Cyclic Linearization of Syntactic Structure
DANNY FOX and DAVID PESETSKY
Abstract
This paper proposes an architecture for the mapping between syntax and phonology
– in particular, that aspect of phonology that determines the linear ordering of
words. We propose that linearization is restricted in two key ways. (1) the relative
ordering of words is fixed at the end of each phase, or ‘‘Spell-out domain’’; and (2)
ordering established in an earlier phase may not be revised or contradicted in a later
phase. As a consequence, overt extraction out of a phase P may apply only if the
result leaves unchanged the precedence relations established in P. We argue first
that this architecture (‘‘cyclic linearization’’) gives us a means of understanding
the reasons for successive-cyclic movement. We then turn our attention to more
specific predictions of the proposal: in particular, the e¤ects of Holmberg’s
Generalization on Scandinavian Object Shift; and also the Inverse Holmberg Ef-
fects found in Scandinavian ‘‘Quantifier Movement’’ constructions (Ro
¨
gnvaldsson
(1987); Jo
´
nsson (1996); Svenonius (2000)) and in Korean scrambling configura-
tions (Ko (2003, 2004)). The cyclic linearization proposal makes predictions that
cross-cut the details of particular syntactic configurations. For example, whether an
apparent case of verb fronting results from V-to-C movement or from ‘‘remnant
movement’’ of a VP whose complements have been removed by other processes, the
verb should still be required to precede its complements after fronting if it preceded
them before fronting according to an ordering established at an earlier phase. We
argue that ‘‘cross-construction’’ consistency of this sort is in fact found.
1. Introduction
1
This paper proposes an architecture for the mapping between syntax and
phonology – in particular, that aspect of phonology that determines
Theoretical Linguistics 31 (2005), 1–45 0301–4428/05/0031–0001
6 Walter de Gruyter
1
The present paper is the first installment in a larger project (Fox and Pesetsky (in
prep.)). The inspiration for this work can be traced to a conversation with John
ordering. In Fox and Pesetsky (in prep.), we will argue that this archi-
tecture, when combined with a general theory of syntactic domains
(‘‘phases’’), provides a new understanding of a variety of phenomena
that have received diverse accounts in the literature. This shorter paper
focuses on two processes, both drawn from Scandinavian: the familiar
process of Object Shift and the less well-known process of Quantifier
Movement. We will argue that constraints on these operations can be
seen as instances of the same property of grammar that explains the fact
that movement is local and successive cyclic. We begin by sketching a
model in which locality and successive cyclicity are consequences of the
architecture that we propose, rather than specific facts about movement
itself. We next present our proposal in somewhat greater detail, and
show how it can account for a wide range of apparent limitations on
movement – in particular, superficially contradictory restrictions on
Object Shift and Quantifier Movement. The restrictions on Object Shift
include those grouped under the rubric of Holmberg’s Generalization,
which Quantifier Movement does not seem to obey. We will argue that
Quantifier Movement instead obeys a near mirror-image of Holmberg’s
Generalization (an ‘‘Inverse Holmberg E¤ect’’), but that both Holmberg’s
Generalization and its mirror image are expected if our proposed archi-
tecture is correct.
Our discussion will be for the most part informal, but we will con-
clude by o¤ering a more formal implementation of our proposals. This
Frampton, and to a fruitful additional meeting with John Frampton and Sam
Guttmann. We are grateful to both of them, and also to the students and visitors who
attended our Spring 2003 seminar at MIT, where much of this work was first presented
and developed. We also want to express our gratitude for the contributions to our work
of Elena Anagnostopoulou, Jonathan Bobaljik, Misi Brody, Noam Chomsky, Anders
Holmberg, Sabine Iatridou, Kyle Johnson, Noriko Kawasaki, Heejeong Ko, Alec
Marantz, Andrew Nevins, Øystein Nilsen, Jon Nissenbaum, Christer Platzack, Norvin
Richards, Joey Sabbagh, Peter Svenonius, Shoichi Takahashi, and Sophia Tapio, among
others. We also wish to thank for their comments and questions audiences at Sophia
University (Tokyo); Universite
´
de Paris VII; Indiana University; Kyoto University
(Kaken Ellipsis Workshop); University of Vittoria (Basque Country); University of
Milan-Bicocca; the University of Massachusetts/Amherst; the Workshop on Interfaces
(Cha
ˆ
teau de la Brete
`
che); the 2003 LSA Summer Institute (Michigan State University);
GLOW 2003 (Lund); and the First North American Syntax Conference (Concordia
University, Montreal).
2 Danny Fox and David Pesetsky
implementation will belong to a family of possibilities consistent with our
overall model. We will leave for the fuller presentation a comparison of
these alternatives, along with a range of other emp irical consequences.
2. Successive-cyclicity and linear order
Much attention has been devoted to movement relations that appear to
be long-distance or unbounded. A large body of work has converged on
the conclusion that such relations actually represent the result of the com-
pounding of a series of more local relations. For example, the apparent
long-distance property of wh-movement in a sentence such as (1) seems
to be a consequence of a series of shorter movement relations – at least
through intermediate specifiers of CP and specifiers of VP
2
as show n in
(1):
(1) Successive-cyclic wh-movement
I wonder [which book he [
thinks [
CP
Mary [
CP
read ]]]]
;
;
;
;
Evidence for the possibility of successive cyclic movement through
Spec,CP and Spec,VP includes a variety of reconstruction phenomena
(see Barss (1986), Lebeaux (1991) for CP; Fox (1999) for VP) – as well
as the possibility of stranding wh-related elements (McCloskey (2000))
and perhaps wh-phrases themselves (McDaniel (1989); Grac
ˇ
anin (2004))
in at least some intermediate Spec,CP and Spec,VP positions.
Additional evidence supports an even stronger conclusion. If Chomsky
(1973) and others are correct in linking island e¤ects to the existence
of intermediate landing sites, it must be the case that long-distance
movement does not merely permit, but actually requires the use of these
intermediate positions. We must then ask why. This question has figured
prominently in much discussion over the past quarter century, and a
variety of answers have been suggested.
2
To keep the discussion simple, we will not make a distinction between vP and VP (as
argued for by Hale and Keyser (1993) and Hale and Keyser (2002), among others) in
the first part of this paper. We return to the vP/VP distinction below and in Fox and
Pesetsky (in prep.).
Cyclic Linearization of Syntactic Structure 3
All proposals known to us attribute a special status to particular
syntactic domains, variously called ‘‘bounding nodes’ ’, ‘‘barriers’’ and
‘‘phases’’. Our proposal will not di¤er in this respect. For now, we will
follow Chomsky (2000) in using the term ‘‘phase’’ for the domains in
question, to highlight a similarity between our proposal and his, to which
we return later. Traditionally, phases play two distinct roles – and it is on
this point that our proposal will crucially di¤er from others, including
Chomsky’s. On the one hand, a phase is taken to determine an ‘‘impene-
trable domain’’ from which movement is generally excluded. On the other
hand, it is also taken to determine a specific position as an ‘‘escape
hatch’’. The escape hatch provides the one and onl y legal route for move-
ment out of the phase. Thus, for example, if CP is a phase, it determines
that movement out of CP is impossible from any position except Spec,CP
– its designated escape hatch. This forces all apparent long-dista nce
movement from CP to be analyzed as involving movement through its
escape hatch. More generally, apparent movement from any phase, on
traditional accounts, must involve the escape hatch of that phase as an
intermediate step.
Our approach will have in common with traditional proposals the re-
quirement that movement from a phase proceeds through certain specific
types of positions within that phase. Unlike traditional proposals, how-
ever, our approach does not stipulate particular positions that provide es-
cape hatches for movement – nor is there any concept of ‘‘escape hatch’’
built into the theory. ‘‘Escape hatch e¤ects’’ are, if we are correct, an ar-
tifact of the role played by phases in linearization of the terminal elements
of syntactic structure. These e¤ects are thus a consequence of the map-
ping between syntax and phonology.
3
Consider the derivation in (2), in which wh -movement procee ds
through intermediate specifiers of VP and CP:
4
3
For other proposals that relate island e¤ects (and requirements of successive cyclicity) to
syntax-phonology interactions, see Pesetsky (1998) (with an antecedent in Perlmutter
(1972)); Uriagereka (1999); among others.
4
It is important to our discussion that the intermediate positions indicated in (2) are on
the left side of VP and CP. This proposal is independently supported (for CP) by the
observations of McCloskey (2000). Our paper as a whole will constitute a sustained ar-
gument for this position. See Nissenbaum (2000) for possibly conflicting evidence.
4 Danny Fox and David Pesetsky
(2) Successive-cyclic wh-movement through Spec,CP and Spec,vP
[To who m will he [
say [
CP
that Mary [
VP
gave the book ]]]]?
;
;
;
;
43 2 1
Let us focus on a particular property of the movements seen in (2). Con-
sider first step 1, which takes place entirely within the lowest VP phase.
This step can be seen as ‘‘revising’’ the word order of VP insofar as wh-
movement to Spec,VP causes the wh-phrase to precede the words gave
the book. which it would otherwise follow. Now consider step 2. This
step also ‘‘revises’’ word order, causing the wh-phrase to precede the
words that Mary which, once more, it would otherwise have followed.
Notice, however, that step 2 only revises word order with respect to these
words – not with respect to the words contained within the previous
phase. The wh-p hrase preceded the words gave the book at the end of the
lowest VP phase, and still precedes them at the end of the next phase.
This type of observation can be repeated for each subsequent step.
Thus step 3 revises the order of the wh-phrase with respect to the verb
say, but not with respect to the words that Mary gave the book. Likewise,
step 4 revises the order of the wh -phrase with respect to the words will he,
but not with respect to any of the words dom inated by lower phases.
The generalization is clear. Each time the wh-phrase moves, it leaves
unchanged its linear order with respect to words dominated by previous
phases, but may revise linear order with respect to words introduced in
the current phase. We propose that this property of (2) is not a coinci-
dence. As we will argue, if this property did not obtain, the sentence
could not be linearized. This proposal will constitute the core of our
explanation for the obligatoriness of suc cessive-cyclic movement and will
be the key component of our account of Holmberg’s Generalization and
of Inverse Holmberg E¤ects.
5
5
Our account of Holmberg’s Generalization has significant properties in common with
earlier work that views the phenomenon as a consequence of ordering generalizations
that constrain the independent working of syntactic operations, e.g. Sells (2001),
Williams (2003); Mu
¨
ller (2000); Kathol (2000). Our proposal di¤ers from these works
in a number of respects – most crucially in the fact that linearization is established on a
phase-by-phase basis. This property of our proposal yields a structural distinction be-
tween aspects of linear order that may be a¤ected by a syntactic operation and those
Cyclic Linearization of Syntactic Structure 5