In different forms, the Latin root caput "head" appears with different vowels:

  • a-u: caput (nominative singular);
  • a-i: capitis (genitive singular), capitī (dative singular), capita (nominative plural), and so on;
  • i-i: praecipitem (accusative singular) "headlong", and so on; praecipitō "I throw headlong"
  • e-0: praeceps (nominative singular) "headlong" (coming from -cept-s, as we may assume);
  • i-u: oc-ciput, sin-ciput.

Are these alternations an expression of a general rule of Latin grammar?

Do they reflect some Proto-Indo-European alternation rule?

And from the point of view of universal grammar (linguistic theories with more predictive sterngth/explanatory force, such as optimality theory), can these alternations be explained as a consequence of certain universal (cross-linguistically valid) phonological constraints? (In which other languages are these constraints active too?)

  • Short /u/ - /i/ is a common alternation in Latin. Not sure about the others. Commented Jan 10, 2013 at 1:13
  • @Mechanicalsnail: Hmm where exactly? In theme vowels? Are you suggesting u/i are theme vowels in caput/capit-?
    – Cerberus
    Commented Jan 10, 2013 at 5:27
  • 1
    @Cerberus As for u/i alternation, I read about sonus medius: An intermediate vowel sound (likely [ɨ] or possibly its rounded counterpart [ʉ]) can be reconstructed for the classical period. Such a vowel is found in docvmentvm, optimvs, lacrima (also spelled docimentvm, optvmvs, lacrvma) and other words. It developed out of a historical short /u/ which was later fronted due to vowel reduction. In the vicinity of labial consonants, this sound was not as fronted and may have retained some rounding.(Allen 2004, p.56,59) Commented Jan 10, 2013 at 17:57
  • @Cerberus But in our case of caput/capit-, such an explanation wouldn't explain the clear distribution of the vowels between different forms: a labial consonant is near in all the forms. (Of course, further explanation can be thought of in terms of reduction affecting the different forms differently because of different syllabic (prosodic) structure of the forms. But this is not stated in the quoted excerpt.) Commented Jan 10, 2013 at 18:06
  • @imz--IvanZakharyaschev: Oh, yes, of course, that sound, as in maxumus. I actually didn't know this reduction also occurred in the endings -us, -um, although it makes sense, considering also how -um may be elided. I don't know how it worked exactly, but if Germanic also has i/u Wechsel...
    – Cerberus
    Commented Jan 10, 2013 at 20:25

1 Answer 1


None of these can be explained by synchronic rules of Latin grammar, at least not in the classical age. Nor does this seem to be regular Proto-Indo-European ablaut, where one would expect something like **cp-/cep-/cop-*. Somehow many different allomorphs of caput exist; but allomorphs of other stems are plentiful in Latin and in most other languages. Many combinations of these allomorphs appear to be possible:

  • cap-/cep-/cip- (semantic root)
  • -ut-/-it- (suffix)

As to the nominative singular -ut v. stem -it-, nominatives singular are normally different from the other forms of words or suffixes of the 3rd declension. Homo, hominis; rex, regis; munus, muneris; etc. Each nominative can probably be explained, and there are patterns, but I don't know about the -u- in caput specifically. And -ud/-ut is probably a Proto-Indo-European suffix, probably with its own rules. I have consulted Walde-Hoffmann (1938) and Ernout-Meillet (1959), but I couldn't find a real explanation for -ut- v. -it- in caput, capitis.

Interestingly, the same -u/i- allomorphs seem to exist in a Germanic word that is probably cognate to caput (so says Philippa (2004–2009), although the Oxford English Dictionary (2nd ed.) is sceptical). Dutch hoofd and German Haupt come from Proto-Germanic **haub-id-*, whereas Old English heafod, modern head, and Old Norse haufuþ come from **haub-ud-*. According to the Oxford English Dictionary, this -id/ud- constitutes "suffix ablaut". One would expect the Germanic and Latin suffixes to be cognates too, but, alas, etymological dictionaries rarely explain all the affixes, and I couldn't find it. Note that there is also a Germanic allomorph **hab-* leading to Proto-Germanic **hab-ud-*, Old English hafud, and Old Norse höfuþ.

As to the -cip- in praecipito, an -a- in the first syllable of a verb normally changed into -i- after a prefix, in a certain phase of Latin. The same applies to -ae- and -e-, but not to -o- and -u-, as far as I know. The length of the vowel was generally preserved, so that long -ā- and -ae- become long -ī-, but short -ă- becomes short -ĭ-. Note that vowels did not (always) change in words formed in later phases of Latin. Nor did they change in the perfect stems of verbs. Hence:

Maneo - remaneo - immanens

Paro - reparo - apparans

(Present) capio - recipio - incipiens

(Perfect) cepi - recepi - receptus

(Present) ago - redigo - redigiens

(Perfect) egi - redegi - redactus

This may explain -cip- after many suffixes. However, it is possible that it is simply an older, unexplained allomorph.

As to -e- in praeceps, this appears to be an allomorph that already existed in prehistoric Latin or earlier. (Cep- in the perfect is probably a lengthening of the vowel and explained through other mechanisms, but this is not certain.) The dictionaries above have no real explanation. It is possible that the -e- is due to contamination from words like princeps and auceps, which probably came from the e grade of the root of capio. According to Ernout-Meillet, the root of capio had three allomorphs that seem to display regular Proto-Indo-European ablaut: **kēp-, kōp-, and **kəp-* (from which cap-). Notice that the stem of princeps etc. is -cip-, whereas that of praeceps etc. is -cipit-.

The -e- is only in the nominative singular, and it is not surprising in general that the nominative singular should be irregular. Pontifex (gen. pontificis) and maleficus come to mind, from facio, "make". Somehow both -e- and -i- occur. Compare also pars, partis leading to expers, expertis: the same nominative -e-, but no -i- in the stem.

As to the disappearance of the suffix in anceps, it is possible that it was syncopated, but it is more likely that the suffix was simply not used in all words derived from what may have been Proto-Indo-European **kh₂p*-/**kap-* (the origin of Proto-Indo-European **kap-ut* is not quite certain: **kh₂p*-/**kap-* may have come from a pre-Proto-Indo-European substrate language). This root meant something like "head", and later reflexes include some forms with the suffix -ut/-ud, others with -elo- (Greek kephalê), and probably others again with no suffix at all.

Note also that this root may be related to capio, habeo, and English/Germanic have/etc. (so suggests Walde), which come from a Proto-Indo-European root **kap-* (from **kh₂p-*/**kəp-*), or **gʰabʰ-* (from **gʰh₂bʰ-*) / **gʰebʰ-*, but this is uncertain. (These Proto-Indo-European roots may be variants on and reflexes of a pre-Proto-Indo-European substrate root, so suggests Philippa, Etymologisch Woordenboek van het Nederlands (2004–2009), 'hebben').

Words derived from the capio root exist that mean something like "recepticle, container" (notice cep-), and a semantic shift from pottery to head is a common phenomenon. Cf. Latin testa "(piece of) earthen pot", from torreo "burn" (related toast), which lead to French tête "head" and similar words in other Romance languages. Cf. also Latin cuppa "cup", probably from a non-Proto-Indo-European language (Philippa), which probably led to English cup, Dutch kop "cup; head (informal)", German Kopf "head (informal)". The original Proto-Indo-European root for "head" was **ker-*, "head, horn", from which horn, cerebrum, cervix, unicorn, triceratops; somehow the central word for "head" was replaced often in the successor languages.

  • 2
    Andrew Sihler (Sihler 1995) argues that in Proto-Italic short vowels e, i, o, u "merged nearly completely" into e (in the medial position), which turned later into i before a single consonant: GEN.SG. *kaputes->capitis.
    – Alex B.
    Commented Jan 11, 2013 at 5:17
  • 1
    @AlexB.: Right, that makes sense. So then was the u in caput long, and in the oblique forms short? Because it is now short in caput. Now all we need to know is how PI vowel lengths worked...
    – Cerberus
    Commented Jan 11, 2013 at 5:50
  • @Cerberus The merger brought to our attention by @alex-b is said to occur "in the medial position". The examples from the Q with u are all with u in the last syllable (caput, occiput). I'd understand "medial" as not in the last syllable, and then this matches the data: here, in the last syllables, the old *u didn't undergo this merger. But my understanding of "medial" may be wrong. As for vowel lengths in different forms, of course, that's an interesting direction to look into, because IE languages show some primary or secondary ablaut differences between NomSg and other forms. Commented Jan 8, 2020 at 23:16
  • @Cerberus As for IE languages showing some ablaut differences between NomSg and other forms, a quick link I can give is to the Wikipedia example with declension of "father" in Greek: en.wikipedia.org/wiki/Indo-European_ablaut#Ablaut_grades Their other related, more complex (suffixed) example is with the reconstructed proto-IE word ` *pértus` (assumed to give Eng. "fort", Latin "port"): en.wikipedia.org/wiki/Indo-European_ablaut#Grammatical_function , which I don't like because it doesn't refer immediately to any real data from real languages and has no references. Commented Jan 8, 2020 at 23:23

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.