Minimalist Merge, destructive feature-checking, and sequential unification

In 1967, Ross tried to characterize the upper bound of the set of well‐formed sentences in a natural language focussing on a collection of constraints on, powerful enough, recursive procedures (i.e. rewriting/transformational rules). After 40 years we still face the very same problem: move α and the projection principle before (Chomsky 1965), merge, move and agree now (Chomsky 2005a) are very general and (too) powerful devices that perhaps allow us to figure out very important universal principles, but that are practically insufficient to constrain many relevant empirical phenomena. Within the spirit of the minimalist initiative, in this talk I will try to show that a more restrictive definition of merge can be successfully rephrased in top‐down (phase‐based), left‐right terms, attaining superior results in terms of computational economy and empirical adequacy, at least with respect to a relevant set of phenomena such as argument cluster coordination Vs. fronting/scrambling/clefting asymmetries (Phillips 1996, Choi and Yoon 2006), and “spec”‐head/multiple agreement.