own by word of mouth and even in the present day most of it
could be recovered if all manuscripts and books were lost. The Buddhists
did not, like the Brahmans, make minute regulations for preserving and
memorizing their sacred texts, and in the early ages of the faith were
impressed with the idea that their teaching was not a charm to be learnt
by heart but something to be understood and practised. They nevertheless
endeavoured, and probably with success, to learn by heart the words of
the Buddha, converting them into the dialect most widely understood. It
was then a common thing (and the phenomenon may still be seen in India)
for a man of learning to commit to memory a whole Veda together with
subsidiary treatises on ritual, metre, grammar and genealogy. For such
memories it was not difficult to retain the principal points in a series
of sermons. The Buddha had preached day by day for about forty-five
years. Though he sometimes spoke with reference to special events he no
doubt had a set of discourses which he regularly repeated. There was the
less objection to such repetition because he was continually moving
about and addressing new audiences. There were trained Brahman students
among his disciples, and at his death many persons, probably hundreds,
must have had by heart summaries of his principal sermons.

But a sermon is less easy to remember than a poem or matter arranged by
some method of _memoria technica_. An obvious aid to recollection is to
divide the discourse into numbered heads and attach to each certain
striking phrases. If the phrases can be made to recur, so much the
better, for there is a guarantee of correctness when an expected formula
appears at appropriate points.

It may be too that the wearisome and mechanical iteration of the Pali
Canon is partly due to the desire of the Sinhalese to lose nothing of
the sacred word imparted to them by missionaries from a foreign country,
for repetition to this extent is not characteristic of Indian
compositions. It is less noticeable in Sanskrit Buddhist sutras than in
the Pali but is very marked in Jain literature. A moderate use of it is
a feature of the Upanishads. In these we find recurring formulae and also
successive phrases constructed on one plan and varying only in a few
words[620].

But still I suspect that repetition characterized not only the reports
of the discourses but the discourses themselves. No doubt the versions
which we have are the resul