332 Dik Bakker and Anna Siewierska
between ‘raw’ UR material (functions, π and ω operators, and inherent fea-
tures of predicates such as Animacy and Gender) on the one hand, and μ
operators on the other hand. In order to be accessible for inheritance, it will
be assumed that any UR feature has to be transformed into a μ operator,
either in a one-to-one or a more-to-one fashion, as in the case of portman-
teau operators. An implication of this is that only those features of URs
qualify as such if they are turned into a μ operator at some stage during ex-
pression. Apart from this there may be universal or language-(type)-
specific downward barriers that make features inaccessible to lower nodes.
Barriers to inheritance may be either functional or formal in nature, i.e. ei-
ther determined by a layer or substructure in the UR or determined by
some syntactic or morphological boundary, i.e. a specific node in the tree.
Principle 5: Operators may percolate, i.e. move upwards to higher
nodes. For percolation there may be universal and language-(type)-specific
upward barriers that make them inaccessible to higher nodes in the tree.
Barriers to percolation are only formal in nature.
These five principles should contribute to the (cognitive and typologi-
cal) adequacy of the dynamic model. They have two interesting implica-
tions for the global working of the model. On the one hand, linguistic
forms are produced precisely in the order in which they are uttered by the
speaker in a live setting, thus giving the model a real-time procedural fla-
vour. On the other hand, several constraints follow from the principles,
which restrict the formal power of the expression component. Indeed, it
may now be determined what information has to be available at what stage
in the tree expansion process. In this way, it may be determined more pre-
cisely what information necessarily has to come in from the UR as opposed
to arising in the course of expression, and thus, which elements have predi-
cational status in the language under consideration and which ones do not.
Furthermore, these constraints provide a means of determining the order in
which grammatical material has to be available in the first place. For ex-
ample, a case-marking postposition is needed on some node in the tree
before the elements on which the corresponding case is expressed are cre-
ated. Since grammatical forms may come into existence at several stages in
the expansion process, we also get an instrument to distinguish between
partially and fully grammaticalized elements. The former will appear rela-
tively early (‘high’) and the latter relatively late (‘low’) in the constituent
structure. Arguably, this distribution may reflect the length of their dia-
chronic path through the language, typically leading from a full predicate
to a fully grammatical entity, and their relative role in the syntax and mor-
phology.