This page has not been fully proofread.

72
 
INTRODUCTION
 
reasonably correct source. In the third group, stanzas found in a single
determined version are generally clear, while the vast majority of these strays
occur in "deplorably incorrect" or, as in the case of Wai 2, quite hopelessly
corrupt MSS. Therefore, I have had to print them in a heavily emended
form whenever parallel citations from anthologies or other sources gave
correct readings. The original corrupt readings have generally not been given
for lack of space. However, where the correct form was in doubt, the original
stanza, no matter how meaningless it seemed, has been reproduced as it stood
with a few suggested emendations or a question mark [which could have
occurred far oftener], in brackets. A considerable number of these stanzas
look as if some optimist had written them himself after a six-week
correspondence course in Sanskrit, the most atrocious being perhaps our 700.
Our major problem, therefore, lies in determining the text of group I stanzas,
and of those in II which occur in almost all sources. Theoretically, the
question is simple enough. Every stanza was composed by someone, the
editor's task being to restore its original text. In practice, one has to deal
with hundreds of divergent sources, all of which are over a thousand years
later than the "original". So simple a text as our Yi with its commentary,
less than two centuries old, preserved by a single family in a peripheral
region, cannot be determined with absolute certainty from its three
representatives. The text of the much more popular version E is correspond-
ingly more diffuse, at times to the point of indeterminacy. It would need a
remarkably solf-confident editor to claim the restoration of the original text
from such material.
 
1
 
4.2. Nature of the commonest variants. A groat part of the actual
variation in readings arises from slips of pen or of the tongue. The one real
ambiguity in the devanagari alphabet rava-kha is excluded by the metrical
nature of our text. For the rest, there is an endless chain of mislections.
The commonest slip, both of pronunciation and writing, is vaba, while a little
projection carries into ca, so that eitta and citta are confused throughout.
The lotters pa and ya are written almost alike, while a little pinching makes
ya look like ma, which can sometimes bo misread as sa, and mispronounced as
sa which in turn is close to the written form of tra. On the other hand,
sa=kha is a common interchange arising from Rajashani specchi, like ya ja.
Moro complicated are the changes due to tongue-twisters; jatharahariņāḥ in
239, rüvaṇaḥ in 49 are simple metathesis. The interchange su sva (whence
sujana =svajana) probably goes back to the oldest form of devanagari, when
the consonant va was joined below the line like the u vowel mark.
Writing the u sign in the line probably causes the variant viphala for
vipulu[ 29ª]. The readings punyair and puspair in 181" are epigraphic
variants as are ittham iccham, arthaargha.
 
E
 
Fortunately, not all the possible variants of this sort occur in any single
manuscript, and they can generally be corrected by noting special tendencies
manifested by the source; occasionally, copying from an older codex with
pṛṣṭhamātrās has also to be compensated. The greater number of these
present no serious problem, and have generally been corrected silently when
the original reading was not in doubt. Our purpose here being not only to