Latin phonology and orthography

Latin phonology continually evolved over the centuries, making it difficult for speakers in one era to know how Latin was spoken in prior eras. A given phoneme may be represented by different letters in different periods. This article deals primarily with modern scholarship's best reconstruction of Classical Latin's phonemes and the pronunciation and spelling used by educated people in the late Roman Republic. This article then touches upon later changes and other variants. Knowledge of how Latin was pronounced comes from Roman grammar books, common misspellings by Romans, transcriptions into other ancient languages, and from how pronunciation has evolved in derived Romance languages.
Latin orthography is the spelling of Latin words written in the scripts of all historical phases of Latin from Old Latin to the present. All scripts use the Latin alphabet, but conventional spellings may vary from phase to phase. The Roman alphabet, or Latin alphabet, was adapted from the Old Italic script to represent the phonemes of the Latin language. The Old Italic script had in turn been borrowed from the Greek alphabet, itself adapted from the Phoenician alphabet. The Latin alphabet most resembles the Greek alphabet around 540 BC, as it appears on the black-figure pottery of the time.

Letterforms

The forms of the Latin alphabet used during the Classical period did not distinguish between upper case and lower case. Roman inscriptions typically use Roman square capitals, which resemble modern capitals, and handwritten text often uses old Roman cursive, which includes letterforms similar to modern lowercase.
This article uses small caps for Latin text, representing Roman square capitals, and long vowels are marked with acutes, representing apices. In the tables below, Latin letters and digraphs are paired with the phonemes they usually represent in the International Phonetic Alphabet.

Letters and phonemes

In ancient Latin spelling, individual letters mostly corresponded to individual phonemes, with three main exceptions:

The vowel letters a, e, i, o, u, y represented both short and long vowels. The long vowels were often marked by apices during the Classical period ⟨Á É Ó V́ Ý⟩, and long i was written using a taller version ⟨I⟩, called i longa "long I": ⟨ꟾ⟩; but now long vowels are sometimes written with a macron in modern editions, while short vowels are marked with a breve in dictionaries when necessary.
Some pairs of vowel letters, such as ae, represented either a diphthong in one syllable or two vowels in adjacent syllables.
The letters i and u - v represented either the close vowels and or the semivowels and.

In the tables below, Latin letters and digraphs are paired with the phonemes that they usually represent in the International Phonetic Alphabet.

Consonants

This is a table of the consonant sounds of Classical Latin. Sounds in parentheses are allophones, sounds with an asterisk exist mainly in loanwords and sounds with a dagger are phonemes only in some analyses.

Vowels

Monophthongs

Latin has ten native vowels, spelled a, e, i, o, u. In Classical Latin, each vowel had short and long versions: and. The long versions of the close and mid vowels e, i, o, u had a different vowel quality from the short versions, so that long were similar to short. Some loanwords from Greek had the vowel y, which was pronounced as by educated speakers but approximated with the native vowels u and i by less educated speakers.

Long and short vowels

Each vowel letter represents at least two phonemes. a can represent either short or long, e represents either or, etc.
Short mid vowels and close vowels were pronounced with a different quality from their long counterparts, being also more open:,, and. This opening made the short vowels i u similar in quality to long é ó respectively. i é and u ó were often written in place of each other in inscriptions:

trebibos for tribibus
minsis for mensis
sob for sub
punere for pōnere

Short most likely had a more open allophone before and tended toward near-open.
Short and were probably pronounced closer when they occurred before another vowel. mea was written as mia in inscriptions. Short before another vowel is often written with i longa, as in dīes, indicating that its quality was similar to that of long and is almost never confused with e in this position.

Adoption of Greek upsilon

y was used in Greek loanwords with upsilon Υ. This letter represented the close front rounded vowel, both short and long:. Latin did not have this sound as a distinctive phoneme, and speakers tended to pronounce such loanwords with in Old Latin and in Classical and Late Latin if they were unable to produce.

An intermediate vowel sound, called sonus medius, can be reconstructed for the classical period. Such a vowel is found in documentum, optimus, lacrima and other words. It developed out of a historical short, later fronted by vowel reduction. In the vicinity of labial consonants, this sound was not as fronted and may have retained some rounding. It was sometimes spelled by the claudian letter Ⱶ ⱶ.

Vowel nasalization

Vowels followed by a nasal consonant were allophonically realised as long nasal vowels in two environments:

Before word-final m:
* monstrum >.
* dentem >
Before nasal consonants followed by a fricative:
* censor >
* consul >
* inferōs >

Those long nasal vowels had the same quality as ordinary long vowels. In Vulgar Latin, the vowels lost their nasalisation, and they merged with the long vowels. This is shown by many forms in the Romance languages, such as Spanish costar from Vulgar Latin cōstāre and Italian mese from Vulgar Latin mēse. On the other hand, the short vowel and was restored in French enseigne and enfant from insignia and infantem, likely by analogy with other forms beginning in the prefix in-.
When a final -m occurred before a plosive or nasal in the next word, however, it was pronounced as a nasal at the place of articulation of the following consonant. For instance, tan dūrum was written for tam dūrum in inscriptions, and cum nōbīs was a double entendre, possibly for cunnō bis.

Diphthongs

ae, oe, au, ei, eu could represent diphthongs: ae represented, oe represented, au represented, ei represented, and eu represented. ui sometimes represented the diphthong, as in cui and huic. The diphthong ei mostly had changed to ī by the classical epoch; ei remained only in a few words such as the interjection hei.
If there is a tréma above the second vowel, both vowels are pronounced separately: aë, aü, eü and oë. However, disillabic eu in morpheme borders is traditionally written without the tréma: meus 'my'.
In Old Latin, ae, oe were written as ai, oi and probably pronounced as, with a fully closed second element, similar to the final syllable in French. In the late Old Latin period, the last element of the diphthongs was lowered to, so that the diphthongs were pronounced and in Classical Latin. They were then monophthongized to and, starting in rural areas at the end of the Republican period. The process, however, does not seem to have been completed before the 3rd century AD in Vulgar Latin, and some scholars say that it may have been regular by the 5th century. Cross-linguistically, the diphthong is rare, but it is frequent in e.g. the Khmer language, where it occurs even in the self-designation ខ្មែរ, which is pronounced.

Vowel and consonant length

Vowel and consonant length were more significant and more clearly defined in Latin than in modern English. Length is the duration of time that a particular sound is held before proceeding to the next sound in a word. In the modern spelling of Latin, especially in dictionaries and academic work, macrons are frequently used to mark long vowels:, while the breve is sometimes used to indicate that a vowel is short:.
Long consonants were usually indicated through doubling, but ancient Latin orthography did not distinguish between the vocalic and consonantal uses of i and v. Vowel length was indicated only intermittently in classical sources and even then through a variety of means. Later medieval and modern usage tended to omit vowel length altogether. A short-lived convention of spelling long vowels by doubling the vowel letter is associated with the poet Lucius Accius. Later spelling conventions marked long vowels with an apex or, in the case of long i, by increasing the height of the letter ; in the second century AD, those were given apices as well. Distinctions of vowel length had become less important in later Latin and have ceased to be phonemic in the modern Romance languages, in which the previous long and short versions of the vowels have been either lost or replaced by other phonetic contrasts. However, distinctions of consonant length is still phonemic in the Romance language of Italian, as the Italian word nono means "the ninth" while the word nonno means "grandfather".
A minimal set showing both long and short vowels and long and short consonants is ānus , annus , anus .

Table of orthography

The letters b, d, f, h, m, n are always pronounced as in English,,,,, respectively, and they do not usually cause any difficulties. The exceptions are mentioned below:

Latin grapheme	Latin phone	English approximation
,		Always hard as k in sky, never soft as in cellar, cello, or social. is a letter coming from Greek, but seldom used and generally replaced by c.
		As ch in chemistry, and aspirated; never as in challenge or change. Transliteration of Greek ⟨χ⟩.
		Always hard as g in good, never soft as g in gem.
		As ngn in wingnut.
		Sometimes at the beginning of a syllable, as y in yard, never as j in just.
		Doubled between vowels, as y y in toy yacht.
		When doubled or before, as clear l in link.
		In all other positions, as dark l in bowl.
		As n in man.
		As p in spy, never aspirated.
		As p in party, always aspirated; never as in photo when being pronounced in English. Transliteration of Greek ⟨φ⟩.
		Similar to qu in quick, never as qu in antique.
		There were two trends: the educated and popular pronunciation. Within educated circles it was pronounced, evoking the Old Latin pronunciation ; meanwhile, within popular circles it was pronounced .
		As r in Italian and several Romance languages.
		As r in Italian and several Romance languages ; e.g. diarrhoea ⟨διάῤῥοια⟩; but it is voiceless.
		As s in say, never as s in rise or measure.
		As t in stay, never as t in nation.
		As th in thyme, and aspirated; never as in thing, or that. Transliteration of Greek ⟨θ⟩.
		Sometimes at the beginning of a syllable, or after and, as w in wine, never as v in vine.
		as in some English accents one is pronounced, but without the nasal sound: parvus, vivunt. The spelling vu is post-classical.
		A letter representing +, as well as + : as x in English axe, never as x in example.
		As in zoom, never as in pizza. Transliteration of Greek ⟨ζ⟩.

Latin grapheme	Latin phone	English approximation
		As in aisle. Transliteration of Greek ⟨αι⟩.
		As in out. Transliteration of Greek ⟨αυ⟩.
		As to ey in they. Transliteration of Greek ⟨ει⟩ in some cases.
		As in Portuguese eu, approximately to the British pronunciation of low. Transliteration of Greek ⟨ευ⟩.
		As in boy. Transliteration of Greek ⟨οι⟩.
		As in Spanish muy, approximately to hooey.
		Transliteration of the Greek diphthong ⟨υι⟩.

Syllables and stress

Old Latin stress

In Old Latin, as in Proto-Italic, stress normally fell on the first syllable of a word. During this period, the word-initial stress triggered changes in the vowels of non-initial syllables, the effects of which are still visible in classical Latin. Compare for example:

faciō 'I do/make', factus 'made'; pronounced and in later Old Latin and Classical Latin.
afficiō 'I affect', affectus 'affected'; pronounced and in Old Latin following vowel reduction, and in Classical Latin.

In the earliest Latin writings, the original unreduced vowels are still visible. Study of this vowel reduction, as well as syncopation in Greek loan words, indicates that the stress remained word-initial until around the time of Plautus, in the 3rd century BC. The placement of the stress then shifted to become the pattern found in classical Latin.

Classical Latin syllables and stress

In Classical Latin, stress changed. It moved from the first syllable to one of the last three syllables, called the antepenult, the penult, and the ultima. Its position is determined by the syllable weight of the penult. If the penult is heavy, it is accented; if the penult is light and there are more than two syllables, the antepenult is accented. In a few words originally accented on the penult, accent is on the ultima because the two last syllables have been contracted, or the last syllable has been lost.

Syllable

To determine stress, syllable weight of the penult must be determined. To determine syllable weight, words must be broken up into syllables. In the following examples, syllable structure is represented using these symbols: C, K, R, and V, VV.

Nucleus

Every short vowel, long vowel, or diphthong belongs to a single syllable. This vowel forms the syllable nucleus. Thus magistrārum has four syllables, one for every vowel, aereus has three, tuō has two, and cui has one.

Onset and coda

A consonant before a vowel, or a consonant cluster at the beginning of a word, is placed in the same syllable as the following vowel. This consonant or consonant cluster forms the syllable onset.

fēminae
vidēre
puerō
beātae
graviter
strātum

After this, if there is an additional consonant inside the word, it is placed at the end of the syllable. This consonant is the syllable coda. Thus if a consonant cluster of two consonants occurs between vowels, they are broken up between syllables: one goes with the syllable before, the other with the syllable after.

puella
supersum
coāctus
intellēxit

There are two exceptions. A consonant cluster of a stop p t c b d g followed by a liquid l r between vowels usually goes to the syllable after it, although it is also sometimes broken up like other consonant clusters.

volucris or
Heavy and light syllables

As shown in the examples above, Latin syllables have a variety of possible structures. Here are some of them. The first four examples are light syllables, and the last six are heavy. All syllables have at least one V. A syllable is heavy if it has another V or a VC after the first V. In the table below, the extra V or VC is bolded, indicating that it makes the syllable heavy.
Thus, a syllable is heavy if it ends in a long vowel or diphthong, a short vowel and a consonant, a long vowel and a consonant, or a diphthong and a consonant. Syllables ending in a diphthong and consonant are rare in Classical Latin.
The syllable onset has no relationship to syllable weight; both heavy and light syllables can have no onset or an onset of one, two, or three consonants.
In Latin a syllable that is heavy because it ends in a long vowel or diphthong is traditionally called syllaba nātūrā longa, and a syllable that is heavy because it ends in a consonant is called positiōne longa. These terms are translations of Greek συλλαβὴ μακρά φύσει and μακρὰ θέσει, respectively; therefore positiōne should not be mistaken for implying a syllable "is long because of its position/place in a word" but rather "is treated as 'long' by convention". This article uses the words heavy and light for syllables, and long and short for vowels since the two are not the same.

Stress rule

In a word of three or more syllables, the weight of the penult determines where the accent is placed. If the penult is light, accent is placed on the antepenult; if it is heavy, accent is placed on the penult. Below, stress is marked by placing the stress mark before the stressed syllable.

volucris	fēminae	puerō

CV.CV.CCVC	CVV.CV.CVV	CV.V.CVV

Iambic shortening

Iambic shortening or brevis brevians is vowel shortening that occurs in words of the type light–heavy, where the light syllable is stressed. By this sound change, words like egō, modō, benē, amā with long final vowel change to ego, modo, bene, ama with short final vowel.

Elision

Where one word ended with a vowel and the next word began with a vowel, the former vowel, at least in verse, was regularly elided; that is, it was omitted altogether, or possibly pronounced like the corresponding semivowel. When the second word was est or et, a different form of elision sometimes occurred : the vowel of the preceding word was retained, and the e was elided instead. Elision also occurred in Ancient Greek, but in that language, it is shown in writing by the vowel in question being replaced by an apostrophe, whereas in Latin elision is not indicated at all in the orthography, but can be deduced from the verse form. Only occasionally is it found in inscriptions, as in scriptust for scriptum est.

Modern conventions

Spelling

Modern usage, even for classical Latin texts, varies in respect of I and V. During the Renaissance, the printing convention was to use I and i for both vocalic and consonantal, to use V in the upper case and in the lower case to use v at the start of words and u subsequently within the word regardless of whether and was represented.
Many publishers have adopted the convention of using I and i for both and, and V and u for both and.
An alternative approach, less common today, is to use i and u only for the vowels and j and v for the approximants.
Most modern editions, however, adopt an intermediate position, distinguishing between u and v but not between i and j. Usually, the non-vocalic v after q or g is still printed as u rather than v, probably because in this position it did not change from to in post-classical times.
Textbooks and dictionaries usually indicate the length of vowels by putting a macron or horizontal bar above the long vowel, but it is not generally done in regular texts. Occasionally, mainly in early printed texts up to the 18th century, one may see a circumflex used to indicate a long vowel where this makes a difference to the sense, for instance, Româ compared to Roma . Sometimes, for instance in Roman Catholic service books, an acute accent over a vowel is used to indicate the stressed syllable. It would be redundant for one who knew the classical rules of accentuation and made the correct distinction between long and short vowels, but most Latin speakers since the 3rd century have not made any distinction between long and short vowels, but they have kept the accents in the same places; thus, the use of accent marks allows speakers to read a word aloud correctly even if they never heard it spoken aloud.

Pronunciation

Post-Medieval Latin

Since around the beginning of the Renaissance period onwards, with the language being used as an international language among intellectuals, pronunciation of Latin in Europe came to be dominated by the phonology of local languages, resulting in a variety of different pronunciation systems. See the article Latin regional pronunciation for more details on those.

Loan words and formal study

When Latin words are used as loanwords in a modern language, there is ordinarily little or no attempt to pronounce them as the Romans did; in most cases, a pronunciation suiting the phonology of the receiving language is employed.
Latin words in common use in English are generally fully assimilated into the English sound system, with little to mark them as foreign, for example, cranium, saliva. Other words have a stronger Latin feel to them, usually because of spelling features such as the digraphs ae and oe, which both denote in English. The digraph ae or ligature æ in some words tend to be given an pronunciation, for example, curriculum vitae.
However, using loan words in the context of the language borrowing them is a markedly different situation from the study of Latin itself. In this classroom setting, instructors and students attempt to recreate at least some sense of the original pronunciation. What is taught to native anglophones is suggested by the sounds of today's Romance languages, the direct descendants of Latin. Instructors who take this approach rationalize that Romance vowels probably come closer to the original pronunciation than those of any other modern language.
However, other languages—including Romance family members—all have their own interpretations of the Latin phonological system, applied both to loan words and formal study of Latin. But English, Romance, or other teachers do not always point out that the particular accent their students learn is not actually the way ancient Romans spoke.

Ecclesiastical pronunciation

Because of the central position of Rome within the Catholic Church, an Italian pronunciation of Latin became commonly accepted, but this was not the case until the latter part of the 19th century. This pronunciation corresponds to that of the Latin-derived words in Italian. Before then, the pronunciation of Latin in church was the same as the pronunciation as Latin in other fields and tended to reflect the sound values associated with the nationality of the speaker. Other ecclesiastical variations are still in use, especially outside the Catholic Church.
The following are the main points that distinguish modern Italianate ecclesiastical pronunciation from Classical Latin pronunciation:
The letters b, d, f, m, n are always pronounced as in English,,,, respectively, and they do not usually cause any difficulties. The exceptions are mentioned below:

Latin grapheme	Latin phone	English approximation
hard,,		As in sky before a, o, u.
soft		As in change before ae, e, i, oe, y.
hard		As in gate before a, o, u.
soft		As in gem before ae, e, i, oe, y.
		As gn in Italian, or ñ in Spanish.
	∅,	Usually mute; occasionally in words like "mihi" and "nihil".
		Sometimes at the beginning of a syllable, as y in yard, never as j in just.
		Doubled between vowels, as y y in toy yacht.
		l exilis and l pinguis merge.
		As n in man.
		As p in spy, never aspirated.
		As in photo when being pronounced in English.
		Similar to qu in quick, never as qu in antique.
,		As r in Italian and several Romance languages.
		Strictly as s in hiss as in the Spanish of the Americas, though increasingly inter-vocally pronounced -s- as in the Italian language. Pronounced when being -ss-.
,		As sh in shade.
,		As t in stay, never as t in nation.
		Only if followed by a vowel, not word-initial or stressed, and not preceded by s, t, or x.
		In Ecclesiastical pronunciation, only represents a consonant.
		A letter representing +, as well as + : as x in English axe. is algo a common allophone intervocalically.
,		Used in words like excellens.
		Never as in pizza. Transliteration of Greek ⟨ζ⟩.

Latin grapheme	Latin phone	English approximation
		As in out. Transliteration of Greek ⟨αυ⟩.
		Similar to ey in they. Transliteration of Greek ⟨ει⟩ in some cases.
		As in Portuguese eu, similar to the British pronunciation of low. Transliteration of Greek ⟨ευ⟩.
		As in Spanish muy, approximately to hooey.

Vowel length is not phonemic. As a result, the automatic stress accent of Classical Latin, which was dependent on vowel length, becomes a phonemic one in Ecclesiastical Latin.
Word-final m and n are pronounced fully, with no nasalization of the preceding vowel.

In his Vox Latina: A guide to the Pronunciation of Classical Latin, William Sidney Allen remarked that this pronunciation, used by the Catholic Church in Rome and elsewhere, and whose adoption Pope Pius X recommended in a 1912 letter to the Archbishop of Bourges, "is probably less far removed from classical Latin than any other 'national' pronunciation"; but, as can be seen from the table above, there are, nevertheless, very significant differences. The introduction to the Liber Usualis indicates that Ecclesiastical Latin pronunciation should be used at Church liturgies. The Pontifical Academy for Latin is the pontifical academy in the Vatican that is charged with the dissemination and education of Catholics in the Latin language.
Outside of Austria, Germany, Czechia and Slovakia, it is the most widely used standard in choral singing which, with a few exceptions like Stravinsky's Oedipus rex, is concerned with liturgical texts. Anglican choirs adopted it when classicists abandoned traditional English pronunciation after World War II. The rise of historically informed performance and the availability of guides such as Copeman's Singing in Latin has led to the recent revival of regional pronunciations.

Pronunciation shared by Vulgar Latin and Romance languages

Because it gave rise to many modern languages, Latin did not "die"; it merely evolved over the centuries in different regions in diverse ways. The local dialects of Vulgar Latin that emerged eventually became modern Italian, Spanish, French, Romanian, Portuguese, Catalan, Romansh, Dalmatian, Sardinian, and many others.
Key features of Vulgar Latin and Romance languages include:

Almost total loss of and final unstressed.
Conversion of the distinction of vowel length into a distinction of height, and subsequent merger of some of these phonemes. Most Romance languages merged short with long and short with long.
Monophthongization of into and into.
Loss of marginal phonemes such as aspirates, which became tenues, and the close front-rounded vowel, which became unrounded.
Loss of before but this influence on the later development of Romance languages was limited from written influence, analogy, and learned borrowings.
Palatalization of before and , probably first into and then before it finally developed into or.
Palatalization of before and, and of, into and then further into in some Romance varieties.
Palatalization of followed by a vowel into. It merged with in dialects in which had developed into this sound, but it remained separate elsewhere.
Palatalization of and followed by a vowel into and. also coalesced to become.
The change of and between vowels, into.
Examples

The following examples are both in verse, which demonstrates several features more clearly than prose.

From Classical Latin

's Aeneid, Book 1, verses 1–4. Quantitative metre. Translation: "I sing of arms and the man, who, driven by fate, came first from the borders of Troy to Italy and the Lavinian shores; he much afflicted both on lands and on the deep by the power of the gods, because of fierce Juno's vindictive wrath."

Ancient Roman orthography
:ARMA·VIRVMQVE·CANÓ·TRÓIAE·QVꟾ·PRꟾMVS·ABÓRꟾS
:ꟾTALIAM·FÁTÓ·PROFVGVS·LÁVꟾNIAQVE·VÉNIT
:LꟾTORA·MVLTVM·ILLE·ETTERRꟾS·IACTÁTVS·ETALTÓ
:Vꟾ·SVPERVM·SAEVAE·MEMOREM·IV́NÓNIS·OBꟾRAM