Professional Documents
Culture Documents
What Can The Presence or Absence of A-Bar Gaps in Adjunct Clauses Tell Us About Constituency and Movement in The Surrounding Environment?
What Can The Presence or Absence of A-Bar Gaps in Adjunct Clauses Tell Us About Constituency and Movement in The Surrounding Environment?
What can the presence or absence of A-bar gaps in adjunct clauses tell us about constituency
and movement in the surrounding environment?
The ability of an infinitival clause in English to contain a non-subject gap is often dependent on
the syntactic environment.
(1) Some snow [(for me) to shovel (up)] is in the driveway. Infinitival relative
cf. *Some snow (for me) to shovel it ...
(2) This type of snow is easy [(for me) to shovel ] Tough-movement
cf. *This type of snow is easy (for me) to shovel it ...
These facts should follow from standard assumptions about semantic composition. The infinitival
clause contains a gap if and only if its sister is a predicate of individuals.
(5) a. This snow is light enough [for me to shovel (it)] too/enough clauses
b. Frank is too angry [for me to talk to (him) now]
cf. *#This snow is light for me to shovel (it)
. *#Frank is angry for me to talk to (him)
• The lightness of the snow in (5a) is below the threshold that would make shoveling it
impossible; and
• Frank’s anger in (5b) is above the threshold that makes my talking to him possible.
Strikingly, the infinitival clause licensed by too/enough allows a gap, but this gap is optional.
In the second part of the talk, I will try to show (reporting joint work with Bernhard Schwarz) that
even here the optionality is an illusion — the two versions (gapped and gapless) have different
surrounding structures once we scratch below the surface.
The picture becomes murkier still with infinitival VP adjuncts, in which the gap truly appears to
be optional:
Gapped and gapless VP-adjuncts
(6) a. Someone left this snow here [for me to shovel ] Purpose clause
b. Someone left this snow here [for me to shovel it] Rationale clause
• The gapped version (6a) — dubbed “purpose clause” by Faraci 1974 — implies only that
the speaker needs to shovel the snow that has been left (perhaps by happenstance).
• The gapless (6b)—a “rationale clause”—entails something about the intent of the agent.
Unlike (6a), the snow in (6b) can’t have been left by happenstance; it can only have been
left as part of the agent’s plan for the speaker to end up shoveling it.
Puzzles:
• What is responsible for this meaning difference, and why should it correlate with the presence
vs. absence of a gap?
In the remainder of this talk, I will report on progress toward unification of these constructions.
Part Two will present evidence for something similar with too/enough — the ”attachment site”
is key, in that the presence of a gap always disambiguates toward a surface scope reading.
In Part One, I will summarize a body of work (going back to Faraci) showing a difference in the
attachment sites of Prupose and Rationale clauses.
Specifically, Purpose Clauses like (6a) are attached low, internal to the VP, while Rationale
Clauses (6b) are VP-external, likely adjoined to the vP:
(7) vP
✭✭✭✭✭❤❤❤❤❤
✭ ❤❤❤❤
✭✭✭✭ ❤
vP Rationale Clause
✭✭✭❤❤❤❤❤❤
✭✭✭✭✭✭ ❤❤❤
✭ ❤
subject V object Purpose Clause
This suggests the possibility of an explanation for the syntactic part of the alternation — i.e. that
there is something about the attachment site that determines whether an infinitival VP-adjunct
can contain a gap. If so, the alternation in (6) is not exceptional after all.
I will attempt to demonstrate that both the obligatory gap in a Purpose Clause and the agentive
entailment of a Rationale Clause — that is, both the syntactic and the semantic parts of the
alternation — follow from the difference in attachment site.
The argument builds upon recent advances in our understanding of the articulated structure
within the verb phrase. I will suggest that if we assume a highly articulated VP structure,
together with a corresponding theory of decomposition of verbal meanings, the differences (and
similarities) between the two types of infinitival adjunct can be attributed in full to the meanings
(and semantic types) of the constituent to which each type of adjunct attaches.
1. A few preliminaries about the distinction between PCs and RCs
• The two types of infinitival VP adjuncts are similar, but they have distinctive properties.
[See Faraci 1974, Huettner 1989, Jones 1985.]
⇒ Purpose clauses are VP-internal, containing a gap bound to the matrix object.
(8) a. Max brought his cat2 here [for me to admire 2 ] Purpose clauses
b. His cat2 was brought t2 here [for me to admire 2 ]
c. Max brought his cat2 here [ 2 to sniff me]
⇒ Rationale clauses are external to the VP, and are not dependent on the matrix object.
(9) a. Max brought his cat here [(in order) for me to admire it] Rationale clauses
b. Max brought his cat here [(in order) for me to cheer up]
(10) a. Max brought his cat2 here [(*in order) for me to admire 2]
b. Max brought his cat2 here [(*in order) 2 to sniff me]
c. Max1 brought his cat2 here [(in order) 1 to annoy me]
• Purpose clauses must have a gap. In contrast, the only allowable gap in a Rationale Clause is
pro controlled by the matrix subject.
A restatement of the puzzle: Why should the presence/absence of a gap, and the meaning
difference, correlate with low–vs.–high attachment?
(11) CP
✦❛❛
✦✦ ❛
OP1 C
✟❍
✟✟ ❍❍
C TP
✏PP
✏✏
✏ PP
for me to rake t1
• Internally, then, PCs have the syntax of infinitival relative clauses, but modify some constituent
other than a NP.
(12) a. Here are [some leaves [OP for you to rake t]] Infinitival relatives
b. [Some leaves [OP for you to rake t]] are right here
(13) I left them (right here) [OP for you to rake t] Purpose clause
(14) [Someone [OP t to rake the leaves]] is right here Infinitival relative
2.2. Low vs. high attachment—Evidence for the correlation
(7) vP
✭✭❤❤❤❤
✭✭✭✭ ❤❤❤❤
✭✭✭✭ ❤❤
vP Rationale Clause
✭✭✭✭✭✭❤❤❤❤❤❤
✭ ❤❤❤
✭✭✭ ❤
subject V object Purpose Clause
(16) a. They1 brought Max2 along [(in order) pro1 to amuse themselves]
(17) They brought Max along [OP pro to introduce {themselves1 /*himself2 } to t]
§ Relative ordering
(18) a. They brought Max along [ to talk to himself] [(in order) to amuse themselves]
(19) a. George put that gun on the table [for me to shoot him with ] [in order to prove
I’m a coward]
b. * George put that gun on the table [in order to prove I’m a coward] [for me to shoot
him with ]
§ VP-fronting
(20) a. * I said I’d invite Max over. . . and [invite Max over] I did [ to talk about himself]
b. I said I’d invite Max over. . . and [invite Max over] I did [ to amuse myself]
(21) a. * I said I’d invite Max over. . . and [invite Max over] I did [for you to talk to ]
b. I said I’d invite Max over. . . and [invite Max over] I did [for you to talk to him]
• Given these assumptions, our initial puzzle should be restated somewhat, and broken
into two pieces.
⇒ Puzzle one: Why is null operator movement obligatory in a VP-internal infintival adjunct?
And why is it impossible in a VP-external adjunct?
⇒ Puzzle two: How and why exactly does the meaning of an infinitival adjunct change
depending on whether it is VP internal or VP-external?
3. Some differences (and similarities) in meaning
• The meanings of purposive adjuncts are modal.
(They express something about purposes/goals/designs/intent.)
(22) a. Max built that house for his kids to inherit ] Purpose clause
b. Max built that house for his kids to inherit it Rationale clause
(23) a. Mary put these papers on the desk [for you to sign ] Purpose clause
b. Mary put these papers on the desk [for you to sign them] Rationale clause
(24) a. Someone left this snow here [for me to shovel ] Purpose clause
b. Someone left this snow here [for me to shovel it] Rationale clause
• A difference in entailments about the agent’s intention can often be teased out.
♦ Rationale clauses express the intended purpose of an action (by the actor)
♦ Purpose clauses can express a more abstract intention—not necessarily that of an agent
⇒ An abstract desire or aim that pertains to a result state?
Passives, unaccusatives and adjectives bring out the difference more sharply
(26) Quote from the stork in Dumbo (after delivering baby Dumbo to his mother)
“Straight from heaven, up above,
Here is a baby for you to love.” (Infinitival relative)
• Specific parts of this basic intuition that I’d like to flesh out:
⇒ Some vPs describe actions, and Rationale clauses are understood as describing the agent’s
intended purpose in carrying out the action.
⇒ Some verbs evoke result states as part of their meanings, and Purpose clauses express
something about goals/intentions that relate to these states (rather than to the events
that cause them).
⇒ The reason that the meanings differ only subtly in many cases (and are sometimes not
distinguishable) is that when the result state is taken to be the direct, intended consequence
of an action, the most salient goal that can be expressed about the result state is simply
the one held by the agent of the causing event.
⇒ for Ned to talk to me means, essentially, that it is desired that Ned talk to me.
(30) Note about semantic types (and conventions I will adopt for naming variables):
- Variables named ‘s’ range over possible situations/events/states/worlds (type s)
- Those named ‘w’ will be limited to the special case of possible worlds
- Functional types and variables:
‘P ’ ranges over functions of type he,sti; ‘p’ ranges over type hs,ti
(32) The meaning of a for-infinitival adjunct clause with null operator movement:
[[Oi for Ned to talk to t i ]]C x = λxλs.[Ned talks to x in every world accessible from w s
that is compatible with the goals (etc) salient in C]
• In short:
⇒ A plain infinitival adjunct (like a Rationale clause) is a function of type hs,ti, and expresses
a salient purpose related to the ‘world (event, etc.) of evaluation’.
⇒ A purpose clause is just the same thing with a gap, that is, a property of type he,sti.
4. A joint solution to both puzzles
• To begin with, we can now restate the two puzzles, now in minimally more precise terms.
⇒ Puzzle one: Why must a VP-internal infintival adjunct have semantic type he,sti? And
why must a VP-external adjunct have type hs,ti?
⇒ Puzzle two: Why does a VP-external purposive adjunct necessarily express the intentions
of the agent? Why do we get precisely the pattern of entailments we get with VP-internal
purposive adjuncts?
• Baker’s Conjecture:
(33) All verb phrases contain an underlying Adjective at their core. [Baker 2003]
vP
✘✘❳❳❳
✘✘✘
✘ ❳❳
❳
DP v
✘✘❳❳❳
✑✑◗◗ ✘✘✘ ❳❳
subject v VP
✏✏PPP
✏✏ P
CAUSE DP V
★❝ ✦❛❛
★ ❝ ✦✦ ❛
object V Adj
BE
bake
boil
bring
build
...
bring
Ø
build ...
This version of Baker’s conjecture is compatible with Gonsalves’ [2008] suggestion (citing work
of Mora Gutiérrez [2001]) that the meanings of such constructions should be paraphrased along
the lines ‘x causes y to be ADJ by V-ing’
This Adjective-like predicate can be modified by an adjunct of the same type.
(35) ...
VPhsti
✭✭✭✭❤❤❤❤❤❤
✭✭✭ ❤
DP Vhesti
★❝ ✭✭✭✭❤❤❤❤❤❤
★ ❝ ✭✭✭
✭ ❤❤
object Vhhestihestii he, sti
✘✘❳❳❳
✘✘✘
✘ ❳❳
❳
bake he, sti he, sti
boil Result adjoined CP
✦❛❛
bring
Predicate ✦✦
✦ ❛❛
build
⇒ Adjuncts that modify this complement will necessarily have operator movement.
Moreover, if it’s correct that XP-adjunction isn’t permitted to non-maximal projections [Chomsky
1986], we have an explanation for why Rationale clauses can’t have null operator movement—the
semantic type can only be hs,ti.
A simple example:
(37) Phoebe brought Max here
(38) • Phoebe is the agent of a volitional act.
• Max becomes (as the result of Phoebe’s volitional act) transported.
• There is a result state consisting in Max’s being here
b. [[V bring]] = λPhe,sti λxλs. bringing(x)(s) & ∃s′ [s′ ≤p s & P(x)(s′ ) ]
c. [[v cause]] = λphsti λxλs.event(s) & agent(s)(x) & ∃s′ [cause(s′ )(s) & p(s′ )]
• The meaning differences will follow as well — once we take into account that agentive
events (as opposed to states) are goal-directed.
⇒ States do not come with inherent goals. Determining which goal is salient is a more flexible
matter.
So...
If a purposive adjunct (like (36b)) modifies the Result Predicate node, in a sentence like
it will be interpreted as expressing some (contextually salient) purpose that relates to the result
state of Max’s being here.
(41)
hsti
vP
✭✭✭✭✭✭❤❤❤❤❤❤
✭ ❤
DP vhesti
✭✭✭❤❤❤❤❤❤
✁✁❆❆ ✭✭✭✭ ❤
we vhhstihestii VP hsti
✭✭✭✭✭❤❤❤❤❤
✭✭ ❤❤
DP Vhesti
✭✭✭❤❤❤❤❤
✪❡
✪ ❡ ✭✭✭✭
✭ ❤❤
❤
Max Vhhestihestii he, sti
Result
bring Predicate
✥✥✥❵❵❵❵❵
✥✥✥ ❵
he, sti CPhe,sti
✘✘❳❳❳
Result ✘✘
✘ ❳❳
Predicate OP1 C
✘✘❳❳❳❳
✘✘✘
✘ ❳❳
here for Mary to talk to t1
λxλs.[s is the state of x being here, and Mary talks to x in every possible world w s
that is compatible with the salient goals/intentions/desires]
On the other hand... if a purposive adjunct (like (36a)) is adjoined to the highest node:
(43)
hsti
vP
✭✭✭✭✭❤❤❤❤❤
✭✭✭✭ ❤❤❤❤
✭✭✭✭✭✭ ❤❤❤
❤❤❤❤
✭✭ ❤
vPhsti CPhsti
✥✥✥❵❵❵❵❵ ✥✥✥❵❵❵❵❵
✥✥✥
✥ ❵❵ ✥✥✥ ❵
DP vhesti for Mary to talk to him
✥✥✥❵❵❵❵❵
✁✁❆❆ ✥✥✥ ❵
we vhhstihestii VPhsti
✘✘✘❳❳❳❳
✘✘
✘ ❳❳
DP Vhesti
✘✘❳❳❳
✪❡
✪ ❡ ✘✘
✘ ❳❳
Max Vhhestihestii he, sti
Result
bring Predicate
here
...then it can only be construed as expressing a purpose related to the causing event
— i.e., the agent’s purpose.
A sentence like We brought Max here for Mary to talk to him can never get a parse that includes
a constituent like 45, because of the type mismatch:
(45) h??i
Result
Predicate
✥✥✥❵❵❵❵❵
✥✥✥ ❵
he, sti CPhs,ti
✥✥✥❵❵❵❵❵
Result ✥✥✥ ❵
Predicate for Mary to talk to him
here
The type mismatch can’t be overcome by means of the gapless adjunct modifying a small clause
as in (46a). Why?
(46) a. b.
hs, ti ??
✥✥✥✥✥❵❵❵❵❵
Result ✥ ❵
Predicate Vhhestihestii hs, ti
✘✘❳❳❳
✘✘✘ ❳❳ Result
✘ ❳
hs, ti CP hs,ti bring Predicate
✥✥✥✥❵❵❵❵❵ ✭✭✭✭❤❤❤❤❤❤
Result ✥✥ ❵ ✭✭✭✭
✭ ❤❤
❤
Predicate for Mary to talk to him Max here for Mary to talk to him
✟❍
✟
✟ ❍❍
Max here
... It follows from the thesis (Baker’s conjecture, or my variant) that the relevant verbs select
properties, not small clauses. So (46b) is impossible.
⇒ Purposive infinitivals without gaps can only be parsed in construction with a higher part
of the VP than the result predicate.
⇒ Consequently they can only be interpreted as being related to a causing event, not the
result state.
⇒ In contrast, Purposive infinitivals with gaps can only be parsed in construction with the
very lowest part of the VP — namely the result predicate.
5. A remaining problem: Stative Predicates
While this approach to the differences between PCs and RCs explains the patterns of entailments
(about agentive intentions) in agentive sentences like (47):
... it does not explain why the same pattern holds in stative sentences like (48) and (49):
The gapless adjuncts in the (b) sentences above attribute agentive intentions (to someone named
or unnamed).
But these sentences involve plain stative predicates. Why can’t the gapless adjuncts in these
examples express goals relevant to the states described by constituents like [Max here], as in the
following structure:
5.1. A suggestion
⇒ Instead, stative predicates combine with a (quasi-)eventive be (Rothstein 1999), hence are
not saturated in their maximal projections.
(51) VPhsti
✘✘❳❳❳
✘✘
✘ ❳❳
DP Vhesti
❅ ✏✏PPP
❅ ✏✏ P
tM ax Vhhestihestii he, sti
Stative
is Predicate
here
⇒ Rothstein argued that be introduces a situating event argument, and relates it (and the
external argument) to the state-predicate that it selects as its complement.
Note that Baker [2003] comes to essentially the same conclusion.
⇒ Crucially for our purposes, this situating event can be understood as having an
inherent goal (whether or not the goal is held by the external argument).
5.2. Evidence in favor of the Rothstein/Baker suggestion
around
Surprisingly, a gapless purpose clause turns out to be possible in the environment that
Rothstein identified as a true stative small clause:
Imagine the following is part of the conversational background: I love swatting flies, and I get
bored if there are none around. Luckily (for me), I started using a new type of shampoo...
⇒ (54a) can express a desire not held by an agent of any causing event.
(55) V
✘✘❳❳❳
✘✘✘ ❳❳
V APhsti
✘✘❳❳❳
✘✘
✘ ❳❳
keeps APhsti CPhsti
✦❛❛ ✘✘❳❳❳
✦✦ ❛ ✘✘✘ ❳❳
flies around for me to kill them
• These facts suggests that Rothstein and Baker were right — there is no
‘predicate-internal subject’ in a copular stative sentence!
As expected, the deviant examples become good if the adjuncts contain gaps:
Where do we stand?
• We then saw that a few simple assumptions about the syntax of VPs can help explain the
correlations
⇒ Baker’s (2003) hypothesis, derived from work on Edo and Chichewa, that a large class of
transitive VPs have stative roots
⇒ A restrictive, compositional approach to the distribution of clauses with A-bar bound gaps
(null operator constructions)
What’s next?
• In what follows, we will see that there is a way to license a gap in a Rationale Clause
⇒ If there is movement in the main clause, a Rationale Clause can have a non-subject gap
⇒ This so-called parasitic gap construction is licensed by the same restrictive composition
principles that otherwise disallow RCs from containing gaps
• The “epilogue” will be a brief look at the puzzle concerning gapped infinitives with too/enough
7 EXTENSIONS (1) PARASITIC GAPS
There is one well-known exception to the generalization that a high VP-adjunct cannot contain
a gap:
(59) *We brought Max here [in order for Mary to talk to t i ]
(60) Who did you bring t i here [in order for Mary to talk to t i ]
In earlier work (Nissenbaum 1998, Nissenbaum 2000) I have argued that these so-called ”Parasitic
Gaps” (PGs) are derived by the same restrictive composition principles that (as we have already
seen) forces low VP-adjuncts (Purpose Clauses) to contain gaps. Specifically:
⇒ the movement in the matrix clause targets a propositional constituent the vP to derive a
predicate of individuals
(61) vP
✭✭✭✭✭✭❤❤❤❤❤❤
✭✭✭✭
✭
❤❤❤❤
❤
object1 vP
✭✭✭✭✭❤❤❤❤❤
✭✭✭✭ ❤❤❤❤
✭ ❤
vP Rationale
✥❵Clause
✘✘❳❳❳ ✥✥ ❵❵❵
✘✘
✘ ❳❳ ✥ ✥✥ ❵❵
λ1 vP λ2 for Mary to talk to t2
✏✏PPP
✏✏ P
subject V t1
⇒ Notice that these odd and exceptional instances of gapped high adjuncts provide
rather striking support for the general approach and conclusions of Part One.
⇒ The mechanisms that we seemingly must attribute to the mental grammar in order to
explain the distribution and interpretation of non-subject gaps in Purpose Clauses, predict
this exceptional behavior in Rationale Clauses (and other high-attached adjuncts.
8 Extensions (2): TOO and ENOUGH
Recall that the degree operators too and enough license an infinitival clause that would otherwise
not be able to appear with adjectival predicates:
(62) a. This snow is light enough [for me to shovel (it)] too/enough clauses
b. Frank is too angry [for me to talk to (him) now]
cf. *#This snow is light for me to shovel (it)
. *#Frank is angry for me to talk to (him)
(Since too and enough pattern together, I will stick with too for the remainder of the section.)
⇒ The puzzle with which we began is that these degree infinitivals seem to have completely
optional gaps.
⇒ Given the conclusions that we have drawn thus far, this optionality is not just puzzling it
should be downright impossible. Is our ’restrictive’ compositional theory wrong??
No!! Complex degree phrases headed by too show scope ambiguities with intensional operators
(Heim 2001). In joint work with Bernhard Schwarz (Nissenbaum and Schwarz 2011), we showed
that these scope ambiguities are limited to gapless degree phrases. The gapped versions are
restricted to an in situ interpretation.
Here’s some corroborating evidence that the optionality is merely an illusion that is somewhat
simpler to illustrate than the scope ambiguities, and which is compatible with the reasoning that
we saw in previous sections.
Faraci (1974) observed that there is a quite sharp restriction on the gaps that can appear in
degree phrases. Namely:
⇒ The gap in an infinitival with too must be anaphoric to the subject of the AP containing
too.
In (63b), Otis is the subject of the VP, not the AP containing too, and is sharply deviant in
comparison to (63a), in accord with Faraci’s generalization.
Similarly, the gap in (64) can only be construed with Frank as its antecedent; it is utterly
impossible to understand (64) with Mary as the antecedent.
8.2 A striking exception to Faraci’s Generalization
It turns out that there is a way to “get around” Faraci’s Generalization: namely, when there is
movement (Nissenbaum and Schwarz 2011):
(65) Mary1 , who John is too angry at t1 [for me to talk to now] ...
⇒ Under the very same conditions that license a parasitic gap, an antecedent that would
be otherwise unavailable for a gap in a too-infinitival becomes possible.
⇒ The gap in an infinitival with too must be anaphoric either to the subject of the AP
containing too, or to a non-subject antecedent that has undergone movement.
Towards an explanation
⇒ If Nissenbaum and Schwarz (2011) were correct to conclude that gapped degree-phrases
with too are restricted to an in situ interpretation, then movement of an otherwise un-
available antecedent, as in (42faraci3), would derive a predicate of individuals and thereby
provide an attachment site for the degree phrase to raise to — in effect, similar to parasitic
gap licensing in a Rationale Clause.
References:
1. E Bach, 1982. “Purpose clauses and control”. In The nature of syntactic representation, ed. P
Jacobson and GK Pullum, Reidel, Dordrecht, pp. 35-57.
2. M Baker, 2003. Lexical categories: Verbs, nouns, and adjectives. Cambridge Studies in Linguistics
102, Cambridge, U.K., Cambridge University Press.
3. M Browning, 1987. Null operator constructions. PhD Diss, MIT. Cambridge, MA: MITWPL.
4. N Chomsky, 1977. “On wh-movement”. In Formal syntax, ed. P Culicover, T Wasow & A Akmajian.
New York: Academic Press.
5. N Chomsky, 1986, Barriers. Cambridge, MA: MIT Press.
6. D Embick, 2004. “The structure of resultative participles in English”. Ling Inquiry 35: 355-392.
7. R Faraci, 1974. Aspects of the grammar of infinitives and for -phrases. PhD diss, MIT. Cambridge,
MA: MITWPL.
8. R Gonsalves, 2008. ‘Constructional meaning and lexical meaning’, in LACUS Forum 34: Speech and
Beyond, ed. P Sutcliffe, L Stanford & A Lommel, Houston, TX, 71-74.
9. J Hintikka, 1969. ‘Semantics for propositional attitudes’, in Philosophical Logic, ed. JW Davis, et
al. Reidel, Dordrecht, 21-45.
10. A Huettner, 1989. Adjunct infinitives in English. PhD diss, Univ of Massachusetts, Amherst.
11. J Huitink, 2005. Anankastic conditionals and salient goals. In Proc. Sinn und Bedeutung 9.
12. C Jones, 1985. Syntax and thematics of infinitival adjuncts. PhD dissertation, University of
Massachusetts. Amherst, MA: GLSA.
13. A Kratzer, 1981. “The notional category of modality”. In Words, Worlds and Kontext, ed. H
Eikmeyer & H Rieser. Berlin, New York: Walter De Gruyter: 825-834.
14. R Larson, 1988. “On the double object construction”. Linguistic Inquiry 19: 335-391.
15. B Levin & M Rappaport, 1986. “The formation of adjectival passives”. Ling Inq 17: 623-661.
16. A Marantz, 1997. “No escape from syntax”. In UPenn Working Papers in Linguistics 4.2, Pro-
ceedings of the 21st Annual Penn Linguistics Colloquium, 201-225.
17. J P Mora Gutiérrez, 2001. “Directed motion in English and Spanish”. Estud Ling Espanola 11.
18. J Nissenbaum, 2000. Investigations of covert phrase movement. PhD dissertation, MIT (published
by MITWPL).
19. J Nissenbaum, 2005. “Kissing Pedro Martinez: (existential) anankastic conditionals and
rationale clauses”. In the Proceedings of SALT XV, ed. E Georgala & J Howell.
20. J Nissenbaum & B Schwarz, 2011. “Parasitic degree phrases”. Nat Lang Semantics 19.1: 1-38.
21. L Pylkkänen, 2002. Introducing arguments. PhD diss, MIT. Cambridge, MA: MITWPL.
22. S Rothstein, 1999. “Fine-grained structure in the eventuality domain: the semantics of predicative
adjective constructions and Be”. Natural Language Semantics 7: 347-420.
23. D Siegel, 1973. “Non-sources for Un-passives”. In Syntax and Semantics II, ed. J Kimball, New
York: Seminar Press.
24. T Wasow, 1977, “Transformations and the lexicon”. In Formal syntax, ed. P Culicover, T Wasow
& A Akmajian. New York: Academic Press.
25. E Williams, 1983. “Against small clauses”. Linguistic Inquiry 14.2: 287-308.