12-21-2005, 05:35 AM
http://www.sepiamutiny.com/sepia/archives/000834.html
Dear Sepia Muntiny,
I'm writing in reponse to you posting on the Indus Script.
Dave Kelley once told me that: "The only thing that is worse that having your research attacked is having it ignored". I was hoping to ignore Farmer et al in the same way I will be ignoring Barua and his Tantric decipherment and the flurry of competing Sanskrit decipherments. Unfortunately, in the light of recent developments I may not be able to. There are several serious error in the Farmer et al argument, so many in fact that it is not possible to discuss them all here. I offer this rather lengthy e-mail as a critique of the most obvious as serious of these errors.
Here is what's wrong with Farmer et al's approach:
1) They makes the argument that because there are so many singletons in the Indus sign list it cannot be writing. They do not command any ancient script so they have no idea of how these scripts work. Out of curiosity I downloaded the Proto-Sumerian sign list from the Cuniform Library Inititive web page. I compared the frequency of signs to that of the Indus script. The sign frequency is nearly identical. In fact the r2 for these distributione is 0.97. The same data is available for the Proto-Elamite script which has an even a high percentage of singletons. It seems that the high frequency of singletons is not proof that the Indus Script is not writting but rather that it is a normal linear scripts from South Asia.
2) They Says: "Inscriptions consist of high frequency signs that rarely repeat even in the longest inscriptions". Then in their proof that the "duck in a pond" texts is not writing (his Case #1) they give the following reason: "The most common Indus sign...shows up no less than three times in this six sign inscription". He is only partly right in both case. There are some examples of the same sign repeated in a given text. I don't know the exact count yet, but I will get my programer to count them. This point is important to Farmer et al because they is assuming the script is "heavily syllabic". I count ââ°Ë 700 signs, way too many for a syllabic script. It is most likely logo-syllabic, with a high probability of at least some determinatives being used. Also In the "duck in a pond" text the two segments of the text are separated by a iconographic element (Bull). It looks to me like two 3 sign texts.
This is a very rare artifact (a silver seal). Not that Farmer et al ever address the bulk of the corpus except in a gross statistical way. They focus instead on rare artifacts types and unusual texts.
3) They say there are no evidence for perishable texts in the Indus Corpus. This is not true. We have the impression on the back of a clay sealing of a text that was carved on a wooden doweling (Mackay, Vol. II, Plate XC:17) and has 9+ signs. The text runs the full length of the sealing and seems to run off in both directions. It is in very rough shape and the sealing picks up only the bottom of the signs. Nevertheless., this is proof of two things. First, Farmer et al have no clue as to what is in the Corpus (or chooses to ignore data that are in opposition to his views) and second, that Indus people carved texts in wood.
4) Farmer et al's treatment of Indus Numbers would be comical if it were not such a serious topic.
a) On the one hand they complain Indus numbers are mostly "2s and 3s in several morphological types". Similar I would guess to 3, 30, 300 etc are in our system. That is, he conflates linear short strokes, stacked short strokes, and linear long strokes into one set of numerals. Strangely, he ignores the very common occurrences of short and long one stroke signs? I have no idea why. He goes on to complain that "Frequently apparent numbers are grouped with other numbers in idiosyncratic ways". He is referring to a seal texts with one long stroke, followed by 3 short stacked strokes, followed by five short stacked strokes. This looks far from "idiosyncratic" to me. In fact it is common in positional notation in all ancient scripts to have strings of numeral which combine systematically to form higher order numbers. In this case one could guess 135 or 1035? I am aware of 20 or so examples of positional notation without really looking.
b) They also completely miss the simple right bracket ")" as five, and several other signs, that from their contexts in replacement sets, may also be numbers or minimally their are functional equivalents.
c) Farmer et al point out that "Certain apparent numerical signs are regularly found in conjunction only with specific non-numerical signs, never with others -- in a way that again seems peculiar for an abstract number system." Again not well thought out. First, "only with specific non-numerical signs" translates from Farmer speak to English as: 16 sign very frequently, and occasionally with 60 additional signs. That is, "only with specific non-numerical signs" = 76 or about 11% of all signs. This is also a common pattern in ancient writing systems: that numbers occur with the nouns the discribe -- 10 sheep, 15 people etc.
5) As if to reinforce their complete lack of knowledge about the Indus Corpus, Farmer et al repeatedly refer to HR3005 (M-0314) as the longest indus text. It is really the forth longest text. It is the longest seal texts, but there are several bas relief tablets with more signs.
6) Farmer et al make much of the terseness of Indus Texts (now about 4.87 signs). Don't they know the average length of a texts from Nissans Uruk data is 6.8 signs? Of course there are some really long ones, but the vast majority are about 3-5 signs long. They consist of noun+number constructions with occasional totals in the longer texts.
7) As to there discussion of the high frequency signs I refer them the recent article by Peter Damerow of the Max-Planck-Institut (Berlin) on Proto-Elamite. As with the signleton signs, the high frequency signs occur in nearly identical proportions in Proto-Elamite, Proto-Sumerian, and the Indus Script.
8) It has also been said that inscribed Indus artifacts occur mostly in garbage heaps and in refuse. This is also a gross exaggeration. In fact there are many good examples of seals clustering in specific houses and near kilns at Mohenjo-daro. There is also a clustering of seals and weights in specific rooms in some of these houses. A careful examination of the excavation data from Mackay allows the tracking of changes of seal use through time based on these distributions at Mohenjo-daro DK.G area. The simple fact is that Farmer et al have no clue about either the archaeology of the Indus Valley, where the seals come from in these deposits, or how inscribed artifacts were used.
MY question is why do publisher continue to print garbage on the Indus script? Couldn't they find a competent referee for the Farmer et al paper? It also suprises me that the basic facts concerning the components and mechanics of ancient scripts seem to be unknown, not just the media but in the academic community as well.
There is a lot of valuable research into the Indus script, unfortunately it is the more bizarre work that attracts the attention of the press. This is too bad because the informed work on Indus Writing is so much more interesting, if somewhat less simplistic than Farmer et al's perception.
Best Regards
Bryan Wells
Traveling Scholar
Harvard University
The arguments of Farmer in this paper are easily
falsified. He claims that the Harappans could not have
had writing because the seal inscriptions are too
short. This opinion is arrogant and Eurocentric. The
size of the inscription does not define its existence
as writing or non-writing.
The research of Farmer et al lacks validity, fails to support their conclusions and is contradicted by their own statistics. For example, Farmer et al make it clear that the mean word length for comparable Egyptian text is 6.94 and Indus text 7.39, this shows no statistical difference and should have alerted the researchersââ¬â¢ to the fallacy of their arguments. In addition, Dr. Gunter Dryer, an Egyptologist, has found Egyptian text with as few as two (2) symbols that phonetically readable. This is evidence that the literature review of the authors does not reflect the actual knowledge base for ancient writing.
For example, Egyptian writing discovered in 1998,
by Dr. Gunter Dreyer, director of the German
Archaeological Institute in Egypt, has one to four
signs with each sign being a CV lexical item. This
writing is recognized as "true writing". Farmer
maintains that no writing can have meaning with only a
limited number of signs, yet these Egyptian clay
tablets number in the hundreds.The presence of a large number of ancient inscriptions written with a limited number of symbols that have a phonemic identity falsifies the theory of Farmer,
since the foundation of his theory rest on the brevity
of the Indus seals and singletons.
Finally, Farmer et al maintain that their work is falsified (p.48) when someone publishes a clear set of rules for Indus writing that can be used to interpret the writing. This was done and was published a decade ago, see:
_________. (1994c). Ancient Dravidian: And introductory
grammar of Harappan with
Vocabularies, Journal Tamil Studies, No.41, 1-21.
_________.(1995a). Ancient Dravidian: The Harappan
signs, Journal Tamil Studies,
No.42, 1-23.
__________.(1995b). Ancient Dravidian: Harappan
Grammar/Dictionary, Journal Tamil Studies, No.43-44,pp.59-130.
You can find an electronic version of these papers at:
http://us.share.geocities.com/olmec982000/HarWRITE.pdf
or
http://geocities.com/olmec982000/HarWRITE.pdf
In Farmer et al's paper they claim that the Dravidian theory for Harappan was rejected. You will notice that they discussed some of the Dravidian based attempts at decipherment, but they failed to discuss mine. Farmer can not claim he does not know about my decipherment because I referred him to my decipherment during our recent on-line debate on one of the Yahoo groups last Spring on his thesis that Indus writing can not be phonetically read. Due to our exchanges on this forum, I know he read what I wrote. This is confirmed by the fact that in his new paper he did not use the Zipf Law to support his theory, because I showed how it did not fit his theory, and supported the view that Indus writing is phonetically readable.
In conclusion the theory of Farmer et al is without foundation. They do not prove that the Harappans were illiterate, because many short inscriptions from other ancient civilizations can be read phonetically.
Dear Sepia Muntiny,
I'm writing in reponse to you posting on the Indus Script.
Dave Kelley once told me that: "The only thing that is worse that having your research attacked is having it ignored". I was hoping to ignore Farmer et al in the same way I will be ignoring Barua and his Tantric decipherment and the flurry of competing Sanskrit decipherments. Unfortunately, in the light of recent developments I may not be able to. There are several serious error in the Farmer et al argument, so many in fact that it is not possible to discuss them all here. I offer this rather lengthy e-mail as a critique of the most obvious as serious of these errors.
Here is what's wrong with Farmer et al's approach:
1) They makes the argument that because there are so many singletons in the Indus sign list it cannot be writing. They do not command any ancient script so they have no idea of how these scripts work. Out of curiosity I downloaded the Proto-Sumerian sign list from the Cuniform Library Inititive web page. I compared the frequency of signs to that of the Indus script. The sign frequency is nearly identical. In fact the r2 for these distributione is 0.97. The same data is available for the Proto-Elamite script which has an even a high percentage of singletons. It seems that the high frequency of singletons is not proof that the Indus Script is not writting but rather that it is a normal linear scripts from South Asia.
2) They Says: "Inscriptions consist of high frequency signs that rarely repeat even in the longest inscriptions". Then in their proof that the "duck in a pond" texts is not writing (his Case #1) they give the following reason: "The most common Indus sign...shows up no less than three times in this six sign inscription". He is only partly right in both case. There are some examples of the same sign repeated in a given text. I don't know the exact count yet, but I will get my programer to count them. This point is important to Farmer et al because they is assuming the script is "heavily syllabic". I count ââ°Ë 700 signs, way too many for a syllabic script. It is most likely logo-syllabic, with a high probability of at least some determinatives being used. Also In the "duck in a pond" text the two segments of the text are separated by a iconographic element (Bull). It looks to me like two 3 sign texts.
This is a very rare artifact (a silver seal). Not that Farmer et al ever address the bulk of the corpus except in a gross statistical way. They focus instead on rare artifacts types and unusual texts.
3) They say there are no evidence for perishable texts in the Indus Corpus. This is not true. We have the impression on the back of a clay sealing of a text that was carved on a wooden doweling (Mackay, Vol. II, Plate XC:17) and has 9+ signs. The text runs the full length of the sealing and seems to run off in both directions. It is in very rough shape and the sealing picks up only the bottom of the signs. Nevertheless., this is proof of two things. First, Farmer et al have no clue as to what is in the Corpus (or chooses to ignore data that are in opposition to his views) and second, that Indus people carved texts in wood.
4) Farmer et al's treatment of Indus Numbers would be comical if it were not such a serious topic.
a) On the one hand they complain Indus numbers are mostly "2s and 3s in several morphological types". Similar I would guess to 3, 30, 300 etc are in our system. That is, he conflates linear short strokes, stacked short strokes, and linear long strokes into one set of numerals. Strangely, he ignores the very common occurrences of short and long one stroke signs? I have no idea why. He goes on to complain that "Frequently apparent numbers are grouped with other numbers in idiosyncratic ways". He is referring to a seal texts with one long stroke, followed by 3 short stacked strokes, followed by five short stacked strokes. This looks far from "idiosyncratic" to me. In fact it is common in positional notation in all ancient scripts to have strings of numeral which combine systematically to form higher order numbers. In this case one could guess 135 or 1035? I am aware of 20 or so examples of positional notation without really looking.
b) They also completely miss the simple right bracket ")" as five, and several other signs, that from their contexts in replacement sets, may also be numbers or minimally their are functional equivalents.
c) Farmer et al point out that "Certain apparent numerical signs are regularly found in conjunction only with specific non-numerical signs, never with others -- in a way that again seems peculiar for an abstract number system." Again not well thought out. First, "only with specific non-numerical signs" translates from Farmer speak to English as: 16 sign very frequently, and occasionally with 60 additional signs. That is, "only with specific non-numerical signs" = 76 or about 11% of all signs. This is also a common pattern in ancient writing systems: that numbers occur with the nouns the discribe -- 10 sheep, 15 people etc.
5) As if to reinforce their complete lack of knowledge about the Indus Corpus, Farmer et al repeatedly refer to HR3005 (M-0314) as the longest indus text. It is really the forth longest text. It is the longest seal texts, but there are several bas relief tablets with more signs.
6) Farmer et al make much of the terseness of Indus Texts (now about 4.87 signs). Don't they know the average length of a texts from Nissans Uruk data is 6.8 signs? Of course there are some really long ones, but the vast majority are about 3-5 signs long. They consist of noun+number constructions with occasional totals in the longer texts.
7) As to there discussion of the high frequency signs I refer them the recent article by Peter Damerow of the Max-Planck-Institut (Berlin) on Proto-Elamite. As with the signleton signs, the high frequency signs occur in nearly identical proportions in Proto-Elamite, Proto-Sumerian, and the Indus Script.
8) It has also been said that inscribed Indus artifacts occur mostly in garbage heaps and in refuse. This is also a gross exaggeration. In fact there are many good examples of seals clustering in specific houses and near kilns at Mohenjo-daro. There is also a clustering of seals and weights in specific rooms in some of these houses. A careful examination of the excavation data from Mackay allows the tracking of changes of seal use through time based on these distributions at Mohenjo-daro DK.G area. The simple fact is that Farmer et al have no clue about either the archaeology of the Indus Valley, where the seals come from in these deposits, or how inscribed artifacts were used.
MY question is why do publisher continue to print garbage on the Indus script? Couldn't they find a competent referee for the Farmer et al paper? It also suprises me that the basic facts concerning the components and mechanics of ancient scripts seem to be unknown, not just the media but in the academic community as well.
There is a lot of valuable research into the Indus script, unfortunately it is the more bizarre work that attracts the attention of the press. This is too bad because the informed work on Indus Writing is so much more interesting, if somewhat less simplistic than Farmer et al's perception.
Best Regards
Bryan Wells
Traveling Scholar
Harvard University
The arguments of Farmer in this paper are easily
falsified. He claims that the Harappans could not have
had writing because the seal inscriptions are too
short. This opinion is arrogant and Eurocentric. The
size of the inscription does not define its existence
as writing or non-writing.
The research of Farmer et al lacks validity, fails to support their conclusions and is contradicted by their own statistics. For example, Farmer et al make it clear that the mean word length for comparable Egyptian text is 6.94 and Indus text 7.39, this shows no statistical difference and should have alerted the researchersââ¬â¢ to the fallacy of their arguments. In addition, Dr. Gunter Dryer, an Egyptologist, has found Egyptian text with as few as two (2) symbols that phonetically readable. This is evidence that the literature review of the authors does not reflect the actual knowledge base for ancient writing.
For example, Egyptian writing discovered in 1998,
by Dr. Gunter Dreyer, director of the German
Archaeological Institute in Egypt, has one to four
signs with each sign being a CV lexical item. This
writing is recognized as "true writing". Farmer
maintains that no writing can have meaning with only a
limited number of signs, yet these Egyptian clay
tablets number in the hundreds.The presence of a large number of ancient inscriptions written with a limited number of symbols that have a phonemic identity falsifies the theory of Farmer,
since the foundation of his theory rest on the brevity
of the Indus seals and singletons.
Finally, Farmer et al maintain that their work is falsified (p.48) when someone publishes a clear set of rules for Indus writing that can be used to interpret the writing. This was done and was published a decade ago, see:
_________. (1994c). Ancient Dravidian: And introductory
grammar of Harappan with
Vocabularies, Journal Tamil Studies, No.41, 1-21.
_________.(1995a). Ancient Dravidian: The Harappan
signs, Journal Tamil Studies,
No.42, 1-23.
__________.(1995b). Ancient Dravidian: Harappan
Grammar/Dictionary, Journal Tamil Studies, No.43-44,pp.59-130.
You can find an electronic version of these papers at:
http://us.share.geocities.com/olmec982000/HarWRITE.pdf
or
http://geocities.com/olmec982000/HarWRITE.pdf
In Farmer et al's paper they claim that the Dravidian theory for Harappan was rejected. You will notice that they discussed some of the Dravidian based attempts at decipherment, but they failed to discuss mine. Farmer can not claim he does not know about my decipherment because I referred him to my decipherment during our recent on-line debate on one of the Yahoo groups last Spring on his thesis that Indus writing can not be phonetically read. Due to our exchanges on this forum, I know he read what I wrote. This is confirmed by the fact that in his new paper he did not use the Zipf Law to support his theory, because I showed how it did not fit his theory, and supported the view that Indus writing is phonetically readable.
In conclusion the theory of Farmer et al is without foundation. They do not prove that the Harappans were illiterate, because many short inscriptions from other ancient civilizations can be read phonetically.