Greek compounds are usually made by combining the roots of two nouns and inserting a linking thematic vowel when it would result in a sequence of two consonants. The vowel is usually -o-, so we have e.g.

claustr-o-phob-ia, electr-o-magnet-ism, ec-o-nom-y, hor-o-scope

When the second root starts with a vowel, the linking vowel -o- is not needed, so we have



mis-anthrop-y, mis-andr-y.

But why is "agoraphobia" with an -a-? It is made from agora and phobia, so one would expect

*agor-o-phob-ia. (but this is not the case)

And, are there more such words that use -a- not -o-?

  • 1
    This is a really good question. Why the votes to close?
    – fdb
    Commented Jan 3, 2017 at 0:05
  • 2
    If I may, there is a proposal to create a Greek Language stackexchange, you can support it by adding your question as an example and/or upvoting the other questions: area51.stackexchange.com/proposals/101509/greek-language
    – nyg
    Commented Jan 5, 2017 at 8:40
  • I'm voting to close this question as off-topic because it belongs on ELU
    – Mitch
    Commented Jan 19, 2017 at 17:03
  • It seems related to historical linguistics and morphophonology, so I think it can stay on this site. Commented Jan 28, 2017 at 1:54
  • @nyg I agree this should go in the Greek stackexchange proposal. Commented May 17, 2017 at 4:19

2 Answers 2


The connecting vowel in Ancient Greek compounds depends on the declension of the first noun: http://www.perseus.tufts.edu/hopper/text?doc=Perseus%3Atext%3A1999.04.0007%3Apart%3D3%3Achapter%3D24

  • If the first noun is first declension, the connecting vowel was originally -ā- : agor-ā-nomos 'market clerk', nik-ē-phoros 'bringing victory' (-ē- is the Ionic for -ā-)
  • If the first noun is second declension, the connecting vowel was originally -o-: log-o-graphos 'speech-writer'
  • If the first noun is third declension, and the stem already ends in a vowel, leave it alone, e.g. ichty-bolos 'catching fish'; otherwise, add -o-, e.g. sōmat-o-phylax 'bodyguard'

The selection of vowels is not random of course: first declension nouns have inflections based on -ā-, and second declension nouns have inflections based on -o-.

This rule has been massively messed up by analogy though: in Ancient Greek, there are plenty of first declension nouns with -o- (dik-o-graphos 'writer of legal speeches'), and second and third declension nouns with -ā- (elaphē-bolos 'deer-shooting'). The -o- linking vowel eventually prevailed, and is the only one used in Modern Greek.

In the case of agoraphobia, the precedent in Ancient Greek was that agora compounds, like agorānomos, always had -ā- as the connecting vowel. The new coinage preserved that precedent.


I believe that the vowel used to link the two words depends on the first one. For example, claustrophobia uses an -o- because it comes from the Latin "claustrum", same with electromagnetism, from Latin "electrum". In the case of misogyny, the Greek word involved is "misos", so an -o- is used, but since the root for agoraphobia is "agora", the -a- is preserved. Another example is triskaidekaphobia, the fear of the number thirteen, which comes from the Greek "triskaideka", which ends in "a". Basically, if it already ends in "a", there's no reason to change it to an "o" when connecting it to another word.

  • 2
    Suggestion: mention that many Latin words in -us and -um actually have -o- as their stem vowel (such as claustrum), since otherwise claustrumclaustro- is non-obvious?
    – Draconis
    Commented Jan 2, 2017 at 20:45
  • 3
    I thought about this, but this is not true for just any word ending in -ā or -ē in Greek. Words like thalassa (sea) and, definitely, psūchē, hōrã, limnē (lake) end in -a, but they have only thalasso-, psycho-, horo-, and limno-. In fact afaik nearly all such Greek words use -o-. So while I really like this idea, I feel we need to zero in on just exactly what words do have -a as a part of its root, not a really removable suffix like an inflection suffix. Commented Jan 3, 2017 at 10:25
  • 4
    I've so far only found two other words that preserve -a in some of its compounds, timē (honor) and skiā (shadow). timē has two such compounds, tima-ochos (honor-having) and tima-oros (>timōros, "avenging"; 2nd element related to "see"), but it has tim-o-krat-iā (timocracy). For skiā, we have the English words sciamachy and sciaphobia, though the second has an alternative form sciophobia. Commented Jan 3, 2017 at 10:50
  • 3
    Triskaidekaphobia is a great and interesting example, but I feel there is really no other possible form. As it goes, tris kai deka is "three and ten", i.e. "thirteen". The -a in deka is just part of the word, as deka (ten) is not declinable and has the only form deka. So when one makes a "triskaideka phobia", they can't really detach -a and add -o- because that -a is not morphologically divisible, but the (last) vowel in the second and last syllable of the morpheme "ten", or the last phoneme thereof. Commented Jan 3, 2017 at 10:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.