proposition that Celtic was not spoken in Silchester. But all
probabilities suggest that it was, at any rate, spoken very little. In
the twenty years' excavation of the site, no Celtic inscription has
emerged. Instead, we have proof that the lower classes wrote Latin for
all sorts of purposes. Had they known Celtic well, it is hardly credible
that they should not have sometimes written in that language, as the
Gauls did across the Channel. A Gaulish potter of Roman date could
scrawl his name and record, _Sacrillos avot_, 'Sacrillus potter', on the
outside of a mould.[1] No such scrawl has ever been found in Britain.
The Gauls, again, could invent a special letter Ð to denote a special
Celtic sound and keep it in Roman times. No such letter was used in
Roman Britain, though it occurs on earlier British coins. This total
absence of written Celtic cannot be a mere accident.

[Footnote 1: One example is _Sacrillos avot form._, suggesting a
bilingual sentence such as we find in some Cornish documents of the
period when Cornish was definitely giving way to English. Another
example, _Valens avoti_ (Déchelette, _Vases céramiques_, i. 302),
suggests the same stage of development in a different way.]

No other Romano-British town has been excavated so extensively or so
scientifically as Silchester. None, therefore, has yielded so much
evidence. But we have no reason to consider Silchester exceptional in
its character. Such scraps as we possess from other sites point to
similar Romanization elsewhere. FVR, for instance, recurs on a potsherd
from the Romano-British country town at Dorchester in Dorset. A set of
tiles dug up in the ruins of a country-house at Plaxtol, in Kent, bear a
Roman inscription impressed by a rude wooden stamp (Fig. 6).[1] In
short, all the _graffiti_ on potsherds or tiles that are known to me as
found in towns or country-houses are equally Roman. Larger inscriptions,