On What We Know
We Dont Know
How Questions Shape Them
On What We Know
We Dont Know
How Questions Shape Them
The University ofChicago Press
Chicago and London
Center for the Study ofLanguage and Information
The University of Chicago Press, Chicago 60637The University of Chicago Press, Ltd., London
c 1992 by the Center for the Study of Language and InformationLeland Stanford Junior UniversityAll rights reserved. Published 1992
Printed in the United States of America99 98 97 96 95 94 93 92 5 4 3 2 1
CIP data and other information appear at the end of the book
To the memory of
Aristides de Sousa Mendes
Portuguese Consul in Bordeaux in June 1940
1 An Approach to Explanation 18
2 A Theory about the Theory of Theory
and about the Theory of Theories 52
3 Why-Questions 75
4 Questions 101
5 Science and the Forms of Ignorance 112
6 Rational Ignorance 128
7 What We Dont Know When We Dont Know Why 145
8 Types and Tokens in Linguistics 170
9 The Ontology of Phonology 209
with Morris Halle
The nine papers in this book cover topics as diverse as the natureof explanation, the ambiguity of the word theory , the varieties of ig-norance, the limits of rationality in the choice of questions, and theontology of linguistics. Each is self-contained and can be read inde-pendently of the others. They are arranged essentially unchangedin the order in which they chronologically appeared. The book as awhole can thus be approached simply as a collection of papers thatwere previously scattered across many publications and are now as-sembled under one cover for convenience.
However, the book can also be approached as a collection of pa-pers that belong naturally together by virtue of a common idea anda common agenda based on that idea. Philosophers have alwaysrecognized that scientific progress consists in more than successiveadditions to a stock of truths and highly confirmed propositions.But what that more comes to has turned out to be surprisinglycontroversial and elusive. The idea common to these papers is thatthe more will be found in the truism that science seeks answersto questions. The common agenda then is simply to examine withsome care what such seeking involves.
We find ourselves, as individuals and as communities, willy-nillycast in a world not of our own making, in which we want to survive,if possible to thrive, and whose features we want to understand.We start out with little prior information about that world, but weare endowed with the ability to come to know that there are thingsabout it that we dont know, that is, with the ability to formulateand to entertain questions whose answers we know we do not know.It is an enormously complex ability derived from many auxiliary
2 On What We Know We Dont Know
abilities. And it induces the wish to know the answer to some ofthese questions. Scientific research represents our most reasonableand responsible way of trying to satisfy that wish. That is its mosttenable defining goal, and not, as others have held, the constructionof doctrines that can be recast as interpreted formal systems, or theachievement of intellectual economy, or the refutation of conjectures,or the solution of puzzles, or the provision of means for practical suc-cess, or the contrivance of communal consensus, or the dominationof certain institutions, or the elaboration of ever more encompass-ing systems, or the making of worlds. Those sorts of achievementsrepresent at best accidental objectives.
However, in seeking its goal science repeatedly runs into difficul-ties. Many of these difficulties are physical in nature and call forthe design of new and more powerful instruments. Others are psy-chological and call for the invention of devices that supplement ourmemory and our computational powers. Still others, and those arethe ones that are relevant here, are intellectual and pertain to ourability to conceive, formulate, consider, connect, and assess ques-tions, and to our ability to conceive, formulate, consider, connect,and assess answers. These sorts of difficulties often call for inspira-tion and creative intelligence. Careful observation and descriptionare not enough.
The idea linking these papers can thus be put more explicitly.It is that this more that scientific endeavors add to truths andwarranted beliefs consists of insights through which inspiration andcreative intelligence manage to overcome such intellectual difficulties.And the agenda is to specify the character of these insights.
In the rest of this introduction I will briefly describe the contentof each paper. I will first describe the specific topic to which it isaddressed for readers who look at it simply as a self-contained paper.I will then describe how it fits in the overall agenda when that is notobvious from the nature of the topic.
An Approach to Explanation belongs to the vast philosophicliterature that seeks to analyze the notion of explanation. What is atstake can best be appreciated by noting an odd disagreement amongphilosophers of science. Some, Pierre Duhem most notably amongthem, hold that empirical science does not and cannot provide expla-nations. Others, and that includes most contemporary philosophersof science, hold that, on the contrary, science can and does provide
explanations. All sides agree that the notion of explanation connotesnorms and desiderata, and all sides also agree that the question iswhether those norms and desiderata are compatible with the primarypositivistic demands for empirical evidence, inductive justification,consistency, generality, productivity, etc. The sides differ on whatthe connoted norms and desiderata come to, and so they reach differ-ent conclusions about their legitimacy. What is at stake, then, goesbeyond what the notion of explanation (or the word that expressesit) connotes and affects our consciousness of scientific desiderata andnorms, and our awareness of what science can or cannot aspire toachieve.
I will not summarize my own analysis of these connotations here,since the whole paper is near at hand, but I can prepare the groundwith a few remarks about its general character.
The paper was written under the influence of so-called ordinarylanguage philosophy. It therefore focuses unashamedly on our useof the verb to explain and of its nominalization explanation on theobvious ground that the most direct way to find out what the notionof explanation connotes is to examine how we apply the words explainand explanation.
The analysis starts from the observation of two facts: first, thepurely grammatical fact that the verb to explain and its cognate ex-planation of admit interrogative sentences (in indirect speech form,i.e., without so-called wh-aux inversion) as their complements; sec-ond, the nongrammatical fact that though they admit a wide rangeof questions as complement, they normally exclude some. They ad-mit most why-questions, how-questions, what-is questions as theirobject, but they do not normally admit, e.g., what time it is. (Wenormally tell but dont explain what time it is.)
The first of these facts, incidentally, is probably responsible forthe vast diversity of analyses that occur in the philosophic literature.Most writers on the subject implicitly limit their attention to somesubfamily of the family of questions admissible as object of explain,and limit themselves to different ones. So Duhem, for instance, limitshimself to the what is the physical structure underlying such andsuch phenomena subfamily; Mill, Hempel and other adherents ofthe covering law view limit themselves to the why subfamily; otherslimit themselves to the what causes subfamily; others still to the howpossible subfamily, or to the what mechanisms are present when so
4 On What We Know We Dont Know
and so occurs subfamily, and so on. They are like the blind men whoeach reported (perhaps correctly) on a different part of the elephant.But unlike the blind men, they follow a reasonable strategy, if oneassumes, as most of them seem to do, that their object should be todisplay truth-conditions distinctive of answers admitted by explainand its cognates. That cannot be done in one fell swoop for such aheterogenous family. It is therefore more reasonable to concentrateon some particularly challenging subfamily. Of course that does notjustify the widespread attitude that only one of these subfamilies islegitimate.
The analysis I propose in An Approach to Explanation shiftsthe focus to what is distinctive about the questions admissible asobjects of to explain and away from what is distinctive, if anything,about the truth-conditions on their answers. I argue that the distinc-tion is fundamentally cognitive in nature, that is, that it has to dowith intellectual states and not with subject matter. Explanations,I argue, express answers to not just any questions but to questionsthat present the kind of intellectual difficulty I call p-predicaments.(I also write of b-predicaments there, but that is immaterial forthis introduction.) A p-predicament, to a first approximation, is thecondition one is in when one is able to entertain a question, has rea-son to think that it is sound, i.e., that it has a right answer, but onecan think only of answers that one knows (or believes) to be false.
Though being in a p-predicament over some question is practi-cally a chronic condition of philosophers, they have seldom acknowl-edged its i