Phoneme inventory arguments and tone

Don Killian donald.killian at HELSINKI.FI
Thu Jun 2 07:20:37 UTC 2011

Dear Peter and the rest,

I would say these should be included as well, of course!  Anything which 
functions contrastively should be considered part of the language's 
phonology.  Atkinson does look at tone levels but did not add tone in 
his segmental count, as far as I'm aware. If someone knows better, 
please correct me.  WALS only mentions complexity or tone levels, not 
number of contrastive tones and how they function in a language.  As 
Larry mentioned, UPSID does include breathy vowels, creaky voice, 
nasalization, etc.. it is only tone, stress, and length which are excluded.

I don't view this as a fault of UPSID.. it was originally created to 
look at relations and patterns of segments in language, and was not 
intended to make broad generalizations based on the number of phonemes 
of a language.  I don't feel that that is an appropriate use of the 

However, if people are intending to write articles as such, we should 
try to have a database which does account for all contrastive aspects in 
a language and include these.  Including tone, length, and all types of 
vocal modifications could increase the phoneme count of some languages 
quite substantially. In Agar Dinka, for instance, Andersen suggests 88 
or 84 contrastive vowels, including length, creaky voice, breathy voice, 
etc, but not tones.  If tones would be included in his count, this would 
increase the language to something rivaling !Xoo's current phoneme count.

Making sure that tone (and all other distinctive aspects) are included 
in all the current referenced databases could also help combat against 
the notion that tone is expendable.

I'd like to also mention one additional thing which Harald Hammarström 
pointed out to me in a personal email. Mark Donohue has a phonological 
inventory database of about 1200 languages, and he does include tone. 
Using segmental inventories alone from the database, Harald got the same 
correlation everyone else is talking about.

However, when including tone (by multiplying the number vowels by the 
contrastive tones), he found no correlation between speaker size and 
inventory.  This is not necessarily the best method; it was done 
informally, and Mark does not include what the TBUs are or what 
combinations of tones are possible. Still, it does give an indication 
that tone is important enough to not be dismissed lightly, and it could 
potentially ruin many of the current studies' conclusions.

Best Regards,


On 06/01/2011 09:37 PM, Peter Bakker wrote:
> *Don Killian <donald.killian at HELSINKI.FI
> <mailto:donald.killian at HELSINKI.FI>> writes:*
> Dear Don Killian,
> I agree that it is not fair that tones are not included in phoneme
> inventories.
> On the other hand, tone is only one of a set of properties that vowels
> can have (and
> sometimes consonants as well). Often in segmental inventories only the
> points of articulation
> are given, whereas tone, length, glottalization and "breathyness" vs.
> plain (and this is
> not an exhaustive list of ways in which vowels with the same place of
> articulation can be modified)
> may increase the number of vocalic contrasts enormously, especially when
> several of these
> can be used with the "same" vowel. If one only adds tone to segmental
> inventories (as Atkinson did),
> one does not grab all possible contrasts that change the meaning of words.
> Peter Bakker
> Dear typologists,
> I apologize for the potentially controversial email, but I was wondering
> one thing about the recent arguments with phoneme inventory sizes, and
> would like some thoughts.
> Many of the arguments lately have been based off of databases such as
> WALS or UPSID, which mention inventory sizes of consonants and vowels.
> However, databases which include tones in phoneme inventories are
> lacking, and I really am wondering how much this is affecting these
> arguments. My current thought is that almost every single study which
> has ignored tones in phoneme inventory questions has flawed enough
> methodology that the conclusions are invalid, irrelevant of whether they
> end up being true or not.
> Why are tones rarely included anywhere, neither in phoneme databases nor
> arguments? I can't imagine almost any modern linguist would argue that
> they are insignificant, but I also find that simply not mentioning tone
> at all, or the fact that the databases are heavily biased in favor of
> non-tonal languages, somewhat frustrating. If ka and ke are significant,
> why not ká and kà? Adding tones to inventory sizes would radically
> change the number of phonemes in quite many languages.
> I sent an email to Søren about his own article in particular, but I'd
> love to hear other comments or responses.
> Best Regards,
> Don
> --
> Don Killian
> Researcher in African Linguistics
> Department of Modern Languages
> PL 24 (Unioninkatu 40)
> FI-00014 University of Helsinki
> +358 (0)44 5016437
> Peter Bakker email: linpb at <mailto:linpb at>
> Department of Linguistics tel. (45) 8942.6553
> Inst. for Anthropology, Archaeology and Linguistics
> Aarhus University tel. institute: (0045)8942.6562
> Nordre Ringgade, buiding 1410 fax institute: (0045)8942.6570
> DK - 8000 Aarhus C room 340
> home page:

More information about the Lingtyp mailing list