PROBLEMS OF LANGUAGE IN BUDDHIST TANTRAbeing translated into the second term but insignificant if first and second terms
are both "twilight." We have arrived at our list-which we have written out with
verbal forms as it appears in the edited texts-by translating consistently from
the initial word order thus: madanam madyam (as it stands, "intoxication
wine"), "wine is ... intoxication." Interestingly the Tibetan version transliterates
one side of the equation and translates the other; such a practice as this suggests
strongly that the transliterated term is a primary term-the one supposedly
found in Sanskrit manuscripts as salf1dhii-for which a "meaning" is given.^49
While there are exceptions, these transliterated terms along with our first terms
tend to be the ones found in the "twilight" song. Thus, there is a tendency to
treat the list's first terms as salf1dhii. This agrees with our earliest impressions
but contradicts our later findings that both first and second terms can be used in
salf1dhiibhiisii.
These materials also call into question our conclusions on ambiguity and
secrecy. We have argued that ambiguous language cannot be known and have
agreed with the position that what cannot be known cannot be said to be secret.
At first glance, the Hevajra's repeated statement that an initiate faces disaster "if
he does not speak in 'twilight language'" appears to support us. This could be a
way of saying that failure to use salf1dhii-bhiisii means a loss of siddhi to which
the ambiguity of special discourse opens up a candidate. And this does provide a
certain logic for our reading also that "twilight" discourse belongs to yoginis
who are "givers of good siddhi" but who-when their language is not used-
reverse roles and "produce their wrath." But the text also says that the "great
symbolism" (mahiisamaya) "is not understood" (na chidrita; literally, "not pene-
trated") by Sravakas, with the double implication that salf1dhii-bhiisii is actually
understood by those who speak it and that this language serves as some kind of
screen. Indeed, Kal)ha glosses mahiisamaya as guptasalf1keta, which may be
rendered "hidden convention"; and the Tibetan renders salf1dhii-bhiisii as dgons
pa/:zi skad ("twilight language") as well as gsmi ba/:zi skad ("secret language").
Any doubt on this issue is dispelled by Vajragarbha, who states explicitly that
"twilight language" is the "great secret" (gsan chen) which keeps outsiders
"baffled" (rmons pa ). That commentary even instructs the initiate to speak with
"twilight language" instead of "ordinary convention" (tha mal pa/:zi tha siiad) as
if these two forms of discourse were interchangeable. But we have argued that
such an interchange is meaningless.
Of course, these additional findings are not at all troublesome for scholars
who take the position that salf1dhii-bhiisii is ambiguous yet at the same time
expresses a certain meaning which can be concealed. But, as we have seen, this
position has its own troubles with the materials analyzed earlier and leaves
untouched its logical inconsistency. Let us note further that the issue of
secrecy-interpreted from whatever materials-has been complicated greatly by
our discovery that the full range of salf1dhii equivalence can find its way into
sal!1dhii-bhiisii. As the issue is frequently discussed, "twilight language" con-
ceals the tantric use of degenerate or controversial elements which are better