Ancient Greek accent


The Ancient Greek accent is believed to have been a melodic or pitch accent. One of the final three syllables of an Ancient Greek word carried an accent. Each syllable contains a vowel with one or two vocalic morae, and one mora in a word is accented; the accented mora was pronounced at a higher pitch than other morae.
The accent cannot come more than three syllables from the end of the word. If the last syllable of a word has a long vowel, or is closed by two consonants, the accent usually cannot come on the antepenultimate syllable; but within those restrictions it is free.
In nouns the accent is largely unpredictable. Mostly the accent either comes as close to the beginning of the word as the rules allow, for example, πόλεμος 'war', or it is placed on the last mora of the word, as in ποταμός 'river'. But in a few words, such as παρθένος 'maiden', the accent comes between these two extremes.
In verbs the accent is generally predictable and has a grammatical rather than a lexical function, that is, it differentiates different parts of the verb rather than distinguishing one verb from another. Finite parts of the verb usually have recessive accent, but in some tenses participles, infinitives, and imperatives are non-recessive.
In the classical period word accents were not indicated in writing, but from the 2nd century BC onwards various diacritic marks were invented, including an acute, circumflex, and grave accent, which indicated a high pitch, a falling pitch, and a low or semi-low pitch respectively. The written accents were used only sporadically at first, and did not come into common use until after 600 AD.
The fragments of ancient Greek music that survive, especially the two hymns inscribed on a stone in Delphi in the 2nd century BC, appear to follow the accents of the words very closely, and can be used to provide evidence for how the accent was pronounced.
Sometime between the 2nd and 4th centuries AD the distinction between acute, grave, and circumflex disappeared and all three accents came to be pronounced as a stress accent, generally heard on the same syllable as the pitch accent in ancient Greek.

Types of accent

The ancient Greek grammarians indicated the word-accent with three diacritic signs: the acute, the circumflex, and the grave. The acute was the most commonly used of these; it could be found on any of the last three syllables of a word. Some examples are:
The circumflex, which represented a falling tone, is found only on long vowels or on diphthongs, and only on the final or penultimate syllable of the word:
The grave is found, as an alternative to an acute, only on the last syllable of a word. When a word such as ἀγαθός 'good' with final accent is followed by a pause, or by an enclitic word such as the weak form of ἐστίν 'is', the accent is written as an acute:
However, when the word does not come before a pause or an enclitic, the acute accent is replaced by a grave:
It is generally assumed that when a word was written with a grave it indicates that there was no rise in pitch, or only a small one.

Placing the accent marks

In Greek, if an accent mark is written on a diphthong or vowel written with a digraph such as ει, it is always written above the second vowel of the diphthong, not the first, for example:
When a word such as a proper name starts with a capital vowel letter, the accent and breathing are usually written before the letter. If a name starts with a diphthong, the accent is written above the second letter. But in ᾍδης 'Hades', where the diphthong is the equivalent of an alpha with iota subscript, it is written in front:
When combined with a rough or smooth breathing, the circumflex goes on top of the breathing, while the acute or grave is written to the right of the breathing, as in the above examples. When an accent is combined with a diaeresis mark, as in νηΐ, the accent is written on top.

Tonal minimal pairs

Whether the accent on a particular syllable is an acute or circumflex is largely predictable, but there are a few examples where a change from an acute on a long vowel to a circumflex indicates a different meaning, for example
There are also examples where the meaning changes if the accent moves to a different syllable:
There is also a distinction between unaccented and fully accented forms in words such as:
The three marks used to indicate accent in ancient Greek, the acute, circumflex, and grave are said to have been invented by the scholar Aristophanes of Byzantium, who was head of the famous library of Alexandria in Egypt in the early 2nd century BC. The first papyri with accent marks date from this time also. In the papyri, at first the accents were used only sporadically, specifically for helping readers to pronounce Greek poetry correctly, and the grave accent could be used on any non-accented syllable. Such accents were useful, since Greek at that time was written without gaps between the words. For example, in one papyrus, the word ὸρὲιχάλκωι 'to brass' is written with grave accents on the first two syllables, in case any reader should mistakenly read the first part of the word as ὄρει 'to a mountain'.
In the following centuries many other grammarians wrote about Greek accentuation. The most famous of these, Aelius Herodianus or Herodian, who lived and taught in Rome in the 2nd century AD, wrote a long treatise in twenty books, 19 of which were devoted to accentuation. Although Herodian's book does not survive in full, an epitome was made of it around AD 400 which still survives. Another important authority was Apollonius Dyscolus, the father of Herodian.
The names of these diacritics in English, and the term ‘accent’, are based on Latin loan-translations of the Greek terms. Latin accentus corresponds to Greek προσῳδία prosōdía "song sung to instrumental music, pitch variation in voice", acūtus to ὀξεῖα oxeîa "sharp" or "high-pitched", gravis to βαρεῖα bareîa "heavy" or "low-pitched", and circumflexus to περισπωμένη perispōménē "pulled around" or "bent". The Greek terms for the diacritics are nominalized feminine adjectives that originally modified the feminine noun προσῳδία and agreed with it in gender.
Diacritic signs were not used in the classical period. They were gradually introduced from the 2nd century BC onwards, but did not become commonly used in manuscripts until after 600 AD.

Origin of the accent itself

The ancient Greek accent, at least in nouns, appears to have been inherited to a large extent from the original parent language from which Greek and many other European and Indian languages derive, Proto-Indo-European. This can be seen by comparing the accent of Greek words with the accent of words in the Vedic hymns. Very often these are the same, for example:
There are also other accentual correspondences between Greek and Vedic, for example:
One difference between Greek and Vedic, however, is that in Greek words the accent is always found in one of the last three syllables, whereas in Vedic it could come anywhere in the word.
The distinction in Greek between circumflex and acute accent appears to be a Greek development, and does not go back to Proto-Indo-European.

Terminology

In all there are exactly five different possibilities for placing an accent. The terms used by the ancient Greek grammarians were:
The word barytone refers to any word which has no accent on the final syllable, that is the 2nd, 3rd and 5th possibilities above.

Pronunciation of the accent

General evidence

It is generally agreed that the ancient Greek accent was primarily one of pitch or melody rather than of stress. Thus in a word like ἄνθρωπος 'man', the first syllable was pronounced on a higher pitch than the others, but not necessarily any louder. As long ago as the 19th century it was surmised that in a word with recessive accent the pitch may have fallen not suddenly but gradually in a sequence high–middle–low, with the final element always short.
The evidence for this comes from various sources. The first is the statements of Greek grammarians, who consistently describe the accent in musical terms, using words such as ὀξύς 'high-pitched' and βαρύς 'low-pitched'.
According to Dionysius of Halicarnassus, the melody of speech is confined to an interval 'of about a 5th'. This statement has been interpreted in different ways, but it is usually supposed that he meant not that it was always a fifth, but that this was the maximum normal difference between high and low syllables. It is thought probable that occasionally, especially at the end of a sentence, the interval was much smaller. Dionysius also describes how a circumflex accent combines high and low pitch on the same syllable, whereas with an acute accent the high and low pitches are in separate syllables.
Another indication that the accent was melodic or tonal is that in the classical period the accents of the words seem to have played no part at all in poetic metres, unlike in languages such as English which have stress-accents. It was not until the 4th century AD that poems began to be written in which the accent played a role.

Evidence from music

An important indication of the melodic nature of the Greek accent comes from the surviving pieces of Greek music, especially the two Delphic hymns, the Seikilos epitaph, and the hymns of Mesomedes. An example is the prayer to Calliope and Apollo written by Mesomedes, court musician to the Emperor Hadrian:
, leader of the delightful Muses, and you, wise initiator into the mysteries, Leto's son, Delian Healer, favour me with your presence."
As can be seen, the accented syllable of a word generally has the highest note within that word, although sometimes the syllables preceding or following the accent are also high.
When the accent is a circumflex, the music often shows a fall from a higher note to a lower one within the syllable itself, exactly as described by Dionysius of Halicarnassus; examples are the words Μουσῶν 'of the Muses' and εὐμενεῖς 'favourable' in the prayer illustrated above. However, sometimes there is no fall within the accented syllable, but the circumflex is set to a single note, as in τερπνῶν 'delightful' or Λατοῦς 'of Leto' above.
If the accent is a grave, there is often no rise in pitch, or else only a small one, as in σοφὲ above.
In this practice of closely imitating the tones of word accents in the melodies of songs, Ancient Greek resembles many living Asian and African languages that have tonal accents. For this reason, the American scholars A.M. Devine and Laurence Stephens have argued that the rises and falls found in Greek music probably give a reasonably good indication of what happened when the words were spoken.
It seems, however, that the music did not always follow the accent exactly. Dionysius of Halicarnassus gives an example from the music written by Euripides for his play Orestes. In the lines which in our modern editions are written as σῖγα, σῖγα, λεπτὸν ἴχνος ἀρβύλας // τίθετε, μὴ κτυπεῖτ᾽ 'Quietly, quietly! Place the tread of your shoe lightly, don't make a noise!', Dionysius reports that in the first three words and the last there was no raised pitch, while in both ἀρβύλας 'of the shoe' and τίθετε 'place' there was a low note followed by two high ones, despite the accent on the first syllable of τίθετε.
However, although the fragments of earlier music sometimes show a mismatch, the Delphic hymns in particular appear to show a very close relationship between the music and the word accents, with all but three of the 180 analysable words matching.
Some more details of the way in which accents were set to music are given below. Note that in the musical examples the pitch is conventional, dating back to a publication by Friedrich Bellermann in 1840. In performance the pitch would have been at least a minor third lower.

Acute accent

When the signs for the notes in Greek music are transcribed into modern musical notation, it can be seen that an acute accent is generally followed by a fall, sometimes extending over two syllables. Usually the fall is only a slight one, as in θύγατρες 'daughters', Ὄλυμπον 'Olympus' or ἔτικτε 'she give birth to' below. Sometimes, however, there is a sharp drop, as in μέλψητε 'you may sing' or νηνέμους 'windless':
Before the accent the rise on average is less than the fall afterwards. There is sometimes a jump up from a lower note, as in the word μειγνύμενος 'mingling' from the second hymn; more often there is a gradual rise, as in Κασταλίδος 'of Castalia', Κυνθίαν 'Cynthian', or ἀνακίδναται 'spreads upwards':
In some cases, however, before the accent instead of a rise there is a 'plateau' of one or two notes the same height as the accent itself, as in Παρνασσίδος 'of Parnassus', ἐπινίσεται 'he visits', Ῥωμαίων 'of the Romans', or ἀγηράτῳ 'ageless' from the Delphic hymns:
Anticipation of the high tone of an accent in this way is found in other pitch-accent languages, such as some varieties of Japanese, Turkish, or Serbian, where for example the word papríka 'pepper' can be pronounced pápríka. It would not be surprising therefore to find that it was a feature of Greek speech also. Devine and Stephens, however, quoting Dionysius's statement that there is only one high tone per word, argue that the norm in Greek words was for unaccented syllables to be low-pitched.
When an acute accent occurs on a long vowel or diphthong, it is generally assumed that the high pitch was on the second mora of the vowel, that is to say, that there was a rising pitch within the syllable. The Greek music sometimes shows exactly this, as with the word αἴθει 'it burns' in the 1st Delphic hymn, or φαίνου 'shine!' in the Seikilos epitaph, or Σελάνα 'the Moon' in the Hymn to the Sun, in which the syllable with the acute is set to a melism of two or three notes rising gradually.
More frequently, however, on an accented long vowel in the music there is no rise in pitch, and the syllable is set to a level note, as in the words Ἅφαιστος 'Hephaestus' from the 1st Delphic hymn or ἐκείνας 'those' or Ῥωμαίων 'of the Romans' from the 2nd hymn:
Because this is so common, it is possible that at least sometimes the pitch did not rise on a long vowel with an acute accent but remained level. Another consideration is that although the ancient grammarians regularly describe the circumflex accent as 'two-toned' or 'compound' or 'double', they usually do not make similar remarks about the acute. There are apparently some, however, who mention a 'reversed circumflex', presumably referring to this rising accent.

Tonal assimilation

Devine and Stephens note that occasionally at the end of a word, the pitch rises again, as though leading up to or anticipating the accent in the following word. They refer to this as a 'secondary rise'. Examples are ἔχεις τρίποδα 'you have a tripod' or μέλπετε δὲ Πύθιον 'sing the Pythian' in the 2nd Delphic hymn. According to Devine and Stephens, it 'probably reflects a genuine process of pitch assimilation in fluent speech'.
In the great majority of cases in the music, the pitch falls on the syllable immediately following an acute accent. However, there are some exceptions. One situation where this can happen is when two words are joined in a plateau or near-plateau, as in the phrases ἵνα Φοῖβον 'so that Phoebus' and πόλει Κεκροπίᾳ 'in the city of Cecrops' in the 2nd Delphic Hymn:
Tonal assimilation or tone sandhi between neighbouring tones is commonly found in tonal languages. Devine and Stephens, citing a similar phenomenon in the music of the Nigerian language Hausa, comment: 'This is not a mismatch but reflects a feature of phrase intonation in fluent speech.'

Circumflex accent

A circumflex was written only over a long vowel or diphthong. In the music, the circumflex is usually set to a melisma of two notes, the first higher than the second. Thus in the first Delphic Hymn the word Φοῖβον 'Phoebus' is set to the same musical notes as θύγατρες 'daughters' earlier in the same line, except that the first two notes fall within one syllable instead of across two syllables. Just as with the acute accent, a circumflex can be preceded either by a note on the same level, as in ᾠδαῖσι 'with songs', or by a rise, as in μαντεῖον 'oracular':
The circumflex therefore appears to have been pronounced in exactly the same way as an acute, except that the fall usually took place within one syllable. This is clear from the description of Dionysius of Halicarnassus, who tells us that a circumflex accent was a blend of high and low pitch in a single syllable, and it is reflected in the word ὀξυβάρεια 'high-low', which is one of the names given to the circumflex in ancient times. Another description was δίτονος 'two-toned'.
Another piece of evidence for the pronunciation of the circumflex accent is the fact that when two vowels are contracted into one, if the first one has an acute, the result is a circumflex: e.g. ὁρά-ω 'I see' is contracted to ὁρῶ with a circumflex, combining the high and low pitches of the previous vowels.
In the majority of examples in the Delphic hymns, the circumflex is set to a melisma of two notes. However, in Mesomedes' hymns, especially the hymn to Nemesis, it is more common for the circumflex to be set to a single note. Devine and Stephens see in this the gradual loss over time of the distinction between acute and circumflex.
One place where a circumflex can be a single note is in phrases where a noun is joined with a genitive or an adjective. Examples are μῆρα ταύρων 'thighs of bulls', Λατοῦς γόνε 'Leto's son', γαῖαν ἅπασαν 'the whole world'. In these phrases, the accent of the second word is higher than or on the same level as that of the first word, and just as with phrases such as ἵνα Φοῖβον mentioned above, the lack of fall in pitch appears to represent some sort of assimilation or tone sandhi between the two accents:
When a circumflex occurs immediately before a comma, it also regularly has a single note in the music, as in τερπνῶν 'delightful' in the Mesomedes' Invocation to Calliope illustrated above. Other examples are κλυτᾷ 'famous', ἰοῖς 'with arrows' in 2nd Delphic hymn, ζῇς 'you live' in the Seikilos epitaph, and θνατῶν, ἀστιβῆ and μετρεῖς in Mesomedes' Hymn to Nemesis.
Another place where a circumflex sometimes has a level note in the music is when it occurs in a penultimate syllable of a word, with the fall only coming in the following syllable. Examples are παῖδα, πᾶσι , λῆξε, σῷζε, and Φοῖβον , and χεῖρα, πῆχυν .

Grave accent

The third accentual mark used in ancient Greek was the grave accent, which is only found on the last syllable of words e.g. ἀγαθὸς ἄνθρωπος 'a good man'. Scholars are divided about how this was pronounced; whether it meant that the word was completely accentless or whether it meant a sort of intermediate accent is unclear. In some early documents making use of written accents, a grave accent could often be added to any syllable with low pitch, not just the end of the word, e.g. Θὲὸδὼρὸς.
Some scholars, such as the Russian linguist Nikolai Trubetzkoy, have suggested that because there is usually no fall after a grave accent, the rise in pitch which was heard at the end of a clause was phonologically not a true accent, but merely a default phrasal tone, such as is heard in languages like Luganda. Other scholars, however, such as Devine and Stephens, argue that on the contrary the grave accent at the end of a word was a true accent, but that in certain contexts its pitch was suppressed.
In the music, a word with a grave frequently has no accent at all, and is set to a single level note, as in these examples from the 2nd Delphic hymn, ὃν ἔτικτε Λατὼ μάκαιρα 'whom blessed Leto bore' and τότε λιπὼν Κυνθίαν νᾶσον 'then, leaving the Cynthian island', in which the words Λατὼ 'Leto' and λιπὼν 'having left' have no raised syllables:
However, occasionally the syllable with the grave can be slightly higher than the rest of the word. This usually occurs when the word with a grave forms part of a phrase in which the music is in any case rising to an accented word, as in καὶ σοφὲ μυστοδότα 'and you, wise initiator into the mysteries' in the Mesomedes prayer illustrated above, or in λιγὺ δὲ λωτὸς βρέμων, αἰόλοις μέλεσιν ᾠδὰν κρέκει 'and the pipe, sounding clearly, weaves a song with shimmering melodies' in the 1st Delphic hymn:
In the Delphic hymns, a grave accent is almost never followed by a note lower than itself. However, in the later music, there are several examples where a grave is followed by a fall in pitch, as in the phrase below, 'the harsh fate of mortals turns', where the word χαροπὰ 'harsh, grey-eyed' has a fully developed accent:
When an oxytone word such as ἀγαθός 'good' comes before a comma or full stop, the accent is written as an acute. Several examples in the music illustrate this rise in pitch before a comma, for example Καλλιόπεια σοφά 'wise Calliope' illustrated above, or in the first line of the Hymn to Nemesis :
There are almost no examples in the music of an oxytone word at the end of a sentence except the following, where the same phrase is repeated at the end of a stanza. Here the pitch drops and the accent appears to be retracted to the penultimate syllable:
This, however, contradicts the description of the ancient grammarians, according to whom a grave became an acute at the end of a sentence just as it does before a comma.

General intonation

Devine and Stephens also note that it is also possible from the Delphic hymns to get some indication of the intonation of Ancient Greek. For example, in most languages there is a tendency for the pitch to gradually become lower as the clause proceeds. This tendency, known as downtrend or downdrift, seems to have been characteristic of Greek too. For example, in the second line of the 1st Delphic Hymn, there is a gradual descent from a high pitch to a low one, followed by a jump up by an octave for the start of the next sentence. The words mean: 'Come, so that you may hymn with songs your brother Phoebus, the Golden-Haired':
However, not all sentences follow this rule, but some have an upwards trend, as in the clause below from the first Delphic hymn, which when restored reads τρίποδα μαντεῖον ὡς εἷλφρούρει δράκων 'how you seized the prophetic tripod which the great snake was guarding'. Here the whole sentence rises up to the emphatic word δράκων 'serpent':
At the end of a clause before a comma, the voice tends to remain raised, and this appears to be true of Greek also. Immediately before a comma, a circumflex accent does not fall but is regularly set to a level note, as in the first line of the Seikilos epitaph, which reads 'As long as you live, shine! Do not grieve at all':
A higher pitch is also used for proper names and for emphatic words, especially in situations where a non-basic word-order indicates emphasis or focus. An example occurs in the second half of the Seikilos epitaph, where the last two lines read 'It is for a short time only that life exists; as for the end, Time demands it'. In the second sentence, where the order is object – subject – verb, the word χρόνος 'time' has the highest pitch, as if emphasised:
Another circumstance in which no downtrend is evident is when a non-lexical word is involved, such as ἵνα 'so that' or τόνδε 'this'. In the music the accent in the word following non-lexical words is usually on the same pitch as the non-lexical accent, not lower than it. Thus there is no downtrend in phrases such as τόνδε πάγον 'this crag' or ἵνα Φοῖβον 'so that Phoebus', where in each case the second word is more important than the first:
Phrases containing a genitive, such as Λατοῦς γόνε 'Leto's son' quoted above, or μῆρα ταύρων 'thighs of bulls' in the illustration below from the first Delphic hymn, also have no downdrift, but in both of these the second word is slightly higher than the first:

Strophe and antistrophe

One problem which has been discussed concerning the relationship between music and word accent is what may have happened in choral music which was written in pairs of corresponding stanzas known as strophe and antistrophe. Rhythmically these always correspond exactly but the word accents in the antistrophe generally do not match those in the strophe. Since none of the surviving music includes both a strophe and antistrophe, it is not clear whether the same music was written for both stanzas, ignoring the word accents in one or the other, or whether the music was similar but varied slightly to account for the accents. The following lines from Mesomedes' Hymn to the Sun, which are very similar but with slight variations in the first five notes, show how this might have been possible:

Change to modern Greek

In modern Greek the accent is for the most part in the same syllable of the words as it was in ancient Greek, but is one of stress rather than pitch, so that an accented syllable, such as the first syllable in the word ἄνθρωπος, can be pronounced sometimes on a high pitch, and sometimes on a low pitch. It is believed that this change took place around 2nd–4th century AD, at around the same time that the distinction between long and short vowels was also lost. One of the first writers to compose poetry based on a stress accent was the 4th-century Gregory of Nazianzus, who wrote two hymns in which syllable quantities play no part in the metre, but almost every line is accented on the penultimate syllable.
In modern Greek there is no difference in pronunciation between the former acute, grave, and circumflex accents, and in the modern 'monotonic' spelling introduced in Greek schools in 1982 only one accent is used, the acute, while monosyllables are left unaccented.

Rules for the placement of the accent

Law of Limitation

The accent may not come more than three syllables from the end of a word.
If an accent comes on the antepenultimate syllable, it is always an acute, for example:
Exception: ὧντινων 'of what sort of', in which the second part is an enclitic word.
With a few exceptions, the accent can come on the antepenult only if the last syllable of the word is 'light'. The last syllable counts as light if it ends in a short vowel, or if it ends in a short vowel followed by no more than one consonant, or if the word ends in -οι or -αι, as in the above examples. But for words like the following, which have a heavy final syllable, the accent moves forward to the penultimate:
The ending -ει always counts as long, and in the optative mood, the endings -οι or -αι also count as long and cause the accent to move forward in the same way:
The accent also cannot come on the antepenultimate syllable when the word ends in -ξ or -ψ, hence the difference in pairs of words such as the following:
Exceptions, when the accent may remain on the antepenult even when the last vowel is long, are certain words ending in -ων or -ως, for example:
If the accent comes on the penultimate syllable, it must be a circumflex if the last two vowels of the word are long–short. This applies even to words ending in -ξ or -ψ :
This rule is known as the σωτῆρα Law, since in the accusative case the word σωτήρ 'saviour' becomes σωτῆρα.
In most cases, a final -οι or -αι counts as a short vowel:
Otherwise the accent is an acute:
Exception 1: Certain compounds made from an ordinary word and an enclitic suffix have an acute even though they have long vowel–short vowel:
Exception 2: In locative expressions and verbs in the optative mood a final -οι or -αι counts as a long vowel:
The third principle of Greek accentuation is that, after taking into account the Law of Limitation and the σωτῆρα Law, the accent in nouns, adjectives, and pronouns remains as far as possible on the same syllable in all the cases, numbers, and genders. For example:
But an extra syllable or a long ending causes accent shift:
There are a number of exceptions to the Law of Persistence.
Exception 1: The following words have the accent on a different syllable in the plural:
The accusative singular and plural has the same accent as the nominative plural given above.
The name Δημήτηρ 'Demeter' changes its accent to accusative Δήμητρα, genitive Δήμητρος, dative Δήμητρι.
Exception 2: Certain vocatives have recessive accent:
Exception 3: All 1st declension nouns, and all 3rd declension neuter nouns ending in -ος, have a genitive plural ending in -ῶν. This also applies to 1st declension adjectives, but only if the feminine genitive plural is different from the masculine:
Exception 4: Some 3rd declension nouns, including all monosyllables, place the accent on the ending in the genitive and dative singular, dual, and plural. Further details are given below.
Exception 5: Some adjectives, but not all, move the accent to the antepenultimate when neuter:
Exception 6: The following adjective has an accent on the second syllable in the forms containing -αλ- :
Oxytone words, that is, words with an acute on the final syllable, have their own rules.

Change to a grave

Normally in a sentence, whenever an oxytone word is followed by a non-enclitic word, the acute is changed to a grave; but before a pause, it remains an acute:
The acute also remains before an enclitic word such as ἐστί 'is':
In the words τίς; 'who?' and τί; 'what? why?', however, the accent always remains acute, even if another word follows:
When a noun or adjective is used in different cases, a final acute often changes to a circumflex. In the 1st and 2nd declension, oxytone words change the accent to a circumflex in the genitive and dative. This also applies to the dual and plural, and to the definite article:
However, oxytone words in the 'Attic' declension keep their acute in the genitive and dative:
3rd declension nouns like βασιλεύς 'king' change the acute to a circumflex in the vocative and dative singular and nominative plural:
Adjectives of the type ἀληθής 'true' change the acute to a circumflex in all the cases which have a long vowel ending:
Adjectives of the type ἡδύς 'pleasant' change the acute to a circumflex in the dative singular and nominative and accusative plural:
The following words have no accent, only a breathing:
However, some of these words can have an accent when they are used in emphatic position. ὁ, ἡ, οἱ, αἱ are written ὃ, ἣ, οἳ, αἳ when the meaning is 'who, which'; and οὐ is written οὔ if it ends a sentence.

The definite article

The definite article in the nominative singular and plural masculine and feminine just has a rough breathing, and no accent:
Otherwise the nominative and accusative have an acute accent, which in the context of a sentence, is written as a grave:
The genitive and dative, however, are accented with a circumflex:
1st and 2nd declension oxytones, such as θεός, are accented the same way as the article, with a circumflex in the genitive and dative.

Nouns

1st declension

Types
Those ending in short -α are all recessive:
Of those which end in long -α or -η, some have penultimate accent:
Others are oxytone:
A very few have a contracted ending with a circumflex on the last syllable:
Masculine 1st declension nouns usually have penultimate accent:
A few, especially agent nouns, are oxytone:
There are also some with a contracted final syllable:
In proparoxytone words like θάλασσα, with a short final vowel, the accent moves to the penultimate in the accusative plural, and in the genitive and dative singular, dual, and plural, when the final vowel becomes long:
In words with penultimate accent, the accent is persistent, that is, as far as possible it stays on the same syllable when the noun changes case. But if the last two vowels are long–short, it changes to a circumflex:
In oxytone words, the accent changes to a circumflex in the genitive and dative, just as in the definite article:
All 1st declension nouns have a circumflex on the final syllable in the genitive plural:
The vocative of 1st declension nouns usually has the accent on the same syllable as the nominative. But the word δεσπότης 'master' has a vocative accented on the first syllable:
Types
The majority of 2nd declension nouns have recessive accent, but there are a few oxytones, and a very few with an accent in between or contracted:
Words of the 'Attic' declension ending in -ως can also be either recessive or oxytone:
Neuter words are mostly recessive, but not all:
Words ending in -ιον often have antepenultimate accent, especially diminutive words:
But some -ιον words are recessive, especially those with a short antepenultimate:
As with the first declension, the accent on 2nd declension oxytone nouns such as θεός 'god' changes to a circumflex in the genitive and dative :
But those in the Attic declension retain their acute:
Unlike in the first declension, barytone words do not have a circumflex in the genitive plural:
Types
3rd declension masculine and feminine nouns can be recessive or oxytone:
Certain names resulting from a contraction are perispomenon:
Masculine and feminine monosyllables similarly can be recessive or oxytone :
3rd declension neuter nouns are all recessive, and monosyllables have a circumflex :
The accent in the nominative plural and in the accusative singular and plural is usually on the same syllable as the nominative singular, unless this would break the three-syllable rule. Thus:
But:
The following are exceptions and have the accent on a different syllable in the plural or the accusative singular:
Words ending in -ευς are all oxytone, but only in the nominative singular. In all other cases the accent is on the ε or η :
In the genitive and dative singular, dual and plural, monosyllables usually move the accent to the final syllable. The genitive plural has a circumflex:
dual: nom./acc. πόδε – gen./dat. ποδοῖν
plural: nom. πόδες, acc. πόδας – gen. ποδῶν, dat. ποσί
plural: νύκτες, νύκτας – νυκτῶν, νυξί
The following are irregular in formation, but the accent moves in the same way:
plural: νῆες, ναῦς – νεῶν, νηυσί
γυνή 'woman' and κύων 'dog' follow the same pattern:
pl. γυναῖκες, γυναῖκας – γυναικῶν, γυναιξί
pl. κύνες, κύνας – κυνῶν, κυσί
The words πατήρ 'father', μήτηρ 'mother', θυγάτηρ 'daughter', γαστήρ 'stomach', ἀνήρ 'man' are similar, but vary in some details:
pl. πατέρες, πατέρας – πατέρων, πατράσι
pl. ἄνδρες, ἄνδρας – ἀνδρῶν, ἀνδράσι
There are some irregularities. The nouns παῖς 'boy' and Τρῶες 'Trojans' follow this pattern except in the genitive dual and plural: παίδων, Τρώων, but dative plural παισί, Τρωσί. The adjective πᾶς 'all' has a mobile accent but only in the singular. Monosyllabic participles, such as ὤν 'being' have a fixed accent. The interrogative pronoun τίς; τί; 'who? what?' also has fixed accent.
Neuter words ending in -ος have a circumflex in the genitive plural:
Concerning the genitive plural of the word τριήρης 'trireme', there was uncertainty. 'Some people pronounce it barytone, others perispomenon,' wrote one grammarian.
Nouns such as πόλις 'city' and ἄστυ 'town' with genitive singular -εως 'city' keep their accent on the first syllable in the genitive singular and plural, despite the long vowel ending:
The vocative singular of some 3rd declension nouns has a different accent from the nominative. Sometimes a final acute changes to a circumflex:
Usually the accent becomes recessive:

Types

Adjectives frequently have oxytone accentuation, but there are also barytone ones, and some with a contracted final syllable. Oxytone examples are:
Recessive:
Paroxytone:
Properispomenon:
Perispomenon:
Comparative and superlative adjectives all have recessive accent:
Adjectives ending in -ής have a circumflex in most of the endings, since these are contracted:
μῶρος 'foolish' is oxytone in the New Testament:
Personal names derived from adjectives are usually recessive, even if the adjective is not:
Unlike in modern Greek, which has fixed accent in adjectives, an antepenultimate accent moves forward when the last vowel is long:
The genitive plural of feminine adjectives is accented -ῶν, but only in those adjectives where the masculine and feminine forms of the genitive plural are different:
But:
In a barytone adjective, in the neuter, when the last vowel becomes short, the accent usually recedes:
However, when the final -ν was formerly *-ντ, the accent does not recede :
The adjective μέγας 'great' shifts its accent to the penultimate in forms of the word that contain lambda :
The masculine πᾶς 'all' and neuter πᾶν have their accent on the ending in genitive and dative, but only in the singular:
The participle ὤν 'being', genitive ὄντος, has fixed accent.

Elided vowels

When the last vowel of an oxytone adjective is elided, an acute appears on the penultimate syllable instead:
This rule also applies to verbs and nouns:
But it does not apply to minor words such as prepositions or ἀλλά 'but':
'the fox knows many things, but the hedgehog one big thing'
The retracted accent was always an acute. The story was told of an actor who, in a performance of Euripides' play Orestes, instead of pronouncing γαλήν᾽ ὁρῶ 'I see a calm sea', accidentally said γαλῆν ὁρῶ 'I see a weasel', provoking laughter in the audience and mockery the following year in Aristophanes' Frogs.

Compound nouns and adjectives

Ordinary compounds, that is, those which are not of the type 'object+verb', usually have recessive accent:
But there are some which are oxytone:
Compounds of the type 'object–verb', if the penultimate syllable is long or heavy, are usually oxytone:
But 1st declension nouns tend to be recessive even when the penultimate is long:
Compounds of the type 'object+verb' when the penultimate syllable is short are usually paroxytone:
But the following, formed from ἔχω 'I hold', are recessive:
Adverbs formed from barytone adjectives are accented on the penultimate, as are those formed from adjectives ending in -ύς ; but those formed from other oxytone adjectives are perispomenon:
Adverbs ending in -κις have penultimate accent:
The first three numbers have mobile accent in the genitive and dative:
Despite the circumflex in εἷς, the negative οὐδείς 'no one ' has an acute. It also has mobile accent in the genitive and dative:
The remaining numbers to twelve are:
Also commonly found are:
Ordinals all have recessive accent, except those ending in -στός :
The personal pronouns are the following:
The genitive and dative of all these personal pronouns has a circumflex, except for the datives ἐμοί, σοί, and σφίσι :
The oblique cases of ἐγώ, σύ 'you ', ἕ, and σφεῖς can also be used enclitically when they are unemphatic, in which case they are written without accents. When enclitic, ἐμέ, ἐμοῦ, and ἐμοί are shortened to με, μου, and μοι :
The accented form is usually used after a preposition:
The pronouns αὐτός 'he himself', ἑαυτόν 'himself ', and ὅς 'who, which' change the accent to a circumflex in the genitive and dative:
Pronouns compounded with -δε 'this' and -τις are accented as if the second part was an enclitic word. Thus the accent of οἵδε does not change to a circumflex even though the vowels are long–short:
The demonstratives οὗτος 'this' and ἐκεῖνος 'that' are both accented on the penultimate syllable. But οὑτοσί 'this man here' is oxytone.
When τίς means 'who?' is it always accented, even when not before a pause. When it means 'someone' or 'a certain', it is enclitic :
The accent on τίς is fixed and does not move to the ending in the genitive or dative.

Prepositions

ἐν 'in', εἰς 'to, into', and ἐκ 'from, out of' have no accent, only a breathing.
Most other prepositions have an acute on the final when quoted in isolation :
The following prepositions were always accented on the first syllable in every context:
Interrogative words are almost all accented recessively. In accordance with the principle that in a monosyllable the equivalent of a recessive accent is a circumflex, a circumflex is used on a long-vowel monosyllable:
Two exceptions, with paroxytone accent, are the following:
The words τίς; and τί; always keep their acute accent even when followed by another word. Unlike other monosyllables, they do not move the accent to the ending in the genitive or dative:
Some of these words, when accentless or accented on the final, have an indefinite meaning:
When used in indirect questions, interrogative words are usually prefixed by ὁ- or ὅς-. The accentuation differs. The following are accented on the second syllable:
But the following are accented on the first:

Types of enclitic

are words which have no accent themselves, but place an accent on the word they follow. Examples in Greek are the following:
The connective τε 'also', 'and':
The emphatic particles:
The pronouns ἐγώ 'I' and ἐμοί 'to me' can combine with γε to make a single word accented on the first syllable:
Indefinite adverbs:
Indefinite pronouns:
But τινές can also sometimes begin a sentence, in which case it is non-enclitic and has an accent on the final.
The present tense of εἰμί 'I am' and φημί 'I say':
These verbs can also have non-enclitic forms which are used, for example, to begin a sentence or after an elision. The verb ἐστὶ 'is' has an emphatic form ἔστι. Judging from parallel forms in Sanskrit it is possible that originally when non-enclitic the other persons also were accented on the first syllable: *εἶμι, *φῆμι etc.; but the usual convention, among most modern editors as well as the ancient Greek grammarians, is to write εἰμὶ and φημὶ even at the beginning of a sentence.
When negative, ἔστι is customarily written with its strong form, but φησί is enclitic:
The strong form ἔστι is also written after εἰ 'if', ὡς 'since', ἀλλ᾽ 'but', τοῦτ᾽ 'this', according to Herodian.
Certain personal pronouns in oblique cases when non-emphatic:
In classical writers, ἑ 'him' and σφας 'them' tend to be used in indirect speech referring to the speaker:
'he ordered the slave-boy to run and ask the man to wait for him'
Some of these pronouns also have non-enclitic forms which are accented. The non-enclitic form of με, μου, μοι 'me', 'of me', 'to me' is ἐμέ, ἐμοῦ, ἐμοί. The accented forms are used at the beginning of a sentence and after prepositions:
When an enclitic follows a proparoxytone or a properispomenon word, the main word has two accents:
When it follows an oxytone word or an accentless word, there is an acute on the final syllable:
When it follows perispomenon or paroxytone word, there is no additional accent, and a monosyllabic enclitic remains accentless:
A two-syllable enclitic has no accent after a perispomenon:
But a two-syllabled enclitic has one after a paroxytone word. After a paroxytone τινῶν has a circumflex:
A word ending in ξ or ψ behaves as if it was paroxytone and does not take an additional accent:
A two-syllable enclitic is also accented after an elision:
When two or three enclitics come in a row, according to Apollonius and Herodian, each passes its accent to the preceding word :
It appears that with certain long-voweled enclitics, such as που, πως, πῃ, πω, Herodian recommended that they should be left unaccented when another enclitic followed. However, most modern editors ignore this second rule, and print εἴ πού τις 'if anyone anywhere' rather than εἴ που τις.

Verbs

In verbs, the accent is grammatical rather than lexical; that is to say, it distinguishes different parts of the verb rather than one verb from another. In the indicative mood it is usually recessive, but in other parts of the verb it is often non-recessive.
Except for the nominative singular of certain participles, a few imperatives such as (εἰπέ 'say', and the irregular present tenses φημί 'I say' and εἰμί 'I am', no parts of the verb are oxytone.

Indicative

In the indicative of most verbs, other than contracting verbs, the accent is recessive, meaning it moves as far back towards the beginning of the word as allowed by the length of the last vowel. Thus, verbs of three or more syllables often have an acute accent on the penult or antepenult, depending on whether the last vowel is long or short :
Monosyllabic verbs, such as βῆ 'he went' and εἶ 'you are', because they are recessive, have a circumflex. An exception is φῄς or φής 'you say'.
A few 3rd person plurals have a contracted ending :
When a verb is preceded by an augment, the accent goes no further back than the augment itself:
Contracting verbs are underlyingly recessive, that is, the accent is in the same place it had been before the vowels contracted. When an acute and a non-accented vowel merge, the result is a circumflex. In practice therefore, several parts of contracting verbs are non-recessive:
Contracting futures such as ἀγγελῶ 'I will announce' and ἐρῶ 'I will say' are accented like ποιῶ.

Imperative

The accent is recessive in the imperative of most verbs:
In compounded monosyllabic verbs, however, the imperative is paroxytone:
The strong aorist imperative active of the following five verbs is oxytone:
However, if plural or prefixed, these imperatives are recessive:
The strong aorist imperative middle of all verbs is perispomenon:
But the following is usually printed with an acute:
As with the active imperative, the plurals always have a recessive accent:
The subjunctive of regular thematic verbs in the present tense or the weak or strong aorist tense is recessive, except for the aorist passive:
It is also recessive in the verb εἶμι 'I go' and verbs ending in -υμι :
But in the aorist passive, in the compounded aorist active of βαίνω 'I go', and in all tenses of other athematic verbs, it is non-recessive:
The optative similarly is recessive in regular verbs in the same tenses. The optative endings -οι and -αι count as long vowels for the purpose of accentuation:
But in the aorist passive, in the compounded aorist active of βαίνω 'I go', and in all tenses of athematic verbs, it is non-recessive:
But ἀπίοι 'he might go away' is accented recessively like a regular verb.

Infinitive

The present and future infinitive of regular thematic verbs is recessive:
But all other infinitives are non-recessive, for example the weak aorist active:
Strong aorist active and middle:
Weak and strong aorist passive:
The aorist active of βαίνω 'I go' when compounded:
The present and aorist infinitives of all athematic verbs:
But the Homeric ἔμμεναι 'to be' and δόμεναι 'to give' are recessive.
The perfect active, middle, and passive:
The present, future and weak aorist participles of regular thematic verbs are recessive:
But all other participles are non-recessive. These include the strong aorist active:
The weak and strong aorist passive:
The compounded aorist active of βαίνω 'I go':
The present and aorist participles of athematic verbs:
The perfect active, middle, and passive:
Two athematic verbs, εἰμί 'I am' and φημί 'I say', are exceptional in that in the present indicative they are usually enclitic. When this happens they put an accent on the word before them and lose their own accent:
But both verbs can also begin a sentence, or follow a comma, or an elision, in which case they are not enclitic. In this case the accent is usually on the final syllable. When it follows an elision, ἐστίν is also accented on the final:
However, the 3rd person singular ἐστί also has a strong form, ἔστι, which is used 'when the word expresses existence or possibility.' This form is used among other places in the phrase οὐκ ἔστι 'it is not' and at the beginning of sentences, such as:
The 2nd person singular εἶ 'you are' and φῄς 'you say' are not enclitic.
The future of the verb 'to be' has its accent on the verb itself even when prefixed:
The verbal adjectives ending in -τέος and -τέον are always paroxytone:
The adjective ending in -τος is usually oxytone, especially when it refers to something which is capable of happening:
Comparison with Sanskrit as well as the statements of grammarians shows that the accent in some Greek words has shifted from its position in Proto-Indo-European.

Wheeler's Law

's Law, suggested in 1885, refers to a process whereby words with a dactylic ending , if they were oxytone in Proto-Indo-European, became paroxytone in Greek. It is also known as the "law of dactylic retraction".
This law is used to explain the paroxytone accent in words such as the following:
Similar words and endings in Sanskrit are regularly accented on the final syllable, and active compounds which do not have a dactylic rhythm often have final accent, e.g. ψυχοπομπός 'soul-escorting'.
There are numerous exceptions to Wheeler's Law, especially words ending in -ικός or -ικόν , which are always oxytone. There are also participles such as δεδομένος or feminine δεδομένη 'given', which have penultimate accent despite not being dactylic. These exceptions are usually explained as being due to analogical processes.

Bartoli's Law

Bartoli's Law, proposed in 1930, aims to explain how some oxytone words ending in the rhythm have become proparoxytone. Another name is the "law of iambic retraction". Examples are:
The existence of such a law has been called into question, however, and it is argued that most or all of the words proposed as examples have other explanations.

Vendryes's Law

Vendryes's Law, proposed in 1945, describes how words of the rhythm, which had penultimate accent in other dialects, came to be pronounced proparoxytone in Attic. This change appears to have taken place about 400 BC, and was known to the Greek grammarians who wrote on accentuation. One ancient commentator on Aristophanes wrote: τροπαῖον should be read as properispomenon in Aristophanes and Thucydides, but as proparoxytone τρόπαιον in later poets.᾽
The law affected words like the following:
The accent shift described by Vendryes's Law seems to have affected mainly adjectives. Verbs such as ἀπῆλθον 'I went away' and participles such as λαβοῦσα 'having taken' were unaffected.

Dialect variations

The ancient grammarians were aware that there were sometimes differences between their own accentuation and that of other dialects, for example that of the Homeric poems, which they could presumably learn from the traditional sung recitation.

Attic

Some peculiarities of Attic, the dialect of Athens, have been noted above under Vendryes's Law.

Aeolic

The Aeolic pronunciation, exemplified in the dialect of the 7th-century BC poets Sappho and Alcaeus from the island of Lesbos, differed in that every major word was pronounced recessively, thus:
But Ἀλκαῖος 'Alcaeus' was apparently pronounced Ἀλκάος in Lesbian.
The Boeotian dialect, although from the same dialect group as Lesbian, did not have this recessive accentuation, and appears not to have differed accentually from common Greek.
The grammarians give no details of the Thessalian dialect but it has been suggested that the dropping of certain vowels in words on inscriptions indicates that it had a stress accent at the beginning of each word.

Doric

The Doric dialect also had certain peculiarities. One was that properispomenon words were pronounced paroxytone. The examples given are 3rd declension nominative plural:
On the other hand, it is reported that the 1st and 2nd declension accusative plural in Doric had a short vowel, leading to accentuations such as:
Another characteristic of Doric was that the endings -οι and perhaps -αι, and in verbs 3rd pl. -ον and -αν counted as long, leading to a paroxytone accent in:
Doric speakers also apparently pronounced a circumflex on certain genitive plurals, which were paroxytone in other dialects:
In Doric the future was also accented non-recessively in all verbs: