4

From this, it follows that much of the methodology of epistemology and of

the methodology of science will be tantamount to the strategic principles of

bracketing. From this, it is in turn seen that a study of uncertain answers is

an enormously complicated enterprise, dif¬cult to achieve an overview of. It

nevertheless promises useful insights. A sense of this usefulness of the inter-

rogative model in dealing with the problems of methodology and inference

can perhaps be obtained by considering suitable special problems of inde-

pendent interest. The two brief essays, “A Fallacious Fallacy” and “Omitting

Data”Ethical or Strategic Problem” (Chapters 9 and 10), illustrate this pur-

pose. The former deals with the so-called conjunctive fallacy. This allegedly

mistaken but apparently hardwired mode of human probabilistic reasoning

is a prize specimen in the famous theory of cognitive fallacies proposed by

Amos Tversky and Daniel Kahneman. The interrogative viewpoint helps to

show that this would-be fallacy is in reality not fallacious at all, but instead

reveals a subtle problem in the Bayesian approach to probabilistic reasoning.

This result cries out for more discussion than can be devoted to the problem

of cognitive fallacies here. Are the other Tversky and Kahneman “fallacies”

perhaps equally dubious?

Omitting observational or experimental data is often considered a serious

breach of the ethics of science. In the second brief essay just mentioned, it is

pointed out, as is indeed fairly obvious from the interrogative point of view,

that such a view is utterly simplistic. Even though data are sometimes omitted

for fraudulent purposes, there is per se nothing ethically or methodologically

wrong about omitting data. Such a procedure can even be required by optimal

strategies of reasoning, depending on circumstances.

But if the basic idea of the interrogative approach to inquiry is this simple

and this old, it might seem unlikely that any new insights could be reached by

its means. Surely its interest has been exhausted long ago, one might expect to

¬nd. The interrogative approach has in fact been used repeatedly in the course

of the history of Western philosophy, for instance in the form of the medieval

obligationes games and in the guise of the “Logic of questions and answers”

in which R. G. Collingwood saw the gist of the historical method. However,

Collingwood™s phrase (taken over later by Hans-Georg Gadamer) indirectly

shows why the elenchus idea has not generated full-¬‚edged epistemological

theories. Collingwood™s “logic” cannot be so-called by the standards of con-

temporary logical theory. In the absence of a satisfactory grasp of the logical

behavior of questions and answers, the idea of “inquiry as inquiry” could not

serve as a basis of successful epistemological theorizing. Such a grasp has only

been reached in the last several years. Admittedly, there have been much

earlier attempts at a logic of questions and answers, also known as “erotetic

logic.” But they did not provide satisfactory accounts of the most important

questions concerning questions, such as the questions about the relation of a

question to its conclusive (desired, intended) answers, about the logical form

of different kinds of questions, about their presuppositions, and so on. One

Introduction 5

might be tempted to blame these relative failures to a neglect of the epistemic

character of questions. For in some fairly obvious sense, a direct question is

nothing more and nothing less than a request for information, a request by the

questioner to be put into a certain epistemic state. Indeed, the speci¬cation

of this epistemic state, known as the desideratum of the question in question,

is the central notion in much of the theory of questions and answers, largely

because it captures much of the essentially (discursive) notions of question

and answer in terms of ordinary epistemic logic.

But the time was not yet ripe for an interrogative theory of inquiry. As is

pointed out in “Second-Generation Epistemic Logic and its General Signi¬-

cance” (Chapter 3), initially modern epistemic logic was not up to the task of

providing a general theory of questions and answers. It provided an excellent

account of the presuppositions and conclusiveness conditions of simple wh-

questions (who, what, where, etc.) and propositional questions, but not of more

complicated questions, for instance of experimental questions concerning the

dependence of a variable on another. However, I discovered that they could

reach the desired generality by indicating explicitly that a logical operator (or

some other kind of notion) was independent of another one. Technically con-

sidered, it was game-theoretical semantics that ¬rst offered to logicians and

logical analysts a tool for handling this crucial notion of independence in the

form of informational independence. These developments form the plot of

Chapter 3.

The interrogative model helps to extend the basic concepts and insights con-

cerning questions to inquiry in general. Some of these insights are examined

in the essay “Presuppositions and Other Limitations of Inquiry” (Chapter 4).

They even turn out to throw light on the earlier history of questioning meth-

ods, including Socrates™ ironic claim to ignorance and Collingwood™s alleged

notion of ultimate presupposition.

Even more radical conclusions ensue from an analysis of the “presupposi-

tions of answers,” which are known as conclusiveness conditions on answers.

They can be said to de¬ne the relation of a question to its conclusive answers.

They are dealt with in the essay “The place of the a priori in epistemology”

(Chapter 5). It quickly turns out that the conclusiveness conditions on answers

to purely empirical questions have conceptual and hence a priori components.

Roughly speaking, the questioner must know, or must be brought to know,

what it is that the given reply refers to. For a paradigmatic example, nature™s

response to an experimental question concerning the dependence of a vari-

able on another can be thought of as a function-in-extension”in other words,

as something like a curve on graph paper. But such a reply truly answers the

dependence question only if the experimental inquirer comes to know what the

function is that governs the dependence between variables”in other (mathe-

matical) words, which function the curve represents. Without such knowledge,

the experimental question has not been fully answered. But this collateral

knowledge is not empirical, but mathematical. Hence, a priori mathematical

Introduction

6

knowledge is an indispensable ingredient even of a purely experimental sci-

ence. Among other consequences, this result should close for good the spurious

issue of the (in)dispensability of mathematics in science.

Since experimental questions are a typical vehicle of inductive inquiry, the

entire problem of induction assumes a new complexion. Inductive reasoning

has not just one aim, but two. It aims not only at the “empirical generalization”

codi¬ed in a function-in-extension or in a curve, however accurate, but also at

the mathematical identi¬cation of this curve. In practice, these two aims are

pursued in tandem. Their interplay is not dealt with in traditional accounts of

induction, even though its role is very real. For instance, if the mathematical

form of the dependence-codifying function is known, an inductive inference

reduces to the task of estimating the parameters characterizing the function

in question. This explains the prevalence of such estimation in actual scienti¬c

inquiry.

In another kind of case, the task of identifying the mathematical function

in question has already been accomplished within the limits of observational

accuracy for several intervals of argument values. Their induction becomes

the task of combining several partial generalizations (and reconciling them as

special cases of a wider generalization). This kind of induction turns out to

have been the dominating sense of inductio and epagoge in earlier discussions,

including the use of such terms by Aristotle and by Newton. (See Hintikka

1993.)

Thus, conclusiveness conditions are seen to play a pivotal role in the epis-

temology of questioning. They are also a key to the logic of knowledge. They

express wh-knowledge (knowing who, what, where, etc.) as distinguished from

knowing that, and show how the former construction can be expressed in terms

of the latter. However, from this expressibility it does not follow that the truth

conditions of expressions such as knowing who also reduce to those govern-

ing knowing that. They do not. The underlying reason is that the measuring

of quanti¬ers depends on the criteria of identi¬cation between different epis-

temically relevant scenarios (possible worlds, possible occasions of use) as

distinguished from criteria of reference. For this reason, we have to distin-

guish an identi¬cation system from a reference system in the full semantics of

any one language, be it a formal language or our actual working language”

called by Tarski “colloquial language.” I have argued for the vital importance

of this distinction in numerous essays, some of which are reprinted in Hintikka

(1999).

The unavoidability of this distinction is highlighted by the intriguing fact that

in our actual logico-linguistic practice, we are using two different identi¬cation

systems in a partnership with one and the same reference system all the time.

This dichotomy means a dichotomy between two kinds of quanti¬ers, public

and perspectival ones.

This dichotomy and its expressions in formal and natural languages have

been explained in my earlier papers. However, what has not been fully spelled

Introduction 7

out is the even more intriguing fact that the two identi¬cation systems are

manifested neuroscienti¬cally as two cognitive systems. This insight is spelled

out and discussed in the essay (written jointly with John Symons, Chapter 6

of this volume) entitled “Systems of Visual Identi¬cation and Neuroscience:

Lessons from Epistemic Logic” in the case of visual cognition. These two

systems are sometimes known as the what system and the where system. It

is known from neuroscience that they are different not only functionally but

anatomically. They are implemented in two different areas of the brain with