Deriving Island Constraints With Searle and Grice. A Pragmatic Account of Bridge Verbs

DERIVING ISLAND CONSTRAINTS WITH
SEARLE AND GRICE.

A PRAGMATIC ACCOUNT OF BRIDGE
VERBS
Sonja M€
uller
Abstract. This paper proposes a pragmatic account of restrictions on extractions

from that-complement clauses. For that-complements, it is well known that their
transparancy for extraction gets influenced by the matrix verb selecting them:
So-called bridge verbs allow extraction from their complement, non-bridge verbs
do not. Analysing the different context changes induced by the extraction
constructions, an account is developed which proves a violation of very basic
pragmatic principles (Searle’s 1969 felicity conditions for questions and Grice’s
1975 Cooperative Principle (plus Conversational Maxims)) under the occurrence of
the negatively influencing factive, implicative and manner-of-speaking verbs.
Evidence for an account of this nature in general as well as the particular one
proposed in this paper comes from the observation that manipulating contextual
factors in the shape of subtypes of the illocutionary force, the questions’ intentions
and intonational patterns has an impact on the structures’ acceptability.
1. Introduction
At least since Ross (1967) formulated his island constraints, it has been
known that long wh-movement (also termed extraction) is restricted to
certain domains in a sentence. Subjects (cf. (1a)) or complex noun
phrases (NP) (cf. (1b)) e.g. are classically counted among Ross’ islands.
(1) a. *Was ist es unwahrscheinlich [dass Peter was kaufte]?
What is it unlikely that Peter what bought
‘*What is it that Peter bought unlikely?’
b. *Was diskutierte sie [[die Behauptung], [dass er was
What discussed she the claim that he what
getan hat]]?
done has
‘*What did she discuss the claim that he had done?’
(English data taken from Erteschik-Shir 1973:100 [(1a)] and
Erteschik-Shir & Lappin 1979:62 [(1b)])
One extraction domain which has raised particular interest in the
research on such constructions are that-complement clauses. It is also
already mentioned in Ross’ (1967) pioneering work in this field of
research that the matrix verb occurring decides whether the domain is
transparent or opaque for extracting a constituent. The generalisation
reached already in the 70s (the decisive piece of work to be named here is
Studia Linguistica 69(1) 2015, pp. 1–57. © 2014 The Editorial Board of Studia Linguistica
Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden,
MA 02148, USA
2 Sonja M€
uller
Erteschik-Shir 1973) and stuck to in the decades of research to come

(cf. e.g. Fukui 1986, Stowell 1986, Rizzi 1990, Kluender 1991, De Cuba
2006, Erteschik-Shir 2007) is that factive as well as manner-of-speaking
verbs prevent moving the wh-phrase to the front. German is no exception
to this assumption: From (2a) to (2b), a degradation can be perceived.
(2) a. Wen meint/ glaubt/ sagt/ will Peter, dass der Chef wen
Who thinks/ believes /says/ wants Peter that the boss who
asst?1
entl€
fires
b. ??Wen ignoriert/ verdr€angt/ fl€
ustert/ schreit/ verursacht/
Who ignores/ represses/ whispers/ shouts/ causes/
beweist Peter, dass der Chef wen entl€asst?2
proves Peter that the boss who fires
As extraction domains in general have predominantly been examined in
terms of syntactic principles, there is also an extensive amount of
syntactic work on this subject in particular (cf. e.g. P€
utz 1975, Cattell
1978, Cinque 1990, Rizzi 1990, De Cuba 2006, Basse 2008).3 However, in
the tradition of less abundantly developed semantic and information
structural accounts (cf. e.g. Erteschik-Shir 1973, 2007, Kluender 1991,
Szabolcsi & Zwarts 1993, Truswell 2007, Abrusan 2008),4 my paper will
1
As is well-known, languages differ as to the acceptability of extraction constructions. In
order to circumvent awkward translations into English, I’m only providing English glosses
and no paraphrases in case no direct translation is possible (due to constraints on selection
[such as (non-)finite complements, tense, mood etc.]). If the reader has the impression that
some examples might be possible in his/her mother tongue, s/he should make sure that the
construction displays the exact reading I’m proposing. It does not suffice to find a
construction which has the same surface structure.
2
Throughout this article, I’m making a distinction between four degrees of acceptability
(*, ??, ? and no sign). However, I don’t want to make any absolute judgements. All I’d like
to claim is that there are better and worse sentences among the sentences looked at. It has
been known for a long time that there is ideolectal, dialectal as well as social variation and
that the differences between extraction constructions are gradient rather than categorial (cf.
e.g. Grewendorf 1988:8, Fanselow 1987:51f., Adli 2004).
3
This does not mean that the accounts do not appeal to semantic/pragmatic ideas.
Cinque’s (1990) concept of referentiality or Cattell’s (1978) predicate classes are certainly
semantic/pragmatic in nature. It is, therefore, more accurate to say that these pieces of work
argue that the semantic/pragmatic differences are reflected in the syntax.
4
Due to lack of space and the focus of this article, I cannot offer a discussion of those
accounts and the many more that have been suggested. I have discussed them somewhere
else, however. The interested reader is, therefore, referred to M€ uller (2011a: chapter 3),
(2011b).
© 2014 The Editorial Board of Studia Linguistica

Deriving island constraints with Searle and Grice 3
present a non-structural approach to (im)possible extractions from that-

complement clauses.5
In particular, my analysis will trace the negative influence of certain
matrix verbs back to the effect that the resulting questions have on
discourse, that is analyses their roles within communicative settings. The
merrits of choosing this perspective are that it does not only allow a
uniform analysis6 of the negative influence that is ascribed to factive and
manner-of-speaking verbs (in the literature they are always treated as two
separate phenomena), but also that it allows broadening the view of
relevant classes of negatively influencing embedders to the class of
implicative verbs (cf. Karttunen 1971, cf. e.g. verursachen (cause), beweisen
(prove) in (2b)) and to smoothly integrate their impact on the extraction
constructions in the suggested argumentation.7 Relying on very basic and
acknowledged pragmatic principles such as Searle’s (1969) felicity
conditions for speech acts as well as Grice’s (1975) Cooperative Principle
(plus Conversational Maxims), the overall claim with respect to degraded
cases of extractions from that-complements is that the questions formed
5
It is a well-known phenomenon that certain extraction domains are transparent for
certain types of moved phrases and opaque for others. The first (more selective) ones are
generally called weak islands, the other (stricter) ones strong islands.
It is also true for the domain of that-complements looked at here that authors consider
them weak islands (cf. e.g. Hegarty 1990, Szabolcsi & Zwarts 1993, Oshima 2007 on English,
Rizzi 1990, Cinque 1990 on Italian). To my knowledge, nobody has ever explicitly stated
this assumption for German data and starting off with the assumption that the object-
extractions in (2) are the ‘good’ cases would simply be wrong. I’ll simplify the picture in the
following and ignore this issue throughout by focusing on object-extractions. Once having
established an analysis for the data in (2), one might go on and see how it might be possible
to integrate those further complications. However, sorting out the data should be the first
aim here, too.
The account presented here thus focuses on properties of the extraction domain and leaves
issues concerning the extractee’s mobility untouched – a topic which is subject in syntactic
(cf. Rizzi 1990, 2004, De Cuba 2006, Basse 2008) as well as non-structural accounts
(cf. Comorovski 1996, Oshima 2007, Abrusan 2009). All I can say right now is that factors
that are mentioned in this context (D-Linking (Rizzi 1990), the claim (made more or less
explicitly in Comorovski 1996, Oshima 2007, Abrusan to appear) that the argument
reference must be > 1 or studies of the extractee’s denotation domain (cf. Szabolcsi & Zwarts
1993, Abrus an 2009)) do at least seem to be compatible with a discourse structural
perspective as it is chosen here.
It is also well-known that the picture is even more complicated when taking into account
different types of A’-movement such as relativisation, topicalisation etc. However, as this
paper shows there are enough questions to be answered on object wh-movement from that-
complement clauses so that unfortunately I cannot provide analyses for all of these other
types and unfortunately I also do not have the space to discuss data for all of these types
either.
6
I call my approach a “uniform analysis” as all the data presented is covered by a
pragmatic analysis (Gricean maxims) – even if my analysis appeals to two different maxims
(quantity and quality).
7
Surprisingly, this class of verbs has not been mentioned in the extraction literature to
play a role at all. This is probably due to the fact that most of the research has been
conducted on English. In English, however, prototypical implicative verbs such as cause or
effect do not select for finite that-clauses.

4 Sonja M€
uller
present inadequate moves in discourse in the following sense: In the case

of factive and implicative verbs, the questions’ uninformativity will be
demonstrated. The occurrence of manner-of-speaking verbs will be
argued to cause the questions’ over-informativity. Crucial evidence for
the hypothesis put forth is that it can be observed that the relative
acceptability of the extractions at issue does not only depend on the verb
constituting the matrix, but that further linguistic factors such as lexical
material (namely, modal particles) or particular intonational contours
(namely, the so-called hat contour) as well as extra-linguistic factors in the
shape of contextual information do have an impact on those constructions
that – going only by their matrix verbs – should be ruled out as
unacceptable. That is, there are questions containing (putative) non-bridge
verbs which are acceptable nonetheless. My account will argue that those
further means contribute to subtly changing a question’s meaning which
again leads to dissolving the uninformativity or, respectively, over-
informativity argued for under default meaning assignment and by
assuming a setting perceived of as a ‘zero-context’.
To develop the analysis, the paper is organised as follows: Section 2
will introduce some background assumptions concerning the formal
modelling of discourse structure (sections 2.1, 2.2). It will introduce the
tools made use of in the actual analysis of (un)acceptable extraction
constructions in section 3, involving a discussion of the impact that the
meaning components of the different verb classes involved in the study
have in discourse structural terms (section 2.3). Section 3.1 deals with
factive and implicative verbs in extraction constructions, section 3.2 and
3.3 introduce the further factors mentioned above which are analysed as
yielding different question intentions. Section 3.4 discusses issues
concerning a reading of the questions under discussion made in section
3.1. Section 3.5 integrates manner-of-speaking verbs into the picture
developed up to this point. Section 4 concludes the discussion.
2. Contexts and their formal modelling

Linguistic utterances obviously do not occur in isolation, but within
larger contexts. Certain persons (speaker, addressee) who are shaped by
their views about the world in general and the topic of the conversation
in particular participate. Conversation occurs at a place, at a certain
time, further utterances already precede. Linguistic phenomena such as
e.g. reference of deictic pronouns or coreference relations provide ample
evidence for this assumption.
2.1. Simple sentences: Assertions and constituent questions

Since Stalnaker (1978) diverse assumptions on a formal representation of
an utterance context have been made (cf. e.g. Roberts 1996, Giannakidou
1998, Bartels 1999, B€uring 2003, Gunlogson 2003, Caponigro & Sprouse
2007, Portner 2005, Farkas & Bruce 2009). The central concept in the
characterisation of a context in Stalnaker (1978) as well as in the other
approaches mentioned here is the Common Ground (CG) which can be
modelled as a set of propositions which represent the assumptions the
discourse participants knowingly mutually agree on. On the one hand,
mutual belief can be achieved by tacit assumption, on the other hand,
speakers come to agree on certain pieces of information in the course of a
conversation. Assuming with Possible World Semantics that a proposi-
tion is associated with the set of worlds in which it is true, the context
also contains the Context Set (CS) which represents the set of worlds in
which all CG-propositions are true.
Stalnaker (1978:322) writes about the purpose of communication: “To
engage in a conversation is, essentially, to distinguish among alternative
possible ways that things may be. The purpose of expressing propositions
is to make such distinctions.” If an assertion is uttered (cf. e.g. (3)), this
aim is attempted directly. Unless the addressee rejects the proposition
expressed, it is added to the CG and worlds in which the CG-propo-
sitions are not valid are eliminated from the CS (cf. (4)).
(3) Anna invites Stephan for coffee. (= p1)
(4) input context output context
CG = { } CG’ = {p1}
CS = W CS’ = CS \ {w2W | p1(w) = 1}
If a constituent question is uttered, it opens up a restricted number of
alternatives which the answer (in most cases an assertion) ideally reduces
to one possibility. Relying on Partition Semantics for the semantic
assumptions on questions (cf. Groenendijk & Stokhof 1984, 1997,
Higginbotham & May 1981, Higginbotham 1991, 1996), a wh-question
such as (5) induces a space of answers as in (6). (For the purpose of the
illustration, the discourse domain is restriced to two individuals here.)8
Each cell corresponds to a proposition (or, respectively, a set of worlds)
which represents a possible complete (within this type of theory strongly
exhaustive) answer.9
8
A partition is a mathematical concept and is defined as follows:
(i) Given a nonempty set A, a partition of A is a collection of nonempty subsets of
A such that
1. For any two distinct subsets X and Y, X \ Y = ∅.
2. The union of all the subsets in the collection equals A. Wall (1972:121)
9
On alternative approaches to the semantics of questions cf. e.g. Karttunen (1977), Heim
(1994), Dayal (1996), Krifka (2001), Reich (2003).

6 Sonja M€
uller
(5) Who does Anna invite for coffee?, D = {Julia, Stephan}

(6)
Anna invites Julia as well as Stephan for coffee. (= p1)
Anna invites Julia, but does not invite Stephan for coffee. (= p2)
Anna invites Stephan, but does not invite Julia for coffee. (= p3)
Anna invites neither Julia nor Stephan for coffee. (= p4)
As (7) illustrates, the CG gets expanded with one possible answer. The
CS can be changed in as many ways as the CG. After an answer has been
given, an assertive context update follows.
(7) input context output context
CG = { } CG’ = CG [ {p1} ∨ CG [ {p2} ∨ CG [ {p3} ∨ CG [ {p4}
CS = W CS’ = CS \ {w2W | p1(w) = 1}∨ CS \ {w2W | p2(w) = 1}∨
CS \ {w2W | p3(w) = 1}∨ CS \ {w2W | p4(w) = 1}
A demand that an adequate proposition has to meet in a context is its

informativity (cf. (8)), i.e. that by adding a proposition, the information
state of discourse is supposed to change ((cf. B€ uring 1997:27).
(8) CS \ {w2W | p(w) = 1}6¼ CS
For an informative assertion, (8) means that it should not intend to add a
proposition that is already in the CG. A wh-question should not open up
any of the possibilities which the CS earlier has already been reduced of
(cf. e.g. (9)).
(9) Informativity: [. . .], don’t ask for known things! (B€
uring 2003:5)
2.2. Complex sentences: two-stage context changes

If a complex sentence such as (10) is introduced into the context, the
context update takes place on two levels: On the level of the main
context, the sentence has the effect that it is added to the CG that Julia
has the belief of Anna inviting Stephan for coffee (= p2) by simulta-
neously deleting those worlds from the CS in which her doxastic system
does not contain this belief.
(10) [p2 Julia thinks [p1that Anna invites Stephan for coffee]].
However, the matrix verb also introduces a further embedded context. In
contrast to (3), the truth of p1 is not evaluated with respect to the
speaker who, therefore, wants it to become information shared by the
discourse participants, but with respect to the individual Julia. Farkas
(1992) refers to the individual with respect to whose system a
proposition’s truth value gets assigned as the individual anchor of a
proposition and she further assumes that the individual anchoring of
propositions happens indirectly: Propositions are anchored in worlds

which are again anchored in individuals. The proposition p1 in (10) is
thus true in the worlds which represent the actual world from Julia’s
perspective, i.e. the matrix subject Julia is the individual anchor of p1
because it is connected to the worlds relative to which p1 is evaluated. In
(3), however, p1 is true in the worlds which represent the actual world
from the speaker’s perspective, that is the speaker is the individual
anchor of p1 because s/he is connected with the worlds relative to which
p1 is evaluated. For the context update induced by a sentence such as
(10), the assumptions by Farkas mean that one has to assume further
sets of propositions besides the CG and the CS, namely, a set of
propositions containing the propositions that constitute Julia’s beliefs
(Di,w [= the set of propositions that an individual i believes in a world w]
in (11)) and corresponding worlds which make up Julia’s belief worlds
(Di,w [= the set of belief worlds of an individual i in a world w] in (11)).
In analogy to a CG/CS-update, p1 is added to Di,w and the worlds in
which p1 is not true are deleted from Di,w.
(11) main context Julia’s doxastic system
CG = { } Di,w = { }
CS = W Di,w = W
CG’ = {p2} D’i,w = {p1}
CS’ = CS \ {w2W | p2(w) = 1} D’i,w = Di,w \ {w2W | p1(w) = 1}
Assuming for complex sentences such as (10) that the context update
takes place on two levels, the context change induced by a complex
question such as (12) can be considered to look like (13).
(12) Who does Julia think that Anna invites for coffee?, D = {Stephan,
Caro}
By asking a question such as (12), the speaker aims at getting that
proposition that should be added to the set constituting Julia’s beliefs,
the proposition answering the root question Who does Anna invite for
coffee? (cf. the partition(s) in (13), cf. section 3.4 on this reading).
(13) main context Julia’s doxastic system
CG = { } Di,w = { }
CS = W Di,w = W
CG’ = CG[ {p1}∨ CG[ {p2}∨ D’i,w = Di,w[ {p5}∨ Di,w[ {p6}∨
CG[ {p3}∨ CG[ {p4} Di,w[ {p7}∨ Di,w[ {p8}
CS’ = CS\ {w2W | p1(w) = 1}∨ Di,w = Di,w\ {w2W | p5(w) = 1} ∨
CS\ {w2W | p2(w) = 1}∨ Di,w\ {w2W | p6(w) = 1} ∨
CS\ {w2W | p3(w) = 1}∨ Di,w\ {w2W | p7(w) = 1} ∨
CS\ {w2W | p4(w) = 1} Di,w\ {w2W | p8(w) = 1}

8 Sonja M€
uller
Differentiating between the effect that the extraction construction has on

the main and the embedded context, on the one hand, the question asks
for the proposition which can be added to the set containing Julia’s
beliefs (Di,w) (and thereby deleting Julia’s non-belief worlds from Di,w) by
opening up the possibilities in (14b). On the other hand, the question
concerns the update of CG and CS with one of the propositions out of
(14a). Of course, the two updates always have to be carried out in
parallel.10
(14) a. Julia thinks that Anna invites Stephan as well as Caro for coffee. = p1
Julia thinks that Anna invites Stephan, but does not invite Caro for coffee. = p2
Julia thinks that Anna invites Caro, but does not invite Stephan for coffee. = p3
Julia thinks that Anna invites neither Stephan nor Caro for coffee. = p4
b. Anna invites Stephan as well as Caro for coffee. = p5

Anna invites Stephan, but does not invite Caro for coffee. = p6
Anna invites Caro, but does not invite Stephan for coffee. = p7
Anna invites neither Stephan nor Caro for coffee. = p8
As will be shown in the next section, – depending on the matrix verbs

occurring – different situations can arise with respect to propositions
having to be anchored in the main and the embedded context.
2.3. Factive, implicative and manner-of-speaking verbs

As has been illustrated in section 1, the matrix verbs occurring have an
impact on the degree of acceptability that is attributed to the structure
resulting from extracting a constituent from the that-clause. A
degradation of the questions results in case the extraction takes place
from complements of factive, implicative and manner-of-speaking
verbs.
Factive verbs presuppose their complement, implicative verbs imply it.
As the negation test (which is one of the classical tests for detecting
10
Assuming that (12) induces the partitions in (14) and the context update in (13) means
attributing a certain meaning to the question which can be paraphrased as in (i).
(i) ‘What is the proposition that can be anchored in Julia’s doxastic system, the
possibilities
all being complete answers to the question: Who does Anna invite for coffee?’
This interpretation, however, is not the only one possible. It is also the reading possible
which can be paraphrased by (ii).
(ii) ‘For which individual is it the case that Julia thinks that Anna invites it for coffee?’
Although those two interpretations do not necessarily make a difference as far as the
meaning of an extraction construction under the occurrence of the matrix verb think is
concerned, the two different interpretations do indeed affect the meaning of such questions
under the occurrrence of different matrix verbs. Cf. sections 3.2.2, 3.3 and 3.4 for discussion
of those two readings.

presuppositions) shows in (15) and (16), the first class of matrix

predicates presupposes the truth of the proposition in the subordinate
clause, whereas the truth of the embedded proposition is implied by the
truth of the complete sentence under the occurrence of verbs of the
second class.
(15) a. Peter ignoriert/ verdr€angt/ verheimlicht, dass der Chef
Peter ignores/ represses/ conceals that the boss
Paul entl€asst. ? Der Chef entl€ asst Paul.
Paul fires the boss fires Paul
‘Peter ignores/represses/conceals (the fact) that the boss fires
Paul. ? The boss fires Paul.’
b. Peter ignoriert/ verdr€angt/ verheimlicht nicht, dass der
Peter ignores/ represses /conceals not that the
Chef Paul entl€ asst. ? Der Chef entl€asst Paul.
boss Paul fires The boss fires Paul
‘Peter does not ignore/repress/conceal (the fact) that the boss
fires Paul. ? The boss fires Paul.’
(16) a. Peter verursucht/ beweist, dass der Chef Paul entl€asst. ?
Peter causes/ proves that the boss Paul fires
Der Chef entl€ asst Paul.
the boss fires Paul
b. Peter verursacht/ beweist nicht, dass der Chef Paul
Peter causes/ proves not that the boss Paul
asst. ↛ Der Chef entl€
entl€ asst Paul.
fires the boss fires Paul
Factoring out manner-of-speaking verbs for the moment, acceptable
extraction constructions contain matrix verbs which neither presuppose
nor imply their complement (cf. (17)).
(17) Peter meint/ sagt/ glaubt/ will (nicht), dass der Chef Paul
Peter thinks/ says/ believes/ wants (not) that the boss Paul
asst. ↛ Der Chef entl€
entl€ asst Paul.
fires the boss fires Paul
‘Peter (does not) think(s)/say(s)/believe(s) that the boss fires
Paul/want(s) the boss to fire Paul. ↛ The boss fires Paul.’
As concerns the context update induced, authors assume that
presuppositions are propositions which have to be true in the CS
that exists before the sentence containing them is uttered and its
proposition considered true in CS’ (cf. Farkas 2003:4, Kadmon
2001:14, Heim 1992:186). In this argumentation, presuppositions are
understood as requirements on the context state before the sentence
containing the presupposition is introduced into the context. For (18),
this means that the CS has to contain worlds in which the embedded

10 Sonja M€
uller
proposition (=p2) is true when the CS gets reduced of the worlds in

which the complex proposition (=p1) is not true (cf. (19)). Otherwise,
the sentence is not defined relative to this context and a context
update cannot apply.
Based on this description, the context change induced by an utterance
such as (18), can be modelled as in (20).11
(18) [p1 Peter ignoriert, [p2 dass der Chef Paul entl€asst]].
(19) CS = {w2W | p2(w) = 1}
CS’ = CS \ {w2W | p1(w) = 1}
(20) main context Peter’s system of ignored
contents
input context a. CG = {p2} a. Ii,w = { }
b. CS = {w2W | p2(w) = 1} b. Ii,w = CS
c. CS p2 c. Ii,w 6 p2
d. CS 6 p1
output context a. CG’ = CG [ {p1} a. I’i,w = Ii,w[ {p2}
= {p1, p2} = {p2}
b. CS’ = CS \ {w2W | b. I’i,w = Ii,w\ {w2CS |
p1(w) = 1} p2(w) = 1}
c. CS’ p1 c. I’i,w p2
d. CS’ p2
A reading under which the proposition p2 (= that the boss fires Paul) is
not true in the worlds in which it is true that Peter ignores that the boss
fires Paul (= p1) is not feasible. The same conditions can be established
for further members of this verb class as e.g. vergessen (forget),
verheimlichen (conceal), verdr€
angen (repress) or bedenken (consider).
A sentence which contains an implication cannot be true without the
implication being true (cf. (21)).12
(21) [p1 Peter verursacht, [p2 dass der Chef Paul entl€asst]], #aber
Peter causes that the boss Paul fires but
der Chef entl€ asst Paul nicht.
the boss fires Paul not
For the context change induced by verbs of this class (cf. (22)), this means
that the embedded proposition is not already part of the CS before the
utterance is made. However, p1 cannot become part of the CS(‘)/CG(‘)
without p2.
11
Ii,w = the set of propositions that an individual i ignores in a world w, Ii,w = the set of
worlds in which all ignored issues of an individual i in a world w are true
12
Vi,w = the set of propositions caused by an individual i in a world w, Vi,w = the set of
worlds in which all propositions caused by an individual i in a world w are true

(22) main context Peter’s ‘cause system’

input context a. CG = { } a. Vi,w = { }
b. CS = W b. Vi,w = W
c. CS 6 p1 c. Vi,w 6 p2
d. CS 6 p2
output context a. CG’ = CG [ {p1, p2} a. V’i,w = Vi,w [ {p2}
= {p1, p2} = {p2}
b. CS’ = CS \ {w2W | p1(w) = 1 b. V’i,w = Vi,w\ {w2W |
∧ p2(w) = 1} p2(w) = 1}
c. CS’ p1 c. V’i,w p2
d. CS’ p2
Further exemplars of implicative verbs are e.g. herbeif€

uhren (bring about)
and beweisen (prove) (cf. Karttunen 1971)) whose occurrence has the
same effect on the context as the use of verursachen (cause).13
The subordinate propositions in (23) neither have to be true in the CS
of the input nor of the output context.
(23) [p1 Peter meint/ sagt/ glaubt/ will, [p2 dass der Chef Paul
Peter thinks/ says/ believes/ wants that the boss Paul
entl€
asst]].
fires
A context update of the kind in (11) describes the effect of an utterance
such as (23). Depending on the matrix verbs, p2 gets added to different
types of embedded contexts (e.g. the matrix subject referent’s beliefs,
reports or wishes) thereby deleting worlds which are not of the respective
kind. On the level of the main context, the information is added that the
respective matrix subject referent’s system is part of the worlds which
make out the current context. The embedded proposition does not
become part of the main context (cf. (11) and (24)).
(24) Peter meint/sagt/glaubt/will, dass der Chef Paul entl€asst,
a. aber der Chef entl€ asst Paul nicht.
but the boss fires Paul not
‘but the boss does not fire Paul.’
b. und der Chef entl€ asst Paul in der Tat.
and the boss fires Paul in the deed
‘and the boss fires Paul indeed.’
Returning the class of manner-of-speaking verbs back into the discus-
sion, sentences such as (25) reveal that they have to be treated in analogy
13
One of the reviewers is absolutely right in questioning my claim by asking for the role
negation and finiteness play. My intuition is that extractions across negated implicative
verbs are still unacceptable although the embedded proposition is not implied in this case.
Extractions from non-finite complements of implicative verbs are fine in English. Further
research is obviously needed at that point.

12 Sonja M€
uller
to the good extraction verbs (cf. the verbs in (24)) as they neither
presuppose nor imply their complement.
(25) [p1 Peter fl€ustert, [p2 dass der Einbrecher noch im
Peter whispers that the burglar still in.the
Haus ist]],
house is
‘Peter whispers that the burglar is still in the house,
a. aber das stimmt nicht. Der Einbrecher ist durchs
but that is.right not the burglar is through.the
Fenster gefl€ uchtet.
window escaped
but this is not right. The burglar escaped through the kitchen
window.
b. und das stimmt. Die F€ uße gucken unter dem Vorhang
and that is.right the feet look under the curtain
heraus.
out
and this is right. The feet peep out from under the curtain.’
The verbs’ context change potential is not different from the more
‘neutral’ verbs of saying (such as sagen (say), behaupten (state), anmerken
(remark)). The propositions expressed by their complements also denote a
set of worlds which contains the worlds in which it is true what Peter says
e.g. This is the CS of a former conversation. The speaker of the complex
sentence reports about this conversation. p2 was added to the main
context in that case and is added to a special system of saying of the matrix
subject (depending on the matrix verb, this might be whispered or shouted
contents e.g.). By uttering (25), p2 is added to the embedded system, p1
becomes part of the main context. As in the case of the verbs in (24), at no
stage of the context update does p2 have to be part of the CS.
3. Deriving the restrictions: Deficient moves in discourse

Section 2 introduced background assumptions concerning the formal
modelling of assertions and questions in discourse structure. Section 2.1
looked at simple sentences, section 2.2 at complex sentences which make
two-stage context changes necessary because the main context as well as
the matrix subject referent’s context is updated. Section 2.3 provided a
discussion of the impact that the meaning components of three different
verb classes (factive, implicative, manner-of-speaking verbs) have in
discourse structural terms.
By relying on the semantic differences between these three verb classes,
in the following section, I will offer an analysis that gives a (discourse)
structural account of the observation that the matrix verbs selecting that-
complement clauses have an influence on the overall acceptability of a
structure that results from wh-extracting a constituent from this very that-
clause. Section 3.1 will develop the analysis first only for factive and
implicative verbs. After an elaboration and extension of the account (based
on further data never mentioned before in the literature) in section 3.2 and
3.3 and a discussion of one meaning attributed to such structures in section
3.4, section 3.5 will transfer the analysis to the case of manner-of-
speaking verbs.
3.1. Uninformativity: The case of factive and implicative verbs

In section 2.3, presuppositions and implications were characterised
semantically in terms of the context update induced by the sentences in
which they are contained. Choosing a more pragmatic approach to
presuppositions means that they are not treated in terms of a relation
between a sentence and a proposition as it is practised within the
semantic view, but that a presupposition is mirrored in the beliefs, in
intentions or expectations of speakers. The approach which is the
intuitively most plausible one is reached when understanding presuppo-
sitions as CG-information. This assumption can e.g. be derived from the
following quotation by Stalnaker (2002:704).
In the simple picture, the common ground is just common or mutual belief,
and what a speaker presupposes is what she believes to be common or
mutual belief. The common beliefs of the parties to a conversation are the
beliefs they share, and that they recognize that they share: a proposition φ
is common belief of a group of believers if and only if all in the group
believe that φ, all believe that all believe it, all believe that all believe that all
believe it, etc.
When uttering a sentence such as (26), the speaker assumes that the
proposition expressed in the complement clause is already part of the
CG, that is the speaker and the hearer know that Mary is pregnant and
that they both know that.14
(26) [p1 Peter verheimlicht, [p2 dass Maria schwanger ist]].
Peter conceals that Mary pregnant is
‘Peter conceals that Mary is pregnant.’
14
There are several factors which complicate the simplified picture presented here. Due to
lack of space and the focus of this paper, I cannot elaborate on those at length. The first
aspect concerns the fact that the assumptions made by speaker and hearer do not necessarily
present true beliefs, but only have to be made for the purpose of the conversation (cf.
Stalnaker 1978:231). This in fact applies to all operations taking part in up-dating the
context and is, therefore, no special issue in the discussion of presuppositions. The second
aspect which is of relevance for the discussion of presuppositions more directly is the insight
that presupposed information does not necessarily have to be old information by being part
of the CG before the utterance containing the presupposition is uttered. This phenomenon
known as accommodation (cf. e.g. Stalnaker 1974:202, Lewis 1979:340, Kadmon 2001:17f.)
also has to be spared out in the illustration here although it has an influence on the analysis.

14 Sonja M€
uller
As the context update in the way illustrated above involves both

evaluating the truth/falsity of propositions with respect to the worlds in
the CS as well as adding the same propositions to the CG, the second
part of the update being left out in (18) (cf. (27)) illustrates exactly the
pragmatic view on the effect on the context referred to in section 2.3.
(27) CG = {p2}
CG’ = CG [ {p1}
= {p2} [ {p1}
= {p1, p2}
The role of the speaker’s attitude towards the presupposed proposition
that is assumed when choosing a pragmatic perspective on the topic can
be backed up by sequences such as (28) to (31).
(28) Peter conceals that the school is on fire, #but I do not consider it
true that the school is on fire.
(29) Peter thinks that the school is on fire, but I do not consider it true
that the school is on fire.
(30) Peter conceals that the school is on fire, #but I do not know whether
the school is on fire.
(31) Peter thinks that the school is on fire, but I do not know whether
the school is on fire.
The speaker can neither commit herself/himself to the (first part of the)
assertion in (28) without committing herself/himself to the presupposed
proposition nor can s/he present the (first part of the) assertion in (30) by
simultaneously denying knowledge of the presupposed content in the
complement. Both attitudes, however, can be advanced in (29) and (31)
under the occurrence of a non-factive verb without any problems.
Such a pragmatic account of presuppositions together with the
semantics of questions introduced in section 2.1 now allows an analysis
of the negative influence that factive verbs have on extractions from that-
complement clauses. In section 2.2, it was assumed that by asking a
question such as (32), a speaker aims at getting that proposition that
should be added to the set constituting Peter’s beliefs, the proposition
answering the root question Who did Mary invite? (cf. the partition in
(33)).15
(32) Wen meint Peter, dass Maria eingeladen hat?, D = {Hans, Fritz}
who thinks Peter that Maria invited has
‘Who does Peter think that Mary invited?’
15
As already mentioned in footnote 10, this is only one interpretation that can be assigned
to this question. Cf. sections 3.2.2, 3.3 and especially 3.4 for further discussion of this issue.
Everyone having doubts about this interpretation be put off to the discussion in section 3.4.

(33)
Peter thinks that Mary invited Hans as well as Fritz.
Peter thinks that Mary invited Hans, but did not invite Fritz.
Peter thinks that Mary invited Fritz, but did not invite Hans.
Peter thinks that Mary invited neither Hans nor Fritz.
Interpreting the less acceptable question in (34) in analogy to (32), that is

the speaker this time wants to know what the proposition that has to be
anchored in the system which contains the contents ignored by Peter
looks like (this proposition being understood as a complete answer to the
question Who did Mary invite?) it opens up the possible complete answers
in (35) (cf. section 2.1 and 2.2 for the background assumptions on the
formal modelling of a question’s impact on discourse).
(34) ??Wen ignoriert Peter, dass Maria eingeladen
who ignores Peter that Mary invited
hat?, D = {Hans, Fritz}
has
(35)
[p5 Peter ignores (the fact) [p1 that Mary invited Hans as well as Fritz]].
[p6 Peter ignores (the fact) [p2 that Mary invited Hans, but did not invite Fritz]].
[p7 Peter ignores (the fact) [p3 that Mary invited Fritz, but did not invite Hans]].
[p8 Peter ignores (the fact) [p4 that Mary invited neither Hans nor Fritz]].
As the proposition expressed by a factive complement clause presents

CG-content, by uttering a question such as (34), the deficient situation
arises that the proposition that should be added to the system which
contains the contents ignored by Peter cannot be different from the one
that is already part of the CG. If it was possible that Peter could ignore
something different from what is the case anyway, (36) should present an
adequate sequence.
(36) Mary invited only Hans, #and Peter ignores (the fact) that Mary
invited Hans as well as Fritz.
This means that by uttering (34), the speaker opens up alternatives
although no alternatives are available. (37) illustrates this situation in
discourse.
(37) main context Peter’s system of ignored contents
CG = {p2} Ii,w = { }
CS = {w2W | p2(w) = 1} Ii,w = CS

16 Sonja M€
uller
CG’ = CG [ {p5}∨ I’i,w = Ii,w[ {p1}∨

CG [ {p6}∨ Ii,w[ {p2}∨
CG [ {p7}∨ Ii,w[ {p3}∨
CG [ {p8} Ii,w[ {p4}
CS’ = CS \ {w2W | p5(w) = 1}∨ I’i,w = Ii,w\ {w2 CS | p1(w) = 1}∨
CS \ {w2W | p6(w) = 1}∨ Ii,w\ {w2 CS | p2(w) = 1}∨
CS \ {w2W | p7(w) = 1}∨ Ii,w\ {w2 CS | p3(w) = 1}∨
CS \ {w2W | p8(w) = 1} Ii,w\ {w2 CS | p4(w) = 1}
Using the factive verb presupposes that the proposition in the
embedded clause (here p2) is part of the CG, so that the question
whether p1, p2, p3 or p4 is part of Peter’s system of ignored contents
simply does not arise. In parallel, adding p5 to p8 to the CG does not
present a serious possibility as the CS’s reduction can only be induced
by p6. Proposition p6 is the only complex proposition which contains
p2. The same problem, however, does not arise with the acceptable
question in (32) as it poses no requirements on the CG and Peter’s
belief system can be updated totally independently from the proposi-
tions in the CG. In principle, Peter can believe in a completely different
possible answer to the question Who did Mary invite? as it is possibly
known that it is the case (cf. (38)). Opening up alternatives is, therefore,
legitimate.
(38) Mary invited only Hans, but Peter thinks that Mary invited Hans
as well as Fritz.
Consequently, by asking for a constituent contained in a that-comple-
ment clause selected by a factive matrix verb, a question results which
does not further progress in discourse, but rather induces a step
backwards because it asks for the form of an issue for which the speaker
simultaneously expresses that s/he (as well as the discourse partner) know
about and know that they both know that. This paper therefore suggests
that the unacceptability of questions such as (34) (when interpreted as in
(35)) is due to the fact that the questions have to be considered fully
uninformative operations in discourse. The questions do not fulfill the
function that is usually associated with questions, namely to open up
alternatives whose reduction leads to an increase in private as well as
consciously shared knowledge.
Relying on the assignment of meaning I assumed for the extraction
constructions so far, extracting a constituent from that-complements
selected by implicative verbs (cf. (39)) in the following sense also leads to
a contradiction: By uttering a sentence containing such a matrix verb, on
the one hand, the speaker expresses that the issue contained in the
subordinate clause becomes CG-information, on the other hand, s/he
expresses that the precise form of the issue is unknown to her/him.

(39) ??Wen verursacht Peter, dass der Chef entl€asst?

Who causes Peter that the boss fires
The proposition expressed in the subordinate clause of a complex
sentence in which an implicative verb makes out the matrix does not
constitute information known in discourse before the sentence is uttered,
but it becomes information of that kind in the course of the utterance. As
illustrated above, implied information logically follows, but is not
presupposed. Due to this state of affairs arising in the case of implicative
verbs, the above argumentation for deriving the negative influence of
factive verbs which relies on the resulting question’s uninformativity
cannot be applied to the case of implicative verbs in its identical form.
The conditions occurring in the case of implicative verbs, however, can
be integrated into the previous argumentation under an expansion of the
Stalnakerian model of the CG by individualising the contextual
parameters. A model that individualises the distributions of knowledge
can be found in Caponigro & Sprouse (2007) (cf. e.g. also Bartels 1999).
Apart from assuming a set of propositions constituting the CG, the
authors also differentiate between the set SB (speaker’s beliefs) and AB
(addressee’s beliefs) which contain the propositions which the speaker
(respectively, the addressee) believes or assumes.
(40) a. SB = {p: p is a belief of the speaker}
b. AB = {p: p is a belief of the addressee}
c. CGS-A = {p: p is mutually believed by the speaker and the
addressee}
(Caponigro & Sprouse 2007:130)
Based on such a threefold distinction of the relevant sets of propositions
in (40), it is the purpose and aim of a conversation to add propositions
from SB/AB to the CG. In the case of posing a constituent question, the
questioner asks the addressee to put the respective piece of information
from her/his AB into the CG. For the conditions of use of ‘ordinary’
wh-questions, Caponigro & Sprouse assume that the proposition asked
for is not part of SB, i.e. the speaker does not know the answer.
Therefore, it follows that the answer is not part of the CG either as the
CG represents the intersection of SB and AB plus consciousness
concerning this intersection. The speaker, however, assumes that they
are part of AB so that the addressee can give an answer which also
becomes part of SB and the CG.
If factive verbs occur in the main clause of a complex sentence, the
subordinate clause’s proposition is part of the CG (and therefore also
part of SB and AB). By extracting a constituent from the that-clause, the
speaker then asks for an issue which – as CG-information – is known to
both discourse partners and which is, therefore, also part of SB (via the
logical connection between CG and SB).

18 Sonja M€
uller
The issue expressed in a subordinate clause embedded under an

implicative verb does not present shared knowledge between speaker and
hearer. However, by presenting the subordinate issue as true, the speaker
nonetheless commits herself/himself to this issue (cf. Simons’ 2007:1043
definition of ‘factive’ verbs). This means that the subordinate clause’s
proposition can be assumed to be part of SB under the occurrence of
such a matrix verb (cf. (41)).
(41) Peter verursacht, dass der Chef sowohl Hans als auch Fritz
Peter causes that the boss both Hans as also Fritz
entl€
asst,
fires
a. #aber ich weiß nicht, ob der Chef sowohl Hans als
but I know not whether the boss both Hans as
auch Fritz entl€asst.
also Fritz fires
‘but I do not know whether the boss will fire Hans as well as
Fritz.’
b. #aber ich weiß nicht, wen der Chef entl€asst.
but I know not whom the boss fires
‘but I do not know whom the boss will fire.’
By referring to the differentiation between SB and CG, the missing
informativity of the degraded question in (39) which opens up the
alternatives in (42) can now be derived indirectly.
(42)
[Peter verursacht, [dass der Chef sowohl Hans als auch Fritz entlässt]p1]p5.
Peter causes that the boss both Hans as also Fritz fires
[Peter verursacht, [dass der Chef Hans entlässt, Fritz nicht entlässt]p2]p6.
Peter causes that the boss Hans fires, Fritz not fires
[Peter verursacht, [dass der Chef Hans nicht entlässt, Fritz entlässt]p3]p7.
Peter causes that the boss Hans not fires, Fritz fires
[Peter verursacht, [dass der Chef weder Hans noch Fritz entlässt]p4]p8.
Peter causes that the boss neither Hans nor Fritz fires
By posing the question in (39), the speaker does not ask for an issue which
is known to both discourse participants, but s/he does ask for an issue
which is known to herself/himself. As the speaker could convey the
proposition that s/he asks the addressee for much more directly than by
asking her/his discourse partner, s/he also does not behave maximally
informatively here. Aiming at informing the discourse partner about an
issue known to the speaker, such a move not only presents a cumbersome
operation, but the speaker also risks arriving at a flow of discourse which is
different from the one originally intended. The discourse partner could
give a different answer than the one the speaker intended to introduce and

which is therefore incompatible with her/his real knowledge. Such a

scenario would need much effort in order to repair the discourse created –
assuming the speaker actually behaves cooperatively and pursues an
honest communicative purpose. The decrease of acceptability of the
extraction construction resulting from asking for a constituent contained
in a that-clause selected by a factive or implicative matrix verb (cf. (34),
(39)) can be traced back to the question not being able to fulfill the
discourse function usually intended by ‘ordinary’ wh-questions, namely to
open up alternatives after whose reduction a growth of information
regarding personal as well as consciously shared knowledge occurs. If an
issue known in discourse or familiar to the speaker is asked for, an
inadequate move in discourse arises. Regarding the force of the move’s
disruption of the flow of discourse, it can be classed between asserting
something that is incompatible with information already known and
asserting something which is already CG-information – two moves that
can also be considered deficient in discourse structural terms. In the first
case, the complete CS is erased and the constitution of shared knowledge is
disrupted. In the second case, one does behave redundantly as neither
progress nor regress (as in the case of poorly acceptable extraction
constructions) arises. Stalnaker (1978:325) writes about such moves which
do not fulfill the function of assertions: “If he failed to conform to the rule,
then he did something that, from the point of view of the conversation, was
unreasonable, inefficient, disorderly, or uncooperative.” Based on the
analysis pursued here, the discourse structural function of the poorly
acceptable extraction constructions containing factive (cf. (34)) and
implicative (cf. (39)) matrix verbs could be characterised in a similar way.
When summing up my approach, one of the reviewers correctly writes
that I predict that violations of Gricean maxims should result in
degradation/ungrammaticality (due to overinformativity/uninformativi-
ty) and that my analysis predicts that other violations of Gricean maxims
should cause degradation/ungrammaticality as well. The reviewer goes
on saying that the character of the degradation/ungrammaticality of my
core question examples seems to be different from other kinds of
pragmatic/Gricean violations. He mentions the following examples: A
speaker might be having a conversation with someone who says “I am
hungry.” and s/he might reply with “You are hungry”. The speaker’s
reply is pragmatically odd. The speaker is not making a legitimate
discourse move, as s/he is simply repeating a proposition that has already
been introduced to the context. In other words, his/her contribution is
uninformative. The reviewer goes on arguing that there is no sense that
the sentence the speaker utters (“You are hungry.”) is degraded/
ungrammatical in any way, even though s/he is violating the Maxim of
Quantity. Another example the reviewer mentions is: “I regret that the
Earth is flat.” By uttering this sentence, the speaker is violating the
Maxim of Quality, but the sentence itself is not degraded. And he states
20 Sonja M€
uller
that one can think of endless other contexts where Gricean maxims are
violated. The point the reviewer makes is clear and I absolutely agree that
my general claim that a pragmatic condition influences the question’s
well-formedness might be considered controversial. The decisive aspect
to keep in mind with respect to my claim, however, is that the reasons
which lead to the assumption that the speaker displays a certain
knowledge state do not result from a particular context being constructed
(of course a perfectly acceptable question also does not degrade when
being asked for the second time in one and the same discourse sequence
although it has already been answered), but are of a much more general
nature: They are anchored in the properties of the linguistic material
employed to yield the respective knowledge states of the participants and
are, therefore, due to the meaning components that are invariably
available for this particular linguistic material (cf. Gajewski’s (2002,
2009) concept of L-analyticity to cope with similar questions when
tracing ungrammaticality back to tautologies and/or contradictions). In
the present case, it is the factive/implicative meaning component of the
verb, in other cases it can also be modal meaning components (cf. M€ uller
2012). Thus, the examples the reviewer mentions do not fall under the
pragmatic conditions which can cause degradations as their uninforma-
tive status cannot be traced back to meaning components which
invariably belong to this linguistic material, but need to be traced back
to uttering the sentence with this lexical material in this particular
context. Uttering the same sentence in a different context does not
necessarily lead to the same uninformative status.
3.2. Repair by changing the question’s intention

My analysis which judges the unacceptable extraction constructions to be
inadequate in discourse semantic terms and, thereby, refers to the
conditions occurring in a particular context, makes the prediction that
the non-occurrence of the factor made responsible for the discourse
structural ‘defect’ (i.e. the contradiction between the cognitive deficit
expressed by posing a wh-question and the knowledge state indicated by
using a factive or implicative matix verb) should have an impact on the
acceptability of the extraction constructions.
3.2.1. Subtypes of question illocution

An analysis of the influence that different subtypes of question illocution
have on the extraction constructions confirms this prediction.16
A detailed analysis of question illocution proves that a cancellation (or
16
Cf. also Erteschik-Shir (1973:61) who already observes that the ‘type’ of question has
an impact on such extractions.

weakening) as well as a reinforcement of the contradiction in discourse

(which concerns the distributions of knowledge) has a positive (or,
respectively, negative) impact on the extraction constructions. Subtypes
looked at are all questions which are not canonical information seeking
questions, but questions the speaker asks himself/herself because s/he has
forgotten some piece of information, pedagogical questions, very strong
and urgent questions as well as deliberative questions.
The data introduced in the following, on the one hand provide
evidence for the general assumption of a discourse semantic constraint I
proposed. On the other hand, they support the concrete modelling of the
extraction data I developed in the last section as the influence observed
can be explained by referring to the parameters that I assume to be
relevant within my analysis. As formulated in Searle’s conditions
concerning the illocutionary type question (cf. Searle 1969:66), cf. (43)),
from a pragmatic perspective, (constituent) questions standardly serve
the purpose of expressing a cognitive deficit on the part of the speaker
who wants to get the answer from the addressee.
(43) Preparatory rule: 1. S does not know ‘the answer’, i.e. [. . .] does
not know the information needed to complete
the proposition truly [. . .].
2. It is not obvious to both S and H that H will
provide the information at that time without
being asked.
Sincerity rule: S wants this information.
Essential rule: Counts as an attempt to elicit this information
from H.
(Searle 1969:66)
In the case of such an information seeking question, regarding the
knowledge states attributed to the discourse participants, the situation
which my analysis is based on so far occurs: Some piece of information is
missing in the speaker’s system of knowledge which, therefore, cannot be
contained in the CG either. The speaker assumes it to be part of the
addressee’s (state of) knowledge so that it can become part of the CG by
being introduced into the context by the addressee and after being
accepted by the questioner. As Searle’s commentary (cf. Searle 1969:66)
already shows, this type of question is not the only one possible.
However, the intention of this illocutionary act can vary.
If e.g. the modal particle doch (sometimes in combination with gleich)
occurs in a constituent question, the speaker expresses that s/he is
searching for an answer which s/he in fact knows, but has temporarily
forgotten. Franck (1980:181ff.) calls this usage of doch “erinnernde[s]
doch” (‘remembering doch’), Luukko-Vinchenzo (1989:30) calls it the
speaker’s attempt to remove his loss of memory. The examples in (44)
illustrate constituent questions which get interpreted exactly along those
22 Sonja M€
uller
lines: The sports club used to be known to the speaker in (44a) as was the
wife’s name in (44b).
(44) a. Nun, wie hieß doch der Sportverein, in dem du warst
Well, how named MP the sports club in the you were
fr€
uher?
earlier
‘Well, what was the name of your former sports club again?’
(Kwon 2005:91)
b. Wie hieß doch gleich seine erste Frau?
How named MP MP his first wife
‘What was his first wife’s name again?’
(Thurmair 1989:269)
It can be observed that extraction constructions which are judged to be
rather bad in isolation improve if the modal particle combination occurs
and the context further supports the particles’ effect. The extraction
construction in (45a) is less acceptable than the structure in (45b).
(45) a. ??Wen impliziert die Aussage, dass der Chef entl€asst?
Who implies the statement that the boss fires
‘Who does the statement imply that the boss will fire?’
b. Peter: I can only remember that several members of staff were
on the verge of being fired. And then there was this silly
statement from the boss which told us in a roundabout way
who would have to go. I do remember the strange atmosphere.
Hm, aber ?wen implizierte die Aussage doch gleich, dass der
Chef entlässt?
INT, but who implied the statement MP MP that the boss fires
‘Well, but who was it again for whom the statement implied
that the boss would fire him?’
The improvement of the less acceptable extraction construction in (45a)
can be traced back within the analysis spelled out in section 3.1 to the fact
that the contradiction between the cognitive deficit expressed by posing
the question and the very knowledge expressed by the presupposition
does not arise: The speaker does not express a general, but only a
temporary cognitive deficit. S/he has to activate an existing piece of
knowledge (presented as such by the factive matrix verb) which is not
available right at the moment, however, which is not missing in principle.
The contradiction between the expression of knowledge and a lack
thereof which I made responsible for the negative influence of certain
verbs can also be dissolved in questions which do not express a cognitive
deficit to be removed on part of the speaker. An example of such a type
of question is the pedagogical question which is characterised by the fact
that the questioner knows the answer and by asking the question aims at

the addressee finding the answer to another question (cf. the second
question in (46)).
(46) Teacher (doing revisions): Why does August have 31 days even
though July has 31 days, too?
Pupil: . . .
Teacher: Na, wer war denn der Kaiser gleich nach
Hey who was MP the emperor immediately after
Julius Caesar?
Julius Caesar
‘Hey, come on, who was the emperor immediately after Julius
Caesar?’
Pupil: Oh, yes, Caesar named July after himself, and emperor
Augustus then named August after himself. It should come imme-
diately after July and should not be shorter than Caesar’s month.
(Truckenbrodt 2004:322)
(context translated by S.M.)
The speaker does not pose the question with the intention to receive the
true answer which enriches her/his own knowledge, but s/he wants to find
out whether the discourse partner knows the answer as well. In the
example looked at here, the answer to the teacher’s question is not even
explicitly given. As the pupil’s answer to the first question is based on the
answer to the second question, the course of the dialogue shows that the
pupil knows the answer to the first question. Therefore, the answer to the
pedagogical question is meant as support in answering the actual
question. This interpretation of the second question in (46) can be used in
order to derive the higher acceptability of otherwise rather worse
extraction constructions in the form of pedagogical questions (cf. (47)).
As between (45a) and (45b), an improvement of the extraction
construction can be perceived between (47a) and (47b).
(47) a. I’m so excited: ??Wen verheimlicht Peter, dass er liebt?
Who conceals Peter that he loves
b. Peter: We’re invited to Hans’ marriage.
Klara: What? Wen heiratet er denn?
Who marries he MP
‘Who does he marry though?’
Peter: Hey, this is so crystal clear!
Klara: I have no idea who this could be!
Peter: Hey, come on! Think about it!
Klara: . . .
Peter (smiling): ?Wen verheimlicht Peter denn seit Jahren,
Who conceals Peter MP for years
dass er liebt?
that he loves

24 Sonja M€
uller
We have met the two of them together several times when we

went for a walk!
Klara: Ah, ok! No, really? He is going to marry the vicar’s
daughter!
Peter: Exactly! It will be a big celebration, I guess.
Assuming for the less acceptable structure in (47a) that a contradiction
arises (constituted by the announcement of a cognitive deficit on part of the
speaker and the simultaneous expression of knowledge of this very issue), it
is exactly the addressee’s acquaintance with the subordinate clause’s
proposition which the speaker actively uses: By posing this question which
the hearer is supposed to be able to easily answer, the speaker gives her/him
a hint regarding the answer to the actual question. By asking for a
presupposed issue in (47b), the speaker, therefore, does not behave
uninformatively, but takes advantage of the distributions of knowledge in
discourse which leads to this question attaining informativeness again.
While the cases referred to so far lead to the extraction constructions’
improvement, it is also possible to achieve the opposite effect by evoking
types of questions in which the question act is particularly strong. Such
questions preserve or even enforce the contradiction between the
knowledge expressed and the lack thereof which is argued to explain
the bad extraction constructions. A speaker’s urgent wish of a question
being answered can again be indicated by certain modal particles. By
referring to the examples in (48) and (49), Kwon (2005) and Thurmair
(1989) show that the question act gets reinforced by the occurrence of
bloß and nur and that the speaker’s interest is directed towards the
question act.
(48) Wo kommt nur auf einmal dieser dicke, fette Hase her?
Where comes MP at once this big fat rabbit from
‘Where does this big fat rabbit come from suddenly?’
(Kwon 2005:132)
(49) “Donnerwetter!” h€ orte Jaga ihn murmeln, “einfach fabelhaft!”
Thunderstorm heard Jaga him mumble simply fabulous
Voller Bewunderung starrte er sie an. “Wo hast du das
Full.of Admiration stared he her on Where have you that
bloß gelernt?”, wollte er wissen.
MP learnt wanted he to.know
“Gosh!” Jaga heard him mumble, “just fabulous!” Full of
admiration, he starred at her. “Where have you learned that, after
all?” he wanted to know.
(Thurmair 1989:241)
In accordance with the analysis suggested here, the occurrence of those
particles in extraction constructions whose acceptability is rather low in

isolation does not lead to an increase in acceptability. The factor made

responsible for the low acceptability of the structures is not neutralised.
On the contrary, it even gets intensified by the reinforcement of the
question act. Introductory phrases (cf. Caponigro & Sprouse 2007:123)
can further enforce the question illocution’s reinforcement (cf. (50), (51)).
(50) I’m really curious: ??Wen impliziert die Aussage bloß, dass
Who implies the statement MP that
der Chef entl€asst?
the boss fires
(51) I really want to know: ??Wen ver€
ubelt Peter seiner Mutter nur,
Who blames Peter his mother MP
dass sie liebt?
that she loves
A further question type which is attributed a reinforcement of the
question illocution is the so-called deliberative question (cf. (52)).
(52) a. Was wird wohl noch alles auf uns zukommen?
What will MP additionally all to us happen
‘(The speaker is wondering: What else will happen to us?’
(Thurmair 1989:144)
b. Was das wohl bedeuten mag?
What this MP mean may
‘(The speaker wondering) What might this mean?’
(Kwon 2005:168)
The typical use of this question is that of a question directed at oneself,
that is this type of question can be considered a pure phrasing of a lack
of information the reduction of which is wished for by the speaker.
Despite its typical use as a self-addressed question, such questions can
also be used as ordinary addressee-oriented questions. In this case, the
obligation for the addressee to answer is abolished (cf. Thurmair
1989:57). Winkler’s (1992:40) assumption that deliberative questions are
not ‘weaker’ questions is important for the argumentation pursued
here. The lack of obligation to answer could lead to this conclusion in
the sense that a question the hearer is not obliged to answer cannot
express a true question act. Winkler argues, however, that especially
such questions do express the speaker’s subjective interest in the
question act and that the speaker whishes to receive an answer. S/he
does not necessarily expect to get it from the hearer, but s/he also does
not always assume that the hearer cannot provide the answer. The
modal particles wohl, bloß and nur and/or the final position of the finite
verb (cf. (52) and (53) [bloß could be substituted by nur without a
change in meaning]) belong to the grammatical means which can code a
deliberative question.

26 Sonja M€
uller
(53) a. Wie ich das bloß am besten mache?

How I that MP on.the best do
b. Wie mache ich das bloß am besten?
How do I that MP on.the best
‘(The speaker wondering) What is the best way to do it?’
(Winkler 1992:40)
Looking at the effect caused by the occurrence of particles which mark
deliberation in extraction constructions (which are otherwise unaccept-
able), one does find out that an improvement of the structures does not
arise.
(54) a. ??Wen bewirkt Peter, dass Maria einl€adt?‘
Who causes Peter that Mary invites
b. Hm. . . Let me think. ??Wen bewirkt Peter bloß, dass Maria
Who causes Peter MP that Mary
einl€adt?
invites
c. ??Wen Peter bloß bewirkt, dass Maria
Who Peter MP causes that Mary
einl€adt? It’s difficult. It’s difficult.
invites
Assuming that a deliberative wh-question displays strong erothetic
illocution, it is no surprise that less acceptable extraction constructions
(interpreted as information-seeking questions in isolation) do not reach
a higher degree of acceptability under the occurrence of particles,
structures and contexts which evoke a deliberative reading of the
wh-question. The contradiction between real knowledge and lack of
knowledge which arises in the case of information-seeking questions
maintains.
The analysis of different question intentions conducted here shows that
mitigating and reinforcing the question act as well as distributions of
knowledge in discourse can be proven to have an impact on the
acceptability of extractions from that-complement clauses. If a reading of
the wh-question is evoked under which the assumed uninformativity of
the question does not arise, i.e. the contradiction between the knowledge
coded and the information deficit expressed does not arise, the structures
are rated more acceptable than it is the case under their interpretation as
an information-seeking question.
3.2.2. A further informative reading

The pragmatic analysis I proposed in sections 3.1 and 3.2.1 makes the
prediction that wh-extraction should improve/degrade whenever the
context allows for an informative interpretation of the wh-question.

The core of the analysis argued for in this paper is the assumption that
a question such as (55) – in analogy to a question as in (57) with the
partition in (58) – induces a partition as in (56).
(55) ??Wen ignoriert Fritz, dass Maria eingeladen
Who ignores Fritz that Mary invited
hat?, D = {Hans, Karl}
has
(56) [Fritz ignoriert,
Fritz ignores
[dass Maria Hans eingeladen hat, Karl eingeladen hat]p1]p5.
that Mary Hans invited has,Karl invited has
[dass Maria Hans eingeladen hat, Karl nicht eingeladen hat]p2]p6.
that Mary Hans invited has,Karl not invited has
[dass Maria Hans nicht eingeladen hat, Karl eingeladen hat]p3]p7.
that Mary Hans not invited has,Karl invited has
[dass Maria weder Hans noch Karl eingeladen hat]p4]p8.
that Mary neither Hans nor Karl invited has
(57) Wen glaubt Fritz, dass Maria eingeladen

Who believes Fritz that Mary invited
hat?, D = {Hans, Karl}
has
(58) [Fritz glaubt,
Fritz believes
[dass Maria Hans eingeladen hat, Karl eingeladen hat]p1]p5.
that Mary Hans invited has,Karl invited has
[dass Maria Hans eingeladen hat, Karl nicht eingeladen hat]p2]p6.
that Mary Hans invited has,Karl not invited has
[dass Maria Hans nicht eingeladen hat, Karl eingeladen hat]p3]p7.
that Mary Hans not invited has, Karl invited has
[dass Maria weder Hans noch Karl eingeladen hat]p4]p8.
that Mary neither Hans nor Karl invited has
That is, the question aims at naming the precise form of the proposition
which should be anchored in Fritz’ system which contains ignored
contents, the embedded proposition corresponding to the strongly
exhaustive answer to the question Wen hat Maria eingeladen? (Who did
Mary invite?). The meaning thus ascribed to the question in (55) is that
the matrix subject’s attitude gets assigned to the complete state of the
world: It asks what it is that Fritz ignores and what he ignores is a

28 Sonja M€
uller
complete description of the actual world regarding the issue of Mary

inviting some people and not inviting others. Although it is surely not
possible under the occurrence of a factive matrix verb to e.g. ignore a
different state than the one existing in the context set, it is still plausible
that the attitude is only displayed towards a part of the state of the
world or even towards no part of it. Under this interpretation, the
extraction construction asks against the background of existing facts (in
this case a known constellation of invited and uninvited guests) for
those aspects of this state which Fritz ignores. One could e.g. imagine a
scenario in which it is known that Mary invited Hans as well as Karl
and the speaker then wants to know what it is that Fritz ignores with
respect to this state. Under such a scenario, Fritz could ignore the fact
that Mary invited Hans, but Karl’s presence gets considered by him or
Fritz could have a problem with both guests, only with Karl, but not
with Hans or also with none of them. The partition mirroring this
meaning is shown in (59).
(59)
[p3 Fritz ignoriert, [p1 dass Maria Hans eingeladen hat]], und
Fritz ignores that Mary Hans invited has and
[p4 Fritz ignoriert, [p2 dass Maria Karl eingeladen hat]].
Fritz ignores that Mary Karl invited has
[p3 Fritz ignoriert, [p1 dass Maria Hans eingeladen hat]], und
Fritz ignores that Mary Hans invited has and
[¬p4 Fritz ignoriert nicht, [p2 dass Maria Karl eingeladen hat]].
Fritz ignores not that Mary Karl invited has
[¬p3 Fritz ignoriert nicht, [p1 dass Maria Hans eingeladen hat]], und
Fritz ignores not that Mary Hans invited has and
[p4 Fritz ignoriert, [p2 dass Maria Karl eingeladen hat]].
Fritz ignores that Mary Karl invited has
[¬p3 Fritz ignoriert nicht, [p1 dass Maria Hans eingeladen hat]], und
Fritz ignores not that Mary Hans invited has and
[¬p4 Fritz ignoriert nicht, [p2 dass Maria Karl eingeladen hat]].
Fritz ignores not that Mary Karl invited has
Ascribing this interpretation to the question in (55) does not evoke the
discourse structural inadequacy argued for above as the speaker does not
ask for the complete state of the world which Fritz ignores. The speaker
knows what the state looks like, but does not know towards which parts
of it Fritz displays the attitude of ignoring it. This reading of the question
is thus fully legitimate and informative.
Assuming that the semantic intuitions with respect to the two
interpretations of the questions (modelled here by assuming two

possibilities of partitioning) are real, the question arises what exactly

differentiates those two readings from each other.
The decisive factor at issue here is in fact where the strong exhaustivity
is anchored. Under the reading in (56), positive and negative instanti-
ations of being invited are opened up, i.e. the exhaustive partitioning
applies in the scope of the conjunction (in the following referred to as the
‘low’ reading). In (59), the partitioning concerns ignored and non-ignored
issues, i.e. the exhaustive partitioning applies above the conjunction
(in the following referred to as the ‘high’ reading). Under the assumption
that both interpretations of the question are available, the account
pursued here predicts a higher degree of acceptability for the factive
construction under the high reading as the construction does not display
the pragmatic anomaly assumed to hold under the low reading.17
As the account pursued here makes the prediction that the questions
which result from extracting a wh-pronoun from a that-clause selected by
a factive, implicative or manner-of-speaking verb are degraded due to a
possible interpretation under which they have to be considered prag-
matically anomalous, evoking the high reading (which has been attested
to be informative) should contribute positively to the questions’
acceptability status. What the account is looking for at this point is
thus linguistic or contextual means which strengthen the reading of the
question that it aims at selecting a set of the CG-propositions which then
make out the (e.g.) ignored issues. The remaining CG-propositions then
are those propositions which are not ignored. In order to make the
account’s predictions even stronger, the low reading should altogether be
17
The reviewer remarks that the ‘high’ reading reminds one of Pesetsky’s (1987) concept
of D-linking and Cinque’s (1990) concept of referentiality. D-linked and referential wh-
phrases can escape weak islands whereas non-D-linked and non-referential wh-phrases
cannot be extracted from the same domains. D-linking and referentiality are certainly
pragmatic characterisations, too. One of the main differences between my account and
Pesetsky (1987) and Cinque (1990) is that I do not intend a syntactic implementation of the
pragmatic factors. On the one hand, I think that (depending on the architecture of grammar
one wants to assume), anchoring pragmatic information in the syntactic derivation is a
problematic assumption in general. On the other hand, this pragmatic aspect which seems so
similar to D-linking and referentiality is only one of the usage-based criteria I propose to
play a role. I believe it is quite implausible to assume a syntactic reflex for different erothetic
illocutionary subtypes. Syntax then had to know (and be interested in) whether a wh-
interrogative clause will be used as a neutral information seeking question, a rhetorical
question, a very strong and urgent question, a pedagogical question etc. Although such
modelling is certainly possible, one still needs to formulate the pragmatic conditions (which
I intend to do with my approach) first. As the reviewer suggests, an advantage of my
account is that I only need to appeal to one module of grammar (pragmatics), while
Pesetsky (1987) and Cinque (1990) need to refer to two modules (syntax and pragmatics). In
addition, I think they need to make a proposal how the two systems can interact with each
other (see my first point above) – especially in those cases in which the pragmatic factors
arise only due to contextual information and not because of certain linguistic material (such
as which-phrases e.g.). At least for generative syntactic approaches, it will be problematic to
argue that syntax knows about the (later) use of the construction.

30 Sonja M€
uller
excluded. In the following, I will argue that the so-called hat contour can
serve this very purpose.
3.3. The hat contour

It has been argued that this intonational contour is associated with a
particular discourse structural meaning. B€ uring (2003) argues that an
assertion displaying this intonational pattern indicates that this sentence
is part of a more complex discourse structure. He builds on the
assumption that discourse is organised in sequences of questions and sub-
questions (cf. also Roberts 1996). In a conversation about an evening
e.g., questions such as (60) (cf. B€uring 2003:4) might arise which get
answered either directly or by splitting the bigger question into sub-
questions which get answered in turn.
(60) question: How was the concert?
sub-question: Was the sound good? No, it was awful.
sub-question: How was the audience? They were
enthusiastic.
sub-question: How was the band?
subsub-question: How was the drummer? Just fantastic.
subsub-question: And what about the singer? Better than ever.
sub-question: Did they play old songs? Not a single one.
question: So what did you do after the concert? . . .
This interplay between question and sub-question aiming at answering

the ‘highest’ question via the different sub-question-sub-answer
sequences is called a strategy in Roberts (1996).
B€uring argues that an English sentence such as (61) serves the purpose
of indicating that it answers a sub-question of such a question-answer-
sequence. (61) e.g. indicates that it is part of a bigger sequence which aims
at answering the overall question Who ate what? by answering the sub-
question What did Fred eat?
\/ \
(61) FRED ate the BEANS.
Moreover, the sentence indicates that there are further questions of the
type What did X eat?, the X being an alternative to the element that
receives the fall-rise in English (namely, the NP Fred in the case at hand)
(cf. (62)).

(62) question: Who ate what?

sub-question: What did Fred eat?
\/ \
sub-answer: FRED ate the BEANS.
sub-question: What did Mary eat?
\/
sub-answer: MARY ate ...
Jackendoff (1972) calls the first accent (the fall-rise) B-accent, the second
one (fall) A-accent. According to B€ uring (2003), the B-accent marks a
contrastive topic, the A-accent the focus.
In German, in such contexts, the hat contour occurs. As the name
suggests, the intonational contour starts with a rise (/ marking a rising
accent), stays constant on the high level reached and falls again at the end
(\ marking a falling accent) (cf. Uhmann 1991, Fery 1993, Mehlhorn 2001
on phonetic/phonological issues regarding this intonational pattern). In
this particular context, the matrix subject noun phrase would mark the
start of the hat contour by carrying the rising accent, the direct object
noun phrase concludes it by being assigned a falling accent. The part in
between is realised on the same pitch level (cf. (63)).
__________________
(63) /FRITZ aß die BOHnen\.
A sentence such as (64) can be interpreted in analogy to (62) as being part
of a more complex discourse structure the topic of which is the question
which person ignored which invitation by Mary (cf. (65)).
(64) B: /FRITZ ignoriert, dass Maria PAUL\ eingeladen hat.
Fritz ignores that Mary Paul invited has
‘Fritz ignores (the fact) that Mary invited Paul.’
(65) question: Wer ignoriert welche Einladung von Maria?
Who ignores which invitation by Mary
‘Who ignores which invitation by Mary?’
sub-question: Wessen Einladung durch Maria ignoriert
Which invitation by Mary ignores
FRITZ?
Fritz
‘Which invitation by Mary does Fritz ignore?’
sub-answer: /FRITZ ignoriert, dass Maria PAUL\
Fritz ignores that Mary Paul
eingeladen hat.
invited has
‘Fritz ignores that Mary invited Paul.’

32 Sonja M€
uller

PEter?
Peter
‘Which invitation by Mary does Peter ignore?’
sub-answer: /PEter ignoriert, dass Maria FRITZ\
Peter ignores that Mary Fritz
eingeladen hat.
invited has
‘Peter ignores that Mary invited Fritz.’
KARL?
Karl
‘Which invitation by Mary does Karl ignore?’
sub-answer: /KARL ignoriert, dass Maria PE\ter
Karl ignores that Mary Peter
eingeladen hat.
invited has
‘Karl ignores that Mary invited Peter.’
Alternatively, the questions in (66) which are judged more acceptable
than the isolated question in (67) are also imaginable.
(66) B: /FRITZ ignoriert, dass Maria PAUL\ eingeladen hat.
Fritz ignores that Mary Paul invited has
‘Fritz ignores that Mary invited Paul.’
A: Ah ok, und ?wen ignoriert /PEter, dass Maria eingeladen hat?
Ah, ok, and who ignores Peter that Mary invited has
B: /PEter ignoriert, dass Maria FRITZ\ eingeladen hat.
Peter ignores that Mary Fritz invited has
‘Peter ignores that Mary invited Fritz.’
A: Aha. Und ?wen ignoriert /KARL, dass Maria eingeladen hat?
Right. And who ignores Karl that Mary invited has
B: /KARL ignoriert, dass Maria PE\ter eingeladen hat.
Karl ignores that Mary Peter invited has
‘Karl ignores that Mary invited Peter.’
(67) ??Wen ignoriert Peter, dass Maria eingeladen hat?
Who ignores Peter that Mary invited has
The intonational pattern’s positive influence on the extraction con-
struction (as illustrated in (66)) can find an explanation within my
account along the following lines: On the one hand, the intonational
contour allows the inference that there is more than one question of the
form Wen ignoriert X, dass Maria eingeladen hat?, i.e. the question

indicates that there are further questions asking for the systems which
contain the ignored contents of different individuals. The whole set of
those questions aims at answering the higher question which intends to
detect all relations of people ignoring each other. This aspect already
seems to suffice to make it a reasonable assumption that there is more
than one choice of ignored contents as the question asks for several
assignments of individuals and the issues ignored by them. Although
everybody could in theory ignore the invitation of the same individuals,
the people who ignore other people’s invitations probably do not ignore
their own invitation. The issues ignored by different individuals,
therefore, should already differ for this reason. This idea is strengthened
by the fact that B€ uring assumes for sequences such as (62) the
conversational implicature that every person who is mentioned in a
question has eaten something else. This means that if Fred has eaten
beans, then Mary has not eaten beans, too. The inference goes along
the following lines: If someone else had eaten beans as well, the speaker
would have been more informative if s/he had already conveyed this
piece of information in the course of speaking about Fred. As the
speaker knows that further people have eaten, but does not say about
them that they ate beans as well, the conclusion can be drawn that they
have eaten something else. Similarly, it can thus be argued for (66) that
the subject referents do not display the same systems of ignored
contents. If everybody ignores something else, this does make it even
more reasonable that there are different subsets of the state of the world
which corrspond to ignored issues as the question can only be answered
with respect to the same state of the world. This means that the systems
cannot differ because of the state of the world being a different one for
each individual. They can only differ by containing different pieces of
the state of the world or of the propositions in the CG respectively.
Against the background of a CG as in (68), 18 different systems of
ignored contents are imaginable. Relying on the implicature triggered
by the answers, the extraction constructions aim at getting to know
which of the 18 possible systems which contain ignored contents apply
to each subject referent.
(68) CG = {that Mary invited Peter, that Mary invited Fritz, that Mary
invited Paul, that Mary invited Karl}
Based on the discourse structural analysis developed in B€ uring (2003),
the hat contour can be considered a linguistic means whose occurrence
evokes the question’s reading under which the strong exhaustivity gets
interpreted above the conjunction. As has been illustrated, under this
reading, the question receives an interpretation under which the conflict
between the speaker’s knowledge (expressed by the factive verb) and
her/his lack thereof (expressed by the constituent question) does not

34 Sonja M€
uller
arise. In accordance with my analysis, the question’s degree of

acceptability rises.
At this point, one might ask why the problematic low reading should
exist at all. However, there are a couple of arguments and observations
which make this assumption necessary. The next section will discuss two
of such aspects.
3.4. Motivating the ‘low reading’

The accounts which model the semantics of questions by relying on the
concept of the partition stand in the tradition of the semantic
characterisation of questions which B€auerle & Zimmermann
(1991:341) call Antwortmengen-Methode (set of answer-method). In this
group, they include accounts whose underlying idea it is to associate the
meaning of questions with the set of their answers. Under this view, a
question denotes a set of propositions which can serve as answers to the
question. Theories differ in that they spell out this core concept
differently. Whereas Hamblin (1958, 1973) assumes the set of possible
answers to a question to constitute the question’s denotation, Karttunen
(1977) chooses a more restrictive formulation and limits the set of
propositions constituting the question’s denotation to the set of true
answers. As described in section 2.1, accounts relying on the concept of
the partition are based on the idea that the logical space is not
unstructured, but forms a partition. The subsets which are built, each
correspond to a strongly exhaustive answer which provides information
on positive and negative instantiations with respect to the state of affairs
asked for.
The aspect which is regarded as the advantage of the partition account
(when compared to e.g. Hamblin’s and Karttunen’s semantic characteri-
sation) and which in a sense motivates the availability of negative
information in a question’s denotation is that the relevance of statements
in relation to answering a question can be determined and that, therefore,
different degrees of answering can be captured under this view (cf.
Higginbotham 1993:688f., Lohnstein 2000:58ff.).
The general function of answers is to reduce the alternatives opened up
by a question, i.e. to reduce classes in the space of answers. Ideally, the
answer reduces the space of answers to one remaining class which makes
up a complete answer because of the single classes’ incompatibility with
each other (cf. section 2 on complete answers). Looking at the answers to
the question in (69), (70a) and (70b) are complete answers. They reduce
the partition in (72) to class 17. The partial answer in (70c) and (70d)
cannot cause the reduction to a single class. Nevertheless, they lead to the
exclusion of certain classes. The statements are compatible with some
classes and incompatible with other classes. In contrast to (70c) and
(70d), (71) does not seem to provide a contribution to the question at all
and is, therefore, irrelevant. Higginbotham (1996:374) assumes that a

relevant answer is incompatible with at least one class of the partition. As
the statement in (71) does not lead to a reduction of the partition, it
presents an irrelevant contribution.
(69) Wen hat Maria eingeladen? D = {Karl, Hans, Fritz, Frieda, Hannah}
Who has Mary invited
‘Who did Mary invite?’
(70) a. Maria hat Karl, Hans und Fritz eingeladen, Frieda
Mary has Karl, Hans and Fritz invited, Frieda
und Hannah nicht.
and Hannah not
‘Mary invited Karl, Hans and Fritz, and didn’t invite Frieda
and Hannah.’
b. Maria hat Karl, Hans und Fritz eingeladen und
Mary has Karl, Hans and Fritz invited and
sonst keinen.
else nobody
‘Maria invited Karl, Hans and nobody else.’
c. Maria hat Karl, Hans und Fritz eingeladen.
Mary has Karl, Hans and Fritz invited
‘Mary invited Karl, Hans and Fritz.’
d. Maria hat weder Frieda noch Hannah eingeladen.
Mary has neither Frieda nor Hannah invited
‘Mary invited neither Frieda nor Hannah.’
(71) Heute hatten meine Studenten so viele Fragen.
Today had my students so many questions
‘My students had so many questions today.’

36 Sonja M€
uller
(72)
class Maria hat ... eingeladen
1 ¬Karl, ¬Hans, ¬Fritz, ¬Frieda, ¬Hannah p1
2 Karl, ¬Hans, ¬Fritz, ¬Frieda, ¬Hannah p2
3 ¬Karl, Hans, ¬Fritz, ¬Frieda, ¬Hannah p3
4 ¬Karl, ¬Hans, Fritz, ¬Frieda, ¬Hannah p4
5 ¬Karl, ¬Hans, ¬Fritz, Frieda, ¬Hannah p5
6 ¬Karl, ¬Hans, ¬Fritz, ¬Frieda, Hannah p6
7 Karl, Hans, ¬Fritz, ¬Frieda, ¬Hanna p7
... ...
17 Karl, Hans, Fritz, ¬Frieda, ¬Hannah p17
...
27 Karl, Hans, Fritz, Frieda, ¬Hannah p27
28 Karl, Hans, Fritz, ¬Frieda, Hannah p28
29 Karl, Hans, ¬Fritz, Frieda, Hannah p29
30 Karl, ¬Hans, Fritz, Frieda, Hannah P30
31 ¬Karl, Hans, Fritz, Frieda, Hannah p31
32 Karl, Hans, Fritz, Frieda, Hannah p 32
(70c) is incompatible with all classes in which less than three persons are
invited and in which Karl, Hans and Fritz are not among the guests.
After uttering (70c), the following four classes remain, so that 28 classes
get reduced although the answer is not a complete answer.
(73)
class Maria hat … eingeladen.
17 Karl, Hans, Fritz, ¬Frieda, ¬Hannah p17
27 Karl, Hans, Fritz, Frieda, ¬Hannah p27
28 Karl, Hans, Fritz, ¬Frieda, Hannah p28
32 Karl, Hans, Fritz, Frieda, Hannah p32
Similarly, (70d) is incompatible with all classes of the partition in which

a woman is invited, however, it is still compatible with all the other
classes.
Whereas a complete answer is compatible with exactly one class of the
partition and incompatible with all the other classes, a partial answer is
incompatible with at least one class of the partition and compatible with
more than one class. Hamblin’s (1958, 1973) and Karttunen’s (1977)
accounts do not provide instruments to account for partial answers. If
the world is such that Mary invited Karl, Hans and Fritz, (69) denotes P
= {p1, p2, p3, p4, p5} (according to Hamblin) and P = {p1, p2, p3}
(according to Karttunen) (cf. (74)). The proposition expressed in (70d) is
not in the set and, therefore, would not be considered an answer.

(74) Maria hat Karl eingeladen. (= p1), Maria hat Fritz eingeladen.
(= p2), Maria hat Hans eingeladen. (= p3), Maria hat Frieda
eingeladen. (= p4), Maria hat Hannah eingeladen. (= p5)
Partial answers can be used to provide evidence for the availability of the
low interpretation of strong exhaustivity. If the question is an extraction
construction such as (75), (76) is an adequate – even if incomplete –
answer.
(75) Wen glaubt Peter, dass der Chef entl€asst?, D = {Paul, Hans}
Who thinks Peter that the boss fires
‘Who does Peter think that the boss will fire?’
(76) Ah, ich weiß es nicht genau, aber er glaubt auf jeden
Well, I know it not exactly but he thinks in any
Fall, dass der Chef nicht Paul entl€ asst.
case that the boss not Paul fires
‘Well, I don’t know it exactly, but in any case he thinks that the
boss does not fire Paul.’
As (70c) reduces the cells of the partition in (72) to four cells, the
answer in (76) excludes two possibilities of all conceivable answers (cf.
(77)).
(77)
[p5 Peter glaubt, [p1 dass der Chef Paul, Hans entlässt]].
Peter thinks that the boss Paul, Hans fires
‘Peter thinks that the boss fires Paul, Hans.’
[p6 Peter glaubt, [p2 dass der Chef Paul, ¬Hans entlässt]].
Peter thinks that the boss Paul, ¬Hans fires
‘Peter thinks that the boss fires Paul, ¬Hans.’
[p7 Peter glaubt, [p3 dass der Chef ¬Paul, Hans entlässt]].
Peter thinks that the boss ¬Paul, Hans fires
‘Peter thinks that the boss fires ¬Paul, Hans.’
[p8 Peter glaubt, [p4 dass der Chef ¬Paul. ¬Hans entlässt]].
Peter thinks that the boss ¬Paul, ¬Hans fires
‘Peter thinks that the boss fires ¬Paul, ¬Hans.’
It does not seem counterintuitive to interpret (76) as partial information

concerning Peter’s opinion (just as (70c) and (70d) are read as partial
answers to the question for the invited guests). After inducing a context
change by this answer, two alternative doxastic models (out of the four
possible previous ones) remain and can be reduced even further – should
the situation arise (cf. (78)).

38 Sonja M€
uller
(78) Ach ja, und er glaubt auch, dass der Chef Hans in der
Oh yes and he thinks also that the boss Hans in the
Tat entl€
asst.
deed fires
‘Oh, and in fact he also thinks that the boss fires Hans indeed.’
Parallel dialogues can also be construed coherently under the occurrence
of different matrix verbs (cf. a verb of saying in (79)).
(79) A: Wen sagt Peter, dass der Chef entl€asst?
Who says Peter that the boss fires
‘Who does Peter say that the boss fires?’
B: Ah, ich weiß nur noch, dass er sagt, dass der Chef
Oh, I know only also that he says that the boss
Paul nicht entl€ asst.
Paul not fires
‘Oh, I can only remember that he says that the boss does not
fire Paul.’
If one assumed that the strong exhaustivity in such an extraction
construction could not be interpreted low, answers such as those in (76)
and (79) could not be considered relevant answers and would have the
same status as a completely irrelevant answer such as (71) (to (69)) which
cannot bring about a reduction of possible answers (abstracting away
from possible pragmatic inferences an answer such as (71) could have
which then still led to a reduction of classes). Under the high
interpretation of the strong exhaustivity, it is excluded to anchor
negative information in the doxastic model or the model containing said
contents. The only possibility to provide negative information as an
answer would be to negate that things have been said or are believed (cf.
(80), (81)).
(80) A: Wen glaubt Peter, dass der Chef entl€asst?
‘Who does Peter think that the boss fires?’
B: Ah, ich weiß nur noch, dass er nicht glaubt, dass der
Well, I know only also that he not think that the
Chef Paul entl€ asst.
boss Paul fires
‘Well, I only remember that he doesn’t think that the boss
fires Paul.’
(81) A: Wen sagt Peter, dass der Chef entl€asst?
Who says Peter that the boss fires
‘Who does Peter say that the boss fires?’

B: Ah, ich weiß nur noch, dass er nicht sagt, dass der
Well, I know only also that he not say that the
Chef Paul entl€ asst.
boss Paul fires
‘Well, I only remember that he doesn’t say that the boss fires
Paul.’
Whereas the meaning of B’s statement in (80) and (76) might possibly be
identical (due to neg-raising),18 this certainly does not apply to the
meaning expressed by B’s statements in (79) and (81). If one wants to
treat answers as in (76) and (79) as valid partial answers to the respective
questions, the negative information has to be ascribed low, i.e. within the
matrix systems. In this argumentation, answers such as (76) and (79)
provide evidence for the availability of the meaning I ascribed to the
question in section 2.1. (82) repeats a question of the type of questions
discussed in section 2. (83) shows again the interpretation I ascribed to
the question.
(82) Wen meint Peter, dass der Chef entl€ asst? D = {Karl, Hans}
‘Who does Peter think that the boss will fire?’
(83)
[Peter thinks [that the boss will fire Hans as well as Karl]p5]p1.
[Peter thinks [that the boss will fire Hans, but will not fire Karl]p6]p2.
[Peter thinks [that the boss will fire Karl, but will not fire Hans]p7]p3.
[Peter thinks [that the boss will fire neither Hans nor Karl]p8]p4.
If one wanted to deny the existence of this meaning, one had to come up
with an explanation for the facts on partial answers (against the
background of their role within the partition account) presented in this
section. From a compositional point of view, the low reading might be
18
Cf. e.g. Bartsch (1973) and Horn (1975) on neg-raising. In case the inference is allowed,
the extraction constructions have the same meaning under both readings. However, this
factor does not have an impact on deriving the extraction constructions’ acceptability as it is
pursued here. In those cases in which the question’s interpretation does have an impact on
the analysis, neg-raising is not possible. (i) and (ii) as well as (iii) and (iv) cannot have the
same meaning.
(i) Peter ignoriert nicht, dass Maria Hans abholt.
Peter ignores not that Mary Hans picks up
‘Peter does not ignore (the fact) that Mary picks up Hans.’
(ii) Peter ignoriert, dass Maria Hans nicht abholt.
Peter ignores that Mary Hans not picks up
‘Peter ignores (the fact) that Mary does not pick up Hans.’
(iii) Peter verursacht nicht, dass der Chef Hans entl€asst.
Peter causes not that the boss Hans fires
(iv) Peter verursacht, dass der Chef Hans nicht entl€asst.
Peter causes that the boss Hans not fires

40 Sonja M€
uller
considered dubious for the reason that one needed to come up with a
mechanism that reconstructs the matrix wh-phrase’s scope in the
subordinate clause (in order to achieve the partitioning of only the
matrix subject’s system).
Although I cannot offer the concrete (technical) solution to this aspect
at this point, in the following, a further well-known German construction
(the so called wh-imperative clause (cf. (84)) will be referred to which also
provides a challenge when discussing the relation between a wh-phrase’s
scope and its syntactic position (a further construction is also mentioned
in footnote 19). Apart from this aspect, the construction provides further
evidence for the availability of the meaning of the question I ascribed in
section 2.1, that is for the meaning of the negative information in the
question’s denotation below the conjunction.19
(84) Wen sag mir doch mal gleich, dass Peter gestern
whom tell me MP MP right away that Peter yesterday
besucht hat.
visited has
‘Tell me right away whom Peter visited yesterday.’
Reis & Rosengren (1992:94)
For sentences such as (84), Reis & Rosengren (1992) argue that the
moved wh-phrase does not scope over the matrix clause (which is why the
sentence is still read as an imperative and not as an interrogative), but
only has scope over the embedded clause. Therefore, (84) has the same
meaning as (85).
(85) Sag mir doch mal gleich, wen Peter gestern besucht hat.
Tell me MP MP right away whom Peter yesterday visited has
‘Tell me right away whom Peter visited yesterday.’
Reis & Rosengren do not give a formal account of the meaning of (84)
(or (85)). However, applying Groenendijk & Stokhof’s and Higginbo-
tham’s semantics for questions to the embedded wh-complement in (85)
(as it is done in Groenendijk & Stokhof 1982), allows paraphrasing (84)
by (86), that is interpreting the [+wh]-clause strongly exhaustively.
19
I’m grateful to Horst Lohnstein for bringing this construction to my attention.

(86)
Sag mir doch mal gleich, dass Peter Hans, Fritz gestern besucht
Tell me MP MP immediately that Peter Hans Fritz yesterday visited
hat.
has
‘Tell me right away that Peter visited Hans, Fritz yesterday.’
Sag mir doch mal gleich, dass Peter Hans, ¬ Fritz gestern
Tell me MP MP immediately that Peter Hans ¬ Fritz yesterday
besucht hat.
visited has
‘Tell me right away that Peter visited Hans, ¬ Fritz yesterday.’
Sag mir doch mal gleich, dass Peter ¬ Hans, Fritz gestern
Tell me MP MP immediately that Peter ¬ Hans Fritz yesterday
besucht hat.
visited has
‘Tell me right away that Peter visited ¬ Hans, Fritz yesterday.’
Sag mir doch mal gleich, dass Peter ¬ Hans, ¬ Fritz gestern
Tell me MP MP immediately that Peter ¬ Hans ¬ Fritz yesterday
besucht hat.
visited has
‘Tell me right away that Peter visited ¬ Hans, ¬ Fritz yesterday.’
There are a couple of other pecularities related to this rather unusual

construction in German (cf. Reis & Rosengren 1992), however, it does
provide evidence for the assumption that the wh-phrase occurring in the
left peripheral surface position does not scope over the rest of the
sentence, but only over the structure below the matrix verb. This is
exactly the domain for which I argue in section 3.2 that the partitioning
can occur there. Consequently, my assumption that the fronted
wh-phrase can principally take narrow scope in the extraction construc-
tions when being extracted from a that-complement clause (i.e. scope
over the subordinate clause) (cf. section 2.2 and 3.1) also has to be
assumed independently for different structures.20
20
Arguing that the fronted wh-phrase’s scope is restricted to the subordinate clause (for
the formal modelling of which additional reconstruction mechanisms might have to be
introduced), seems even less dubious when one takes into consideration that research in
related fields (cf. e.g. sentences such as (i)) sees itself confronted with (in a sense) opposite
puzzles concerning the implementation of necessary scope expansions in order to account
for the semantic intuitions.
(i) Was glaubt Peter, wen der Chef entl€asst?
What thinks Peter who the boss fires
I’m grateful to Horst Lohnstein for mentioning such structures.

42 Sonja M€
uller
3.5. Over-informativity: The case of manner-of-speaking verbs

As I already alluded to in the course of developing the above account,
modelling the negative influence of factive and implicative matrix verbs
on extractions from that-complement clauses can be considered a
violation of the preparatory rule for ordinary questions within classical
Searlean speech act theory (cf. (43)). The question resulting from asking
for a constituent contained in a factive or implicative that-clause (under
certain interpretations of the questions (cf. section 3.1)) can be assigned a
meaning under which it does not fulfill the function that is usually
associated with questions, namely to open up alternatives whose
reduction leads to an increase in private (as well as consciously shared)
knowledge. The criterion sorting acceptable and less acceptable extrac-
tion constructions within the analysis I suggested is a discourse structural
one and predicts the negative rating of extraction constructions in which
factive and implicative matrix verbs occur. By pursuing this line of
argumentation, the analysis, however, does not exclude that further
discourse structural criteria have an impact on the structures. By opening
up another criterion of that nature, in the following, the analysis will be
extended to also cover the non-bridge verb quality of manner-of-speaking
verbs such as fl€ ustern (whisper) or schreien (shout).
Since the first mentioning of extraction data of the sort I’m discussing
in this paper, it has been cited again and again in the literature that
extraction constructions containing manner-of-speaking verbs are judged
to be less acceptable than parallel structures in which ‘pure’ verbs of
saying (e.g. sagen (say)) constitute the matrix. Erteschik-Shir (1973:51)
e.g. illustrates that the acceptability of the sentence in (87) decreases
when the verbs listed in (88) are inserted for V ((88a) > (88b) > (88c)).
(87) What did you V ((to) them) that he had done?
(88) a. say, tell, report, announce
b. grunt, holler, murmur, mumble, roar
c. purr, snarl, editorialise, lisp, animadvert
Kluender (1991:245) refers to the contrast between (89a) and (89b).
(89) a. How angry did Mary say that John was? >
b. How angry did Mary whisper that John was?
Chomsky (1977:85) gives the examples in (90).
(90) a. *What did John complain that he had to do this evening?
b. *What did John quip that Mary wore?
c. ?Who did he murmur that John saw?
And as becomes obvious from the judgements given in (2) (repeated for
convenience in (91)), it can also be shown for parallel cases in German
that the constructions are assigned a lower degree of acceptability.
(91) a. Wen meint/ glaubt/ sagt/will Peter, dass der Chef wen
Who thinks /believes/ says/wants Peter that the boss who
entl€
asst?
fires
‘Who does Peter think/believe/say that the boss will fire?/Who
does Peter want the boss to fire?’
b. ??Wen fl€ ustert/ schreit Peter, dass der Chef wen entl€asst?
Who whispers/ shouts Peter that the boss who fires
‘Who does Peter whisper/shout that the boss will fire?’
At first glance, it might seem difficult to incorporate this class of verbs
into the analysis developed in sections 3.1 and 3.2: As was illustrated in
section 2.3, those verbs denote an assertive speech act in the same way as
ordinary verbs of saying, the difference being that they additionally
specify the mode of saying. Regarding their context change potential,
pure verbs of saying and manner-of-speaking verbs, therefore, cannot be
assumed to differ. The proposition expressed in the complement clause of
any of the verbs in (88) to (91) does not have to be considered part of the
speaker’s or both discourse participant’s knowledge system.
An observation which is also as old as the discussion of such extraction
constructions itself is that questions such as (87) to (91) improve in case
they are embedded in the right context. E.g. Erteschik-Shir & Lappin
(1979:68) assume for the sentence in (92) that it does not have to be
considered deviant (as happens in isolation) when the context is such that
it is uttered among the members of an editorial board.
(92) ??What did the paper editorialize that McGovern had done?
An adequate context does not even have to be constructed extra-
linguistically. Erteschik-Shir (2006:290) mentions (93).
(93) ?What did Truman Capote lisp that he’d do?
Asking about a random person what it is that s/he lisped results in a
question of a rather low degree of acceptability. Asking for the lisped
contents of Truman Capote (for whom it is known that he lisped)
describing his plans, however, leads to the extraction construction
improving. Similarly, the German sentence in (94) which is of a
questionable degree of acceptability when being judged in isolation, loses
its oddity when being uttered in a scenario (e.g.) in which three discourse
participants (one of them is Peter) eavesdrop on a conversation and
cannot speak to each other in a loud voice. Peter says something which the
other discourse partners do not understand. Therefore, they ask what
Peter whispered. In this context, the question in (94) is fully acceptable.
(94) Wen fl€
ustert Peter, dass Maria anruft?
Who whispers Peter that Mary on.calls

44 Sonja M€
uller
What the three examples referred to here have in common (and what will
be an essential part in trying to account for them in discourse structural
terms) is that they all allow using such a (rather unusual) matrix verb. If a
speaker simply wants to know about the issue reported on by the matrix
subject referent, by specifying the mode of saying that explicitly, s/he
behaves too informatively as s/he could simply have used the weaker
form and asked for what was said. However, if it is plausible in the
context why the speaker is asking for what another person e.g. whispered,
the degree of informativeness regarding the particular mode of saying is
legitimate.21 Similarly, it seems quite unusual to ask what someone lisped
if the person posing the question is interested in getting to know what the
individual conveyed e.g. regarding his plans in a previous conversation.
However, if it is known that the matrix subject referent lisps (which
applies to Truman Capote), that is, there is no other mode of expressing
the content asked for for the matrix subject referent, the matrix verb
again does not get considered too informative. Asking in a neutral
context what a paper editorialised (that is referring to a very particular
journalistic mode of reporting) when intending to simply find out what
was written about the issue expressed in the complement clause also
involves an exaggerated degree of explicitness in a situation where a more
general term would do better. However, when the context is such that
such a degree of explicitness is actually called for due to the
circumstances and the person asking really needs to know what has
been written in this particular way, the question again retains its
appropriate degree of informativity needed to present an adequate
contribution to the actual discourse situation. Within Erteschik-Shir’s
own account (which in its general outline has been held by her since her
first piece of work on this subject in 1973), the author interprets the effect
in (92) and (93) as a neutralisation of the additional meaning components
which differentiate manner-of-speaking verbs from pure verbs of saying,
thereby reducing the matrix verbs of their semantic complexity and
allowing a dominant interpretation of the that-clause (which is the
criterion governing (im)possible extractions from such contexts in
Erteschik-Shir 1973) (cf. also Erteschik-Shir 2007:194ff.). Van Valin &
La Polla (1997:630) – more than Erteschik-Shir (although very similar
intuitions can be assumed to be modelled) – choose an explanation which
explicitly builds on Grice’s (1975:45) Cooperative Principle (cf. (92)).
(95) Make your conversational contribution such as is required, at the
stage at which it occurs, by the accepted purpose or direction of the
talk exchange in which you are engaged.
Van Valin & La Polla (1997:630) write: “[. . .] the speaker’s choice of an
informationally richer expression [. . .] over another more neutral
21
I am grateful to Anika Dewald for pointing this out to me.

possibility [. . .] only makes sense in terms of the Cooperative Principle if

the manner of expression is in fact highly relevant to the main point of
the utterance.” Their explanation ties in with my analysis that using
manner-of-speaking verbs presents a choice of lexical material which can
be considered to be over-informative in a neutral context, that is unless
the high degree of informativeness is demanded by the context. Going
one step further, the low degree of acceptability of extractions from that-
complement clauses in which the matrix clause contains a manner-of-
speaking verb can be traced back to a pragmatic inadequacy which arises
due to a violation of the (second part of) Grice’s (1975) Maxim of
Quantity (cf. (96)).
(96) Maxim of Quantity
1. Make your contribution as informative as is required (for the
current purposes of the exchange).
2. Do not make your contribution more informative than is
required.
By relying on over-informativeness in the case of the negative influence
of manner-of-speaking verbs and uninformativeness in the case of factive
and implicative verbs, my account which intends to model the
restrictions on extractions from that-complement clauses – in the case
of all three verb classes – refers to discourse structural properties which
play an essential role in determining the pragmatic adequacy of
utterances in discourse which has, therefore, found its way into well-
known and acknowledged principles such as Grice’s Principle of
Cooperativity and one of its subordinate maxims. Treating the data in
(90), (91b), (92) and (93) as violations of the Maxim of Quantity seems
straightforward.
Intending to unify the reasons for the negative influence of all three
verb classes, it can be argued that question formation resulting from
extracting constituents from factive or implicative that-clauses also
presents a violation of a Gricean maxim – unless a respective context is
required in which the question gains informativity again: As the speaker
knows the answer and, nevertheless, asks the addressee to provide it,
under the interpretation assumed in section 3.1, formulating such
questions can also be considered to violate the Maxim of Quantity, in
this case the first part (cf. (96)). The more adequate (and, therefore,
informative) move in discourse would be to remind the hearer of the
shared knowledge (factive case) or provide the discourse partner with the
respective piece of information (implicative case). Apart from violating
the Maxim of Quantity, such a move in discourse can also be considered
to violate the first submaxime of the Maxim of Quality (cf. (97) taken
from Grice 1975:46).

46 Sonja M€
uller
(97) Maxim of Quality

Try to make your contribution one that is true.
1. Do not say what you believe to be false.
2. Do not say that for which you lack adequate evidence.
By opening up alternatives which in fact do not present alternatives
(either because the answer is already CG-information (factive verbs) or
because the speaker alone already knows the answer (implicative verbs)),
the speaker provides the hearer with possibilities about which s/he knows
that they are false, that is the addressee could in principle choose exactly
one of those answers which do not correspond to what the speaker
knows. The result would be a discourse whose repair would require a lot
of work on part of the speaker because s/he had to explain why s/he
asked such a misleading question if s/he already had the answer present
and intended a cooperative discourse.
Against the background of the argumentation in sections 3.1 to 3.4 in
which I assumed two different interpretations of the extraction construc-
tions, it is interesting that evoking the ‘high’ reading which I argued to
lead to an improvement of the questions (which I considered to be rather
bad without a context in which this particular reading arises) also leads
to an increase in acceptability in the case of manner-of-speaking verbs.
Therefore, apart from finding a general heading for the restrictions
concerning the occurrence of factive, implicative and manner-of-speaking
verbs in the form of the Gricean maxims and violations thereof, I can
also show that the interpretation of the questions argued to play a role
under the occurrence of factive and implicative matrix verbs is also of
relevance when it comes to manner-of-speaking verbs in extractions from
that-complements.
I argued that a question such as (98) whose meaning is expressed by the
partition in (99) leads to a violation of the Maxim of Quantity as a
speaker who utters this question in a neutral context aiming at getting to
know what Peter said, behaves too informatively by describing the mode
of saying in such a detailed way.
ustert Peter, dass der Chef entl€asst?, D = {Hans, Fritz}
(98) ??Wen fl€
Who whispers Peter that the boss fires

(99)
Peter flüstert, dass der Chef sowohl Hans als auch Fritz entlässt.
Peter whispers that the boss both Hans as also Fritz fires
‘Peter whispers that the boss will fire Hans as well as Fritz.’
Peter flüstert, dass der Chef Hans entlässt, Fritz nicht entlässt.
Peter whispers that the boss Hans fires, Fritz not fires
‘Peter whispers that the boss will fire Hans, will not fire Fritz.’
Peter flüstert, dass der Chef Fritz entlässt, Hans nicht entlässt.
Peter whispers that the boss Fritz fires, Hans not fires
‘Peter whispers that the boss will fire Fritz, will not fire Hans.’
Peter flüstert, dass der Chef weder Hans noch Fritz entlässt.
Peter whispers that the boss neither Hans nor Fritz fires
‘Peter whispers that the boss will fire neither Hans nor Fritz.’
As I explained in section 3, the reading of (98) in (99) results under the

‘low’ anchoring of the strong exhaustivity. However, as I assumed for
parallel questions under the occurrence of different matrix verbs, such
questions also allow a reading under which the strong exhaustivity is
anchored ‘high’. The partition induced in this case is presented in (100).

48 Sonja M€
uller
(100)
Peter flüstert, dass der Chef Hans entlässt, und
Peter whispers that the boss Hans fires and
Peter flüstert, dass der Chef Fritz entlässt.
Peter whispers that the boss Fritz fires
‘Peter whispers that the boss will fire Hans and
Peter whispers that the boss will fire Fritz.’
Peter flüstert, dass der Chef Hans entlässt, und
Peter whispers that the boss Hans fires and
Peter flüstert nicht, dass der Chef Fritz entlässt.
Peter whispers not that the boss Fritz fires
‘Peter whispers that the boss will fire Hans, and
Peter does not whisper that the boss will fire Fritz.’
Peter flüstert nicht, dass der Chef Hans entlässt, und
Peter whispers not that the boss Hans fires and
Peter flüstert, dass der Chef Fritz entlässt.
Peter whispers that the boss Fritz fires
‘Peter does not whisper that the boss will fire Hans, and
Peter whispers that the boss will fire Fritz.’
Peter flüstert nicht, dass der Chef Hans entlässt, und
Peter whispers not that the boss Hans fires and
Peter flüstert nicht, dass der Chef Fritz entlässt.
Peter whispers not that the boss Fritz fires
‘Peter does not whisper that the boss will fire Hans, and
Peter does not whisper that the boss will fire Fritz.’
Under such an interpretation of the question, the speaker aims at

learning which issues have been whispered and which ones have not, that
is a contrastive interpretation of different modes of conveying informa-
tion arises. Under such an interpretation, the plausible reason for posing

the question is not that one wants to know what has been said at all, but
one rather wants to know which parts have been conveyed in this way.22
As I illustrated in section 3.3, the interpretation in (100) can be evoked
contextually. In a sequence consisting of a question which aims at
learning in which mode Peter conveys the different dismissals conducted
by his boss, there could be sub-questions which ask for which dismissal
has been uttered in the respective mode (cf. (101)).
(101) question: Auf welche Art und Weise a€ußert Peter wessen
in which way and mode utters Peter whose
Entlassung?
dismissal
‘How does Peter utter whose dismissal?’
sub-question: Was ist mit Fl€ €
ustern? Uber wen
What is with whispering About whom
fl€
ustert er’s?
whispers he.it
‘What about whispering? About whom does he
whisper it?’
€
sub-answer: Peter /FLUStert, dass der Chef HANS\
Peter whispers that the boss Hans
entl€asst.
fires
‘Peter WHISpers that the boss will fire HANS.’

sub-question: Was ist mit Rumschreien? Uber € wen
What is with shouting About whom
schreit er es rum?
shouts he it around
‘What about shouting it around? About whom
does he shout it around?’
22
The conclusion that issues which have not been whispered are conveyed in a different
way is in fact highly plausible, but it is not necessary. Asserting that certain things are not
whispered (cf. (i)) can also mean that Peter does not say those things at all. However, if a
speaker wants to express that Peter does not say at all that the boss will fire Hans (and that
it is not only not the case that Peter does not say it in this mode), the sentence seems as over-
informative as the isolated question in (98).
(i) Peter fl€ ustert nicht, dass der Chef Hans entl€asst.
Peter whispers not that the boss Hans fires
‘Peter does not whisper that the boss will fire Hans.’
As the speaker behaves more informatively than when saying Peter does not say that. . .
under such a precise characterisation of the mode of conveyance and as a cooperative
speaker is assigned a motivation for her/his degree of informativeness, it is plausible to
assume that only the variance between the neutral sagen (say) and the mode of speaking
quietly and voicelessly (and not the oral conveyance at all) gets negated.

50 Sonja M€
uller
sub-answer: Peter schreit ‘/RUM, dass der Chef MaRIA\

Peter shouts around that the boss Mary
entl€
asst.
fires
‘Peter is shouting aROUND that the boss will

fire MAry.’
sub-question: Was ist mit Stammeln? Uber € wen
What is with stammering About whom
stammelt er’s (weil er’s nicht fassen
stammers he.it because he.it not believe
kann)?
can
‘What about stammering? About whom does
he stammer it (because he can’t believe it)?’
sub-answer: Peter /STAMmelt, dass der Chef PAUL\
Peter stammers that the boss Paul
entl€asst.
fires
‘Peter STAMmers that the boss will fire PAUL.’

Instead of the questions in (101), extraction constructions as in (102)
could occur as well.
(102) A: Ok, dann wissen wir nun also: Peter schreit
ok then know we now consequently Peter shouts
/RUM, dass der Chef HANS\ entl€asst, Peter
around that the boss Hans fires Peter
/STAMmelt, dass der Chef PAUL\ entl€asst und was
stammers that the boss Paul fires and what
ist mit fl€ ustern, ?wen /FLÜStert Peter, dass der
is with whispering who whispers Peter that the
Chef entlässt?
boss fires
The question in (102) is more acceptable in this context and with this
intonation than when being uttered in isolation as in (103).
(103) ??Wen fl€
ustert Peter, dass der Chef entl€asst?
Who whispers Peter that the boss fires
The question in (102) (or the preceding answers) express that there are
further questions of the form ‘About whom does Peter express the
dismissal in the mode X?’. In addition, the implicature is triggered that
Peter expresses different dismissals in different modes. After one mode of
expression has been ticked off (i.e. e.g. that the speaker has named all

cases which have been stammered), it is clear that – due to the parallel
questions to the stammered cases – there are further modes of expression
contributing to answering the ‘highest’ question and the inference is
triggered that Peter speaks about the other dismissals in a different mode.
A question such as in (102), therefore, displays a reading under which the
inadequacy assumed for (103) (lack of motivation for asking for a
whispered issue) gets dissolved. In case there is reason to assume that
there are whispered and non-whispered issues, posing such a question is
fully legitimate. If a context evokes the reading of a question which can
be imagined appropriate (and which it is denied to be in isolation), the
result is an increase in the construction’s acceptability. The possibility of
manipulating the factors involved in the extraction constructions
containing manner-of-speaking verbs in such a precise way can be
considered to speak in favour of my pragmatic/discourse-semantic
analysis developed for extraction constructions whose matrix clause
contains factive or implicative verbs in sections 3.1 to 3.4.
4. Conclusion
This paper has proposed an account which traces the non-bridge verb
quality of factive, implicative and manner-of-speaking verbs in extrac-
tions from that-complement clauses back to the effect that the resulting
questions have on discourse. By relying on very basic and well-
acknowledged pragmatic principles such as Searle’s (1969) felicity
conditions for the speech act question as well as Grice’s (1975)
Cooperative Principle (plus the Conversational Maxims of Quantity and
Quality), the overall claim with respect to degraded cases of extractions
from that-complements is that the resulting questions present inadequate
moves in discourse. They involve a violation of felicity conditions of the
speech act involved as well as a violation of the demand for cooperativity
in general. In the case of manner-of-speaking verbs, violating part two of
the Maxim of Quantity can be proven: If a speaker poses a question
which asks for whispered, stammered or shouted contents while actually
pursuing the aim to get to know what the issue reported on by the matrix
subject referent exactly looks like, s/he behaves too informatively as s/he
could have used a weaker description than specifying the mode of saying
so very explicitly (cf. section 3.5). In the case of factive and implicative
complements, violating the (first part of the) preparatory rule of the
illocutionary type question (S does not know ‘the answer’, i.e. [. . .] does
not know the information needed to complete the proposition truly [. . .].)
can be interpreted as violating the first part of the Maxim of Quantity
(Make your contribution as informative as is required (for the current
purposes of the exchange.)) and/or the first submaxim of the Maxim of
Quality (Do not say what you believe to be false.). The proposition
expressed in the complement embedded by a factive matrix verb presents
52 Sonja M€
uller
CG-content (at least in the default case), that is knowledge consciously

shared between speaker and addressee. The proposition expressed in an
implicative that-complement presents knowledge on part of the speaker.
By asking for information of that sort, on the one hand, the speaker asks
although s/he does know the answer, on the other hand, s/he leaves
wrong alternatives for the addressee to choose from (cf. section 3.1). This
analysis of the negative influence of certain matrix verbs on extraction
constructions is presented in section 3.1 (factive and implicative verbs)
and section 3.5 (manner-of-speaking verbs) after section 2 has provided
some background assumptions on the formal modelling of contexts as
well as an interpretation of the meaning components involved in the three
verb classes from a discourse perspective. The main argument for
choosing a discourse structural perspective for modelling the restrictions
is the observation that changing the question illocution or employing a
particular intonation (which is again argued to be associated with a
particular discourse structural interpretation) results in dissolving the
pragmatic inadequacy made responsible for the questions’ degradation
which goes hand in hand with an increase in acceptability. Mitigating the
question illocution or manipulating the distribution of knowledge leading
to the question’s uninformativity can improve the questions’ acceptabil-
ity in case a speaker asks for known contents. Building a context in which
asking for what has been communicated in a very particular way is
actually needed, a question involving a manner-of-speaking verb can be
proven to retain its appropriateness and, thereby, its acceptability.
Discussion of erothetic illocutionary subtypes and different interpreta-
tions of the questions (‘high’ vs. ‘low’ strong exhaustivity) and their
impact on the extraction constructions is part of sections 3.2.1 (factive
and implicative verbs in interaction with the question illocution), 3.2.2,
3.3 and 3.4 (factive and implicative verbs and question interpretation) as
well as section 3.5 (manner-of-speaking verbs and question interpreta-

tion). Section 3.4 provides a motivation for assuming the two readings of
the questions and their corresponding partitions.23
To put the insights of this paper into a nutshell, by analysing the effect
that questions (which result from extracting a constituent from a that-
complement clause) have on a discourse context, it argues that violating
pragmatic conditions such as felicity conditions of the speech act
involved or even more general principles of cooperative moves in
conversations have an influence on the acceptability of linguistic
expressions.
References
, M. 2008. Contradiction and Grammar: The Case of Weak Islands. Ph.D.
ABRUSAN
dissertation, MIT. Accessible via: http://sites.google.com/site/martaabrusan/.
Accessed at: 29/10/2010.
, M. i.E. Presuppositional and Negative Islands. To be published in:
ABRUSAN
Natural Language Semantics. Accessible via: http://sites.google.com/site/
martaabrusan/. Accessed at 29/10/2010.
ADLI, A. (2004). Grammatische Variation und Sozialstruktur. Berlin: Akademie
Verlag.
BARTELS, C. 1999. The intonation of English statements and questions: A
compositional approach. New York/London: Routledge.
BARTSCH, R. 1973. “Negative Transportation” gibt es nicht. Linguistische Berichte
27:1–7.
23
My account stands in the tradition of accounts developed by Erteschik-Shir (1973,
Erteschik-Shir & Lappin 1979, Erteschik-Shir 2007) or Abrusan (2009). I cannot provide a
detailed comparison at this point. However, I would like to make a few remarks on
similarities and differences.
Erteschik-Shir (1973) very early suggests that dominance can account for (un)grammatical
extractions from that-complements. Her ideas are in so far similar to mine in that she also
argues that it is not matrix verbs alone which inherently code the (non-)dominance of their
complement, but that this property can be modified in particular utterance contexts.
However, her concept is of a rather vague and intuitive nature and it might get a more
concrete definition within my account when spelling it out by relying on the status the
complement propositions have in discourse. Non-dominant complements (which block
extraction in Erteschik-Shir’s accounts) might then be complements whose propositions
already belong to the CG (factive complements) or are at least believed by the speaker
(implicative complements). Erteschik-Shir is concerned with a lot of different extraction
constructions and mainly deals with English data. She focuses on the difference between
neutral verbs of saying and manner-of-speaking verbs when discussing that-complements.
However, she does not look as detailed as I do only at extractions from that-complements
describing three negatively influencing verb classes.
Abrus an (2009) looks at different factive constructions, but does not offer an account for
other verb classes when looking at extractions from that-complements. She considers factive
that-complements weak islands. Her account would, therefore, predict the perfect
acceptability of all marked data I presented in this article. Furthermore, her approach is
purely semantic. She can speak about semantic properties of factive environments and
question interpretations. However, she cannot explain (and does not look at) contextual
effects or illocutionary types which I argue to also influence extractions from that-
complements in German.

54 Sonja M€
uller
BASSE, G. 2008. Factive complements as defective phases. Proceedings of the 27th

West Coast Conference on Formal Linguistics, ed. N. Abner & J. Bishop, 54 –
62. Massachussetts: Cascadilla Proceedings Project.
€
BURING , D. 1997. The meaning of topic and focus – The 59th Street Bridge Accent.
London: Routledge.
€
BURING , D. 2003. On D-trees, beans, and B-accents. Linguistics and Philosophy 26
(5): 511–545.
CAPONIGRO, I. & SPROUSE, J. 2007. Rhetorical questions as questions. Proceedings
of Sinn und Bedeutung 11:121–133. Barcelona.
CATTELL, R. 1978. On the source of interrogative adverbs. Language 54:61–77.
CHOMSKY, N. 1977. On wh-movement. Formal Syntax, ed. P. Culicover,
T. Wasow & A. Akmajian, 71–132. New York: Academic Press.
CINQUE, G. 1990. Types of A’-dependencies. Cambridge, Massachusetts: MIT
Press.
COMOROVSKI, I. 1996. Interrogative phrases and the syntax-semantics interface.
Dordrecht: Kluwer.
DAYAL, V. 1996. Locality in WH quantification: Questions and relative clauses in
Hindi. Dordrecht: Kluwer.
DE CUBA, C. 2006. The adjunction prohibition and extraction from non-Factive
CPs. Proceedings of the 25th West Coast Conference on Formal Linguistics, eds.
D. Baumer, D. Montero & M. Scanlon, 123–131. Sommerville, MA: Cascadilla
Proceedings Project.
ERTESCHIK-SHIR, N. 1973. On the nature of island constraints. Ph.D. dissertation,
MIT.
ERTESCHIK-SHIR, N. 1997. The dynamics of focus structure. Cambridge:
Cambridge University Press.
ERTESCHIK-SHIR, N. 2006. Bridge phenomena. The Blackwell Companion to
Syntax, Vol.1, ed. M. Everaert & H. van Riemsdijk, 284–294. Oxford:
Blackwell.
ERTESCHIK-SHIR, N. 2007. Information structure. The syntax-discourse interface.
Oxford: Oxford University Press.
ERTESCHIK-SHIR, N. & LAPPIN, S. 1979. Dominance and the functional
explanation of island pheonomena. Theoretical Linguistics 6:41–84.
FANSELOW, G. 1987. Konfigurationalit€ at. Untersuchungen zur Universalgrammatik
am Beispiel des Deutschen. T€ ubingen: Stauffenburg.
FARKAS, D. 1992. On the semantics of subjunctive complements. Romance
Languages and Modern Linguistic Theory. Papers from the 20th Linguistic
Symposium on Romance Languages (LSRL XX), Ottawa, 10–14 April 1990,
eds. P. Hirschb€ uhler & P. K. Koerner, 69–105. Amsterdam/Philadelphia:
Benjamins.
FARKAS, D. 2003. Assertion, Belief and Mood Choice. Presented at ESSLLI,
Conditional and unconditional modality workshop. Vienna. Accessible via:
people.ucsc.edu/∼farkas/papers/mood.pdf. Accessed at: 06/03/2010.
FARKAS, D. & BRUCE, K. 2009. On reacting to assertions and polar questions. To
appear in Journal of Semantics. Accessible via: http://people.ucsc.edu/~farkas/
pa-pers.html. Accessed at: 07/09/2010.

FERY , C. 1993. German intonational patterns. T€ ubingen: Niemeyer.
FRANCK, D. 1980. Grammatik und Konversation. Scriptor, K€ onigstein, Ts.
FUKUI, N. 1986. A theory of category projection and its applications, Ph.D.
dissertation, MIT.
GAJEWSKI, J. 2002. On analyticity in natural language, Ms. University of
Connecticut, Accessible via: http://gajewski.uconn.edu/research.html. Accessed
at: 21/02/2010.

GAJEWSKI, J. 2008. L-triviality and grammar. Handout UConn Logic Group

27/02/2009. Accessible via: http://gajewski.uconn.edu/research.html. Accessed
at: 04/03/2010.
GIANNAKIDOU, A. 1998. Polarity sensitivity as (non)veridical dependency.
Amsterdam/Philadelphia: Benjamins.
GREWENDORF, G. 1988. Aspekte der deutschen Syntax. Eine Rektions-Bindungs-
Analyse. T€ubingen: Stauffenburg.
GRICE, H. P. 1975. Logic and conversation. Speech Acts, ed. P. Cole & J. Morgan,
41–58. New York: Academic Press.
GROENENDIJK, J. & STOKHOF, M. 1982. Semantic analysis of WH-complements.
Linguistics and Philosophy 5:175–233.
GROENENDIJK, J. & STOKHOF, M. 1984. On the semantics of questions and
the pragmatics of answers. Varieties of Formal Semantics, ed. F. Landman &
F. Veltman, 143–170. Dordrecht: Foris.
GROENENDIJK, J. & STOKHOF, M. 1997. Questions. Handbook of Logic and
Language, eds. van Benthem J. & A. ter Meulen, 1055–1124. Amsterdam:
Elsevier.
GUNLOGSON, C. 2003. True to form. Rising and falling declaratives as questions in
English. New York: Routledge.
HAMBLIN, C. L. 1958. Questions. Australasian Journal of Philosphy 36:159–168.
HAMBLIN, C. L. 1973. Questions in Montague English. Foundations of Language
10:41–53.
HEGARTY, M. 1990. On adjunct extraction from complements. Papers on
Wh-Movement, MIT Working Papers in Linguistics 13, eds. L. Cheng &
H. Demirdash, 101–124. Cambridge, MA.: MIT.
HEIM, I. 1992. Presupposition projection and the semantics of attitude verbs.
Journal of Semantics 9: 183–221.
HEIM, I. 1994. Interrogative semantics and Karttunen’s semantics for Know. The
Proceedings of the Ninth Annual Conference and the Workshop on the Discourse
of the Israel Association for Theoretical Linguistics, ed. R. Buchalla &
A. Mittwoch, 128 –144. Jerusalem: Akademon Press.
HIGGINBOTHAM, J. 1991. Interrogatives I. MIT Working Papers in Linguistics
14:47–76.
HIGGINBOTHAM, J. 1993. Interrogatives. Formal semantics. The essential readings,
eds. P. Portner & B. H. Partee, 687–709. Oxford: Blackwell.
HIGGINBOTHAM, J. 1996. The semantics of questions. The handbook of
contemporary semantic theory, ed. S. Lappin, 361–383. Oxford: Blackwell.
HIGGINBOTHAM, J. & MAY, R. 1981. Questions, quantifiers, and crossing.
Linguistic Review 1:41–80.
HORN, L. 1975. Neg-raising predicates: Toward an explanation. Chicago
Linguistic Society (CLS) 11:279–294.
JACKENDOFF, R. 1972. Semantic interpretation in generative grammar. Cambridge,
Mass./London: MIT Press.
KADMON, N. 2001. Formal pragmatics. Semantics, pragmatics, presupposition and
focus. Oxford: Blackwell.
KARTTUNEN, L. 1971. Implicative verbs. Language 47:340–358.
KARTTUNEN, L. 1977. Syntax and semantics of questions. Linguistics and
Philosophy 1:3–44.
KLUENDER, R. 1991. Deriving island constraints from principles of predication.
Island constraints. Theory, acquisition and processing, eds. H. Goodluck &
M. Rochemont, 223–258. Dordrecht/Boston/London: Kluwer.

56 Sonja M€
uller
KRIFKA, M. 2001. For a structured meaning account of questions and answers.

Audiatur Vox Sapientia. A Festschrift for Arnim von Stechow, eds. C. Fery & W.
Sternefeld, 287–319. Berlin: Akademieverlag.
KWON, M.-J. 2005. Modalpartikeln und Satzmodus. Untersuchungen zur Syntax,
Semantik und Pragmatik der deutschen Modalpartikeln. Ph.D. dissertation,
M€ unchen. Accessible via: Deutsche Nationalbibliothek: http://deposit.ddb.de/
cgi-bin/dokserv?idn=97887653. Accessed at: 19/04/2009.
LEWIS, D. 1979. Scorekeeping in a language game. Journal of Philosophical Logic
8:339–359.
LOHNSTEIN, H. 2000. Satzmodus kompositionell. Zur Parametrisierung der
Modusphrase im Deutschen. Berlin: Akademie Verlag.
LUUKKO-VINCHENZO, L. 1988. Formen von Fragen und Funktionen von Frages€ atzen.
Eine deutsch-finnische kontrastive Studie unter besonderer Ber€ucksichtigung der
Intonation. T€ubingen: Niemeyer.
MEHLHORN, G. 2001. Produktion und Perzeption von Hutkonturen im
Deutschen. Linguistische Arbeitsberichte 77, Leipzig, 31–57.
€
MULLER , S. 2011a. (Un)informativit€ at und Grammatik. Extraktion aus
Nebens€ atzen im Deutschen. T€ubingen: Stauffenburg.
€
MULLER , S. 2011b. Extraktionsinseln. Zur ihrer Syntax, Semantik und
Informationsstruktur. Trier: Wissenschaftlicher Verlag Trier.
€
MULLER , S. 2012. The distribution of knowledge in unacceptable questions.
Modality and Theory of Mind Elements across Languages, eds. W. Abraham &
E. Leiss, 147–210. Berlin: de Gruyter.
OSHIMA, D. 2007. On factive islands: pragmatic anomaly vs. pragmatic infelicity.
New Frontiers in Artificial Intelligence: Joint JSAI 2006 Workshop Post-
Proceedings, eds. T. Washio, K. Satoh, H. Terada & A. Inokuchi, 147–161.
Dordrecht: Springer.
PESETSKY, D. 1987. Wh-in situ: Movement and unselective binding. The
representation of indefiniteness, eds. E. Reuland & G. ter Meulen, 98–129.
Cambridge, MA: MIT Press.
PORTNER, P. 2005. The semantics of imperatives within a theory of clause types.
Proceedings of Semantics and Linguistic Theory 14, eds. W. Kazuha &
R. Young, 235–252. Ithaca, New York: CLC Publications.
€
€ , H. 1975. Uber
PUTZ die Syntax der Pronominalform ‚es’ im modernen Deutsch.
T€ ubingen: Stauffenburg.
REICH, I. 2003. Frage, Antwort und Fokus. Akademie Verlag, Berlin.
REIS, M. & ROSENGREN, I. 1992. What do WH-imperatives tell us about
WH-movement? Natural Language and Linguistic Theory 10(1): 79–118.
RIZZI, L. 1990. Relativized Minimality. Cambridge, Massachusetts/London,
England: MIT Press.
RIZZI, L. 2004. Locality and Left Periphery. Structures and Beyond, ed.
A. Belletti, 223–251. Oxford: Oxford University Press.
ROBERTS, C. 1996. Information structure: Towards an integrated formal theory of
pragmatics. Ohio State University Working Papers in Linguistics 49, ed. J.-H.
Yoon & A. Kathol, 91–136. Ohio: Papers in Semantics.
ROSS, J. R. 1967. Constraints on variables in syntax. Ph.D. dissertation, MIT.
SEARLE, J. R. 1969. Speech acts. An essay in the philosophy of language. New
York: Cambridge University Press.
SIMONS, M. 2007. Observations on embedding verbs, evidentiality, and
presupposition. Lingua 117(6): 1034–1056.
STALNAKER, R. 1974. Pragmatic presuppositions. Semantics and Philosophy, ed.
M. Munitz & P. Unger, 197–213. New York: University Press. Reprinted in

Pragmatics. A Reader, ed. Davis, S. 1991:471–482. New York: Oxford

University Press.
STALNAKER, R. 1978. Assertion. Pragmatics, ed. P. Cole, 315–332. New York:
Academic Press.
STALNAKER, R. 2002. Common ground. Linguistics and Philosophy 25:701–721.
STOWELL, T. 1986. Null antecedents and proper government. Proceedings of
the North Eastern Linguistic Society 16, ed. S. Berman, J.-W. Choe &
J. Mc Donough, 476–493. Amherst.
SZABOLCSI, A. & ZWARTS, F. 1993. Weak islands and an algebraic semantics for
scope taking. Natural Language Semantics 1: 235–285.
THURMAIR, M. 1989. Modalpartikeln und ihre Kombinationen. Niemeyer:
T€ubingen.
TRUCKENBRODT, H. 2004. Zur Strukturbedeutung von Interrogativs€ atzen.
Linguistische Berichte 199: 313–350.
TRUSWELL, R. 2007. Locality of Wh-movement and the Individuation of Events.
Ph.D. dissertation, University College London.
UHMANN, S. 1991. Fokusphonologie. Eine Analyse deutscher Intonationskonturen
im Rahmen der nicht-linearen Phonologie. T€ubingen: Niemeyer.
WINKLER, E. 1992. Modalpartikeln in selbst€andig verwendeten Verbends€ atzen.
ur Phonetik, Sprachwissenschaft und Kommunikationsforschung 45
Zeitschrift f€
(1):30–48.
Received May 25, 2012 Sonja M€uller
Accepted July 30, 2013 Faculty of Linguistics and Literary Studies
Bielefeld University
Box 10 01 31, D-33501 Bielefeld
Germany
[email protected]

Copyright of Studia Linguistica is the property of Wiley-Blackwell and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.

Deriving Island Constraints With Searle and Grice. A Pragmatic Account of Bridge Verbs

Uploaded by

Copyright:

Available Formats

Deriving Island Constraints With Searle and Grice. A Pragmatic Account of Bridge Verbs

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Deriving Island Constraints With Searle and Grice. A Pragmatic Account of Bridge Verbs

Uploaded by

Copyright:

Available Formats

DERIVING ISLAND CONSTRAINTS WITH

SEARLE AND GRICE.

Abstract. This paper proposes a pragmatic account of restrictions on extractions

Erteschik-Shir 1973) and stuck to in the decades of research to come

© 2014 The Editorial Board of Studia Linguistica

present a non-structural approach to (im)possible extractions from that-

© 2014 The Editorial Board of Studia Linguistica

present inadequate moves in discourse in the following sense: In the case

2. Contexts and their formal modelling

2.1. Simple sentences: Assertions and constituent questions

© 2014 The Editorial Board of Studia Linguistica

(5) Who does Anna invite for coﬀee?, D = {Julia, Stephan}

A demand that an adequate proposition has to meet in a context is its

2.2. Complex sentences: two-stage context changes

propositions happens indirectly: Propositions are anchored in worlds

© 2014 The Editorial Board of Studia Linguistica

Diﬀerentiating between the eﬀect that the extraction construction has on

b. Anna invites Stephan as well as Caro for coffee. = p5

As will be shown in the next section, – depending on the matrix verbs

2.3. Factive, implicative and manner-of-speaking verbs

© 2014 The Editorial Board of Studia Linguistica

presuppositions) shows in (15) and (16), the ﬁrst class of matrix

© 2014 The Editorial Board of Studia Linguistica

proposition (=p2) is true when the CS gets reduced of the worlds in

© 2014 The Editorial Board of Studia Linguistica

(22) main context Peter’s ‘cause system’

Further exemplars of implicative verbs are e.g. herbeif€

© 2014 The Editorial Board of Studia Linguistica

3. Deriving the restrictions: Deficient moves in discourse

3.1. Uninformativity: The case of factive and implicative verbs

© 2014 The Editorial Board of Studia Linguistica

As the context update in the way illustrated above involves both

© 2014 The Editorial Board of Studia Linguistica

Interpreting the less acceptable question in (34) in analogy to (32), that is

As the proposition expressed by a factive complement clause presents

© 2014 The Editorial Board of Studia Linguistica

CG’ = CG [ {p5}∨ I’i,w = Ii,w[ {p1}∨

© 2014 The Editorial Board of Studia Linguistica

(39) ??Wen verursacht Peter, dass der Chef entl€asst?

© 2014 The Editorial Board of Studia Linguistica

The issue expressed in a subordinate clause embedded under an

© 2014 The Editorial Board of Studia Linguistica

which is therefore incompatible with her/his real knowledge. Such a

3.2. Repair by changing the question’s intention

3.2.1. Subtypes of question illocution

© 2014 The Editorial Board of Studia Linguistica

weakening) as well as a reinforcement of the contradiction in discourse

© 2014 The Editorial Board of Studia Linguistica

© 2014 The Editorial Board of Studia Linguistica

We have met the two of them together several times when we

© 2014 The Editorial Board of Studia Linguistica

isolation does not lead to an increase in acceptability. The factor made

© 2014 The Editorial Board of Studia Linguistica

(53) a. Wie ich das bloß am besten mache?

3.2.2. A further informative reading

© 2014 The Editorial Board of Studia Linguistica

(57) Wen glaubt Fritz, dass Maria eingeladen

© 2014 The Editorial Board of Studia Linguistica

complete description of the actual world regarding the issue of Mary