Quintilian, Institutio Oratoria, Book 1
Harold Edgeworth Butler, Ed.

7. Having stated the rules which we must follow in speaking, I will now proceed to lay down the rules which must be observed when we write. Such rules are called orthography by the Greeks; let us style it the science of writing correctly. This science does not consist merely in the knowledge of the letters composing each syllable (such a study is beneath the dignity of a teacher of grammar), but, in my opinion, develops all its subtlety in connexion with doubtful points. [2] For instance, while it is absurd to place a circumflex over all long syllables since the quantity of most syllables is obvious from the very nature of the word which is written, it is all the same occasionally necessary, since the same letter involves a different meaning according as it is long or short. For example we determine whether mains is to mean an “apple tree” or a “bad man” by the use of the circumflex; [3] palus means a “stake,” if the first syllable is long, a “marsh,” if it be short; again when the same letter is short in the nominative and long in the ablative, we generally require the circumflex to make it clear which quantity to understand. [4] Similarly it has been held that we should observe distinctions such as the following: if the preposition ex is compounded with specto, there will be an s in the second syllable, while there will be no s if it is compounded with pecto. [5] Again the following distinction has frequently been observed: ad is spelt with a d when it is a preposition, but with a t when it is a conjunction, while cum is spelt quum when it denotes time, but cum when it denotes accompaniment. [6] Still more pedantic are the practices of making the fourth letter of quidquid a c to avoid the appearance of repeating a question, and of writing [p. 137] quotidie instead of colidie to show that it stands for quot diebus. But such practices have disappeared into the limbo of absurdities.

[7] It is often debated whether in our spelling of prepositions we should be guided by their sound when compounded, or separate. For instance when I say optinuit, logic demands that the second letter should be a b, while to the ear the sound is rather that of p: or again take the case of immunis: [8] the letter n, which is required by strict adherence to fact, is forced by the sound of the m. which follows to change into another m. [9] We must also note when analysing compound words, whether the middle consonant adheres to the preceding syllable or to that which follows. For example since the latter part of haruspex is from spectare, the s must be assigned to the third syllable. In abstemius on the other hand it will go with the first syllable since the word is derived from abstinentia temeti, “abstention from wine.” [10] As for k my view is that it should not be used at all except in such words as may be indicated by the letter standing alone as an abbreviation.¹ I mention the fact because some hold that k should be used whenever the next letter is an a, despite the existence of the letter c which maintains its force in conjunction with all the vowels.

Orthography, however, [11] is also the servant of usage and therefore undergoes frequent change. I make no mention of the earliest times when our alphabet contained fewer letters² and their shapes differed from those which we now use, while their values also were different. For instance in Greek the letter o was sometimes long and short, as it is with us, and again was sometimes used to express the syllable [p. 139] which is identical with its name.³ [12] And in Latin ancient writers ended a number of words with d, as may be seen on the column adorned with the beaks of ships, which was set up in the forum in honour of Duilius.⁴ Sometimes again they gave words a final g, as we may still see in the shrine of the Sun, close to the temple of Quirinus, where we find the word uesperug, which we write uesperugo (evening star). [13] I have already spoken of the interchange of letters⁵ and need not repeat my remarks here: perhaps their pronunciation corresponded with their spelling. [14] For a long time the doubling of semivowels was avoided,⁶ while down to the time of Accius and beyond, long syllables were indicated by repetition of the vowel. [15] The practice of joining e and i as in the Greek diphthong ει lasted longer: it served to distinguish cases and numbers, for which we may compare the instructions of Lucilius:

The boys are come: why then, their names must end
With e and i to make them more than one; and later—

If to a thief and liar (mendaci furique) you would give,
In e and i your thief must terminate.

But this addition of e is quite superfluous, since t can be long no less than short: [16] it is also at times inconvenient. For in those words which end in i and have e as their last letter but one, we shall on this principle have to write e twice: I refer to words such as aurei or argentei and the like. [17] Now such a practice will be an actual hindrance to those who are learning to read. This difficulty occurs in Greek as [p. 141] well in connexion with the addition of an iota, which is employed not merely in the termination of the dative, but is sometimes found in the middle of words as in λῄστης, for the reason that the analysis applied by etymology shows the word to be a trisyllable⁷ and requires the addition of that letter. The diphthong ae now written with an e, was pronounced in old days as ai; [18] some wrote ai in all cases, as in Greek, others confined its use to the dative and genitive singular; whence it comes that Vergil,⁸ always a passionate lover of antiquity, inserted pictai uestis and aquai in his poems. [19] But in the plural they used e and wrote Syllae, Galbae. Lucilius has given instructions on this point also; his instructions occupy quite a number of verses, for which the incredulous may consult his ninth book. [20] Again in Cicero's days and a little later, it was the almost universal practice to write a double s, whenever that letter occurred between two long vowels or after a long vowel, as for example in caussae, cassus, diuissiones. That he and Vergil both used this spelling is shown by their own autograph manuscripts. [21] And yet at a slightly earlier date iussi which we write with a double s was spelt with only one. Further optimnus maximus, which older writers spelt with a u, appear for the first time with an i (such at any rate is the tradition) in an inscription of Gaius Caesar.⁹ [22] We now write here, but I still find in manuscripts of the old comic poets phrases such as heri ad me uenit,¹⁰ and the same spelling is found in letters of Augustus written or corrected by his own hand. [23] Again did not Cato the censor spell dicam and faciam as dicem [p. 143] and faciem and observe the same practice in words of similar termination? This is clear from old manuscripts of his works and is recorded by Messala in his treatise on the letter s. Sibe and quase are found in many books, but I cannot say whether the authors wished them to be spelt thus: [24] I learn from Pedianus that Livy, whose precedent he himself adopted, used this spelling: to-day we make these words end with an i. [25] What shall I say of uorlices, uorsus and the like, which Scipio Africanus is said to have been the first to spell with an e? [26] My own teachers spelt seruus and ceruus with a uo, in order that the repetition of the vowel might not lead to the coalescence and confusion of the two sounds: to-day however we write these words with a double u on the principle which I have already stated: neither spelling however exactly expresses the pronunciation. It was not without reason that Claudius introduced the Aeolic digamma to represent this sound.¹¹ It is a distinct improvement that to-day we spell cui as I have written it: [27] when I was a boy it used to be spelt quoi, giving it a very full sound, merely to distinguish it from qui.

[28] Again, what of words whose spelling is at variance with their pronunciation? For instance C is used as an abbreviation for Gaius, and when inverted stands for a woman, for as we know from the words of the marriage service women used to be called Gaiae, just as men were called Gaii.¹² Gnaeus [29] too in the abbreviation indicating the praenomen is spelt in a manner which does not agree with its pronunciation. We also find columnas¹³ and consul spelt without an n, [p. 145] while Subura when indicated by three letters is spelt Suc.¹⁴ I could quote many other examples of this, but I fear that I have already said too much on so trivial a theme.

[30] On all such subjects the teacher must use his own judgment; for in such matters it should be the supreme authority. For my own part, I think that, within the limits prescribed by usage, words should be spelt as they are pronounced. [31] For the use of letters is to preserve the sound of words and to deliver them to readers as a sacred trust: consequently they ought to represent the pronunciation which we are to use. [32] These are the more important points in connexion with writing and speaking correctly. I do not go so far as to deny to the teacher of literature all part in the two remaining departments of speaking and writing with elegance and significance, but I reserve these for a more important portion of this work, as I have still to deal with the duties of the teacher of rhetoric.

[33] I am however haunted by the thought that some readers will regard what I have said as trivial details which are only likely to prove a hindrance to those who are intent upon a greater task; and I myself do not think that we should go so far as to lose our sleep of nights or quibble like fools over such minutiae; for such studies make mincemeat of the mind. But it is only the superfluities of grammar that do any harm. [34] I ask you, is Cicero a less great orator for having given this science his diligent attention or for having, as his letters show, demanded rigid correctness of speech from his son? Or was the vigour of Gaius Caesar's eloquence impaired by the publication of a treatise on Analogy? [35] Or the polish [p. 147] of Messala dimmed by the fact that he devoted whole books to the discussion not merely of single words, but of single letters? Such studies do no harm to those who but pass through them: it is only the pedantic stickler who suffers.