deposition in the bud, I have never yet been able to imitate them artificially. In spite of all the methods employed I have never succeeded, after many days of observation, in seeing even the commencement of yolk segmentation, although in eggs taken from buds I could always reckon with certainty on seeing this take place after twenty hours. I have already explained the important part which the egg-stalk plays at the outset of embryonic evolution.

The larval stage presents differences in the two generations, but they only concern the duration of development, which is very variable. The larvae agree perfectly in their structure and organization; as they all live under similar conditions, no opportunity is offered for any special or varied adaptation to environment. The structure of the mandibles alone differs in some species; thus the larva of Neuroterus has strongly toothed mandibles, while that of Spathegaster has them simple and unindented. It is the character of the galls that regulates these modifications. When the gall substance is hard, as in the Neuroterus galls, the larva is found provided with strong mandibles; if on the contrary the galls are succulent and their walls soft, as in the Spathegaster galls, the mandibles of the larva will be found to be simple and less powerful.

The duration of larval evolution varies much in the two generations. In the summer species the larva increases rapidly, and when full grown passes straight into the pupa state. In the winter species the larval stage is infinitely longer, and presents the following varieties:

(1) The larva developes during the same year and acquires its full growth; then it rests a year, or even more, in the gall (species of genus Aphilotrix).

(2) The larva during the first year undergoes only partial development, passes the winter, and does not complete its evolution until the following year.

(3) Larval development undergoes a period of complete suspension after the larva has left the egg and commenced to form the gall. It remains dormant some months, and only begins to grow again when the gall falls to the ground (Neuroterus).

The prolonged larva rest of certain species is very curious; and that which is especially remarkable is, that we often see them remain unchanged for three years before they enter the pupa state. Even among species which have no alternate generation, there are some galls from which the fly does not emerge until the third year. From the regularity with which this occurs, it seems to be perfectly clear that there are individual differences in the duration of development, consequently we find that in the same species some individuals will complete their development in one year, while others will require two. The prolongation of the larval stage is a remarkable phenomenon; one would rather have imagined that a short stage would have been of greater advantage to the species, since the gall would have been exposed for a less time to danger from changes of temperature. It is possible for the two generations to complete their evolution in one year, as in the Neuroterus-Spathegaster and the Dryophanta-Spathegaster groups. It is interesting to find among the genera which have habitually

a two-year cycle, a species in which the greater part of the individuals complete their evolution in one year.

It may perhaps be taken as an indication that at some former time, when conditions of climate were different, a long larval stage was the rule, but that little by little it has diminished, in some already to its full extent, in some only partially, and in others as yet not at all. We must also admit the same view to hold good with regard to the species which have no alternate generation; with a certain number of individuals the cycle has become annual, with others it is still biennial.

CHAPTER VI.

ALTERNATING GENERATIONS IN OAK GALL-FLIES. THE RELATIONSHIP OF THE PARTHENOGENETIC TO THE SEXUAL GENERATION. HOW IS THIS GENERATION-CYCLE TO BE EXPLAINED?

91

It only remains, in conclusion, to review generally the subject of alternation of generations in oak gall-flies, but I would premise that I have chosen the term 'alternating generations "1 without intending in any way to prejudge the question. It only indicates the possession of cyclical propagation; while the different terms used with regard to this form of propagation, such as alternating generation, heterogenesis, metagenesis, are, although all closely related, each used in a different sense. Lubbock requires, as a necessary condition of alternating generation, that one generation should propagate itself by budding, like Aphides; but with gall-flies propagation by budding does not take place. Although parthenogenesis and budding may not differ in principle, yet there is this important distinction between them, that in the former, embryonic development runs its entire. course outside, and in the latter, inside the ovarium. Among gall-flies both generations develop in exactly the same manner. For this reason I cannot agree with 1 Generationswechsel.

Lichtenstein, who so ably investigated the Phylloxera, in assuming that the agamous generation of gall-flies holds a subordinate position to the sexual generation, in the way that the budding generations of Phylloxera and Aphides do to their winged and sexual generations.

The question of the mutual relationship of the two generations to each other, is of fundamental importance in the inquiry into the origin of alternation of generations. It is necessary therefore next to consider parthenogenesis, in so far as it has a bearing on alternation of generations.

When I first discovered the alternation of generations in gall-flies, I believed in the existence of a definite law, in accordance with which the gall-flies of one generation invariably propagated themselves parthenogenetically, and those of the next generation always sexually. But more extended observation convinced me that there was no such universal law. I soon found that some species continue to propagate themselves by an annual parthenogenesis. This new light led me to investigate more fully parthenogenesis as it exists in other families of Hymenoptera, and I will state briefly the result in so far as of interest to the preceding question.

Parthenogenesis has frequently been observed among saw-flies. Professor v. Siebold showed, by accurate observations made on Nematus ventricosus, that parthenogenesis very frequently occurs, although an equal number of males and females exists in this species. I have myself closely investigated another species, Nematus Vallisnierii.

In the autumn of 1876, I collected a large quantity of