This paper proposes that the notion 'language instinct' appropriately chara
cterizes nonnative language (L2) acquisition in two distinct ways. I argue
that like native language (L1) development, L2 development, even by adults,
relies on language instincts - despite L1-L2 differences at intermediate s
tages and in ultimate attainment - and that a primary source of L1-L2 diffe
rences is differences in their respective initial states. A variety of acqu
isition data, from the L2 child, the L2 adolescent and the L2 adult, are us
ed to illustrate and assess three models that adopt this general characteri
zation of L2 acquisition: Minimal Trees (Vainikka and Young-Scholten, 1994)
, Weak Transfer (Eubank, 1993/94) and Full Transfer/Full Access (Schwartz a
nd Sprouse, 1996), These proposals differ on the extent of L1 influence, i.
e., on the representation of the L2 initial state, and I show that the L2 d
ata support Full Transfer/Full Access.