One of the most widely accepted facts about word meaning is that many
words share meaning in common, grouping words into classes, but that
within each class there's something unique and idiosyncratic about
each word. It is commonly assumed that the grammar of a language only
"sees" the regular bits --- the idiosyncratic parts of a given words
meaning are not really linguistically very significant, so much so
that studying their individual content is not likely to be very
fruitful. In this talk I explore this issue in detail, and suggest
that hidden within the idiosyncrasies of individual words are types of
regularity --- sometimes the very same regularities that define whole
word classes --- that are significant for how the word is used
grammatically. I draw on two primary case studies, one a broad,
typological study of the ways subtly different cause/effect scenarios
are described across a balanced sample of languages, and one on the
diverse patterns of expressions of giving in English and the Bantu
language Kinyarwanda. I conclude by suggesting that regularity and
idiosyncrasy are not so different from one another, and form more of a
continuum than a sharp division.