Ancient Romans called the letter R littera canina as the R in Classical Latin was trilled to sound like a growling dog! Was the letter R trilled to sound like a growling dog when the ancient Romans spoke Old Latin too and did Latins (the Italic tribe that spoke Old Latin ) trill the letter R too to sound like a growling dog before the Ancient Romans?

Click on Single consonants on https://www.stilus.nl/uitspraak-Latijn/index.htm#31 to hear a sound clip of Thomas Bervoets demonstrating the trilled R in different words!


2 Answers 2


Old Latin /r/ probably had non-trill allophones: at minimum, a tap/flap [ɾ], and likely approximant and fricative allophones as well.

The existence of [ɾ] in Old Latin is supported by rare cases of /d/ changing into /r/, such as meridies from medius + dies (usually explained as involving dissimilation of the two /d/ sounds) and old spellings of certain ad-prefixed words such as arfuise for adfuisse and arvorsum for adversum. The voiced stop /d/ is more acoustically similar to [ɾ] than it is to [r], so when these changes of /d/ to /r/ occurred, it's likely that the tap/flap [ɾ] was a fairly common realization of /r/, at least in these positions (between vowels and at the end of syllables).

Actuually, in modern languages that have a phoneme transcribed and described as the alveolar trill /r/, the /r/ phoneme frequently has a range of allophones that includes non-trilled sounds. This is because it requires a fair amount of articulatory "effort" to produce an acoustic alveolar trill, so any "weakening" (lenition) of the sound is liable to turn it into something other than a trill. Therefore, evidence for the existence of non-trilled pronunciations of /r/ in Old Latin doesn't rule out the presence of a fortis trill [r] as well. However, I don't know of any positive evidence for a trill in Old Latin. If a trill was used, it may have only been a minority variant, or only predominant over other allophones in specific contexts. The contexts that we would most expect to favor a trill pronunciation are at the absolute start of a word, and when /r/ was doubled (geminated) as /rr/.

In Classical Latin as well, it's plausible a trill may have been a preferred realization only in certain contexts. Many of the Romance languages have a tap/flap as one reflex of Latin /r/.

"Rhotic Variation in Tuscan Italian", by C. Celata, A. Vietti, and L. Spreafico, describes the variety of realizations of /r/ in present-day Italian, where approximant, trill, tap, and fricative realizations can all be found.

enter image description here

Obviously, this table has no direct applicability to Old Latin, but my impression is that Old Latin data is consistent with there being a range of realizations similar to the range we see in present-day Italian. Aside from being related languages, the phonemic and phonotactic behavior of /r/ in Old Latin is similar to that of /r/ in present-day Italian: both languages have one rhotic phoneme, /r/, which is always a singleton consonant when in word-initial position or before or after another consonant, but can be singleton (short) or geminate (long) in intervocalic position.

  • 1
    The distribution of sounds in Spanish is also a little screwy. There is a morpheme /rr/ that contrasts with /r/ medially between vowels, but word initially, only /rr/ can appear and preconsonantally, either can appear optionally in the same word without any difference in meaning. Commented Feb 4, 2022 at 20:23
  • 1
    Brazilian Portuguese is even screwier than Spanish. It has the same phonemic differences as Spanish, but both /r/ and /rr/ have a dazzling number of alternative pronunciation based on dialect and position in the word, ranging over [r] , [ɾ], [ɹ̠], [χ], [h] or complete deletion. Commented Feb 4, 2022 at 20:35
  • 1
    @Vegawatcher: Yes, Spanish has a distinction, although the behavior of Spanish -r- and -rr- is different from in Latin since adding a vowel-final prefix to an r-initial word in Spanish creates intervocalic -rr- while in Latin it creates intervocalic -r-
    – Asteroides
    Commented Feb 4, 2022 at 20:49
  • 2
    @Vegawatcher: e.g. Spanish has prerromano, where the consonant after the prefix is spelled with "rr" and pronounced with the same sound as the morpheme-internal "rr" in tierra,, whereas Latin has derogat, where the consonant after the prefix is spelled with single "r" and presumably pronounced differently from the morpheme-internal "rr" in Latin terra.
    – Asteroides
    Commented Feb 4, 2022 at 21:22
  • 3
    @Vegawatcher: Another line of evidence for Latin pronunciation of "r" is scansion. There are lots of examples of a word-final short vowel before word-initial "r" scanning as a short syllable; it's harder to give examples word-internally since no monosyllabic prepositional prefix ends in a short vowel, but there are a couple of *bi-*prefixed words (biremis, birotus) and some compounds (semirasus, semireductus) where the intervocalic r should also show the same metrical behavior. I don't know how Romance lines up
    – Asteroides
    Commented Feb 4, 2022 at 21:35

Probably not, but it's hard to be sure.

As far as I've been able to tell, we don't have any grammarians describing the sounds of Latin letters before the Classical period. The phrase littera canīna appears in Persius but I haven't found it anywhere earlier.

We do, however, have some linguistic evidence. In Old Latin, the /z/ sound shifted to be written R, in a process called "rhotacism". This change makes more sense if Old Latin R was something more like the English R, an alveolar approximant.

Now, there are plenty of other possibilities as well. Maybe /z/ shifted to an alveolar approximant which then merged with R, which was always a trill /r/, for example. Or maybe there was a back approximant involved somewhere as an intermediary between /z/ and /r/. But the simplest hypothesis seems to be that /z/ merged directly with R, which was an approximant, and R then later shifted to a trill.

  • 2
    I dunno, I don't buy the Z to alveolar approximant. I think Z to trill/flap makes much more sense to me.
    – cmw
    Commented Feb 1, 2022 at 3:34
  • 2
    @cmw z > ɹ just involves opening the closure a little, z > r involves completely changing the articulation mechanism.
    – Tristan
    Commented Feb 1, 2022 at 15:25
  • 2
    ɹ is not labialised. The English phoneme often transcribed as ɹ is, but ɹ as that symbol is actually defined has no labial element
    – Tristan
    Commented Feb 1, 2022 at 16:52
  • 3
    There are many examples of languages where the rhotacism z>r occured, but I'm aware of no language for which it has been suggested that the main allophone had to pass through an approximant stage like in English. Sardinian is an example of a language where /s/ is realised as [r] before voiced consonants, and it has no approximant allophone at all. The rhotic phoneme is notorious for admitting of allophony, including fricative and approximant realisations, especially in the coda position or in gemination (Turkish, Egyptian Arabic, Sicilian), and this is clearly the simpler explanation. Commented Feb 2, 2022 at 9:09
  • 2
    I agree with @Unbrutal_Russian -- rhotacism doesn't require an English-style approximant rhotic (if it did, it would be much less common). [z] > [ɾ] (i.e. to flap not trill) seems like an easy articulatory change -- basically you're just shortening the duration of alveolar constriction until you're tapping it.
    – TKR
    Commented Feb 2, 2022 at 18:17

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.