LINGUIST List 12.213

Sat Jan 27 2001

Disc: Origins of Human Language

Editor for this issue: Karen Milligan <>


  1. Herb Stahlke, Re: 12.206, Disc: Origins of Human Language
  2. Zylogy, Re: 12.206, Disc: Origins of Human Language

Message 1: Re: 12.206, Disc: Origins of Human Language

Date: Fri, 26 Jan 2001 13:47:04 -0500
From: Herb Stahlke <>
Subject: Re: 12.206, Disc: Origins of Human Language

Jess Tauber writes:

> The signal inversion hypothesis extrapolates feature density until
even "phonemes" can be considered as entire utterances some time in
the distant ancestry of man. With just 16 binary feature pairs, with
one choice always instantiated for each pair, will generate well over
50000 separate units.

Larry Trask responds: 

>OK. But I would like to know just what those 16 binary features
are.Can we be assured that all 65,536 combinations of these features
are physiologically possible and auditorily distinct? I find this
hard to believe. Just to start with, I find that any single nasal
stop, produced in isolation, is usually next to impossible to
distinguish from any other nasal stop. I also find it very hard to
distinguish one voiceless plosive from another, without the aid of
neighboring vowels.

Herb Stahlke writes:

Beyond the paradigmatic question of how many feature combinations
there can be, there is the syntagmatic question of what's a
possible sequence and the further question, that linguists rarely
ask, of how many cells in a particular matrix will be filled in a
particular language. Yoruba, for example, defines the verb as a
CV syllable with a tone. Given 17 consonants, 10 vowels, and 3
tones, the three-dimensional matrix would contain 510 cells. By
the time you factor out phonologically impermissible forms, like
nasal consonants followed by oral vowels, the matrix is reduced
only slightly, to about 470. Of these, based on an exhaustive
search of several Yoruba dictionaries, a total of 332 actually
occur. These are distributed throughout the matrix with some
random gaps. Similar studies of a number of languages with
similar constraints show that the Yoruba ratio .7 is actually
quite high. Consider possible English CVC monosyllables, where
the ration is less than .5. Clear it's not a matter of straight
combinatorics. Rather, distinctiveness within the mental lexicon
is also a factor.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 12.206, Disc: Origins of Human Language

Date: Fri, 26 Jan 2001 14:27:52 EST
From: Zylogy <>
Subject: Re: 12.206, Disc: Origins of Human Language

Responding to Larry Trask's posting:

I don't have any particular prejudice regarding the actual number of
coarticulations that might actually be possible in nonhuman
communication- it could very well be lower than 16 pairs of features,
and it could also be that not all feature pairs are always
instantiated in any utterance. Just a demonstration of a "what-if" to
show that a holistic "phoneme/morpheme/phrase/clause/etc." might be
able to contain quite a bit of information. But apes do appear to be
able to do things we can't (such as utilize supralaryngeal air sacs,
perhaps have several different bilabial articulations, engage the
glottis with the nasal system to different degrees, etc.), just as
they cannot (at all or easily) do many of the things we can.

But let's suppose a multidimensional feature-string matrix were
complex, and ideophone-like in that iconic organization is at a
premium, so that neighboring cells in the matrix would have a minimum
semantic difference between them in parallel to the phonological
distance. Precision would be a real necessity, IF the difference
between members within any dimension were small, or IF there was
(near) overlap qualitywise between members of different
dimensions. Potential difficulties of this sort are supposed to be
responsible for the optimization of the human oral tract to produce
maximally different phonological contrasts.

But we're still thinking in terms of HUMAN phonemes- with syllable
structure, oral articulatory position, etc. A holistic scheme may be a
bit more alien than that.

Featural density per segment may be less of a problem if one increases
the frequency range used for distinctive signalling. We use formants
0,1,2,3. What if you could use higher ones, and with a broader range
than we do? Primates are supposed to utilize some of these higher
formants for individuation/identity marking, while low pitched grunts
are utilized to mark group solidarity (self effacement?). Fair
enough. Can these higher formants be tweaked to create contrasts
independently of lower ones?

And maybe I'm being unfair to linguists generally as regards holistic
communication- true, human language doesn't have much of anything like
that. Kind of limiting, though. I can just imagine such limitation
getting us into a shooting war if and when we ever meet space aliens
for real.

Ideophones. Harummph. Kinda reminds me of the apocryphal aside by
Huxley on hearing Soapy Sam's remark about his lineage at the Darwin
debate. Ideophones may be peripheral in some languages, but in others
they are pretty darned important. One scholar recently told me that
expressive roots comprise more than 30% of the total in Uralic
languages. Various Mon-Khmer languages are swimming in thousands of
them, as are Japanese and Korean. On the basis of percentage of
attested roots in the language, these items will rate very high,
either as free forms with their own word class, or percolated
historically into the regular lexical root stock. Indo-European
languages have very many cases of such expressive roots hidden within
regular lexemes.

Indeed, the manner component of verbal semantics would be severely
impoverished if not for the enrichment from this source, among
others. It is interesting that it is in just those languages with
reduced stocks of "real" verb roots that we find a proliferation of
ideophone/expressive forms. Altaic verbs, for instance, appear for the
most part to derive either from nouns/adjectives or expressives plus
auxiliaries, of which there is a rather short list. Similar derivation
is found in many Australian languages. And I could go on (and on).

Finally, as to economy of expression- primates don't have much of a
manufactured world, so don't require the largish vocabulary segment
for this as we do. Combinatoriality is at a minimum socially as well,
so all the high words many cultures have are out. Primary colors, for
everything. And without the need for indirection (which we excel in),
what is the point of drawing messages out. Anything of relevance is
here, now, in your face. The natural world tends to be relatively
constant, with natural classes of items/beings there for everyone to
see. So unless individuation is necessary, use class
terms. Generalize. Be blunt. Get to the point and stop wasting my
time. Life is hard enough.

So just how much vocabulary would we need in such circumstances? I've
recently heard that common chimps have at least NINE food
calls. Wouldn't one have been sufficient? Vervet monkeys have their
three predator calls. Then there is the Morton continuum of calls for
challenge, appeasement, etc., possibly quantized. The question is
whether calls from any semantic domain are systematized, and whether
the system is iconically organized. Nobody will look unless asked,
that's clear. A hierarchical organization could help distinguish basic
classes of call semantic domains, and then further distinctions could
be handled by extra features. Don't have a clue how much nonacoustic
signalling accompanies or how much context comes into play.

Jess Tauber
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue