It had been ascertained in fact that the rust fungus is not, as was at first supposed, a single organism, but comprises, according to Eriksson, thirteen distinct species, each with physiological varieties, and that those which are destructive to some grasses and cereals, are incapable of attacking others. This necessitated a scrutiny of the nature of grass-immunity. In a paper communicated to the Cambridge Philosophical Society in 1902, Ward announced a conclusion which was as important as it was unexpected. He had more and more made use of the graphical method for presenting to the eye at a glance the result of a mass of separate observations. In this case he uses it with striking effect. He shows conclusively, as far as rust in brome-grasses is concerned, that: "The capacity for infection, or for resistance to infection, is independent of the anatomical structure of the leaf, and must depend on some other internal factor or factors in the plant." Finally, he is led to the conclusion that "it is in the domain of the invisible biological properties of the living cell that we must expect the phenomena to reside." He pointed out the probability that light would be thrown on this from the action of chemotaxis, on the one hand, and from that of toxins and antitoxins in animal organisms on the other. This is a most fertile conception, which would, however, have required a good deal of verification, and this, unhappily, he did not live to attempt. But with characteristic ingenuity he pointed out the analogy between the infective capacity of uredospores and the prepotency of pollen, which had previously engaged the attention of Darwin. In a paper published in the following year in the Berlin _Annales Mycologici_, he announced a no less significant result. With his usual thoroughness in research he had cultivated side by side at Cambridge more than two hundred species and varieties of Bromus, and had watched the degree to which they were infected by rust under identical conditions. He found that though in the brome-grasses the rust peculiar to them is specifically identical its _forms_ are highly specialised. The form which attacks the species of one group will not attack those of another. Host and parasite are mutually "attuned." He termed this "adaptive parasitism." This raised the problem, which had first occurred to him in Ceylon, of how a parasite adapted to species of "one circle of alliance" can pass to those of another. Occasionally it happens that a uredo-form will infect a species where it ordinarily fails. In such a case "its uredospore progeny will thenceforth readily infect that species." Ward regarded this as a case of education. Working on this principle, he succeeded by growing the parasite _successively_ on a series of allied species which were imperfectly resistant, to ultimately educate it to attack a species hitherto immune. He called these "_bridgeing species_." He established, in fact,