A cross-modal view of VP ellipsis
Carlo Cecchetto
University of Milan-Bicocca and
Structures Formelles du Langage (CNRS & Université Paris 8)
Based on joint work with: Alessandra Checchetto, Carlo Geraci, Mirko
Santoro & Sandro Zucchi
1
PART I.
VP ELLIPSIS EXISTS IN LIS
2
SEARCHING FOR VP ELLIPSIS
VP ellipsis was initially identified only in English:
John broke a vase, and Mary did too
Subsequent research (cf. Goldberg 2005) claimed that other languages
(Hebrew, Irish, and Swahili a.o.) show VP ellipsis, although it is
superficially different from the English type. In English, as we just saw,
the verb is missing together with other VP internal material. In Hebrew, in
which both V-to-I movement and VP deletion occur, a finite verb survives
VP ellipsis.
I am going to claim that LIS has the same type of VP ellipsis as English.
3
CANDIDATES FOR VP ELLIPSIS IN LIS
In LIS a constituent can go un-uttered if a suitable antecedent is present.
This elliptical construction involves typically, although not uniquely, the
use of an adverbial sign glossed here as SAME, meaning “too/as well”.
(1) DINING-ROOM GIANNI VASE BREAK, MARIO SAME
“Gianni broke a vase in the dining room and Mario did so, too”
The same happens with negation:
(2) DINING-ROOM GIANNI VASE BREAK. PIERO SAME NOT
“Gianni broke a vase in the dining room while Piero did not”
4
STRIPPING?
However (1) and (2) are only weak evidence for a VP ellipsis analysis. An
analysis in terms of stripping, captured by the English translation in (ii). In
principle cannot be excluded:
(1) DINING-ROOM GIANNI VASE BREAK, MARIO SAME
(i) “Gianni broke a vase in the dining room and Mario did so, too”
(ii) “Gianni broke a vase in the dining room. Mario, too”
5
IT IS NOT STRIPPING, I
We will use three diagnostics to set apart VP ellipsis and stripping. The
first diagnostic is the presence of an auxiliary-like element in the elliptical
clause. For example, in (3), an auxiliary sits in the elliptical clause, much
like in classical VP ellipsis cases.
(3) GIANNI BOOK BUY MUST. MARIA MUST SAME
‘Gianni must buy a book and also Maria must (buy a book)’
6
IT IS NOT STRIPPING, II
In English VP ellipsis, but not stripping, can occur in subordination (cf.
Lobeck 1995):
(4)
(5)
John left. Bill thinks that Piero did too
* John left. Bill thinks that Piero too
In LIS, predicate ellipsis can occur in a subordinated clause, consistent
with a VP ellipsis analysis.
(6) GIANNI MARIAi LOVE. IX-3i THINK PIERO SAME
'Gianni loved Maria. She thinks that Piero did too'
7
IT IS NOT STRIPPING, III
In English VP ellipsis, but not stripping, allows for backward anaphora (cf.
Lobeck 1995).
(7)
(8)
John didn’t, but Mary bought books
*Mary too and John bought books
In LIS, predicate ellipsis can be licensed by backward anaphora,
consistent with a VP ellipsis analysis.
(9)
8
____________if
IF PIERO NOT GIANNI GO
'If Piero does not, Gianni will go'
TWO BIG ISSUES IN THE VP
ELLIPSIS LITERATURE
9
FIRST ISSUE. PHONOLOGICAL DELETION OR SEMANTIC
COPYING?
 Phonological deletion approach: a full-fledged VP is present in syntax,
although it is deleted at PF.
 Semantic copying approach: a silent proform is generated in syntax
and is interpreted in the semantic component as having the same
meaning as the antecedent VP (the ellipsis site contains no structure)
10
SECOND ISSUE: RECOVERABILITY OF THE CONTENT OF
ELLIPSIS.
 Semantic Identity Condition: recoverability is licensed by identity in
meaning.
 Syntactic Identity Condition: recoverability is licensed by identity in
morpho-syntactic form. Under this approach a category can go
unuttered only if it has the same syntactic structure and the same
lexical composition as its linguistic antecedent. It should be clear that,
if the identity in form approach is assumed, semantic identity still
holds, but it is only derivative.
11
THE TWO ISSUES ARE NOT ONE AND THE SAME
The issue of recoverability and the issue whether the ellipsis site contains
structure are connected but distinct. One can argue that recoverability
requires identity in syntactic structure only if (s)he is willing to concede
that the ellipsis site contains structure. But one can argue that for
recoverability identity in meaning suffices even if the ellipsis site contains
structure.
12
DECIDING ISSUE 1, LONG DISTANCE DEPENDENCIES IN ENGLISH
It is possible to extract a wh-phrase from an ellipsis site (cf. Baltin 2012,
Chao 1987, Fiengo and May 1994 and Tancredi 1992):
(10) I know which person Mary talked to and which person Bill didn't.
This is a challenge for the semantic copying approach. If the ellipsis site
contains a null anaphora, it is difficult to explain why (10) is grammatical,
since (overt) anaphors do not have internal structure, as confirmed by the
ungrammaticality of (11).
(11) *I know which person Mary talked to and which person Bill didn't do
so
On the other hand, the phonological deletion approach can easily explain
why only (10) is grammatical. If the VP is present in the syntactic
component, it can contain the trace of the wh-phrase.
13
DECIDING ISSUE 1, LONG DISTANCE DEPENDENCIES IN LIS
In LIS as well the ellipsis site can contain a trace.
___wh
___wh
(12) IN-THE-PAST GIANNI MEET WHO I-KNOW BUT FUT WHO I-KNOW NOT
'I know who Gianni met in the past but I do not know who he will meet in the
future'
(12') … BUT [CP [IP pro FUT [VP ti MEET] ] WHOi ] I-KNOW NOT

PHONOLOGICAL DELETION APPROACH IS FAVORED
14
PART III.
WHAT CAN WE LEARN FROM LIS
ABOUT VP ELLIPSIS IN
GENERAL?
IDENTITY IN MEANING IS NOT
ENOUGH
15
ADVERB INCORPORATION AS A TEST
In many sign languages, including LIS, manner adverbs can either be
incorporated in the verb (cf. 13a) or appear as an independent sign (cf.
13b). In (13a) the movement of the dominant hand towards the mouth of
the signer, characteristic of the sign EAT, is repeated and is articulated
more rapidly than in the citation form of the verb.
(13) GIANNI EAT-QUICKLY
(14) GIANNI EAT QUICKLY
Gianni eats quickly
16
ADVERB INCORPORATION SHOWS THAT IDENTITY IN MEANING IS
NOT ENOUGH
If identity in meaning were enough to license SAME-ellipsis, sentences
like (15) and (16) should be on a par with each other, since the
antecedent clause in (15) and (16) expresses the same meaning, despite
the fact that (15) contains an independent sign for QUICKLY while in (16)
QUICKLY is incorporated. However, (16) is sharply ungrammatical.
(15) MARIO MEAT EAT QUICKLY. GIANNI SAME SLOWLY
“Mario eats meat quickly. Gianni does that slowly”
(16) *MARIO MEAT EAT-QUICKLY. GIANNI SAME SLOWLY
If identity in form is required, there is an explanation for why (16) is out:
since QUICKLY is incorporated into the verb in the antecedent clause,
there is a clash in meaning (one cannot eat slowly and quickly at the
same time). In (15) there is no clash because the ellipsis site is identical
to the verb EAT alone.
17
DIGRESSION: SIMILAR CASES IN SPOKEN LANGUAGES?
A similar point can be made for English:
(17) Abby took the exam again while Ben did for the first time
(18) ?? Abby retook the exam while Ben did (that) for the first time
(Jason Merchant, p.c.)
18
ANOTHER CASE WHERE IDENTITY IN MEANING IS NOT ENOUGH:
ROLE SHIFT AND THE STRICT/SLOPPY READING
In simple cases, both the strict and the sloppy reading are easily
detectable in VP ellipsis in LIS.
(19) GIANNIi SECRETARY HISi VALUE. PIERO SAME
Strict reading:
 Gianni values his own secretary. Piero values Gianni’s secretary
Sloppy reading:
 Gianni values his own secretary. Piero values Piero’s secretary
The situation is more complex in sentences involving role shift.
19
ROLE SHIFT
Role shift is a strategy common across sign languages in which the
signer shifts into the role and adopts the perspective of the quoted person
by slightly shifting the body position, changing the position of the head,
and breaking the eye contact with the addressee.
(20) and (21) are synonymous but role shift takes place only in (23). As a
consequence, the subject of the embedded clause in (23) has the form of
a first person pronoun.
(20) GIANNIi SAY IX-3i MARIA KISS
“Gianni said that he kissed Maria”
________________________rs
(21) GIANNIi SAY IX-1_rs MARIA KISS
“Gianni said that he kissed Maria”
20
ROLE SHIFT AND THE STRICT/SLOPPY READING
The elliptical clause that follows an antecedent clause without role shift is
ambiguous between a strict and a sloppy reading, much as its English
counterpart.
(22) GIANNIi SAY IX-3i MARIA KISS. PIEROj SAME
“Gianni said that he kissed Maria. Piero did too”
Strict reading:  Gianni said that Gianni kissed Maria. Piero said that
Gianni kissed Maria
Sloppy reading:  Gianni said that Gianni kissed Maria. Piero said that
Piero kissed Maria
21
ANOTHER CASE WHERE IDENTITY IN FORM IS REQUIRED: ROLE
SHIFT BLOCKS THE STRICT READING
If role shift takes place in the antecedent clause, only the sloppy reading
surfaces.
________________________rs
(23) GIANNIi SAY IX-1_rs MARIA KISS. PIEROj SAME
Strict reading: *Gianni said that Gianni kissed Maria. Piero said that
Gianni kissed Maria
Sloppy reading: Gianni said that Gianni kissed Maria. Piero said that
Piero kissed Maria
This is another argument showing that identity in meaning is not enough.
The elliptical predicate must take as an antecedent a predicate which has
the same form of the antecedent, namely a predicate where role shift has
taken place. This happens even if the predicate with and without role shift
have the same meaning.
22
PART IV.
WHAT CAN WE LEARN FROM LIS
ABOUT VP ELLIPSIS IN
GENERAL?
IDENTITY IN FORM IS NOT
ENOUGH
23
SPEED AND AMOUNT READINGS
(24) GIANNI BEANS EAT-QUICKLY
Sentences like (24) in which the adverb is incorporated into the verb are
ambiguous.
 “Speed reading”: Gianni eats beans quickly (but he does not
necessarily eats a big quantity of beans).
 “Amount reading”: Gianni eats a lot of beans.
If the adverb is not incorporated, only the “speed reading” is selected.
24
ADVERB INCORPORATION SHOWS THAT IDENTITY IN FORM IS
NOT ENOUGH
Crucially, in (24) the reading that obtains in the antecedent clause must
obligatorily obtain in the SAME clause.
(24) GIANNI BEANS EAT-QUICKLY. PIERO SAME
If the “speed reading” is selected in the antecedent clause, the SAME
clause receives the “speed reading” (Piero eats quickly).
If the “amount reading” is selected in the antecedent clause, the SAME
clause receives the “amount reading” (Piero eats a lot of beans).
A semantic parallelism that goes beyond identity in form seems to be
required here.
25
PART V.
WHEN IDENTITY IN FORM IS NOT
THAT STRICT…
26
VEHICLE CHANGE IN ENGLISH, I
(25) *Mary admires Johni and hei does, too.
(25) is ungrammatical, since the elided VP contains a referential
expression (“John”) c-commanded by a coindexed pronoun, triggering a
Principle C violation.
However, (26) is OK, suggesting that the referential expression “John”
can undergo vehicle change, namely a pronoun replaces the referential
expression “John” in the elided VP but preserves its indexical information:
(26) Mary admires Johni, and hei thinks that Sally does, too.
(26’) Mary admires Johni, and hei thinks that Sally does admire himi, too
27
VEHICLE CHANGE IN ENGLISH, II
Note that vehicle change cannot save (24). If vehicle change applies, the
Principle C violation is obviated but a Principle B violation is triggered:
(24) *Mary admires Johni and hei does, too.
(24’) *Mary admires Johni and hei does admire himi, too
28
VEHICLE CHANGE IN LIS
(27)
a. IX-1 GIANNI MARIA PIERO MEET …
‘I met Gianni, Maria and Piero'
b. *GIANNI MARIAi PRAISE. IX-3i SAME
c.  GIANNI MARIAi PRAISE. IX-3i THINK PIERO SAME
“Gianni praised Maria and she thinks that Piero did too”
(27c) is grammatical and can mean that Gianni praised Maria and she
thinks that Piero praised her (=Maria). This reading requires vehicle
change:
(27c’) GIANNI MARIAi PRAISE. IX-3i THINK PIERO IX-3i PRAISE SAME
29
CONCLUSION
We just started scratching the surface of the investigation of the
properties of VP ellipsis in the visuo-spatial modality. Still, our preliminary
exploration suffices to show that:
(i) sign languages replicate the complex pattern which is presented by the
phenomenon in spoken languages.
(ii) some diagnostics that can be applied to sign languages only (or can
be applied to sign languages to an higher degree than they can be
applied to sign languages) may elucidate questions on the nature of VP
ellipsis that are difficult to decide if sign languages are left out of the
picture.
30
FOR FULLER REFERENCES AND A MORE ACCURATE ACCOUNT:
C. Cecchetto, A. Checchetto, C. Geraci, M. Santoro & S. Zucchi, (2015)
“The Syntax of Predicate Ellipsis in Italian Sign Language (LIS)”, Lingua,
166: 214-235.
FOR VIDEOS OF THE LIS SENTENCES
https://sites.google.com/site/thegrammaroflis/home/scientificpapers/syntax/the-syntax-of-ellipsis-in-lis
31