are considered to be a flock because they are seen to be like. To do James Mill justice, he drops the argument as soon as he has struck it out. It is only worth notice as showing his aim. 'Likeness' seems to imply a relation dependent on the ideas themselves; not purely external and arbitrary. If we could get rid of likeness, all association would ultimately be 'contiguity.' 'The fundamental law of association,' as he says elsewhere,[522] 'is that when two things have been frequently found together, we never perceive or think of the one without thinking of the other.' The two ideas are associated as two balls are associated when they are in the same box. So far as they are themselves concerned, they might be separated without any alteration in their own properties. What, then, corresponds to the 'box'? Association depends upon relations of time and space. Things are associated by occurring in succession or together; the red colour of a rose is in the same place with the shape of the leaf; the scent is perceived at the same time with the colour. The thunder follows the lightning. What, then, he might ask, are 'time' and 'space'? Are they 'ideas' or 'sensations' or qualities of the objects? or, in any case, as supplying the ultimate principle of association, do they not require investigation? Before coming to that problem, however, we have to settle other knotty points. We must clear away illusions which seem to introduce something more than association. Elements of thought not at first sight expressible simply in terms of sensations and ideas must be analysed to show that they are only disguises for different combinations of the facts. Reasoning, according to most logicians, supposes, first, concepts, and therefore some process of classification of the objects of thought; and, secondly, some process of combining these concepts to bring out hitherto unknown truths. What, then, is the meaning of the general or abstract symbols employed in the process? Mill's provision of raw materials consists so far of sensations and ideas, which are worked up so as to form 'clusters' (the word is taken from Hartley) and 'trains.' This corresponds to synchronous and successive associations. How does the logical terminology express these 'clusters' and 'trains'? Mill answers by a theory of 'naming.' Language fulfils two purposes; it is required in order to make our ideas known to others; and in order to fix our own ideas. Ideas are fluctuating, transitory, and 'come into the mind unbidden.' We must catch and make a note of these shifting crowds of impalpable entities. We therefore put marks upon the simple sensations or upon the 'clusters.' We ticket them as a tradesman tickets bundles of goods in his warehouse, and can refer to them for our own purposes or those of others. As the number of objects to be marked is enormous, as there are countless ideas and clusters and clusters of clusters of endless variety to be arranged in various