Structural case, locality and cyclicity

Abstract

It is argued that long-distance case assignment does not actually exist. Unlike longdistance agreement, it would necessarily violate the Phase Impenetrability Condition. All examples that have been analyzed in terms of long-distance case actually involve the appearance of the nominative as the default case. Cases that are actually assigned – the accusative and the non-structural cases – always obey locality. Yet assignment of the accusative does appear to violate cyclicity. This problem can however be solved as well with the right view on when case is assigned, once the dependency between structural accusatives and other DPs is properly understood. As a result, the alternative solution put forward by Sigurðsson (2006b), requiring that subjects first-merge lower than objects, is shown to be unnecessary. 1 Phases and long-distance agreement Apparent long-distance agreement and case-assignment, as exemplified in the Icelandic example 1a (from Sigurðsson, 2006b) and the English example 1b (from Legate, 2005b), raise important questions for the locality of syntactic operations.1 (1) a. Það there voru were:PL taldir believed hafa have verið been veiddir caught fjorir four laxar. salmon:N.PL ‘People believed there to have been four salmon caught.’ b. There seem to have arrived ten trains into this station today. On the one hand, nominative case on the DP fjorir laxar ‘four salmon’ in the embedded non-finite clause would appear to be assigned or licensed from finite T in the matrix clause. On the other, the case and phi-features on fjorir laxar must trigger nominative plural agreement on veiddir ‘caught’, taldir ‘believed’ and voru ‘were’ (similarly in the English example). Crucially, the dependencies here are in principle unbounded, as arbitrarily many non-finite raising clauses can intervene between the finite verb and the clause containing the DP. In Chomsky (2000, 2001, 2005) this is handled by defining the units of syntactic representation relevant for locality in such a way that no boundaries intervene between the low nominative DP and the agreeing finite T in such sentences. Specifically, only C and v* (the flavor of v which introduces an external argument) define phases, and neither of these is present at the relevant levels. Crucially, plain v (the version which appears in passive and unaccusative clauses like in 1 and does not introduce an external argument) is not assumed to be a phase head.2 However, the distinction between the two types of v is suspicious, especially if phases are really supposed to be “the closest syntactic counterpart to a proposition: either a verb phrase in which all θ-roles are I would like to thank Artemis Alexiadou, Berit Gehrke, Kirsten Gengel, Nino Grillo, Kleanthes Grohmann, Alec Marantz, Florian Schafer, Halldor Sigurðsson and Sandhya Sundaresan for comments and helpful discussion. In some formulations of the theory, vP is considered a phase, but not a strong phase. Since what matters for the locality of operations under those formulations is the strong phase, this point is not relevant for our concerns.