The rigid poetic meters used by ancient poets strongly indicate that elision is done (almost) every time one word ends in a vowel and the next one begins with another — with the usual exceptions regarding h- and -m. In prose there are no clear rhythmical hints as to whether elision was there or not.
Do we know if elision was widely or systematically used in classical Latin outside poetry? Is there a reason to believe that elision was or was not specific to verse?
First, though it is indicated by an apostrophe in modern texts, elision also occurs in ancient Greek poetry. The rules were different from Latin, though. I quote Smyth for them:
- Elision is the expulsion of a short vowel at the end of a word before a word beginning with a vowel. An apostrophe ( ’ ) marks the place where the vowel is elided. ἀλλ’(ά) ἄγε, ἄδωκ’(α) ἐννέα, ἐφ’ ( = ἐπὶ) ἑαυτοῦ (64), ἔχομ’ (ι) ἄν, γένοιτ’(ο) ἄν.
- a. Elision is often not expressed to the eye except in poetry. Both inscriptions and the Mss. of prose writers are very inconsistent, but even where the elision is not expressed, it seems to have occurred in speaking; i.e. ὅδε εἶπε and ὅδ’ εἶπε were spoken alike. The Mss. are of little value in such cases.
- Elision affects only unimportant words or syllables, such as particles, adverbs, prepositions, and conjunctions of two syllables (except περί, ἄχρι, μέχρι, ὅτι 72 b, c), and the final syllables of nouns, pronouns, and verbs. a. The final vowel of an emphatic personal pronoun is rarely elided.
- Elision does not occur in
- a. Monosyllables, except such as end in ε (τέ, δέ, γέ).
- b. The conjunction ὅτι that ὅτ’ is ὅτε when).
- c. The prepositions πρό before, ἄχρι, μέχρι until, and περί concerning (except before ι).
- d. The dative singular ending ι of the third declension, and in σι, the ending of the dative plural.
- e. Words with final υ.
- Except ἐστί is, forms admitting movable ν (134 a) do not suffer elision in prose. (But some cases of ε in the perfect occur in Demosthenes.)
- αι in the personal endings and the infinitive is elided in Aristophanes; scarcely ever, if at all, in tragedy; its elision in prose is doubtful. οι is elided in tragedy in οἴμοι alas.
Second, elision was not specific to verse. As usual for this type of thing, Sydney Allen is the first place to go to:
The pronunciation of a final vowel and a following initial vowel,l each with its syllabic value, has the Latin title of 'hiatus'. But, as we know, this type of junction was generally avoided in Latin verse, except at strong pauses, i.e. at verse ends and infrequently at main caesurae. In prose, Quintilian (ix, 4, 33 if.) is not opposed to occasional hiatus (' nonnumquam hiulca etiam decent faciuntque ampliora quaedam, ut "pulchra oratione acta'''); Cicero (Or., 150, 152; cf. Her., iv, 18) seems less tolerant, but his practice does not altogether support his precept.
The grammarians speak of complete loss of the final syllable in such cases (e.g. †Marius Victorinus, K. vi, 66) ;2 and one (Sacerdos, K. vi, 448) actually cites sequences such as menincepto, monstrhor- (for mene incepto, monstrum horrendum). The general term given to such loss is 'elisio', corresponding to the Greek ἔκθλιψις, though the grammarians mostly refer to it as 'synaliphe' (συναλοιφή). Such 'elision' is specifically contrasted with 'contraction' †('episynaliphe' or συνεκφώμησις), as in aeripedem for aëripedem.
In spite of these statements, various modern writers have refused to believe that the final vowel could be completely lost in such cases, since this would be likely to obscure the meaning. This is not altogether a valid argument; towards the end of a word sounds tend to become more 'redundant', i.e. predictable in terms of what has already been uttered; and even in the case of grammatical inflexions the sense is often inferable from other factors in the context (it is thus very common in the IndoEuropean languages to find that the final syllable is phonetically weak and liable to assimilation, reduction, or loss). In the first hundred lines of the Aeneid, for example, it has been suggested that elision could cause ambiguity in only two cases-and that neither of these 'would perceptibly alter the meaning of the passage'. Moreover, Plautus (Cure., 691) seems deliberately to introduce an ambiguity by this means in order to pun on cum catello ut accubas and cum catella ut accubas.
W. Sidney Allen, Vox Latina 78–79
A fun little example is Tablet 234 of the Vindolanda tablets records a soldier writing home to buy good shoes because of the bothersome "storms and winter". In Latin, that's tempestates et hiems (or hiemps), but this fellow wrote (or more likely dictated to a scribe) tempestates etiam.
I quote the online digital commentary on it:
The reading of tempestates is supported by the erasure [[et hiem]] which indicates that the scribe began to write et hiemes. TLL VI 2774.12-20 notes that grammarians and glossaries equate hiems and tempestas and gives the meaning uentus fortis (s.v.IA); the only citation with the words used together is Rufinus, Hist.mon.epil. 7, p.462a. The supralinear substitution of etiam strongly suggests that this was a phonetic dictation error (for which see Milne and Skeat (1938), 52, 55, and cf. Turner (1987), p.17).
However, this is for common speech. There is a debate whether this occurs in e.g. the orations of Cicero. Cicero himself (and others) remarks on the matter.
We are not allowed to make a pause between vowels, even if we should wish to do so. This is shown by those famous though slightly uncouth speeches of Cato.
Cic. Orat. 152
Cato, it should be remembered, wrote oratory. He continues:
Consonants were frequently omitted in contractions, for greater smoothness, as, for example, multi' modis, in vas' argenteis, palm et crinibus, tecti' fractis...Moreover, words are often contracted not for convenience but merely for the sake of sound. How was the name of your ancestor changed from Axilla to Ala except from a desire to avoid a harsh-sounding letter? The same letter is removed by refined Latin speech from maxillae, taxilli, vexillum and pauxillum. They were also ready to combine words by blending, e.g. sodes for si audes, and sis for si vis. You may even find three words in one—capsis.
Cic. Orat. 153–154; these passages translated by Hubbell for the LCL
This indicates that indeed elision/contraction was common. In fact, a little further down he questions why some were against it despite its regular use in antiquitas.
For further reading, I recommend Andrew M. Riggsby 1991, "Elision and Hiatus in Latin Prose," Classical Antiquity 10.2: 328–343.
Likewise, elision wasn't uncommon in certain Greek prose works:
ἀλλ᾽ ἦ, τὸ λεγόμενον, κατόπιν ἑορτῆς ἥκομεν καὶ ὑστεροῦμεν;
Plato Gorgias 447a
