Saturday, November 19, 2011

An Indus Manuscript in Afghanistan?

Tablet M-1405 A and B: POT-HATTED BEARER, side A, on right above,
followed by an unusual form of anthropomorph -- a sign or part of the icon?
An equally odd anthropomorph, with both arms raised, appears on a
purported manuscript from Kabul, Afghanistan.  Is MAN intended?

I have not posted recently as I have been working on my database on Indus inscriptions.  I have been going through the concordance of Koskenniemi and Parpola (1982), looking for each of their noted inscriptions in my database.  It is a slow and tedious process, because I look up many of their inscriptions in two or three different ways due to our often different “readings.”  It will be some time before I completely finish the process.

The purpose of doing this is, first, to correct errors that I have made (especially frequent on the difficult-to-“read” tablets) and, second, to establish whether similar symbols are indeed independent signs or variants of a single sign.  Since I began my classification by counting the strokes required to draw each sign, minor differences in the number of stripes (as in the STRIPED TRIANGLE, STRIPED VEST, etc.) or prongs (as in the TRI-FORK, QUAD-FORK, etc.) included in a symbol distinguished independent signs.  It is now clear that most of this type of variation is not meaningful.  For example, the STRIPED TRIANGLE appears in essentially the same contexts regardless of the number of internal stripes.
Seal M-58 with inscription: CIRCLED VEE / BI-QUOTES //

As a result, I no longer distinguish the CIRCLED BI-FORK from the CIRCLED TRI-FORK, and so on.  All are now encoded CIRCLED FORK.  In fact, a short post that is encircled (PACMAN) is merely another variant of the same sign.  A circle without any stripes at all remains distinct from the striped variety.  This is a general rule, as striped versions of a sign appear in a set of contexts that differ from those in which the unstriped versions appear.  Additionally, a striped symbol occasionally occurs in the same inscription as an unstriped symbol of similar shape.  This indicates two independent signs, one striped and one empty.  Generally, the seal carvers seem to have included more stripes/prongs on the larger seals containing few symbols.  On smaller items and where a relatively large number of signs are crowded together, there are fewer prongs and stripes.
Tablet H-187A with inscription (right to left): STRIPED TRIANGLE (4 stripes) /
TRIPLE TRIANGLES / POT / COMB (note the first sign differs in stroke count
from that in the previous inscription, but pairs with the same following sign, which
indicates that the number of stripes does not distinguish independent signs).

In a few cases I retained the distinct names where it is not entirely clear to me that Koskenniemi and Parpola’s groupings are correct.  For example, they consider two common motifs to be essentially one sign, CIRCLED VEE and VEE IN DIAMOND.  Both occur in the same positions, especially in the prefix.  This tends to show variants of one sign.  However, these two “versions” both appear together on rare occasions, an indication that they are independent signs (M-855 and M-196).  These scholars are undoubtedly aware of the general rule that two versions appearing together indicates independent signs, but the more common occurrences in the same contexts must have convinced them to ignore the anomalous inscriptions (1982: 196-200). 

In most cases, though, Koskenniemi and Parpola’s groupings of signs are clearly correct.  They do not distinguish the versions of PANTS, for example, since all variations appear in much the same contexts.  They do not distinguish BELTED FISH from SLASH IN FISH or BACKSLASH IN FISH.  I am changing my own database to reflect this, noting all as MARKED FISH.  Similarly, they do not usually find mirror images of a sign to be distinct, so the LOOP ARMED MAN HOLDING SLASH (or BACKSLASH) now has the slightly simpler name, LOOP MAN HOLDING SLASH.  The SKEWERED CHEVRON and ANKH signs should probably be merged, as these two are grouped together into one section of the KP concordance.  In fact, in several cases, I identified a particular sign as SKEWERED CHEVRON but it is shown as ANKH in KP, or vice versa.  This fact also suggests that these two are really just variants of a single sign. 
Detail from seal M-103 with inscription: CARTWHEEL / BI-QUOTES // FOUR QUOTES / TRI-FORK.
This and other common prefixes fail to appear in the manuscript from Kabul, just as apparent numerals
comprising short strokes (QUOTES) do not occur there -- two suspicious findings.

Working through the concordance in this painstaking manner assists in recognizing pairs of signs that often go together, as noted.  I only mentioned a few of these in earlier posts, based on the observations of M. Korvink.  The sign pairs that include apparent numerals are the ones I find most intriguing.  For example, several types of “numeral” commonly appear before FISH, and many of the same ones also frequently appear before one or another type of FORK.  This regularity is part of the evidence that various researchers use to conclude that numerals – and other modifiers – must precede substantives in the language underlying the Indus script.  But some of the “numerals” occur most often following another sign.  For example, STACKED SEVEN normally follows EF TOPPED EXIT or BATTERY.  And the CUPPED SPOON/CUPPED POST group frequently comes before THREE POSTS.  Thus, there is not a clear-cut preference for “numerals” as a group to either precede or to follow.  It depends on the particular sign pair.  Given such equivocal evidence, one cannot really declare the Harappan language to be “right-branching” or “left-branching.”  As is so often the case with this script, the overall impression is that the script is not closely coupled to any particular language.
Round seal from those in the British Museum, this one found in 1883.
DOUBLE POTS (the only instance of this doubled terminal).

This brings us to the matter of a supposed manuscript containing Indus symbols that is displayed in a museum in Kabul, Afghanistan.  This “manuscript” surfaced around the same time that Farmer, Sproat, and Witzel published their famous paper entitled “The Collapse of the Indus Script Thesis” (2004).  That timing arouses suspicion immediately.  Now a paper has become available online describing this manuscript and comparing the symbols found on it with those in the Indus script (Zuberbühler 2009).  Zuberbühler finds 230 signs, of which 58 are too damaged to be read.  These include between 62 and 76 independent signs depending on how they are analyzed.  Of the 62 most securely identified symbols, 26 are nearly identical to Indus signs listed in Parpola (2009: 70-78).  The author adds another sign to this group, the LOOP or finless fish, based on its appearance among graffiti on pot shards.  Nine of the manuscript signs are mirror images of symbols in Parpola’s list, while eight are simplified variations containing one or more stroke less than the forms shown in Parpola’s list. Four more are probable matches but, being partly obliterated by defects in the manuscript, cannot be seen clearly.  There are 11 more possible matches, generally similar in shape to Parpola’s signs but with more significant variations.  Finally, one sign resembles the SPEAR but contains an additional element not found among Parpola’s listed variants, perhaps to be added to the dubious category.  Thus, of the 62 securely identified symbols in the manuscript, just three are unlike Indus signs.

After comparing the individual signs, Zuberbühler counts the frequency of each.  Her conclusions are that frequencies are broadly similar: “23 of the top 30, 27 of the top 40, 30 of the top 50 and 35 of the top 60 Indus symbols are present on the Kabul birch bark” (2009: 26).  In comparison, the unrelated Easter Island script shares 10 of the top 30, 11 of the top 40, 14 of the top 50 and 16 of the top 60 Indus signs.  These figures suggest a relationship between the Kabul signs and the Indus script.  But what relationship?  Farmer, Sproat, and Witzel issued a challenge shortly after their paper was published, offering to pay $10,000 to the discoverer of an Indus manuscript containing 50 or more signs.  The purpose of their challenge was to underline their confidence that no such manuscript would ever be discovered, because the Indus “proto-writing” system just was not used to create such things.  Should they now prepare to pay up?

Before they consider such a payment, certain conditions must be met.  The first step, clearly, is to determine whether the “relationship” of the Kabul manuscript is simply that of a clever forgery. The bark manuscript could be carbon tested to determine its age, for example.  If it dates to the 20th or 21st century – as most experts would expect – then it is a forgery and no payment is due. 

But let us suppose that our clever forger had discovered, not an ancient manuscript, but an ancient piece of birch bark.  Using an appropriately aged surface, he might have created his fake manuscript by writing on the old bark with a modern felt-tipped pen, carefully counting the signs so as to make them appear in the proper frequencies.  Perhaps the ink, too, would have to be tested.

But even without such tests, which would have to be conducted by reputable labs at well-known institutions, one can test the validity of such a manuscript.  Having worked with the Koskenniemi and Parpola concordance, I would expect at least some of the familiar patterns of Indus signs to appear in any genuine manuscript.  But even a cursory glance at the supposed manuscript shows that they do not.  Korvink, for example, established three segments occurring in authentic Indus inscriptions: prefix, medial section, and terminal.  I looked for these three segments in the data shown in Zuberbühler’s paper, with disappointing results.

No prefixes at all appear in the Kabul manuscript.  There may be three occurrences of BI-QUOTES, but only once alongside one of its most common variables, the VEE IN DIAMOND.  And in this one instance, since the direction of the script is from right to left, the VEE IN DIAMOND is on the "wrong" side.  Likewise, there are two possible occurrences of the PINCH.  Of these, one occurs between two terminal signs, something it never does in authentic inscriptions.  In the other instance, it is so oddly written that its identity is not clear.  There remains the SINGLE QUOTE which seems to appear only once, alongside one of the symbols that has no clear match among the Indus signs.  There are odd inscriptions in the familiar Corpus (including a single instance of a doubled POT), but none contains as many different peculiarities as the Kabul manuscript.

There are a few terminal signs, as Zuberbühler identifies the POT and SPEAR on the Kabul artifact.  Among the Indus terminal signs, though, some of the more common ones are altogether missing from the Kabul manuscript.  Those that do appear are clearly not behaving as they do in known Indus inscriptions.  In one area, three different terminal signs are bunched together in an odd fashion (a single POT-HATTED BEARER is immediately followed by SPEAR in the second row, while a SPEAR / POT sequence occurs in the seventh).  Elsewhere, the POT simply seems to pop up now and then -- usually after two or three signs.  More peculiarly, the COMB seems to be confused with a multiply striped rectangle that does not occur in the Indus script.  Thus, although some of the Indus terminal signs are present, there is no clear terminal function in the Kabul manuscript.
Tablet H-216 A and B sides: STRIPED CHEVRON / STRIPED VEST / POT
(A side, reading from left to right, which is unusual); CUP / THREE POSTS
(B side, showing the most common grouping of one to four posts, with CUP).

The medial section remains to be discussed, an element more difficult to analyze because it is more variable.  Among the more numerous signs that occur in the medial section of known inscriptions, the various types of FISH are prominent on authentic artifacts.  Some fish-like symbols do occur in the Kabul manuscript, drawn a bit oddly, while others that we would expect to see are not present (no MARKED FISH, for example).  More damning, the ones that do appear do not follow the standardized sequences that Korvink outlined. 

Similarly, there are a few apparent numerals in the Kabul manuscript as well.  But again, not all of the more common types found in Indus inscriptions make an appearance here.  In particular, there is only one “stacked” form, the STACKED TWELVE.  There are no “shorts” aside from the possible SINGLE QUOTE and BI-QUOTES.  Of the few “longs” that occur, frequency statistics from Indus inscriptions lead us to expect SINGLE POST, TWO POSTS, THREE POSTS, and possibly FOUR POSTS (up to six occur on the rare tablet and seven are even more rare on pot shards).  However, the only reason these “numerals” appear among the 60 most common Indus signs is because of their frequent occurrence on tablets alongside CUP.  In spite of this close association, the CUP is missing from the Kabul manuscript.  A peculiar version of CUPPED POST occurs several times on the Kabul artifact, but it does not pair with THREE POSTS.  An oddly angular FISH appears on the Kabul artifact as well, but it does not pair with TWO POSTS.  Thus, expectations are not met even if we look only at expected pairs with the four apparent numerals that do occur in the Kabul manuscript.
As a result of the lack of expected patterning, most scholars who have weighed in on the matter consider the Kabul manuscript a forgery.  Farmer, Sproat, and Witzel are not going to pay for that.  Nor should they.