It has long been noticed that discourse functions can be encoded
configurationally, usually accompanied by a particular intonation
pattern.
In LFG, this
can be captured by annotations on the c-structure which provide both a
discourse function and a grammatical function to the constituent.
This is shown in (1) which states that YP has the discourse
function DF (e.g., TOPIC or FOCUS) and that this DF is identical to
some grammatical function indicated by the uncertainty equation BODY
BOTTOM (Kaplan and Zaenen 1988 and 1989). That is, there is exact
sharing of information between the discourse function and the
grammatical function.
(1)
For example, Huang 1992 proposes the equations in (2) for
topicalization in English. The left dislocated XP is assigned the TOP
discourse function. The TOP is then equated with any grammatical
function other than COMP (i.e., GR-COMP) at any level of COMP or
XCOMP (i.e., {COMP|XCOMP} 
). So, in a sentence like
(3), the TOP is also the XCOMP's OBJ, as seen by the f-structure
in (3).
(2)
(3) F-structures I can format easily.
In King 1995 I proposed a similar account for Russian li
questions in which the constituent preceding the complementizer
li is the focus of a yes-no question. The c-structure rule in
(4) places the fronted constituent in Q-FOC and
equates it with a grammatical function from any number of XCOMPs
(COMPs do not allow extraction in Russian). So, in a question like
(5), the fronted NP, knigu `book', is the Q-FOC and the OBJ
of the top level f-structure.
(4)
(5)
Thus, the encoding of discourse functions via annotations on the c-structure tree works relatively well for simple arguments and adjuncts, assigning them both a grammatical function and a discourse function in the f-structure.