On Indus Civ. signs

Michael Witzel witzel at FAS.HARVARD.EDU
Fri Apr 24 21:44:36 UTC 2009

Dear All,

A propos yesterday's SCIENCE article <http://www.sciencemag.org.>

  "Entropic Evidence for Linguistic Structure in the Indus Script."
By Rajesh P. N. Rao, Nisha Yadav, Mayank N. Vahia, Hrishikesh  
Joglekar, R. Adhikari and Iravatham Mahadevan. Science, Vol. 324  
Issue 5926, April 24, 2009.  (In the Brevia section: Published Online  
April 23, 2009; Science DOI: 10.1126/science.1170391)

This less than 2 page paper is based on *invented* data;  however,  
this fact appears only if you actually read the additional materials,  
not easily available, unless you have a subscription:

Its conclusions about 'script', language, etc. therefore are  
baseless, wrong and misleading:
Garbage in, garbage out, as  has been reported by www.newscientist.com.

Instead, see our (S. Farmer, R. Sproat, M. Witzel) brief refutation,  
published on the same day as the Rao paper, at:



Rao et al. somehow managed to get through the review process at  
Science, though it took them 4 months to do so.

They did so by failing to indicate in their  paper proper that their  
"representative examples of nonlinguistic signs"  are *made-up*  
corpora.  These "non-linguistic signs" lie at the center of their  
argument (i.e., that Indus signs are *not*  nonlinguistic).  But,  
their Type 1 and Type 2 systems of signs (tokens) are radically  
different from anything found in the real world.

If they had said that openly in their paper the paper would never  
have been published.
And, if the press releases had noted that people would not  have been  

Instead, they barely indicate the "assumed" nature of their data, and  
this only in their online "Supplemental Information", which very few  
people will see -- and certainly not those who merely follow the  
current news and internet tsunami.

If they had calculated the 'conditional entropy' (certain signs  
necessarily following others) of ANY *real* nonlinguistic symbol  
system, they would instead have found that there are frequent  
statistical overlaps with linguistic systems. Not unexpectedly also  
with the Indus symbols. In fact, real world nonlinguistic signs will  
fall somewhere in the middle: no sign system is either totally  
disordered or totally disordered.

We (Farmer, Sproat, WItzel) have already shown precisely that for  
symbol frequencies in our 2004 paper that they supposedly refute.  
See:  <http://www.safarmer/fsw2.pdf> (see the chart on p. 27)

Again, if they had calculated the entropy of any genuine, not made  
up, "representative nonlinguistic symbols" they would have found that  
they looked much like writing as well. Take a look the  Scottish  
heraldic signs in our paper.


There are many more technical arguments than the ones we list in our  
little Refutation -- involving gross misuse of the concept of  
conditional entropy, language structure, attestation and localization  
of ancient Sumerian, Vedic and Tamil, etc.  -- that we can lay out in  
a little piece later.

In the margin: the  Rao, et al. paper is depends on an article by  
Claude Shannon "A Mathematical Theory of Communication," in The Bell  
System Technical Journal 27 (1948), pp 379-423 and 623-656. See the  
of the phrase: "conditional entropy."
Rao, et.al., however, do not even bother to give the title of this  
article in their bibliography. The Shannon article is available here  
(see esp. pp. 14-15 of that study):

More to come SOON.


Amusingly, even A. Parpola, who has spent most of his career on the  
'decipherment' of the Indus signs, the  so-called Indus script,  

"It's a useful paper," said University of Helsinki archaeologist Asko  
Parpola, an authority on Indus scripts, "but it doesn't really  
further our understanding of the script."

Parpola said the primary obstacle confronting decipherers of  
fragmentary Indus scripts — the difficulty of testing their  
hypotheses — remains unchanged." (see:  <http://blog.wired.com/ 

And the Guardian, in rather garbled fashion, has him say:
""Language is one of the hallmarks of a literate civilisation. If  
it's real writing, we have a chance to know their language and to get  
to know more about their religion and other aspects of their culture.  
We don't have any literature from the region that can be understood."

To my mind at least, language is a hallmark of ANY human culture,  
whether hunter-gatherer or state society...

Others however, who too have  been heavily involved in this futile  
exercise, such as the Indus archaeologist M. Kenoyer, are even less  

"J. Mark Kenoyer, a linguist at the University of Wisconsin-Madison,  
says Rao's paper is worth publishing, but time will tell if the  
technique sheds light on the nature of Indus script. "At present they  
are lumping more than 700 years of writing into one data set," he  
says. "I am actually going to be working with them on the revised  
analysis, and we will see how similar or different it is from the  
current results." <http://www.newscientist.com/article/dn17012- 

None of them, apparently, has read the supplementary materails  
carefully. O si tacuisses...

I will tell him so in our May Round Table at Kyoto, where Indus  
specialists will get together from Japan, S. Asia, America. Perhaps  
he will then rethink this "working together" with them...

By the way,  Kenoyer is an earth digging archaeologist, not a  
linguist --  not  by any stretch of imagination :^)

Conclusion: read the "footnotes" carefully, not the hype.

Michael Witzel
witzel at fas.harvard.edu

Dept. of Sanskrit & Indian Studies, Harvard University
1 Bow Street,
Cambridge MA 02138, USA

phone: 1- 617 - 495 3295 (voice & messages), 496 8570, fax 617 - 496  
my direct line:  617- 496 2990

More information about the INDOLOGY mailing list