As any scientist knows, a theory must not be judged by its plausibility or lack of it, but by how well it can anticipate new discoveries. The most important things for us to summarize, therefore, are the new things to look for. We need to know a great deal more about the developmental periods of non-periodical cicadas. On purely demographic grounds, a long pre-reproductive period should be a severe disadvantage; other things being equal, it should be strongly selected against. In view of the data in Table 2, one must postulate some counter-advantages for the long developmental periods characteristic of cicadas. As one hypothesis, we suggest in Table 3 a counter-advantage connected with loud song, large size, and obligately slow feeding. This idea could be exploded by finding a large cicada with a short life cycle, provided its feeding habits are typical of cicadas in general. The current work of Anderson and Shorey with Diceroprocta apache on asparagus roots (see earlier footnote) may provide exactly the disproof that is needed. With the advent of nearly-perfect periodicity, the rules are considerably changed. Given that there already is a long life cycle, the above-ground predators become satiated, and the below-ground ones deprived of food, by every periodical emergence (Lloyd and Dybas, 1966). The prey can now evolve predator-foolhardy behavior, especially as old behavior patterns may be incompatible with the demands of reproduction at enormously elevated population densities. The most severe selective penalties are now levied against those individuals that fail to emerge along with the majority; either a lengthening or shortening of the life cycle would be selected against under these circumstances. As between two competing cicada populations with equally perfect periodicities, however, the shorter life cycle should still have an advantage. Thus, we anticipate that 13-year broods should now be displacing 17-year ones wherever their ranges are contiguous, as, for example, XIX and XXIII supplanting IV in the Ozarks (cf. Figs. 7A-B and 6G). This idea can obviously be tested by detailed mapping of future emergences and historical records. In the course of evolution towards periodicity, phenotypic variability in the developmental period would acquire a maladaptive significance that it could never have had before. The fact that immature periodical cicada nymphs are as variable in size as they are (Fig. 1) must mean that these species have some special physiological mechanism (i.e., something other than a very uniform growth rate) which insures that the nymphs will emerge together after 13 or 17 years-some mechanism like "counting" diapauses, for example. The elucidation of this physiological mechanism is a worthwhile research objective in its own right. The ecological theory predicts that no such mechanism will be found in any non-periodical cicada, because it would have no selective advantage without the periodicity. We need, therefore, to know something about the variability, as well as the mean duration, of developmental periods in non-periodical cicadas. Looking back to an evolutionary stage before the physiological timing mechanism and the periodicity were perfected, we can imagine that strong selection would have been necessary in order to prevent the incipient periodicity from disappearing of its own accord. How strong would depend, of course, on how great was the variability in the developmental period (Fig. 2). We can also imagine population interactions taking place between "protoperiodical" cicadas and their predators that one would not expect to see taking place now. If cicadas were to appear above ground for several consecutive years (with a hiatus in between), then populations of parasitoids with life cycles of one year (living on alternative hosts during the hiatus) would build up, as shown in Fig. 3. This idea predicts that the alternative hosts of parasitoids that attack periodical cicada eggs or adults should suffer much greater mortality rates from parasitoid attack in the year after a periodical cicada emergence than in the year prior to one. Unfortunately, this would be a difficult thing to measure, since the alternative hosts will generally be comparatively rare. Another testable prediction is that wherever two periodical broods one year out of phase have adjacent ranges (see I-II, III-IV, V-VI, VIII-IX, IX-X, XXII-XXIII, Figs. 6 and 7), the leading brood should gradually be supplanting the lagging brood. Moreover, this should be occurring even where the population densities are so low that underground competition for feeding sites is not likely to be important, i.e., at densities of fewer than five mature nymphs per square yard. If this is, in fact, a tenable idea, then it creates difficulties with the theory, because the interaction with year-to-year parasitoids would produce a strong selective pressure favoring ever-shorter developmental periods in protoperiodical cicadas (see Fig. 3). We could accommodate this by assuming that the protoperiodical ancestor had a life cycle even longer than 17 years but, if that were true, one would expect to find life cycles of this length in present-day non-periodical cicadas. None have turned up so far, and the pattern of developmental periods relative to body size (Table 2) suggests that none are likely to be found. As an alternative explanation, we are led to postulate something which can never be verified directly: a parasitoid with a life cycle of many years duration, nearly synchronized with the protoperiodical cicadas and exerting a counter-selection pressure favoring a longer life cycle. Since there is no trace of such a parasitoid now, we have to assume that it became extinct. We cannot assume that this hypothetical parasitoid was one that attacked the cicada nymphs, because we would then have no explanation for the apparent absence of this kind of life cycle in parasitoids of present-day non-periodical cicadas. We are left with a hypothetical parasitoid attacking eggs or adults, which implies a long period of dormancy underground. The only documented case of this that we know about is the report of Tillyard (1926) concerning two- or three-year dormancy in larvae of the Australian cicadahunting wasp, Exeirus lateritius. However, the startling data of Barnes (Fig. 4) suggest that long dormancy underground is possible, even for small insects. The testable prediction the theory makes, then, is that prolonged underground dormancy of this kind will prove to be a great deal more common and widespread among insect parasitoids than is generally realized. An alternative hypothesis to account for the evolution of exceedingly slow developmental rates in protoperiodical cicadas assumes very high population densities, severe competition for food, and a spectacular ability on the part of certain genotypes to withstand prolonged periods of starvation. There is no apparent reason why periodical nymphs should not still retain this ability, so this idea is testable also. In addition to the very long life cycles themselves, any comprehensive theory must be able to account for both the 13-year and 17-year life cycles (both perfectly periodical), for the various broods within each life cycle (Figs. 6 and 7), and for the three sibling species (or species pairs, see Table 1)-species that coexist sympatrically within each brood, within either life cycle. In Fig. 6E, we adopt the suggestion of Alexander and Moore (1962) that adjacent broods one year out of step arose by summation of diapauses, resulting perhaps from unseasonable cold in a particular year, and we add to this the further hypothesis that accelerations of four years can also take place. This leads us to postulate that the 17-year cicadas differ from the 13-year ones by the possession of a supernumcrary sixth instar. There is at present no convincing direct evidence for this sixth instar in 17-year cicadas, but the assumption that it exists, and can be omitted from the life cycle in response to some environmental shock, enables us to postulate a mechanism for deriving a fully-periodical 13-year life cycle directly from a fully-periodical 17-year one, without losing the periodicity in the process. Fortunately, the question of a supernumerary sixth instar is directly testable. Present efforts are being directed towards that end. As one would expect, the habitat preferences of the three species differ. Dybas and Lloyd (1962) described these differences for Magicicada septendecim and M. cassini. A future paper will treat all three species, with both 13-year (Brood XXIII) and 17-year (Broods II, III, IV) life cycles. We find the same habitat differences among the three species, whatever the life cycle, whatever the brood. The ecological tie that has kept these three species in perfect synchrony with one another, through all this evolution of different broods and life cycles is, by hypothesis, the same mechanism that was originally responsible for selecting in favor of periodicity-the possibility of satiating predators and, given that predators are satiated, the selective penalties against (predator-fool-hardy) individuals that do not emerge in synchrony with the group. A few such individuals still occur, sometimes even several years out of synchrony. These "stragglers" are probably seldom reported in the literature, presumably because their significance is not generally appreciated.