Children improve at word-learning during the second year of life -
sometimes dramatically. This fact has suggested a change in
mechanism, from associative learning to a more referential form of
learning. This paper presents an associative exemplar-based model
that accounts for the improvement without a change in mechanism. It
provides a unified account of children's growing abilities to: (i)
learn a new word given only one or a few training trials (``fast
mapping''); (ii) acquire words that differ only slightly in
phonological form; (iii) generalize word meanings preferentially along
particular dimensions, such as object shape (the ``shape bias''); and
(iv) learn second labels for already-named objects, despite a
persisting resistance to doing so (``mutual exclusivity''). The model
explains these improvements in terms of increased attention to
relevant aspects of form and meaning, which reduces memory
interference. The interaction of associations and reference in
word-learning is discussed.