In this book, Ian Roberts proposes to account for head movement in the narrow syntax, arguing that head movement is simply a type of Agree.
In chapter 1, Roberts tackles what is arguably the most important argument Chomsky gives for the idea of relegating head-movement to PF—that it does not cause LF effects. Roberts shows, using evidence from various Romance languages and English, that contra Chomsky, head movement has semantic effects after all and is thus still required within narrow syntax. [End Page 287]
Chapter 2 addresses the differences (and similarities) between head movement and pied-piping as a result of phrasal movement. Roberts notes here that while Agree and Merge are well-defined operations, Pied-Pipe is left with no formal definition. This leads him to explore the mechanisms within Pied-Pipe, and ultimately to argue that the principles of Structure Preservation and Chain Uniformity cannot force this type of movement, while the A-over-A Condition (Rackowski and Richards 2005) can. The A-over-A Condition not only explains the examples of pied-piped movement, but also allows for head movement since it can be configured for both minimal and maximal category movement, as in (1).
(1). A-over-A condition and categorial targets:
“A goal α is the closest one to a given probe if there is no distinct goal β such that for some X (X a head or maximal projection), X c-commands α but does not c-command β.”(p. 34)
The difference between maximal and minimal projections for Roberts is that pied-piping occurs when a maximal category is the target of the probe, while head movement occurs when a minimal category is the target. I present this idea in further detail in Table 1.
The third chapter contains the crux of many of Roberts’ arguments and is therefore by far the longest. This chapter is devoted to the many types of clitics found (primarily) in Romance. An in-depth look at the various phenomena associated with clitics allows Roberts to substantiate at least one of the three conditions he suggests restricting head movement to.
(2). Restricted environments where head movement is expected/allowed:
a. β lacks internal structure, that is, it is βmin/max.
b. βP lacks a specifier [… ], that is, the structure is [βmax βmin Y].
c. Spec, βP is not a goal for P[robe] [… ] while βmin is, that is,
… P+F … [β P+F XP+G [… βmin +F …(p. 39)
Although Roberts ultimately argues that (2b) is merely a theoretical possibility and is not likely to exist because Edge Features mandate that phase edges are necessarily filled, he argues that (2a) is the structure of clitics, and that (2c) is instantiated in verb movement, both V/v-to-T and V/v-to-C, which is the subject of chapter 4.
Chapter 5, unlike the others, does not attempt to motivate the need for head-movement as an operation in narrow syntax per se, but instead details how it can be incorporated into the theory of movement more generally. Here, Roberts argues that additional ad hoc mechanisms would be required to remove head-movement from narrow syntax, that head-movement is a natural result of the theory of movement, and that by maintaining it as part of the narrow syntax we end up with a more minimal and elegant theory. Roberts discusses the logical extensions of the theory of movement as it exists (p. 208) which I adapt here into Table 1.
Beyond the empirical evidence adduced throughout the previous sections in support of head-movement as a syntactic operation, rows b. and h. show that, even on theoretical grounds, this type of movement belongs in the narrow syntax. As Roberts [End Page 288] shows here, head-movement is a natural consequence of the theory of movement, and to remove it would require additional ad hoc machinery. In sum, this would lead to a more cumbersome theory and one which would not...