Tuesday, January 11, 2011

Is There a Banyan Tree in the Indus Script?

In his book The Indus Script: A Positional-Statistical Approach, M. Korvink uses the term “banyan” for a sign that I call TRI-FORK TOPPED POT (2007: 28).  The sign resembles the letter “U” with trident-like projections at the tops of each side.  The sign appears in other writing on the Indus script with numerical designations (sign 332{a} in Koskenniemi and Parpola 1982: 21; sign 303 in Wells 1998: appendix; J-4 in Fairservis 1992: 173; my VI51, indicating that it is the fifty-first in my list of six-stroke signs).  Korvink notes that there are variants of this “banyan,” but does not show them.  He analyzes occurrences of his “banyan” based on information in I. Mahadevan’s concordance, to which I lack access.  Korvink also refers to inscriptions by the code numbers assigned to them in the same concordance, codes which do not match the item numbers in the Corpus of Indus Seals and Inscriptions (Vol. 1, Joshi and Parpola 1987; Vol. 2, Shah and Parpola 1991).  This makes my evaluation of his findings a bit difficult.  What Korvink has to say about other, better identified Indus signs seems valid, though, so it is worthwhile researching this topic further. 

Broken Indus seal L-64 with end of inscription: TRI-FORK TOPPED POT / POT.
Korvink states that the “banyan” occurs most frequently before the “jar” sign (2007: 30, 32).  The term “jar” is ambivalent.  It could refer to either my CUP or my POT (the “U” shaped sign or the “U” shape with double prongs on each side, respectively).  In this case, Korvink writes out several inscriptions, clarifying the reference as meaning POT.  POT is a terminal sign, in his analysis, as is the “banyan.”  Other terminals include MAN, SPEAR, COMB, and various types of BEARER (my terms, not his).  He subdivides the terminal class (as he does his prefix class) into constants and variables.  An inscription may or may not have a terminal, so, in that sense, the whole terminal is variable.  For Korvink, a variable is a sign that may or may not appear alongside a constant, but it will not occur unless there is another terminal sign following it.  In contrast, the constant may appear  in the terminal alone, unaccompanied by another terminal.  Thus, one finds inscription with and without terminals.  And in those inscriptions that do contain terminals, constants sometimes appear with and sometimes without variables, whereas variables appear only with constants, never without them. 
Prefixes, like terminals, include constants and variables.  In the case of the prefix, the constant sign does not occur in initial position, but causes the variable to precede it.  Thus, a constant must follow a variable, but a particular constant is not invariably paired with the same variable.  The constants are few: SINGLE QUOTE, DOUBLE QUOTE, PINCH.  The variables, on the other hand, are many and often may occur with more than one constant.  Further, the constant does not occur outside this end-of-prefix position.  However, a sign that sometimes occurs in a prefix as a variable may also occur in medial or final position in an inscription, as long as the constant is not there.  The situation is a bit more complicated than that, but that is the gist of Korvink’s thesis. 
In the terminal portion of the inscription, the signs follow a certain sequence.  Korvink lays out this sequence in the form of a grid.  From beginning to end, the terminal follows the following sequence: (1) BANYAN, (2) COMB (“b” position), (3) SPEAR, (4) POT, (5) MAN, (6) CHEVRON-HATTED BEARER, (7) POT HATTED BEARER, (8) (SIMPLE) BEARER, (9) COMB (“a” position).  No inscription contains all of these terminals.  Typically, an inscription has one, two, or three terminals in sequence, but that may be the limit. 
In Korvink’s grid, the “banyan” (shown as my TRI-FORK TOPPED POT) occurs once apiece before a number of terminal constants: the simple BEARER, the CHEVRON HATTED BEARER + POT HATTED BEARER pair, the COMB, and the MAN.  The “banyan” appears before the SPEAR eight times.  The most frequent pairing is before the POT, which appears 110 times.  In all, the “banyan” occurs 118 times in the corpus of inscriptions, according to Korvink’s figures, which do not add up, to my mind (110 + 8 + (4 x 1) = 122).  Each of the signs that follows the “banyan” is a constant terminal that occurs elsewhere without the preceding “banyan.”  The “banyan,” though, does not occur as a terminal without a following constant (Korvink’s analysis again). 
Note that, by this analysis, the opposite sequence should not occur, i.e., one of the constant terminals should not precede the variable, “banyan.”  However, Korvink admits that there is one inscription with the “banyan” in which the reverse order does appear, i.e., POT + “banyan” (which I find to be on a tablet from Harappa, H-302).  This inscription has a medial portion preceding the terminal section and another terminal sign following the out-of-sequence “banyan,” i.e., COMB (Joshi and Parpola 1987: 225).  The inscription reads from right to left: BI-RAKE / FISH / POT / TRI-FORK TOPPED POT / COMB.  Here, the terminal portion contains three signs (POT / TRI-FORK TOPPED POT / COMB), following two medial signs, hence, this instance cannot be a matter of reading in the wrong direction.  Korvink suggests that it could easily be due to “scribal error” (2007: 30).
In one other anomalous inscription, M-1103 on a seal from Mohenjo daro, the “banyan” occurs in medial position.  Korvink does not give an explanation for this, but it is presumably similar to that described earlier in his book for signs in the prefix.  When variables appear in the prefix, e.g., CARTWHEEL, CIRCLED VEE, or VEE IN DIAMOND preceding the common constant BI-QUOTES, both signs are classified as parts of the prefix.  In each case, the first sign is the variable and the second the constant.  Each of these variable signs also occurs outside the prefix, i.e., without the BI-QUOTES following it.  His explanation is that the presence of the constant is what causes the variable to take the preceding position (2007: 22). 
Seal K-43 with inscription (reading from top to bottom and left to right): CRAB / REVERSED CRAB / POTTED TWO //
Presumably, there is a similar explanation for the terminal portion of inscriptions.  The constant in the terminal causes the variable to take its position.  In the case of inscriptions other than M-1103, the POT (or SPEAR or MAN) causes the “banyan” to take the first position in the terminal, immediately after the rest of the inscription.  But when the POT (or SPEAR or MAN) is not there, the “banyan” can appear elsewhere.  As a result of this principle, in M-1103 the TRI-FORK TOPPED POT is in medial position: BLANKET / CUPPED STRIPED SPOON / TRI-FORK TOPPED POT / STRIPED MALLET / CARTWHEEL / SINGLE QUOTE (reading from left to right; Shah and Parpola 1991: 116).  In another inscription, on seal K-40 from Kalibangan, the same sign is in final position: FIVE TOED FOOT / BATTERY / DOUBLE CIRCLES / VEE IN DIAMOND / BI-QUOTES / TRI-FORK TOPPED POT (left to right again; Joshi and Parpola 1987: 305).  This inscription is analyzed as comprising a prefix of five signs ending with the constant BI-QUOTES, followed by a medial section comprised of a single sign, the TRI-FORK TOPPED POT.  Although this sign is the last one in the inscription, it is not a terminal element in this case, because it lacks a companion sign that can be classified as a terminal constant.
Thus far, Korvink’s analysis seems valid.  I do have a minor quibble, based on terminology.  While Korvink does not state that these kinds of positional regularities necessarily indicate grammatical elements, these seem to be implied by the terms he uses, especially by the word “prefix.”  To his credit, he notes that previous researchers clearly confused these two levels of analysis.  He also makes a case for not making this mistake and states that his analysis is only positional.  I think his case would be improved by using somewhat different terminology which would not seem to imply a grammatical analysis.  In studies of proto-Elamite, a contemporary symbol system, Damerow and Englund use the term “header” for a group of symbols that appear at the beginning of some texts (the “hairy triangle”; 1989).  I think it would be better to use this same term in place of “prefix.”  That suggests the opposite term “footer” instead of “terminal.”
In my view there are a couple of other anomalies.  Although Korvink does not say this, I get the impression from reading his book that he saw the medial segment of the inscription as the one part that was obligatory, while both his prefix and terminal were optional.  Other scholars, who claim to have discerned the meaning of specific signs (something Korvink does not do), certainly claim that the terminals are suffixes.  Neither of these appears to be correct.  According to my own observations, although there is some validity to Korvink’s three segments, an inscription may contain any one of these.  In other words, all three are optional.  The medial segment is there most often.  If we abbreviate the three sections as P (for prefix), M (for medial), and T (for terminal), I can state my findings more succinctly and clearly.  I see inscriptions of the following types: P, M, T; PM, MT, PT; PMT, and longer inscriptions that combine two of these (e.g., PMTM and so on).  The odd ones that don’t seem to follow Korvink’s analysis are not particularly common, but they are there.  Korvink himself specifically notes some of these instances.  Other researchers often fail to account for anomalies, such as instances where described as a suffix occurs alone.  Suffixes should not do that.  At the very least, such a sign would seem to require at least two meanings, only one of which can be suffixal.
That aside, I still need to determine which of the other signs in various lists fall under the rubric of “banyan.”  Korvink states that the latter sign occurs 118 times in the corpus of inscriptions, based on Mahadevan’s figures.  Wells gives the frequency of the TRI-FORK TOPPED POT as 30 occurrences, so there must be a large number of other inscriptions not accounted for.  The first possibility for a variant “banyan” is my BI-FORK TOPPED POT (Koskenniemi and Parpola 332{c}, Wells 305, Fairservis J-9, my VI53).  This is another “U” shape with only two prongs at the top of each “arm” similar to the letter “Y.”  This is almost certainly one of  the “banyan” variants, since a few inscriptions on tablets are duplicates but for this difference.  For example, on tablet H-879 the inscription reads (from left to right): STRIPED DOOR & KNOB / BI-FORK TOPPED POT / POT (V) / COMB (5).  The next tablet, H-880, reads (left to right): STRIPED DOOR & KNOB / TRI-FORK TOPPED POT / POT (V) / COMB (6).  Next is H-881 with the same inscription as 879 (with BI-FORK TOPPED POT, though the COMB has only 4 teeth now).  H-882 duplicates H-880 (with TRI-FORK TOPPED POT, plus the COMB has only 4 teeth).  The next one, H-883, might have been a duplicate, since it ends with POT / COMB.  But the first part is broken off.  A tiny bit at the top of the right side remains of the previous sign.  But it is not enough to clearly determine whether it had two or three prongs.  My guess would be that it is another TRI-FORK TOPPED POT.  Either way, it fits the general pattern of these others.  These near duplicates strongly suggest that the two symbols, BI-FORK TOPPED POT and TRI-FORK TOPPED POT, are variants of one sign, the “banyan.”  When Korvink uses the term variant, this does not indicate that these two symbols necessarily have the same lexical meaning (footnote 2007: 30).  Such replacement patterns may indicate this, but it is also possible that the symbols only served a similar syntactical function.  As Wells observes, this kind of replacement pattern could also indicate that the lexical meanings of the two signs are in the same semantic class (1998: REF).  For example, words (or signs) for two different types of grain, such as barley and wheat, could be expected in the same or similar contexts in an inscription.
Wells gives the frequency of the BI-FORK TOPPED POT as 23 occurrences.  Adding these together with the 30 he lists for the TRI-FORK TOPPED POT yields a total number of “banyan” occurrences of 53.  This is still a good ways short of Korvink’s 118 occurrences.  Once I obtained the first two volumes of the Corpus, I made my own count of each of the symbols.  I first devised an identifying name for each sign and then wrote out in transcribed form all of the inscriptions that I could read (some are illegible).  I entered this information into a computer database in Microsoft Excel, which allows me to search out and identify the appearances of the signs.  My list of inscriptions containing the BI-FORK TOPPED POT and TRI-FORK TOPPED POT follow:
TRI-FORK TOPPED POT / POT: Banawali B-31 {broken pot shard, tops only, R-L}; Harappa H-13, 30, 101, 161, 287, 302 {*reverse order!}, 469, 593, 599, 852, 880, 882, 903, 973; Kalibangan K-24, 43 {over pot}; Lothal L-5, 64 {? broken}, 95 {? broken}, 112 {? odd, tri- + post topped pot?}, 211; Mohenjo daro M-47, 149, 239, 245, 275, 284, 314 {medial? in long inscription}, 456, 733, 735, 843 {?broken; final sign prob. Bearer}, 862, 1010 {?broken}, 1070 {? or tri- + post topped pot}, 1071, 1078 {?br}, 1079, 1100, 1155, 1163, 1177, 1400B {?unclear}; Nindowari damb Nd-1 = 45 subtotal with 8 uncertain, followed by POT
OTHERS: TRI-FORK TOPPED POT (alone); M-1103 {medial, no pot}, M-1579 (alone on bangle); K40 {FINAL, no pot}; Kot Diji Kd-10 {? broken pot shard, unclear fragment beside clear}) = subtotal 3 +/- 1
BANYAN “A” total 48 +/- 9
BI-FORK TOPPED POT) / POT: Harappa H-199, 362, 376 (?), 879, 881, 912, 975, 994; Lothal L-14, 278; Lohunjo daro Lh-1; Mohenjo daro M-44, 55(?), 112, 162, 165, 198, 221, 274, 623 (?), 646, 754, 758, 818, 859, 957, 986, 1019, 1602 (?); Rakhi garhi Rgr-1; Rahman dheri Rhd-238 (?) = 31 +/- 5 subtotal BANYAN “Bb”; BI-FORK TOPPED POT / SPEAR: Mohenjo daro M-0026 = 1 subtotal BANYAN “Ba”
BANYAN “B” total 32 +/- 5
[COMPOSITES, NOT BANYAN, and LIGATURES: bi-rake & trident prong yu / pot (H-0182, H-0203, H-0365, M-0543, M-0654, M-0688, M-1181, C-0022) = 8 of type A [NOT BANYAN]; bi-rake & flange prong yu / 4 toe foot [medial ?] / pot / man (M-0899) = 1 of type C [NOT BANYAN]; bi-rake & wy prong yu (or bi-fork prong yu) / pot (C-0021) = 1 of type B [NOT BANYAN]; bi-rake & (?) prong yu / pot (?) (H-0148, H-0999); trident prong yu with attached post / pot (M-0622) [LIGATURED BANYAN “A”]  = total ODD POTS (BI-RAKE ATTACHED TO x); H-300 (Loop-topped DUBYA) [are Dubyas also Banyans?]; M-234 is not DUBYA (or CUP) but CORN HOLDER; M-594 prob. POTTED ONE(broken) / zee / crossroads ex / pot]
This brings a possible total of “banyan” signs (A and B above) to a maximum of 70, using my figures.  There remains a substantial gap between this frequency and that cited by Korvink.  Another sign, the FLANGE TOPPED POT, may be a third variant of Korvink’s “banyan” (Koskenniemi and Parpola 332{b}, Wells 306, Fairservis J-7, my VI54).  Like BANYANS “A” and “B,” this symbol also occurs most frequently preceding POT, a fact that identifies it as another variable in a compound terminal. 
flange prong yu (FLANGE TOPPED POT) / POT: Harappa H-47, 58, 477 {striped, in prefix without pot}, 743, 744, 745, 746; Mohenjo daro M-30, 394 {striped}, 397, 551, 552, 553, 554, 555, 556, 557, 558, 559, 560, 561, 562, 563, 564, 565 {15 duplicates}, 678 {?}, 707 {?}, 1275, 1327 {?broken, no pot}, 1503, 1504, 1505, 1506, 1507, 1508, 1509, 1510, 1511, 1512, 1513 {11 duplicates}, 1650 {rod, reading right to left} = BANYAN “Ca” subtotal 40 +/- 3, “Cb” subtotal 1]
BANYAN “C” total 41 +/- 3
In one of the inscriptions including the FLANGE TOPPED POT, the possible “banyan” appears in the absence of a following POT.  The inscription reads, from left to right: SPACESHIP / STRIPED FLANGE TOPPED POT / STRIPED MALLET / SINGLE QUOTE / MAN WITH DOUBLE ANKLETS.  In this case, H-477, the sign in question is in the prefix, since it appears among the symbols preceding SINGLE QUOTE.  Presumably, the absence of the POT is what “allows” the supposed “banyan” here to leave the terminal portion of the inscription.  The other STRIPED FLANGE TOPPED POT appears immediately before the POT, as expected for a variant of the “banyan” (M-394) 
I count a maximum total of 48 TRI-FORK TOPPED POTS (BANYAN “A”), 32 BI-FORK TOPPED POTS (BANYAN “B”), and 41 FLANGE TOPPED POTS (BANYAN “C”) of which 17 are questionable.  Adding these together yields a minimum total of 104, deleting all uncertain readings, and a maximum total of 121 “banyans,” assuming all uncertain readings are accepted.  Two further questionable possibilities include a sign I call EL TOPPED POT (Rhd-169 and Rhd-170 from Rahman-Dheri).  These are rather different graphically from all previous examples in that they are angular, with square bottoms.  There is a single horizontal prong extending outward on each side, an element I am terming an EL.  Each of these two examples appears on a pot shard, where it might be intended as a slightly abbreviated POT (a sign which normally has two such prongs on each side).  If this is the case, they cannot be variants of the “banyan.”  A third possibility is that these only exist as pot marks and are intended as something different from any other sign.  Including them as additional variants brings the maximum total to 122, while leaving them out of consideration keeps the minimum total of “banyans” at 104.  The total number cited by Korvink, 118 occurrences, falls within this range of possibilities (i.e., 118 ≈ 104-122).  We can safely assume that at least a few of questionable items are misidentified in my maximum total, accounting for the difference.
One has to wonder, though, whether there is any way to test my conclusions.  Perhaps the form of certain ligatures (composite symbols made up of parts of two or more other symbols) involving these proposed “banyans” will tend to confirm the hypothesis that there are three or four variants of this one sign.  There are two occurrences of FAT CHEVRON IN TRI-FORK TOPPED POT, K-18 from Kalibangan and H-484 from Harappa, the latter broken and somewhat uncertain.  Originally, I thought the broken occurrence at Harappa might indicate a CARTWHEEL inside, since only the peaked, bisected top is visible.  But there is no other CARTWHEEL in such a position, while there is another FAT CHEVRON.  Hence, the latter now seems more likely.  Still, there are many signs that occur only once, in this script, so each of these composite signs might be a singleton.  As it happens, there is no variant *FAT CHEVRON IN BI-FORK TOPPED POT, nor is there a *FAT CHEVRON IN FLANGE TOPPED POT.  Thus, this ligature cannot help determine the likelihood that the three possible BANYANS really are variants of one sign.

Seal H-14 (detail, with bovine horn at left) with inscription: DOUBLE STACKED TWELVE / RAKE / PINCH // WHISKERED FISH / BI-RAKE AND BI-FORK TOPPED POT // POT. 

Another ligature has a BI-RAKE on the left, a vertical stem beneath two comb-like elements stacked one over the other (the whole thing resembling a telephone pole with two crosspieces).  This element is attached to another by a curving line to the right.  The part on the right is the focus of our investigation at this point.  In my terms, there are three variations on this theme: (1) BI-RAKE & TRI-FORK TOPPED POT, (2) BI-RAKE & BI-FORK TOPPED POT, and (3) BI-RAKE & FLANGE TOPPED POT.  All 16 instances of such ligatures occur either just before POT or next to a break where there could have originally been such a sign at the end:

BI-RAKE & TRI-FORK TOPPED POT: Mohenjo daro M-5, 118, 162, 706, 750*, 780, 804(?), 903, 924, 958, 979, 1146; Harappa H-14 = total 13 +/- 2 LIGATURE “A” < BI-RAKE + BANYAN “A”
BI-RAKE & BI-FORK TOPPED POT: Mohenjo daro M-705 = total 1 LIGATURE “B” < BI-RAKE + BANYAN “B”
BI-RAKE & STRIPED FLANGE TOPPED POT: Mohenjo daro M-360; Harappa H-478(?) = total 2 +/- 1 LIGATURE “C” < BI-RAKE + BANYAN “C”
BI-RAKE & BANYAN LIGATURE total 16 +/- 3
(* M-750 actually has a QUAD-FORK top on the right; cf. H-1010 BI-FORK & QUAD-FORK TOPPED POT which may even be a QUINT-FORK, as the reading is somewhat uncertain.)
The fact that there appear to be three variants of the BI-RAKE AND BANYAN LIGATURE, along with the fact that each of the BANYAN variants appears in this ligature, tend to support the identification of these three types of BANYAN.  The hypothesis is not proved, as the total number of ligatures is extremely small.  It is only tentatively supported.  Future discoveries may provide additional information on this topic.

Moving now to the subject of parallel symbols in other cultural spheres, I find few examples.  For the TRI-FORK TOPPED POT (BANYAN "A"), which I enumerate VI51, proto-Elamite provides the only such parallel.  The Iranian symbol is "V" shaped, as is occasionally the case with Indus signs that are more frequently "U" shaped.  At the ends of the prongs, the proto-Elamite sign has a similar trident-like element (M061, see reference below for the website).
Proto-Elamite sign M061.

 The "Y" shaped prongs of the BI-FORK TOPPED POT (my VI53) sometimes resemble the hands of the anthropomorphic figures found on a number of seals.  This calls to mind the Egyptian ka, one aspect of the human soul, which was written with a hieroglyph formed of two arms reaching upward (D28 in Gardiner's list).  The base of the glyph is squared off, unlike the Indus sign, but the simplified hands are quite similar.
Proto-Elamite sign M285, a parallel to Indus sign BI-FORK TOPPED POT.

The third apparent variant (FLANGE TOPPED POT, VI54) lacks a proto-Elamite parallel.  It does resemble a proto-cuneiform sign, however (ZATU831).  The meaning of this symbol is unknown, which is the case with most proto-Elamite signs as well.  A second parallel occurs in Luwian hieroglyphs.  The symbol here represents a syllable, sa, and appears upside-down compared to the Indus sign (Cambel 1999: 90).

Proto-cuneiform sign ZATU831, parallel to Indus sign FLANGE TOPPED POT.


Cambel, Halet. 1999. Corpus of Hieroglyphic Inscriptions, Vol. II, Karatepe-Aslantas. New York: Walter de Gruyter.

Damerow, Peter and Robert Englund. 1989. The Proto-Elamite Texts from Tepe Yahya. Cambridge: Harvard University Press.

Fairservis, Walter. 1992. The Harappan Civilization and its Writing: A Model for the Decipherment of the Indus Script. Leiden: E.J. Brill.

Joshi, Jagat Pati and Asko Parpola, eds. 1987. Corpus of Indus Seals and Inscriptions 1. Collections in India. Helsinki: Suomalainen Tiedeakatemia.

Korvink, Michael. 2007. The Indus Script: A Positional-Statistical Approach.  Gilund: Amazon.com.

Koskenniemi, Kimmo and Asko Parpola. 1982. A Concordance to the Texts in the Indus Script. Helsinki: Department of Asian and African Studies, University of Helsinki.

Shah, Sayid Ghulam Mustafa and Asko Parpola, eds. 1991. Corpus of Indus Seals and Inscriptions 2. Collections in Pakistan. Helsinki: Suomalainen Tiedeakatemia.

Proto-cuneiform sign list online:

Proto-Elamite sign list online:


  1. Fascinating orthographic analyses, Diwiyana.

    The orthographic style of ligaturing may explain the signs and signs variants you mention.

    For e.g., Mahadevan Sign 347 can be read as glyphic elements: POT, two tri-forks (or sprouts) on either end. Thus, POT, SPROUT, TWO can be read rebus as glyphics and rebus substantives. This has been attempted in my Indus Script Cipher published through Amazon. http://tinyurl.com/4gaev9c Best wishes, I am sending by mail the sign variants and other signs similar to Sign 347 of Mahadevan concordance. Kalyanaraman, Ph.D.

  2. Thank you for your comment and for your subsequent e-mail. I must point out that use of the rebus principle is not universal at the proto-writing stage and that, at best, is the stage at which the Indus script remained when it disappeared -- according to the well-argued thesis of Farmer, Sproat, and Witzel.

  3. Hi


    the Indus script had certainly reached the syllabic stage and longer texts certainly agreed .read my paper Sujay