The UPenn IE Tree (a test)

X99Lynx at X99Lynx at
Thu Aug 26 05:56:45 UTC 1999

(First let me say that I've looked at the chart in Larry Trask's Historical
Linguistics (p.369) so many times the page is starting to fall out.  And I am
aware that the one used by Sean Crist is a slightly different version.)

<< So here at this first juncture:

>                                     PIE
>                                    /   \
>                                   /   Anatolian

> Does this mean that PIE co-exists with Anatolian?  It would have to,
> wouldn't it?

No.  An ancestral language cannot co-exist with its own descendant.>>

Now I am not saying I have a problem with this.  I'm sure there are good
methodological reason for the rule that "an ancestral language cannot
co-exist with its own descendant."  (But if it is only terminological, I
might ask why it is necessary.)

However, from a historical viewpoint, from the view of any modern science
whose subject is considered in time, the specific system reflected in the
Stambaum seems odd. (Repeating again that I'm not challenging the linguistic
basis of the chart.)

Sometimes you learn something about systems by putting them through some
extreme applications.  Like sticking a thermometer in a bucket of ice.  I'd
like to try a simple little test.

Let me give a pure hypothetical.  The names of the languages may be the same
as in the Stammbaum, but otherwise this is nothing but an imaginery situation.

Take our Stammbaum and lets arbitrarily put "Celtic" at the top where PIE is
now.  "Celtic" is the first ancestor on our tree.

Now add a factual premise.  That - throughout the relative time period
represented in all the branchings of the Stammbaum to the present - a
"language" that its speakers called "Celtic" at the onset persisted in its
original form without change, from day one to the present.

Throughout this whole time, there was complete continuity and zero
innovations in that "language", and therefore no "innovations" were shared
with any of the other languages that are represented in the branches.  Lets
call this group of identical languages "Celtic 1...Celtic 6"

Now lets say the branchings are pretty much the same as they were in the
original Stammbaum - so we now have:
                                   /   \
                                  /   Anatolian
                                /  \
                               /   Tocharian
                             /  \
                            /    \
                           /       \
                          /          \
                         /         Italic
                       /  \
                      /    \
                     /     /\
                    /     /  \
                   /   Greek Armenian
                 /  \
                /    \
               /      \
              /        \
             /         /\
            /         /  \
           /         /    \
          /         /      \
        Indo-   Balto-    Germanic
       Iranian  Slavic

(I'm not sure about I-Ir, which seems to be part of the stem.)
But every branch reflects a set of innovations placed at a relative point in
time.  Celtic 1....Celtic 6 however has no innovations and starts first in

According to LT, we cannot call Celtic 'Celtic' anymore after Anatolian
branches because a daughter exists (Anatolian).  So lets call it C2.  C2 is
identical in every way shape and form to Celtic.  All innovations (in
relation to Celtic) only exist in Anatolian.

Same scenario with Tocharian.  Tocharian innovates and becomes a different
language than C2 - which remained identical to Celtic.  We also know that
Tocharian is solely decended from C2.  The resulting C3 on the other hand
remains identical to C2 and Celtic.

This continues to the current day, so that the present C6 is identical to the
parent Celtic, sharing none of the innovations of any of the branches that
have emerged along the line.

How would these daughter languages - identical to the original parent except
in name - be indicated on this Stammbaum?  I don't know.   Would they get a
branch or are they the stem?  I don't know.  Or would it look exactly as it
does above?

(NOTE:  Putting the rule about co-existing ancestors aside for the moment, a
historian or everyday person might call Celtic through C6 by one name, since
all are identical.  But the existence of those other daughter branches made
us rename "Celtic" each time a new daughter arrived.)

Now to test the system.  Some new assumptions:

1.  We have no record of the original language - Celtic.

2.   We have no record of C2, C3, C4, C5.

3.  Only C6 is recorded, datable to the latest branch off (I-Ir, B-S, Ger?)
in relative time.

4.   We have a record of all other languages and the relative times are about
the same as those reflected in the original Stammbaum.

Some things that should be clear from this little experiment:

1.  Generally, we would have no way of knowing that the apparent shared or
totally unique 'innovations' in C6 are not "innovations" at all, but reflect
entirely the first parent.

2. In fact we would probably identify any unique attributes in C6 as recent
unshared innovations.  They would not be found in the other recorded
languages, but in fact they would be original attributes lost in the
innovating daughter languages.

3. Our assumption was that Celtic...C5 (identical languages) were the
immediate parents of each branch-off language.   But because we have no
record of those Celtic....C5 languages we might find parent-daughter or
subgroup relations where there are none (based on inherited commonalities
lost in some of the innovating daughters)

4. The number of apparent 'shared innovations' between C6 and some of the
other branching languages might suggest a closer relation with some rather
than others.  But our assumption was that in fact all were equidistant.

5.  We would not reconstruct the parent anywhere close to C6, assuming it to
be a recent and maybe odd daughter.  But in fact that would be the only
accurate reconstruction.

It really isn't pertinent to say that this could never happen.  We could
always adjust the scenario enough to make it more possible.

Whether it is probable or not does not matter.  The point is that if it did
happen, the Stammbaum with its given assumptions, would not be able to
reflect these events accurately.  (But it would give the appearance of an
accurate solution.)

I think that this happens because the system is based on innovations but not
conservations.  If you only measure the vectors of change, the vectors of
continuity become invisible.  A little like reconstructing the lineage of
dinosaurs by assuming they are all similar, and then measuring how much they
differed.  Rather than trying to find a way to rationally measure the
similarities in the first place.

On the other hand, I have no better system to offer.

This does I hope explains my awkward question about how long PIE could
continue to co-exist with its daughter languages.  And how the answer might
affect the direction of reconstructions.

Steve Long

More information about the Indo-european mailing list