Thursday, September 15, 2011

Indus Script: Beyond Signs to Sequences

Indus seal from Gonur Depe with elephant and inscription, showing contact between Indus Valley
and Bactria-Margiana culture: EGG ON NEST / BI-QUOTES / FISH UNDER CHEVRON / FISH /
MAN HOLDING DEE-SLASH / MALLET / TRI-FORK / POT (Kohl 2007: 199, fig. 5.12a).


I have now discussed all of the Indus signs in my list, a list based on signs found in other lists, namely those of Fairservis, of Wells, and of Koskenniemi and Parpola.  Iravatham Mahadevan has also compiled a list along with a concordance of inscriptions (1977).  I only have an unclear copy of his list, not the work itself, so I have not mentioned Mahadevan’s sign numbers in my discussion.  Both Wells and Parpola have also published revised lists, in which both the total number of signs differs from those I quoted here previously, and the numbers designating the individual signs differ.  The Koskenniemi and Parpola list that I used includes 396 signs (1982), while Parpola’s revised list includes 398 (2009).  Wells originally listed 588 signs, later revising that to 958 (1998 and 2011).  Fairservis, of course, is at the low end with 230 (1992).  My own list, with 694 separate notations, splits several groups of variants into an unnecessarily large number of categories.  Were I to revise it, the number of signs would shrink, though perhaps the total would remain a little higher than Parpola’s total.  In Deciphering the Indus Script, Parpola mentions Mahadevan’s list, which includes 417 signs (2009: 78).

The total number of signs cannot be pinned down precisely, as of yet, because no two researchers count them in exactly the same way.  But if we average the above results, we get a more or less accurate total of 526.  That’s lower than either of Wells’ totals (which are probably too high, as is my own total).  It’s also higher than Fairservis’ total (which is too low).  One might wonder what difference it makes what the precise total is.  But it is significant, even without greater precision because, theoretically, this information can suggest what type of writing the Indus system is. 
Plaque from Agrab in the Diyala Valley with humped bull, an animal often
pictured on Indus seals (Kohl 2001: 225, fig. 9.12).

There are too many signs for it to be an alphabet, to begin with.  The English alphabet, for example, has 26 letters, the Cyrillic alphabet of the Russians contains 33, and the Mkhedruli of Georgia (a Caucasian country formerly part of the Soviet Union) has 40 (Nakanishi 1980: 20 and 24).  These are much smaller figures than even the lowest estimate of Indus symbols.  Even if we double these alphabetic totals, counting capital letters and small letters as separate symbols – yielding 52 letters in our alphabet, 66 in Cyrillic, and 80 in Mkhedruli – the numbers remain below 100. and far below the number of Indus signs by anyone's count.  We can safely conclude, then, that Indus script is not alphabetic.

A syllabary easily may have a larger numbers of symbols than an alphabet.  Thai script contains 50 consonant + vowel symbols, as well as 21 additional diacritical marks to indicate vowels other than the inherent –o, plus four signs to represent tones, yielding a total of 75 symbols (Nakanishi 1980: 76).  The Japanese writing system is more complicated, with 77 katakana signs (or 53, depending on how one counts them) for writing foreign words, 73 hiragana signs (or 50, again depending on how one counts them), for a total of 150 (or 103) (Nakanishi 1980: 94).  Each of these syllabic symbols represents either a vowel (V) or consonant + vowel group (CV), with the exception of n.  More complex syllabaries may also include symbols representing CVV or a consonant plus diphthong (e.g., dwa, as found in Linear B), VC (e.g., an, as in Sumerian cuneiform), or CVC (dur, as in Sumerian cuneiform).  The inclusion of these additional types of syllables can increase the number of signs substantially.
Ceremonial axe head with eagle design from Tepe Yahya Per. IVB
(Kohl 2001: 216, fig. 9.6).

The kana form only part of the Japanese writing system, which also makes use of Chinese characters or kanji.  The latter belong to another type of script, typically termed logographic, in which each word is written with a distinct symbol (or, more precisely, each morpheme is written with a unique symbol).  In such a writing system, the total number of symbols is in the thousands.  The modern Chinese writing system contains tens of thousands of characters, in part due to its long history, but about 5,000 are commonly used for newspaper writing.  The earliest Chinese writing is found on so-called oracle bones, dating to the Shang period (about 1200 B.C.E.).  This system may have contained as many as 5,000 characters total, though precision is just as difficult as with numbering Indus signs (Keightley 1985: 59).  Scholars can read approximately half of that total.  Whatever the precise figure, whatever the period, Chinese writing has many more characters than Indus script has signs.
Indus seal in stepped diamond shape with eagle motif from Harappa (H-166); compare the motif to that
on the axe head from Tepe Yahya.  These bear a resemblance to seals and motifs from Altyn Tepe also.

The Bronze Age scripts contemporary with the Indus civilization are mostly mixed systems, combining a syllabary with logographs.  This complicates the matter of determining a total number of symbols, especially when a given symbol has more than one interpretation.  For example, Middle Egyptian makes use of 700 to 800 glyphs, with more variation than currently exists in Chinese (Gardiner 1976).  Many glyphs have more than one use, sometimes functioning as ideographs (or determinatives), sometimes serving to indicate (consonantal) sounds.  Sumerian cuneiform includes a comparable number of signs, many of which also have multiple interpretations.  Generally speaking, the earlier the script, the higher the total number of signs.  The totals for proto-cuneiform and proto-Elamite signs sometimes rise above 1,000, for example, while the earliest cuneiform contains several hundred symbols.  As cuneiform was adopted by people other than the Sumerians, the number of signs often decreased.  Elamite cuneiform, for example, includes about 600 signs, while the much later Old Persian contains around 50 (36 appear in Budge 1920; closer to 50 are included in the unicode, which contains phonetic elements, word symbols, punctuation marks, and numerical symbols).  Thus, going by symbol count alone, these facts would seem to indicate that the Indus script is logo-syllabic.  In fact, this is the conclusion drawn by some researchers (e.g., Parpola 2009: 85).

However, the frequency statistics suggest that the Indus script may be something else.  Mahadevan provides a table showing the frequency distribution of the 417 signs that he includes in his list (1977: 17; after Parpola 2009: 78).  Using Wells’ latest data, we can see which signs occur most frequently, those which may be most amenable to statistical study (2011: 166-233).  This information reveals the fact that over 50% of all sign occurrences are due to appearances of 33 signs, with all the other signs appearing fewer than 100 times.  As frequency drops, the number of signs increases, with the result that singletons and signs that appear fewer than 10 times form the largest group by far.


Frequency
range
Number
of signs
Total sign
occurrences
Percentage
of total
1000+
1
1,395
10.43
999-500
1
649
4.85
499-100
31
6344
47.44
99-50
34
2381
17.81
49-10
86
1833
13.71
9-2
152
648
4.92
1
112
112
0.84
Total
417
13,372
100

Table 1. Frequencies of Indus signs (from Mahadevan 1977).

The most frequent sign, the only one that occurs more than a thousand times, is POT, which appears most often at the end of an inscription (i.e., as a terminal sign).  The second most common sign is BI-QUOTES, which Korvink characterizes as the constant (or final) element in the prefix of inscriptions (2007: 18).  Looking just at this data, some researchers conclude that these two signs must be grammatical markers, phonetic symbols, or both (e.g., Fairservis considers POT an honorific ending and BI-QUOTES a locative case suffix). 

According to Wells’ figures, two other signs appear more than 500 times, namely CUP and TWO POSTS.  This analysis is problematic, however, because many of the occurrences are on tablets, either those made in molds and so providing multiple copies, or on copper tablets, again with multiple copies.  If one discounts the copies, the frequency of both these signs falls.  Still, CUP appears frequently enough to be among the common signs.
Tablet H-216A and B sides with inscriptions: STRIPED FAT CHEVRON / STRIPED VEST / POT //
(B) 3 POSTS / CUP (many tablets have CUP plus two to four POSTS on the B side).

Additional signs with frequencies over 400 include COMB, THREE POSTS (same problem as with TWO POSTS), and FISH.  Those appearing between 200 and 300 times include WHISKERED FISH, SPEAR, CIRCLED FORK, VEE IN DIAMOND, FISH UNDER CHEVRON, CARTWHEEL, CIRCLED VEE, PINCH, MALLET, HAIRY HUNCHBACK (same problem as with CUP), CUPPED POST, and POTTED ONE.  Those with frequencies between 100 and 199 include SINGLE QUOTE, SINGLE POST, BELTED FISH, POT-HATTED BEARER (combining all types), MAN, RAKE, BOAT, FOUR POSTS (same problem as with TWO POSTS), CRAB, GRID (combining all types again), WINGED MAN, FORK-TOPPED POT (combining the TRI-FORK TOPPED POT and BI-FORK TOPPED POT), FORK (combining BI-FORK, TRI-FORK, QUAD-FORK, and QUINT-FORK), PRAWN, FOOTED STOOL, OVERLAPPING CIRCLES, BLANKET (combining those containing 4 to 10 “ticks” as well as any with additional internal marks), FAT-LEGGED LAMBDA, CROSSROADS EX, and LOOP-ARMED MAN WITH SLASH.  This gives a maximum total of 40 symbols that one might profitably study with statistical methods (although there are fewer such symbols if we follow Wells’ criteria for separating signs).  The other 300 to 900 (depending on the total number of signs) appear too seldom for meaningful statistical measures.

Such frequency data are quite different from the frequencies of letters in alphabetic texts.  We know that in English “e” is the most frequent letter, while “q” and “z” are relatively rare.  The majority of letters have a frequency in between the two extremes; a frequency table for the English alphabet would not show such a large clustering of letters at the bottom.  One might wonder, then, whether the frequency data for Indus signs resembles that for glyphs in Middle Egyptian texts or for signs in Akkadian cuneiform.  Based on my experience with such writing systems, I expect to find multiple appearances of symbols representing phonetic elements, while the logographs (signs or glyphs representing words or determinatives in words) are less frequent.  In Egyptian, for example, the glyphs in the “alphabet” (i.e., those used to represent a single consonant sound) are the most common, so that virtually any text will have many bread loaves (t) and chicks (w), owls (m) and ripples of water (n).  A brief glance at a passage from the Book of the Dead demonstrates this:
Part of a passage from the Book of the Dead, in which one can see 3 repetitions of the
bread loaf(here resembling a thick hyphen) just in the far left column, 3 repetitions
of the chick (once in the left-most column, twice more in the second from the left),
and 3 repetitions of the seated god with the beard (once in each column). 
Other symbols repeat as well, including the djed-column (shish kebab)
which appears doubled twice and singly once.

But is a section of a book – even an ancient one – equivalent to the content of seals and very small tablets?  It is possible that the Indus inscriptions mostly contain names and titles, as Parpola suggests on the analogy of Near Eastern seals (2009: 117).  So a better comparison with Egyptian hieroglyphs might be to the cartouches and serekhs containing the pharaoh’s name, sometimes with an accompanying title.  Since Egyptian hieroglyphs continued to be used a good deal longer than Indus script, I compare only the royal cartouches and serekhs from the 1st through the 11th dynasties, dating from about 2650 BCE to 2000 BCE.  This includes the Old Kingdom, First Intermediate Period, and the first dynasty of the Middle Kingdom, a time during which the Indus civilization passed through its Early to Mature Harappan Transition and Mature Harappan periods.

Frequency
range
Number
of signs
Total sign
occurrences
Percentage
of total
20-35
4
118
29.57%
11-19
5
70
17.54%
6-10
14
92
23.06%
2-5
27
78
19.55%
1
41
41
10.28%
Totals
91
399
100%

Table 2. Frequencies of Egyptian hieroglyphs in cartouches and serekhs
of Dynasties I - XI (data from Budge 1978: 917-922).

We can see that singletons and rare signs (those appearing fewer than 10 times) form the majority, while a smaller set repeats frequently.  The two most common signs in these royal names are theophoric elements, the name of the sun (N5, 35 occurrences) and of part of the soul, the ka (D28, 33 occurrences).  Other common signs are mostly phonetic elements, including the ripple of water spelling n (N35, 22 occurrences), the bread loaf spelling t (X1, 28 occurrences), folded cloth spelling s (S29, 13 occurrences), and chick spelling w (G43).  Other glyph found frequently in royal names are nfr, which means “good, beautiful” (F35, 18 occurrences) and the sun rising over a hill (N28, 14 occurrences). 

A title that often appears in a cartouche refers to the Two Lands, a name for Egypt (N16, 14 occurrences).  This glyph demonstrates one of the problems with counting symbols.  In all of the royal names in my sample where this symbol appears, it is doubled.  This might suggest the doubled element is a single glyph if we did not know other uses.  As it happens, the glyph also occurs singly; it just does not do so in this set of royal names.  If we take the doubled glyph as a single glyph, though, that would halve its occurrences.  That, in turn, would drop it from the frequent group. 
The Egyptian hieroglyphic "alphabet" (uniliterals),
the signs that frequently repeat in inscriptions
(from an old postcard).

There is, then, the question of the reed (M17).  It appears singly 9 times, doubled 7 times.  I have separated these occurrences into two categories, since the single reed spells i and the doubled ones y.  But if we combine these two sets and count each reed, there are 23 occurrences.  Counting this way, the reed moves from the infrequent to the frequent group.  Some Egyptian glyphs also occur in groups of three.  In this group of royal names, the ba bird (G29, representing another part of the soul) appears singly as a singleton, but when tripled it occurs twice.  When there are three in a row, this represents bau, in Budge’s transcription, the plural form of ba.  Although not seen in this sample of names, the ka is also tripled in some cases, e.g., in Menkaure (a Fourth Dynasty pharaoh).  The kau in the middle of this name – and this element in other names – can be spelled out with the glyph of the ka plus the chick, or it can be represented graphically as a tripling of the ka glyph.  Although rare, one occasionally finds a quadrupling as well.  The name of the New Kingdom official Neferneferuaten may be written with four nfr signs, three of them representing the plural form neferu, the fourth representing the repetition of nefer.
Tablet M-495A with inscription (from right): PACMAN / CRAB / HAIRY HUNCHBACK /
POT / BEARER // TRIPLE CUPS / SINGLE POST (note that tripling is unusual).

These kinds of problems of counting also occur with Indus signs.  Some only appear in doubled form, including the QUOTE UNDER CHEVRON.  Others appear both singly and doubled, such as variants of GRID.  Occasionally, a sign is tripled, as on two tablets from Mohenjo daro and two from Harappa, each bearing TRIPLE CUPS (M-494, M-495, H-764, H-765; also in graffiti on Rhd-53-56).  Further repetition is almost exclusively found in apparent numerals (QUOTES, POSTS).  There are just two signs other than “quotes” and “posts” that repeat four times in a row: QUADRUPLE CEES (M-136) and QUADRUPLE TRI-FORKS (M-1123).  Parpola takes these quadrupled symbols as equivalents of FOUR POSTS/QUOTES + sign (2009: 81).  This seems reasonable, although the Egyptian example of Neferneferuaten indicates quadrupling can have a non-enumerative meaning.

In Egyptian names as in other texts, it is the phonetic elements that repeat in non-consecutive positions.  For example, the chick appears twice but non-contiguously in the famous Fourth Dynasty Khufu’s name, spelling out the u after kh and after f.  The reed both begins and ends one of the names of the Fifth Dynasty king, Tet-ka-Ra Assa, with a doubled bolt between them (issi in modern transcription).  The bread loaf appears twice in the name of the New Kingdom pharaoh Tutankhamen, representing both t sounds.  These uniliterals – glyphs that represent a single consonant – are found in initial, medial, and final positions from name to name. 
King Tut's names, at left as on a cartouche-shaped box, and at right as
in embossed gold glyphs on the cover of the book Tutankhamun and
the Golden Age of the Pharaohs (Hawass n.d.).  Note the repetition of
the semi-circular bread loaf by the chick for the two t's in this name.

This type of non-consecutive or random repetition is less common in the Indus inscriptions than simple doubling.  Examples of it include M-396: FOOTED STOOL / PINCH // FISH UNDER CHEVRON / BLANKET WITH 4 TICKS / RAYED CIRCLE / DOUBLE CEES / CIRCLED DOT / BACKSLASH IN FISH / BLANKET WITH 4 TICKS / ANT (6 LEGS); M-627: FISH / TABLE (LONG LEG) / GRID / POT (or POTTED ONE) // STRIPED LOOP UNDER CHEVRON / DOUBLE GRIDS; M326B: CARTWHEEL / BI-QUOTES (?) // EX / STRIPED MALLET // BARBELL ON POST / TWO POSTS & BACKSLASH / BARBELL ON POST; M-634: CARTWHEEL / LOOP ARMED MAN HOLDING SLASH / CARTWHEEL / CIRCLED BELTED RECTANGLE / CARTWHEEL / SINGLE QUOTE // RAKE / FISH // POT // EX UNDER TABLE; and H-369: POTTED ONE / TRI-FORK / POTTED ONE / SINGLE POST / CUPPED SPOON / THREE POSTS / SPEAR // (B side) THREE POSTS / CUP (counting A and B sides as a single inscription would make THREE POSTS repetitious here).
Seal M-627 with inscription showing two kinds of sign repetition:
FISH / PINCH // GRID / POT // STRIPED LOOP UNDER CHEVRON/
DOUBLE GRIDS (the POT may be a POTTED ONE).

The relative paucity of the random type of repetition suggests to some researchers that sound is not encoded in the Indus inscriptions (Farmer, Sproat, and Witzel 2004).  If there really is no phonetic information in the script, then by the definition of many linguists, it is not a fully developed writing system.  We will explore this possibility in greater detail in the next post.

REFFERENCES

Budge, E.A.W. 1978. An Egyptian Hieroglyphic Dictionary with an Index of English Words, King List and Geographical List with Indexes, List of Hieroglyphic Characters, Coptic and Semitic Alphabets. Vol. II. New York: Dover (orig. 1920 by John Murray, London).

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

Farmer, S., R. Sproat, and M. Witzel. 2004. "The Collapse of the Indus-Script Thesis: The Myth of a Literate Harappan Civilization," in Electronic Journal of Vedic Studies 11 (2): 19-57.

Faulkner, R.O. 1994. The Egyptian Book of the Dead: The Book of Going Forth by Day. San Francisco: Chronicle Books.

Hawass, Z. n.d. Tutankhamun and the Golden Age of the Pharaohs. Washington, D.C.: National Geographic.

Joshi, J.P. and A. Parpola. 1987. Corpus of Indus Seals and Inscriptions. 1. Collections in India. Helsinki: Suomalainen Tiedeakatemia.

Korvink, M.P. 2007. The Indus Script: A Positional Statistical Approach. Gilund Press. (Amazon.com)

Kohl, P.L. 2001. "Reflections on the Production of Chlorite at Tepe Yahya: 25 Years Later," in Excavations at Tepe Yahya, Iran 1967-1975. The Third Millennium, C.C. Lamberg-Karlovsky, ed. Cambridge, Massachusetts: Peabody Museum of Archaeology and Ethnology, Harvard University. Pp. 209-230.

Kohl, P.L. 2007. The Making of Bronze Age Eurasia. Cambridge: University Press.

Mahadevan, I. 1977. The Indus script: Texts, concordance and tables. Memoirs of the Archaeological Survey of India, no. 48, Delhi.

Nakanishi, A. 1990 and 1980. Writing Systems of the World. Boston: Tuttle Publishing.

Parpola, A. 2009 and 1994. Deciphering the Indus Script. Cambridge: University Press.

Shah, S.G.M. and A. Parpola. 1991. Corpus of Indus Seals and Inscriptions. 2. Collections in Pakistan. Helsinki: Suomalainen Tiedeakatemia.

Wells, B. 1998. An Introduction to Indus Writing: A Thesis.  and

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

2 comments:

  1. This comment has been removed by a blog administrator.

    ReplyDelete
  2. This comment has been removed by a blog administrator.

    ReplyDelete