[…] the syntactic component of the grammar generates an infinite set of abstract structures - call them "S-structures" that are assigned a representation in phonetic form (PF) and in LF (read: "logical form"[...]).
- Chomsky (1993), a pag.4

The base generates D-structures (deep structures) which are associated with S-structures by the rule Move-α. To clarify terminology, I will use the term "surface structure" in something like its original sense, referring to the actual labelled bracketing of an expression at the level PF. Changing usage over the years has given rise to a fair degree of confusion. The term "surface structure" has been used in much recent work to refer to a more abstract representation than the actual labelled bracketing of an expression.
- Chomsky (1993), a pag.18

[…] we may therefore think of S-structure as an enriched D-structure incorporating the contribution of D-structure to LF.
- Chomsky (1993), a pag.45

S-structure is the fundamental level of syntactic representation, receiving its interpretation at the levels of PF and LF by rules of the PF- and LF-components, respectively. At S-structure, each NP apart from trace is assigned a function chain (GF1..., GFn), whether it is an argument (including PRO) or a non-argument (idiom chunks, pleonastic "it" and "there", etc.) [...] We have decomposed S-structure into two factors: D-structure, which is a representation of GF-θ, and a rule adding GFs to a function chain.
- Chomsky (1993), a pag.137