“01wisniewski”

i i 2003/6/9
page 5
i i

Logique & Analyse 173–174–175 (2001), 5–43

QUESTIONS AND INFERENCES

ANDRZEJ WIŚNIEWSKI

Abstract
Central concepts of inferential erotetic logic are discussed. Search
scenarios and search rules are defined in terms of the logic of ques-
tions.

1. Introduction

The aim of this paper is to present the basic ideas of a certain approach to the
logic of questions. We call it here inferential erotetic logic (IEL for short).
IEL focusses the attention of the logic of questions on erotetic inferences
and thus forms an alternative to the received view, that is, roughly, to the
approach which puts the structure of questions and the problem of question-
answer relationship at the center of attention of erotetic logic. 1 At the same
time IEL presents an alternative to the “Interrogative Model of Inquiry” de-
veloped by Hintikka and his associates.2
We shall concentrate here on basic concepts, underlying intuitions and ex-
amples, leaving aside proofs and more technical results (these can be found
in the papers and books listed in the “Bibliography”). The last sections con-
tain some new developments.

2. Erotetic Inferences and Validity

IEL starts with a trivial observation that before a question is asked or posed,
a questioner must arrive at it. In many cases arriving at questions resembles
coming to conclusions: there are premises involved and some inferential
thought processes take place. If we admit that a conclusion need not be

1
The logic of questions is also called erotetic logic (from Greek “erotema” = question).
2
Hintikka’s book Inquiry as Inquiry: A Logic of Scientific Discovery (Kluwer, Dordrecht
/Boston /London 1999) includes most of his papers devoted to the subject.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 6
i i

6 ANDRZEJ WIŚNIEWSKI

“conclusive”, we can say that sometimes questions play the role of conclu-
sions. But questions can also perform the role of premises: it often happens
that we arrive at a question when we are looking for an answer to another
question. Thus the concept of an erotetic inference can be introduced. As a
first approximation an erotetic inference may be defined as a thought process
in which we arrive at a question on the basis of some previously accepted
declarative sentence or sentences and/or a previously posed question. There
are, then, erotetic inferences of (at least) two kinds: the key difference be-
tween them lies in the type of premises involved. In the case of erotetic infer-
ences of the first kind the premises are declarative sentences. The premises
of an erotetic inference of the second kind consist of a question and possibly
some declarative sentence(s); erotetic inferences which do not involve any
declarative premises can be regarded as a special case of erotetic inferences
of the second kind. From the syntactical point of view, an erotetic inference
of the first kind can be identified with an ordered pair <X, Q>, where X
is a finite and non-empty set of declarative sentences (the premises) and Q
is a question (the conclusion). Similarly, an erotetic inference of the second
kind can be identified with an ordered triple <Q, X, Q1 >, where Q, Q1 are
questions and X is a finite (possibly empty) set of declarative sentences; the
question Q is the interrogative premise (we shall call it an initial question),
the elements of X are declarative premises and Q1 is the conclusion. Erotetic
inferences construed syntactically are also called erotetic arguments.
IEL proposes some conditions of validity for erotetic inferences.
As long as we are concerned with “standard” inferences, validity amounts
to the transmission of truth: if the premises are all true, the conclusion must
be true as well. But it is doubtful whether it makes any sense to assign
truth or falsity to questions. On the other hand, a question usually has many
possible answers, and each of them may be true or false. Moreover, there
are questions which have well-defined sets of direct (i.e. possible and just-
sufficient) answers. Assume that Q is a question which has a well-defined
set of direct answers. We say that Q is sound if and only if at least one direct
answer to Q is true, and unsound otherwise. It seems natural to impose the
following necessary condition of validity on erotetic inferences of the first
kind:
(C1 ) (transmission of truth into soundness) if the premises are all true,
then the question which is the conclusion must be sound.

Is this sufficient? Certainly not. For if (C1 ) were sufficient, the following
inferences would be valid:

i i

i i
 “01wisniewski”
i i 2003/6/9
page 7
i i

QUESTIONS AND INFERENCES 7

(I) Andrew is rich.

Andrew is happy.
Is Andrew happy?

(II) If Andrew is rich, then he is happy.

Andrew is rich.
Is Andrew happy?

What is wrong with the above inferences? The question which is the con-
clusion has a (direct) answer which provides us with information which is
already present (directly or indirectly) in the premises. In other words, the
question which is the conclusion is logically redundant and thus not informa-
tive. So IEL imposes the following additional necessary condition of validity
on erotetic inferences of the first kind:
(C2 ) (informativeness) a question which is the conclusion must be infor-
mative with respect to the premises.

Informativeness is then explicated as the lack of entailment of any direct

answer from the premises; the applied concept of entailment need not be
classical (see below).
Let us now turn to erotetic inferences of the second kind. The natural gen-
eralization of the standard condition of validity is:
(C3 ) (transmission of soundness/truth into soundness) if the initial ques-
tion is sound and all the declarative premises are true, then the question
which is the conclusion must be sound.

Again, (C3 ) is only a necessary condition. If (C3 ) were sufficient, the fol-
lowing would be valid inferences:
(III) Is Andrew a logician?
Is Andrew married?

(IV) Where did Andrew leave for: Chicago or New York?

Did Andrew leave for the US?

The problem here is that the questions which are conclusions have (direct)
answers which are cognitively useless: these answers, if accepted, do not
contribute in any way to the process of finding answers to initial questions.
IEL imposes the following additional condition of validity on erotetic in-
ferences of the second kind:

i i

i i
 “01wisniewski”
i i 2003/6/9
page 8
i i

8 ANDRZEJ WIŚNIEWSKI

(C4 ) (open-minded cognitive usefulness) each direct answer to the ques-

tion which is the conclusion is potentially useful, on the basis of the
declarative premises, for finding an answer to the initial question.

Condition (C4 ) is then clarified by requiring that each direct answer to

the question which is the conclusion should, together with the declarative
premises, narrow down the class of possibilities offered by the initial ques-
tion. In the majority of cases this amounts to entailment of a direct or partial
answer to the initial question (see below). Note that we do not require that
an answer to the question which is the conclusion should, together with the
declarative premises, yield a single answer to the initial question: we only
require that the class of possibilities should be narrowed down. Yet, when
the class is narrowed down to a singleton class, a single answer to the initial
question is forthcoming.
Conditions (C1 )–(C4 ) are expressed in rather vague terms. Of course,
IEL does not stop here. The following semantical concepts are introduced:
(i) evocation of questions by sets of declarative sentences, and (ii) (erotetic)
implication of questions by questions and possibly declarative sentences. Va-
lidity of erotetic inferences of the first kind is then defined in terms of evo-
cation, whereas validity of erotetic inferences of the second kind is defined
by means of erotetic implication. The proposed definitions of evocation and
erotetic implication are explications of the relevant notions of the arising of
questions (cf. Wiśniewski 1995a, Chapter 1). By defining the semantical
concept “a question Q is evoked by a set of declarative sentences X” we ex-
plicate the concept “a question Q arises from a set of declarative sentences
X”, whereas the definition of “a question Q1 is implied by a question Q
on the basis of a set of declarative sentences X” gives us an explication of
the notion “a question Q1 arises from a question Q and a set of declarative
sentences X.” Thus, although conditions (C1 ), (C2 ) on the one hand, and
conditions (C3 ), (C4 ) on the other seem diverse at first sight, the analysis of
validity of erotetic inferences proposed by IEL is based on a certain general
idea.

3. Questions and Answers

IEL presupposes a certain analysis of questions and answers.

The approaches to questions proposed by various logicians and formal
linguists can be divided into reductionist and non-reductionist positions.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 9
i i

QUESTIONS AND INFERENCES 9

Within the reductionist approach, in turn, radical and moderate standpoints

can be distinguished.3
According to the radical view, questions are not linguistic entities. Ques-
tions are identified with: sets of sufficient answers, or sets of possible an-
swers, or sets of true answers, or functions defined on possible worlds, or
functions from categorial answers to propositions, or speech acts of a spe-
cial kind, etc.
The moderate reductionist view considers questions as linguistic entities
which, however, can be reduced to expressions of some other categories. It
is claimed that each question can be adequately paraphrased as an expression
belonging to some other syntactic category and then formalized within a cer-
tain logic which, although not primarily designed as the logic of questions,
can thus be regarded as providing us with the foundations of erotetic logic.
Sometimes questions are identified with declarative sentences of a strictly
defined form (Harrah). More often questions are identified with imperatives
of a special kind. But the most popular approach is the imperative-epistemic
approach, with Lennart Åqvist and Jaakko Hintikka as its most eminent rep-
resentatives.4
According to the non-reductionist approach, questions are specific expres-
sions of a strictly defined form; they are not reducible to expressions of
other syntactic categories. The most widespread proposal here is to regard
a question as an expression which consists of an interrogative operator and
a sentential function (declarative formula with free variable(s)). This idea
is most widely and thoroughly elaborated on in the works of Tadeusz Ku-
biński.5 But Nuel D. Belnap’s theory of questions and answers is the most
well known.6 Belnap distinguishes between: (a) natural language questions,
3
For a general overview, see Wiśniewski 1995a, Chapter 2. See also the survey paper
of David Harrah ‘The logic of questions’, included in the second edition of Handbook of
Philosophical Logic (Kluwer, Dordrecht/Boston/London 2002, Volume 8, pp. 1–60), or (for
more linguistically oriented approaches) the paper of Jeroen Groenendijk and Martin Stokhof
‘Questions’, published in Handbook of Logic and Language (Elsevier, Amsterdam 1996,
pp. 1055–1125).
4
Cf. Lennart Åqvist A New Approach to the Logical Theory of Interrogatives (Almqvist
& Wiksell, Uppsala 1965); Jaakko Hintikka The Semantics of Questions and the Questions
of Semantics (North-Holland, Amsterdam 1976).
5
Cf. e.g. Tadeusz Kubiński An Outline of the Logical Theory of Questions (Akademie-
Verlag, Berlin 1980) or a survey paper: A. Wiśniewski ‘Kubiński’s theory of questions’, in:
A. Wiśniewski and J. Zygmunt (eds.), Erotetic Logic, Deontic Logic and Other Logical Mat-
ters. Essays in Memory of Tadeusz Kubiński (Wydawnictwo Uniwersytetu Wrocławskiego,
Wrocław 1997, pp. 9–50).
6
Cf. e.g. Nuel D. Belnap and Thomas B. Steel The Logic of Questions and Answers
(Yale University Press, New Haven 1976).

i i

i i
 “01wisniewski”
i i 2003/6/9
page 10
i i

10 ANDRZEJ WIŚNIEWSKI

(b) interrogatives and (c) questions understood as abstract (set-theoretical)

entities. Interrogatives are expressions of some formalized languages. They
are not only formal counterparts of natural language questions, but they also
express questions understood as abstract entities. The basic idea of Belnap’s
approach is that an interrogative “presents” a set of alternatives together with
some suggestions or indications as to what kind of choice or selection among
them should be made. A simple interrogative consists of the question mark
?, the lexical subject and the lexical request. The function of a lexical subject
is to offer the relevant (nominal) alternatives, whereas the role of a lexical
request is to characterize the required kind of selection. A direct answer to
an interrogative is then, roughly, a declarative formula made up of the rel-
evant alternative(s) which “satisfies” the conditions imposed by the lexical
request.
IEL accepts the non-reductionist approach to questions, but does not fore-
judge its particular form. What we need is a formalized language which has
both declarative formulas and questions (the distinction between interroga-
tives and questions is disregarded) as meaningful expressions, where ques-
tions are not declarative formulas. Languages of this kind can be defined in
various ways, and questions can be introduced, e.g. according to Kubi ński’s
approach or Belnap’s approach (to mention only the richest sources of ideas).
Yet, in the majority of published papers on IEL a specific attitude has been
adopted. But before this is presented, let us say a few words about answers
to questions.
Most logical theories of questions pay at least as much attention to answers
to questions as to the questions themselves. It is usually assumed that a ques-
tion may have many answers and thus the phrase “answer to a question” is
not tantamount to “the true answer to a question.” In other words, the ana-
lyzed answers are usually possible answers; a possible answer may be true or
false (or even have no logical value at all). But it is not the case that all pos-
sible answers are equally interesting to erotetic logicians. The standard way
of proceeding is to define some basic category of possible answers. They are
called direct answers (Åqvist, Belnap, Harrah, Kubiński), sufficient answers
(Stahl), conclusive answers (Hintikka), etc. Those “principal” possible an-
swers (let us use this general term here) are supposed to satisfy some general
conditions, usually expressed in pragmatic (in the traditional sense of the
word) terms. For example, direct answers in Belnap’s sense are the answers
which “are directly and precisely responsive to the question, giving neither
more nor less information than what is called for.”7 In the light of Hin-
tikka’s theory a reply is called a conclusive answer if it completely satisfies

7
N.D. Belnap ‘Åqvist’s corrections-accumulating question sequences’, in: J.W. Davis et
al. (eds.), Philosophical Logic (D. Reidel, Dordrecht 1969), p. 124.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 11
i i

QUESTIONS AND INFERENCES 11

the epistemic request of the questioner. Although the above conditions are
formulated in pragmatic terms, logical theories of questions usually define
the principal possible answers to questions or interrogatives of formalized
languages in terms of syntax and/or semantics. For example, Kubi ński as
well as Belnap (in the case of elementary interrogatives) proceed syntacti-
cally, whereas Hintikka’s approach is twofold: possible replies and so-called
conclusiveness-conditions are characterized in syntactic terms, and a possi-
ble reply is a conclusive answer if this reply together with the description of
the questioner’s state of knowledge entails (in the sense of some underlying
epistemic logic) the desideratum of the question, that is, an epistemic for-
mula which describes the cognitive state of affairs the questioner wants to be
brought about.
IEL proceeds syntactically: it is assumed that the syntax of a considered
formalized language assigns to a question a set of sentences, which are then
called directs answers to the question. As long as we are pursuing general
considerations, the details of this assignment need not be decided on; we
only assume that the following conditions are fulfilled:
(C5 ) direct answers are sentences, i.e. declarative formulas with no in-
dividual or higher-order free variables (since sentential functions are not
definite enough in order to answer anything);

(C6 ) each question has at least two direct answers (since a necessary con-
dition of being a question is to present at least two “alternatives” or con-
ceptual possibilities among which some selection can be made);

(C7 ) each finite and at least two-element set of sentences is the set of
direct answers to some question.

Direct answers to a question of a formalized language are then viewed

as the possible and just-sufficient answers. Yet, the following have to be
carefully distinguished: (a) direct answers to a question of a formalized lan-
guage, and (b) possible and just-sufficient (i.e. direct) answers to a natural-
language question. As many philosophers and linguists have pointed out,
there are cases in which it is strongly context-dependent what sentences may
be counted as the possible and just-sufficient answers to a natural-language
question. Moreover, there are natural-language questions for which it makes
no sense at all to speak about sets of possible and just-sufficient answers;
why-questions and how-questions are often recalled in this context. Yet, we
can still claim that questions of a formalized language in which the question-
answer relationship is defined syntactically represent questions of a natural
language. The relation of representation we have in mind can be character-
ized as follows: a question Q of a formalized language represents a question

i i

i i
 “01wisniewski”
i i 2003/6/9
page 12
i i

12 ANDRZEJ WIŚNIEWSKI

Q∗ of a natural language construed in such a way that the possible and just-
sufficient answers to Q∗ have the logical form of direct answers to Q. If a
natural-language question has many readings, it has many representations.
The richer the underlying formalized language, the more we can represent
within it.
Since there is no (and in many cases cannot be) one-to-one correspon-
dence between natural-language questions and questions of a formalized lan-
guage, we can go further and define questions of a formalized language as
expressions whose form does not resemble their natural-language counter-
parts. One possibility is to define a question (of an object-level formalized
language) as an expression which consists of the question mark ? and an
expression of the object-language which is equiform to the expression of
the metalanguage which designates the set of direct answers to the question.
For example, when we add the question mark ? and the brackets { } to the
vocabulary of a first-order language (or a propositional language), we can
supplement the language with questions of the form:

(#) ? {A1 , ..., An }

where n > 1 and A1 , ..., An are syntactically distinct sentences (i.e. closed
well-formed formulas); these sentences are direct answers to the question.
A question of the form (#) represents those natural-language questions (or
their clarified counterparts) whose sets of possible and just-sufficient an-
swers consist of sentences which have the logical form of A 1 , ..., An , ex-
clusively. Note that an expression of the form (#) still belongs to the object-
language, whereas the expression {A1 , ..., An } is equiform to a metalan-
guage-expression which designates the set of direct answers to the question.
One advantage of this solution is that it is now extremely easy to say what
counts as a direct answer to the question, and what natural-language ques-
tions are represented by it. This procedure can also be applied to (possible
representations of) some wh-questions (cf. Wiśniewski 1995a, Chapter 3).
When we go in this direction, we arrive at a level of abstraction at which
the so-called conclusiveness conditions in Hintikka’s sense are no longer
needed. Yet, let us stress that the above approach to questions is not essen-
tial to IEL: if only conditions (C5 ), (C6 ) and (C7 ) specified in the previous
section are fulfilled, questions can be introduced in the way proposed by Bel-
nap, or by Kubiński, or by other logicians who adopt the non-reductionist
standpoint.

In what follows we assume the von Neumann-Gödel-Bernays version of

set theory. We shall use the standard set-theoretical terminology and nota-
tion. The expression “iff” abbreviates “if and only if.”

i i

i i
 “01wisniewski”
i i 2003/6/9
page 13
i i

QUESTIONS AND INFERENCES 13

4. Minimal erotetic semantics

Let L be a formalized language of the required kind. Thus well-formed

formulas (wffs for short) of L split into two disjoint categories: declarative
well-formed formulas (d-wffs), defined in a standard way, and erotetic well-
formed formulas (e-wffs), that is, questions of L defined according to some
pattern, but in such a way that the conditions (C5 ), (C6 ) and (C7 ) of the pre-
vious section are fulfilled. In order to go further we now have to supplement
L with a semantics. But, surprisingly enough, in order to pursue general in-
vestigations within IEL, the semantics need not be very specific with respect
to the erotetic part of L.
As a matter of fact, it suffices to suppose that the declarative part of L
is supplemented with a semantics which is rich enough to define a certain
concept of truth for d-wffs of L. Depending on the particular form of the
semantics, truth can be defined in terms of models, or algebraic structures,
or Boolean valuations, or games, etc. Let DL stand for the set of all d-wffs
of L, and EL for the set of all e-wffs of L. Of course, DL ∩ EL = ∅. By a
partition of L we mean a partition of DL , that is, an ordered pair <T, U >,
where T ∩ U = ∅ and T ∪ U = DL . Intuitively, T consists of all the wffs
which are true, and U is made up of all the d-wffs which are untrue. Then
it is assumed that in the class of all partitions a subclass of admissible parti-
tions is distinguished. In the general case it is only required that admissible
partitions should comply with the underlying semantics and that the inter-
section of the first elements of all admissible partitions forms a non-empty
set (elements of this intersection may be thought of as valid formulas). Yet,
for the purposes of this paper we will also assume that admissible partitions
are directly determined by the accepted semantics of the declarative part of
L. The following examples should clarify this:

Example 1:

Assume that L results from the language of Classical Propositional Calculus

(CPC) by enriching it with questions. Let DL be the set of all CPC-formulas.
Assume that the semantics for DL is based on the notion of a Boolean valu-
ation. The class of admissible partitions of L is determined by Boolean val-
uations of DL : a partition <T, U > of L is admissible iff there is a Boolean
valuation v such that T = {A ∈ DL : v(A) = 1}.

Example 2:

Assume that L is a first-order language enriched with questions. Assume

that the concept of truth of a d-wff A of L in an interpretation M =<M, f >
of the declarative part of L (in symbols: M  A) is defined in the standard

i i

i i
 “01wisniewski”
i i 2003/6/9
page 14
i i

14 ANDRZEJ WIŚNIEWSKI

way. A partition <T, U > of L is admissible iff there is an interpretation M

of the declarative part of L such that T = {A ∈ DL : M  A}.

Example 3:

Assume again that L is a first-order language enriched with questions, and

that the declarative part of L is supplemented with a standard model-theoreti-
cal semantics. Assume also that in the class of all interpretations of the
declarative part of L a proper subclass of normal interpretations was dis-
tinguished (for instance, one can stipulate that an interpretation <M, f > is
normal iff each element of M is the value of some closed term of L; of
course, there are also other options). A partition <T, U > of L is said to be
admissible iff there is a normal interpretation M of the declarative part of L
such that T = {A ∈ DL : M  A}.

Let dQ denote the set of direct answers to a question Q. The concept of

soundness of a question in a partition is defined as follows:

Definition 1: A question Q is sound in a partition <T, U > iff dQ ∩ T 6= ∅.

When the semantics of the declarative part of L becomes more specific,

soundness in a partition can be replaced by soundness in an interpretation,
or soundness with respect to a Boolean valuation, etc. The underlying idea
is always the same: a sound question is a question which has at least one true
direct answer.

We need two concepts of entailment: the standard concept of single-conclu-

sion entailment and the concept of multiple-conclusion entailment. 8 The
latter is a semantical relation between sets of d-wffs.

Definition 2: A set of d-wffs X entails a d-wff A (in symbols: X  A) iff for

each admissible partition <T, U >:
(*) if X ⊂ T, then A ∈ T.

Definition 3: A set of d-wffs X multiple-conclusion entails a set of d-wffs Y

(symbolically: X ||= Y ) iff for each admissible partition <T, U >:
(**) if X ⊂ T, then Y ∩ T 6= ∅.

8
Cf. D.J. Shoesmith and T.J. Smiley Multiple-conclusion logic (Cambridge University
Press, Cambridge 1978).

i i

i i
 “01wisniewski”
i i 2003/6/9
page 15
i i

QUESTIONS AND INFERENCES 15

Note that both entailment and multiple-conclusion entailment (mc-entailment

for short) are defined by means of admissible partitions, and these, in turn,
have sets of all true d-wffs (true in the sense of underlying semantics) as the
first elements. The idea which lies behind the concept of multiple-conclusion
entailment is: if all the d-wffs in X are true, then at least one d-wff in Y must
be true. Or, to put it differently, the truth of all the X-es guarantees the exis-
tence of at least one truth in Y .
If L fulfills the conditions of Example 2, we get:
(i) X  A iff the following condition holds:
(*0 ) for each interpretation M of the declarative part of L: if all the
d-wffs in X are true in M, then A is true in M.

(ii) X ||= Y iff the following holds:

(**0 ) for each interpretation M of the declarative part of L: if all
the d-wffs in X are true in M, then at least one d-wff in Y is true
in M.

When admissible partitions are defined in terms of Boolean valuations or

normal interpretations (cf. examples 1 and 3 above), the relevant conditions
refer to Boolean valuations and normal interpretations, respectively. This is
as it should be.
Note that entailment can be defined in terms of mc-entailment. But the
converse does not hold: it can happen that X mc-entails Y although no
element of Y is (single-conclusion) entailed by X.
The concept of multiple-conclusion entailment has proved its usefulness
in the logic of questions in many ways. In particular, both evocation and
erotetic implication can be defined in terms of mc-entailment (cf. sections 4
and 6).
For conciseness, we write A ||= Y instead of {A} ||= Y , and X ||= A
instead of X ||= {A}.
Following Belnap, we shall now introduce the concept of a presupposition
of a question.

Definition 4: A d-wff A is a presupposition of a question Q iff A is entailed

by each direct answer to Q.

The set of presuppositions of a question Q will be referred to as PresQ.

Some questions, but not all, have prospective presuppositions.

Definition 5: A d-wff A is a prospective presupposition of a question Q iff

(i) A is a presupposition of Q, and (ii) A ||= dQ.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 16
i i

16 ANDRZEJ WIŚNIEWSKI

In general, it is possible that all the presuppositions of a question are true,

but the question is not sound, that is, has no true direct answer. But if a
question has a prospective presupposition, then the question is sound given
that a prospective presupposition of it is true.
By means of the concepts introduced above further concepts of erotetic
semantics can be defined (cf. Wiśniewski 1995a, 1996, 1997). This, how-
ever, would go beyond the scope of the present paper. We only introduce the
following:

Definition 6: A question Q is:

(a) normal iff PresQ ||= dQ;
(b) regular iff for some A ∈ PresQ: A ||= dQ.

For intuitions, see Wiśniewski (1995a, 1997). Let us stress that the fact
that a question has any of the above characteristics is dependent upon both
the form of the question and the properties of an underlying semantics. In
particular, it may happen that a given question is normal in the light of one
semantics, but not normal from the standpoint of another. Note also that
since mc-entailment need not be compact, normality is not tantamount to
regularity.
So far we have introduced only one category of answers. Yet, the concep-
tual apparatus characterized above allows us to define further categories. For
example, we can say that a d-wff A is a just-complete answer to a question
Q if and only if there is a direct answer B to Q such that B entails A and A
entails B. Roughly, just-complete answers are equivalent to direct answers;
this concept is not superfluous since direct answers are defined syntactically.
Partial answers are defined by:

Definition 7: A sentence A is a partial answer to a question Q iff A is not a

just-complete answer to Q and for some non-empty proper subset Y of the
set of direct answers to Q we have:
(i) A ||= Y , and
(ii) B  A for each B ∈ Y .

Thus a partial answer to a question is a sentence which is not equivalent to

any direct answer to the question, but which is true if and only if a true direct
answer belongs to some specified proper subset of the set of all the direct
answers to the question.9
Since we have assumed that each question has at least two direct answers,
Definition 7 yields that a question which has exactly two direct answers has
9
For a discussion, see Wiśniewski (1995a, pp. 114–115).

i i

i i
 “01wisniewski”
i i 2003/6/9
page 17
i i

QUESTIONS AND INFERENCES 17

no partial answers. This is not to say, however, that binary questions have
only direct and just-complete answers; these questions, as well as other ques-
tions, have also incomplete answers, corrective answers, etc.

5. Definition of Evocation

Looked at a formal point of view, evocation of questions is a rather simple

notion. As we have said, it can happen that no d-wff in a set of d-wffs Y is
entailed by a set of d-wffs X, but still X mc-entails Y . Now think of Y as of
the set of direct answers to a question. Direct answers are the “just-sufficient
possibilities” offered by a question. It may happen that none of them holds.
Yet, suppose that X mc-entails the set of direct answers to a question Q,
but does not entail any particular direct answer to Q. Thus if X consists
of truths, we have a guarantee that a true direct answer to Q exists. On the
other hand, since no direct answer to Q is entailed by X, the problem as to
which direct answer to Q is true cannot be resolved in a logically legitimate
way by means of X only. In other words, Q is sound relative to X and
informative with respect to X. In such a situation we say that X evokes Q.
More formally, we introduce:

Definition 8: A set of d-wffs X evokes a question Q (in symbols: E(X, Q))

iff
(i) X ||= dQ, and
(ii) for each A ∈ dQ: Xnon||= A.10

Yet, it is neither assumed nor denied that an evoking set consists of truths.
Let P =<T, U > be an arbitrary but fixed admissible partition of (the declar-
ative part of) the language. Assume that E(X, Q). The following tables
describe possible connections:

X Q
X⊂T sound in P
Xnon⊂ T sound in P or unsound in P

Table 1.

10
Of course, clause (ii) is equivalent to:

(ii0 ) for each A ∈ dQ: Xnon A.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 18
i i

18 ANDRZEJ WIŚNIEWSKI

Q X
sound in P X ⊂ T or Xnon⊂ T
unsound in P Xnon⊂ T

Table 2.

For properties of evocation, see Wiśniewski (1991a, 1995a, 1996).11 Let

us only mention the following:

Fact 1: If X evokes Q, then X entails each presupposition of Q.

Fact 2: Assume that Q is normal. Then X evokes Q iff X entails each

presupposition of Q, but does not entail any direct answer to Q.

Fact 3: Assume that Q is regular. Then X evokes Q iff X entails some

prospective presupposition of Q, but does not entail any direct answer to Q.

We say that an erotetic inference of the first kind <X, Q> is valid iff
E(X, Q).

Generation of questions is a special case of evocation: we say that a set

of d-wffs X generates a question Q iff X evokes Q and ∅ non||= dQ. The
underlying intuition is: a generated question is made sound by the gener-
ating set12 and is informative with respect to this set. For generation see
Wiśniewski (1989, 1991a, 1995a).13

11
In Wiśniewski (1991a) evocation is called “weak generation”.
Chapter 5 of Wiśniewski (1995a) is devoted to properties of evocation in first-order lan-
guages enriched with questions and supplemented with a standard model-theoretic seman-
tics. A more general approach (in terms of the minimal erotetic semantics) is adopted in
Wiśniewski (1996). Meheus (cf. Meheus 1999) proposes some generalizations which also
cover the case of evocation from inconsistent sets of premises.
12
If ∅ ||= dQ, then Q is sound in each admissible partition of the language and thus clause
(ii) of Definition 8 is fulfilled for trivial reasons. Generation excludes this.
13
See also Meheus (1999) for the inconsistent case.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 19
i i

QUESTIONS AND INFERENCES 19

6. Examples of Evocation

In order to present examples let us consider two languages of the kind ana-
lyzed above.

6.1. Language L1

Language L1 results from the language of Classical Propositional Calculus

(CPC) by enriching it with questions of the form (#). To be more precise,
the declarative part of L1 is the language of CPC; we shall use the letters
p, q, r, s, t, u, p1 , .... for propositional variables and the symbols ¬, ⊃, ∧,
∨, ≡ for negation, implication, conjunction, disjunction and biconditional,
respectively. By d-wffs of L1 we shall mean the CPC-formulas. The vo-
cabulary of the erotetic part of L1 consists of the following signs: ?, {, }.
Questions of L1 are of the form:

(#) ? {A1 , ..., An }

where n > 1 and A1 , ..., An are syntactically distinct d-wffs of L1 ; these

d-wffs are called direct answers to the question.
A question of the form (#) can be read: “Is it the case that A1 , ... , or is
it the case that An ?”. Yet, sometimes other readings can be recommended.
For example, questions of the form (called simple yes-no questions):

(6.1) ? {A, ¬A}

can be read “Is it the case that A?” and abbreviated as:

(6.2) ?A

Questions of the form:

(6.3) ? {A ∧ B, A ∧ ¬B, ¬A ∧ B, ¬A ∧ ¬B}

are called binary conjunctive questions and abbreviated as:

(6.4) ? ± | A, B |

The semantics for the declarative part of L1 is based on the concept of

Boolean valuation; the concepts of Truth and Falsehood of a d-wff under a
Boolean valuation are understood in the standard way. Boolean valuations
determine admissible partitions of L1 (cf. Example 1 in Section 4); the re-
maining semantical concepts are defined according to the pattern presented

i i

i i
 “01wisniewski”
i i 2003/6/9
page 20
i i

20 ANDRZEJ WIŚNIEWSKI

in Section 4.

The following present simple examples of evocation in L1 ; for trans-

parency, we write A1 , ..., An `E Q instead of E({A1 , ..., An }, Q):

(6.5) p ∨ ¬p `E ? p,

(6.6) p ∨ q `E ? p,

(6.7) p ∨ q `E ? q,

(6.8) p ⊃ q `E ? p,

(6.9) p ⊃ q `E ? q,

(6.10) p ≡ q `E ? {p ∧ q, ¬p ∧ ¬q},

(6.11) p ∨ q `E ? {p, q},

(6.12) p ∨ q `E ? {p ∧ q, p ∧ ¬q, ¬p ∧ q},

(6.13) p ⊃ q ∨ r, p `E ? {q, r},

(6.14) p ∧ q ⊃ r, ¬r `E ? {¬p, ¬q},

(6.15) (p ∨ q) ∨ r `E ? {p, q ∨ r}.

6.2. Language L2

Language L2 results from the language of Monadic Predicate Calculus with

identity (MPC= ) by enriching it with questions of the form (#) as well as
with existential and general wh-questions. We assume that the vocabulary of
the declarative part of L2 contains individual constants, but does not contain
function symbols. The letters P , R, T will be used as metalinguistic vari-
ables for predicates, and the letters c, c1 , ... will be metalinguistic variables
for individual constants. By d-wffs of L2 we shall mean the well-formed
formulas of the language of MPC= , defined in the standard manner. A sen-
tence is a d-wff with no free variable(s); freedom and bondage of variables
are defined as usual. By terms of L2 we shall mean the individual variables
and individual constants. An expression of the form Ax refers to d-wffs of
L2 which have x as the only free variable. The result of the substitution of

i i

i i
 “01wisniewski”
i i 2003/6/9
page 21
i i

QUESTIONS AND INFERENCES 21

an individual constant c for a variable x in a d-wff Ax will be designated by

A(x/c); needless to say, A(x/c) is always a sentence.

The vocabulary of the erotetic part of L2 consists of the signs: ?, {, }, S,

U. First, L2 contains questions of the form:

(#) ? {A1 , ..., An }

where n > 1 and A1 , ..., An are nonequiform (i.e. syntactically distinct)

sentences of L2 . Direct answers to questions of the form (#) are defined as
above; we accept here analogous notational conventions as in the case of L 1 .
Second, L2 contains questions falling under the schemata:

(##) ? S(Ax),

(###) ? U(Ax),

where x stands for an individual variable and Ax is a d-wff of L 2 which has

x as the only free variable. A direct answer to a question of the form (##)
is a sentence of the form A(x/c), where c is an individual constant. Direct
answers to questions of the form (###) fall under the schema:

(!) A(x/c1 ) ∧ ... ∧ A(x/cn ) ∧ ∀x(Ax ⊃ x = c1 ∨ ... ∨ x = cn )

where n ≥ 1 and c1 , ...., cn are distinct individual constants.

Questions of the form ? S(Ax) can be read “Which x is such that Ax?”.
We shall call them existential wh-questions. Questions of the form ? U(Ax)
can be read “What are all of the x’s such that Ax?” and thus we call them
general wh-questions.
Wh-questions of L2 are counterparts of some natural-language wh-ques-
tions. Yet, there are natural-language wh-questions which can be formalized
by means of questions of the form (#). Note that a question of the form
? {P c1 , ..., P cn } can be read “Which of the c1 , ..., cn has the property P ?”.
The symbols S and U belong to the vocabulary of the object-level lan-
guage L2 . Yet, we can introduce them to the metalanguage as well (but
with different meanings). We assume that on the metalanguage level an ex-
pression S(Ax) designates the set of all the sentences of the form A(x/c),
whereas U(Ax) designates the set of all the sentences of the form (!). We
adopt this solution for two reasons. First, we will be dealing with sets of
direct answers to questions of L2 and it is convenient to have a specific no-
tation for them. Second, we can now say that each question of L 2 consists
of the sign ? followed by an (object-language) expression which is equiform

i i

i i
 “01wisniewski”
i i 2003/6/9
page 22
i i

22 ANDRZEJ WIŚNIEWSKI

to a metalanguage expression that designates the set of direct answers to the

question.
The semantics for the declarative part of L2 is a model-theoretic one.
By an interpretation of the declarative part of L2 we mean an ordered pair
<M, f >, where M is a non-empty set (the domain) and f is the interpre-
tation function defined on individual constants and predicates of L 2 in the
standard way. The concept of truth of a d-wff in an interpretation is defined
in the standard manner.
Admissible partitions of L2 are determined by the class of all interpreta-
tions (cf. Section 4, Example 2) and the remaining semantical concepts are
defined accordingly (cf. Section 4).
We are now ready to give some examples of evocation in L 2 . In what
follows we assume that distinct metalinguistic variables represent distinct
object-language entities.

(6.16) P c1 , ..., P cn `E ? ∀xP x,

(6.17) P c1 ∧ Rc1 , ..., P cn ∧ Rcn `E ? ∀x(P x ⊃ Rx),

(6.18) P c1 ∧ Rc1 , ..., P cn ∧ Rcn , P cn+1 `E ? Rcn+1 ,

(6.19) P c1 , P c2 `E ? c1 = c2 ,

(6.20) ∃xP x `E ? {P c, ∃x(P x ∧ x 6= c)},

(6.21) ∃x(P x ∧ (x = c1 ∨ ... ∨ x = cn )) `E ? {P c1 , ..., P cn },

where n > 1,

(6.22) ¬∀x(x = c1 ∨ ... ∨ x = cn ⊃ P x) `E ? {¬P c1 , ..., ¬P cn },

where n > 1,

(6.23) ∃x(P x ∧ (x = c1 ∨ ... ∨ x = cn )) `E ? S(P x),

where n > 1,

(6.24) ¬∀x(x = c1 ∨ ... ∨ x = cn ⊃ P x) `E ? S(¬P x),

where n > 1,

i i

i i
 “01wisniewski”
i i 2003/6/9
page 23
i i

QUESTIONS AND INFERENCES 23

(6.25) ∀x(P x ≡ x = c1 ∨ ... ∨ x = cn ) ∨ ...∨

∀x(P x ≡ x = c01 ∨ ... ∨ x = c0k ) `E ? U(P x),

where n > 1 and k > 1.

We have considered only relatively simple languages enriched with rather

simple questions. For further examples of evocation, in particular evoca-
tion of complex questions in more sophisticated languages see Wiśniewski
(1995a), Chapter 5.

7. Erotetic Implication

Unlike evocation, erotetic implication is a complex notion.

Assume again that L is a language which fulfills the conditions specified
in Sections 3 and 4. Erotetic implication is characterized by:

Definition 9: A question Q implies a question Q1 on the basis of a set of

d-wffs X (in symbols: Im(Q, X, Q1 )) iff
(i) for each A ∈ dQ: X ∪ {A} ||= dQ1 , and
(ii) for each B ∈ dQ1 there exists a non-empty proper subset Y of dQ
such that X ∪ {B} ||= Y .

If Im(Q, X, Q1 ), then Q1 is said to be the implied question, Q is the im-

plying question and the elements of X are called auxiliary d-wffs.
Let us now analyze the intuitive content of the proposed definition.
According to clause (i), the set of direct answers to an implied question is
multiple-conclusion entailed by each set made up of the auxiliary d-wffs and
a direct answer to the implying question (recall that a question has at least
two direct answers). Thus erotetic implication warrants the transmission of
soundness/truth into soundness: if an implying question is sound (i.e. has
at least one true direct answer) and all the auxiliary d-wffs are true, then
the implied question is sound as well, and this condition is fulfilled for each
admissible partition of the language. Yet, clause (ii) yields (but does not
amount to!) the reverse: if an implied question is sound and all the auxiliary
d-wffs are true, then the implying question is sound (again, this condition
is fulfilled for each admissible partition). Of course, it is neither assumed
nor denied that Q and/or Q1 are sound, and that X consists of truths. The
following table presents possible connections; we assume that Im(Q, X, Q 1 )

i i

i i
 “01wisniewski”
i i 2003/6/9
page 24
i i

24 ANDRZEJ WIŚNIEWSKI

and that P =<T, U > is an arbitrary but fixed admissible partition of the
language:

Q X Q1
sound in P X⊂T sound in P
unsound in P X⊂T unsound in P
sound in P Xnon⊂ T sound in P or unsound in P
unsound in P Xnon⊂ T sound in P or unsound in P

Table 3.

Q1 X Q
sound in P X⊂T sound in P
unsound in P X⊂T unsound in P
sound in P Xnon⊂ T sound in P or unsound in P
unsound in P Xnon⊂ T sound in P or unsound in P

Table 4.
Thus erotetic implication is an “almost equivalence” with respect to sound-
ness. Yet, one cannot reduce the intuitive content of the analyzed concept
to such an equivalence. This is due to the second clause of the proposed
definition.
According to clause (ii) of Definition 9, each set made up of a direct an-
swer to an implied question and the auxiliary d-wffs multiple-conclusion
entails a certain non-empty proper subset of the set of direct answers to the
implying question. The words italicized above are crucial. Let us start from
the last. The set of direct answers to a question can be viewed as the class of
possibilities offered by the question. Thus when we have a proper subset of a
set of direct answers, the class of possibilities is narrowed down. The clause
(ii) requires that for each direct answer B to the implied question Q 1 there
exists a certain non-empty proper subset Y of the set of direct answers to the
implying question Q such that Y is multiple-conclusion entailed by X∪{B}.
Thus if B is true and all the auxiliary d-wffs in X are true, we have a guar-
antee that a true direct answer to the implied question Q is among those in
the restricted class Y . Now observe that clause (ii) speaks about each di-
rect answer to the implied question. Consequently, each couple <B, X>,
where B ∈ dQ1 , fixes a certain restricted class of possibilities, and, given
that X consists of truths, a couple which contains a true direct answer to Q 1

i i

i i
 “01wisniewski”
i i 2003/6/9
page 25
i i

QUESTIONS AND INFERENCES 25

(if there is any) determines the real possibilities (that is, a class of answers
to Q which must contain at least one true answer). Thus an implied question
is potentially cognitively useful with respect to the implying question and,
since clause (ii) speaks about each direct answer to the implied question, the
usefulness of an implied question can be called “open-minded”. Using the
philosophical jargon, erotetic implication defined above has a “teleological
feature” and this feature distinguishes it from other proposals.
Of course, different answers to an implied question usually point at differ-
ent sets of possibilities. Moreover, it may happen that the appropriate proper
subsets are simply singleton sets, as happens in the example presented be-
low:

Example 4:

Q: How old is Andrew?

X: Andrew is as old as Peter.
Q1 : How old is Peter?

It may also happen that no single direct answer to the implying question is
forthcoming (but still subclasses are fixed), viz.:

Example 5:

Q: Where does Andrew live?

X: Andrew lives in a university town in Western Poland.
Q1 : Which towns in Western Poland are university towns?

The implied question is construed here as a “complete list” question: a

direct answer to it provides us with a list of towns together with the clause
that they are the only university towns in Western Poland.
A mixed situation is also possible. Let us consider:

Example 6:

Q: Where did Andrew leave for: Paris, London, or Moscow?

X: If Andrew left for Paris, London or Moscow, then he departed in the
morning or in the evening.
If Andrew departed in the morning, then he left for Paris or London.
If Andrew departed in the evening, then he left for Moscow.
Q1 : When did Andrew depart: in the morning, or in the evening?
/ \
[morning] [evening]
[Paris, London] [Moscow]

i i

i i
 “01wisniewski”
i i 2003/6/9
page 26
i i

26 ANDRZEJ WIŚNIEWSKI

Yet, irrespective of the possible practical benefits which may be gained,

erotetic implication is still a well-defined semantical concept. We say that an
erotetic inference of the second kind <Q, X, Q1 > is valid iff Im(Q, X, Q1 ).
The properties of erotetic implication have been characterized elsewhere
(cf. Wiśniewski 1990a, 1994a, 1995a, 1996). Let us only mention here a few
facts. First, in the general case erotetic implication is not “transitive”, i.e. if
Im(Q, X, Q1 ) and Im(Q1 , X, Q2 ), then not always Im(Q, X, Q2 ). But if
Q1 regularly implies Q2 on the basis of X, then transitivity holds. Regular
erotetic implication is a special case of erotetic implication. We have:

Definition 10: A question Q regularly implies a question Q1 on the basis of

a set of d-wffs X iff
(i) for each A ∈ dQ: X ∪ {A} ||= dQ1 , and
(ii) for each B ∈ dQ1 there exists A ∈ dQ such that X ∪ {B} entails A.

The difference lies in the fact that single direct answers to an implying ques-
tion are pointed at (i.e. entailed) by couples made up of X and a direct
answer to the implied question (cf. Example 4).
Second, observe that when the following condition is fulfilled:
(*) for each B ∈ dQ1 : X ∪ {B} entails some direct or partial answer to
Q

then clause (ii) of Definition 9 is fulfilled as well. Moreover, condition (*)

is equivalent to clause (ii) of Definition 9 given that mc-entailment in a lan-
guage is compact.
Third, observe that if Im(Q, X, Q1 ), then each presupposition of the im-
plied question Q1 is entailed by any set made up of X and a direct answer to
the implying question Q.
Fourth, if Q and Q1 are normal questions (cf. Section 4), then clause (i)
of Definition 9 is equivalent to the following:
(**) for each C ∈ PresQ1 : PresQ ∪ X entails C.

Moreover, if Q and Q1 are regular questions, then clause (i) of Definition 9

is equivalent to:
(***) there exist: a prospective presupposition D of Q, and a prospective
presupposition C of Q1 such that X ∪ {D} entails C.

Recall, however, that erotetic implication is defined by means of two clauses.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 27
i i

QUESTIONS AND INFERENCES 27

8. Erotetic Implication: Examples

In what follows we shall use the letters A, B, C, D, E, F , with subscripts if

needed, as metalinguistic variables for d-wffs of L1 and L2 (cf. sections 6.1
and 6.2). When no quantifiers or symbols referring to individual variables or
constants occur, the formulas presented below refer both to erotetic implica-
tion in L1 and in L2 ; otherwise only L2 is taken into consideration. We shall
supplement the examples with comments included in square brackets, which
will indicate the type of erotetic implication involved.

8.1. Pure erotetic implication

Definition 9 neither assumes nor denies that X is non-empty. If Im(Q, ∅, Q 1 ),

then we say that Q implies Q1 and we write Im(Q, Q1 ). In this case we have
pure erotetic implication. Analytic erotetic implication is a special case of
pure erotetic implication.

Definition 11: A question Q analytically implies a question Q1 iff Im(Q, Q1 )

and each immediate subformula of a direct answer to Q1 is a subformula14
of a direct answer to Q or a negation of a subformula of a direct answer to
Q.

For transparency, we write Q `Im Q1 instead of Im(Q, Q1 ). The follow-

ing hold:
(8.1) ? ¬A `Im ? A, [regular, analytic]

(8.2) ? A `Im ? ¬A, [regular, analytic]

(8.3) ? {A, B} `Im ? {B, A}, [regular, analytic]

(8.4) ? {A, B ∨ C} `Im ? {A, B, C}, [regular, analytic]

(8.5) ? {A, B, C} `Im ? {A, B ∨ C}, [analytic]

(8.6) ? (A ⊗ B) `Im ? ± | A, B |, [regular, analytic]

where ⊗ is any of the connectives: ∧, ∨, ⊃, ≡.

14
We assume that subformulas are defined in the standard manner; in particular, a d-wff
A is a subformula of A. Moreover, we assume that negation is present in a language.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 28
i i

28 ANDRZEJ WIŚNIEWSKI

(8.7) ? ± | A, B | `Im ? (A ⊗ B), [not analytic if ⊗ 6= ∧]

where ⊗ is any of the connectives: ∧, ∨, ⊃, ≡.

(8.8) ? ± | A, B | `Im ? A, [analytic]

(8.9) ? ± | A, B | `Im ? B, [analytic]

(8.10) ? A `Im ? ± | A, B |, [regular, not analytic]

(8.11) ? B `Im ? ± | A, B |, [regular, not analytic]

8.2. Erotetic implication on the basis of non-empty sets of d-wffs

We write Q, A1 , ..., An `Im Q1 instead of Im(Q, A1 , ..., An , Q1 ). The fol-

lowing hold:
(8.12) ? (A ⊗ B), αA `Im ? B, [regular]

where ⊗ is any of the connectives: ∧, ∨, ⊃, ≡, and αA is equal to ¬A or to

A,

(8.13) ? (A ⊗ B), αB `Im ? A, [regular]

where ⊗ is any of the connectives: ∧, ∨, ⊃, ≡, and αB is equal to ¬B or to

B,

(8.14) ? {¬A, A ∧ B, A ∧ ¬B} `Im ? A,

(8.15) ? {A1 , ..., An }, A1 ∨ ... ∨ An `Im ? Ai , [regular if n = 2]

where 1 ≤ i ≤ n.

(8.16) ? {A1 , ..., An }, A1 ∨...∨An , ¬(A1 ∧...∧An ) `Im ? {¬A1 , ..., ¬An },
[regular if n = 2]

(8.17) ? {A1 , ..., An }, A1 ≡ B1 , ..., An ≡ Bn `Im ? {B1 , ..., Bn },

[regular]

i i

i i
 “01wisniewski”
i i 2003/6/9
page 29
i i

QUESTIONS AND INFERENCES 29

(8.18) ? {A1 , ..., An }, B ⊃ A1 ∨ ... ∨ Ai , ¬B ⊃ Ai+1 ∨ ... ∨ An `Im ? B,

[regular if n = 2]
where 1 ≤ i < n.

(8.19) ? {A1 , ..., An }, B ⊃ A1 ∨ ... ∨ Ai , C ⊃ Ai+1 ∨ ... ∨ An , B ∨ C `Im

? {B, C},
[regular if n = 2]
where 1 ≤ i < n.

(8.20) ? {A1 , ..., An }, A1 ∨...∨Ai ≡ B, Ai+1 ∨...∨An ≡ C `Im ? {B, C},

[regular if n = 2]
where 1 ≤ i < n.

(8.21) ? {A1 , ..., An }, A1 ∨ ... ∨ Ai , ⊃ B, Ai+1 ∨ ... ∨ An ⊃ ¬B, A1 ∨ ... ∨

An `Im ? B,
[regular if n = 2]
where 1 ≤ i < n.

(8.22) ? A(x/c), ∀x(Ax ≡ Bx ∧ Cx) `Im ? ± | B(x/c), C(x/c) |,

[regular]

(8.23) ? S(Ax), ∀x(Ax ≡ Bx) `Im ? S(Bx), [regular]

(8.24) ? U(Ax), ∀x(Ax ≡ Bx) `Im ? U(Bx), [regular]

(8.25) ? S(Ax), ∀x(Ax ⊃ x = c1 ∨ ... ∨ cn ) `Im ? {A(x/c1 ), ..., A(x/cn )},

[regular]
where c1 , ..., cn (n > 1) are distinct individual constants,

(8.26) ? S(Ax), ∃xAx, ∀x(Ax ⊃ x = c1 ∨ ... ∨ cn ) `Im ? A(x/ci ),

where c1 , ..., cn (n > 1) are distinct individual constants and 1 ≤ i ≤ n.

It is no accident that wh-questions of L2 occur in the above examples only

a few times. It may be argued that their proper analysis requires a semantics
in which all the elements of a domain have names (i.e. are values of some
closed terms). Moreover, in order to analyze erotetic implication of gen-
eral wh-questions of L2 and other “complete-list” questions one needs the
quantifier “there exist finitely many” (cf. Wiśniewski 1990a) or at least nu-
merical quantifiers (cf. Wiśniewski 1995a). For further examples of erotetic
implication, in particular in more sophisticated languages and pertaining to
more complex questions, see Wiśniewski (1990a) and Wiśniewski (1995a),
Chapter 7.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 30
i i

30 ANDRZEJ WIŚNIEWSKI

9. Rules and Scenarios

9.1. E-rules and Im-rules

Erotetic inferences occur in almost every process of reasoning. Yet, the com-
mon attitude has been to regard them as belonging to the “pragmatics” of
reasoning, in the pejorative sense of pragmatics as referring to something
that is not subjected to any objective rules. No doubt, there are erotetic in-
ferences of this kind. But there are also erotetic inferences which have a
well-established structure due to the fact that evocation or erotetic impli-
cation hold between their premises and conclusions. IEL regards them as
valid. Of course, validity is a normative concept and the appropriate concept
of validity is neither given by God nor by Tradition (even Jaakko Hintikka
is silent here). But even if one has other intuitions concerning validity of
erotetic inferences, the semantic warrants connected with IEL-valid erotetic
inferences still remain. Moreover, if we have a formula which says that a
question of a given form is evoked by a finite set of d-wffs made up of d-
wffs of strictly defined forms (cf. examples in Section 6), we can formulate a
rule each application of which leads to the question which is both sound and
informative with respect to the premises. Similarly, when a formula of the
form Q, A1 , ..., An `Im Q1 (or of the form Q `Im Q1 ) holds (cf. examples
in Section 8), we can formulate a rule each application of which leads to the
question which is both sound with respect to the premises and cognitively
useful with respect to the initial question. These rules are called E-rules and
Im-rules, respectively.
Yet, IEL does not stop here. The logical tools elaborated within IEL make
an analysis of search scenarios possible. As a result of the analysis, search
rules are characterized.

9.2. Erotetic search scenarios

What are erotetic search scenarios (e-scenarios for short)? A few examples
will help to clarify matters here.

Example 7:

Let us imagine that we are looking for an answer to the question:

(9.1) Did John meet Mary and talk to her?

Question (9.1) is construed here in such a way that the following are direct
answers to it:

i i

i i
 “01wisniewski”
i i 2003/6/9
page 31
i i

QUESTIONS AND INFERENCES 31

(9.2) John didn’t meet Mary.

(9.3) John met Mary and talked to her.
(9.4) John met Mary, but didn’t talk to her.

The natural way of proceeding is to ask the following question first:

(9.5) Did John meet Mary?

If the negative answer to (9.5) holds, the answer (9.2) to (9.1) is found. If,
however, the affirmative answer to (9.5) holds, the following question should
be asked:
(9.6) Did John talk to Mary?

The affirmative answer to (9.6) together with the (already established) affir-
mative answer to (9.5) give the answer (9.3) to the principal question (9.1),
whereas the negative answer to (9.6) together with the affirmative answer to
(9.5) yield the answer (9.4) to the principal question.
The following diagram presents the e-scenario involved:

Figure 1.
? {¬p, p ∧ q, p ∧ ¬p}
  ? pP PP
 P
p ¬p

? qP
 PP
 P
q ¬q
p∧q p ∧ ¬q

Example 8:

Let us now imagine that we are looking for an answer to the following prin-
cipal question:
(9.7) Where did John leave for: Paris, London, Kiev, or Moscow?

on the basis of the following initial premises:

(9.8) If John departed in the morning, then he left for Paris or London.
(9.9) If John did not depart in the morning, then he left for Moscow.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 32
i i

32 ANDRZEJ WIŚNIEWSKI

(9.10) John left for London if he took his famous umbrella; otherwise he
did not leave for London.

We can start with asking the following question:

(9.11) Did John depart in the morning?

If the negative answer to (9.11) is true, then by (9.9) John left for Moscow.
If, however, the affirmative answer to (9.11) holds, then the following ques-
tion arises:
(9.12) Where did John leave for: London or Paris?

Yet, question (9.12) need not be asked. The following question:

(9.13) Did John take his famous umbrella?

is now raised by question (9.12) on the basis of the affirmative answer to

(9.11) and the premises (9.8) and (9.10). The affirmative answer to (9.13)
together with premise (9.10) yield that John left for London, whereas the
negative answer to (9.13) together with the affirmative answer to (9.11) and
premise (9.8) yield that John left for Paris.
The following diagram presents the appropriate e-scenario:

Figure 2.
? {p, q, r, s}
t⊃p∨q
¬t ⊃ s
q≡u
?t
 P
 PP
 P
t ¬t
? {p, q} s
  ? uP PP
 P
u ¬u
q p

Example 9:

Our third example is more abstract. Assume that P1 , P2 , P3 are one-place

predicates of L2 and that a is an individual constant of L2 . Let us imagine

i i

i i
 “01wisniewski”
i i 2003/6/9
page 33
i i

QUESTIONS AND INFERENCES 33

that we are looking for an answer to a question of the form:

(9.14) ? P1 a

(“Does a have the property P1 ?”) on the basis of the following initial premise:
(9.15) ∀x(P1 x ≡ P2 x ∧ P3 x)

Figure 3 shows a possible e-scenario:

Figure 3.
? P1 a
∀x(P1 x ≡ P2 x ∧ P3 x)
? ± | P2 a, P3 a |
? P2 a
 P
 PP
 P
P2 a ¬P2 a
? P3 a
 P
¬P1 a
 PP
 P
P3 a ¬P3 a
P1 a ¬P1 a
The principal question (9.14) on the basis of (9.15) implies:
(9.16) ? ± | P2 a, P3 a |

(cf. (8.22)). Question (9.16), in turn, implies:

(9.17) ? P2 a

(cf. (8.8)). Question (9.17) is now asked and answered. Since it has two di-
rect answers, the scenario branches. The right branch contains the negative
answer ‘¬P2 a’ to (9.17) and, since ‘¬P2 a’ and (9.15) entail ‘¬P1 a’, ends
with the negative answer to the principal question (9.14). The left branch
contains the affirmative answer ‘P2 a’ to (9.16) followed by the question:
(9.18) ? P3 a

which is implied by question (9.17) (cf. (8.9)). Question (9.18) is asked and
answered, and the scenario branches again. The leftmost branch contains the
affirmative answer to (9.18) and ends with the affirmative answer to the prin-
cipal question (9.14), which is now entailed by the results established so far
and the premise (9.15). The right sub-branch contains the negative answer
to (9.18) and ends with the negative answer to the principal question (9.14);

i i

i i
 “01wisniewski”
i i 2003/6/9
page 34
i i

34 ANDRZEJ WIŚNIEWSKI

this answer is entailed by the answer just received (i.e. ‘¬P 3 a’) and premise
(9.15).

Although e-scenarios can be represented by downward trees, they are de-

fined as families of mutually interconnected erotetic derivations (e-deriva-
tions for short). An e-derivation corresponds to a branch of a tree: it starts
with the principal question and ends with a direct answer to it. An auxil-
iary question which occurs in an e-derivation is either asked (is a query) or
serves as a premise; each auxiliary question must be implied by a certain
question which occurs earlier, possibly on the basis of some earlier d-wffs.
A d-wff can enter an e-derivation for three possible reasons: (a) as an ini-
tial premise, or (b) as a direct answer to an auxiliary question, and (c) as a
consequence of some d-wff(s) which satisfy the conditions (a) or (b). It is
assumed that if an e-scenario involves an auxiliary question which is then
answered in some way in an e-derivation, then the e-scenario contains also
e-derivations in which the auxiliary question is answered in all the other
possible ways; these e-derivations are identical up to the point at which the
auxiliary question is asked, but start to differ at the level of answers to the
auxiliary question (thus an e-scenario is “fair” with respect to queries).
To be more precise, e-scenarios can be defined as follows.
Let L be a formalized language which fulfills the conditions specified in
sections 3 and 4. We shall use the lower-case Greek letters ϕ, ψ, γ, with
subscripts if needed, as metalinguistic variables for wffs (i.e. d-wffs and
questions) of L. We have:

Definition 12: A finite sequence e = ϕ1 , ..., ϕn of wffs is an erotetic deriva-

tion of a direct answer A to the question Q on the basis of a set of d-wffs X
iff ϕ1 = Q, ϕn = A and the following conditions hold:
(1) for each question ϕk of e such that k > 1:
(a) dϕk 6= dQ, and
(b) ϕk+1 is either a question or a direct answer to ϕk ;
(2) for each d-wff ϕj of e:
(a) ϕj ∈ X, or
(b) ϕj is a direct answer to ϕj−1 , where ϕj−1 6= Q, or
(c) ϕj is entailed by a certain set of d-wffs such that each element
of this set precedes ϕj in e;
(3) for each question ϕk of e such that ϕk 6= Q: ϕk is implied by a
certain question ϕi which precedes ϕk in e on the basis of the empty
set, or on the basis of a set of d-wffs such that each element of this
set precedes ϕk in e.

Note that by “precedes” we do not mean “immediately precedes”.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 35
i i

QUESTIONS AND INFERENCES 35

An element ϕk (where 1 < k < n) of an e-derivation e = ϕ1 , ..., ϕn is a

query of e if ϕk is a question and ϕk+1 is a direct answer to ϕk . Note that
e-derivations can involve auxiliary questions which are not queries and serve
as erotetic premises only (recall that erotetic implication is not “transitive”).

Definition 13: A finite family Φ of sequences of wffs is an erotetic search

scenario for a question Q relative to a set of d-wffs X iff each element of Φ
is an e-derivation of a direct answer to Q on the basis of X and the following
conditions hold:
(1) dQ ∩ X = ∅;
(2) Φ contains at least two elements;
(3) for each element e = ϕ1 , ϕ2 , ..., ϕn of Φ, for each index k such that
1 ≤ k < n:
(a) if ϕk is a question and ϕk+1 is a direct answer to ϕk , then for
each direct answer B to ϕk , the family Φ contains a certain
e-derivation e0 = ψ1 , ψ2 , ..., ψm such that ψj = ϕj for j =
1, ..., k, and ψk+1 = B;
(b) if ϕk is a d-wff, or ϕk is a question and ϕk+1 is not a direct
answer to ϕk , then for each e-derivation e0 = ψ1 , ψ2 , ..., ψm in
Φ such that ψj = ϕj for j = 1, ..., k we have ψk+1 = ϕk+1 .

The e-derivations which are elements of an erotetic search scenario Φ for

Q relative to X are called paths of Φ, the question Q is the principal question
of Φ, and the elements of the set X are called initial premises. If a path e of
Φ has a direct answer A to Q as its last element, we say that e leads to A. If
Φ is an e-scenario for Q relative to the empty set, we simply say that Φ is an
e-scenario for Q.
An e-scenario for Q relative to X is said to be complete iff each direct an-
swer to Q is the endpoint of some path of the scenario, and incomplete oth-
erwise. Figures 1 and 3 present examples of complete e-scenarios, whereas
Figure 2 shows an incomplete e-scenario.
A query of an e-scenario is a query of a path of the e-scenario.
Roughly, a branching point of an e-scenario is a wff which occurs in the e-
scenario in such a way that the tree-like diagram which presents this scenario
branches immediately below the occurrence of the wff. To be more precise,
we say that a wff γ is a branching point of an e-scenario Φ iff there are: an
index k > 1 and a subset Ψ of Φ such that Ψ contains at least two elements
and for any e = ϕ1 , ϕ2 , ..., ϕn , e0 = ψ1 , ψ2 , ..., ψm (where e 6= e0 ) in Ψ we
have: (i) γ = ϕk = ψk , (ii) ϕj = ψj for j = 1, ..., k, and (iii) ϕk+1 6= ψk+1 .
The properties of e-scenarios were analyzed elsewhere (cf. Wiśniewski
(2003)). Let us only mention the following:

i i

i i
 “01wisniewski”
i i 2003/6/9
page 36
i i

36 ANDRZEJ WIŚNIEWSKI

Fact 4: Only questions can play the role of queries of e-scenarios.

Fact 5: Each query of an e-scenario is a branching point of the scenario;

no d-wff can be a branching point.

Fact 6: Each e-scenario involves at least one query; each e-scenario has
a major branching point, which is the first query of every path of the e-
scenario.

Fact 7: A direct answer to the principal question cannot enter an e-scenario

as the second element of any of its paths.

If P =<T, U> is an admissible partition of L and a d-wff A belongs to T,

we say that A is true in P. One can prove:

Golden Path Theorem: Let Φ be an e-scenario for a question Q relative to

a set of d-wffs X. Let P be an admissible partition of L such that each d-wff
in X is true in P and Q is sound in P. Then the scenario Φ contains at least
one path e such that:
(a) each d-wff of e is true in P; and
(b) each question of e is sound in P; and
(c) e leads to a direct answer to Q which is true in P.

The above theorem was proved in Wiśniewski (2003) for the propositional
case. The proof for the generalized case goes along similar lines. Note
that the Golden Path Theorem holds both for complete and incomplete e-
scenarios.
For further examples of e-scenarios see Wiśniewski (2003).

9.3. Search rules

In order to introduce the concept of a search rule we need some auxiliary

concepts.
We say that an e-scenario Φ for Q relative to X is in the canonical form
iff there is no element of X which occurs in Φ after the first branching point
of Φ.
One can prove that for each e-scenario Φ there exists an e-scenario Φ∗ in
the canonical form which results from Φ by moving all the initial premises
(i.e elements of X) occurring on paths of Φ after the first branching point to
places before the first branching point. The e-scenarios presented above are
already in the canonical form.
An e-scenario Φ for Q relative to X is concise iff Φ is in the canonical
form and any d-wff which occurs on a path of Φ is: (i) an element of X, or

i i

i i
 “01wisniewski”
i i 2003/6/9
page 37
i i

QUESTIONS AND INFERENCES 37

(ii) an answer to a query immediately succeeding the query, or (iii) a direct

answer to Q which occurs at the end of the path.
Each e-scenario in the canonical form can be transformed into a corre-
sponding concise e-scenario by deleting all the d-wffs which do not fulfil the
clauses (i), (ii) and (iii) specified above (one may prove that erotetic implica-
tion is retained). The e-scenarios presented in figures 1, 2 and 3 are already
concise.
An e-scenario can involve auxiliary questions which are not queries. These
questions are not requests for information and serve as erotetic premises
only. Let Φ be a concise e-scenario for Q relative to X. The imperative
counterpart of Φ results from Φ by deleting all the auxiliary questions which
are not queries of Φ. Note that imperative counterparts are defined for con-
cise e-scenarios only, and these, in turn, are in the canonical form.
Imperative counterparts of concise e-scenarios need not be e-scenarios
(their constituents need not be e-derivations, since clause (3) of Definition 12
can be violated). Yet, the analogue of Golden Path Theorem holds for such
counterparts.
The following shows the imperative counterpart of the e-scenario pre-
sented in Figure 3:

Figure 4.
? P1 a
∀x(P1 x ≡ P2 x ∧ P3 x)
?
 P
P2 a
 PP
 P
P2 a ¬P2 a
? P3 a
 P
¬P1 a
 PP
 P
P3 a ¬P3 a
P1 a ¬P1 a
Assume that we have a tree-like diagram D of the imperative counterpart
of a concise e-scenario. Let us now replace the non-logical constants/proposi-
tional variables in D with appropriate metalinguistic variables; of course,
different non-logical constants/propositional variables should be replaced by
distinct metalinguistic variables. The result can be viewed as a search rule.
Here are examples of search rules based on the e-scenarios presented above:

i i

i i
 “01wisniewski”
i i 2003/6/9
page 38
i i

38 ANDRZEJ WIŚNIEWSKI

Figure 5.
? {¬A, A ∧ B, A ∧ ¬B}
  AP
?
PP
 P
A ¬A
  BP
?
PP
 P
B ¬B
A∧B A ∧ ¬B
Figure 6.
? {A, B, C, D}
E ⊃A∨B
¬E ⊃ D
B≡F
?
 P
E
 PP
 P
E ¬E

? FP D
 PP
 P
F ¬F
B A
Figure 7.
?Pc
∀x(P x ≡ Rx ∧ T x)
?
 P
Rc
 PP
 P
Rc ¬Rc

? TP
c ¬P c
 PP
 P
Tc ¬T c
Pc ¬P c
But what justifies us in saying that diagrams of the kind presented above
can be called search rules? Let us take a look at the diagram presented in
Figure 7. It conveys the following information: given that the initial premise
holds, in order to answer a principal question of the form ? P c an answer to
question ‘? Rc’ is needed first. If the established answer is a negative one,
the negative answer to the principal question is obtained. If, however, the es-
tablished answer is the affirmative one, an answer to question ‘? T c’ is also
needed. If the established answer to ‘? T c’ is affirmative, then the affirma-
tive answer to the principal question is found. If one establishes the negative

i i

i i
 “01wisniewski”
i i 2003/6/9
page 39
i i

QUESTIONS AND INFERENCES 39

answer to ‘? T c’, then the negative answer to the principal question holds. In
other words, the above diagram provides us with a set of conditional instruc-
tions which tell us both what auxiliary questions should be asked and when
they should be asked in order to answer the principal question; moreover, the
diagram shows what should be done if such-and-such a direct answer to an
auxiliary question appears to be acceptable and does so with respect to any
direct answer to each auxiliary question. The same can be said about other
search rules of the kind considered.
Erotetic search scenarios differ with respect to their complexity and as do
the corresponding search rules. As long as we are concerned with practi-
cal applications, relatively simple search rules seem more useful. Note that
there are search rules which do not involve any initial declarative premises;
Figure 5 gives us an example. Here are more examples 15 (in order to save
space, we present the diagrams in a “horizontal” form):

Figure 8.
A, ¬¬A


? ¬A, ? A P
PP
P
¬A
Figure 9.
B, A ∧ B


A, ? B P
  PP
 P
? (A ∧ B), ? A P ¬B, ¬(A ∧ B)
PP
P
¬A, ¬(A ∧ B)

15
The following does not hold (⊗ stands here for any of the connectives ∧, ∨, ⊃, ≡):

(i) Im(? (A ⊗ B), ? A).

But the following hold:

(8.6) ? (A ⊗ B) `Im ? ± | A, B |,
(8.8) ? ± | A, B | `Im ? A.

The search rules presented in Figures 9–12 result from e-scenarios which have instances of
(8.6) and (8.8) “between” principal questions and first queries; of course, these formulas do
not occur in their imperative counterparts.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 40
i i

40 ANDRZEJ WIŚNIEWSKI

Figure 10.
B, A ⊃ B


A, ? B P
 PP
 P
? (A ⊃ B), ? A P ¬B, ¬(A ⊃ B)
PP
P
¬A, A ⊃ B
Figure 11.
A, A ∨ B


? (A ∨ B), ? A P B, A ∨ B
PP 
P 
¬A, ? B P
PP
P
¬B, ¬(A ∨ B)
Figure 12.
B, A ≡ B


A, ? B hh
hh
 ¬B, ¬(A ≡ B)

? (A ≡ B), ? A P
P PP B, ¬(A ≡ B)
((((
¬A, ? B P
P PP
¬B, A ≡ B

10. Related problems and applications

The analysis of search scenarios has given rise to a certain proof method,
called Synthetic Tableaux Method (STM). STM was developed by Urba ński
(cf. Urbański 2001, 2002, and (forthcoming)) for CPC and some non-classi-
cal propositional calculi; completeness results were proved. Looking from
the erotetic point of view, a STM-style proof proceeds by answering the
operative questions which are implied by the principal question (which has
the form of a simple yes-no question) on the basis of the results obtained at
previous steps. Yet, the reference to erotetic implication is only implicit: an
STM-style proof is a well-defined proof-theoretical entity.
The problem of reducibility of questions to sets of (“operative”) questions
is important both to IEL and to STM. A semantical concept of reducibility is

i i

i i
 “01wisniewski”
i i 2003/6/9
page 41
i i

QUESTIONS AND INFERENCES 41

defined in Wiśniewski (1993); numerous results can be found in Wiśniewski

(1994b), (1995a) and (forthcoming). In particular, the problem of reducibil-
ity of evoked questions to sets of questions which are implied by them on
the basis of evoking sets was addressed and some general results were ob-
tained (cf. Wiśniewski 1995a, Chapter 7.6). Leśniewski considered a more
general notion of reducibility, namely the reducibility of a single question to
a set of questions on the basis of a non-empty set of declarative premises; he
received elegant and important results (cf. Leśniewski 1997, 2000).
Some concepts borrowed from IEL have been applied in the philosophy
of science. Kuipers & Wiśniewski (1994) present an analysis of the so-
called explanation by specification in terms of erotetic inferences, question
generation and erotetic implication. Wiśniewski (1999a) introduces a rela-
tivized (with respect to background knowledge and an abnormic hypothesis
in Bromberger’s sense16 ) concept of a possible correct answer to a why-
question; some IEL-based procedures for looking for answers of this kind
are analyzed. Tworak (1993) addresses the problem of the cognitive impor-
tance of research questions by using, int. al., evocation and erotetic impli-
cation. Sady (1990) applies the concept of question generation in rational
reconstructions of some scientific discoveries.
As we have seen, IEL does not require Classical Logic (CL) as a basis.
Yet, most of the exemplifications of the central concepts of IEL are based on
CL and thus inherit all the merits and shortcomings of CL. For example, as
long as the informativeness of a question is defined in terms of lack of CL-
entailment (of any direct answer from the relevant body of knowledge), the
logical omniscience paradox comes into play and no question is informative
in an inconsistent environment. In practice, however, a scientist is usually
unaware of many CL-consequences of his/her insights, and is able to operate
in the presence of an inconsistency. Meheus (1999) makes an important step
towards adapting evocation and question generation to the inconsistent case.

Institute of Philosophy
University of Zielona Góra
Wojska Polskiego 71a
65-762 Zielona Góra
Poland
E-mail: [email protected]

16
Cf. Sylvain Bromberger On What We Know We Don’t Know: Explanation, Theory,
Linguistics, and How Questions Shape Them, The University of Chicago Press, Chicago and
London, and Center for Study of Language and Information, Stanford 1992.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 42
i i

42 ANDRZEJ WIŚNIEWSKI

REFERENCES

Kuipers, T.A.F. and A. Wiśniewski (1994), ‘An erotetic approach to expla-

nation by specification’, Erkenntnis 40, No. 3, pp. 377–402.
Leśniewski, P. (1997), Zagadnienie sprowadzalności w antyredukcjonisty-
cznych teoriach pytań (The Problem of Reducibility of Questions in
Non-Reductionistic Theories of Questions), Wydawnictwo Naukowe In-
stytutu Filozofii UAM, Poznań.
Leśniewski, P. (2000), ‘On the generalized reducibility of questions’, in:
J. Nida-Rumelin (ed.), Rationality, Realism, Revision. Proceedings of
the 3rd International Congress of the Society for Analytical Philosophy,
Walter de Gruyter, Berlin/New York, pp. 119–126.
Leśniewski, P. and A. Wiśniewski (2001), ‘Reducibility of questions to sets
of questions: Some feasibility results’, Logique et Analyse, this volume.
Meheus, J. (1999), ‘Erotetic arguments from inconsistent premises’, Logique
et Analyse 165–166, pp. 49–80 (appeared in 2002).
Meheus, J. (2001), ‘Adaptive logics for question evocation’, Logique et
Analyse, this volume.
Sady, W. (1990), Racjonalna rekonstrukcja odkryć naukowych (Rational
Reconstruction of Scientific Discoveries), Wydawnictwo Uniwersytetu
Marii Curie-Skłodowskiej, Lublin 1990.
Tworak, Z. (1993), ‘Zagadnienie doniosłości poznawczej pytań
badawczych’ (The problem of cognitive importance of research
questions), Studia Metodologiczne 28, pp. 65–85.
Urbański, M. (1998), ‘Some remarks concerning modal propositional logic
of questions’, Logic and Logical Philosophy 6, pp. 187–196.
Urbański, M. (2001), ‘Remarks on synthetic tableaux for Classical Proposi-
tional Calculus’, Bulletin of the Section of Logic 30, No. 4, pp. 194–204.
Urbański, M. (2001a), ‘Synthetic tableaux and erotetic search scenarios: ex-
tension and extraction’, Logique et Analyse, this volume.
Urbański, M. (2002), Tabele syntetyczne a logika pytań (Synthetic Tableaux
and the Logic of Questions), Wydawnictwo Uniwersytetu Marii Curie-
Skłodowskiej, Lublin.
Urbański, M. (forthcoming), ‘Synthetic tableaux for Łukasiewicz’s logic
Ł3’, Logique et Analyse.
Wiśniewski, A. (1985), ‘Propositional logic and erotetic inferences’, Bulletin
of the Section of Logic PAN 14, No. 2, pp. 72–78. Erratum, ibid., 15,
No. 1, 1986, p. 35.
Wiśniewski, A. (1989), ‘The generating of questions: a study of some
erotetic aspects of rationality’, in: L. Koj and A. Wiśniewski, Inquiries
into the Generating and Proper Use of Questions, Wydawnictwo Uniw-
ersytetu Marii Curie-Skłodowskiej, Lublin, pp. 91–155.

i i

i i
 “01wisniewski”
i i 2003/6/9
page 43
i i

QUESTIONS AND INFERENCES 43

Wiśniewski, A. (1990a), ‘Implied questions’, Manuscrito 13, No. 2, pp. 23–

38.
Wiśniewski, A. (1990b), Stawianie pytań: logika i racjonalność (The Pos-
ing of Questions: Logic and Rationality), Wydawnictwo Uniwersytetu
Marii Curie-Skłodowskiej, Lublin.
Wiśniewski, A. (1991a), ‘Erotetic arguments: a preliminary analysis’, Studia
Logica 50, No. 2, pp. 261–274.
Wiśniewski, A. (1991b), ‘Multiple-conclusion consequence and erotetic im-
plication’, Logic Group Bulletin, Warsaw Scientific Society, Supplement
to No. 1, pp. 9–11.
Wiśniewski, A. (1993), ‘Reducibility of questions to sets of yes-no ques-
tions’, Bulletin of the Section of Logic 22, No. 3, pp. 119–126.
Wiśniewski, A. (1994a), ‘Erotetic implications’, Journal of Philosophical
Logic 23, No. 2, pp. 173–195.
Wiśniewski, A. (1994b), ‘On the reducibility of questions’, Erkenntnis 40,
No. 2, pp. 265–284.
Wiśniewski, A. (1995a), The Posing of Questions: Logical Founda-
tions of Erotetic Inferences, Kluwer Academic Publishers, Dor-
drecht/Boston/London 1995.
Wiśniewski, A. (1995b), ‘Some results on the reducibility of questions’, Bul-
letin of Symbolic Logic, vol. 1, No. 1, p. 137.
Wiśniewski, A. (1996), ‘The logic of questions as a theory of erotetic argu-
ments’, Synthese 109, No. 2, pp. 1–25.
Wiśniewski, A. (1997), ‘Some foundational concepts of erotetic semantics’,
Poznań Studies in the Philosophy of the Sciences and the Humanities 51,
pp. 181–211 (special issue: Knowledge and Inquiry: Essays on Jaakko
Hintikka’s Epistemology and Philosophy of Science, ed. by Matti Sinto-
nen).
Wiśniewski, A. (1999a), ‘Erotetic logic and explanation by abnormic hy-
potheses’, Synthese 120, No. 3, pp. 295–309.
Wiśniewski, A. (1999b), ‘Questions and complete search scenarios’, Ab-
stracts of the 11th International Congress of Logic, Methodology and
Philosophy of Science, Kraków 1999, p. 148.
Wiśniewski, A. (2003), ‘Erotetic search scenarios’, Synthese 134, No. 3,
pp. 389–427.
Wiśniewski, A. (forthcoming), ‘Reducibility of safe questions to sets of
atomic yes-no questions’, Poznań Studies in the Philosophy of the Sci-
ences and the Humanities, special issue: The Lvov-Warsaw School: Sec-
ond Generation.

i i

i i