(13) LAYERING PRINCIPLE: When a rule assigns features from a paradigmatic dimension D to a morphosyntactic representation R that already contains values from D, the result is that the previous values are made hierarchically subordinate to the new values. [...] our theory of morphosyntactic representation is essentially that these are unordered feature complexes, whose only internal structure is that which results from principle (13). This suggest a rather resricted picture of just how the syntax interacts with these objects: In fact, we can propose that the only effect of the syntax on the morphosyntactic representation is limited by the constraint in (17). (17) The only operation a syntactic rule can perform on a morphosyntactic representation is to add features to it. In addition to this constraint on the effect syntactic operations can have on morphosyntactic representation, we would clearly like to constrain the accessibility of these latter to the syntax. Ideally, something like (18) might be mantained. (18) Only a rule that adds features to one morphosyntactic representation can examine the content of another.
- Anderson (2004), a pag.40

A minimal theory of this MORPHOSYNTACTIC REPRESENTATION would be that it consists simply of an unordered bundle of features, including all and only those feature values that (in a particular language) are referred to, introduced, or manipulated by the syntax. This is essentially the line Chomsky took in "Aspects" [Chomsky, N., 1965, Aspects of the Theory of Syntax, MIT Press], where the complex symbols that constituite the terminal nodes of phrase markers play the same role as our morphosyntactic representation.
- Anderson (2004), a pag.37