| EDITORS: Aloni, Maria; Butler, Alastair; Dekker, Paul
TITLE: Questions in Dynamic Semantics
SERIES: Current Research in the Semantics/Pragmatics Interface 17
Paul Isambert, Université de Paris 3, Sorbonne-Nouvelle, Paris, France.
This book is a collection of papers illustrating what the editors ''immodestly
call the Amsterdam style,'' which aims at accounting for semantic issues in a
totally explicit fashion, using more or less simple extensions of first-order
predicate logic. As the title indicates, such classical papers as Groenendijk
and Stokhof (1991) or Veltman (1996) are taken as starting points. Here the
issue concerns questions and related subjects like topic and focus or exhaustivity.
Paul Dekker, Maria Aloni and Alastair Butler's ''The Semantics and Pragmatics of
Questions'' is a general introduction without much formal apparatus to the
recurrent topics of the book. Their main goal is to explain the logic of
questions of Groenendijk and Stokhof (1997), upon which all contributions are
based. Questions create partitions on the context, i.e. possible worlds are
grouped together in blocks that are different answers to the question under
discussion. The differences between worlds in the same partition are immaterial
when it comes to answering. For instance, if knowing who is in Amsterdam is at
stake, and admitting that there are only two individuals John and Mary, then
there are four partitions: one containing the worlds where nobody is in
Amsterdam, one where both are, one where only John is and one where only Mary
is. In the latter partition, for instance, the difference between worlds where
Mary loves John and worlds where she doesn’t is not relevant to the question.
Note that two questions may be equivalent in a context, even if they aren’t with
respect to all possible worlds, if the context is such that their difference
concerns only worlds that have been eliminated. For instance, if we assume that
John and Mary would visit Amsterdam only if the sun shines, then the question
''Does the sun shine?'' and ''Who is in Amsterdam?'' are equivalent. But we could
also be mistaken concerning the conditions of John and Mary’s trip, and the
actual world could be outside the domain of this equivalence: this is the
pragmatics of questions, which are interpreted against some background
Part I: ''Update Semantics''
Jeroen Groenendijk's ''The Logic of Interrogation'' (originally published as
Groenendijk, 1999) designs a ''game of interrogation'', with one interrogator
asking questions and a witness answering them. The language of this game is
first order predicate logic, where questions behave as previously described.
More formally, updating a context with a question yields all the pairs of worlds
<v,w> (belonging to the context) such that they have the same answer to the
question. Thus, questions do not eliminate worlds as assertions do. Then,
Groenendijk reinterprets Grice's Maxim of Quality as consistency: the witness
should not contradict herself, that is she should not make an assertion leading
to the absurd state. Moreover, following the Maxim of Quantity, she should not
make redundant contributions (and this goes for the interrogator too), i.e.
assertions already entailed by the context. Finally the Maxim of Relation is
reinterpreted as: if an assertion P eliminates some world v, and v belongs to
the pair <v,w>, then P should also eliminate w. That is, assertion should not
eliminate worlds at random but instead entire partitions, thus resolving issues
at hand. Thus, if ''Alf rescued Bea and no one else'' is uttered after the
question ''Whom did Alf rescue?'', then ''and no one else'' means that Alf rescued
only Bea, and not that nobody besides Alf rescued Bea. For if the latter
proposition was meant, then it would answer the question ''Who rescued Bea?''. But
this question has not been raised, and there is no corresponding partition on
the context; answering it, then, would eliminate worlds belonging to different
partitions, and this would not be relevant.
In the third chapter, Balder ten Cate and Chung-chieh Shan provide an
axiomatization for the Logic of Interrogation. They also show that when
questions are not taken into account, this logic boils down to classical
first-order logic. Finally, they try out an alternative to this logic by adding
varying domains to it: each possible world has its own domain. Although this
extension is not welcome in the linguistic theory, ramifications in mathematics,
computer theory and philosophy are discussed.
Paul Dekker's ''Optimal Inquisitive Discourse'' relaxes constraints on the game of
interrogation. Questions are defined in a similar way to Groenendijk's paper,
but usual interlocutors replace the interrogator and the witness and Grice's
maxims are modified accordingly. Relevance is interpreted as the fact that the
questions of the interlocutors must be answered as the discourse proceeds
(assuming that interlocutors spell out their questions), while Quality obtains
when the discourse is supported by the interlocutors' (inner) states. Finally,
Dekker discusses the ''mention some'' problem about exhaustivity: some questions
obviously do not ask for an exhaustive answer, as in ''How can I get to the
station?''. In that case, one can assume that the hearer understands that the
question does not reflect the speaker's ''decision problem'', but is an easier
formulation of it (the hard formulation being something like a series of
questions of the kind ''Will I reach the station if I take this road?'', taking
into account that these roads are probably infinitely many).
Part II: ''Topic and Focus''
This section starts with Gerhard Jäger's ''Only Updates. On the Dynamics of the
Focus Particle 'Only''' (originally published as Jäger, 1996). It is a refinement
of Groenendijk's logic to handle the contextual restriction on the meaning of
'only', while sticking to compositionality (unlike Rooth, 1992). Assertions
yield proper updates only if they eliminate entire partitions, i.e. if they
address an issue previously raised by an interrogative. Thus ''Only Socrates is
wise'' will not mean the same thing, depending on whether ''Who is wise?'' or
''Which Athenians are wise?'' has been asked. In the latter case, one cannot
conclude that Zeus in unwise, since only Athenians' wisdom is at stake, thus
accounting for the contextual properties of 'only'.
In ''The Dynamics of Topic and Focus'', Maria Aloni, David Beaver, Brady Clark and
Robert van Rooij define a context as a pair consisting of an usual information
state and an environment, which is a sequence of information states representing
what is under discussion (i.e. topics). Instead of partitioning the current
state, questions are stacked in the environment under the form of their true
answers. Moreover, focus constituents presuppose that the corresponding question
is present in the environment. Refining Groenendijk's notion of entailment
thanks to the notion of support, the authors propose the notion of congruence,
according to which a sentence is congruent if its presupposition are met, thus
accounting for the troubles that appear when the wrong constituent is focused.
Paul Dekker's ''Nobody (Anything) Else'' introduces a semantics where the
satisfaction relation takes into account sequences of individuals, which work as
witnesses for anaphoric relations and existential quantification. The denotation
of questions (also considered as topics) is the set of those sequences that
yield true answers. On this basis, a restricted quantification is defined such
that an existential formula is true relative to some sequence of individuals.
Finally, the interpretation of ''else'' targets an individual different from all
the individuals previously mentioned, thanks to restricted quantification,
explaining how expressions like ''John and somebody else'' work when used to
answer a question.
Part III: ''Implicatures and Exhaustiveness''
In ''Exhaustivity, Questions and Plurals in Update Semantics'' (a revised version
of Zeevat, 1994), Henk Zeevat defines an operator q such that q(P) yields the
exhaustive interpretation of P. This operator is applied to form questions,
which thus denote their exhaustive answer, but which are also updated in a
special way: they create an alternative information state updated as usual with
the denotation of the question. If the answer is positive, this state becomes
the current one, otherwise it is negated with respect to the main state. In case
of a wh-question, the answer is interpreted as exhaustive (and the information
state is updated accordingly). Finally, focus is interpreted as presupposing the
related question, thus accounting for exhaustivity effect with focus.
In '''Only': Meaning and Implicatures'', Robert van Rooij and Katrin Schulz
criticize what they call the ''focus alternative approach'' (e.g. Rooth, 1992),
according to which ''Only John came'' means that there's no individual such that
he also came. This approach leads to the exclusion of sentences that may in fact
be true. Instead, the authors propose a ''background alternative approach'' such
that the same sentence means that the set denoted by the predicate is the
smallest one including John. They also consider that the fact that John came is
not presupposed but only implicated (since it can be cancelled, as in ''Only John
came, if even he did'') because the speaker does not say less than he knows
(following the Maxim of Quantity) and is maximally competent (an adaptation of
the Maxim of Quality).
In a similar vein, Benjamin Spector's ''Scalar Implicatures: Exhaustivity and
Gricean Reasoning'' proposes that exhaustivity and implicatures be derived from
the assumption that the speaker makes only optimal answers, i.e. answers that
are not entailed by any other answers with the same truth conditions and
belonging to the same alternative set. However, contrary to the classical
approach, this set is built as the set of all positive answers to the question
under discussion, and it is shown that, for instance, although an answer like
''(A or B) or C'' is entailed by ''(A and B) or C'' and thus should exclude it and
hence exclude C, this does not happen since, in the relevant alternative set
(containing no negation), there is no other answer available with a possible
exclusive reading for ''or''. Moreover, the author shows that only positive
answers are exhaustified because negative ones are not compared to all their
Part IV: ''Intonation and Syntax''
The last part of the book begins with ''Nuclear Accent, Focus, and Bidirectional
OT'' by Maria Aloni, Alastair Butler and Darrin Hindsill. The authors aim at
accounting for the relative distribution of nuclear accent and focus (in its
semantic sense) and propose the ranking of three classical constraints, namely
the ''Nuclear Stress Rule'' according to which the most embedded constituent bears
the accent, ''Destress'', which causes predictable words to be unstressed, and the
''Focus Set Rule'' which states that the focus should target a constituent
containing the nuclear accent. Classical Optimality Theory does not suffice,
however, since one must take into account both the speaker and the hearer to
resolve possible ambiguities, hence the bidirectional stance. Quite
interestingly, the ''Nuclear Stress Rule'', which is a syntactic constraint and
thus generally taken to be exceptionless, happens to be the most violable one.
In ''Counting (on) Usage Information: WH-Questions at the Syntax-Semantics
Interface'', Alastair Butler shifts the account of so-called intervention effects
in interrogatives from syntax to semantics. Taking over Dekker's ideas, the
author shows that wh-questions semantically imply wh-binders and matching usage
information. The latter should not be embedded under negation, with the extra
assumption that the least oblique wh-argument may bear as much information as
necessary, thus accounting for the difference between ''Which person did not read
which book?'' and ''*Which book didn't which person read'' (where the subject is
embedded under the negation), as well as other data from French, Korean, German
Finally, Marie Šafářová's ''Nuclear Rises in Update Semantics'' analyzes final
rise as the epistemic diamond (the symbol for ''it is possible that'' in
intensional logic), such that any sentence P with a final rise is interpreted as
''it is possible that P''. This combines with a question operator with the same
semantics as in Groenendijk's paper. Thus it is explained that rising
declaratives often behave like interrogatives because if the speaker utters ''it
is possible that P'', then her interlocutor is likely to reply ''P'' or ''not P'' if
she knows about it. Lack of commitment and other so-called affective meanings
are similarly accounted for. However, with an adequate formalization of Grice's
maxims, the author also explains why such rising declaratives may be understood
as simple ones.
This book is definitely not a handbook where each chapter would address a
well-delimited issue. Rather, these are working papers with overlapping subjects
and alternative solutions. Thus the reader might feel at the end of each
contribution that there's far more to the issue than what is addressed. But the
reader might also figure out that the next chapter deals with some remaining
problems. No definitive solution stems from the whole book, but one nonetheless
ends with a nice picture of the ''Amsterdam style'', and with an interesting view
on the problems at hand too.
The title might be misleading. Indeed, barring Groenendijk's opening paper,
whose reprint is welcome, most contributions take the semantics of questions as
a background to address related subjects, like topic, focus or exhaustivity.
This is quite an interesting move that shows how one issue leads to other ones;
accounting for their relatedness is more far-reaching than simply sticking to
questions over and over, all the more as those related issues often shed some
new light on the original problem. For instance, the pervasiveness of Grice's
Maxims leads to several tentative formalizations, thus avoiding any artificial
separation between some core semantics and a peripheral pragmatics. What does it
mean for a speaker to be relevant, for instance, is a problem that should not be
explained away by (and confused with) a mere stipulative statement like ''the
speaker should be relevant.'' And indeed it is of higher importance when it comes
to questions, since for instance it must enable one to distinguish between a
wrong answer and an irrelevant one. More generally, trying to guess what the
speaker should think or do is not welcome from a linguistic point of view, and
one of the merits of formal semantics as illustrated by this book is to avoid
such psychological assertions in favor of a strict linguistic model.
As the editors put it in their preface, the Amsterdam style is both formal and
fully explicit. This means that the reader won't be spared any technical detail,
and at the same time will be able to check each and every step of the
development. Needless to say, anyone uncomfortable with logic or unwilling to
struggle with a definition should avoid this book. Those interested in
linguistic insights more than in machinery at work should also step back. But
there is no need to be particularly fond of mathematical formalization to find
some interest in this book; although there won't be any new data to chew on, the
interest lies in the connections the authors are able to draw between various
subjects. For instance, it is striking how investigating questions and
exhaustivity offers new insights on the issue of discourse coherence (although
hardly mentioned as such). One should not forget that the matter here is dynamic
semantics, not simply formal semantics. This means that online speech is at the
heart of it (not just representing content) and that such complex objects as
discourse or dialogue are investigated more deeply than might seem at first
sight. Anyone interested in those topics should have a look at this book, if not
give it a thorough reading.
Groenendijk, Jeroen. 1999. ''The Logic of Interrogation. Classical Version''. In
T. Matthews and D. L. Strolovitch, eds., _The Proceedings of the Ninth
Conference on Semantics and Linguistic Theory_. CLC Publications.
Groenendijk, Jeroen and Martin Stokhof. 1991. ''Dynamic Predicate Logic''.
_Linguistics and Philosophy_, 14(1).
Groenendijk, Jeroen and Martin Stokhof. 1997. ''Questions''. In J. van Benthem and
A. ter Meulen, eds., _Handbook of Logic and Language_. Elsevier Science Publishers.
Jäger, Gerhard. 1996. ''Only Updates. On the Dynamics of the Focus Particle
'Only'''. In M. Stokhof and P. Dekker, eds., _Proceedings of the Tenth Amsterdam
Colloquium_. University of Amsterdam.
Rooth, Mats. 1992. ''A Theory of Focus Interpretation''. _Natural Language
Veltman, Frank. 1996. ''Defaults in Update Semantics''. _Journal of Philosophical
Zeevat, Henk. 1994. ''Questions and Exhaustivity in Update Semantics''. In H.
Bunt, R. Muskens and G. Rentier, eds., _Proceedings of the International
Workshop on Computational Semantics_. University of Tilburg.
ABOUT THE REVIEWER
Paul Isambert is a PhD student at the University of Paris 3, France. He's
currently working on grammaticalization and discourse structure, especially
concerning topic shifts and anaphora.