Wednesday, 7 April 2010

exotic dental fricatives

Adam Brown asked me if I knew what percentage of the world’s languages have the sound [θ].

I referred him to The World Atlas of Language Structures, which is now available on-line.

Go to the WALS website, search under “features” and look for chapter 19: Presence of Uncommon Consonants. This chapter covers clicks, labial-velars, pharyngeals, and ‘th’ sounds. (It does not distinguish between voiceless θ and voiced ð.) The database covers 2650 different languages.

Dental or alveolar non-sibilant fricatives are just as rare as labial-velar plosives, occurring in just 43 (or 7.6%) of the languages surveyed, but the distribution of these languages is practically worldwide. They are found in languages as varied in location and family affiliation as Modern Greek, Albanian, Spanish and English (Indo-European), Kabardian (Northwest Caucasian), Mari and Nganasan (Uralic), Burmese and Sgaw Karen (Sino-Tibetan), Lakkia and Yay (Tai-Kadai), Swahili and Moro (Niger-Congo), Dahalo (Afro-Asiatic), Berta and Murle (Nilo-Saharan), Fijian, Yapese and Drehu (Austronesian), Ngiyambaa (Pama-Nyungan), Rotokas (West Bougainville), Aleut (Eskimo-Aleut), Chipewyan (Athapaskan), Acoma (Keresan), Maricopa (Yuman), Cubeo (Tucanoan), Huastec (Mayan), Mixtec languages and Mezquital Otomí (Oto-Manguean), Amahuaca (Panoan), Tacana (Tacanan), Cochabamba Quechua and Mapudungun (Araucanian).

We might also mention Welsh and Icelandic (both I-E, and not included in the database).
Here is part of the map that comes up. The presence of a th sound or sounds is marked by a red circle.
This means that yes, for most learners of EFL the English dental fricatives are difficult, exotic sounds. It’s not surprising that they are sounds that advocates of teaching English as a Lingua Franca want to not bother with (but that, conversely, anyone who aspires to a higher standard than ELF cannot afford not to master).


  1. Strange that there's no mention of Semitic languages there, most prominently many Arabic dialects and Standard Arabic.

  2. That's an interesting website. I found it funny that it says that English has no glottalised consonants. It does in Britain!

  3. Athel Cornish-Bowden7 April 2010 at 16:45

    The quotation is right, of course, in saying that Spanish has a th sound, but the unvoiced sound is only used in Spain (and not all of Spain, at that). Whether it has a voiced th sound is something for a better phonetician than I am to answer -- I'm not sure if the first d in "Madrid" is really the same sound as the th in English "the". However, something that I've observed many times and that I find surprising, is that all the Spanish speakers I've ever encountered in Latin America or the Canaries can pronounced an unvoiced th without any difficulty even though they never use it when speaking Spanish (unless imitating a Madrileño). (This is something I have plenty of opportunity to notice, as I have an unvoiced th in my name.) When I've commented on this to Latin Americans they say that they've all been taught to pronounce Spanish "properly" at school and that's why they have no difficulty. I'm not convinced, however. All British people of my age were taught to pronounce a French r at school, but virtually none of them succeed in producing one without a lot of effort.

  4. Does that list only include languages with /T/ or /D/ as phonemes? In Tuscan accents of Italian, [T] is an allophone of /t/ in post-vocalic positions and the same (AFAIK) applies to [D] for /d/ in Spanish, and I don't have any good reason to believe this must be an exceedingly rare phenomenon cross-linguistically, so if the list also included languages with non-phonemic [T] and/or [D] I'd expect the percentage to be somewhat higher than 7.6%.

  5. Expanding on my last post: There even are English speakers which use an alveolar non-sibilant fricative for post-vocalic syllable-coda /t/, e.g. John Horton Conway and "Ireland Eleven" on IDEA, the latter especially in pre-pausal positions.

  6. English speakers learning Welsh are generally intimidated by /ɬ/, considering it a weird and exotic sound. But according to Ian Maddieson's Patterns of Sounds, it's actually more common cross-linguistically (or at least among the languages in that database) than /θ/.

  7. As alluded above, some Semitic languages have theta. Even Proto-Semitic itself is reconstructed with the sound and it was preserved in Ugaritic.

  8. Does anyone think that /θ/ and /ð/ are on the way out in England? People of my age (mid-30s) who grew up south of approximately Gloucester/Oxford tend to have /f/, /v/ in my personal experience.

    They seem pretty secure in the North America, though.

  9. They seem pretty secure in the North America, though.

    Pretty secure in America except for in BVE (Black Vernacular English) where fronting is quite common (I had black friends that had th-fronting quite frequently in their speech) and also in foreign language accents of course (my Polish neighbor had a lot of trouble with /θ/). I'm not sure how Canadian blacks talk though. I think they tend to sound more like Canadian whites.

  10. Glen, if you include earlier stages of languages, you'd get a much, much higher percentage. But even among today's Semitic languages and dialects, /θ/, /ð/ and a pharyngealized /ð/ aren't rare. As a rule they're continuing the Proto-Semitic consonants or - still since some two millenia ago - they are the result of spirantisation, similar to Spanish /Vd/ but phonemic by now.

  11. >>We might also mention Welsh and Icelandic (both I-E, and not included in the database).

    And Cornish.

  12. WALS is wrong about Ngiyambaa having a dental fricative. The phoneme written "dh" in Tamsin Donaldson's practical orthography of this language is an interdental stop, the same as occurs in most Australian languages.

    Elsewhere in Australia, Bandjalang /d/ has a fricative or affricate allophone non-initially (something like the situation in Spanish), but this is not the case with Ngiyambaa. Several languages in Cape York, thousands of kilometers away, do have a phoneme /ð/ though.

    I emailed WALS about this some years ago. They replied that they'd sort it, but it seems they haven't.