350 Dik Bakker and Anna Siewierska
tion: the Ag argument, which happens also to have Topic function. This
leads to the binding of the Subject-related variables in (17), which gives us
(18).
(18) NODE 1 (partially instantiated)
Lab: sentence
Config: [decl (P 1 )S (P 2 )A (C 6 : [(x 1 )TOP (x 8 )FOC (f 1 )])],
[(past e 1 : [hab f 1 : come [V] (def x 1 )AG,SUBJECT]
(gen sg x 8 : class [N])ILLAT ) ] )
FncFtrs: decl, past, hab, [V], sg, 3
FrmFtrs: x 1
The next step is the selection of the right template, for which the actual
contents of Config are scanned. This has been done in (19), where three
slots are introduced. Note that there is a position for the subject rather than
a general P1 position, since it is ‘known’ that the subject will land here in
this case. The other positions are for the verbal cluster and the satellite.
(19) NODE 1 (fully instantiated)
Lab: sentence
Config: [decl (P 1 )S (P 2 )A (C 6 : [(x 1 )TOP (x 8 )FOC (f 1 )])],
[(past e 1 : [hab f 1 : come [V] (def x 1 )AG,SUBJECT]
(gen sg x 8 : class [N])ILLAT ) ] )
FncFtrs: decl, past, hab, [V], sg, 3
FrmFtrs: x 1
SubCat: subj, vc, sat
This will now be expressed further via the process sketched in Section- This will lead to the expression of (12), on the basis of the partial tree
depicted in Figure 4 below (the numbers next to the arrows give the order
of creation; we have left out the upward-pointing percolation arrows). Note
that the tree below is in fact completely ‘flat’, i.e. there are no ‘unneces-
sary’ nodes and branches. We think that structure is represented in the first
place on the RL and IL levels and is not necessarily reflected iconically
and without further arguments in the structure of morphosyntax. Further
elaboration of the theory of expression should make clear whether more
formal structure and more formal categories are necessary to represent the
linguistic facts beyond some (relative or absolute) minimum.^21