Part nine of a tour through Greek inflectional morphology to help get students thinking more systematically about the word forms they see (and maybe teach a bit of general linguistics along the way).


A Tour of Greek Morphology: Part 9 and more...

A Tour of Greek Morphology: Part 9

Part nine of a tour through Greek inflectional morphology to help get students thinking more systematically about the word forms they see (and maybe teach a bit of general linguistics along the way).

In part 8 we saw, amongst other things, that the present active infinitive has a spurious diphthong ει from ε+ε whereas the the present active second and third person singulars have a ει that is a true ε+ι diphthong.

This somewhat justifies our observation of the ις and ι pattern in the second and third person singulars across all the present actives we’ve seen so far.

If we show the “inert” part of the endings separated from the vowel that interacts with a preceding stem vowel to form the circumflex verbs, we get something like this:

INFε ε νε σθαι
1SGω -ο μαι
2SGε ιςη ι (sometimes ε ι)
3SGε ιε ται
1PLο μενο μεθα
2PLε τεε σθε
3PLου σι(ν)ο νται

You can see the predominance of initial ε and ο with three exceptions:

  • the ω of the ACT 1SG
  • the ου of the ACT 3PL
  • the η of the MID 2SG

We now know to ask the question: is ου in ACT 3PL a spurious diphthong (from ο+ο) or a true diphthong (from o+υ)? If υ works the same way as ι in our contraction rules, it must be a spurious diphthong.

There’s additional evidence for this:

  • In the Western Greek dialects (like Doric) we find -οντι
  • It was not uncommon for Attic-Ionic to have σι for τι in other dialects (we’ll encounter more examples later)
  • Dentals like ν drop out in Attic-Ionic when followed by σ and this generally causes the preceding vowel to lengthen (what is called compensatory lengthening)

So it seems our ουσι(ν) was originally from the -οντι preserved in Doric.

This introduces interesting parallels with the -ονται in the middle.

What about the ῃ in the MID 2SG? We don’t need to go to another dialect to see traces of what’s going on. In the NT we have the PM-4 circumflex verb:


with ᾶσαι for ᾷ. The ᾶσαι can be explained as the stem vowel α interacting with the ending εσαι. The ᾷ can be explained simply through the σ dropping out (and similarly the ῃ in the PM-1 and PM-2 and so on) plus our contraction rules.

Interestingly, later Greek restored the uncontracted ending and we find it again in Modern Greek.

And so we have the reconstructed endings:

INFε ενε σθαι
1SGω -ο μαι
2SGε ιςε σαι > ῃ
3SGε ιε ται
1PLο μενο μεθα
2PLε τεε σθε
3PLο ντι > ουσι(ν)ο νται

There are some tantalising patterns here, especially in the middle: the αι in 5 out of 7 cells; the μ/σ/τ in the 1st/2nd/3rd person.

As usual I want to emphasize the reconstructed forms in this table help explain things historically but should not necessarily be taken as an indication of a process that went on syncronically in the minds of native speakers. I’m not aware of any evidence that native speakers would have, for example, thought of ουσι as being an underlying οντι, or ῃ as being an underlying εσαι.

We haven’t yet explained what’s going on with the ACT 1SG nor why ει would have been an alternative for ῃ in the MID 2SG.

But other than the ACT 1SG, all other endings start with either an ε or ο. We’ll talk more about this later (including why this vowel is called the thematic vowel) but note that which of the two vowels is used is completely predictable by what follows.

If the following segment is nasal (μ or ν), the vowel is ο. If the following segment is ε, ι, σ, or τ, the vowel is ε. Most descriptions consider the ε the default and the nasal context leading to ο being the exception. But we could also look for features that ε, ι, σ, and τ have that μ and ν don’t (other than just being NON-nasal).


A Tour of Greek Morphology: Part 8

Part eight of a tour through Greek inflectional morphology to help get students thinking more systematically about the word forms they see (and maybe teach a bit of general linguistics along the way).

So far, just for the active, we’ve suggested the following contraction rules.

  • έει > εῖ
  • έω > ῶ
  • έε > εῖ
  • έο > οῦ
  • έου > οῦ
  • άω > ῶ
  • άε > ᾶ
  • άει > ᾷ (in the indicative) and ᾶ (in the infinitive)
  • άο > ῶ
  • άου > ῶ
  • όω > ῶ
  • όε > οῦ
  • όει > οῖ (in the indicative) and οῦ (in the infinitive)
  • όο > οῦ
  • όου > οῦ
  • ήω > ῶ
  • ήε > ῆ
  • ήει > ῇ (in the indicative) and ῆ (in the infinitive)
  • ήο > ῶ
  • ήου > ῶ

In this post I want to explain why these aren’t just an arbitrary set of sound changes and that they are really quite systematic. We’ll say a little bit about Greek orthography and build a model using some simple phonological features that explains the core contraction rules quite compactly.

Before I do that, though, I want to emphasize again that I’m not suggesting these “rules” need to be learned by the language learner. They are historical explanations for the spelling of circumflex verb endings in certain dialects and I’m discussing them to give people a flavour for linguistic description. The best way to learn the circumflex verbs is to produce and read them in context. It really doesn’t take long to just intuitively know that ἀγαπᾷς is a second person singular or that ἀγαπᾶν is an infinitive. You don’t need to know the contraction rules or how to model them with phonological features.

But if you’re interested in WHY the forms are ἀγαπᾷς and ἀγαπᾶν (including why one has an iota subscript and the other doesn’t) keep reading!


You’ve probably been told that ε and o are always short vowels. As far the LETTERS themselves go, in our standard Greek orthography, that is true. But a long ε and a long o existed as sounds in Classical Greek and earlier. Different dialects wrote these differently. Some just wrote Ε and Ο regardless of whether they were long or short. This is similar to Α, Ι, or Υ, which could be used for both the short and long variants. The Ionians, however, used the digraphs ΕΙ and ΟΥ for the long-Ε and long-Ο respectively. At the time, this was NOT the same sound as the diphthongs ΕΙ and ΟΥ, despite being written the same. It is likely that the long ε and long ο were pronounced with the tongue a little higher up (hence closer to the way ι and υ were pronounced) to reduce any confusion with η and ω which were pronounced with a lower tongue, closer to α. The digraphs ΕΙ and ΟΥ, when used for the long ε and long ο are sometimes called “spurious diphthongs” because they weren’t actually diphthongs at all, they were long monophthongs.

The Greeks started to standardize on the Ionian spelling and, in 403 BC, Athens officially adopted the Ionian spelling.

This purely orthographic convention explains why εε > ει and οο > ου. That doesn’t mean ALL occurences of ει are long ε or all occurences of ου are long ο. ει and ου CAN be true diphthongs, but when they come from ε+ε or ο+ο respectively, they are just long monophthongs.

Now as already mentioned, both short and long α was just written as α and so αα > α is a similarly straightforward contraction (the result being a long α). If you have a circumflex or an iota subscript, the α must have been long.

Basic Contractions

So the diagonals of this contraction table make sense:


Now ε+ο and o+ε both result in a long ο (written ου). The order doesn’t matter. The ο wins out over the ε and the ε assimilates to ο resulting in the equivalent to ο+ο.

Both α+ο and ο+α result in ω and again order doesn’t matter. At the time of the spelling standardization, ω was effectively in between α and ο so this makes sense.

Note, however, that α+ε and ε+α don’t behave the same way in our table above. α+ε results in α but ε+α results in η. We might expect both to be η given how α+ο and ο+α behaved. It seems that order matters in some cases but not others.

Phonological Features

One way we can model all this is by assigning each of the vowels binary features of low, back, and round and making generalisations about those categories.

In other words:


Note that not all combinations are possible and +round implies +back.

(We haven’t included ι or υ here as they don’t play a part in this analysis.)

Now all the ε, ο, α contractions can be explained in terms of assimilation of +low and +round and partial assimilation of +back, as follows:

  • the output is +low if either input vowel is +low
  • the output is +round if either input vowel is +round
  • the output is +back if the first input vowel is +back
  • the output is +back if it is +round

The rules also explain why any vowel + ω goes to ω. In fact, if you work them through, these simple rules explain all 23 contractions in our list at the top of the post (and more that haven’t come in to play yet) with just one additional rule:

  • if you have more than two vowels, the contraction is left associative

There are likely other solutions with other features and rules but my analysis roughly follows that of Sommerstein in The Sound Pattern of Ancient Greek, that of Bubeník in The Phonological Interpretation of Ancient Greek: A Pandialectal Analysis (which also considers differences in things like the Doric dialect), and apparently that of Lejeune in Phonétique historique du mycénien et du grec ancien (on which Bubeník’s is based). This style of analysis is typical of the early second half of the twentieth century so I’m not claiming it’s in any way state-of-the-art. But it demonstrates that the contraction rules are very systematic.

The Difference in the Infinitive vs Indicative

There is one final thing we haven’t explicitly addressed but which is fully explained by these simple rules on features: why is άει sometimes ᾷ and sometimes ᾶ (and likewise why is όει sometimes οῖ and sometimes οῦ)?

The answer is simply that if the ει is a spurious diphthong (i.e. actually just a long εε) then our simple rules will result in long ᾶ but if it’s a true diphthong, the result is long α + ι which is written ᾷ. Similarly in the case of όει, a spurious diphthong will result in οῦ (from οεε > οοε > οο > ου) but a true diphthong in οῖ (οει > οοι > οι)).

What this tells us is that the ει in the ειν ending in the infinitive is a spurious diphthong but the ει in εις and ει in the second and third person singular actives are true diphthongs.


A Man Walks Into A Bar

I’ve thought for a while that “A man walks into a bar” jokes are a great example of how definiteness works in English. I mentioned this to Jonathan Robie in Cambridge and he seemed to like the example too so I thought I’d share it more broadly.

Consider the standard joke form:

A man walks into a bar. The bartender says X. The man says Y.

Notice this has two indefinite articles and two definite articles. When do we use the indefinite article and when do we use the definite article?

In our sentence above, we’ve neither been introduced to the man nor the bar before. And so we use the indefinite article.

We can’t say “* The man walks into a bar” unless he’s been introduced before. Likewise we can’t say “* the bar” unless the bar’s been introduced before. For example,

Chris is one crazy guy! The man walks into a bar…

is fine if we take the man to be Chris. Similarly,

You know that bar on 52nd Street? A man walks into the bar…

works if the bar in the joke is the one on 52nd Street.

If we were telling a second joke, we could use the to indicate the man (or the bar) was the same but notice we’d have to use something like another and NOT a for introducing a second bar (or man):

Later, the man walks into another bar…


Later, another man walks into the bar…

Notice in our original joke, the third sentence starts “The man”. This makes sense because that man has already been introduced. We wouldn’t say “* The man walks into a bar. The bartender says X. A man says Y.” Even it were a different man, we’d probably use something like “Another man”.

But notice we did use the with the bartender even though he or she has NOT been introduced yet. The reason is our frame for a bar is that it has a bartender. The existence of the bartender has effectively been set up by us having a bar and that’s the bartender we want to reference so it’s not a completely new reference. Saying “* A man walks into a bar. A bartender says X” would be odd. Notice also that even if the bartender is a man, the following “The man says Y” is unambiguous.

Even if there were more than one bartender (certainly possible, although not prototypical for the frame) we’d have to say something like “One of the bartenders says X”.

This can be demonstrated with an example where we EXPECT multiple instances.

A man walks into a classroom. One of the students says X.

In this case, it would be odd to say “* A student says X” and even odder to say “* the student says X”. We want definiteness (because the classroom frame has already established the likelihood of a group of students and that’s the group we want to reference a member of) but because it’s a group, we need to say “one of” to call out an individual.

“One of the” calls out an indefinite member of a definite group.


A Tour of Greek Morphology: Part 7

Part seven of a tour through Greek inflectional morphology to help get students thinking more systematically about the word forms they see (and maybe teach a bit of general linguistics along the way).

κλῶμεν in 1Co 10.16 is clearly ACT 1PL but we can’t tell from just that if it’s a PA-4 or PA-5. In authors like Galen and Hippocrates we find the MID 3SG κλᾶται which we’ve called PM-4, which strongly suggests it’s a PA-4 in the active.

If that’s the case, we’d expect an ACT 2SG of κλᾷς, an ACT 3SG of κλᾷ, and an ACT 3PL of κλῶσι(ν).

But in various authors we can find the respective forms κλάεις, κλάει, and κλάουσι.

This suggests that α plays the same role in PA-4 and PM-4 as ε did in PA-2 and PM-2.

For this to work,

  • άω > ῶ
  • άε > ᾶ
  • άει > ᾷ (in the indicative) and ᾶ (in the infinitive)
  • άο > ῶ
  • άου > ῶ

We’ll discuss the άει issue in the next post.

What about PA-3 and PM-3? We’re basically trying to solve for x given:

  • xω > ω
  • xε > ου
  • xει > οι (in the indicative) and ου (in the infinitive)
  • xο > ου
  • xου > ου

It’s difficult to find examples in the present verb forms of other dialects and texts, but even in the New Testament it’s not difficult to find cases where οε and οο are alternatively spelled ου (e.g. ἀγαθοεργ- in 1 Tim and ἀγαθουργ- in Acts). This makes ο a possible candidate for x and note, in particular, the ACT 3SG forms have so far all been quite transparent in what vowel ends the stem.

So we appear to have:

  • όω > ῶ
  • όε > οῦ
  • όει > οῖ (in the indicative) and οῦ (in the infinitive)
  • όο > οῦ
  • όου > οῦ

And although a proper argument will get us quite far afield (maybe one day), it turns out PA-5 and PM-5 can be explained by:

  • ήω > ῶ
  • ήε > ῆ
  • ήει > ῇ (in the indicative) and ῆ (in the infinitive)
  • ήο > ῶ
  • ήου > ῶ

So, in summary, the circumflex verbs can be explained through a historical interaction (generally referred to as a contraction) between a vowel at the end of the original stem and the vowel at the start of what is added to it.

  • PA-2 and PM-2 come from a stem originally ending in έ
  • PA-3 and PM-3 come from a stem originally ending in ό
  • PA-4 and PM-4 come from a stem originally ending in ά
  • PA-5 and PM-5 come from a stem originally ending in ή

Often circumflex verbs are referred to as contract verbs but, while contraction is indeed the historical explanation for how the circumflex verbs got their forms, I like the name circumflex verbs because it describes an actual synchronic characteristic of the verb forms rather than an explanation of how they happened to get like that. It’s interesting that ancient grammarians like Dionysius Thrax called them perispomenon verbs (the term for words with a circumflex on the last syllable) and called PA-1/PM-1 verbs barytone verbs (the term for words with NO ACCENT on the last syllable).

In the next post, we’ll explore why the contraction rules are not random but, in fact, are quite systematic. We’ll also touch on why the contractions don’t seem to work quite the same way in the infinitive.


A Tour of Greek Morphology: Part 6

Part six of a tour through Greek inflectional morphology to help get students thinking more systematically about the word forms they see (and maybe teach a bit of general linguistics along the way).

Every form we’ve seen of λύω so far starts with λυ, unchanged except for accent. Also, all the forms that start with λυ (or λύ) have been forms of λύω.

Every form we’ve seen so far that’s active first person plural ends with μεν. Also, all the forms that end with μεν have been active first person plural.

Put another way, the λύ in λύομεν has nothing to do with being active first person plural and the μεν in λύομεν has nothing to do with being a form of λύ (at least based on every paradigm we’ve seen so far).

What about the ο in between them? It cannot (at least at the moment) be said to only depend on the fact we have a form of λύω nor can it be said to only depend on the fact we have an active first person plural form. The vowel seems to depend BOTH on the lexical item AND the morphosyntactic properties of voice, person, and number.

Similarly with ποιεῖτε. The initial ποι indicates and only indicates the lexical item. The final τε indicates and only indicates the active second person plural. The fact we have εῖ rather than ο (or ε or οῦ or any other vowel) is because of BOTH the lexical item and the morphosyntactic properties.

What is happening here becomes very clear when we look at some older texts or texts in more conservative dialects. For example, in Herodotus, written in the Ionic dialect, we don’t find ποιεῖτε but instead ποιέετε. In fact, here’s what we find:

ACT INFποιέειν
ACT 1SGποιέω
ACT 2SGποιέεις
ACT 3SGποιέει
ACT 2PLποιέετε

There are a couple of things about this that are remarkable. Firstly, if we split off the common part (now ποιέ rather than ποι) then our distinguishers are all IDENTICAL to those of λύω. Secondly, this restores the accent placement to be properly recessive.

Our ποιῶ and ποιεῖτε are so accented (and not *ποίω or *ποίειτε) because the accent has remained on the same mora (relative to the start) as the older form.

The vowels are thus explained by noting that historically:

  • έει > εῖ
  • έω > ῶ
  • έε > εῖ

Even without finding the necessary forms in Herodotus, we can infer (assuming the ποιέ is consistent and the distinguishers are those of λύω) the forms missing above and hence the following additional historical vowel changes:

  • έο > οῦ
  • έου > οῦ

And making the same assumption about the middle forms add:

  • έῃ > ῇ

All the PA-2 and PM-2 endings can now be explained by:

  • the verb-specific common part (the stem) ending in ε
  • the voice / person / number endings originally being identical to those of λύω
  • the six historical vowel changes listed (referred to as contractions)

In the tour’s next post, we’ll see if we can similarly explain the other forms we’ve seen. Then, in a subsequent post, we’ll come back to these vowel changes and see what’s systematic about them.

I want to close by emphasizing that I am only trying to describe HOW the circumflex verbs came about, not suggest anything about how native speakers processed or generated the contracted forms. As an analogy: it might be interesting to learn why the English words foot and feet are spelled the way they are relative to how they are pronounced but that explanation doesn’t bear much, if any, relation to what’s going on in the minds of native speakers nor is it necessarily of any use to people learning English as a second language. I’ll touch on that again in a few posts time, but you can also read my 2015 post The Dangers of Reconstructing Too Much Morphophonology.

Click here to safely unsubscribe from "J. K. Tauber: at the intersection of computing, linguistics, biblical greek and learning science."
Click here to view mailing archives, here to change your preferences, or here to subscribePrivacy
Email subscriptions powered by FeedBlitz, LLC, 365 Boston Post Rd, Suite 123, Sudbury, MA 01776, USA.