A vowel followed by Tdenotes its own length. (4)This means: “aT” denotes a, “a ̄T” denotes a ̄, “uT” denotes u, etc.
Ka ̄tya ̄yana, another early grammarian, objected to (4), that “long” and
“short” are merely habits of speech: some people speak fast and others slow.
Patañjali retorted that this objection pertains to dhvani, “speech sound,” not to
the subject matter of linguistics which is sphot.a, the meaningful unit of expres-
sion. Patañjali is right because, in Sanskrit, as in many other languages, the dif-
ference between short and long may affect meaning. Patañjali’s concept of
sphot.ais concerned with competenceof the language; speakers’ habits such as
rapid or slow speech belong to the psychological domain ofperformance. This is
one of several cases where rules of Pa ̄n.ini’s grammar have logical, psychologi-
cal, or philosophical implications.
We are ready to return to the metalinguistic “indicatory markers” indicated
by the capitals “N.,” “K,” “N
.
,” etc. in (3). Their use is explained by metarule (5):An initial sound joined to a final (indicatory) sound (denotes the
intervening sounds as well). (5)Thus, aN. denotes “a i u,” aK denotes “a i u r..l,” iK denotes “i u r..l,” yaN.
denotes “ya va ra la,” and aC denotes vowels. The expression aN.may denote “a
i u” or the class of all vowels and semivowels because the indicatory marker “N.”
occurs twice as the reader will have noted. I shall not discuss whether this is a
flaw or a particular merit of the system but Patañjali and other commentators
pay plenty of attention to such problems. They do not assume that Pa ̄n.ini was
perfect and reject what he said if they find that it contradicts the data or does
not work.
Pa ̄n.ini needs the notations aN., aK, yaN.etc. as elements of his artificial lan-
guage, because phonology requires many combinations of sounds that cannot
be simply or naturally expressed in the varga system of (1) or in any form of (3)
without metalinguistic markers. The notations of (5) enable him to express how
two vowels combine into a single long one as follows:
aK is lengthened when followed by a homorganic vowel. (6)Another rule is needed to express the fact that the vowels i,u,r.andl.are
replaced by the semivowels y,v,randl, respectively, when followed by a het-
erorganic (i.e., nonhomorganic) vowel: e.g., dadhi+atra >dadhyatra(“milk
here!”),madhu+atra >madhvatra(“honey here!”), etc. (compare the contrast in
English between pennilessandpenny arcade). Using (5), Pa ̄n.ini’s first step may
have been to formulate this fact as: “iK is replaced by yaN.when a heterorganic
aC follows.” But “heterorganic” is omitted by default because (5) already took
care of the homorganic cases.
Why does dadhi+atrabecomesdadhyatraand not *dadhvatra? I have expressed
it by using the English term “respectively.” Pa ̄n.ini uses a metarule:
354 frits staal