Review of Formal Pragmatics. Semantics, Pragmatics, Presupposition ...

Review of Formal Pragmatics. Semantics, Pragmatics,
Presupposition, and Focus, N. Kadmon, Blackwell
Publishers, Oxford, 2001, 430 pages, ISBN 0-631-20120-3
Robert van Rooij
Institute for Logic, Language, and Computation
Nieuwe Doelenstraat 15
1012 CP Amsterdam
Formal Pragmatics discusses various topics that are at the borderline of the semantics
and pragmatics of natural language, in particular the treatment of indefinites and anaphora,
presuppositions, and the analysis of focus. Enormous progress has been made in the last two
decades in the analysis of these phenomena. It is time for this body of work to be summarized
and described in a relatively easy accessible way such that students could become familiar
with it. Kadmon’s 430 page long book does exactly that. It covers a substantial body of
formal work on linguistic phenomena, and – according to the back cover – serves both as
a textbook, as well as a reference or research book. It presupposes only a very minimal
knowledge of formal semantics, but still manages to lead students through recent formal
developments in the subjects treated in the book. However, this book is not written as a
neutral overview or textbook: on controversial issues Kadmon always makes clear which
analyses she prefers, and sometimes (especially on the analysis of focus) the book reports
on novel research by herself and others. I have not used this book in my classes on formal
pragmatics. Thus, I will evaluate this book mostly as a reference and research monograph.
Having firm views regarding most of the subjects covered in this book, I cannot help it that
these opinions sometimes influence my following discussion.
The first part of the book deals with dynamic semantics. After a useful introductory
discussion on conversational implicatures, presuppositions, and the interplay between con-
text and content, the book consists of three parts: (i) dynamic semantics and its treatment
1

---

of indefinites and pronouns; (ii) the treatment of presupposition within dynamic semantics;
(iii) the analysis of focus-dependent interpretation.
The first part is about dynamic semantics and contains three chapters. In chapter 2,
Kamp’s Discourse Representation Theory (DRT) and Heim’s File Change Semantics (FCS)
are introduced, and chapters 3 and 4 mainly report on Kadmon’s own work on numerical
determiners and anaphora she did for her PhD dissertation. Although the part of dynamic
semantics is perhaps not complete— Groenendijk & Stokhof’s Dynamic Predicate Logic
(DPL), for instance, is not discussed at all—, the phenomena discussed here (and in the rest
of the book) are always covered in a way easily accessible to a wide audience. The main
ideas are always first discussed in an informal way, followed later by the formal definitions.
Chapter 3 exemplifies one of the perhaps most important insights of dynamic semantics
that is relevant for the analysis of pronouns and the pragmatic analysis of conversational
implicatures: that we can assume that although Leif has four chairs and Leif has at least four
chairs give rise to the same truth conditions, they still have a different semantic meaning.
This is especially important for pragmatic theories of conversational implicatures, because
here it has always been a problem how to account for the intuition that only the former type
of example gives rise to an ‘exactly’-implicature, if both sentences have the same semantic
meaning.
In chapter 4 Kadmon argues in favor of the uniqueness assumption of (singular) anaphora.
Though I agree to a large extent with her argumentation, I find it somewhat strange that
first DRT/FCS is introduced as the theory to deal with indefinites and anaphora, and that
later the perhaps most crucial assumption behind its analysis of the anaphoric dependency
is argued against. As far as I can see, the perhaps most crucial claim of DRT/FCS is that
Evans’ E-type analysis of unbound pronouns gives rise to the wrong predictions and thus
that contexts should be modeled by sets of possibilities that are finer grained than possible
worlds (even when a token analysis is assumed). But by adopting a uniqueness assumption
of anaphora, it is exactly this aspect of dynamic semantics that Kadmon proposes to give
up. One wonders why we then should still use standard dynamic semantics.
The second part of the book, chapters 5 to 11, deals with presuppositions. Chapter 5
basically discusses the satisfaction theory of presuppositions due to Karttunen, Stalnaker,
and Heim. The first two assume that to account for presuppositions of complex sentences,
we should not look at the truth conditional content of a sentence (as was proposed in par-
tial logic approaches to presuppositions) but rather at the way we process or interpret the
2

---

sentence, i.e., in terms of the way what is presupposed changes during the interpretation
of a sentence. On the basis of this, plus the assumption that we interpret/process conjunc-
tive sentences sequentially, it is predicted that such sentences behave asymmetrically with
respect to presuppositions. Kadmon (pp. 123-125) finds the Karttunen/Stalnaker approach
explanatory inadequate because it can’t derive the presupposition of a conjunctive sentence
in terms of the content of this sentence. Dynamic semantics (especially the versions of Heim
and Groenendijk & Stokhof) can, because it equates meaning/content with context change.
But I don’t think that using dynamic semantics makes the analysis any more (or less) ex-
planatory adequate: the explanatory force of both the Karttunen/Stalnaker account and
the dynamic semantics account of presuppositions comes from the sequential way we (by
assumption) interpret conjunctive sentences, and it is quite irrelevant as far as explanatory
adequacy is concerned whether we identify this dynamic change of context with the content
of the sentence (as proponents of dynamic semantics do) or not (Stalnaker never did). But,
of course, Kadmon is right in claiming that these two-dimensional theories missed something
that dynamic theories improved upon: they didn’t discuss in enough detail the interaction
between meaning and context change, which is especially crucial once we look below the
level of the clause.
In chapter 6 the satisfaction theory of presuppositions is mainly contrasted with the can-
cellation approach mainly due to Gazdar (1979). Whereas the satisfaction theory basically
assumes that what is presupposed by a sentence should (in the ideal or normal case) already
be common ground, according to the cancellation approach a presupposition is just an item
of information one can (by default) infer from a sentence, very much like a conversational
implicature.
Though Kadmon finds the satisfaction approach intuitively more appealing, it is also
problematic in that we sometimes use a sentence that triggers a presupposition, although
the content of this presupposition is not yet common ground. A popular way to get rid
of some of such examples, and as extensively discussed in chapters 9 and 10 of Kadmon’s
book, is by making use of presupposition accommodation: we accommodate the context such
that the presupposition is satisfied in the context after all. Presupposition accommodation
should be thought of, I believe, as a form of Gricean exploitation. A sentence gives rise
to a presupposition because it can be used appropriately only in a context in which this
presupposition is met, or satisfied. But once it becomes common knowledge that a sentence
can be used appropriately only in certain kinds of contexts, this common knowledge can be
3

---

exploited by a speaker: she can already use the sentence even if this contextual requirement
is not met, because she can rely on the hearer’s ability to recognize what kind of context is
required and to produce one by standard abductive reasoning. I think it is very natural to
assume such a mechanism, at least, if presupposition accommodation is the exception rather
than the rule (in the latter case, the motivation for working in the satisfaction framework
seems blurred), and if it is limited to what is known as ‘global accommodation’. Based on
some suggestions of Heim, and defended more fully by authors such as Roberts, van der
Sandt, and Geurts, Kadmon proposes that many problematic examples for the satisfaction
analysis noticed by Gazdar and others concerning, for instance, negation, The king of France
is not bold, because there is no king of France, disjunction, I will meet either the king or the
president of Bessarabia, and possibility statements, It is possible that John used to smoke and
it is possible that he stopped doing so should be accounted for by non-global accommodation.
I agree that once one allows for non-global accommodation, these examples cease to be a
problem. However, with authors like Heim and Beaver I would be very reluctant to use a
notion like local accommodation. If we would allow for non-global accommodation all too
easily, we would give up the distinction between what is presupposed and what is asserted
(and/or conversationally implicated) too easily. By doing so, we also threaten to give up
the main idea behind the satisfaction analysis of presuppositions: that with the use of a
presupposition trigger, we indicate something about the context’s common ground with
respect to which we make the utterance containing this trigger.
Consider an example similar to one she discusses herself. Suppose we observe a woman
that we have never seen before behaving very strangely. We try to figure out why she
behaves the way she does. I tell you: Perhaps she stopped smoking. For such cases Kad-
mon is obviously correct: I can say this without taking it to be common ground that this
woman used to smoke. Moreover, I also don’t want to claim that she smoked. Globally
accommodating the context with the information that the woman used to smoke gives rise
to the wrong prediction as well: although this information is not incompatible with what is
presupposed, we don’t infer from the sentence that the woman actually used to smoke. In
these cases, Kadmon proposes, local accommodation is the natural alternative: the sentence
just means that Perhaps (the woman used to smoke and has stopped smoking). For reasons
explained above, I don’t think this is the way we should go. I would prefer an analysis
where the sentence still gives rise to a gobal presupposition, but one that is weaker than just
the factual information that the woman used to smoke. In fact, this weaker presupposition
4

---

is just Perhaps the woman used to smoke. Thus, here and elsewhere, we can avoid local
accommodation by globally accommodating a weaker presupposition.
Chapter 10 deals with presuppositions of quantified sentences.
It deals partly with
the binding problem faced by Karttunen & Peters’ well-known two-dimensional analysis of
presuppositions. Heim (1983) was the first to see that once we account for presuppositions
in dynamic semantics, we are not faced with the binding problem anymore, because we allow
for the fact that the interpretation of presupposition and assertion can be (anaphorically)
dependent on one another. Though I agree with Kadmon that the universal presupposition
she predicted for existential sentences was too strong, I also think that Heim’s universal
presupposition for quantified sentences like Every German loves his Mercedes is much better
than Kadmon suggests: It is well known that the interpretation of quantificational sentences
depend on a contextually given domain of quantification. But this means that Heim only
predicts a universal reading with respect to such a domain, which sometimes can be globally
accommodated to the initial context of interpretation (perhaps as a topic, as suggested by
David Beaver). In this way we predict the same readings for quantified sentences as favored
by Kadmon – our example sentence is true if and only if every (male) German who has a
Mercedes loves it, without saying anything about the poor Germans without a Mercedes –
although the sentence does put a presuppositional constraint on the initial context in which
the sentence could be used appropriately, just as the satisfaction account would have it.
Chapter 11 deals with the standardly neglected issue whether what is presupposed by
a simple clause is conventionally coded in the semantic meanings of the lexical items con-
tained, or whether it also depends on features of the context. Kadmon argues convincingly
that at least sometimes a presupposition is triggered conversationally, and that these presup-
positions give rise to projection behavior quite similar to the standard ones (triggered by a
lexical item like stop, for instance). Adopting the satisfaction theory of presuppositions, she
rightly argues against Levinson’s (1983) suggestion that presupposition and conversational
implicature are basically the same thing: while presuppositions help us (in the ideal case)
to remember what was already commonly known when the sentence was uttered, a scalar
implicature, for instance, tells us only something about what the speaker knew at this prior
state.
The third part of the book – almost 200 pages long – deals with the prosodic mani-
festation, and the model-theoretic meaning and function of focus. Prosodic manifestation
is dealt with in chapter 12 and section 13.4. The latter section addresses the relationship
5

---

between (lack of) prosodic prominence and (lack of) new information.
The standard analyses (especially Rooth, 1984) of ‘only’ as discussed in chapter 14
propose that (1a) should be interpreted as (1b) (ignoring now the distinction between pre-
supposition and assertion, and where [A] stands for the denotation of A and Alt(A) for the
set of alternatives of A.)
(1) a. John only painted [apples]F .
b. [only φ] = {w ∈ [φ]|∀p ∈ Alt(φ) : p is true in w → [φ] = p}
As discussed in chapter 15, the following examples are well-known to be problematic for
the above interpretation rule: (1b) would falsely rule out (the possibility) that John painted
apples for all three examples.
(2) a. John only painted [apples and pears]F .
b. John only painted [apples or pears]F .
c. John only painted [a still-life]F .
Kadmon suggests that replacing the identity relation ‘=’ in (1b) by entailment, or better
(to account for (2c)) by Kratzer’s notion of lumping helps to solve these problems. Though I
agree that lumping helps, I believe that making use of it obscures rather than clarifies what
is going on with the analysis of only. The following somewhat different analysis gives rise
to the same predictions as Rooth’s (1b) when it is proper names that are focussed, but also
solves the above problems concerning (2a) and (2b):
(3) [only φ] = {w ∈ [φ]|¬∃v ∈ [φ] : {ψ ∈ Alt(φ)|v ∈ [ψ]} ⊂ {ψ ∈ Alt(φ)|w ∈ [ψ]}}
In fact, as explained in van Rooij & Schulz (submitted), this rule is basically the same
as Groenendijk & Stokhof’s (1984) rule of exhaustive interpretation. To account for (2c)
we don’t need Kratzer’s world-dependent entailment relation she called ‘lumping’, but just
limit the worlds under consideration to those that satisfy the meaning postulate saying that
for every individual x if x paints something, x also paints all of its parts. For suppose that
the painted still-life in w is a complex a + b containing a (apples) as one of its parts. Then,
obviously, there is no world alternative to w in a (modal) model that satisfies the meaning
postulate where John painted still-life a + b but where he didn’t paint a. It follows that from
(2c) we cannot conclude anymore that John didn’t paint apples.
6

---

According to Rooth’s alternative semantics, the focus-semantic value of an expression
has (at least) two roles to play: (i) it helps to determine the domain over which a focus-
senstive operator like only quantifies, and (ii) it helps to determine whether an assertion is
a congruent answer to a question. Rooth (1992) argued that in the ideal case, no lexical
item should ever have to refer directly to the notion of focus, meaning that the semantic
use of focus, i.e. (i), should follow from its pragmatic use, (ii). This is not exactly what
Rooth (1992) in the end defends, but more recent work of especially Roberts (1996) and
Schwarzschild (1997) have tried to work out this appealing view and Kadmon adopts these
analyses in chapter 17 of her book.
Kadmon follows Roberts and Schwarzchild in claiming that only doesn’t have to asso-
ciate with focus (partly based on Partee’s examples involving so-called ‘second occurrence
expressions’), but in case that it does the standard interpretation results because of the prag-
matic congruence constraint on contexts induced by the focus semantic value of the whole
sentence including only. Both claims are controversial, however. First, Krifka (1995) and
others have argued that only cannot associate with so-called ‘weak pronouns’, and Beaver &
Clark (2003) argue that experimental results suggest that also in Partee’s second occurrence
examples there is prosodic prominence on the term only ‘associates’ with. Although – not
being a phonologist – I cannot reliably judge the strength of these arguments, they might
put some doubt on Kadmon’s claim that only doesn’t have to associate with focus. Equally
worrying, it seems to me, is that it is completely unclear how Schwarzchild, Roberts, and
Kadmon would like to account for sentences involving a word like even. Would they like
to propose – in analogy with the case of only – that the focus semantic value of the whole
sentence including even has to be congruent with the context? If so, I want to see how this
should be worked out. If not, it suggests that we have an operator that associates with
focus after all, and their completely pragmatic analysis of focus should be given up at least
for this particle. Although I agree (just like Rooth) that in the ideal case no operator ‘as-
sociates’ semantically with focus – who wants to say, for instance, that negation associates
with focus? –, once we assume that even does, it doesn’t seem to be so terrible anymore to
assume the same for only. Let me stress again that I would like the radical project proposed
in chapters 16 and 17 of Kadmon’s book to be successful, but I am just not sure how to
work things out, nor that all empirical phenomena speak in its favor.
Chapters 18 and 19 contain a useful discussion on whether we need a recursive analysis
of focus semantic values and on whether focus is always ‘scoped’ out of the clause in which
7

---

it occurs. In both cases Kadmon’s answer is ‘No’.
Chapter 20 discusses the interpretation of topical accent. Based on the analyses of topic
by Roberts and B¨
uring, Kadmon proposes that (4a) presupposes (among others) (4b).
(4) a. [Larry]T kissed [Nina]F .
b. For each individual, who did that individual kiss?
I think Kadmon’s analysis of topical accent is appealing: it accounts for the intuition
that the only difference between focussed and topicalized constituents is that whereas the
former is ‘the last element to be filled in’, the latter is the ‘penultimate element to be filled
in’, or ‘the last element to be filled in in the last QUD’.
Still, her analysis of topical accent is limited in the same way as the standard analyses of
focus interpretation are: it works well in case it is proper names that are accentuated, but
once more complex noun phrases are taken into account the analysis is at best incomplete.
For instance, I don’t think that [One boy]T kissed [Nina]F presupposes a question of the form
Who did one boy kiss? with a quantificational (at least) interpretation of the noun phrase.
On that interpretation, things even go surely wrong if we assume that the phrases in focus
are (by default) interpreted exhaustively (as proposed in Groenendijk & Stokhof, 1984), or
for a sentence like [One boy]T kissed only [Nina]F , where exhaustivity is explicitly demanded:
it falsely predicts that no boy kissed anyone else than Nina. But in fact, dynamic semantics
comes to the rescue here: we can interpret only not w.r.t. the quantifier one boy, but w.r.t.
the particular individual introduced by ‘one boy’. Thus, I believe that by integrating the
first and third part of the book more closely, more could have been said about phrases with
topical accent than proposed in the analyses discussed in Kadmon’s book.
In the final chapter of the book, chapter 21, Kadmon discusses an important issue that
so far hasn’t had the attention it deserved in pragmatics: how do focal and normal pre-
suppositions interact? Kadmon first argues convincingly that, although it would solve some
problems in neat ways, we better not assume that focus gives rise to an existential presup-
position, but rather that a question is under discussion. Then she takes up an old discussion
triggered by Strawson, and claims that a definite NP does not have to give rise to an existen-
tial presupposition in case it is part of the focus. The last section of the book, section 21.3,
contains a very interesting discussion on ‘altered’ presuppositions: standard presuppositions
are weakened in case the presupposition trigger itself is in focus.
8

---

In this critical review I have stressed some points where I cannot follow Kadmon’s treat-
ment: I don’t see why we should adopt standard dynamic semantics once we assume unique-
ness for anaphoric pronouns; I don’t think local accommodation fits in nicely with the sat-
isfaction theory of presupposition; and I believe that Kadmon should have chosen another
analysis of ‘only’. But these critical remarks should not give the impression that I don’t
like the book. Although I believe that the book could have been more compact, and some-
what less ‘Amherst-oriented’, it does indeed accomplish what the back cover promises: it
presents the way the semantics-pragmatics interface has come to be viewed today from the
point of view of many model-theoretic semanticists. In fact, I think that Kadmon’s book
manages very well to make the recent formal developments on the analysis of indefinites and
anaphora, presuppositions, and the analysis of focus accessible to a wide audience. More-
over, if one compares this book with the, perhaps, still standard textbook in pragmatics, i.e.
Levinson’s (1983) Pragmatics, it becomes very clear how much progress has been made on
certain topics in the field. Pragmatists better not ignore the theories reviewed and expanded
upon in Kadmon’s book, if they want to discuss any of the above mentioned topics.
References
[1] Beaver, David & Brady Clark (2003), ‘Always and only: Why not all focus sensitive
operators are alike’, Natural Language Semantics, 11, 323-362.
[2] Gazdar, Gerald (1979) Pragmatics, New York: Academic Press.
[3] Groenendijk, Jeroen & Mertin Stokhof (1984), Studies on the Semantics of Questions
and the Pragmatics of Answers, Ph.D. dissertation, University of Amsterdam.
[4] Heim, Irene (1983), ‘On the projection problem for presuppositions, In Proceedings of
WCCFL 2. pp. 114-25.
[5] Kratzer, Angelika (1989), ‘An investigation of the lumps of thought’, Linguistics and
Philosophy, 12: 607-53.
[6] Krifka, Manfred (1995), ‘Focus and/or Context: A second look at second occurrence
expressions’, In: Kamp, H. & B. Partee (eds.), Context-dependence in the Analysis of
Linguistic Meaning, Prague.
[7] Levinson, Stephen (1983), Pragmatics, Cambridge University Press, Cambridge.
9

---

[8] Roberts, Craige (1996), ‘Information structure in discourse: towards an integrated for-
mal theory of pragmatics’, In: J. Hak Yoon and A. Kathol (eds.), Ohio State University
Working Papers in Linguistics, volume 49.
[9] Rooth, Mats (1985), Association with focus, PhD dissertation, Amherst.
[10] Rooth, Mats (1992), ‘A theory of focus interpretation’, Natural Language Semantics,
1: 75-116.
[11] Rooij, Robert van & Katrin Schulz, submitted, ‘Pragmatic meaning and non-monotonic
reasoning: The case of exhaustive interpretation’.
[12] Sandt, Rob van der (1992), ‘Presupposition projection as anaphora resolution’, Journal
of Semantics, 9: 223-267.
[13] Schwarzschild, Roger (1997), ‘Why some foci must associate’, (ms), Rutgers University.
Robert van Rooij is a fellow of the Royal Netherlands Academy of Arts and Sciences
(KNAW) stationed at the Department of Philosophy of the University of Amsterdam. He
earned his PhD at the Institut f¨
ur Maschinelle Sprachverarbeitung at the University of
Stuttgart in 1997 with a thesis called ‘Attitudes and Changing Contexts’. His working area
includes the formal semantics and pragmatics of natural language, especially the analyses
of propositional attitudes, indefinites and pronouns, presupposition, focus, questions, and
conversational implicatures. He has published a number of papers on all these subjects in,
among others, the Journal of Semantics, the Journal of Pragmatics, and Linguistics and
Philosophy.
10

---