Available via license: CC BY-NC
Content may be subject to copyright.
Abstract
In this brief article, I point out the predicted existence of a second mode of movement (along-
side Internal Set-Merge) that should be sanctioned under Chomsky’s Strong Minimalist Thesis,
namely Internal Pair-Merge, essentially yielding ‘movement by adjunction’. I then suggest one
possible area in which the computational system makes use of this mode of movement – name-
ly, it allows us to implement Agree-less Move to phase heads (aka ‘pure EPP-movement’, or
Chomsky’s (2007, 2008) Edge-Feature movement) without compromising the Activity Condition,
thus potentially yielding a truly minimalist, narrow-syntactic analysis of optional, ‘discourse-dri-
ven’, ‘stylistic’ movement operations like topicalization and focus-movement. Not only does
Internal Pair-Merge emerge as a theoretical possibility implied by the SMT (section 1), but it is
also a device that is fully exploited by the computational system (section 2).
Key words: Merge, Internal Merge, Set-Merge, Pair-Merge; optional movement, EPP, Edge
Feature; Activity Condition; Strong Minimalist Thesis.
1. What emerges from Merge?
As first discussed within a minimalist context in Chomsky 1995 (henceforth MP),
an operation that forms larger units out of those already constructed is conceptually
necessary for any recursive system of hierarchic discrete infinity. This operation,
which Chomsky dubs Merge, can in that sense be said to “come for free”; its ini-
tial formulation is given in (1).
* I would like to thank the audience at the 5th GLOW in Asia conference, held at Jawaharlal Nehru
University (New Delhi) in October 2005, for their stimulating comments and questions on this
material, as well as an anonymous reviewer for helpful suggestions.
Catalan Journal of Linguistics 8, 2009 55-73
Internal Pair-Merge:
the Missing Mode of Movement
*
Marc Richards
University of Leipzig. Institute of Linguistics
Beethovenstr. 15, 04107 Leipzig. Germany
richards@uni-leipzig.de
Table of Contents
1. What emerges from Merge?
2. IPM as Agree-less Move
3. Conclusions
References
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 55
(1) Merge(α, β) → K = {H(K), {α, β}}
This states that merger of two syntactic objects creates an unordered set, K, of
those two items, and that one of the two members is privileged in the sense that it
provides the label for the larger object K thus created, the label being a head. Given
Merge, the operation Move is recast as Merge of a copy (of which various inter-
pretations have been offered in the subsequent literature, including remerge, mul-
tidominance, and copy-plus-deletion algorithms). As such, the earlier distinction
between substitution and adjunction as fundamentally different operations, fun-
damental to GB theories of movement, largely disappears – all we have is Merge.
Instead, the (presumably real) distinction between specifiers/complements and
adjuncts is captured via the label, as in (2).
(2) a. ‘Substitution’ (projection, formation of “a new category” [MP:248])
Merge(α, K) → L = {H(K), {α, K}}
b. ‘Adjunction’ (formation of a multi-segment category)
Merge(α, K) → L = {<H(K), H(K)>, {α, K}}
Only in (2a), substitution, is the label identical to the head of K; in the case of
(2b), adjunction, the label is an ordered pair. As a technical consequence of this
view of XP-adjunction as forming a multi-segment category, YP adjoined to XP
is excluded from the checking domain of X, and thus cannot be subject to Last
Resort (since feature-checking cannot be at stake in its triggering). Therefore,
Chomsky (MP:324ff.) suggests that XP-adjunction might be relegated to a post-
syntactic “stylistic” component instead of forming part of the computation of LF (i.e.
narrow syntax). In short, the existence of syntactic XP-adjunction is rejected (though
see Kidwai 2000 for discussion and an alternative view which reconciles XP-adjunc-
tion with the morphosyntax), and the phenomena described in terms of XP-adjunc-
tion (such as scrambling, topicalization, extraposition, right-node raising, etc.)
belong to PF or some such external or interface component. In this way, the option-
ality and semantic vacuity of these operations can be kept apart from core syntac-
tic feature-driven movement, such as Move-DP (Case-driven A-movement), Move-
wh (A-bar operator movement) and the like.
The picture changes again, however, in Chomsky 2000 (henceforth MI), in
which a basic, structural, syntactic difference is revived between substitution and
adjunction, now reconceived as Set-Merge (SM) and Pair-Merge (PM), respec-
tively. Whereas SM is a symmetrical operation creating a binary unordered set,
PM is asymmetric, its output an ordered pair:
(3) a. Set-Merge (‘substitution’)
Merge(α, β) → {α, β} = ‘symmetric’
b. Pair-Merge (‘adjunction’)
Merge(α, β) → <α, β> = ‘asymmetric’ (α adjoins to β)
56 CatJL 8, 2009 Marc Richards
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 56
As further elaborated in Chomsky 2004 (henceforth BEA), PM is required in
order to accommodate adjuncts (nonargument modifiers) into the structure, yield-
ing predicate composition in the semantic component, SEM. If the need to express
this relation is imposed by the interface with SEM, then this requirement has the sta-
tus of an interface condition; any minimal device that affords its satisfaction can
then be said to follow from (or at least conform to) the Strong Minimalist Thesis,
SMT (see BEA, Chomsky 2007, Chomsky 2008, for discussion of this point). In that
sense, both SM and PM are compatible with the SMT. However, (XP-)adjunction,
as a distinct structure-building movement operation, remains excluded, as things
stand.
Alongside the Set vs. Pair dimension of Merge, there is a more fundamental
distinction in the manner in which Merge may apply. Up through Chomsky 2001
(henceforth DbP), Move is regarded as an ‘imperfection’, something beyond the
conceptually necessary, as its absence from other symbolic systems would seem
to confirm. That is, Move seems unique to human language and, as such, would
appear to be an “unexplained property of UG” in the sense of Chomsky 2008. In
BEA, however, Chomsky offers a reconceptualization of Move that removes this con-
ceptual hurdle. Now reconceived as Internal Merge (IM), Move is simply one of the
two logical possibilities for merger of α and β: either α (merged to β) is separate
from β, or else α is a subpart of β:
(4) Two logical possiblities for Merge(α, β) → {α, β}
a. Subcase 1: α is distinct from/outside of β→External Merge (EM)
b. Subcase 2: α is part of/contained inside β → Internal Merge (IM)
In the latter case, we have an operation, IM, which yields the displacement
property. Movement, qua internal merge, is thus an entailment of the SMT: it would
have to be stipulated not to exist (and any such stipulation would be a departure
from SMT), as asserted by Chomsky in the following quote:
(5) “[D]isplacement is not an ‘imperfection’ of language; its absence would be
an imperfection.” (BEA: 8)
As a result of this shift in perspective, both (4a)/EM and (4b)/IM come equal-
ly free, as summed up in (6).
(6) EM and IM both conform to SMT.
We have therefore arrived at a system in which Merge comes in three ‘types’,
each with distinct interpretive properties at SEM (each thus meeting a distinct inter-
face need). Firstly, there is EM, yielding argument structure (“deep”/“base” seman-
tics); then there is IM, yielding discourse/information structure and scope/binding
relations (what Chomsky terms “surface” or “edge” semantics); then, finally, there
is PM, yielding adjuncts (predicate modification and composition). The former
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 57
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 57
two (EM vs. IM) stand in an opposition that corresponds to argument vs. discourse
structure – the “duality of semantics” that Chomsky views as a requirement of
SEM (Chomsky MI, BEA, 2008). If IM is free, then it is expected (under SMT)
that the language system will make use of the free device to implement and satis-
fy this interface need (rather than innovating any new or special devices). Standing
apart from these is PM: whereas EM and IM create unordered sets (cf. (4)), PM
creates ordered pairs.
However, as set out in (7), there is a striking asymmetry in this system, and an
unexpected one if EM and IM really are just a single operation (Merge). The set/pair
dichotomy applies only to the external dimension of Merge. Thus only three of the
four logical possibilities implied by the two dimensions Set vs. Pair and External
vs. Internal are part of the current theory:
(7)
Unless we make stipulations to the contrary, the combination of the SM−PM and
EM−IM dimensions of Merge should actually yield a symmetrical typology under
SMT. That is, IM should also be split along the set/pair dimension so that, in addi-
tion to two kinds of EM, there are also two kinds of IM (movement) implied by
the system.
(8)
A fourth formal possibility thus also ‘comes free’, namely Internal Pair-Merge
(IPM).
(9) ESM, ISM, EPM and IPM all conform to SMT.
Set-Merge Pair-Merge
External
ESM EPM
Internal
ISM
IPM
Set-Merge Pair-Merge
External
EM PM
Internal
IM
58 CatJL 8, 2009 Marc Richards
1. What IPM is not, then, is I(S)M of a pair-merged unit. The latter is discussed in BEA (pp.15-18)
in the context of a cyclic implementation of anti-reconstruction effects with adjuncts (Freidin/Lebeaux
effects):
(i) [
wh
Which <[
β
picture of Bill
i
], [
α
that John
j
liked]>] did he
j/*i
buy t
wh
In (i), the adjunct/modifier relative clause that Bill liked is pair-merged cyclically in the base
position of the wh-phrase, creating a PM unit <α, β> that is then wh-moved in the usual way (i.e.
ISM) to spec-CP. The lack of reconstruction follows on the assumption that α (containing John)
adjoined to β is “on a separate plane” (i.e. not part of the “simple”, set-merged structure), so that
he can only c-command β and not α – thus α in the lower copy of <α, β> is ‘invisible’ for the
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 58
The predicted existence of this fourth mode of Merge – that is, Merge that
yields the ordered pair <α, β> wherein α adjoined to β is part of β
1
– seems to
have gone unnoticed in the literature, but there is nothing in the current theory that
rules it out (technically or conceptually – see section 2.3). With this main point
made, the question that arises is empirical – does the system actually make use of
this formal option? I would like to suggest that it does, and indeed quite exten-
sively. Let us therefore briefly consider how IPM might be expected to work, what
its properties might be, and thus what kinds of phenomena we might usefully
describe in terms of this new formal mechanism.
2. IPM as Agree-less Move
The first possibility that suggests itself as perhaps the most obvious potential appli-
cation of IPM is that it might provide us with a minimalist reworking of syntactic
XP-adjunction, with which the parallels are clear – both, essentially, implement
move-by-adjunction. That is, XP-adjunction, qua IPM, now emerges from Merge
as a natural, necessary component of narrow syntax, contra the concerns raised by
Chomsky in MP (see previous section). Since IPM is simply PM applied internal-
ly, we might expect it to share with PM the property of applying completely freely
(cf. MI:134). As such, IPM would seem to provide just the analytical tool we need
in order to accommodate within narrow syntax the kind of truly optional, ‘dis-
course-driven’ movement operations previously analysed in terms of (X/XP-)adjunc-
tion and/or relegated to PF (the ‘stylistic component’). As noted in the previous
section, these might include topicalization (10a), long-distance (A-bar) scrambling
(10b), and focus-movement (10c); all three share the property of being semanti-
cally vacuous in the sense that they are ‘undone’ at SEM/LF, undergoing radical
reconstruction and thus failing to feed binding. (Other possibilities for optional
and/or vacuous adjunction-type movements that may be analysed in terms of IPM
include head movement, cf. DbP:37ff., and successive-cyclic, ‘escape-hatch’ move-
ment, cf. the Edge Feature (EF) movement to phase edges of Chomsky 2008. See
also section 2.3 below and footnote 11 for more on EF-movement and the EF−IPM
connection.)
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 59
purposes of binding/reconstruction (though see Chametzky 2003:206 for criticism of the reason-
ing behind this). Whether we accept this anti-reconstruction story or not, the point to be made here
is that the movement operation in (i) is not IPM; rather, it is ISM of an EPM object. IPM means that
the derived copy/position, not the base copy/position, is created by PM.
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 59
(10) ‘Semantically vacuous’ optional movement
[a = English; b = Japanese; c = Dutch]
a. (i) Herself
i
Mary
i
has always liked t
i
(ii) *Mary
i
, herself
i
has always liked t
i
(Thráinsson 2001:176)
b. (i) Nani-o
i
John-ga [Mary-ga e
i
katta ka] sitteiru
2
what-
ACC
John-
NOM
Mary-
NOM
bought Q knows
‘John knows what Mary bought.’ (Boškovic´ & Takahashi
1998:353)
(ii) [Hotondo-no uta-o]
i
dareka-ga saiwaini t
i
utatta
most-
GEN
song-
ACC
someone-
NOM
fortunately sang
‘Someone fortunately sang most of the songs.’
some >> most, *most >> some (Miyagawa 2003:188)
(iii)*Karera-o
i
[otagai-no sensei-ga [Hanako-ga t
i
hihansita
them-
ACC
each.other-
GEN
teacher-
NOM
Hanako-
NOM
criticized
to] itta]
that said
*‘Each other’s teachers said that Hanako criticized them.’
(Kitahara 2002:179)
c. ...dat [zo’n foto van haarzelf
i
]
j
zelfs Jan deze actrice
i
niet graag t
j
that such picture of herself even Jan this actress not gladly
toont
shows
(Neeleman 1994:399)
IPM, like its XP-adjunction predecessor, would thus be a ‘costless’ operation
unconstrained by Last Resort (LR), hence truly optional and semantically inert
(see, amongst many others, Saito 1989, Fukui 1993, Saito & Fukui 1998, Takano
1998 on the LR-evading nature of XP-adjunction).
Nevertheless, such true optionality sits uneasily with current minimalist prin-
ciples of interface economy, such that we might want to think twice before allow-
ing IPM to apply in the absence of a local, featural trigger (i.e. in violation of LR).
For one thing, the movements in (10) are not, in fact, entirely semantically vacuous.
Whilst it may be true that they do not affect logico-semantic interpretation (bind-
ing, reference, truth and the like), they clearly do have an effect on discourse struc-
ture (that is, the information packaging of the structure into theme/rheme, topic/focus,
new/old, etc.). It is therefore arguable that there is no such thing as truly optional,
vacuous movement: all ‘optional’ movement has some effect on interpretation at
the interface, whether it be the scope/binding or the discourse/informational kind
60 CatJL 8, 2009 Marc Richards
2. Radical reconstruction is observed here in that the wh-phrase nani-o can only be interpreted as
taking scope over the interrogative, embedded clause (Boškovi´c & Takahashi 1998:353).
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 60
of ‘edge semantics’. Since the operations in (10) yield interpretive outcomes at
SEM, they do in fact conform to minimalist notions of interface economy, the logic
of which is informally defined in (11), and are thus amenable to description in
terms of LR after all. (See DbP:34, BEA:10, Reinhart 1995, Fox 2000 on inter-
face economy, and Kidwai 2000, Bailyn 2001, 2003, Miyagawa 1997, 2003 on the
point that no truly optional movement exists and that all movement conforms to
LR, affecting interpretive outcomes).
(11) Interface economy
All movement must be triggered (= Last Resort);
all triggered movement is obligatory (only the trigger may be optional);
this trigger must be cashed out in terms of extra effects at SEM (= Full
Interpretation: ‘no superfluous symbols’).
Within the Probe-Goal-Agree incarnation of checking theory (MI et seq.), there
is just a single formal trigger for movement. This is the EPP-feature, construed as
an extra requirement of the probe. Agree, yielding valuation of a probe’s features,
may thus take place without concomitant Move insofar as the EPP-feature is option-
al (thus checking and agreement are no longer contingent on movement). In line
with (11), optional EPP-features are motivated insofar as they have an effect on
outcome (DbP: (60)), be it scopal or informational or both. New discourse inter-
pretations such as those arising in (10) are thus the effect (function), and not the
cause (triggering mechanism), of movement in this system, allowing us to dispense
with ‘greedy’, pseudo-semantic features of the topic/focus kind.
The upshot, then, is that even IPM can and must be triggered, namely by an
(optional) EPP-feature, like any other instance of movement. As such, the inter-
face economy principles of Last Resort and Full Interpretation apply to IPM and ISM
alike. It may well be that the movements in (10) involve IPM; however, we can-
not define IPM and distinguish it from ISM on the basis of triggers, since both
ISM and IPM must be triggered and there is, we assume, only one type of trigger:
the EPP-feature. What, then, is the defining property of IPM that allows us to tell
it apart from ISM?
The claim I would like to make is that what yields the IPM/ISM distinction is
not the (presence/type of) trigger, but rather the presence/absence of Agree. That is,
whereas IM of the Set-Merge kind (ISM) always involves a component of Agree (cf.
MI, DbP), IPM is Agree-less (or ‘pure-EPP’) movement of the kind that has been
brought to light by a number of authors in recent years (see below). This claim is
summarized in (12).
(12) a. Move fed by Agree = ISM
b. Move independent of Agree = IPM
The notion that movement must be parasitic on Agree in the Probe-Goal system
is bound up in the Activity Condition on Agree, the requirement that both probe
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 61
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 61
and goal be active (visible) for Agree by virtue of unvalued (uninterpretable) fea-
tures. Probes just are unvalued (φ-)features, whereas goals (interpretable (φ-)sets)
must be rendered visible for Agree by means of activation features (Case, for the
φ-system). In this way, the Activity Condition (AC) captures, directly, the idea that
Case and φ-agreement are two sides of the same coin (cf. Schütze 1997:126, Martin
1999:16, Boeckx 2003, Rezac 2004:Chapter V, and many others) – you cannot
value one without valuing the other. This yields ‘Inverse Case Filter’ effects, i.e.
derivations that crash due to the probe going fatally unvalued when the goal has
already been Case-valued (and thus is no longer active for a second application of
Agree):
(13) a. * It seems [t was told Mary [that Bill is a liar]]
b. * A lot of people seem [t are intelligent]
c. * There seem [that a lot of people are intelligent]
d. * There seems to a strange man [that it is raining outside]
As (13c-d) show, these effects occur irrespective of movement, obtaining equal-
ly with ‘pure’ (long-distance) Agree, and so cannot be reduced to the EPP require-
ment. Therefore, the fact that movement fails in (13a-b) can be taken to follow
from the failure of Agree (as in (13c-d)); that is, movement is impossible without
prior agreement.
In MI and BEA, Chomsky suggests that the Agree component of Move (where
Move is a composite operation comprising Agree + Piedpipe + (Re)Merge) serves
to identify the goal which Move then shifts to satisfy the EPP-feature on the probe.
Move, in short, comprises and is parasitic on Agree:
(14) Move → Agree
At least until the revised system of Chomsky 2008, we therefore have a sys-
tem in which Agree can take place without Move (yielding long-distance agree-
ment effects such as those which obtain in expletive-associate existentials, e.g.
There appears to be a man in the garden) whereas the reverse is not true. Not least,
this view that ‘Move implies Agree’ (i.e. (14)) captures a basic but nontrivial fact
about movement: namely, that whatever moves to spec-Probe is, for the most part,
the same category with which the probe agrees (and not some other random XP) –
thus nominative subjects raise to spec-TP, accusative objects raise to spec-vP (Object
Shift), and wh-phrases raise to interrogative spec-CP, etc. It follows that, for the
implication in (14) to hold, EPP-features cannot be independent, probing features
in their own right; rather, they are simply movement ‘diacritics’ associated with
(φ-)probes (cf. Roberts & Roussou 2003).
However, there is a major problem confronting this view of the Move–Agree
relation (a problem to which IPM, I propose below, offers a simple and effective
solution). As pointed out by Nevins & Anand (2003), there is plenty of crosslin-
62 CatJL 8, 2009 Marc Richards
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 62
guistic evidence that Agree and EPP-satisfaction (movement) are doubly disso-
ciated in many languages. That is, not only can Agree take place without move-
ment, but movement must also be allowed to take place in the absence of Agree.
Thus, as numerous authors have observed in recent years, languages such as
Russian, Japanese and Hindi exhibit structures in which a previously Case-val-
ued argument raises to a vP-external A-position (assumed to be spec-TP) despite
the fact that such an argument should be inactive for Agree (and thus raising) as
a consequence of lacking an unvalued Case feature (see Bailyn 2003, Lavine &
Freidin 2002 on Russian; Miyagawa 2001, 2003, 2004, 2005 on Japanese; Nevins
& Anand 2003 on Hindi). The data set in (15-17), for example, illustrates that
the satisfaction of T’s EPP-feature need not involve an agreed-with nominative
argument.
(15) Russian (Generalized Inversion – Bailyn 2003; adversity impersonals – Lavine
& Freidin 2002)
a. Soldata ranilo pulej
soldier-
ACC
wounded-
3PNS
3
bullet-
INSTR
‘A soldier was wounded by a bullet.’
b. (i) Milicionerov
i
ranilo puljami prinadležaš?imi drug
militiamen-
ACC
wounded-
3PNS
bullets-
INSTR
belonging each
drugu
i
other-
DAT
(ii)*Puljami prinadležaš?imi drug drugu
i
ranilo
bullets-
INSTR
belonging each other-
DAT
wounded-
3PNS
milicionerov
i
militiamen-
ACC
‘Militiamen were wounded by each other’s bullets.’
(L&F 2002:280(42))
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 63
3. 3PNS = 3rd person neuter singular (‘default’ agreement).
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 63
(16) Japanese (Miyagawa 2001, 2003, 2004, 2005; Kitahara 2002)
a. [Hotondo-no uta-o]
i
dareka-ga t
i
utatta
most-
GEN
song-
ACC
someone-
NOM
sang
‘Someone sang most of the songs.’
some >> most, most >> some (Miyagawa 2003:188)
b. (?) Karera-o
i
[otagai-no sensei]-ga t
i
hihansita
4
them-ACC each other-GEN teacher-NOM criticized
*‘Each other’s teachers criticized them.’ (Kitahara 2002:167)
c. ‘Focus agreement’ (Miyagawa 2004, 2005):
(i) Taro-ga nani-o kai-mo sina-kat-ta
Taro-
NOM
what-
ACC
buy-MO do-
NEG-Past
(ii) *Nani-o
i
Taroo-ga t
i
kai-mo sina-kat-ta
5
what-
ACC
Taro-
NOM
buy-MO do-
NEG-Past
‘Taro didn’t buy anything (at all).’
(17) Hindi (Nevins & Anand 2003)
kisii šaayer-ne har ghazal lik
h
ii
some poet-
ERG
every song-
NOM
write-
f.PERF
(some >> every; *every >> some)
‘Some poet wrote every song.’ (Nevins & Anand 2003: 370)
The structures in (15)-(17) share two key properties. Firstly, they all involve
A-movement, i.e. standard raising-to-subject (spec-TP) as in English, as attested
by the fact that these movements feed scope/binding and have no effect on
focus/discourse structure (they are felicitous in ‘out-of-the-blue’, ‘what happened?’
contexts); cf. Lavine & Freidin 2002:270ff, Bailyn 2003:176 (note 8). Secondly,
they all lack morphological (φ-)agreement, indicating an absence of Agree between
T and the raised DP. Indeed, Agree(T, DP) is barred in all of these examples due to
failure to meet the Activity Condition, as follows.
In Russian (15), the previously Case-valued accusative argument raises to spec-
TP despite the fact that it should be inactive for Agree with T (and thus raising) as
a consequence of having already been valued (to accusative) by v. In Hindi (17), the
ergative DP is similarly inactive for Agree with T, assuming ergative either to be lex-
ical/inherent (as in Nevins & Anand 2003) or to be assigned by v (cf. Müller 2004)
and thus lacking an unvalued structural Case feature; further, if T values nomina-
64 CatJL 8, 2009 Marc Richards
4. This sentence is judged acceptable, unlike the similar binding attempt in (10biii). The lack of
reconstruction in all the ‘pure-EPP’ (Agree-less) A-movements in (15)-(17) is captured under
Nevins & Anand’s (2003) PEPPER principle (roughly, P
ure EPP Excludes Reconstruction) – see
their paper for discussion.
5. This example illustrates a lack of reconstruction in that the indeterminate reading of the wh-item
nani is only possible in scope of -mo particle (see Miyagawa 2004, 2005).
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 64
tive on the unraised object, T too should be inactive and thus unable to initiate fur-
ther probing, thus the AC is again violated here. Finally, in Japanese (16), T has
already agreed with the subject, valuing the (φ-)probe on T and nominative on the
DP. Not only should the accusative DP be inactive for the movement-feeding
Agree(T, DP) operation, as in (15), but T should itself be inactive for a second
instance of Agree (and thus raising), as in (17).
6
Again, these structures should not
be possible if the movements in question are subject to the AC.
Clearly, the strong position that Move implies Agree (cf. (14)) cannot be main-
tained: If Agree is subject to the AC, then raising to spec-TP in (15)-(17) cannot
involve Agree and must therefore be instances of Move without Agree. There is
little question, then, that we have to accommodate Agree-less movement into the
computational system.
7
The question is how best to do this. Ideally, we want to
achieve the necessary weakening of (14) (‘Move implies Agree’) without exclud-
ing the familiar kind of movement that is parasitic on Probe-Goal Agree and that
is associated with morphological agreement (cf. (13)). That is, we have to recon-
cile (15)-(17) with the AC.
There are two main approaches to achieving this aim in the literature. The first
is to weaken the AC so that inactive elements are visible to probes for certain oper-
ations (specifically, for Match, a subcomponent of Agree). The second is to aban-
don the AC altogether, so that items remain permanently available for Move/Agree
(there is thus no active–inactive distinction). Both approaches are problematic. Let
us briefly consider each of these possibilities in turn, before suggesting IPM as a less
problematic alternative in section 2.3.
2.1. Match-driven Move (Kitahara 2002, Boeckx 2003)
According to the first approach, the AC is a condition only on Agree (i.e. valua-
tion). Probe-Goal can still occur without resultant Agree/valuation whenever the
AC doesn’t hold: this is simply Match. Match alone is then claimed to be suffi-
cient for feeding Move.
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 65
6. The Japanese derivation sketched here follows the analysis in Kitahara 2002. Alternatively, Miyagawa
2004, 2005 proposes ‘focus agreement’ in topic-prominent languages like Japanese, so that T has
a Focus probe instead of a φ/Agree probe in such languages.
7. As a reviewer notes, this is a long-standing observation, going back at least to Belletti & Rizzi’s
(1988) analysis of Romance psych-verb construcions, in which non-nominative experiencers raise
to the canonical subject position (spec-TP, in our terms) despite a lack of agreement with the finite
verb (which agrees with the internal argument). These and other plausible cases of Agree-less
movement to the subject position, such as locative inversion and certain impersonal constructions,
may well provide additional instances of IPM. It is worth clarifying, in this connection, that my
claim here is not that Agree-less movement is a previously unnoticed phenomenon in the litera-
ture (the ‘missing mode of movement’ of the title); rather, I am claiming that such movements are
plausibly characterized as instantiations of a previously unnoted (‘missing’) formal possibility,
namely IPM. That is, I am claiming only to have identified the formal mechanism behind these
movements. This mechanism then allows Agree-less movement to be naturally accommodated
within the minimalist architecture, something which has so far proven problematic (for the rea-
sons given in the following subsections).
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 65
Match-driven Move is thus premised on a fundamental distinction between
Match and Agree. This is the same distinction that underlies Chomsky’s (MI:123)
notion of defective intervention (DI), as defined in (18): it is precisely on the
assumption that inactive DPs are visible to probes for Match (but not for Agree)
that we expect unvalued features to intervene for Agree without themselves being
a potential goal for Agree (i.e. without themselves being able to value the probe
in question), yielding DI effects.
(18) Defective intervention (MI:123 (42), 129; DbP; Chomsky 2008; Rezac 2004;
Hiraiwa 2005)
In structure α > β > γ, where > is c-command and β and γ both match probe
α, inactive β blocks matching (and thus Agree) between α and γ.
Insofar as DI is a real phenomenon, we have independent empirical evidence
for the Match-Agree distinction and thus for the weakening of the AC that this
approach to (15)-(17) entails. In fact, Chomsky offers just three empirical arguments
for DI: it yields superraising and (certain) MLC effects (cf MI:128-9 (47)); it yields
the lack of long-distance nominative agreement across in-situ quirky subjects in
Icelandic (cf. MI:130-1 (51)); and it provides an empirical argument for the claim that
T (unlike C and v*) is not a phase head (see BEA:21-2, Chomsky 2008: 152 for
details). The validity of the Match/Agree distinction thus hinges on just these three
cases. However, as discussed in Richards 2008, all three have simpler, arguably
more plausible explanations (in particular the Phase Impenetrability Condition,
PIC). For reasons of space, I will not reproduce the relevant counterarguments here;
instead, let us focus on a possible conceptual argument for DI. It could be argued,
to this end, that Probe-Goal should apply freely (cf. BEA:13), and that only (valued)
φ-sets should matter to φ-probes (that is, Case features should be irrelevant); see
Rezac 2004 for such an argument. If correct, then intervention by inactive φ-sets
(i.e. Match/DI) is the null assumption: it could only be excluded by stipulation.
Alternatively, however, we could just as well argue against DI on conceptual
grounds. Ideally, either a feature is visible as a potential goal (i.e. it is active) or it
is not; any further distinctions (such as ‘visible for Match but not for Agree/Value’,
‘visible for Match and Move but not for Agree/Value’, both of which are invoked
in Chomsky’s three empirical cases mentioned above) are surely then a departure
from optimality and the SMT. On such a view, only active elements should be vis-
ible to probes and thus act as interveners; this, in turn, enables the AC and Case
features to do the job they were originally designed to do. After all, if active and inac-
tive DPs alike are visible to probes, then why would the system have Case/activation
features (and the AC) in the first place? Why not dispense with them altogether
(as in Nevins 2004)? From the minimalist perspective, there seems to be no rea-
son for rendering DPs active, via Case features, if they are visible anyway; there-
fore, if DI is real, then Case is a departure from the SMT, serving no purpose from
the perspective of the interfaces. Recall from above that such a rationale for Case/AC
is, in fact, offered by Chomsky (see DbP (4:(2c)), BEA (14:(10ii))) – Case serves
to identify and determine goals, ensuring convergence by rendering goals visible to
66 CatJL 8, 2009 Marc Richards
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 66
probes (which could not otherwise be valued, violating Full Interpretation). However,
it is precisely this rationale that Match/DI weakens – goals (interpretable φ-sets)
are visible to probes anyway, whether they have unvalued Case features or not.
Match/DI thus deprives us of a conceptually necessary role for Case.
Such a role for Case can only be maintained if, instead of DI, the null assump-
tion is actually as in (19), which elevates the AC to the status of an interface con-
dition.
(19) Feature visibility (syntax) / ‘Strong Activity Condition’
Only unvalued features are visible to the syntax (Probe-Goal etc.)
The tenability of (19) is of course open to empirical scrutiny. From the con-
ceptual perspective, however, it allows for a fully lookahead-free syntax that oper-
ates exclusively on uninterpretable (unvalued) formal features, as seems desirable
if the syntax can neither know nor care about matters of (LF-)interpretability.
Interpretable/valued feature sets should, on such grounds, be invisible to the syn-
tax, indistinguishable from pure semantic and phonological features and therefore
equally irrelevant to the syntactic computation – this is arguably the null assump-
tion that (19) captures. If (19) indeed holds, then it now follows that Agree is impos-
sible unless we have a feature that renders interpretable φ-sets visible to the syntax.
Such a feature (Case) now simply follows from the SMT, and specifically from the
need to ensure that Full Interpretation is met (at SEM); goal-activeness is thereby
deduced.
As stated above, this line of reasoning is denied if we assume DI and the
Match/Agree distinction. Case and the AC only make sense if there is no DI/Match,
i.e. if (19) holds. If we accept this line of argumentation, then (15)-(17) cannot be
analysed as Match-driven Move.
2.2. Pure-EPP movement (Nevins 2004; also Lavine & Freidin 2002,
Nevins & Anand 2003, Chomsky 2005)
Another possibility is to dispense with the AC and/or Case altogether (rather than
with Match/DI). Pursuing this strategy, Nevins (2004) argues that the movements
in (15)-(17) obtain purely for the EPP-feature (i.e. they do not involve Agree and
the valuation of uninterpretable features), and that the existence of such pure EPP-
movement denies the AC, which he defines as follows:
(20) Activity Condition (version in Nevins 2004)
Inactive elements are not accessible for further operations
[Nevins 2004:9 (29)]
Crucially, for Nevins, “further operations” encompasses Move as well as Agree.
Thus (20) bars movement of inactive elements, rendering this version of the AC
logically incompatible with pure EPP-movement. It thus follows that the activity con-
dition in the form of (20) cannot hold and must be eliminated.
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 67
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 67
Whilst this is an interesting and carefully reasoned conclusion, and whilst
Nevins’s proposal that the movements in question (cf. (15)-(17)) are purely EPP-
driven is surely the correct one, I would contest the other premise on which the
argument is based, namely (20). As we saw in our discussion of (13), the AC for
Chomsky is a condition only on Agree, not on Move (it thereby captures the Case-
Agreement connection). Only derivatively does the AC constrain Move, insofar as
Move is parasitic on Agree (i.e. insofar as (14) holds). Since the AC applies only
to Agree, the existence of pure EPP-movement of inactive elements is logically
incompatible only with (14) (i.e. the claim that Move implies Agree); it is not,
however, incompatible with the AC per se (defined as a condition on probes and
goals, i.e. on Agree). All that follows is that pure EPP-movement cannot involve
Agree (thus the AC holds vacuously in these cases), which is exactly the desired
result given the lack of Agree(ment) in (15)-(17), as discussed above. Therefore, only
the version of the AC in (20) must be rejected; there is no need to eliminate or
reject the AC on Agree.
Indeed, the desired conclusion that pure EPP-movement cannot involve Agree
(contra (14)) only follows if we maintain the AC on Agree. That is, the lack of
morphological agreement with the movements in (15)-(17) is unexpected if the
AC is renounced. This is because syntactic elements would remain, in effect, per-
manently active and thus able to enter into multiple, indeed unlimited, Agree oper-
ations. As we saw in (13), Inverse Case Filter effects attest that this cannot be cor-
rect.
8
The key property of pure EPP-movement, namely that it is Agree-less
movement, is exactly what the AC in fact predicts and ensures.
In sum, we do not need (or want) to abandon the AC in order to allow for pure
EPP-driven movement. Such an abandonment would only be necessary if move-
ment of inactive elements did involve Agree(ment). Since it doesn’t (cf. (15)-(17)),
pure EPP-movement tells us nothing about the AC as a condition on Agree.
2.3. A third possibility: IPM
The movement phenomena in (15)-(17) show us that we have to allow for Move
without Agree. That is, pure EPP-movement is an inescapable empirical fact.
Numerous questions remain, however, as to whether and how pure EPP-movement
can be implemented in the grammar in a manner conforming to the SMT. Clearly,
the EPP-feature cannot operate as a probe in such constructions. If inactive ele-
ments are invisible to probes (i.e. if we renounce DI and maintain (19)/AC, as pro-
posed above), then EPP-probes cannot exist, precisely because they do not seek
68 CatJL 8, 2009 Marc Richards
8. Other mechanisms can be invoked to reinstate ICF effects in an AC-less system. Thus Nevins pos-
tulates a “Single Case Constraint” to the effect that “a DP that is valued with more than one case
feature is illegible to PF”. However, this only captures one half of the AC, as probes, too, cannot
be fully valued more than once under the AC (hence ‘Case Filter’ effects, e.g. *It seems John to
be happy). Further, the very existence of Case features in the first place remains something extra
(i.e. unexplained from the SMT perspective) without (19), and the latter cannot be maintained on
the AC-less approach.
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 68
active goals (contra the Activity Condition on Probe-Goal relations); therefore,
Agree-free movement cannot involve Probe-Goal Agree.
9
What, then, is the for-
mal relation between EPP-features and the inactive elements that move to satisfy
them in (10)/(15)-(17)?
It seems clear that this relation cannot be one of Set-Merge (SM). If only active
heads can project, and SM implies projection of the probe/head (cf. (2a)), then
pure EPP-movement (which involves inactive heads) cannot involve Set-Merge.
Yet movement, as currently conceived (cf. internal merge in (7)), is Set-Merge.
This would seem to exclude Agree-free movement from the theory. The challenge,
then, is to come up with a formally precise mechanism that allows Agree-free
movement in a principled and restrictive manner. I would like to submit that, once
we recognize that the internal/external dimension should apply to Pair-Merge as
well as to Set-Merge (i.e. the symmetrical typology in (8)), exactly such a mecha-
nism emerges: IPM.
Firstly, IPM does not involve projection of the head (cf. (2b): XP-adjunction
creates a multi-segment category rather than a change of category). Inactive heads
– those that lack Agree probes, as manifested by default agreement in Russian (15),
or whose probe is already valued, such as the adjunction sites in (10) – may thus
freely undergo IPM to satisfy EPP-features. Further, since only inactive items may
undergo IPM, it follows that IPM must occur after ISM (i.e. after Agree-driven
movement), implying an intrinsic ordering between the two merge types.
10
IPM thus allows Agree-free and Agree-constrained movement to harmonious-
ly co-exist, as desired:
(21) a. IM with Agree ([+ projection]) = ISM
… optionally followed by…
b. IM without Agree ([− projection]) = IPM
In short, we can have our cake and eat it: IPM allows us to abandon the idea
that Agree is involved in (15)-(17) (hence the lack of morphological agreement)
without weakening or abandoning the AC (and with it the simplest explanation of
Case Filter and Inverse Case Filter effects in terms of activation features, and the
SMT-conforming rationale for the latter in terms of (19)).
There is at least one further useful application of IPM as Agree-less Move
worth mentioning here, and that is that it offers a means of describing (and imple-
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 69
9. This ‘probe’ would be unique in a further respect, namely in being able to be satisfied/valued via
EM (i.e. via Merge of an expletive) rather than via Agree(/IM). This expletive-related anomaly is
a familiar and long-standing problem that has appeared in many guises throughout the develop-
ment of minimalist checking theory; see, e.g., MP:311-2 for the strong-feature precursor.
10. This is, of course, in line with earlier analyses of the relation between A-movement, A-bar move-
ment and interpretation, in which binding theory was taken to apply after the former but before
the latter (hence the reconstruction property of A-bar movement) – see, for one version of this, the
‘NP-structure’ of Williams 1994.
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 69
menting) successive-cyclic movement through phase edges under the PIC. Such
movements are notoriously problematic in terms of Probe-Goal Agree: not only
do they involve lookahead to avoid crash of goal/activation-features at the lower
phase level (cf. Chomsky 2008, Franks 2005 and others), but there is simply no
evidence for Agree at the relevant intermediate stages (cf. Boeckx 2003, Boškovi?
2004). Again, these problems are solved in a maximally simple way if no Agree
is involved at the intermediate phase-levels, i.e. if this is just another case of move-
ment purely for EPP (viz. IPM).
11
We conclude, then, that Agree-less, pure-EPP movement is Internal Pair-Merge.
3. Conclusions
A (re)consideration of the status of Merge in current minimalist theorizing has
led to the following straightforward conclusions. (i) Movement, qua Internal
Merge, does not have to be Set-Merge; it can also be Pair-Merge – that is, Internal
Pair-Merge is predicted to exist. (ii) IPM (we have argued) is best regarded as
characterizing Agree-less movement (i.e. pure EPP-movement), a kind of move-
ment long recognized in the literature; this coexists with movement fed by Agree,
which is constrained by the Activity Condition – that is, Internal Set-Merge. (iii)
IPM is fully exploited by the computational system, offering a simple and trans-
parent analysis of a wide range of phenomena, including: pure EPP-movement of
inactive elements in Russian, Hindi and Japanese; ‘vacuous’ A-bar operations
such as topicalization and long-distance scrambling; successive-cyclic move-
ment through intermediate phase edges; and, plausibly (though not explored
here), an alternative Agree-free approach to topic/focus-prominence (cf.
Miyagawa’s (2004, 2005) “focus-agreement”; see footnote 6) – and maybe also
70 CatJL 8, 2009 Marc Richards
11. Chomsky 2008 (p. 151), too, suggests that we solve the problem of intermediate phase-edge move-
ment by treating such movement as driven by the Edge Feature (EF) of the phase head rather than
by Agree. This raises the question of to what extent IPM, as proposed here, differs from EF-move-
ment as proposed by Chomsky in that paper. There are several substantive differences, though the
two are by no means incompatible. Firstly, EF (for Chomsky) is a movement trigger (akin to the
EPP-feature of previous works), yielding A-bar positions, rather than a type of movement per se.
IPM may well, therefore, equate with that kind of movement that is triggered by EF, without redun-
dancy (it is a nontrivial claim that EF triggers IPM rather than ISM). On the other hand, EF-move-
ment and IPM do not denote the same operations: the former, for Chomsky, yields A-bar move-
ment to the phase edge (thus contrasting with Agree-driven movement to phase-internal positions,
which is A-movement), yet IPM clearly instantiates both A- (cf. (15)-(17)) and A-bar (cf. (10))
types of movement. The A-IPM cases in (15)-(17), then, would be unlikely candidates for EF-
movement in Chomsky’s system; further, there are cases of A-bar movement that might be better
analysed as involving probe-driven, Agree-type movement (such as matrix wh-movement; see
Rizzi 2006 on “criterial freezing”: essentially (in our terms) AC applied to the A-bar system), and
which would therefore be ISM on the analysis presented here but (presumably) EF-movement for
Chomsky. In short, there is both (i) IPM that corresponds to Agree- and not to EF-movement, and
(ii) EF-movement that corresponds to ISM and not to IPM. (It is worth noting, tangentially, that
Chomsky 2007 proposes that EF is unerasable, hence the availability of multiple specifiers. If so,
then it is not even clear that EF is sufficient any longer as a movement trigger, since movement
does not result in feature deletion here. I leave this issue for consideration in future work.)
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 70
(for future research) head movement, locative inversion, experiencer raising (cf.
footnote 7), Icelandic stylistic fronting (Dennis Ott, p.c.), and so forth. (iv) IPM
is optional, insofar as its trigger (EPP-feature) is optional. (v) IPM can poten-
tially bring under the fold of ‘optimal design’ many of the robust, pervasive phe-
nomena that have previously been analysed in terms of optional XP-adjunction
but which have so far eluded minimalist analysis on grounds of the AC, inter-
face economy and Last Resort.
In sum, XP-adjunction in the form of IPM as a second mode of syntactic move-
ment would, to borrow Chomsky’s refrain, have to be stipulated not to exist.
References
Bailyn, John (2001). “On Scrambling: A Reply to Boškovi? & Takahashi”. Linguistic
Inquiry 32: 635-658.
Bailyn, John. (2003). “Does Russian Scrambling Exist?” In: Karimi, S. (ed.). Word
Order and Scrambling. Oxford: Blackwell, pp. 156-176.
Belletti, Adriana; Rizzi, Luigi (1988). “Psych-verbs and Theta Theory”. Natural
Language and Linguistic Theory 6: 291-352.
Boeckx, Cedric (2003). Islands and Chains: Resumption as Stranding. Amsterdam:
Benjamins.
Boškovi?, Željko. (2004). “On the Locality of Move and Agree: Eliminating the
Activation Condition, Generalized EPP, Inverse Case Filter, and Phase-
Impenetrability Condition.” University of Connecticut, manuscript.
Boškovi?, Željko; Takahashi, Daiko (1998). “Scrambling and Last Resort”. Linguistic
Inquiry 29: 347-366.
Chametzky, Robert (2003). “Phrase Structure”. In: Hendrick, R. (ed.) Minimalist Syntax.
Oxford: Blackwell, pp. 192-225.
Chomsky, Noam (1995). The Minimalist Program. Cambridge, Mass: MIT Press.
Chomsky, Noam (2000). “Minimalist Inquiries: the Framework”. In: Martin, R.; Michaels,
D.; Uriagereka, J. (eds). Step by step. Cambridge, Mass.: MIT Press, pp. 89-156.
Chomsky, Noam (2001). “Derivation by Phase”. In: Kenstowicz, M. (ed.). Ken Hale: A
Life in Language. Cambridge, Mass.: MIT Press, pp. 1-52.
Chomsky, Noam (2004). “Beyond Explanatory Adequacy”. In: Belletti, A. (ed.).
Structures and Beyond. The Cartography of Syntactic Structures (volume 3). Oxford:
Oxford University Press.
Chomsky, Noam (2007). “Approaching UG from below”. In: Interfaces + Recursion
= Language? Chomsky minimalism and the view from Syntax-semantics,
V. Sauerland and H-M. Gärtner (eds.), Berlin: Mouton de Gruyter, pp. 1-30.
Chomsky, Noam (2008). “On Phases”. In: Foundational Issues in Linguistic Theory.
Essay in Honor of Jean-Roger Vergnaud, C. Otero et al. (eds.), Cambridge, Mass:
MIT Press, pp. 134-166.
Fox, Daniel (2000) Economy and Semantic Interpretation. Cambridge, Mass: MIT Press.
Franks, Steven (2005). “Agnostic Movement”. Abstract for poster presented at NELS 36.
Fukui, Naoki (1993). “Parameters and Optionality”. Linguistic Inquiry 24: 399-420.
Hiraiwa, Ken (2005). Dimensions of Symmetry in Syntax: Agreement and Clausal
Architecture. MIT, doctoral dissertation.
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 71
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 71
Kidwai, Ayesha (2000). XP-Adjunction in Universal Grammar. Scrambling and Binding
in Hindi-Urdu. Oxford/New York: Oxford University Press.
Kitahara, Hisatsugu (2002). “Scrambling, Case, and Interpretability”. In: Epstein, S.
D.; Seely, T. D. (eds.). Derivation and Explanation in the Minimalist Program.
Oxford: Blackwell, pp. 167-183.
Lavine, James S.; Freidin, Robert (2002). “The Subject of Defective T(ense) in Slavic”.
Journal of Slavic Linguistics 10: 253-289.
Martin, Roger (1999). “Case, the Extended Projection Principle, and Minimalism”. In:
Epstein, S. D.; Hornstein, N. (eds.). Working Minimalism. Cambridge, Mass: MIT
Press, pp. 1-25.
Miyagawa, Shigeru (1997). “Against Optional Scrambling”. Linguistic Inquiry 28: 1-25.
Miyagawa, Shigeru (2001). “The EPP, Scrambling, and Wh-in-Situ”. In: Kenstowicz, M.
(ed.). Ken Hale: A Life in Language. Cambridge, Mass.: MIT Press, pp. 293-338.
Miyagawa, Shigeru (2003). “A-Movement Scrambling and Options without
Optionality”. In: Karimi, S. (ed.). Word Order and Scrambling. Oxford: Blackwell,
pp. 177-200.
Miyagawa, Shigeru (2004). “On the EPP”. In: McGinnis, M. & Richards, N. (eds.).
Proceedings of the EPP/Phase Workshop (MITWPL).
Miyagawa, Shigeru (2005). “Unifying Agreement and Agreement-less Languages”. In
Proceedings of WAFL2 (MITWPL).
Müller, Gereon (2004). “Argument Encoding and the Order of Elementary Operations”.
University of Leipzig, manuscript.
Neeleman, Ad. (1994). “Scrambling as a D-structure phenomenon”. In: Corver, N.;
van Riemsdijk, H. (eds.). Studies in Scrambling: Movement and Non-Movement
Approaches to Free Word-Order Phenomena. Berlin: de Gruyter, pp. 387-429.
Nevins, Andrew (2004). “Derivations without the Activity Condition”. In: McGinnis,
M.; Richards, N. (eds.). Proceedings of the EPP/Phase Workshop (MITWPL).
Nevins, Andrew; Anand, Pranav (2003). “Some AGREEment Matters”. In: Garding, G.;
Tsujimura, M. (eds.). WCCFL 22 Procedings. Somerville: Cascadilla, pp. 370-383.
Reinhart, Tanya (1995). “Interface Strategies”. OTS Working Papers in Linguistics.
Utrecht.
Rezac, Milan (2004). Elements of Cyclic Syntax: Agree and Merge. University of
Toronto, doctoral dissertation.
Richards, Marc (2008). “Quirky Expletives”. In: d’Alessandro, R.; Hrafnbjargarson, G.;
Fischer, S. (eds) Agreement Restrictions. Berlin: Mouton de Gruyter, pp. 181-213.
Rizzi, Luigi (2006). “On the Form of Chains: Criterial Positions and ECP Effects”. In:
Cheng, L.; Corver, N. (eds.). Wh-Movement: Moving On. Cambridge, Mass: MIT
Press.
Roberts, Ian; Roussou, Anna (2003). Syntactic Change. A Minimalist Approach to
Grammaticalization. Cambridge: Cambridge University Press.
Saito, Mamoru (1989). “Scrambling as semantically vacuous A’-movement”. In: Baltin,
M.; Kroch, A. (eds.). Alternative Conceptions of Phrase Structure. Chicago:
University of Chicago Press, pp. 182-200.
Saito, Mamoru; Fukui, Naoki (1998). “Order in Phrase Structure and Movement”.
Linguistic Inquiry 29: 439-474.
Schütze, Carson (1997). INFL in child and adult language: Agreement, Case, and
Licensing. MIT, doctoral dissertation.
72 CatJL 8, 2009 Marc Richards
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 72
Takano, Yuji (1998). “Object Shift and Scrambling”. Natural Language and Linguistic
Theory 16: 817-889.
Thráinsson, Höskuldur (2001). “Object Shift and Scrambling”. In: Baltin, M.; Collins,
C. (eds.). The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell,
pp. 148-202.
Williams, Edwin (1994). Thematic Structure in Syntax. Cambridge, Mass: MIT Press.
Internal Pair-Merge: the Missing Mode of Movement CatJL 8, 2009 73
CatJourLing 8 001-189:CatJL 11/1/10 13:18 Página 73
