One of the distinctions among languages is the tonal/atonal distinction. Dediu & Ladd (2007) suggest that this split between tonal and atonal languages is related to a recent mutation in the ASPM gene. In particular, people with the newer variant of this gene correspond closely to groups with atonal languages. The new variant of this gene is believed to have appeared between 14,100 and 500 years ago (most likely about 6k years ago). The mutation fixed itself very quickly in certain sub-populations, suggesting a selective sweep. If the new variant of this gene is strongly related to atonality in language, then one would also expect a divergence of the atonal languages from a tonal ancestor around the same time (14,100 to 500 years ago).

Is there evidence that atonal languages have tonal languages as ancestors? If so, did the seperation appear in the time frame of 14,100 to 500 years ago?

  • 2
    I don't agree that the distinction between tonal/atonal is sharp. Maddieson in WALS claims it is, but then acknowledges that the 'pitch-accent' cases are perhaps only marginally tonal, but includes them (and there are lots of them). 'Tonality' can cover a range of phenomena and it's not clear where/if a line should be drawn. Sep 7, 2012 at 4:50
  • 1
    BTW one example of tonal->atonal change is Ancient Greek, which was tonal (pitch-accent) while Modern Greek is not. Sep 7, 2012 at 4:53
  • 1
    Some people sharply disagree with this distinction, in fact such people are in control at Wikipedia and will prevent any attempt to add "Tonal language" categories. (Or did so last time I looked a couple of years ago.) Sep 7, 2012 at 8:18
  • Is ASPM the gene that mutated to produce Merge?
    – jlawler
    Sep 7, 2012 at 16:04
  • 1
    @GastonÜmlaut I removed my assertion about the 'sharpness' of the distinction, since it seems to have been controversial and distracting from the main question of if atonal languages have tonal ancestors and if so when the split happened. Sep 8, 2012 at 4:29

2 Answers 2


There are a number of examples known of tonal systems being lost in daughter languages, a couple are described below. Given that it is thought that the majority of the world's languages make use of some kind of tonal contrast in their phonology, that tone seems to be a feature likely to be conserved diachronically, and that tonogenesis seems to happen fairly readily, we shouldn't expect there to be many examples of the loss of tone.

According to Downing (2004) proto-Bantu is reconstructed as having had two tones. While most present-day Bantu languages have tonal contrasts a few have lost them, retaining only a stress system, eg Chimwi:ni, Kituba, Nyakyusa and Swahili.

It has been argued by Indo-Europeanist scholars (including Christian Lehmann) that an early stage of proto-Indo-European had a pitch-accent system which was lost in most daughter branches. It was preserved into Ancient Greek (although lost in Modern Greek) and continues in present-day Lithuanian and some Slavic languages. The tone systems found in some North Germanic languages are thought to be relatively recent innovations.

  • if tonogenesis is more frequent than tone loss, as is implied in the first paragraph, one might expect a more skewed distribution in terms of tonal/atonal languages.
    – user483
    Sep 10, 2012 at 1:15
  • 1
    @jlovegren Isn't that what we see in the world's languages? Of course it may be that tonogenesis is simply easier to detect than tone loss. Sep 10, 2012 at 3:15

The examples given in GastonUmlaut's answer are sufficient, so I want to only add a note addressing the main methodological issue here. It concerns the relative strengths of historical factors and biological factors in typology. We can say, simplifying things, that a language is the way it is because:

  1. It is a reproduction of an earlier stage of the language (Historical factors)
  2. It must be learnable and usable by humans (Biological factors)

A large number of biological factors have been identified by comparative studies in phonology, morphology and syntax. To give a few examples: voiced fricatives are dispreferred because they are more difficult to produce than voiceless fricatives; suffixes are preferred over prefixes, probably because this facilitates processing, and a consistent head--modifier ordering between verbs and objects, and between adpositions and their objects, is also preferred for processing reasons.

However, these biological factors are necessarily weak when compared to historical factors. If the factor disfavoring voiced fricatives were strong, then it would be simply too difficult for humans to learn to incorporate this kind of sound into running speech, and no language would have them.

Now if biological factors is discovered which is different across human populations, we would expect for it to be weak compared to historical factors, because so far there is no evidence that any human population is constitutionally unable to learn some other's language.

I think if this point is kept in mind then Dedieu & Ladd's finding should not be especially controversial, since it is not at all surprising that there are minor genetic differences across human populations.

  • A good point, and important that it be made. Sep 8, 2012 at 23:43

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.