¹ The scope of this paper is arbitrarily restricted to the discussion of clauses introduced by the definitely recognized case forms of the pronoun; but the main points of the argument are applicable to the clauses introduced by the adverbial forms.

² ‘Der lateinische Relativsatz’, Glotta 3. 1–18.

³ Stolz-Schmalz Lateinische Grammatik⁵ 708 ff. This work will hereafter be referred to as Hofmann.

⁴ See Hofmann 706; Bräunlich, The Indicative Indirect Question in Latin 191 ff.; Buck, A Grammar of Oscan and Umbrian 143.

⁵ The quality is necessarily a participial quality, but that fact makes no difference in the classification.

⁶ The term ‘noun’ is here used in a sense which includes substantive and adjective.

⁷ Collections of examples of qui-clauses may be found in the following works : Hale, The Cum-constructions; Dittmar, Studien zur Lateinischen Moduslehre; Gaffiot, Le Subjonctif de Subordination en Latin; Gaffiot, Pour le Vrai Latin; Methner, Bedeutung und Gebrauch des Konjunktivs.

⁸ The pronoun here is not anaphoric. On the distinction between anaphoric and deictic pronouns see Wackernagel, Vorlesungen über Syntax² 84 ff.

⁹ Compare Poutsma, Grammar of Late Modern English 417 ff., 985 ff.; Jespersen, Philosophy of Grammar 103 ff.

¹⁰ See Hale and Buck, Latin Grammar 300–2.

¹¹ In general the substantive is to be recognized through the absence of an ‘antecedent’ expressed or understood. If the qui-clause precedes the main clause, there can be no antecedent; in other cases if no antecedent is expressed and if the antecedent in case it were expressed would be in an oblique case, no antecedent is understood. For qui-clauses without an antecedent, see Hofmann 707. Kroll in his article, Glotta 3. 1–18, recognizes as old that type of sentence in which the qui-clause precedes; but the position of the clause is an incidental matter. Neither Hofmann nor Kroll seems to recognize the existence of a substantive qui-clause.

¹² The substantive clause occurs in the Italic dialects; as a ‘generalizing’ clause it is common in the Oscan and Umbrian legalistic inscriptions.

¹³ We even have some examples of the substantive clause depending on a preposition. Cic, Att. 5. 11. 6: nunc redeo ad quae mihi mandas. The construction is common enough in English.

¹⁴ For an example in Oscan, see Conway, The Italic Dialects, No. 42; Buck, A Grammar of Oscan and Umbrian 240.

¹⁵ For examples see Deecke, De Usu Pronominis Relativi apud Poetas Veteres Latinos Quaestiones Syntacticae 53 ff.

¹⁶ Since it was the stem quo- which was regular in the adjectival use of the indefinite and interrogative pronoun, it was natural that this stem should appear in the qui homo type of substantive clause and so as a relative with a noun antecedent. We may conjecture that it displaced the stem qui- as a relative with is as antecedent and that then it displaced the same stem used pronominally in the substantive clause.

¹⁷ On the nature of hypothesis in general one may compare Vaihinger, The Philosophy of ‘As If’ (English translation by Ogden), Chapter xxi.

¹⁸ For examples see Methner, Bedeutung und Gebrauch des Konjunktivs 61–77.

¹⁹ The theory that the use of the modes in qui-clauses in early Latin differed from that in later Latin is to a large extent the result of the failure to distinguish the definite from the indefinite in sentences of the est qui type.

²⁰ For other examples see Gaffiot, Pour le Vrai Latin 117–32. Such clauses are to be sharply distinguished from the indefinite hypothetical clauses.

²¹ The true character of the indefinite hypothetical clause becomes clearer if one compares such sentences as the following. Trin. 363: nam sapiens quidem pol ipsus fingit fortunam sibi. Here sapiens is an indefinite hypothetical designation.

²² Concerning asyndetic clauses existing in Latin, see Juret, Système de la Syntaxe Latine 304 ff. Concerning the part played by asyndesis in the history of English relative clauses, see Curme, Journal of English and Germanic Philology 11. 10 ff., 180 ff., 355 ff.

²³ It may be that a number of clauses in Latin are to be interpreted as asyndetic substantive indefinite hypothetical clauses. Juret (Systéme de la Syntaxe Latine 308) interprets as an asyndetic conditional clause Cic., Parad. 6. 44: filiam quis habet, pecunia opus est. But the sentence is probably equivalent to qui filiam habet ei pecunia opus est. Probably we should think of quivis and quilibet as stereotyped asyndetic clauses. This raises the question whether the clause in the following example is not asyndetic. Amph 396: quid tibi lubet fac. Compare Merc. 991, Bacch. 866, Pers. 398, Trin. 570. One may even go so far as to suggest as a possibility that forms of the stem qui- were never subordinating in Latin or the dialects. For Latin examples see Bräunlich The Indicative Indirect Question in Latin, 193ff. But the clause in sentences of the non (nihil) habeo quid scribam type whether asyndetic or not are indefinite real. For Ciceronian examples see Gaffiot, Pour le Vrai Latin 73ff.

²⁴ Quoted by Curme, Language 4. 146, from Lucy Furnam.

²⁵ Concerning ‘incompleteness’ see Mendell, Latin Sentence Connection, Chapters iv and v.

²⁶ For this and other examples, see Poutsma, Grammar of Late Modern English 1002.

²⁷ Compare Sturtevant, Linguistic Change 133ff.—See also Leumann, ‘Zum Mechanismus des Bedeutungswandels’, IF 45. 105–18. While this article deals with changes in the meaning of words, the principles set forth are applicable to the mechanism of the formation and development of subordinate clauses.

²⁸ As a probable example of a sentence of this type but used to express as a hypothesis the existence of an action, may be cited Eun. 252: negat quis; nego. In the example from Cic., Parad. 6. 44, cited in note 23, if quis is not subordinating, it is not impossible to take filiam quis habet as an independent sentence expressing the hypothetical existence of a person.

²⁹ Of course, when the clause was once established in one type of sentence, then the clause itself became linguistic material for use in other types. It is possible, for example, that the clause originated in the est qui sentences and that it was thence taken into the other types. We should, however, do well to think of the existence of the clause in one type as an influence making easier the independent formation of the clause in the others.

³⁰ It is the belief of the present writer (1) that the indefinite real clause in Latin always has the subjunctive and (2) that the use of the subjunctive in this clause was original, that is, that the asyndetic substantive clause and the substantive sentence from which this was formed had the subjunctive. Other theories concerning the subjunctive in this clause are not a few. But theories concerning the mode have little or no bearing on the theory of origin advanced in this paper. That theory has to do with the process by which the qui subordinate clause was formed and by which qui acquired subordinating function; and that process was not conditioned by the mode used.

³¹ That the type was Italic is indicated by the Oscan example referred to in note 14.

³² For other examples see Methner, Bedeutung und Gebrauch des Konjunktivs 41–67.