Generative SLA views language as a symbolic system, autonomous from cognition, and too complex to be acquired by training or through inductive or deductive learning from the input.
According to researchers working in the generativist paradigm, language acquisition is guided by an innate language-specific learning device (Universal Grammar) which is activated by exposure to input and requires little or no negative evidence in the form of feedback from interlocutors or instruction.
For second language acquisition, generativism assumes that interlanguage representations are constrained by Universal Grammar. This is demonstrated when interlanguages show properties which are absent in the L1 and are also difficult to detect in the L2, giving rise to the same logical problem of acquisition identified by Chomsky for L1.
The full-transfer, full-access hypothesis posits that first and second language acquisition involve the resetting of parameterized universal principles, triggered only by positive evidence (i.e., input), and that there is no role for negative evidence (e.g., instruction concerning what is not possible in the L2) [...W]hen the L2 is a proper subset of the L1 with respect to a certain aspect of language, L2 learners will have to retreat from the overly general hypothesis that emanates from their L1 (White, 1987, 1991), something which cannot be done on the basis of positive evidence alone.
For generativists, because the problem of language acquisition is so acute, it cannot be solved by interaction with the environment. Accordingly, generativists accord little or no place for instruction in the acquisition process.
By the UG SLA accounts, then, instruction is either entirely or largely unnecessary.