Prior to the seminal work of Devine and Stephens (The Prosody of Greek Speech, Oxford University Press, 1994), there were too many unanswered questions about how the Greek pitch accent work for a Homeric performance to be more than sheerest conjecture. We knew that the acute accent was a high pitch, and that the circumflex was a falling one (starting from about the level of the acute), and there was some indication that an interval of about a fifth was important, but this wasn't enough to deliver a convincing performance (what one tended to get was a sort of unexciting two-tone chant).

And furthermore, what was the grave? A low pitch? A lowered high? Or what? There are a lot of grave accents in Greek, so you cannot finesse the issue of what to do about it.

What Devine and Stephens did was to combine theoretical ideas and observations about language typology from modern linguistics, with observations of how the accents in the text of the Delphic Hymns correlated with the tunes. These correlations had of course been investigated before (See W.S. Allen, Accent and Rhythm, Cambridge University Press, 1973), but not in the light of specific hypotheses about how pitch accent systems work, based on observations of living languages. In this way they established some basic results which I think put a performance on a somewhat sounder footing, although much of course still has to be invented (even if the paleontologist has got the skeleton right, an artist who wants to draw the animal has to make a lot up).

First we look at their specific results, then some general conclusions.

Specific Results

Result 1 (Negative):

In some languages, properties of consonants and vowels have a significant effect on pitch. For example it might be lowered before a voiced consonant, such as a 'b' (as opposed to unvoiced 'p'). This is not a significant effect in Ancient Greek.

Result 2 (Word Melody):

The basic melody of a word is Mid-Hi-Low, with the Hi located at the acute, or beginning of circumflex. The pitch ramps up to the Hi, then falls down to the Low, gradually over the (at most two) syllables after the Hi:

(of course the pitch fall is during the vowels, not, to any significant extent, the consonants). Putting a high pitch before the rhythmic peak (called variously ictus, thesis, princeps or longum of the foot; in this word it's the long vowel) is the basic trick of this technique; if you can do that, the rest might fall into place.

The circumflex also has a falling pitch, and these will also tend to be rhythmic peaks, so it might be useful to listen to what it would sound like with a circumflex on the second-to-last syllable:

You should listen for the ramp up to the high beginning of the circumflex, in the form of a slightly hightened pitch on the precedeeding /i/.

There is some evidence presented in Allen (1973) that the Ancient Greeks really did associate falling pitch with 'stress' (scare quotes because we don't really know how 'stress' = rhythmic emphasis was implemented).

Result 3 (The Grave):

The grave turns out to have been a rise, but less of one than the acute. The following syllable started on a slightly higher pitch than the acute:

Of course the evenness of the rise from the beginning to the acute is just to make the picture easier to create. The details cannot be known.

Result 4 (Catethesis):

This is the second biggie, after the grave. Work on pitch in languages in the seventies and eighties revealed that there tends to be a process of called 'catethesis', whereby, whithin a small group of words pronounced together as a phrase, as the phrase continues, the pitch range falls and contracts. Devine and Stephens showed that this happened in Greek as well as other languages that had been investigated (such as Tokyo Japanese), and the consequence is that the first of two acutes will normally be at a higher pitch than the second, if they are in words that naturally group together. This is immensely important because it gives us some of what we need to make real tunes rather than a boring 2-tone chant. There are relatively few half-lines or substantial phrases where catethesis is not modified by the effects of the grave, but here's one:

The details of the rise up to the second acute are somewhat conjectural; I think it sounds nice to have a bit of a noticeable increase on the preceeding /i/.

Catathesis was probably triggered by the fall after the High pitch rather than the High itself; it is unclear what happens when a vowel after an acute is elided, but the evidence is consistent with there being no catethesis under these circumstances. Acutes on 'grammatical words' (articles, particles, minor adverbs) as opposed to 'lexical words' (nouns, verbs, adjectives, major adverbs) also seems not to have triggered catethesis.

A homeric line normally consists of two or three phrases in which catethesis can apply (the caesura seems pretty clearly to have been a boundary between such phrases). You actually don't have to worry about it too much, it seems to be part of 'Chomsky's Universal Grammar', and if you can get the acutes to be high and the circumflexes falling, and things feeling reasonably natural, catethesis will happen of its own accord.

Result 5 (Anathesis):

This is actually my term, but it refers to the very import Devine and Stephens discovery that the grave accent not only suspends catethesis, but works like a sort of anti-catathesis, causing the pitch range to increase after it. If the idea that the grave increases the pitch range is true, then we would expect a series of graves to produce a crescendo effect, like this:

There's a little bit of evidence to the effect that sequences of expecially lexical graves might have produced this kind of 'terracing' effect (see for example Devine and Stephens p. 336), but it's not really convincing. However I think the effect works esthetically. The basic phenomenon of Anathesis was not predicted from theory, but rather discovered in the actual data.

Result 6 (Secondary Rise):

And another discovery is a phenomeon whereby the rise to an early High in a word can start in the previous word, if they are closely linked. Esthetically, this has the effect of making things hang together, which is what you'd expect, since ramping up to the High is something that does happen within a word, and Secondary Rise says that it can happen in the previous word if they are forming a phonological phrase:

Result 7 (Freedom to emphasize):

Emphasis can cancel the catethesis and other 'downtrend' effects (in addition to the catethesis within closely knit phrases, there are larger scale tendencies for the pitch to fall during an utterance). Proper names in particular tend to have higher pitches than would otherwise be expected.


Once you've decided where the phrasing breaks are, these rules can specify when the voice should rise and when it should fall, but not by how much, although movements of more than a fifth are probably not appropriate in ordinary circumstances (however I find that in a word like héloi in Od 19:511, I want to have a very sharp transition between the High on the acute and a Low on the following syllable). My intuitive perception is that there actually is a correct pitching for each line, and that finding it is a sort of puzzle solving, and performing it accurately quite difficult. If this is true, it would follow that competing rhapsodes would have been judged on the basis of their ability to hit the right notes, among other things. On the other hand perhaps everyone followed their own vision, and the one that sounded best overall prevailed.

Created by: Avery Andrews
Maintained by: Avery Andrews
Last modified: 22 July, 2002