- Permission is given to repost this message in other Lists.
What a way to wake up! Time magazine now has a little piece on the
"Indus script," based apparently entirely on press releases by Rao et
al. (They didn't contact us for sure, or any specialists we know.)
> In 2004, perhaps out of befuddlement and frustration, a group ofWe have already showed that the methods they used in their Science
> scholars declared that the script marked only rudimentary pictograms
> and that the Indus Valley people were functionally illiterate. That
> hypothesis, which caused a minor uproar in the world of Indus Valley
> researchers, was recently rejected by a team of mathematicians and
> computer scientists, assembled from institutions in the United
> States and India. Their study, published initially in April in
> Science and more extensively in August in the Proceedings of the
> National Academy of the Sciences, employed computer modeling to
> prove that the Harappan script communicated language, and has
> reinvigorated attempts to crack what is one of the lingering puzzles
> of ancient history.
paper are useless in distinguishing writing from nonlinguistic
symbols. Data were also presented on that in Kyoto and Singapore -- in
the latter case in a keynote address by Sproat at the biggest annual world
conference in computational linguistics. For the evidence, see:
For unambiguous statistical evidence further backing that Refutation,
and for links to discussions of other computational linguists who
have independently debunked Rao's work, see:
Do these reporters ever do anything besides rewrite press releases
from Rao et al.? Do they ever read any of these papers or phone other
researchers for their opinions?
On the PNAS paper, I had this to say when it came out a few weeks
ago (also contains a link to the abstract of Richard Sproat's keynote
address in Singapore):
In brief: the PNAS paper tells us exactly what we've known since the
1920s, as Michael also has pointed out: that Indus symbols (like ALL
nontrivial symbol systems, nonlinguistic and linguistic alike) have
some kind of positional order in them. No one has questioned the fact
that Indus signs come in clusters, etc., ever since Hunter published his
dissertation on the "script" in 1929. The Soviets and Finns also used
computational methods to "prove" that 45 years ago (!), and it has been
"proven" in many ways by others since.
It has absolutely nothing to do with whether or not the symbols are
linguistic, and there is nothing at all original about the work of Rao
et al. besides the fact that they used *different* statistical methods
than used by many before them to "prove" what no one has
questioned since the 1920s.
Now also on my webpage, announcing a new paper in which we expand on
this and provide new statistical evidence, in a note I put up a few days ago:
> Note added 30 August: A new article by the same group making similarSee also what we already said about the inadequacy of simple statistical
> claims has now been published [in late August] in PNAS. All their
> new paper does is show again what has been unquestioned since the
> 1920s -- and which was shown in fact using other computational
> methods by the Soviets and Finns 45 years ago!: that there is
> "order" of some sort in Indus symbols. But that obvious conclusion
> has nothing to do with whether the symbols are speech-encoding
> [i.e., true writing] or not, since all non-trivial symbol systems
> have the same property, as we have shown now in many publications.
> We're currently writing a paper that will hopefully kill off this
> line of reasoning once and for all -- providing further detailed
> evidence that simple statistical measures like those used by Rao et
> al. are useless in distinguishing writing from garden-variety
> religious and political symbols, alchemical formulae, medieval
> heraldic symbols, or military medals, etc. Our reductio ad absurdum
> of the general approach of Rao et al. will be published in September
methods to distinguish writing from nonlinguistic sytstems in our 2004
paper. The reason we discuss so much overlapping evidence in that
paper (archaeological, statistical, linguistic) is precisely because
simple statistical methods aren't capable of making such a distinction.
Rao et al. don't discuss a single inscription but base their work
*totally* on a 1977 (!) concordance of inscriptions by Mahadevan that
*already* contained tables of symbol clusters in an appendix.
Our paper again:
As noted above, we are currently writing a formal response to the PNAS
paper that not only points this out but demonstrates it in respect to
nonlinguistic sign systems. But we don't have the resources to respond
to every news story of this nature, which propagate like
rabbits in heat. Our paper will be published later this month and
posted everywhere we can. These stories in mass circulation magazines
have their impact, however, and they NEVER publish retractions.
The quality of popular science stories has ironically gotten much
worse since Internet arrived on the scene. Explaining that is a bigger
puzzle that needs to be "cracked."
Frustrated and befuddled indeed,