Reprojections

It is hard to decide whether a proposal should be couched in derivational or representational terms, or some combination of the two. The best kind of argument for derivations involves the need for representations whose information is required piecemeal–being vital at one point and unwanted at another. In such cases, one would favor analyses that involve partial representations whose properties are not conserved as the derivation unfolds. This “loss of information” scenario fits poorly into models which make use of enriched representational codings (including traces, indices, and similar elements) because whatever information may have been lost in the course of the derivation can be reconstructed through some abstract coding.

What follows offers a candidate “loss of information” phenomenon. It is expressed within one particular interpretation of the Minimalist Program which drastically reduces notational options …