Wednesday, July 4, 2012

Testing a Hypothesis on Indus Inscriptions

Seal M-1107 with inscription: FAT EX IN DIAMOND / BI-QUOTES //
(the constant in this prefix is the most common of the rare ones, as
it occurs before BI-QUOTES nine times).

In an earlier post I noted the possibility that the prefix (P) found in many inscriptions contains one type of information, while the medial segment (M) contains another type, and the terminal (T) could contain still a third type.  If the Indus symbols represent identity markers as in Near Eastern cylinder seals, one might hypothesize that the three distinct segments contain something akin to names of owners, occupational terms or titles, and relational information such as “son of X” or “servant of deity Y” (not necessarily in that order).  Not only is such information borne by early cylinder seals from Mesopotamia, but very similar information is found on many ancient Egyptian scarabs (sometimes used as stamp seals).

Focusing on the Indus inscriptions, we have seen that the T segment is generally short; usually it consists of a single sign, but sometimes includes two or three symbols.  The content of this segment is also highly constrained compared to the other segments, with fewer than a dozen signs demonstrably falling into this category.  These include POT, FORK-TOPPED POT (2 variants), FLANGE-TOPPED POT, COMB, SPEAR, MAN, BEARER, CHEVRON-HATTED BEARER, POT-HATTED BEARER, and PINWHEEL.  Because of the limited size of the terminal set, it is unlikely that these symbols represent individual owners. 

In proto-Elamite, as noted in previous posts, the many singletons and other rare symbols most likely represent owners, presumably because each owner needs a distinct symbol (or two) that will not be confused with that of another owner (Damerow and Englund 1989: 11-13).  In societies that used tamgas, especially Central Asian societies using ownership marks, scholars have generally found more than 10 distinct emblems (Pim et al 2010).  For example, there are more than 20 examples used in Arabia for marking camels (2010: 84), more than 70 Mongolian tamgas (2010: 98), over 60 nishan signs in one Mazdean sanctuary in southern Kazakhstan (2010: 135), and 60-plus linear symbols found on pyramidal and other seals in western Anatolia dating to the Classical period (2010: 154).  So it is probably safe to conclude that the T segment of Indus inscriptions does not indicate an owner.
Seal M-272 with inscription: CIRCLED CROSS / BI-QUOTES
(an unusual inscription with a singleton sign as the variable in
the prefix; also unusual in that there is nothing but a prefix).

It also seems unlikely that these signs represent social groups based on (real or fictive) kinship such as families, clans, or sodalities, for the same reason -- tamgas are marks of family/clan ownership at least as often as individuals’ marks.  It is conceivable that terminals represent deities, perhaps patrons of the seal owners.  It is also possible that these signs indicate social roles, the equivalent of the titles on cylinder seals and scarabs.  If the latter is the case, there were not many different roles in Indus society or else owners of inscribed objects came from only a few social statuses. 

The other two segments – P and M – may contain a single sign just as the terminal may.  But P and M also occur frequently with two, three, or even more symbols.  Does this indicate that P and M contain the same or similar types of information?  I am thinking specifically of the common Near Eastern pattern of early seal inscription, with a personal name followed by a relation to another personal name (in the following, a personal name is abbreviated PN, OCC indicates an occupational term or title, REL represents a social relationship, and DN indicates a divine name):

·         PN:  Ishma-ilum (Semitic ruler of Kisik during Period II 3000-2334 BCE; Collon 2005: 31, no. 84)

·         PN +OCC + DN:  HE-kung-sig, priestess of Pabilsag (Ur, Period II; Collon 2005: 31, no. 93)

·         PN + REL + PN:  Ninmelila, wife of Ur-Dada (Period III 2334-2000 BCE; Collon 2005: 35, no. 110)

·         PN + OCC + LOC + PN + OCC + REL:  Gudea, governor of Lagash: Abba the scribe (is) your servant (Period III; Collon 2005: 36, no. 114)

·         PN + REL + PN + OCC:  E-gissu, son of Bibbi, the carpenter (Period III; Collon 2005: 36, no. 116)

·         PN + REL + DN:  Danni, servant of the god Nergal (Period IV 2000-1500; Collon 2005: 47, no. 158)

One way to examine the possibility of similar information types appearing in both P and M is to compare the specific Indus signs that occur in these two segments.  If signs found in P generally do not appear in M (and vice versa), then the two segments most likely contain two different types of information.  If both segments generally contain the same set of symbols, then this is evidence that similar information appears in both segments.  If this proves to be the case, the constant sign that ends the prefix (SINGLE QUOTE, BI-QUOTES, PINCH, or, rarely, a combination of two of these) may contain no actual information itself, merely serving to separate P from M.

I.  Prefixes containing a single variable sign plus the constant:
1.  29 symbols appear only rarely as the single variable before BI-QUOTES in a prefix (between 1 and 9 occurrences each).
2.  Three single-sign variables are very common before BI-QUOTES
a.  CARTWHEEL (73 inscriptions)
b.  CIRCLED VEE (93 inscriptions)
c.  VEE IN DIAMOND (99 inscriptions)
d.  A few of these inscriptions include a second constant sign
i.  BI-QUOTES + SINGLE QUOTE (8 inscriptions)
ii.  BI-QUOTES + BI-QUOTES (2 inscriptions)

1.  Common: FOOTED STOOL (20), FAT EX (39), CARTWHEEL (27), BOAT (54).
2.  Rare: 10 signs, each appearing in 1 to 4 inscriptions.
3.  A few inscriptions combine PINCH with another constant in the variable: PINCH + SINGLE QUOTE (2 inscriptions); PINCH + BI-QUOTES (2 inscriptions).

1.  Rare: 11 signs, each appearing in 1 to 3 inscriptions.
2.  Common: CARTWHEEL (14 inscriptions).
3.  A few inscriptions include two constants in the variable: SINGLE QUOTE + BI-QUOTES (1 inscription); BI-QUOTES + SINGLE QUOTE (included in section on BI-QUOTES above -- 8 inscriptions); SINGLE QUOTE + SINGLE QUOTE (2 inscriptions).

II.  Two or more variables before the constant in the prefix
1.       Two variables appear in 133 inscriptions.
2.       Three variables appear in 74 inscriptions.
3.       Four variables appear in 17 inscriptions.
4.       Five variables appear only in 2 somewhat doubtful inscriptions; in both cases, one symbol occurs twice, such doubling being counted as a single sign by Wells (which would make these additional instances of four variables).
5.       Six variables appear in 2 inscriptions.
6.       Seven variables appear in 1 inscription (KP2109).
7.       Embedding apparently occurs in 8 inscriptions.

B.      PINCH
1.       Two variables appear in 19 inscriptions.
2.       Three variables appear in 6 inscriptions.
3.       Embedding apparently occurs in 9 inscriptions.

1.       Two variables appear in 34 inscriptions.
2.       Three variables appear in 24 inscriptions.
3.       Four variables appear in 10 inscriptions.
4.       Five variables appear in 4 inscriptions (H-642, M-634, M-1057, M-1103).
5.       Embedding apparently occurs in 26 inscriptions.

Looking only at the prefix variable consisting of a single sign, several characteristics are clear.  Of the signs that occur in P in the variable slot, few are singletons.  By definition, singletons do not occur anywhere else, so they appear just in P.  In contrast, the vast majority of signs that appear as a variable in P also occur in other inscriptions in M.  Some, in fact, occur as the whole medial segment; i.e., the middle section includes only this one sign.  Thus, by and large, the signs found in P do not differ remarkably from those found in M.  This would seem to suggest that P and M do contain similar types of information.  So far, so good for my hypothesis.

However, this is only true when the variable portion of P contains only one sign.  The picture differs in the longer prefixes, those containing two or more variable signs.  Many sign combinations that occur frequently in M do not appear in P.  For example, CUPPED SPOON (or CUPPED POST) + 3 POSTS is a frequent pair, but only in M, never in P.  One of the constituent signs – 3 POSTS – does occur in the occasional P, but the other constituent, CUPPED SPOON, does not.  Similarly, the pair 2 POSTS + FISH occurs frequently in M, but not in P.  The first part, 2 POSTS, occurs occasionally in a prefix, but the second part, FISH, does not.  The main overlap in content between P and M where the variable portion contains two or more signs is the set of doubled signs (e.g., DOUBLE GRIDS, DOUBLE CUPS).  Such doubled symbols can occur in either P or M in an inscription – but not both in one and the same inscription.

This finding suggests that the informational content of P is not of the same type as that found in M, after all, despite some superficial similarities.  This, in turn, suggests that Indus inscriptions do not contain information similar to the Near Eastern type PN + REL + PN (e.g., Ninmelila, wife of Ur-Dada; or E-gissa, son of Bibbi).  Finally, if this conclusion is correct, then the prefix constants (SINGLE QUOTE, BI-QUOTES, PINCH) probably do not function simply as marks separating P from M. 


Collon, D. 1987 & 2005. First Impressions: Cylinder Seals in the Ancient Near East. London: The British Museum Press.

Damerow, P. and R.K. Englund. 1989. The Proto-Elamite Texts from Tepe Yahya. Cambridge, Mass.: Peabody Museum of Archaeology and Ethnology, Harvard University.
Pim, J.E., S.A. Yatsenko, and O.T. Perrin. 2010. Traditional marking Systems: A Preliminary Survey. London: Dunkling Books.

Tuesday, July 3, 2012

Seals Vs. Tablets: The Case of "Hairy Hunchback"


What information is likely to be included in Indus inscriptions?  No one knows for sure but many researchers have provided hypotheses.  Some of these focus on the presumed economic function of inscribed objects.  For example, M. Korvink suggests, “The script at the top of a seal could conceivably express the commodities, their quantity, locations, and the proprietor, etc. while the motif was ‘guarantor of the transaction’” (2008: 76-77).

B. Wells also considers an economic function likely: “Seals can be demonstrated to function as closures for doors, boxes and other packages enclosed in reeds and textiles” (2011: 162).  As an example of more specific meaning, he cites a particular jar (designated Jar G) found at Mohenjo daro, Moneer southeast area.  This jar contained many fragments of stoneware bangles, which Wells estimates derived from 17 or 18 original bangles.  The jar had a sealing closing it and on this sealing were the two signs SINGLE POST / STACKED SEVEN, which Wells takes to be a way of writing “seventeen,” the (approximate) number of bangles inside.  From this, he concludes that one function of seals was “to facilitate the cycle of production” (2011: 161).  If he is correct in his reconstruction of the bangles in this jar and if he is also correct about the meaning of SINGLE POST / STACKED SEVEN, then the inscription does seem to refer to the quantity if not the identity of the items inside the jar.

Because of the parallel use of Near Eastern cylinder seals to close various containers and doors, M. Coe posits an economic function as well (1995: 394).  That is, he says, the inscriptions on seals “in all likelihood name the owner or owners of the goods sealed.”  So, assuming an economic function is correct, we find more than one possibility for the content of inscriptions.  These could identify owners or they could identify the objects of commerce, the commodities.

In proto-Elamite, as I have noted before, the rare signs typically indicate owners, while common signs that represent commodities are paired with numerals (Damerow and Englund 1989: 11-13).  As discussed at length in previous posts, I do not interpret the apparent numerals in Indus script as modifyers that actually enumerate.  As a result, I consider it unlikely that Indus inscriptions on seals parallel proto-Elamite economic tablets in their content.  Instead, I hypothesize that the Indus inscriptions on seals are largely owners’ marks, perhaps similar to Turkic tamgas (e.g., see Pim, Yatsenko, and Perrin 2010).

Even if this hypothesis is correct, it does not necessarily mean that all the Indus inscriptions have the same type of content.  In particular, the group of objects termed tablets may bear something different from what appears on seals.  The former group comprises three different types of objects, actually.  The bas-relief tablets may have been created from impressions of seal inscriptions, in which case the inscriptional content of both objects should be essentially of the same type.  But there are also incised tablets, some on baked clay, some on metal.  Many of the former have an inscription on both sides, with the second side usually containing CUP plus an apparent numeral (most often from SINGLE POST to FOUR POSTS, rarely SIX POSTS).  It is conceivable that this type of artifact had a function different from that of the seals (and perhaps different from bas-relief tablets).

So let us examine the inscriptions on tablets with those on seals to see whether there are indeed differences.  One way of doing this is to look at specific signs.  For example, my first impression is that HAIRY HUNCHBACK appears more often on tablets than on seals.  Wells divides variations of this symbol into four separate signs in his latest work (2011: 180).  The “head” is on the right and there are four “hairs” on W176; the “head” is on the left and there are three “hairs” on W177; an angular version of the first is W178; and a variation of the second with an open “head” is W179.  Together, these variants occur 201 times by Wells’ count.  Of these occurrences, 123 are on tablets while 72 are on seals (6 additional occurrences on pot shards or other objects).  That is, using Wells’ figures, 61% of occurrences are on tablets versus 36% on seals (3% other).  Many of the tablets bear inscriptions that are duplicated on other tablets, however; if each inscription is counted just once no matter how many duplicates there are, the imbalance between tablets and seals is not as pronounced.

It is still interesting to note what contexts this sign appears in, on the different objects.  Among the inscriptions recorded in the KP concordance, I find the following on tablets (underlining indicates tablets that are incised, i.e., those we are most interested in here):

·         1 of HAIRY HUNCHBACK alone (actually placed between two circles, which may or may not be intended as signs) on H-220-222A (bas relief)

·         1 of HAIRY HUNCHBACK in final position on M-1418 (bas relief)

·         1 of HAIRY HUNCHBACK + POT-HATTED BEARER on H-220-222 (bas relief)

·         4 of HAIRY HUNCHBACK + WINGED MAN on tablets on H-179 (bas relief), H-740-742 (bas relief), M-543-546 and M-1497-1502 (incised)

·         1 of HAIRY HUNCHBACK + VEST on H-837 (round bas relief tablet)

·         3 of HAIRY HUNCHBACK + STRIPED VEST on H-218 (bas relief), H-341 (triangular bas relief), H-892-893 (incised)

·         13 of HAIRY HUNCHBACK + POT on M-511-512 and M-550 (incised); M-1448-1451, M- 1453, and M-502-503 (incised); M-1452 (incised); H-699 and H-171 (bas relief); M-494-495 (bas relief); M-1460 and M-509-510 (incised); H-315 (incised); H-932, H-934-935, H-937, H-959, H-961-963, H-978-981, H-309, H-311, H-316-318, and H-352-357 (incised) as well as H-233 (bas relief); H-987 (incised); H-933, 936, 960, 964, 308, and 312-314 (incised); H-853 (bas relief); H-232 (“shield” shaped bas relief); H-811 (bas relief)

Other inscriptions in the concordance are either on seals or on objects I cannot identify in my database.

In comparison, I find the following on seals:

·         2 + one of the various BEARER signs (BEARER, CHEVRON-HATTED B., POT-HATTED B.)

·         4 + WINGED MAN



·         1 + RAKE

·         1 + DOUBLE CEES

·         1 + HUNCHBACK (not “hairy”)

·         1 + SPACESHIP

·         4 + STRIPED VEST

·         2 + CUPPED SPOON

·         28 + POT
Aside from the multiple co-occurrences of HAIRY HUNCHBACK and POT, there is no clear pattern and this particular "pair" is common to both tablets and seals.  Hence, analysis of this sign indicates no significant difference between media.  Whatever the inscriptions once meant, they seem to mean the same regardless of whether they appear on tablets or seals.  However, this tentative conclusion only pertains to this particular sign thus far.  Ideally, we would analyze occurrences of the other common signs in the same way before drawing a firm conclusion.


Coe, M.D. 1995. “On not breaking the Indus code” in Antiquity 69 (1995): 393-5.

Damerow, P. and R.K. Englund. 1989. The Proto-Elamite Texts from Tepe Yahya. Cambridge, Mass.: Peabody Museum of Archaeology and Ethnology, Harvard University.

Korvink, M.P. 2008. The Indus Script: A Positional Statistical Approach. Gilund Press.

Pim, J.E., S.A. Yatsenko, and O.T. Perrin. 2010. Traditional Marking Systems: A Preliminary Survey. London: Dunkling Books.

Wells, B.K. 2011. Epigraphic Approaches to Indus Writing. Oxford and Oakville: Oxbow Books.