stigma are ripe. Then come flowers in which the upper pair of anthers are ripe, the lower pair and stigma still immature. Then all four anthers are found ripe, the stigma still remaining as before. Then, lastly, come flowers in which the bifid stigma is open and bent forwards, but in which the two upper anthers have discharged all their pollen, while the two lower ones have not yet exhausted their store. Now, why is it that the upper anthers are thus before the lower

ones in their development? It cannot, I think, be for any other reason than that they are in closer proximity; in fact, close against the stigma, and that there would be great risk of selffertilisation were they not to discharge their pollen before the viscid surface was mature; whereas the lower anthers lie at a distance, and are much less dangerous. Were this not the case, there is no reason why the upper should precede the lower. In fact, when a useful purpose can be served, we find the lower pair preceding the upper. This is the case with woundwort (Stachys). Here the lower pair ripen first, shed their pollen, and then retire laterally, one to the right, the other to the left. This they do in order to make room for the style to bend forwards, and come into proper position.

In examining the successive blossoms of a spike of foxglove, another fact will be noticed, which, as well as the general didynamous structure, testifies to the importance of a vertical arrangement of the anthers. Each of these has two large lobes, which in the immature blossoms are placed horizontally, so as to cover a considerable breadth of the corolla. Were they to ripen in this position, most of their pollen would be wasted, for it would adhere to the sides of the humble bee, in parts entirely out of reach of the centrally placed stigmas. To avoid this, the lobes, as they ripen, change their direction and become vertical, lying close to the median line. (These successive changes are shown in figs. 5, 6, 7.) Their range is thus made to coincide with that of the stigma; that is to say, a bee entering a freshly opened flower will have its back daubed with pollen in the same median line as will come into contact with the stigma when it visits a more mature blossom.

The imperative necessity of keeping the anthers in the same line with the stigma may perhaps explain the anomalous structure of the stamens in Prunella. In this flower, curious buttresses project from the outer side of the filaments against the inner surface of the wide hood. These serve, I imagine, to fix the anthers more securely in the median line and to prevent any lateral divergence.

This

There remains yet another point of interest. Why are there only four stamens in these didynamous flowers, even when calyx and corolla have five divisions? and why is the missing stamen invariably the posterior one? The reason is obvious. stamen, were it present, would occupy the position which is wanted for the style. It is necessary that this shall lie against the wall of the corolla, so as to leave a free passage for insects. It is necessary that it shall lie against the posterior wall and in the median line, or its range will not coincide with that of the anthers, and the stigma will not strike the part of the bee on which the pollen is smeared. In fact, there is no other position

but that of the missing stamen in which the style can be placed.

The best thing, therefore, that can happen is that this stamen shall not be developed at all; and this is what does occur in didynamous flowers. The next best thing is that, when developed, it shall get out of the way and make place for the style. This is what occurs in Pentstemon. Here four of the stamens and the style are arranged as in ordinary didynamous flowers. The fifth stamen, however, is present, but makes place for the style, by running from its insertion in the posterior median line right across the tube to the opposite side of the corolla. It is less in the way here than if it occupied its natural position. Still it is perfectly useless, and in some degree an obstruction. It is not therefore to be wondered at that, as a rule, it produces no pollen, and that not very rarely it is altogether absent. The Pentstemon would thus appear to be on its way to become didynamous.

Doubtless there are not a few didynamous flowers the structure of which may seem to be more or less in contradiction with the preceding remarks. The apparent contradictions are, however, I think, always capable of explanation. Some of these exceptions to the ordinary arrangement I will now consider.

Scrofularia.-In this genus the posterior stamen is still the missing one, although the style and other stamens are ranged against the anterior wall. The absence of the fifth stamen cannot therefore in this case be explained by its place being required for the style. It can, however, be accounted for on another ground. The stamen would be perfectly useless, for its anther would strike an insect on the back, while the stigma only comes into contact with the under surface. The anther is absent therefore on grounds of economy. But as its entire disappearance is not required to make room for the style, it is not surprising to find that it is not so completely absent as in other didynamia, but is represented by a scaly staminode of considerable size.

Gesneria.-Here we have a corolla in general form like that of a foxglove. The style also with its stigma occupies a like position as in that flower; that is, it runs in close contact with the posterior wall and in the median line, occupying the place of the absent stamen. The anthers of the four remaining stamens, instead of being arranged in pairs longitudinally, as in foxglove and most didynamia, lie in a horizontal line immediately in front of the stigma. In the case of Teucrium, I said that such an arrangement would not answer the requirements, because the two external anthers would smear their pollen on parts of the bee which would be quite out of

range of the centrally set stigma. In Gesneria, however, this difficulty is overcome. In the first place, the stigma is very broad, so as to cover a large range.

Then the anthers are

individually very narrow, and, moreover, adhere to each other by their edges, so that the four together form a disk no broader than the stigma, immediately in front of which it lies (figs. 9, a, and 10). The dehiscence of these coherent anthers is, as usual, on the anterior face, that is, on the side turned away from the stigma. Thus, a bee entering the corolla soon after its expansion will get a smudge of pollen on the median line of its back. It will not touch the stigma, for this is shielded by the anther-disk. But when the flower has been some time open, a change occurs in the position of the parts. The filaments bend forwards, and carry the disk away from the stigma right across the corolla, until it comes into contact with the anterior lip, where it remains fixed (fig. 9, b). It is plain that when a bee visits a flower in this later condition, the same spot on its back, which in the earlier stage came in contact with the anther-disk, will now come in contact with the stigma: for this occupies the same position as did the disk, and is, moreover, of the same breadth. Thus in Gesneria-as in so many other cases the more mature flowers are fertilised by the pollen of the less mature.

Antirrhinum.-The general arrangement of style and stamens in snapdragon is the same as in foxglove; but the persistent closure of the mouth of the corolla might seem a certain proof that the fertilisation is independent of insects. Any one, however, who watches the flowers with a little patience will soon see that the closure is not sufficient to exclude bees. Sometimes a bee will be seen trying in vain to force an entrance; but in such case the flower is, I believe, invariably immature. So soon as the anthers are open, the tightness of the closure relaxes sufficiently to allow the bees to force their way in without any great exertion, and I have repeatedly seen them do so.

That such visits are required for due fertilisation I have, moreover, found on experiment. I covered a large Antirrhinum with a tent of gauze, so arranged as to exclude bees. The plant flowered abundantly; but though in other Antirrhinums close by, which were not so protected, scarcely a single flower failed to be fertilised, only two small capsules were produced from all the numerous flowers of the protected plant; and even of these two it was doubtful whether the fertilisation was not due to an accidental rent in the gauze that occurred towards the end of the experiment, and was not immediately mended; so that possibly a bee may have got in at that period.

I was much struck in this and a few similar experiments by