Phonological changes from Classical Latin to Proto-Romance

General changes

• is lost without a trace in all positions.
• *If that results in a collision of identical short vowels, they fuse into the corresponding long vowel, as in >.
• Final is lost without a trace in polysyllabic words, as in >.
• *In monosyllables it tends to survive as, as in > > Spanish quién.
• Clusters consisting of a stop followed by a liquid consonant draw the stress position forward, as in >.
• *Two apparent counterexamples are and, judging by the Old French outcomes palpres and poltre.
• is lost before fricatives, leaving the preceding vowel lengthened, as in >.
• * is often retained or later restored if it belongs to a prefix or to a word which has forms where a fricative does not follow, as in > French défense, thanks to related forms such as the infinitive > French défendre.
• Sequences of two generally merge to a single long, as in >.
• In some outlying rural areas, the diphthongs and reduce to and respectively in Classical times. Influence from such dialects made a number of Latin words acquire monophthongized variants early on, as in or. Most words, however, remain unaffected by this.
• *Later, 'mainstream' Latin experiences a general monophthongization of to, and of to, and remains intact in most cases, as in >.
• turns to the fricative, as does original in intervocalic position, as in >.
• *Intervocalic in contact with a rounded vowel tends to disappear, as in > >.
• **It is often restored if other forms of the word have a non-rounded vowel following, as in the nominative plural.
• In hiatus, unstressed front vowels become, and unstressed back vowels become, as in >.
• *The same process also affects stressed front and back vowels in hiatus if they are antepenultimate. When is produced, primary stress shifts to the following vowel, but when is produced, primary stress shifts instead to the preceding syllable, as in >.
• *If is formed after a geminate consonant, it is deleted, as in > >.
• * is deleted before unstressed back vowels, as in > >.
• ** is occasionally deleted before unstressed non-back vowels as well, as in > >.
• **Similarly, is delabialized to before back vowels, whether they are stressed or not, as in >.
• *If those changes result in sequences of or, they merge to and respectively, as in > >.
• *If forms after, the resulting simplifies and delabializes to, as in > >.
• raises before or, as in > > Italian cui, fui.
• before vocalizes to, as in >.
• is reduced to before or after a consonant, as in >, or at the end of words of more than one syllable.
• *Intervocalically, it sometimes metathesizes to, as in >.
• Words beginning with receive an initial supporting vowel, unless they are preceded by a word ending in a vowel, as in >.
• *The earliest unambiguous attestations occur in inscriptions of the second century AD. In some Romance languages, such as Spanish, word-initial remains phonologically forbidden to this day. In other languages, such as Romanian, the supporting vowel seems to have been abandoned early on, resulting in restoration of initial. Although there is barely any direct inscriptional evidence of the supporting vowel in Latin inscriptions in the Balkans, its development and subsequent loss is considered to be indirectly attested by the dropping of word-initial before in cases in which it was not originally a supporting vowel, as in Romanian spulbera 'to dust', from *ex-pulverāre. Compare also > * > Italian scala, scadere; French échelle, échoir.
• and before are raised, respectively, to and, as in > > Italian biscia, uscio.
• Compound verbs stressed on a prefix are usually reconstructed according to their prefixless equivalent, with their stress shifted forward from the prefix, as in > *, by analogy with the simplex form.
• * simply yields, perhaps because the verb, while recognisable as a compound, was not easy to identify with the original.
• *Some words such as 'fasten' are apparently not recognised as compounds at all and so remain unchanged.
• Monosyllabic nouns ending in a consonant receive an epenthetic final, as in > > > French rien.
• Phonemic vowel length gradually collapses via the following changes :
• *Long vowels shorten in unstressed syllables.
• *Long vowels shorten in stressed closed syllables.
• *Short vowels lengthen in stressed open syllables.
• On account of the above, the vowel inventory changes from to, with pre-existing differences in vowel quality achieving phonemic status and with no distinction between original and. Additionally:
• *Unstressed and merge into and respectively.
• *In the second syllable of words with the structure, and merge into and respectively.
• Word-internal merges into a preceding consonant and palatalises it, as in > > > Italian.

Sporadic changes

• Vowels other than are often syncopated in unstressed word-internal syllables, especially in contact with liquid consonants or, to a lesser extent, nasal consonants or, as in >.
• *In a few words, unstressed initial syllables followed by experience syncope, as in >.
• *If syncope results in being followed by a consonant, it may vocalize to, as in > > * > Italian fòla.
• *If syncope results in, the cluster is generally replaced by, as in >.
• In cases where a long vowel precedes a geminate consonant, one of the elements often shortens unpredictably, sometimes leading to such doublets as > > Spanish copa, cuba; French coupe, cuve.
• *Long vowels sometimes shorten early on in closed syllables even if followed by two different consonants, leading to variations such as > > Italian undici, Spanish once.
• *Conversely, the cluster may lengthen preceding vowels early on, as in > >.
• Pretonic vowels sporadically assimilate to or dissimilate from the stressed vowel of the following syllable.
• * can dissimilate to before a following, as in >.
• * can dissimilate to before a following, as in >.
• * can dissimilate to before a following, as in >.
• * can dissimilate to before a following back vowel, as in >.
• * can assimilate to a following, as in >.
• * can assimilate to a following, as in * > *.
• * can assimilate to a following, as in > *.
• and may yield a low-mid vowel if followed by, as in > > Italian uovo, Sardinian colòra.
• may yield a mid vowel if preceded by, as in > *.
• assimilates to a following in a number of cases, as in >.
• *After a long vowel, the resulting reduces to, as in >.
• Initial and sometimes voice, as in >.
• *This is particularly frequent with borrowings from Greek. κρυπτή, καμπή > * > Italian grotta, gamba.
• sometimes assimilates to, as in the alternation grundīre~grunnīre.
• There is occasional loss or assimilation of final, but it is nowhere regular until a much later period.
• When two neighbouring syllables each contain, one frequently dissimilates to or is deleted.