their dialects. Their history is from the beginning a history of decay rather than of growth, and hence the unmistakable family-likeness which pervades every one even of their latest descendants. The language of the Sepoy and that of the English soldier are, strictly speaking, one and the same language. They are both built up of materials which were definitely shaped before the Teutonic and Indic branches separated. No new root has been added to either since their first separation; and the grammatical forms which are of more modern growth in English or Hindustání, are, if closely examined, new combinations only of elements which existed from the beginning in all the Aryan dialects. In the termination of the English he is, and in the inaudible termination of the French il est, we recognize the result of an act performed before the first separation of the Aryan family, the combination of the predicative root as with the demonstrative root ti; an act performed once for all, and continuing to be felt to the present day.

It was the custom of Nebuchadnezzar to have his name stamped on every brick that was used during his reign in erecting his colossal palaces. Those palaces fell to ruins, but from the ruins the ancient materials were carried away for building new cities; and on examining the bricks in the walls of the modern city of Baghdad on the borders of the Tigris, Sir Henry Rawlinson discovered on each the clear traces of that royal signature. It is the same if we examine the structure of modern languages. They too were built up with the materials taken from the ruins of the ancient languages, and every word, if properly examined, displays the visible stamp impressed upon it from the

first by the founders of the Aryan and the Semitic empires of speech.

The relationship of languages, however, is not always so close. Languages may diverge before their grammatical system has become fixed and hardened; and in that case they cannot be expected to show the same marked features of a common descent as, for instance, the Neo-Latin dialects, French, Italian, and Spanish. They may have much in common, but they will likewise display an after-growth in words and grammatical forms peculiar to each dialect. With regard to words we see that even languages so intimately related to each other as the six Romance dialects, diverged in some of the commonest expressions. Instead of the Latin frater, the French frère, we find in Spanish hermano. There was a very good reason for this change. The Latin word frater, changed into fray and frayle, had been applied to express a brother or a friar. It was felt inconvenient that the same word should express two ideas which it was sometimes necessary to distinguish, and therefore, by a kind of natural elimination, frater was given up as the name of brother in Spanish, and replaced from the dialectical stores of Latin, by germanus. In the same manner the Latin word for shepherd, pastor, was so constantly applied to the shepherd of the people or the clergyman, le pasteur, that a new word was wanted for the real shepherd. Thus berbicarius from berbex or vervex, a wether, was used instead of pastor, and changed into the French berger. Instead of the Spanish enfermo, ill, we find in French malade, in Italian malato. Languages so intimately related as Greek and Latin have fixed on dif ferent expressions for son, daughter, brother, woman,

man, sky, earth, moon, hand, mouth, tree, bird, &c.1 That is to say, out of a large number of synonymes which were supplied by the numerous dialects of the Aryan family, the Greeks perpetuated one, the Romans another. It is clear that when the working of this principle of natural selection is allowed to extend more widely, languages, though proceeding from the same source, may in time acquire a totally different nomenclature for the commonest objects. The number of real synonymes is frequently exaggerated, and if we are told that in Icelandic there are 120 names for island, or in Arabic 500 names for lion,2 and 1,000 names for sword,3 many of these are no doubt purely poetical. But even where there are in a language only four or five names for the same objects, it is clear that four languages might be derived from it, each in appearance quite distinct from the rest.

The same applies to grammar. When the Romance languages, for instance, formed their new future by placing the auxiliary verb habere, to have, after the infinitive, it was quite open to any one of them to fix upon some other expedient for expressing the future. The French might have chosen je vais dire or je dirvais (I wade to say) instead of je dirai, and in this case the future in French would have been totally distinct from the future in Italian. If such changes are possible in literary languages of such long standing as French and Italian, we must be prepared for a great deal more in languages which, as I said, diverged before any definite settlement had taken place either in their

1 See Letter on Turanian Languages, p. 62.

Renan, Histoire des Langues sémitiques, p. 137.

3 Pococke, Notes to Abulfaragius, p. 153; Glossology, p. 352.

grammar or their dictionary. If we were to expect in them the definite criteria of a genealogical relationship which unites the members of the Aryan and Semitic families of speech, we should necessarily be disappointed. Such criteria could not possibly exist in these languages. But there are criteria for determining even these more distant degrees of relationship in the vast realm of speech; and they are sufficient at least to arrest the hasty conclusions of those who would deny the possibility of a common origin of any languages more removed from each other than French and Italian, Sanskrit and Greek, Hebrew and Arabic. You will see this more clearly after we have examined the principles of what I call the morphological classification of human speech.

As all languages, so far as we can judge at present, can be reduced in the end to roots, predicative and demonstrative, it is clear that, according to the manner in which roots are put together, we may expect to find three kinds of languages, or three stages in the gradual formation of speech.

1. Roots may be used as words, each root preserving its full independence.

2. Two roots may be joined together to form words, and in these compounds one root may lose its independence.

3. Two roots may be joined together to form words, and in these compounds both roots may lose their independence.

What applies to two roots, applies to three or four or more. The principle is the same, though it would lead to a more varied subdivision.

The first stage, in which each root preserves its in

dependence, and in which there is no formal distinction between a root and a word, I call the Radical Stage. This stage is best represented by ancient Chinese. Languages belonging to this first or Radical Stage, have sometimes been called Monosyllabic or Isolating. The second stage, in which two or more roots coalesce to form a word, the one retaining its radical independence, the other sinking down to a mere termination, I call the Terminational Stage. This stage is best represented by the Turanian family of speech, and the languages belonging to it have generally been called agglutinative, from gluten, glue. The third stage, in which roots coalesce so that neither the one nor the other retains its substantive independence, I call the Inflectional Stage. This stage is best represented by the Aryan and Semitic families, and the languages belonging to it have sometimes been distinguished by the name of organic or amalgamating.

The first stage excludes phonetic corruption altogether.

The second stage excludes phonetic corruption in the principal root, but allows it in the secondary or determinative elements.

The third stage allows phonetic corruption both in the principal root and in the terminations.

A few instances will make this classification clearer. In the first stage, which is represented by Chinese, every word is a root, and has its own substantial meaning. Thus, where we say in Latin baculo, with a stick, we say in Chinese ŷ ćáng.1 Here might be taken for a mere preposition, like the English with. But in Chinese this is a root; it is the same word which,

1 Endlicher, Chinesische Grammatik, p. 223.