This page has not been fully proofread.

74
 
not necessarily hold for the present collection. The best example of this
is the verb form śikṣatu in 48%. In school editions like M. R. Kāļe's,
this is emended to the correct śikṣatüm, which I have found in just one
MS, namely the late and worthless Y-type fragment Nagpur 801. The
remaining verb-forms in the stanza are yātu, jayatu, prayātu, so that one
might see here the beginning of some form of assonance within the stanza.
But the reading itself cannot possibly be contested, seeing the remarkable
uniformity of the MSS. The form gacchatäm in 25ª is just the opposite
type of error, though with correct variants-none of which can be accepted
on MS evidonce alone. Nabhyarthito jaladharo'pi is not a good construction
in 64; nor smitam kiñcid vaktram in 93%, though emended to mugdham
in most of S and explained as vaktram kiñcid smitam by most N commenta-
tors. Plenty of other examples could be found to prove our point.
 
INTRODUCTION
 
The second main principle is that strong variation always indicates
some difficulty which the scribes tended to smooth out in different ways.
In 1092, the reading -sastrișu rajyeta hah is certainly not in doubt, but
the word sastri is comparatively rare in classical Sanskrit in spite of its
use by Patanjali [on Pan. 2. 1. 55] and definition as a knife in Amarakośa
2. 8. 92. Common enough in Prakrit as satti, about the only well-known
classical Sanskrit work that gives it is Māgha's Siśupāla vadha 4. 44. The
confusion particularly evident in otherwise correct southern codices is thus
clearly explained.
 
The text in general determines itself by the concordance of the
overwhelming majority of the MSS, but in the really difficult places the
Valentinian law of citations cannot be applied. We have first to note
that the N recension as a whole is conservative, preserving archaisms and
solecisms, whereas S tends to paraphrase in correct Paṇinian Sanskrit whenever
necessary. The N text, in the final analysis, rests mainly upon A-E, as
the other versions are not well-determined, or are contaminated by foreign
readings while the Y collations made by others are unreliable. Whenever
A, or even Ao.1 shows agreement with any Malayalam source, I have
taken the reading as original, for it could hardly have arisen otherwise at
the two extremes of Bhartrhari territory. When, however, N and S differ
en bloc, with nothing to choose between them, I have taken the N reading
provisionally as a stopgap, because Bhartrhari after all was a northerner
according to all traditions, and the MS tradition certainly originated in
the north. Finally, the rather intricate metres make restoration easier in
some ways, for one can then constitute the text letter by letter. The
wavy line is used to call attention to a strong variant which could as
well have been accepted for the main reading without much violence to
the canon. It could have been used much oftener, particularly in 273,
where every version has its own strongly divergent readings so that the
task of restoration seems quite hopeless.
 
4. 4. Applications of the method. The reading -gunāḥ samvāsats
jayate at the end of 33 is certainly ungrammatical, giving a singular verb
with a plural subject. The major alternative in N is to take the singular
-guņaḥ, which violates the rule for compounds. Here the meaning is not