Phase Theory EGG Summer School, Banja Luka ⋅ 06/08/2018

Class 1: What are phases?

Andrew Murphy
Universität Leipzig
[email protected]

1 What are phases? 2 Classical Phase Theory (Chomsky 2000, 2001)

• The basic architecture of the Minimalist Program (Chomsky 1995) is the Y-Model of grammar: 2.1 Merge-over-Move
(1) ‘Inverted’ Y-Model of Grammar:
• The original motivation for phases in Chomsky (2000) comes from the following paradigm:
Lexicon Numeration {love, Mary, John}

(3) a. [ TP2 A man seems [ TP1 to [ vP be the garden ]]]

b. [ TP2 There seems [ TP1 to [ vP be a man in the garden ]]]
Narrow c. *[ TP2 There seems [ TP1 a man to [ vP be in the garden ]]]
John
Syntax
love Mary (4) Generalization:
If expletive there is to be merged in Spec-TP2 , it must also be merged in Spec-TP1 .
Spell-Out
• This can be captured by the following constraint (also see Frampton & Gutmann 1999):

(5) Merge-over-Move (informal version):

At a given derivational step, if both Move of α and Merge of β are possible, then Merge of
β is preferred.
Logical Phonological
Form Form (6) a. Step Σ:
Numeration: {there, C, T[EPP] , v, seem}
love′ (John, Mary) /dZ6n l2vz me@ri/ [ TP1 to[EPP] [ vP be [ DP a man ] the garden ]]]

• When does Spell-Out happen? b. Step Σ+1 (Move):

Numeration: {there, C, T, v, seem}
– Option 1: We build the entire syntactic structure and then send it to the interfaces. [ TP1 [DP a man ] [ T′ to[EPP] [ vP be in the garden ]]]
– Option 2: We build only part of entire syntactic structure, send that part to the interfaces,
and then continue to build the rest of the structure.
c. Step Σ+1 ′ (Merge):
• The latter is the so-called Multiple Spell-Out model (Uriagereka 1999; also cf. Bresnan 1978).
Numeration: {C, T[EPP] , v, seem}
(2) Numeration: {love, John, Mary} [ TP1 there [ T′ to[EPP] [ vP be [ DP a man ] the garden ]]]
a. Create SO [ love Mary ]
b. Send SO to interfaces (Spell-Out) loves Mary • The preference for Merge-over-Move expressed in (5) means that only (6c) is possible.
c. Continue to build [ John loves Mary ] • When the matrix clause is merged, only movement of there is possible (due to Minimality):
d. Send SO to interfaces John loves Mary

1
 (7) a. [ TP2 T[EPP] [ vP seems [ TP1 there [ T′ to[EPP] [ vP be [ DP a man ] the garden ]]]]] Suppose we select LA as before [. . . ] Suppose further that at each stage of the derivation a
b. [ TP2 There [ T′ T[EPP] [ vP seems [ TP1 [ T′ to[EPP] [ vP be [ DP a man ] the garden ]]]]]] subset LAi is extracted, placed in active memory (the ‘work space’), and submitted to the
procedure L. When LAi is exhausted, the computation may proceed if possible. Or it may
return to LA and extract LAj , proceeding as before.
b.′ *[ TP2 A man [ T′ T[EPP] [ vP seems [ TP1 there [ T′ to[EPP] [ vP be the garden ]]]]]] (Chomsky 2000:106)

(13) a. Numeration: {a, b, c, d, e, f}

b. Numeration with lexical subarrays: {{a, b}, {d, e}, {e, f}}
• If we hadn’t blocked the step in (6b), the undesirable continuation in (8) would be possible:
• Only one lexical subarray is accessible at a time (i.e. lexical access is cyclic), so this reduces the
(8) a. Numeration: {there, C}
computational burden (Ott 2009).
[ TP2 T[EPP] [ vP seems [ TP1 a man [ T′ to[EPP] [ vP be in the garden ]]]]]
b. Numeration: {C} • This also helps us solve our undergeneration problem in (9):
*[TP2 There [ T′ T[EPP] [ vP seems [ TP1 a man [ T′ to[EPP] [ vP be in the garden ]]]]]]
(14) a. There exists [evidence [that a man was in the garden]]
• The following example poses a problem for Merge-over-Move, however (see e.g. Obata 2006): Numeration:
LA1 : {C, exist, there, T[EPP] , v}
(9) a. There exists [evidence [that a man was in the garden]] LA2 : {evidence, that, T[EPP] , v, a man, be, in, the, garden}
b. [Evidence [that there was a man in the garden]] exists b. [Evidence [that there was a man in the garden]] exists
Numeration:
• Both of these sentences have the same numeration: LA1 : {C, exist, T[EPP] , v}
LA2 : {evidence, that, there, T[EPP] , v, a man, be, in, the, garden}
(10) {C, T, there, v, evidence, T[EPP] , exist, that, a, to[EPP] man, v, be, the, in, garden}
• In (14a), there and a man belong to different lexical arrays and therefore do not compete.
• The problem is that, when we build the relative clause for (9a), the Merge-over-Move preference
A phase of a derivation is a syntactic object derived . . . by choice of LAi .
will block movement
(Chomsky 2000:106)
(11) There exists evidence that a man was in the garden Some problems for the MoM argument:
[ TP T[EPP] [ vP be a man in the garden ]]
Numeration: {C, T, there, v, evidence, T[EPP] , exist, that} • If Move and (External) Merge are actually just different modes of application of the same basic
operation merge (Internal vs. External), how can a constraint differentiate them?
a. [ there [ T′ T[EPP] [ vP be a man in the garden ]]] (Merge)
b. *[ TP a man [ T′ T[EPP] [ vP be in the garden ]]] (Move) • Deal (2009:302ff.) points out the too-many-‘there’s-problem. If we had two expletives in the
numeration, then what prohibits (15a)?

(15) a. *[ TP2 there seems [ TP1 there to [ vP be a man in the garden ]]]
• As the derivation proceeds, at TP2 there is no expletive in the numeration that can be merged
b. [ TP2 there seems [ TP1 to [ vP be a man in the garden ]]]
(12a). The only solution is to move the complex DP to check the EPP.
To rule this out, it seems we would need a Merge-over-Move preference at TP1 , but the opposite
(12) a. [ TP2 T[EPP] [ vP exist [ DP evidence [ CP that [ TP1 there [ vP was a man in the garden ]]]]]]
preference (Move-over-Merge) at TP2 .
b. [ TP2 [DP evidence [ CP that there was a man in the garden ]] [ T′ T[EPP] [ vP exists ]]]
• There is evidence that expletive there merges low in Spec-vP (see e.g. Deal 2009; Wu 2017).
• It has been argued that there is no need to move to the specifier of TP1 (Castillo et al. 2009).
• The problem here is that we cannot derive the sentence in (9a) (an undergeneration problem).
• There have been both empirical (Shima 2000; Murphy & Puškar to appear) and conceptual
• Solution: Chomsky (2000) proposes that numerations should be divided into lexical subarrays, arguments (Chomsky 2013:41) for the opposite preference, Move-over-Merge. Do both exist?
i.e. smaller sub-numerations: How is competition between them regulated? (see Broekhuis & Klooster 2007 for a suggestion).

2
2.2 Phase heads and the PIC (18) XP
• If a phase is defined as a lexical subarray in the numeration, how are LAs defined?
X HP
• Chomsky (2000:106) talks about phases/LAs constituting a ‘natural syntactic object that is
relatively independent in terms of interface properties’
α H′
– For LF, a phase should be a ‘semantically complete/natural’ object:
On the ‘meaning side’, perhaps the simplest and most principled choice is to take SO
phase
[syntactic object] to be the closest syntactic counterpart to a proposition: either a verb H YP
edge
phrase in which all θ-roles are assigned or a full clause including tense and force. Call
these objects propositional.
(Chomsky 2000:106) phase Y ZP
head
– At PF, a phase should constitute a ‘phonologically complete/natural’ object: Spell-Out
The choice of phases has independent support: [. . . ] they have a degree of phonetic domain
independence (as already noted for CP vs TP). [. . . ] The same is true of vP constructions
generally.
(Chomsky 2001:12)
• For a head X, only elements at the phase edge (α and H) are accessible for X.
• YP (or phrases contained in YP) are no longer accessible to syntactic operations because they
• Chomsky (2000:106) suggests that we should assume that vP and CP constitute phases.
have already been transferred to the interfaces (LF and PF) after HP was complete.
• However, Chomsky (2000) assumed that only transitive vP (with an external argument) were
• The only way for a phrase such as ZP to be accessible for X is for it to first move to the phase
‘complete’ enough to constitute a phase. To clarify this, we can refer to transitive v’s as v*P.
edge.
• We now have some options when it comes to defining phases:
(19) XP
(16) a. Phases are convergent objects.
b. Phases are objects that determine points of Spell-Out (or Transfer).
X HP
• Option (16a) is undesirable because it requires Look-Ahead (i.e. we cannot know if an object is
convergent until it is sent to the interfaces).
ZP H′
• Instead, we can try to determine phases as the point in which syntactic structure is sent to the
interfaces in a MSO model.
α H′
• One way of achieving this is to combine the phases CP and v*P with the following condition:

(17) Phase Impenetrability Condition (PIC1 ) (Chomsky 2000:108): H YP

In a phase α with head H, the domain of H is not accessible to operations outside α, only [EPP]
H and its edge are accessible to such operations. Y

• The head of a phase (e.g. C, v*) is then a phase head. Spell-Out

• The edge of a phase includes the phase head and its specifier(s). domain

• The complement of the phase head is the Spell-Out domain that is sent to interfaces once the
phase is complete.
• Chomsky (2000) envisages the possibility of freely adding an EPP-feature to phase heads to
• This gives rise to the following abstract structure: facilitate this kind of movement.

3
 (20) CP • If unaccusative vP were a phase, then the DP should not be accessible for agreement/case
assignment from T, given the definition of the PIC in (17).

C TP (24) TP

T vP
T vP
Spell-Out
[φ:]
domain
DP v′ DP v′

v VP v VP

V DP
V DP
Spell-Out arrive
domain
a train
• Notice that the direct object of a transitive is inside the Spell-Out domain of the vP phase.
• In order for it to move out, e.g. in an object wh-question, it must first move to the edge of vP.
• Arguably, there may also have to be a relation between the expletive and its associate DP.
(21) Who does Peter like? Evidence for this comes from what is known as the ‘definiteness restriction’ on expletive
associates (see Milsark 1974 for English and Bobaljik & Jonas 1996 for Icelandic).
a. [ vP Peter [ v ′ v [ VP like who ]]] (Create vP)
b. [ vP Peter [ v ′ v [EPP] [ VP like who ]]] (EPP-insertion)
(25) Definiteness restriction:
c. [ vP who [ v ′ Peter [ v ′ v [EPP] [ VP like ]]] (Movement to vP edge)
a. There is a man in the room.
d. [ vP who [ v ′ Peter [ v ′ v [EPP] [ VP like ]]] (Spell-Out) b. There are three men in the room.
e. [ CP who C[wh] [ TP Peter T [ vP [v′ DP [ v ′ v [EPP] [ VP like ]]]]]] c. *There is the man in the room.
(Movement to Spec-CP) d. *There are those men in the room.

• Another consequence is that long distance movement proceeds through the edge of each phase Problem:
along the way (CP and vP): • Treating a defective v head as not being a phase potentially undermines the diagnostic of phases
being semantically and phonologically complete syntactic objects.
(22) Successive-cyclic movement:
• For example, just because a vP lacks an (overt) external argument does not mean that they are
[ CP Who do you [ vP think [ CP that Mary [ vP [ VP likes ]]]]] ?
incomplete in some way (all θ-roles assigned, propositional).
• Shouldn’t the same logic apply to unergatives that lack an overt internal argument?
2.3 Strong vs. weak phases • Furthermore, there is no clear sense in which unaccusative vPs are any less phonologically
complete than transitive v*Ps.
• Given the PIC in (17), we can see why unaccusative/passive vP was not assumed to be a phase. • To address this point, Chomsky (2001, 2004) assumes that all vPs are phases based on the
criteria of being a ‘convergent SO’, however introduced the notion of strong and weak phases.
(23) There arrived a train (at the station).

• In (23), the DP a train is clearly in its base-positionas the direct object of arrive.

4
 vP with nontransitive v is relatively isolted and [. . . ] cannot be phase for Spell-Out. (28) a. Henni höfðu leiðst þeir
Call these weak phases. then the strong phases are those that have an EPP-position her.dat had.3pl bored.at they.nom
as an escape hatch for movement and are, there, the smallest constructions that ‘She had found them boring.’ (Sigurðsson 2002:692)
qualify for Spell-Out. b. Marii podobała się ta książka
(Chomsky 2004:124) Maria.dat please refl this book.nom
‘Maria liked this book.’ (Citko 2014:35)
• Only strong phases trigger cyclic Spell-Out and provide an intermediate landing site for
movement (in class 3, we will see empirical arguments that this is wrong; Legate 2003). • There is evidence that this nominative is genuinely a syntactic object (see e.g. Zaenen et al.
• But it is unclear whether this distinction is really helpful. To all intents and purposes, ‘weak’ 1985; Taraldsen 1995 for Icelandic).
phases lack any discernable properties of phasehood. • Thus, we have the following structure:
[. . . ]if an XP does not trigger cyclic transfer, it is pointless to call it a phase.
(29) a. [ TP DPdat [ T′ T[φ:] [ vP dat [ v ′ v [ VP V DPnom ]]]]]
(Gallego 2010:168)

As far as we can tell, [the strong-weak] distinction plays no role in the theory, except that
it restricts phase properties to strong phases. Weak phases act as if they weren’t phases; in • Given the PIC in (17), the complement of the v phase head would become inaccessible after
particular, they don’t count as domains for the application of Spell-Out or PIC. the vP is complete.
(Boeckx & Grohmann 2007:215) • Consequently, the object DP is no longer accessible for agreement.
• There is also another potentially serious problem with the strong/weak phase distinction,
(30) a. [ vP DPdat [ v ′ v [ VP V DP ]]]]] (Spell-Out)
namely we can have unbounded A-dependencies involving raising TPs:
b. [ TP T[φ:] [ vP DPdat [ v ′ v [ VP V DP ]]]] (Merge TP)
(26) [ TP John [ vP seemed [ TP to [ vP appear [ TP to [ vP have been offended ]]]]]] 7
• Problem: Appealing to weak phases here would severely undermine the motivation for phases
(if we were to start allowing construction-specific exceptions).
• While there may or may not be EPP driven movement to these intermediate positions (Bošković • To address this issue, Chomsky (2001) proposes a slightly weaker definition of the PIC, which
2002; Castillo et al. 2009), if all of these non-transitive, defective vP were to constitute weak (i.e. delays Spell-Out of a phase complement until the next highest phase head is merged.
non-)phases, then the search space available to matrix T would rather large (even potentially
unbounded): (31) Phase Impenetrability Condition (PIC2 ) (Chomsky 2001:14):
Given the structure [ZP Z . . . [HP α [H′ H YP ]]], where H and Z are phase heads,
(27) [ TP T[EPP] [vP seemed [ TP to [ vP appear [ TP to [ vP have been offended John ]]]]] the domain of H is not accessible to operations at ZP; only H and its edge are accessible
to such operations.
search space of T
• This is sometimes referred to as the ‘Weak PIC’ (Gallego 2010:58) or as PIC2 (Müller 2004:291).
• This seems to severely undermine the idea that phases and cyclic Spell-Out are motivated by • Under this definition, the complement of vP is not sent to Spell-Out until the next highest
reducing computational burden (more on this below). phase head (C) is merged.

2.4 PIC1 vs. PIC2 (32) a. [ vP v DP ]

b. [ vP DPdat [ v ′ V DP ]] (No Spell-Out under PIC2 !)
• The conception of phases in Chomsky (2000) faced a much more serious problem, however. c. [ TP T[φ:pl] [ vP DPdat [ v ′ V DP ]]] (Agree possible with DO)
• Recall that the problem of agreement into a phase domain with unaccusatives/passives was
solved by assume that these constitute weak/non-phases. d. [ CP C [ TP DPdat [ T′ T[φ:pl] [ vP [ v ′ V DP ]]]]] (Merge CP)
e. [ CP C [ TP DPdat [ T′ T[φ:pl] [ vP [ v ′ V DP ]]]]] (Spell-Out of vP)
• However, there are languages in which there is agreement into the domain of a transitive vP:

5
• If we have this view of the PIC, then we do not really need a strong/weak phase distinction. (35) Phase Impenetrability Condition (PICLA ) (Richards 2011:81):
The problem of T agreeing with the direct object in unaccusatives/passives disappears (at least Given [ZP Z [XP X [HP α [H′ H YP ]]]], with H and Z belonging to separate LAs,
in these simple cases). the domain of H is not accessible to operations ‘outside’ [the maximal phrase projected
from] LAH ; only H and its edge α are accessible to such operations
where LAH = the lexical subarray defined by head H,
• However, the PIC2 weakens the locality of syntactic operations significantly (Müller 2004;
with either X ∈ LAH
Richards 2011) and this undermines one of the main conceptual motivations for phases (com-
(= PIC2 )
X ∉ LAH
putational efficiency).
or (= PIC1 )
• In the case of T agreeing with a direct object DP, the search space available to T is increased
into the domain of the phase head under the PIC2 (33b). • Here, the PIC is defined depending on which LA non-phase head such as T are assigned to.

(33) a. [ CP C [ TP T [vP DP [ v ′ v [ VP V DP ]]]]] • Depending on how we construct our LAs, we get either PIC1 or PIC2 :

search space of T (PIC1 ) (36) a. PIC1 = {C, T}, {v, V}, . . . [CP C [ TP T [vP DP [ v ′ v [ VP V DP ]]]]]
b. [ CP C [ TP T [vP DP [ v ′ v [ VP V DP ]]]]] Max. phrase projected from LA
search space of T (PIC2 ) b. PIC2 = {C}, {T, v}, {V}, . . . [CP C [TP T [ vP DP [ v ′ v [ VP V DP ]]]]]
Max. phrase projected from LA
• While in this example it may seem like a minimal extension, depending on the syntactic
structure involved, the domain accessible could be larger: • Consequence: We remove the concept of phase heads/phase edges (see Richards 2011 for what
to replace it with).
(34) I told John who I’d like to meet.
a. [ CP C [ TP T [vP DP [ v ′ v [ VP DP [ V′ V [ CP DP [ C′ C TP ]]]]]]]]
2.5 Feature inheritance and further developments
search space of T (PIC1 )
• There were further developments in Phase Theory after Chomsky (2001).
b. [ CP C [ TP T [vP DP [ v ′ v [ VP DP [ V′ V [ CP DP [ C′ C TP ]]]]]]]]
• For example, it has been argued that phase heads can be defined as the locus of uninterpretable
search space of T (PIC2 ) features (see e.g. Chomsky 2004, 2008; Gallego 2010:60; Miyagawa 2011:1273; Legate 2012).
• The concept of Feature Inheritance was introduced by Chomsky (2007, 2008), where uninter-
• PIC1 : Reduced search space, problem with DAT-NOM constructions pretable features on C and v are passed down to the next non-phase head, i.e. T and V.
• PIC2 : Increased search space, no problem with DAT-NOM constructions
(37) Feature Inheritance (Chomsky 2007, 2008):
The indeterminacy in PIC formulations is thus identical with and equivalent to an a. [ CP C[uF] [ TP T [ vP DP [ v ′ v [uF] [ VP V DP ]]]]]
indeterminacy in T’s search space—nothing else changes or is at stake. The question b. [ CP C [ TP T[uF] [ vP DP [ v ′ v [ VP V[uF] DP ]]]]]
of why two divergent definitions of the PIC are possible (as opposed to just a single
PIC, or indeed three, four, or 34 PICs) thus reduces to the question of why T’s (i.e.,
the nonphase head’s) search space is undetermined by phase theory in this way. • Question: What kind of operation is FI? It doesn’t look like Agree. It must be conceived of as a
If we can find a principled answer to this latter question, then we can remove the kind of ‘feature sharing’ (Richards 2007) or perhaps copying (Gallego 2014).
stipulation of T’s search space (i.e., (8a)) from the definition of the PIC itself and • Furthermore, feature inheritance requires that operations at the phase level take place ‘simulta-
thus reduce the two PICs to a single, fundamental formulation – a unified PIC. neously’ (Richards 2007).
(Richards 2011:80f.) • Among other reasons, the EPP and φ-feature probes on T actually originate on C.
• Richards (2011) suggests that the two PICs can be unified if phases are defined over lexical • One such simple case discussed by Chomsky (2004) involves an apparent problem with inter-
arrays, rather than categories (PICLA ): vention by moving operators (Citko 2014:48f.). Consider (38).

6
 (38) Who have the boys talked to? 2.6 Some criticism of Classical Phase Theory
(39) TP 2.6.1 Phase ≠ Spell-Out domain
• Often interface properties of the phase-defining category (C, v) are taken to be relevant, for
T vP example if it is a ‘complete’ phonological or semantic object.
[φ:]
• However, this logic is undermined by the fact that interfaces only see the phase complement,
DP v ′ which undergoes Transfer (cf. some diagnostics in Legate 1998 and Matushansky 2005).
who • For this reason, a transitive VP should count as ‘propositionally complete’, which it is not (at
[φ:3sg]
DP v ′ least intuitively).
the boys
Chomsky argues that TP is not a phase. Interestingly, the criterion for phasehood he adopts
[φ:3pl] v VP –propositionality– would classify the embedded TP [. . . ] as a phase.
(Bošković 2002:182)

7 V PP • So a raising TP such as (40a), for example, would constitute a phase given this criterion. If this
talk were the case, then the ban on ‘improper movement’ would be mysterious (40b) (*Ā → A).
P
to (40) a. John1 seems [ TP 1 to be happy ]
b. *John1 seems [ CP 1 that [ TP 1 is happy ]]

• The argument goes that T should not be able to agree with the subject because the intermediate
• The same holds for CP, where force and discourse-related features are hosted.
subject intervenes.
• The same confusion holds for phonological diagnostics, too. Chomsky (2001) points to the
• If arguments apply ‘simultaneously’, i.e. movement to Spec-CP and Agree between T and the
phonetic independence of CP vs. TP, for the phasehood of the former rather than the latter.
boys, then this problem does not arise (apparently).
However, it precisely TP that is sent to the interfaces as a single unit (and not CP+TP, for
Objection: example).
• Whether or not there is intervention at the phase edge ultimately depends on the timing of • Interface diagnostics should perhaps be taken with a pinch of salt if something like non-
intermediate movement and merge of the subject (see e.g. Heck & Himmelreich 2017). simultaneous Spell-Out is possible (e.g. Marušič 2009).
• Chomsky (2008) provides other arguments based on CED effects (cf. Gallego 2011) and opaque
intervention in Icelandic (cf. Bošković 2012), however these all have plausible alternative expla- 2.6.2 The status of the ‘spelled out’ phase complement
nations. Ultimately, it seems that there are no really convincing arguments for simultaneous
• Another motivation for cyclic Spell-Out in phases comes from efficient computation, as we
application of operations at the phase level.
have seen.
• There are also conceptual objects regarding the seeming abandonment of cyclicity and deriva-
• Syntactic material that has been sent to the interfaces is no longer accessible for later operations
tionalism (cf. Epstein & Seely 2002).
and can be ‘forgotten’ (Chomsky 2001:12f.).
• One criticism of this is that, while the internal structure of the phase complement may be inac-
Note:
cessible, the phase complement itself (VP, TP) must still be accessible for syntactic operations
• There have been some further developments related to labelling that I will not discuss, i.e. the (i.e. movement).
POP+ framework (Chomsky 2013, 2015). As far as I can see, their assumptions diverge too
• This has been proposed to be a desirable result. For example, Abels (2003, 2012) proposed that
much from Classical Phase Theory to be considered a continuation of the same theory.
phase heads cannot be stranded

(41) Stranding Generalization (Abels 2003:09; Abels 2012:89):

The complement of a phase head may not be moved.

7
• This actually follows from the PIC coupled with a more general constraint on ‘anti-local’ (45) Base-generation: (46) Movement:
movement, i.e. movement that is too short. a. [ CP kong [ TP . . . ] ] a. [ CP kong [ TP . . . ] ]
b. [ CP [ TP . . . ] kong ] b. [ CP [TP . . . ] kong TP ]
(42) Anti-Locality (Abels 2003; Grohmann 2003):
YP
• Evidence for movement in this construction comes from tone sandhi (marked by ●). Tone
Y′ sandhi is a rule-governed tonal change on a syllable when preceded by another tone-bearing
syllable. The exact change is unimportant here, if tone sandhi has occurred, the syllable is
followed by ●.
Y ZP
• The fact that tone sandhi requires a preceding tone-bearing syllable is shown by the fact it does
7 not occur sentence-finally:

(47) No tone sandhi clause-finally (Simpson & Wu 2002:73):

• If Y is a phase head, then in order for ZP to be moved, it must be present at the edge of YP in A●sin chin● ho. A●hui ma● chin● ho.
order (given the PIC). A-sin very good A-hui also very fine
• However, movement from the complement position to the specifier position violates Anti- ‘A-sin is very well. A-hui is also very well.’
Locality (42).
• Thus, stranding of phase heads is predicted to be ungrammatical. • If we go back and look at the examples in (44), we see that tone sandhi is triggered on the
complementizer kong in (44a) since it precedes the TP.
• Is this empirically borne out?
• Crucially, tone sandhi is found on kong in final position in (44b), although this is not generally
• In English, CPs can be moved (43a), but TPs cannot (43b). possible (47). This only makes sense if the TP-complement is moved, as in (46b).1

(43) a. [ CP that Mary is smart ] was believed • This independent tone sandhi diagnostic provides a strong argument for the movement analysis
CP by all.
b. *[ TP Mary is smart ] was believed [ CP that and shows that phase complements must, in some cases, be moveable.
CP ] by all.

• However, it has been argued that sentence-final particles are derived by movement of TP to 2.6.3 Locality of Move vs. Agree
Spec-CP. • Bošković (2003, 2007) argues that Agree can reach into a finite clause in violation of the PIC,
• This could follow from our background assumptions on linearization, i.e. Kayne’s (1994) Linear especially in Algonquian (also see Lee 2003; Legate 2005):
Correspondence Axiom, where head-final structures have to be derived by ‘roll-up’ movement.
• Possibly the most convincing case for movement of a phase complement is discussed by (48) a. Chuckchee (Bošković 2007:613):
Simpson & Wu (2002). @nan q@lGil˛u l@N@rk@-nin-et [ CP iNqun Ø-r@n@mN@v-nen-at qora-t ]
he regret-3-pl that 3sg-lost-3-pl reindeer-pl
• They discuss the complementizer kong in Taiwanese that can occur either before or after the
TP complement. ‘He regrets that he lost the reindeers.’
b. Blackfoot (Legate 2005:150):
(44) Final and initial-kong in Taiwanese (Simpson & Wu 2002:79,81): kits-íksstakk-a [ CP omá n-oxkó-wa máxk-itáp-aapiksistaxsi kiistóyi omí
a. A●hui liau●chun● [ CP kong● [ TP A●sin si● tai●pak● lang ]] 2obj-want-3subj my 1-son-3 3-might-toward.throw you det
A-hui thought kong Asin is Tapei person pokón-i ]
‘A-hui thought that A-sin is from Tapei.’ ball-4
b. [ CP [ TP A●sin si● tai●pak● lang ] kong● ] ‘My son wants to throw the ball to/at you.’
Asin is Tapei person kong
‘A-sin is from Tapei.’ 1 There are some issues here regarding the exact nature of Spell-Out. It seems we may have to assume some interleaving
of syntax and postsyntax (see Martinović to appear).

8
• This is even problematic for the PIC2 (but cf. Polinsky & Potsdam 2001 on Tsez). 4.1 Successive-cyclicity
• I suggested that PIC2 might obviate the need for the strong/weak phase distinction:
• The primary syntactic diagnostic for phases pertains to successive-cyclic movement.
(49) [ CP C [ TP T [vP DP [ v ′ v [ VP V DP ]]]]] • Recall that the PIC means that moving elements must pass through the edge of each phase, if
not generated there.
search space of T (PIC2 )
(55) Successive-cyclic movement:
• This is only true for monoclausal cases, however. (50) is problematic if all vPs are phases. [ CP Who do you [ vP think [ CP that Mary [ vP [ VP likes ]]]]] ?
(50) a. [ TP Ten trains [ vP seem(*-s) [ TP to have [ vP arrived into the station today ]]]]
b. [ TP There [ vP seem(*-s) [ TP to have [ vP arrived ten trains into the station today ]]]] • We can therefore posit the following diagnostics for successive-cyclicity.
(Legate 2005:148)
(56) a. Intermediate pronunciation:
• The problem is that T needs to agree with the direct object, however the complement of the Can (part of) a moving phrase be pronunced at an intermediate landing site?
phase head v 1 (if a strong phase), becomes inaccessible after v 2 is merged. b. Intermediate interpretation:
Can a moving phrase be interpreted at an intermediate landing site?
(51) [ TP There T [vP2 v 2 seem [ TP to have [ vP1 v 1 [ VP arrived ten trains ] into the station c. Intermediate licensing:
today]]]] Do certain licensing properties of a moved item hold at an intermediate landing site?
search space of T

3 An alternative to the PIC (Fox & Pesetsky 2005) 4.2 PF diagnostics

• Fox & Pesetsky (2005) propose an alternative way to derive successive-cyclic movement without
• If the phase complement and phase edge are transferred separately, we expect the following:
the PIC.
• They argue that phases (vP and CP) are spelled-out upon completion, and a set of linearization (57) Null Hypothesis:
statements is generated. Items in the phase edge should not interact phonologically with items in the phase com-
plement (assuming domains of phonological computation match Spell-Out domains).
(52) Order Preservation (Fox & Pesetsky 2005:6):
Information about linearization, once established at the end of a given Spell-out domain, • This seems to be wrong in many cases, e.g. tone sandhi in Taiwanese (Bošković 2017).
is never deleted in the course of a derivation.
(53) [CP Who does [ TP Mary [vP tMary [ v ′ v [ VP like ⟨who⟩ ]]] ]] 4.3 LF diagnostics
• Less obvious: Should phases be of a particular type (e.g. ⟨t⟩, ⟨e⟩)?
who ≺ Mary ≺ like Mary ≺ like ≺ who
(54) [CP Who does [ TP Mary [vP ⟨who⟩ [ v ′ tMary [ v ′ v [ VP like ⟨who⟩ ]]]] ]]
• It has been proposed that Quantifier Raising targets phases. . . but why?
who ≺ Mary who ≺ Mary ≺ like

• This avoids the stipulative nature of the PIC (viz. edge).

• However, it requires that ‘spelled out’ material is freely accessible, so arguably does not reduce References
computational burden in any obvious way. Abels, Klaus (2003). Successive Cyclicity, Anti-Locality and Adposition Stranding. PhD thesis, University of
Connecticut.
4 Diagnostics for phasehood Abels, Klaus (2012). Phases: An Essay on Cyclicity in Syntax. de Gruyter: Berlin.
Bobaljik, Jonathan David & Dianne Jonas (1996). Subject Positions and the Roles of TP. Linguistic Inquiry
In order for Phase Theory to be a viable research programme, we need to have a set of diagnostics 27(2). 195–236.
that can be used to identify phases. What could these be? Boeckx, Cedric & Kleanthes K. Grohmann (2007). Putting Phases In Perspective. Syntax 10(2). 204–222.

9
Bošković, Željko (2002). A-Movement and the EPP. Syntax 5(3). 167–218. Benjamins: Amsterdam.
Bošković, Željko (2003). Agree, Phases and Intervention Effects. Linguistic Analysis 33. 54–96. Heck, Fabian & Anke Himmelreich (2017). Opaque Intervention. Linguistic Inquiry 48(1). 47–97.
Bošković, Željko (2007). On the Locality and Motivation of Move and Agree: An Even More Minimal Kayne, Richard (1994). The Antisymmetry of Syntax. MIT Press: Cambridge, MA.
Theory. Linguistic Inquiry 38(4). 589–644. Lee, Jaecheol (2003). Phase Sensitivity in Wh-Dependencies. Korean Journal of Linguistics 28. 67–89.
Bošković, Željko (2012). Don’t Feed Your Movements When You Shift Your Objects. In M. Uribe-Etxebarria Legate, Julie Anne (1998). Verb Phrase Types and the Notion of a Phase. Ms. MIT.
& V. Valma (eds). Ways of Structure Building. Oxford University Press: Oxford. 245–252. Legate, Julie Anne (2003). Some Interface Properties of the Phase. Linguistic Inquiry 34(3). 506–516.
Bošković, Željko (2017). Tone Sandhi in Taiwanese and Phasal Spell-Out. Journal of Linguistic Society of Legate, Julie Anne (2005). Phases and Cyclic Agreement. In M. McGinnis & N. Richards (eds). Perspectives
Japan 152. 31–58. on Phases. MITWPIL: Cambridge, MIT. 147–156.
Bresnan, Joan (1978). Contraction and the Transformational Cycle in English. Indiana University Linguis- Legate, Julie Anne (2012). The Size of Phases. In Á. Gallego (ed.). Phases: Developing the Framework. de
tics Club, Bloomington. Gruyter: Berlin. 233–250.
Broekhuis, Hans & Wim Klooster (2007). Merge and Move as Costly Operations. Groninger Arbeiten zur Martinović, Martina (to appear). Interleaving Syntax and Postsyntax: Spell-Out Before Syntactic Move-
germanistischen Linguistik 45. 17–37. ment. Syntax.
Castillo, Juan Carlos, John E. Drury & Kleanthes K. Grohmann (2009). Merge over Move and the Marušič, Franc Lanko (2009). Non-Simultaneous Spell-Out in the Clausal and Nominal Domain. In K. K.
Extended Projection Principle: MOM and the EPP Revisited. Iberia 1(1). 53–114. Grohmann (ed.). InterPhases: Phase-Theoretic Investigations of Linguistic Interfaces. Oxford University
Chomsky, Noam (1995). The Minimalist Program. MIT Press: Cambridge, MA. Press: Oxford. 151–181.
Chomsky, Noam (2000). Minimalist Inquiries: The Framework. In R. Martin, D. Michaels, J. Uriagereka Matushansky, Ora (2005). Going Through A Phase. In M. McGinnis & N. Richards (eds). Perspectives on
& S. J. Keyser (eds). Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. MIT Press: Phases. MITWPL: Cambridge, MA. 157–181.
Cambridge, MA. 89–155. Milsark, Gary L. (1974). Existential Sentences in English. PhD thesis, MIT.
Chomsky, Noam (2001). Derivation by Phase. In M. Kenstowicz (ed.). Ken Hale: A Life in Language. MIT Miyagawa, Shigeru (2011). Genitive Subjects in Altaic and the Specification of Phase. Lingua 21. 1265–1282.
Press: Cambridge, MA. 1–52. Müller, Gereon (2004). Phase Impenetrability and Wh-Intervention. In A. Stepanov, G. Fanselow &
Chomsky, Noam (2004). Beyond Explanatory Adequacy. In A. Belletti (ed.). Structures and Beyond: The R. Vogel (eds). Minimality Effects in Syntax. de Gruyter: Berlin. 289–325.
Cartography of Syntactic Structures. Vol. 3. Oxford University Press: Oxford. 104–191. Murphy, Andrew & Zorica Puškar (to appear). Closest Conjunct Agreement is an Illusion. Natural
Chomsky, Noam (2007). Approaching UG from Below. In U. Sauerland & H. Gärtner (eds). Interfaces + Language and Linguistic Theory
Language = Recursion?. de Gruyter: Berlin. 1–29. Obata, Miki (2006). Phase and Convergence. University of Pennsylvania Working Papers in Linguistics
Chomsky, Noam (2008). On Phases. In R. Freidin, C. Otero & M.-L. Zubizarreta (eds). Foundational 12(1). 279–292.
Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. MIT Press: Cambridge, MA. Ott, Dennis (2009). The Conceptual Necessity of Phases: Some Remarks on the Minimalist Enterprise.
133–165. In K. K. Grohmann (ed.). Explorations of Phase Theory: Interpretation at the Interfaces. de Gruyter:
Chomsky, Noam (2013). Problems of Projection. Lingua 130. 33–49. Berlin. 253–275.
Chomsky, Noam (2015). Problems of Projection: Extensions. In E. Di Domenico, C. Hamann & S. Mat- Polinsky, Maria & Eric Potsdam (2001). Long-Distance Agreement and Topic in Tsez. Natural Language
teini (eds). Structures, Strategies and Beyond: Studies in Honour of Adriana Belletti. John Benjamins: and Linguistic Theory 19(3). 583–646.
Amsterdam. 1–16. Richards, Marc D. (2007). On Feature Inheritance: An Argument from the Phase Impenetrability Condi-
Citko, Barbara (2014). Phase Theory: An Introduction. Cambridge University Press: Cambridge. tion. Linguistic Inquiry 38(3). 563–572.
Deal, Amy Rose (2009). The Origin and Content of Expletives: Evidence from ‘Selection’. Syntax 12(4). Richards, Marc D. (2011). Deriving the Edge: What’s in a Phase?. Syntax 14(1). 74–95.
285–323. Shima, Etsuro (2000). A Preference for Move over Merge. Linguistic Inquiry 31(2). 375–385.
Epstein, Samuel D. & Daniel T. Seely (2002). Rule Applications as Cycles in a Level-Free Syntax. In Sigurðsson, Halldór Ármann (2002). How to be an Oblique Subject: Russian vs. Icelandic. Natural
S. Epstein & T. Seely (eds). Derivation and Explanation in the Minimalist Program. Blackwell: Oxford. Language and Linguistic Theory 20(4). 691–724.
65–89. Simpson, Andrew & Zoe Wu (2002). IP-Raising, Tone Sandhi and the Creation of S-Final Particles:
Fox, Danny & David Pesetsky (2005). Cyclic Linearization of Syntactic Structure. Theoretical Linguistics Evidence for Cyclic Spell-Out. Journal of East Asian Linguistics 11(1). 67–99.
31(1–2). 235–262. Taraldsen, Knut Tarald (1995). On agreement and nominative objects in Icelandic. In H. Haider, S. Olsen
Frampton, John & Sam Gutmann (1999). Cyclic Computation, A Computationally Efficient Minimalist & S. Vikner (eds). Studies in Comparative Germanic Syntax. Kluwer: Dordrecht. 307–327.
Syntax. Syntax 2(1). 1–27. Uriagereka, Juan (1999). Multiple Spell-Out. In S. D. Epstein & N. Hornstein (eds). Working Minimalism.
Gallego, Ángel J. (2010). Phase Theory. John Benjamins: Amsterdam. MIT Press: Cambridge, MA. 251–282.
Gallego, Ángel J. (2011). Successive Cyclicity, Phases and CED Effects. Studia Linguistica 65(1). 32–69. Wu, Dafeng (2017). Ellipsis as a diagnostic of movements in expletive there and it sentences. Ms. MIT.
Gallego, Ángel J. (2014). Deriving Feature Inheritance from the Copy Theory of Movement. The Linguistic Zaenen, Annie, Joan Maling & Höskuldur Thráinsson (1985). Case and Grammatical Functions: The
Review 31(1). 41–71. Icelandic Passive. Natural Language and Linguistic Theory 3(4). 441–483.
Grohmann, Kleanthes K. (2003). Prolific Domains: On the Anti-Locality of Movement Dependencies. John

10