[6] The still more complex combinations--which are so clearly expressed in the crystallo-hydrates, double salts, and similar compounds--although they may be regarded as independent, are, however, most easily understood with our present knowledge as aggregations of whole molecules to which there are no corresponding double compounds, containing one atom of an element R and many atoms of other elements RX_{_n_}. The above types embrace all cases of direct combinations of atoms, and the formula MgSO_{4},7H_{2}O cannot, without violating known facts, be directly deduced from the types MgX_{_n_} or SX_{_n_}, whilst the formula MgSO_{4} corresponds both with the type of the magnesium compounds MgX_{2} and with the type of the sulphur compounds SO_{2}X_{2}, or in general SX_{6}, where X_{2} is replaced by (OH)_{2}, with the substitution in this case of H_{2} by the atom Mg, which always replaces H_{2}. However, it must be remarked that the sodium crystallo-hydrates often contain 10H_{2}O, the magnesium crystallo-hydrates 6 and 7H_{2}O, and that the type PtM_{2}X_{6} is proper to the double salts of platinum, &c. With the further development of our knowledge concerning crystallo-hydrates, double salts, alloys, solutions, &c., in the _chemical sense_ of feeble compounds (that is, such as are easily destroyed by feeble chemical influences) it will probably be possible to arrive at a perfect generalisation for them. For a long time these subjects were only studied by the way or by chance; our knowledge of them is accidental and destitute of system, and therefore it is impossible to expect as yet any generalisation as to their nature. The days of Gerhardt are not long past when only three types were recognised: RX, RX_{2}, and RX_{3}; the type RX_{4} was afterwards added (by Cooper, Kekulé, Butleroff, and others), mainly for the purpose of generalising the data respecting the carbon compounds. And indeed many are still satisfied with these types, and derive the higher types from them; for instance, RX_{5} from RX_{3}--as, for example, POCl_{3} from PCl_{3}, considering the oxygen to be bound both to the chlorine (as in HClO) and to the phosphorus. But the time has now arrived when it is clearly seen that the forms RX, RX_{2}, RX_{3}, and RX_{4} do not exhaust the whole variety of phenomena. The revolution became evident when Würtz showed that PCl_{5} is not a compound of PCl_{3} + Cl_{2} (although it may decompose into them), but a whole molecule capable of passing into vapour, PCl_{5} like PF_{5} and SiF_{4}. The time for the recognition of types even higher than RX_{8} is in my opinion in the future; that it will come, we can already see in the fact that oxalic acid,