All this is perfectly intelligible, both with regard to nominal and verbal themes. Curtius admits four kinds of verbal themes as the outcome of his Fourth Period. He had assigned to his Third Period the simple verbal themes ἐσ-τί, and the reduplicated themes such as δίδω-σι. To these were added, in the Fourth Period, the following four secondary themes:—

(1) πλέκ-ε-(τ)-ιSanskrit lipa-ti
(2) ἀλείφ-ε-(τ)-ι  „  laipa-ti
(3) δείκ-νυ-σι  „  lip-nau-ti
(4) δάμ-νη-σι  „  lip-nâ-ti.

He also explains the formation of the subjunctive in analogy with bases such as lipa-ti, as derived from lip-ti.

Some scholars would probably feel inclined to add one or two of the more primitive verbal themes, such as

limpa-tirumpo
limpana-tiλαμβάνε(τ)ι

but all would probably agree with Curtius in placing the formation of these themes, both verbal and nominal, between the radical and the latest inflectional period. A point, however, on which there would probably be considerable difference of opinion is this, whether it is credible, that at a time when so many nominal themes were formed,—for Curtius ascribes to this Fourth Period the formation of such nominal bases as

λόγ-ο, intellect,= lipa-ti
λοίπ-ο, left,= laipa-ti
λιγ-νύ, smoke,= lip-nau-ti
δάφ-νη, laurel,= lip-nâ-ti —

the simplest nominal compounds, which we now call nominative and accusative, singular and plural, were still unknown; that people could say dhṛsh-nu-más, we dare, but not dhṛsh-ṇú-s, daring-he; that they had an imperative, dhṛshṇuhí, dare, but not a vocative, dhṛshṇo? Curtius strongly holds to that opinion, but with regard to this period too, he does not seem to me to establish it by a regular and complete argument. Some arguments which he refers to occasionally have been answered before. Another, which he brings in incidentally, when discussing the abbreviation of certain suffixes, can hardly be said to carry conviction. After tracing the suffixes ant and tar back to what he supposes to have been their more primitive forms, an-ta and ta-ra, he remarks that the dropping of the final vowel would hardly be conceivable at a time when there existed case-terminations. Still this dropping of the vowel is very common, in late historical times, in Latin, for instance, and other Italian dialects, where it causes frequent confusion and heteroclitism.[32] Thus the Augustan innocua was shortened in common pronunciation to innoca, and this dwindles down in Christian inscriptions to innox. In Greek, too, διάκτορος is older than διάκτωρ; φύλακος older than φύλαξ.

Nor can it be admitted that the nominal suffixes have suffered less from phonetic corruption than the terminations of the verb, and that therefore they must belong to a more modern period (pp. 39, 40). In spite of all the changes which the personal terminations are supposed to have undergone, their connection with the personal pronouns has always been apparent, while the tracing back of the nominal suffixes, and, still more, of the case-terminations to their typical elements, forms still one of the greatest difficulties of comparative grammarians.[33]

Professor Curtius is so much impressed with the later origin of declension that he establishes one more period, the fifth, to which he assigns the growth of all compound verbal forms, compound stems, compound tenses, and compound moods, before he allows the first beginnings of declension, and the formation even of such simple forms as the nominative and accusative. It is difficult, no doubt, to disprove such an opinion by facts or dates, because there are none to be found on either side: but we have a right to expect very strong arguments indeed, before we can admit that at a time when an aorist, like ἔδεικ-σα, Sanskrit a-dik-sha-t was possible, that is to say, at a time when the verb as, which meant originally to breathe, had by constant use been reduced to the meaning of being; at a time when that verb, as a mere auxiliary, was joined to a verbal base in order to impart to it a general historical power; when the persons of the verb were distinguished by pronominal elements, and when the augment, no longer purely demonstrative, had become the symbol of time past, that at such a time people were still unable to distinguish, except by a kind of Chinese law of position, between “the father struck the child,” and “the child struck the father.” Before we can admit this, we want much stronger proofs than any adduced by Curtius. He says, for instance, that compound verbal bases formed with yâ, to go, and afterwards fixed as causatives, would be inconceivable during a period in which accusatives existed. From naś, to perish, we form in Sanskrit nâśa-yâmi, I make perish. This, according to Curtius, would have meant originally, I send to perishing. Therefore nâśa would have been, in the accusative, nâśam, and the causative would have been nâśamyâmi, if the accusative had then been known. But we have in Latin[34] pessum dare, venum ire, and no one would say that compounds like calefacio, liquefacio, putrefacio, were impossible after the first Aryan separation, or after that still earlier period to which Curtius assigns the formation of the Aryan case-terminations. Does Professor Curtius hold that compound forms like Gothic nasi-da were formed not only before the Aryan separation, but before the introduction of case-terminations? I hold, on the contrary, that such really old compositions never required, nay never admitted, the accusative. We say in Sanskrit, dyu-gat, going to the sky, dyu-ksha, dwelling in the sky, without any case-terminations at the end of the first part of the compound. We say in Greek, σακέσ-παλος, not σάκοσ-παλος, παιδοφόνος, not παιδαφόνος, ὀρεσ-κῷος, mountain-bred, and also ὀρεσί-τροφος, mountain-fed. We say in Latin, agri-cola, not agrum-cola, fratri-cīda, not fratrem-cīda, rēgĭfugium, not regis-fugium. Are we to suppose that all these words were formed before there was an outward mark of distinction between nominative and accusative in the primitive Aryan language? Such compounds, we know, can be formed at pleasure, and they continued to be formed long after the full development of the Aryan declension, and the same would apply to the compound stems of causal verbs. To say, as Curtius does, that composition was possible only before the development of declension, because when cases had once sprung up, the people would no longer have known the bases of nouns, is far too strong an assertion. In Sanskrit[35] the really difficult bases are generally sufficiently visible in the so-called Pada, cases, i.e., before certain terminations beginning with consonants, and there is besides a strong feeling of analogy in language, which would generally, though not always (for compounds are frequently framed by false analogy), guide the framers of new compounds rightly in the selection of the proper nominal base. It seems to me that even with us there is still a kind of instinctive feeling against using nouns, articulated with case-terminations, for purposes of composition, although there are exceptions to that rule in ancient, and many more in modern languages. We can hardly realize to ourselves a Latin pontemfex, or pontisfex, still less ponsfex instead of pontifex, and when the Romans drove away their kings, they did not speak of a regisfugium or a regumfugium, but they took, by habit or by instinct, the base regi, though none of them, if they had been asked, knew what a base was. Composition, we ought not to forget, is after all only another name for combination, and the very essence of combination consists in joining together words which are not yet articulated grammatically. Whenever we form compounds, such as railway, we are still moving in the combinatory stage, and we have the strongest proof that the life of language is not capable of chronological division. There was a period in the growth of the Aryan language when the principle of combination preponderated, when inflection was as yet unknown. But inflection itself was the result of combination, and unless combination had continued long after inflection set in, the very life of language would have become extinct.