Five reasons for building phrase structures top-down from left to right

In this paper I will summarize five arguments supporting an explicit formalization of a minimalist grammar which is derivational and directional: derivational since constituents and dependencies are built dynamically, piecemeal, using structure building operations such as merge and move in a phase-tailored computation; directional in the sense that these structure building operations operate strictly from left-to-right as proposed by Colin Phillips (1996, 2003) and top-down (Chesi 2004). This directionality issue is obviously the main difference with respect to standard minimalist approach 1 . The issues I will summarize are both formal and empirical and are tightly related to the revised definition of merge and move: first of all I will go trough Phillips’ (2003) arguments, showing how merge should be redefined in a left-to-right manner, merge-right, based on the evidence that intermediate constituents can be targets for coordination. This will allow us to capture interesting conflicting results among constituency and c-command tests (section 3, Phillips 2003) in a cross-linguistic perspective (Choi and Yoon 2006). Two formal arguments will be presented, then, in order to introduce the directionality shift with respect to the move operation and the notion of derivation by phase: unmotivated intermediate steps for movement (section 4, Chesi 2004a) and growing complexity with respect to cross-phasal dependencies (section 6, Chesi 2004b) suggest that top-down, left-to-right derivations are interesting solutions to be explored systematically: the complexity argument, in fact, isolates a natural class of constituents which exactly correspond to strong islands; left-to-right derivations are able to provide an unified account of such constituents, as in Huang (1982), without loosing the ability to discriminate important empirical differences (section 5, Chesi 2004a). This is especially true for parasitic gaps constructions: we will be able to recast the connectedness effect (Kayne 1983) in derivational terms without stipulating any complex-NP constraint (section 7, Bianchi and Chesi 2005).