majority of chemists in favour of the theoretical, because it is thought to be more likely that an undetected source of error has attended the experiments than that the statement of the law should be at fault.

The element beryllium is a metal which was formerly supposed, chiefly on the evidence of the specific heat (445 at 10° to 100°), to have the atomic weight 13.5, but it was observed that there is no place in the scheme for an element having the combination of properties exhibited by beryllium associated with such a value for the atomic weight. Then it was discovered that the specific heat of beryllium is anomalous in the same sense as the specific heats of boron, carbon, and silicon especially, and, in a minor degree, of several other elements of comparatively low atomic weight. Redeterminations of the specific heat of beryllium at higher temperatures (62 at 400° to 500°), and applications of these values in the formula expressing the law of Dulong and Petit, led to the value 91 for the atomic weight, and this is now universally adopted. This value is confirmed by the vapour density of the chloride. Hence beryllium is now recognised as the first term of the series to which magnesium belongs.

When, therefore, the equivalent of an element is known, and a question is proposed as to its atomic weight, the answer may be obtained not only by referring to the specific heat, but by recalling the

chief properties of the element, and then considering whether, having such properties, it can possibly occupy the position which would be assigned to it if the equivalent is accepted as equal to the atomic weight. Take the case of magnesium, the equivalent of which is 121. Magnesium is a light metal, forming one salifiable oxide which is insoluble in water; it forms a sulphate which crystallises in prisms containing water of crystallisation, and having the same form as sulphate of zinc. On referring to the table we see that there is no place between carbon and nitrogen available for any element, and that there is no element in any of the columns headed by boron, carbon, or nitrogen which exhibits properties similar to those of magnesium. But the isomorphism of the sulphate with sulphate of zinc at once gives the clue, and we see that the equivalent must be doubled in order to supply the value of the atomic weight. Magnesium with the value 242 or thereabouts then naturally falls into place as the first term of the series of which zinc and cadmium are the succeeding members.

The confidence placed in this table of the elements is based upon the belief that the properties of the elements are intimately dependent upon their atomic weights, that the number of elements is limited, and that their atomic weights can have only certain values. All experience in the past certainly tends to support this view. Thus in 1871 the places in the table now occupied by gallium,

germanium, and scandium were vacant. These three elements have since been recognised with the properties attributed to them from a consideration of their position in the table.1

There is a strong analogy between this manner of displaying the connection between the elements and their atomic weights and the process of grouping the compounds of carbon into homologous series. Fifty years ago a considerable number of hydrocarbons, alcohols, aldehyds, bases &c. were known, but for the most part they were known only as individual substances without recognisable relations to one another. Then it was remarked, first by Schiel, and afterwards by Dumas, that the radicles of the alcohol and of the fatty acids exhibited a regularity of composition, and that the properties of the substances themselves which formed the series were only gradually modified in passing from term to term. Hence wood spirit, common alcohol, and fusel oil were found to be related to another, and even such apparently dissimilar substances as formic acid and stearic acid were recognised as terms of the same 'homologous' series.

The arrangement of the elements themselves into groups of closely related members of the same type, with properties gradually modified through successive terms, was a discovery of the same order,

1 For

a fuller exposition, see Mendeleeff's 'Principles of

Chemistry,' vol. ii.

and since, as pointed out by Dumas, the values of the atomic weights exhibit the same kind of relations to one another as the atomic weights of the radicles in successive terms of a homologous series, the analogy is very remarkable. A single example will suffice:

Now the several short series of elements when placed in parallel columns, as in the table given above, stand towards another in much the same position as heterologous series of carbon compounds; except, of course, that while the latter can be transformed into one another, the former cannot.

The periodic law, then, is based upon and includes the results of the earlier attempts at classification of the elements. All classification is founded upon the recognition of resemblances and differences, upon finding similarity in the midst of diversity; but to be useful a system must be based upon a record of as many points of resemblance as possible. Classification, therefore, implies, first, the practical process of observation, and, secondly, the survey of many observations, with a view to finding resemblances, so constituting the logical process of induction. In looking for indications of relationships, some considerable experience is