Few measurements of the osmotic pressure of salts have been made, owing to the difficulty in producing a membrane which shall allow water to pass, and which shall be impermeable to salts. But very numerous measurements of the depression in freezing-point and the rise in boiling-point of solutions of salts have been made ; and it has been already explained that these quantities are proportional to the osmotic pressure of the dissolved substances. It has been experimentally discovered that in all such cases the fall in freezing-point, or the rise in boiling-point is too great for the supposed molecular weight of the salt. It must be concluded that the osmotic pressure would also be increased, were it possible to measure it. But the fall in freezing-point or the rise in boiling-point does not imply a doubled osmotic pressure, when there is reason to expect it, unless the solution is very dilute. Now, if the pressure were doubled, we might argue from such cases as ammonium chloride that dissociation into two portions had occurred ; but in moderately concentrated solutions, as the pressure is not doubled, it must be concluded that the dissociation is not complete ; it is only in very dilute solutions that complete dissociation can be imagined to have taken place. Cases are known where substances in the state of gas undergo gradual dissociation, and then the pressure does not attain its maximum until the temperature has been sufficiently raised or the pressure sufficiently reduced. The reason that this is not noticed with ammonium chloride is that the temperature of complete dissociation has been reached before the substance turns to gas.

Common salt is chloride of sodium ; its formula is NaCl ; and for long the suggestion that it dissociated into an atom of sodium and an atom of chlorine on being dissolved in water was received as too improbable to be worth consideration. There is, of course, another way out of the difficulty ; it is to suppose that a molecule of salt has the formula Na₀Cl₂ ; in that case, 117 grams of salt—(2x23)+ (2x35.5)—dissolved in 10,000 grams of water should produce the normal lowering of freezing-point ; or, if it produced a larger lowering, it might be supposed that these complex molecules had split more or less completely into the simpler molecules, NaCl. But though the explanation suggested might account for this instance, it is incapable of accounting for the fact that chloride of barium, which is known to possess the formula BaCl₂ (or a multiple thereof), gives, in sufficiently dilute solution, a depression three times that which one would have expected from its supposed molecular weight, or that ferricyanide of potassium and ferrocyanide of potassium, the formulae of which are respectively K₃Fe(CN)_f and K₄Fe(CN)₆, should give four and five times the expected depression. But these results are quite consistent with the hypothesis that NaCl + Aq decomposes into Na. Aq and CI. Aq ; BaCl₂ + Aq decomposes into Ba. Aq and CI. Aq, and CI. Aq ; K₃Fe(CN)_fi + Aq decomposes into K.Aq + K.Aq + K.Aq, and Fe(CN)₆.Aq; and K₄Fe(CN)₆ + Aq decomposes into K.Aq + K.Aq + K.Aq + K.Aq, and Fe(CN)₆.Aq. (The symbol " Aq " stands for an indefinite but large amount of water—"aqua.") Here again we are face to face with facts and an attempted explanation. The facts are that certain compounds, which have long been known as " salts," give too great a depression of the freezing-point or too great a rise of boiling-point of the solvent in which they are dissolved, corresponding to too great an osmotic pressure. It has been observed that when the dilution is sufficient the depression in each case reaches a maximum, and that that maximum is two, three, four, or five times what might be expected ; and in each case it is possible to divide the salt into two, three, four, or live imaginary portions, which often consist of atoms, though frequently of groups of atoms.