LINGUIST List 13.2768

Sat Oct 26 2002

Disc: Last Post/Darwinism & Evolution of Language

Editor for this issue: Karen Milligan <>


  1. H.M. Hubey, Re: 13.2735, Disc: 2nd to Last Posting - Darwinism & Evolution
  2. Nemonemini, Re: 13.2721, Disc: Darwinism & Evolution of Lang

Message 1: Re: 13.2735, Disc: 2nd to Last Posting - Darwinism & Evolution

Date: Wed, 23 Oct 2002 21:41:51 -0400
From: H.M. Hubey <>
Subject: Re: 13.2735, Disc: 2nd to Last Posting - Darwinism & Evolution

I feel that I should make an attempt to respond to three emails of Mr.
Landon. But I will have to be brief because this is an extremely
complex topic. However, for those interested in related topics I will
point out this reference:

Hubey, H.M. "Evolution of Intelligence: Direct modeling....",
Kybernetes, Vol 31, No. 3/4, 2002.

>Date: Mon, 21 Oct 2002 22:09:07 EDT
>Subject: Re: 13.2704, Disc: Darwinism & Evolution of Lang
>>Actually this is simply a matter of definition. It goes like this:
>>random + deterministic = random
>>random*deterministic = random
>>The former is "additive noise" and the latter "multiplicative noise",
>>and both are random. "Random" is not equal to "uniforrmly random",
>>thus one can find broadbrush patterns in random phenomena.
>I recommend a good biography of Isaac Newton,to explain
> both reality and the methodology with respect to reality at
>the same time. You will discover that science as we know is monkey see
>monkey do plus ingenious math tricks. 
I see no reason to pursue this line. You are welcome to join the
"Language" list and pursue your arguments there. To join, send email
to this address:, and in the body
of the message put "subscribe language" (without the quotes).

What is randomness and what is deterministic is very clearly defined
and I wrote it above. No need for confusion.

>Date: Mon, 21 Oct 2002 23:43:49 EDT
>Subject: Re: 13.2721, Disc: Darwinism & Evolution of Lang
>RE Linguist 13.2721
>>I realize I have to expand what I wrote.
>>It is best explained via equations. Let f(t) be some function of time,
>>and r(t) be a random function of time. Then if y(t)= r(t)*f(t) and
>>x(t)=r(t)+f(t), both x(t), and y(t) are random processes. This is due
>>simply to definition of randomness. To apply directly to evolution,
>>let the "evolution" of something very simple be given by the equation
>> dz(t)/dt + a(t)*z(t) = f(t)
>I will study this but I should say that.... 

>The flaw, to me, can be seen in the old edition of Hartl
>on pop gen, where the 'force' is brought in relation to natural
>selection. The problem is that natural selection is not a force, and
>evolution is not physics, and the equations don't really explain
>evolution. What is this 'force', i.e. what's our game in terms of
>foundational concepts? We can't say until we have the facts. 

I explained all this. And soon I will be working on "population
genetics" models with a colleague whose specialty is dynamics (in
fact, population dynamics) and we will be working on language
evolution. To add more, the partial differential equations for the
pdf (prob density function) of random processes are the so-called
diffusion models, and I have already done work in this field and
understand it reasonably well. You can also read about them in both of
these books:

Hubey, H.M. 1999, Mathematical Foundations of Linguistics, Lincom
Europa Hubey, H.M. 1994 Mathematical and Computational Linguistics,
Mir Domu Tvoemu, Moscow (this later published by Lincom Europa in

Longer and more detailed versions of these books (with examples and
exercises suitable for teaching) will be eventually published.

>Consider finally a puzzle. The pieces are scrambled, and seem
>random. Then we see the relation of certain pieces to each other, and
>solve a corner of the puzzle. Then we see the pieces cohere, and are
>not random. Thus my usage is quite ordinary, and takes common sense
>derivations. The term random is hardly a numerical issue at all. 

It can't be anything else. It is exactly because it is so difficult to
explain that so much precision and detail are needed. There are tons
of mathematical materials on probability theory, stochastic processes,
plausible reasoning (See Jaynes' book), etc. It is a very rich field
and also very powerful. Today most of datamining, model-building,
search for patterns, knowledge discovery are all tied together within
the view point of probability theory, namely Bayesian reasoning. See
the book by Hastie, Tibshirani, and Feldman for the last 30 years of
research in this field.

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

Message 2: Re: 13.2721, Disc: Darwinism & Evolution of Lang

Date: Thu, 24 Oct 2002 00:59:13 EDT
From: Nemonemini <>
Subject: Re: 13.2721, Disc: Darwinism & Evolution of Lang

Re Linguist 13.2721

> It is best explained via equations. Let f(t) be some function of time,
> and r(t) be a random function of time. Then if y(t)=3D r(t)*f(t) and
> x(t)=3Dr(t)+f(t), both x(t), and y(t) are random processes. This is due
> simply to definition of randomness. To apply directly to evolution,
> let the "evolution" of something very simple be given by the equation
> dz(t)/dt + a(t)*z(t)= f(t)

On second thought....I find this intriguing. Give me a reference to
something here. There is always an outside chance some kind of
stochastic model like this might have a partial application to the
type of thing pointed to in the eonic model. However... Consider a
case where a similar phenomenon has an effective method, e.g. the
study of economic cycles. Instead of modelling some 'economic
phenomenon' with a predictive or control model, we stand as immersed
agents in the present looking backwards at any structure
characterizing free activity, economic cycles. Whatever their 'causes'
we remain, within limits, free agents able to reapply this information
about the past form the present onwards, e.g. modulate a slump. We
have the ability to modify the system based on new information. This
simple fact puts the subject forever beyond the standard type of
analysis. I was reading a recent book on long waves and
Kondradieff. The author complained that Greenspan wrecked his
predictions! Apparently Greenspan saw it coming and changed
course. Etc...

The eonic model is like that. So we need a model that clearly
partitions the past and present. How do that? My type of
discrete-continuous model is built around that, all dynamics is about
the past, then the model changes character in the present and we have
free action. Such things aren't weird, but ordinary. Consider a
computer (determinate program) and a user with a mouse. Exact type of
situation. How does it work?

The determinate system alternates with the user, or free agent, whose
input is a series of options, while the system responds with
determinate programming. This mixed model is clearly evident in
history, as we look backwards. Note that we can study the history of
this interaction, it would be determinate interval, free option
interval, one after the other, with the user's present openended.

The eonic model reflects that. Lo and beyond if we examine world
history we see this alternation of degrees of freedom between a higher
determination, and a relatively free optionality. There are two things
at work, then. And the free agent does not attempt to predict his
future based on the system, but to increase his freedom in that system
by using information he has learned from experience, information he
will use to _exit_ that system. It is an odd type of model, at
first. We simply aren't in the realm of physics here, although cousins
of physics models are evident. But they reflect the exact statement,
by the way, of Kant's Third Antinomy, which see. That antinomy is
broken into two parts, one determinate, one relatively free!! It's a
small world. And this is quite natural, but the core is a
contradiction. Now we see why social science never starts, never
could start, because its basis is a contradiction, an antinomy. And
yet it can start at any time, by embracing that contradiction.

This may be the last series the moderator wants here on this. Let me
say I am gratified by the response to the webpage on the eonic
model. It takes a knack to get used to it. The original relevance to
linguistics was indicated in the language transformatin at a very high
level indeed in the Greek Archaic, justy to remind ourselves of the
fact that we are still inside evolution, and that even things at the
level of art show, not deterministic generation, but just this kind of
combined determinate-optionality indicated. The core pivot of action
then is the creative gesture of the evolutionary transition, i.e some
distinction of consciousness and self-consciousness. Determinate
models don't really work because it is like sheep herding, pushing a
crowd, so to speak. We see evolution over a stupendous scale doing
sheepherding 'soft push' dynamics. It operates at a minimum of
interaction. Science Fiction? Look at the data in the light of a very
simple scheme of periodization. Then what used to be called the Axial
Age stands out for what it is. But its implications are
remarkable. Genuine macrodynamics right under our noses, looking

Again, let me express my skepticism about Darwinian explanations of
the emergence of language. They probably don't work. In fact we have
no data. The example of the Greek Archaic shows that highspeed
evolution leaves poets in its wake. So we must have missed something.

Thanks. Lots to say here. But enough. I hope the model indicated at will prove helpful.

John Landon
Website on the eonic effect
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue