Agree in syntax, agreement in signs

This paper explores the idea that abstract Agree is a precondition on Merge and an integrated part of it. That is, an element F merges with the structure X only if the relation of Agree holds between the two (the Agree Condition on Merge). The relation of Agree holds between F and X iff X contains an active feature fx that matches F. Move is forced by an inactive intervener ?x between F and fx, which, if not crossed by fx, would block matching, F ↔ fx. It follows that Move and Merge are fundamentally different, Move tucking in, as a ‘rescuing operation’ in an already existing structure, whereas Merge adds information to structure, thereby expanding it.
Whenever Merge applies, the possibility of agreement arises, i.e. languages make parametric (PF) choices whether or not to signal each instance of Merge morphologically, that is, agreement is in effect a ‘sign of compositionality’. The various agreement phenomena of Icelandic illustrate that agreement involves feature copying processes that take place exclusively in PF. Thus, morphological agreement is quite distinct from (albeit preconditioned by) abstract syntactic Agree. In addition, the Icelandic facts discussed suggest that also ‘head movement’ is confined to PF.
If this is on the right track, PF is a multilayered and a highly complex system, producing strings that can be radically different from underlying syntactic structures.