Iranian Avesta). The Padapa ̄t.ha “word-for-word” analysis separated words and
other elements, insuring their proper pronunciation, including accents (which
I omit from our discussion) and orally fixed the corpus or “text” thus contribut-
ing to canon formationduring the late Vedic period in eastern India (Videha
around 600–500 bc).
There are several stages in the development of this analysis which depict a
gradual extension or generalization, i.e., scientific progress. At first the separa-
tion between words or stems and suffixes was marked by a brief pause in the
recitation as in the above example from Rigveda10.127.2 or in r.s.i-bhih., “by the
seers.” This method was extended to nominal compounds: saptaputram >sapta-
putram “seven sons.” It was called avagraha, “separation” and we express it by a
hyphen though no writing was known in India at this period. Since “separation”
was used in each analysis only once, a problem arose: many compounds consist
of more than two members.
S ́a ̄kalya’s solution was inspired by semantics: das ́apramatı ̄mis analyzed as
das ́a-pramatı ̄m“ten protectors” (i.e., fingers), not das ́apra-matı ̄mwhich would
correspond to the equally meaningless “tenpro-tectors.” More complex cases
were taken care of with the help of two other methods of analysis: pragraha,
“marking,” e.g., saptaputram iti; and parigraha, “marking-and-separation,” e.g.,
saptaputram iti sapta-putram.
The Padapa ̄t.ha is an analysis of theSam.hita ̄pa ̄t.haor “continuous recitation,”
but by isolating words from each other, it facilitated the opposite of what was
intended: the forgettingof single words. Special patterns of repetition or vikr.tis
were constructed to minimize this risk. The question arises whether these pro-
cedures were prescriptiveordescriptive– a relevant question for the historian of
science since science is presumably descriptive. How can we find out?
If we represent the words of the Sam.hita ̄pa ̄t.ha as: 1 2 3 4... and the
Padapa ̄t.ha as: 1 / 2 / 3 / 4 /..., the next two vikr.tivariations may be repre-
sented as follows:
Kramapa ̄t.ha“Step-by-step recitation”: 1 2 / 2 3 / 3 4 /...
Jat.a ̄pa ̄t.ha“Plaited recitation”: 1 2 / 2 1 / 1 2 / 2 3 / 3 2 / 2 3 /...TheKramapa ̄t.hadoes not introduce any new sandhicombinations: all its expres-
sions belong to the Vedic Sam.hita ̄ and are a ̄rs.a, i.e., they were used by the “seers”
(r.s.i); but the Jat.a ̄pa ̄t.ha(called after plaited or matted hair) introduces “2 1” and
“3 2” which are reversals of the original order: they introduce something that
was not in the Veda and therefore “not from the seers” (ana ̄rs.a).
TheKramapa ̄t.haof the first words of our first example of Sanskrit sandhiis
straightforward:
a ̄+uru / uru +apra ̄h./ apra ̄h.+amartya ̄>oru / urvapra ̄h./ apra ̄ amartya ̄/...TheJat.a ̄pa ̄t.hais plaited. It begins: a ̄+uru +uru +a ̄+a ̄+uruwhich becomes:
orvurvoru – a real tangle. Worse, it contains an element that does not come from
the Sam.hita ̄ of the Vedic seers: uru +a ̄>urva ̄. However, urva ̄is not an artificial
350 frits staal