Nicholas Rescher - Epistemetrics (2006)

P1: JZZ
0521861209pre CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:42
This page intentionally left blank
ii
P1: JZZ
Epistemetrics
Epistemetrics is a scholarly discipline waiting to be born. With regard

to scientific information there is the discipline of scientometrics, rep-
resented by a journal of that name. Science, however, does not have
a monopoly on knowledge; although it is one of our most impor-
tant cognitive resources, it is not our only one. While scientometrics
is a centerpiece of epistemetrics, it is far from being the whole of
it. Nicholas Rescher’s endeavor to quantify knowledge is not only of
interest in itself, but it is also instructive in bringing into sharper relief
the nature of and the explanatory rationale for the limits that unavoid-
ably confront our efforts to advance the frontiers of knowledge. In
pursuing this objective, Rescher’s book takes the measure of both the
vast extent and the ultimate limitations of human knowledge.
Nicholas Rescher is University Professor of Philosophy at the Uni-

versity of Pittsburgh, where he is also Chairman of the Center for
Philosophy of Science. He has received fellowships from the Ford,
Guggenheim, and National Science Foundations. Author of more
than one hundred books ranging over many areas of philosophy, he
is the recipient of six honorary degrees from universities on three
continents and was awarded the Alexander von Humboldt Prize for
Humanistic Scholarship in 1984.
i
P1: JZZ
ii
P1: JZZ
Epistemetrics
NICHOLAS RESCHER
University of Pittsburgh
iii
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo
Cambridge University Press

The Edinburgh Building, Cambridge cb2 2ru, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521861205
© Nicholas Rescher 2006
This publication is in copyright. Subject to statutory exception and to the provision of

relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
First published in print format
isbn-13 978-0-511-14650-3 eBook (EBL)

isbn-10 0-511-14650-7 eBook (EBL)
isbn-13 978-0-521-86120-5 hardback

isbn-10 0-521-86120-9 hardback
Cambridge University Press has no responsibility for the persistence or accuracy of urls
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
P1: JZZ
To James R. Wible
v
P1: JZZ
vi
P1: JZZ
Contents
Preface page ix
1 Asking for More Than Truth: Duhem’s Law of Cognitive

Complementarity 1
The Security/Definiteness Trade-off and the Contrast Between
Science and Common Sense 1
Science versus Common Sense 3
Further Ramifications 5
Knowledge in Perspective 6
2 Kant’s Conception of Knowledge as Systematized
Information 9
Distinguishing Knowledge and Information 9
Kant on the Systematicity of Knowledge 10
The Hierarchical Textualization of Knowledge 13
3 Spencer’s Law of Cognitive Development 15
Spencer’s Law: The Dynamics of Cognitive
Complexity 15
How Cognitive Taxonomy Has Grown More
Complex 18
Evolving Complexity 24
A Quantitative Perspective 26
4 Gibbon’s Law of Logarithmic Returns 29
Kant’s Principle of Questions Propagation 29
Edward Gibbon and Logarithmic Returns 32
Diminishing Returns and Logarithmic Retardation 39
Planck’s Principle 42
vii
P1: JZZ
viii Contents
5 Adams’s Thesis on Exponential Growth 45

Scientific Progress 45
Adams’s Thesis 46
Dimensions of the Phenomenon 48
The Linear Growth of Knowledge 51
The Lesson 54
Bibliographic Appendix 54
6 Quality Retardation 57
The Centrality of Importance 57
Levels of Cognitive Importance 59
Quality Retardation 62
Elites 64
Cognitive Importance as Reflected in Space Allocation 66
Cognitive Importance as Reflected in Citation 68
7 How Much Can Be Known? A Leibnizian Perspective
on the Quantitative Discrepancy Between Linguistic
Truth and Objective Fact 73
How Much Can a Person Know? Leibniz on Language
Combinatorics 73
The Leibnizian Perspective 77
Statements Are Enumerable, As Are Truths 80
Truths versus Facts 81
The Inexhaustibility of Fact 81
Facts Are Transdenumerable 83
More Facts Than Truths 86
Noninstantiable Properties and Vagrant Predicates 87
Musical Chairs Once More 90
Coda: Against Cognitive Nominalism 92
Appendix: Further Implications 93
8 On the Limits of Knowledge: A Kantian Perspective
on Cognitive Finitude 95
Limits of Knowledge 95
Cognitive Finitude 97
Surd Facts and Unknowability 99
Larger Lessons: Isaiah’s Law 101
Conclusion 105
Bibliography 107
Index of Names 111
P1: JZZ
Preface
When you can measure what you are speaking about, and express it in
numbers, you know something about it; but when you cannot measure
it, when you cannot express it in numbers, your knowledge is of a meager
and unsatisfactory kind: it may be the beginning of knowledge, but you
have scarcely, in your thoughts, advanced to the stage of science.
– William Thomson, Lord Kelvin (1824–1907), English Physicist
This book develops the theory of knowledge from a quantitative per-

spective that serves to throw light on the scope and limits of human
knowledge. It seeks to provide theorists of knowledge in philosophy,
information theory, cognitive studies, communication theory, and cog-
nate disciplines with the conceptual tools required for a quantitative
treatment of the products of inquiry.
Kelvin’s dictum is an exaggeration that takes things too far. I have
never thought for a moment that if you cannot say it with numbers that
it just is not worth saying. But all the same, I do firmly believe that where
you cannot put numbers to work you will understand the matter better
and more clearly for being able to explain why. So it seems well worth-
while to see what a qualitative approach to knowledge can do for us.
The discipline represented by the domain of inquiry to which the
present book is addressed does not as yet exist. Epistemetrics is not yet a
scholarly specialty. To be sure, as regards scientific information in spe-
cific there is the discipline of scientometrics, represented by a journal
of that name. But while this book too will keep scientific knowledge in
ix
P1: JZZ
x Preface
the foreground, various of its key principles – for example Duhem’s

Law of Chapter 1, or the Principle of Quality Retardation of Chapter 6 –
hold every bit as much for our knowledge of everyday-life matters as
they do for the natural and human sciences. After all, science does
not have a monopoly on knowledge: while it is doubtless our most
important cognitive resource it is not our only one. And so while sci-
entometrics is a centerpiece of epistemetrics, it is not the whole of it.1
Again the measurement of intelligence is a large and flourishing
industry. But intelligence is no more than the capacity for producing
and handling knowledge and does not as such address the product
itself. Then too there are all sorts of tests and quiz shows and even
games (on the order of Trivial Pursuit) that compare the knowledge
of different individuals. But of course such comparisons do not address
the epistemetric issue of cognitive measurement at large. And so, see-
ing that epistemetrics is not as yet an established field of inquiry, the
present discussion can do no more than offer a preliminary glimpse
into the nature of such a discipline.
Autobiographically speaking, I now believe that my brief time of
service at the RAND Corporation some 50 years ago convinced me
of the power and utility of the quantitative point of view. And I am
persuaded that it affords us the basis for a deeper understanding of
the nature and prospects of the processes at issue with the accession
and development of knowledge.
To be sure, many factors conspire to make the measurement of
knowledge into a vexed and complicated enterprise. A head count of
particular items of information is of course pointless because one item
can informatively encompass many others. Thus someone who knows
a general truth is in a position to infer all of its potentially innumerable
instances. One generalization can encompass a zillion particularities.
And this points to the further issue that information is not created
equal – some items deal with large and important matters, some with
trivialities.
Given such complications it may seem unexpected that any-
thing instructive can be accomplished in the epistemetric project of
quantifying knowledge. But something can indeed be done, and the
1 My personal effort in the domain of scientometrics is represented by my Scientific

Progress (Oxford: Blackwell, 1978). It will be obvious to anyone who knows that book
that the present work is heavily indebted to it.
P1: JZZ
Preface xi
present discussion will endeavor to synthesize and coordinate some of

the useful suggestions that have been offered along these lines over
the years.
Rather surprisingly, epistemology, that is, the theory of knowledge,
has generally not made much of the distinction between various modes
of cognitively available material – between reasonable conjecture and
plausible supposition, for example, or between reliable information
and actual knowledge. But it is this latter distinction – between mere
information and genuine knowledge – that is pivotal for the present
deliberations, which will seek to address the qualitative issue of knowl-
edge versus information in a quantitative perspective.
The book’s venture into the quantification of knowledge will pro-
ceed by a series of principles which, for reasons of mnemonic vividness
can be indicated by association with various scientists and philoso-
phers. Sometimes their work suggests rather than formulates what is
at issue. But no matter! For what matters more than historical piety
is that the pieces all fit together to make up a coherent and cohesive
story regarding the limitations of inquiry and the limits of knowledge.
Ironically, the business of cognitive quantification is so problem-
atic that one could say with some justice that the present treatment
of quantities is qualitative in nature, and that rather than providing
a properly quantitative assessment of our knowledge it only offers a
qualitative perspective on it. But even if that were so, such an exer-
cise is instructive in bringing into sharper relief the nature of and
the explanatory rationale for the limits that unavoidably confront our
ongoing efforts to advance the frontiers of knowledge. And the quan-
titative perspective developed here – in however rough and sketchy a
way – affords some instructive insights into the nature of cognition that
no merely qualitative deliberations would make available. In particu-
lar, the project manages to throw some vivid light on the unavoidable
limitations of human knowledge.
Erik Angner and C. J. Thomas read my manuscript and spotted
various corrigenda. And I am grateful to Estelle Burris for her highly
competent assistance in seeing the book through the process of getting
into print.
Pittsburgh, Pennsylvania
May 2005
P1: JZZ
xii
P1: KAE
0521861209c01 CUNY338B/Rescher 0 521 83164 2 December 7, 2005 12:9
Asking for More Than Truth

Duhem’s Law of Cognitive Complementarity
(1) Duhem’s Law of cognitive complementarity holds that inquiry is subject to

a complementary relationship between security and confidence on the one hand,
and definiteness and detail on the other, so that s × d ≤ const. (2) Among other
things, this relationship serves to characterize the difference between science and
common sense, seeing that these two domains take a very different stance regarding
security and definiteness. (3) Duhem’s Law engenders an impetus to vagueness in
matters where truth is paramount. (4) Moreover, security/detail complementarity
has important lessons for the conduct of inquiry, and in particular means that
knowledge is more than correct information as such.
The Security/Definiteness Trade-off and the Contrast

Between Science and Common Sense
It is a basic principle of epistemology that increased confidence in
the correctness of our estimates can always be secured at the price of
decreased accuracy. For in general an inverse relationship obtains between
the definiteness or precision of our information and its substantiation: detail
and security stand in a competing relationship. We estimate the height of
the tree at around 25 feet. We are quite sure that the tree is 25 ± 5 feet
high. We are virtually certain that its height is 25 ± 10 feet. But we can
be completely and absolutely sure that its height is between 1 inch and 100
yards. Of this we are “completely sure” in the sense that we are “abso-
lutely certain,” “certain beyond the shadow of a doubt,” “as certain as
we can be of anything in the world,” “so sure that I would be willing to
stake my life on it,” and the like. For any sort of estimate whatsoever
1
P1: KAE
2 Duhem’s Law of Cognitive Complementarity
increasing
s × d = c (constant)
security
( s)
increasing
detail
(d )
display 1.1. Duhem’s Law: the complementarity trade-off between security
and definiteness in estimation. Note: The shaded region inside the curve repre-
sents the parametric range of achievable information, with the curve indicating
the limit of what is realizable. The concurrent achievement of great detail and
security is impracticable.
there is always a characteristic trade-off relationship between the evi-

dential security of the estimate, on the one hand (as determinable on
the basis of its probability or degree of acceptability), and on the other
hand its contentual detail (definiteness, exactness, precision, etc.).
And so a complementarity relationship of the sort depicted in
Display 1.1 obtains. This was adumbrated in the ideas of the French
physicist Pierre Maurice Duhem (1861–1916) and may accordingly be
called “Duhem’s Law.”1 In his classic work on the aim and structure of
physical theory,2 Duhem wrote as follows:
A law of physics possesses a certainty much less immediate and much more
difficult to estimate than a law of common sense, but it surpasses the latter by
1 It is both common and convenient in matters of learning and science to treat ideas
and principles eponymously. An eponym, however, is a person for whom something is
named, and not necessarily after whom this is done, seeing that eponyms can certainly
be honorific as well as genetic. Here at any rate eponyms are sometimes used to make
the point that the work of the person at issue has suggested rather than originated the
idea or principle at issue.
2 La théorie physique: son objet, et sa structure (Paris: Chevalier and Rivière, 1906); tr.
by Philip P. Wiener, The Aim and Structure of Physical Theory (Princeton: Princeton
University Press, 1954). This principle did not elude Neils Bohr himself, the father of
complementarity theory in physics: “In later years Bohr emphasized the importance of
complementarity for matters far removed from physics. There is a story that Bohr was
once asked in German what is the quality that is complementary to truth (Wahrheit).
After some thought he answered clarity (Klarheit).” Steven Weinberg, Dreams of a Final
Theory (New York: Pantheon Books, 1992), p. 74 footnote 10.
P1: KAE
Science versus Common Sense 3
the minute and detailed precision of its predictions. . . . The laws of physics can
acquire this minuteness of detail only by sacrificing something of the fixed and
absolute certainty of common-sense laws. There is a sort of teeter-totter of balance
between precision and certainty: one cannot be increased except to the detriment of the
other.3
In effect, these two factors – security and detail – stand in a relation of

inverse proportionality, as shown in the picture of Display 1.1.
Note that the relationship at issue envisions the boundary of realiz-
able information as set by a curve of the form x × y = c, or equivalently
y = c /x. Accordingly, the sum total of the area of accessibility lying

under this curve is given by y d x = c d x/x ≈ log x. On this basis
the overall size of the body of high-quality information that combines
security and definiteness to an acceptable extent is given by a logarith-
mic measure. It will be useful to bear this in mind as the discussion
proceeds.
Science versus Common Sense

Duhem emphasized that this relationship has important implications
for the standing of the exact sciences where, after all, we always aim
at the maximum of achievable universality, precision, and exactness.
Thus in physics when we make the assertion, “The melting point of
lead is 327.7 degrees Celsius,” we are claiming that all pieces of (pure)
lead will unfailingly melt at exactly this temperature. We certainly do
not mean to assert that most pieces of (pure) lead will probably melt at
somewhere around this temperature. (And in this regard, there would be
a potential problem, should it turn out, for example, that there is no
melting point at all and that what is actually at issue is the center of a
statistical distribution.) In aspiration always and in practice generally,
the theoretical claims of science involve no hedging, no fuzziness,
no incompleteness, and no exceptions; they are strict: precise, wholly
explicit, exceptionless, and unshaded. Here we operate at the lower
right-hand side of the Display 1.1 curve.
After all, in intent and in aspiration science aims to characterize
nature as it really is. And since (as we certainly believe) nature is fully
definite and detailed, science endeavors to infuse these characteristics
3 Duhem, La théorie physique, pp. 178–79. Italics supplied.

P1: KAE
into the claims it stakes regarding the world. It scorns the very idea of
claiming that matters stand roughly thus-wise or that things function
something like such-and-such. Unlike everyday-life communication,
the exact sciences stand committed not just to truth but to accuracy
and exactness as well. And this, their seeming strength, is their Achilles’
heel as well.
By contrast, the situation of ordinary life is very different; when
we assert that “peaches are delicious” we are maintaining something
like “most people will find the eating of suitably grown and duly
matured peaches a rather pleasurable experience.” Such a statement
has all sorts of built-in safeguards on the order of “more or less,” “in
ordinary circumstances,” “by and large,” “normally,” “when all things
are equal,” “rather plausible,” and so on. They are not really laws in
the usual sense, but rules of thumb, a matter of practical lore rather
than scientific rigor. But this enables them to achieve great security.
For there is safety in vagueness: a factual claim can always acquire secu-
rity through inexactness. Take “there are rocks in the world” or “dogs
can bark.” It is virtually absurd to characterize such everyday life cont-
entions as fallible: their security lies in their very indefiniteness and
imprecision.
And there is good reason for adopting this resort to vagueness in
everyday life, for protecting our claims to reliability and trustworthi-
ness becomes crucial in personal interactions. We proceed in cognitive
matters in much the same way that lenders such as banks proceed in
financial matters. We extend credit to others, doing so at first to only
a relatively modest extent. When and if they comport themselves in a
manner that shows that this credit was well deserved and warranted,
we proceed to give them more credit and extend their credit limit.
By responding to trust in a responsible way, one improves one’s credit
rating in cognitive contexts much as in financial contexts. The same
sort of mechanism is at work on both sides of the analogy: creditworthy
comportment engenders a reputation on which further credit can be
based; earned credit is like money in the bank, well worth the measure
needed for its maintenance and for preserving the good name that
is now at stake. Thus we constantly rely on experts in a plethora of
situations, continually placing reliance on doctors, lawyers, architects,
and other professionals. But they, too, must so perform as to estab-
lish credit, not just as individuals but, even more crucially, for their
P1: KAE
Further Ramifications 5
profession as a whole.4 And much the same sort of thing holds for
other sources of information. (The example of our senses is a partic-
ularly important case in point.) In everyday life, in sum, we prioritize
correctness over accuracy.
However, while everyday-life common sense trades definiteness for
security, science does the very reverse, with the result that its claims
become subject to greater insecurity. As Duhem put it:
A law of physics is always provisional and relative. It is provisional also in that

it does not connect realities but symbols, and that is because there are always
cases where the symbol no longer corresponds to reality; the laws of physics can-
not be maintained except by continual retouching and modification. . . . One
might be tempted to draw the strange conclusion that the knowledge of the
laws of physics constitutes a degree of knowledge inferior to the simple knowl-
edge of the laws of common sense.
Science decidedly prioritizes accuracy and detail over security. So

as Duhem himself maintained, his principle both characterizes and
explains the profound differences between the nature of our knowl-
edge in science and in the matters of everyday life.
Further Ramifications
Duhem’s Law of Security/Detail Complementarity has substantial
implications for the modus operandi of inquiry. Thus one of its fun-
damental implications is represented by the following observation:
THESIS 1: Insofar as our thinking is vague, truth is accessible even in the face of error.
Consider the situation in which you correctly accept P-or-Q. But –

so let it be supposed – the truth of this disjunction is entirely rooted
in P, while Q is flatly false. However, you accept P-or-Q only because
you are mistakenly convinced of the truth of Q , while it so happens
that P is something you actually disbelieve. Nevertheless, despite your
error, your belief is entirely true.5 Consider a concrete instance. You
4 Compare H. M. Vollmer and D. L. Mills, eds., Professionalization (Englewood Cliffs, NJ:

Prentice Hall, 1966). This credit, once earned, is generally safeguarded and main-
tained by institutional means: licensing procedures, training qualifications, profes-
sional societies, codes of professional practice, and the like.
5 Examples of this sort indicate why philosophers are unwilling to identify knowledge
with true belief, even where belief is justified.
P1: KAE
believe that Smith bought some furniture because he bought a table.

However it was, in fact, a chair that he bought, something you would
flatly reject because you believe he bought a table. All the same, your
belief that he bought some furniture is unquestionably correct. The
error in which you are involved, although very real, is yet not so grave
as to destabilize the truth of your belief.
Ignorance is a matter of inability to answer questions properly. But
one has to be careful in this regard. Answering a question informatively
is not just a matter of providing a correct answer but also a matter
of offering an exact answer. Thus consider the question “What is the
population of Shanghai?” If I respond “More than ten and less than
ten billion” I have provided a correct answer, albeit one that is not
particularly helpful.
So the irony of it is that insofar as our ignorance of relevant matters
leads us to be vague in our judgments, we nevertheless may well man-
age to enhance the likelihood of being right. The fact of the matter is
that we have this:
THESIS 2: By constraining us to make vaguer judgments, ignorance enhances our

access to correct information (albeit at the cost of less detail and precision).
For example, if I have forgotten that Seattle is in Washington State,

then if “forced to guess” I might well erroneously locate it in Ore-
gon. Nevertheless, my vague judgment that “Seattle is located in the
Northwestern United States” is quite correct.
This state of affairs means that when the truth of our claims is critical
we generally “play it safe” and make our commitments less definite and
detailed. And in practical matters in particular, such rough guidance is
often altogether enough. We need not know precisely how much rain
there will be to make it sensible for us to take an umbrella. Nevertheless
in those matters where exactness counts this pathway to truth is rather
problematic.
Knowledge in Perspective
Duhem’s Law of Cognitive Complementarity means that it is going to
be a fact of life in the general theory of estimation that the harder
we push for certainty – for security of our claims – the vaguer we will
have to make these claims and the more general and imprecise they
P1: KAE
Knowledge in Perspective 7
will become. And so if we want our scientific claims to have realistic

import – taking them to provide an account of how matters actually
stand – we have to reconstrue them loosely. Take the atomic theory.
We should not – cannot – say that atoms are in every detail as the sci-
ence of the day holds them to be: that the “Atomic Theory” section
of our Handbook of Physics succeeds in every jot and tittle in character-
izing reality as it actually is. But if we “fuzz things up” – if we claim
merely that physical reality is granular and that atoms exist and have
roughly such-and-such features – then what we say is no longer subject
to (reasonable) doubt.
The complementarity of security and detail accordingly carries
important lessons for the realization of knowledge. For one thing,
seeing that informativeness is a pivotal factor here, it means that knowl-
edge calls for more than mere correctness. And, for another, it means
that knowledge is something difficult, something we do not achieve
all that easily. For on its basis, knowledge qualifies as such not only
through its claims to truth but also through its informativeness. The
fact that quality is going to be a key factor here means that the quantity
of information cannot be equated with the quantity of knowledge.
In the pursuit of knowledge we seek and demand more than the
mere truth about things. For truth comes to us too cheap and easy
when it is secured at the price of uninformativeness. Knowledge does
not issue from trivial truth, it must contribute to our understanding
of things. And so, both security and informativeness figure among the
essentials for knowledge. But it lies in the nature of our human situa-
tion as finite inquiring beings that only so much can be accomplished
along these lines. Duhem’s cognitive complementarity law constrains
us to make choices: we cannot “have it both ways.” And just these con-
siderations lead to a question that will set the theme for the rest of the
book: Just what are the ramifications and implications that such limits
pose for the development of knowledge?
P1: KAE
8
P1: KAE
Kant’s Conception of Knowledge as

Systematized Information
(1) Knowledge is not just a matter of information as such, but of information that
is coherently and cohesively systematized. (2) This view of knowledge as properly
systematized information – in effect, information as structured in an idealized
expository treatise – goes back to Immanuel Kant. (3) Cognitive systematization
is hierarchical in structure because a systemic organization of the exposition of
the information at issue into successively subordinate units becomes paramount
here. And, viewed in this light, structure will of course reflect significance with
larger units dominating over subordinate ones.
Distinguishing Knowledge and Information

The interplay between knowledge and information is pivotal for the
present deliberations. Actual information (in contrast with misinfor-
mation) requires little more than truth. But knowledge is something far
more demanding: it calls for information that is organized, purified,
systematized. It makes no sense to say “It is known that p, but it may
possibly not be so (or . . . “there are considerations that lead to doubt
about it”).” From the cognitive point of view, knowledge is money in
the bank. It must fit together coherently. The very concept of knowl-
edge is such that what is known must be systemically consolidated: the
matter of quality will also play a crucial role. For items of information
are not created equal. Some are minute and trivial, others large and
portentous. So there is little point to merely doing a nose count here.
Only information that is scrutinized, verified, coordinated, and sys-
tematized can plausibly qualify to be regarded as knowledge. Whatever
9
P1: KAE
10 Knowledge as Systematized Information
else it is, knowledge is information of substantial quality. And this leads

to the question: How is knowledge related to information in strictly
quantitative terms?
Since factual contentions are formulated symbolically, the quantita-
tive assessment of raw information can be made by measuring exposi-
tory text – that is, by looking to the amount of text expended in stating
the matter. Information can thus be assessed – in a first approximation
at least – in terms of sheer textuality, subject to the idea that the ampler
and fuller its exposition, the more information this text account con-
veys.
Knowledge, on the other hand, is something very different, and
mere information – unreconstructed textuality – does not do what is
needed there. For knowledge does not consist in information as such
but only in appropriately systematized information. And accordingly,
knowledge is not a matter of the extent of text actually devoted to
the issue, but rather requires an assessment of systemic enmeshment.
To view knowledge in textual terms would require a radically different
approach, one that looks not merely to sheer quantity of text but rather
to the textual role of the information at issue.
But just how is this idea of knowledge and its systematization to be
implemented?
Kant on the Systematicity of Knowledge

In the eighteenth century Immanuel Kant (1724–1804) had elo-
quently argued that the mission of rational inquiry is the systemic
organization of knowledge: its coordination into one coherent struc-
ture under the guiding aegis of unifying principles.
If we consider in its whole range the knowledge obtained for us by the under-
standing, we find that what is peculiarly distinctive of reason in its attitude
to this body of knowledge, is that it prescribes and seeks to achieve its sys-
tematization, that is, to exhibit the connection of its parts in conformity with
a single principle. . . . This unity of reason always presupposes an idea, (or
plan), namely, that of the form of a whole of knowledge – a whole which is
prior to the determinate knowledge of the parts and which contains the con-
ditions that determine a priori for every part its position and relation to the
other parts. . . . This idea accordingly demands a complete [organic] unity in
the knowledge obtained by understanding by which this knowledge is to be
P1: KAE
Kant on the Systematicity of Knowledge 11
not a mere contingent aggregate, but a system connected according to nec-

essary laws. We may not say that this idea is a concept of the object, but only
of the thoroughgoing unity of such concepts, in so far as that unity serves
as a rule for the understanding. These concepts or reason are not derived
from nature; on the contrary, we interrogate nature in accordance with these
ideas, and consider our knowledge as defective so long as it is not adequate
to them. (Critique of Pure Reason, A645 = B673)
The paradigm of system that lay before Kant’s eyes was that of sci-
ence – of Euclid’s systematization of Geometry, Archimedes’ system-
atization of statics, and Newton’s systematization of celestial mechan-
ics. And his model of rational systematization was that exemplified
in the work of the great seventeenth-century rationalist philosophers:
Descartes, Spinoza, and also Leibniz as expounded by the subsequent
members of his school, especially Christian Wolff.1
As Kant saw it, adequate understanding can be achieved only
through the systemic interrelating of facts. The mission of human
reason is to furnish a basis for the rational comprehension of what
we know and this can be accomplished only by positioning these facts
as integral parts of an organic whole. Kant developed his biological
analogy of system in the following terms:
[O]nly after we have spent much time in the collection of materials in some-
what random fashion at the suggestion of an idea lying hidden in our minds,
and after we have, indeed, over a long period assembled the materials in a
merely technical manner, does it first become possible for us to discern the
idea in a clearer light, and to devise a whole architectonically in accordance
with the ends of reason. Systems seem to be formed in the manner of lowly
organisms, through a generatio aequivoca from the mere confluence of assem-
bled concepts, at first imperfect, and only gradually attaining to completeness,
although they one and all have had their schema, as a original germ, in the
sheer self-development of reason. Hence, not only is each system articulated
in accordance with an idea, but they are one and all organically united in a
system of human knowledge, as members of one whole.
(Critique of Pure Reason, A834 = B862)
1 Leibniz’s theory of cognitive systematization is detailed in the author’s essay entitled

“Leibniz and the Concept of a System” in his Leibniz’s Philosophy of Nature (Dordrecht:
D. Reidel, 1981), pp. 29–41. On the broader issues see the author’s Cognitive System-
atization (Oxford: Basil Blackwell, 1979).
P1: KAE
The nature of systematization was accordingly explained by Kant

along the lines of a fundamentally biological analogy:
In accordance with reason’s legislative prescriptions, our diverse modes of

knowledge must not be permitted to be a mere rhapsody, but must form a
system. Only so can they further the essential ends of reason. By a system I
understand the unity of the manifold modes of knowledge under one idea.
This idea is the concept, provided by reason, of the form of a whole . . . [which]
determines a priori not only the scope of its manifold content, but also the posi-
tions which the parts occupy relatively to one another. The scientific concept
of reason contains, therefore, the end and the form of that whole which is
congruent with this requirement. The unity of the end to which all the parts
relate and in the idea of which they all stand in relation to one another, makes
it possible for us to determine from our knowledge of the other parts whether
any part be missing, and to prevent any arbitrary addition, or in respect of its
completeness [to discover] any indeterminateness that does not conform to
the limits which are thus determined a priori. The whole is thus an organised
unity (articulatio), and not an aggregate (coacervatio). It may grow from within
(per intussusceptionem), but not by external addition (per appositionem). It is thus
like an animal body. (Critique of Pure Reason, A832–3 = B860–61)
In reasoning along these lines, Kant put us on the right tack regard-
ing the nature of cognitive systematization. Here completeness and
comprehensiveness become paramount desiderata, and a body of
information constitutes systemic knowledge when its articulation is
characterized by such organic unity. Thus as Kant sensibly saw it, only a
body of information coherently systematized on principles of organic
interlinkage can be regarded as constituting knowledge. Authentic
knowledge must form part of a coherently integrated system, and
every part of such a system must serve in the role of a contribu-
tory sub-system: an organ of the overall organism. And so, despite
various historical anticipations, this idea – that to qualify as authen-
tic, cognitively significant knowledge, informative contentions must be
part of a system – deserves to be called Kant’s Principle of Cognitive
Systematicity thanks to the prominence and centrality that he gave to
this idea in insisting that knowledge, in its qualitative and honorific
sense, is a matter of the extent to which information is coherently
systematized.
P1: KAE
The Hierarchical Textualization of Knowledge 13
10 chapters per book

10 sections per chapter
10 paragraphs per section
10 sentences per paragraph
10 words per sentence
display 2.1. A Hypothetical Treatise.
The Hierarchical Textualization of Knowledge

Kant’s line of thought can be carried a step further to the idea that in
a well-designed systemic exposition of information, systemic role mirrors
cognitive status. In a deductive systematization such as that of Euclid, the
axioms have priority over the theorems.2 Here the order of exposition
reflects cognitive fundamentality and thereby dominance. However,
in a more discursive exposition it is not so much order as structure
that is determinative. Some ideas are chapter correlative and require
substantial development; others deserve a mere section – or paragraph
or sentence. Dominance of hierarchical placement in an analytical
table of contents reflects cognitive status showing that length and detail
of treatment come into correlation.
Now in the ideal case, the cognitive significance of themes and
topics can be assessed via the organizational status dedicated to their
treatment. And so the systemic structure of a book’s discussion domain
in terms of its constituent elites could be mapped out by dividing it
into successive units as per the hypothetical treatise of Display 2.1. And
this in turn intends a systemic structure with layers of components at
different hierarchical levels as per Display 2.2 (which for simplicity
depicts the situation based on 3 rather than 10).
In this light, the hierarchical exposition of information through
its exfoliation into successively subordinate units becomes paramount
for the quantitative assessment of knowledge. Some items of infor-
mation represent important ideas that are chapter-worthy; others are
2 Kant had in view not only Euclid’s Elements and Newton’s Principia but also the philo-
sophical systematizations of the scholastic and neo-scholastic eras of Aquinas, Scotus,
Suarez and Wolff. For further details on these matters see the author’s Cognitive Sys-
tematization (Oxford: Basil Blackwell, 1979).
P1: KAE
display 2.2. The systemic structure of a domain via units of successive expos-
itory subordination.
smaller scale, worthy of a sentence – or a paragraph at most. And this

sort of thing would (ideally) be reflected in the taxonomic depth of
an analytically well-designed table of contents. And, of course, this
book analogy could be replicated on a larger scale in such a sequence
as cognitive domain, discipline, substantive, specialty, problem-area,
problem, problem-component.
What we have here is a picture of nested aggregation with findings
at a given level of quality in systemic significance encompassing a mul-
titude of others at lower levels of quality. Quantity and quality are thus
brought into correlation. In an appropriately structured exposition,
the table of content status of ideas and theses is bound to reflect their
cognitive importance.
But just how does such an hierarchical expository architectonic
develop as we elaborate a cognitive system by pressing our efforts ever
further in the course of rational inquiry?
P1: KAE
Spencer’s Law of Cognitive Development
(1) Herbert Spencer’s “Law of Cognitive Development” through taxonomic exfo-

liation depicts an important phenomenon in the growth of cognition. (2) This
is amply illustrated by the history of science, which tells an ongoing story of tax-
onomic complexification. (3) Such increasing complexity is the price exacted by
scientific progress. (4) On the basis of Spencer’s Law it can be seen that the actual
knowledge inherent in the mass of information stands not as the volume thereof,
but merely as its logarithm.
Spencer’s Law: the Dynamics of Cognitive Complexity

Already well before Darwin, the English philosopher Herbert Spencer
(1820–1903) argued that organic evolution is characterized by Karl
Ernst von Baer’s (1792–1876) law of development “from the homoge-
neous to the heterogeneous” and thereby produces an ever-increasing
elaborateness of detail and complexity of articulation.1 As Spencer
saw it, organic species in the course of their development confront
a successive series of environmental obstacles, and with each success-
ful turning along the maze of developmental challenges the organ-
ism becomes selectively more highly specialized in its bio-design, and
thereby more tightly attuned to the particular features of its ecological
context.2
1 Herbert Spencer, First Principles, 7th ed. (London: Appleton’s, 1889); see sect. 14–17
of Part II, “The Law of Evolution.”
2 On the process in general see John H. Holland, Hidden Order: How Adaptation
Builds Complexity (Reading, MA: Addison Wesley, 1995). Regarding the specifically
15
P1: KAE
16 Spencer’s Law of Cognitive Development
Now this view of the developmental process may be of limited appli-

cability in biological evolution, but there can be little question about its
holding good in cognitive evolution. For rational beings will of course
try simple things first and thereafter be led step by step toward an
ever-enhanced complexification. In the course of rational inquiry we
try the simple solutions first, and only thereafter, if and when they
cease to work – when they are ruled out by further developments –
do we move on to the more complex. Matters go along smoothly until
an oversimple solution becomes destabilized by enlarged experience.
For a time we get by with the comparatively simpler options – until the
expanding information about the world’s modus operandi made possible
by enhanced new means of observation and experimentation insists
otherwise. And with the expansion of knowledge those new accessions
make ever-increasing demands.
And this state of affairs means that the course of empirical inquiry
moves historically in the direction of ever-increasing complexity. The
developmental tendency of our intellectual enterprises – natural sci-
ence preeminent among them – is generally in the direction of
greater complication and sophistication. In this way, cognitive evo-
lution in its various forms – whether of animal species or of literary
genres – ongoingly confronts us with products of greater and greater
complexity.3
In scientific inquiry, in particular, we look to the most econom-
ical theory accommodation for the amplest body of currently avail-
able experience. Induction – here short for “the scientific method”
in general – proceeds by constructing the most straightforward and
economical structures able to house the available data comfortably
while yet affording answers to our questions.4 Accordingly, economy
and simplicity serve as cardinal directives for inductive reasoning,
whose procedure is that of the precept: “Resolve your cognitive prob-
lems in the simplest, most economical way that is compatible with
evolutionary aspect of the process see Robert N. Brandon, Adaptation and Environment
(Princeton: Princeton University Press, 1990).
3 On the issues of this paragraph compare Stuart Kaufmann, At Home in the Universe:
To Search for the Laws of Self-Organization and Complexity (New York: Oxford University
Press, 1995).
4 For further details see the author’s Induction (Oxford: Basil Blackwell, 1980).
P1: KAE
The Dynamics of Cognitive Complexity 17
a sensible exploitation of the information at your disposal.” But we

always encounter limits here. Simple solutions take us only so far.
An inner tropism toward increasing complexity is thus built into the
very nature of the scientific project as we have it. And this circumstance
leads to what may be called Spencer’s Law of Cognitive Development:
As a body of information on any given topic grows in size the complexity of

that body – its inner variation and diversification – also increases.
What we have here is a thesis to the effect that cognitive progress

is accompanied by and can be measured in terms of the taxonomic
complexity of the information manifold at hand.
Our cognitive efforts manifest a Manichaean-style struggle between
complexity and simplicity – between the impetus to comprehensive-
ness (amplitude) and the impetus to system (economy). We want our
theories to be as extensive and all-encompassing as possible and at the
same time to be elegant and economical. The first desideratum pulls
in one direction, the second in the other. And the accommodation
reached here is never actually stable. As our experience expands in
the quest for greater adequacy and comprehensiveness, the old the-
ory structures become destabilized – the old theories no longer fit the
full range of available fact. And so the theoretician goes back to the
old drawing board. What he comes up with here is – and in the circum-
stances must be – something more elaborate, more complex than what
was able to do the job before those new complications arose (though we
do, of course, sometimes achieve local simplifications within an overall
global complexification). We make do with the simple, but only up to
the point when the demands of adequacy force additional complica-
tions upon us. Be it in cognitive or in practical matters, the processes
and resources of yesteryear are rarely, if ever, up to the demands of the
present. In consequence, the life-environment we create for ourselves
grows increasingly complex. The Ockham’s Razor injunction, “Never
introduce complications unless and until you actually require them,”
accordingly represents a defining principle of practical reason that is
at work within the cognitive project as well. And because we try the
simplest solutions first, making them do until circumstances force us
to do otherwise, it transpires that in the development of knowledge –
as elsewhere in the domain of human artifice – progress is always a
P1: KAE
matter of complexification. In fact inherent impetus toward greater

complexity pervades the entire realm of human creative effort. We
find it in art; we find it in technology; and we certainly find it in the
cognitive domain as well.5
How Cognitive Taxonomy Has Grown More Complex

It is instructive to examine somewhat more closely the role of com-
plexity in the domain of cognition, now focusing on science in partic-
ular. Progress in natural science is a matter of dialogue or debate in
a reciprocal interaction between theoreticians and experimentalists.
The experimentalists probe nature to discern its reactions, to seek
out phenomena. And the theoreticians take the resultant data and
weave about them a fabric of hypotheses that is able to resolve our
questions. Seeking to devise a framework of rational understanding,
they construct their explanatory models to accommodate the findings
that the experimentalists put at their disposal. Thereafter, once the
theoreticians have had their say, the ball returns to the experimental-
ists’ court. Employing new, more powerful means for probing nature,
they bring new phenomena to view, new data for accommodation.
Precisely because these data are new and inherently unpredictable on
the basis of earlier knowledge, they often fail to fit the old theories.
Theory extrapolations from the old data could not encompass them;
the old theories do not accommodate them. A disequilibrium thus
arises between available theory and novel data, and at this stage, the
ball reenters the theoreticians’ court. New theories must be devised to
accommodate the new, nonconforming data. Accordingly, the theo-
reticians set about weaving a new theoretical structure into which the
new data will fit. They endeavor to restore the equilibrium between
theory and data once more. And when they succeed, the ball returns
to the experimentalists’ court, and the whole process starts over again.
5 An interesting illustration of the extent to which lessons in the school of bitter expe-
rience have accustomed us to expect complexity is provided by the contrast between
the pairs: rudimentary/nuanced; unsophisticated/sophisticated; plain/elaborate;
simple/intricate. Note that in each case the second, complexity-reflective alterna-
tive has a distinctly more positive (or less negative) connotation than its opposite
counterpart.
P1: KAE
How Cognitive Taxonomy Has Grown More Complex 19
With the enhancement of investigative technology, the “window”

looking out on nature’s parametric space becomes constantly
enlarged. Scientific theory formation is, in general, a matter of spot-
ting a local regularity of phenomena in parametric space and then
projecting it “across the board,” maintaining it globally. But the the-
oretical claims of science are themselves never small-scale and local.
They stipulate – quite ambitiously – how things are always and every-
where. In developing natural science we use the window of technolog-
ical capability to scrutinize parametric space, continually augmenting
our database and then generalizing on what we see. But what we have
here is not a lunar landscape where once we have seen one sector we
have seen it all, and where theory-projections from lesser data gener-
ally remain in place when further data come our way. Instead it does
not require a sophisticated knowledge of history of science to realize
that our worst fears are usually realized – that our theories seldom if
ever survive intact in the wake of substantial extensions in our cognitive
access to new sectors of nature’s parametric space.
The history of science is a sequence of episodes of leaping to the
wrong conclusions because new observational findings indicate mat-
ters are not quite so simple as heretofore thought. As ample experi-
ence indicates, our ideas about nature are subject to constant stresses –
often radical and change-demanding – as we “explore” parametric
space more extensively. The technologically mediated entry into new
regions of parameter space constantly destabilizes the attained equi-
librium between data and theory. Physical nature exhibits a very dif-
ferent aspect when viewed from the vantage point of different levels
of sophistication in the technology of nature-investigator interaction.
The possibility of change is ever-present. The ongoing destabilization
of scientific theories is the price we pay for operating a simplicity-
geared cognitive methodology in an actually complex world. The nat-
ural dialectic of scientific inquiry ongoingly impels us into ever-deeper
levels of sophistication.6
6 On the structure of dialectical reasoning see the author’s Dialectics (Albany: State
University of New York Press, 1977), and for the analogous role of such reason-
ing in philosophy see The Strife of Systems (Pittsburgh: University of Pittsburgh Press,
1985).
P1: KAE
And in this context it is important to recognize that growth in the tax-

onomical structure of science is an integral part of the progress of science
itself.7 As the French physicist Pierre Auger has written:
At the time of Auguste Comte, the sciences could be classified in six of seven
main categories known as disciplines, ranging from mathematics to sociology.
Since then, during the nineteenth century and at the beginning of the twenti-
eth, there has been what might be described as an intro-disciplinary dismem-
berment, each of the main categories splitting up into increasingly specialized
fields, each of which rapidly assumed comparable importance to that of the
actual disciplines from which it sprang. Chemistry, for example, in the days of
Lavoisier formed a reasonably homogeneous entity, but chemists were soon
obliged to choose between inorganic and organic chemistry; within the latter,
a distinction arose during the second half of the nineteenth century between
the chemistry of aromatic compounds and that of aliphatic compounds, the
latter shortly being further subdivided into the study of saturated compounds
and that of unsaturated compounds. Finally, at the present time, a chemist
can devote a most useful research career entirely to a single chemical family.
The same process can be discerned in physics and in biology.8
In the days of St. Thomas Aquinas, the whole of learning consisted

of five or six areas, each of which had five or six sub-fields.9 The his-
tory of science is an endlessly repetitive story of simple theories giv-
ing way to more complicated and sophisticated ones. The Greeks had
four elements; in the nineteenth century Mendeleev had some sixty;
by the 1900s this had gone to eighty, and nowadays we have a vast
series of elemental stability states. Aristotle’s cosmos had only spheres;
Ptolemy’s added epicycles; ours has a virtually endless proliferation
of complex orbits that only supercomputers can approximate. Greek
7 The scientific study of the taxonomy and morphology of science itself is a virtually
nonexistent enterprise. Philosophers used to deal with these matters, but they aban-
doned them after the late 19th century, when science began to change too fast for
those concerned to look at truth sub specie aeternitatis. (A good survey of the historical
situation is given in Robert Flint, Philosophy as Scientia Scientiarum: A History of Classi-
fications of the Sciences [Edinburgh and London: W. Blackwood and Sons, 1904].) In
more recent days the subject has been left to bibliographers. For the older, “classical”
attempts see Ernest Cushing Richardson, Classification: Theoretical and Practical, 3rd
ed. (New York: H. W. Wilson, 1930).
8 Pierre Auger, Current Trends in Scientific Research (Paris: UNESCO Publications, 1961),
pp. 15–16.
9 See Joseph Mariétan on Problème de la classification des sciences d’Aristote à St. Thomas
(St. Maurice: Fribourg, 1901).
P1: KAE
science was contained on a single shelf of books; that of the Newto-

nian age required a roomful; ours requires vast storage structures filled
not only with books and journals but with photographs, tapes, floppy
disks, and so on. Of the quantities currently recognized as the funda-
mental constants of physics, only one was contemplated in Newton’s
physics: the universal gravitational constant. A second was added in
the nineteenth century, Avogadro’s constant. The remaining six are
all creatures of twentieth-century physics: the speed of light (the veloc-
ity of electromagnetic radiation in free space), the elementary charge,
the rest mass of the electron, the rest mass of the proton, Planck’s
constant, and Boltzmann’s constant.10
It would be naive – and quite wrong – to think that the course of
scientific progress is one of increasing simplicity. The very reverse is
the case: scientific progress is a matter of complexification because
over-simple theories invariably prove untenable in a complex world.
Our ever-more-sophisticated searches invariably engender changes of
mind, thereby moving us in the direction of an ever-more-complex
and sophisticated picture of the world. In this regard our commitment
to simplicity and systematicity, though methodologically necessary, is
ontologically unavailing, for our methodological commitment to sim-
plicity should not and does not preclude the substantive discovery of
complexity.
This complexification of scientific knowledge is also manifested
vividly in the fact that our cognitive taxonomies are bursting at the
seams. Consider the example of taxonomic structure of physics. We
may assume a three-layer taxonomy: the field as a whole, the branches
thereof, and the sub-branches of the branches. The taxonomic situ-
ation prevailing toward the beginning of this century is pictured in
Display 3.1. It is interesting to contrast this picture of the taxonomic
situation in physics with the picture of the situation in subsequent
decades as given in Display 3.2.
These tables tell a significant story. In the 11th (1911) edition of the
Encyclopedia Britannica, physics is described as a discipline composed
of 9 constituent branches (e.g., “Acoustics” or “Electricity and Mag-
netism”), which were themselves partitioned into 15 further specialties
10 See B. W. Petley, The Fundamental Physical Constants and the Frontiers of Measurement
(Bristol: Hilger, 1985).
P1: KAE
Astronomy
—Astrophysics
—Celestial Mechanics
Acoustics
Optics
—Theoretical Optics
—Spectroscopy
Mechanics
Heat
—Calorimetry
—Theory of Radiation
—Thermodynamics
—Thermometry
Electricity and Magnetism
—Electrochemistry
—Electrokinetics
—Electrometallurgy
—Electrostatics
—Thermoelectricity
—Diamagnetism
—Electromagnetism
Pneumatics
Energetics
Instrumentation
display 3.1. The taxonomy of physics in the 11th edition of the Encyclopedia
Britannica (1911). Note : Adapted from the Classified List of Articles at the end
of Vol. XXIX (index volume).
(e.g., “Thermoelectricity” and “Celestial Mechanics”). The 15th

(1974) version of the Britannica divides physics into 12 branches whose
sub-fields are – seemingly – too numerous for listing. (However the
14th, 1960s edition carried a special article entitled “Physics, Articles
on,” which surveyed more than 130 special topics in the field.) When
the National Science Foundation launched its inventory of physical
specialties with the National Register of Scientific and Technical Per-
sonnel in 1954, it divided physics into 12 areas with 90 specialties. By
1970 these figures had increased to 16 and 210, respectively. And the
process continues unabated to the point that people are increasingly
reluctant to embark on this classifying project at all.
In the nineteenth century, a different situation prevailed. Then, all
of the great figures of the philosophy of science of this era – Comte,
Whewell, Mill, Spencer, Peirce – made substantial examinations of the
P1: KAE
1954 1970
Astronomy (16 specialties) Astronomy

Acoustics (7 specialties) —Solar-Planetary Relationships
Optics (8 specialties) (9 specialties)
Mechanics and Heat (13 specialties) —Planetology (6 specialties)
Electromagnetism (6 specialties) —11 Further Astrophysical
Solid State (8 specialties) Specialties
Atomic and Molecular Physics Acoustics (9 specialties)
(5 specialties) Optics (10 specialties)
Nuclear Physics (9 specialties) Mechanics (10 specialties)
Theoretical Physics: Quantum Physics Thermal Physics (9 specialties)
(4 specialties) (= Elementary Electromagnetism (8 specialties)
Particles and Fields) Solids (25 specialties)
Theoretical Physics: Classical Fluids (9 specialties)
(3 specialties) Atmospheric Structure and Dynamics
Electronics (7 specialties) (16 specialties)
Instrumentation and Miscellaneous Atoms and Molecules (10 specialties)
(4 specialties) Nuclei (3 specialties)
Elementary Particles and Fields
(6 specialties)
Physical Chemistry (25 specialties)
Biophysics (6 specialties)
Solid Earth Geographics (10 specialties)
Instrumentation (28 specialties)
display 3.2. Physics specialties in the “National Register of Scientific and
Technical Personnel” for 1954 and 1970. Data from American Science Manpower:
1954–1956 (Washington, DC: National Science Foundation Publications,
1961) and “Specialties List for Use with 1970 National Register of Scientific
and Technical Personnel” (Washington, DC: National Science Foundation
Publications, 1970).
systematic structure of science as a whole, including the inventory and

arrangement of its component branches and subdivisions.11
The rapid expansion of interest in the taxonomy of knowledge dur-
ing the nineteenth century (in the wake of the growth of historical
11 The better older surveys of the historical situation are Julius Pelzholdt, Bibliotheca
Bibliographica (Leipzig: W. Engelmann, 1866). Charles W. Shields, Philosophia Ultima,
vol. II, The History of the Sciences and the Logic of the Sciences or the Science of the Sciences (New
York: Scribner’s, 1888–1905), 3. vols; Robert Flint, Philosophy as Scientia Scientiarum
and History of the Classification of the Sciences (Edinburgh and London: Blackwood,
1904); and Ernest C. Richardson, Classification: Theoretical and Practical, 3rd ed. (New
York: H.W. Wilson, 1930).
P1: KAE
interest) is a phenomenon as striking as its modern neglect. No major

nineteenth-century philosopher of science failed to give extensive con-
sideration to the problem of classifying the sciences, and no major
twentieth-century philosopher of science has touched it.12 Substan-
tially the same story can be told for every field of science. The emer-
gence of new disciplines, branches, and specialties is manifest every-
where. And as though to negate this tendency and maintain unity, one
finds an ongoing evolution of interdisciplinary syntheses – physical
chemistry, astrophysics, biochemistry, and so on. The very attempt to
counteract fragmentation produces new fragments. Indeed, the phe-
nomenology of this domain is nowadays so complex that some writers
believe the idea of a “natural taxonomy of science” must be abandoned
altogether.13 The expansion of the scientific literature is so great that
natural science has in recent years been disintegrating before our very
eyes. An ever-larger number of ever-more refined specialties has made
it increasingly difficult for experts in a given branch of science to
achieve a thorough understanding about what is going on, even in the
specialty next door.
Evolving Complexity
The ongoing refinement in the division of cognitive labor as the infor-
mation explosion continues has resulted in a literal dis-integration of
knowledge. The “progress of knowledge” has been marked by a prolif-
eration of increasingly narrow specialties – and no single person can
keep up with it. A person’s understanding of matters outside his or
her immediate bailiwick is bound to become superficial. No counting
scientist now has a detailed understanding of his or her own field
beyond the boundaries of a narrow subspecialty. At their home base
scientists know the details, nearby they have an understanding of gen-
eralities, but at a greater remove they can be no more than informed
amateurs.
The “unity of science” to which many theorists aspire may possibly
come to be realized at the level of highly general concepts and theories
12 For an analysis see R. G. A. Dolby, “Classification of the Sciences: The Nineteenth-

Century Tradition” [unpublished study issued for its author by the University of Kent
at Canterbury ca. 1975].
13 See John Dupré, The Disorder of Things: Metaphysical Foundations of the Disunity of Science
(Cambridge, MA: Harvard University Press, 1993).
P1: KAE
Evolving Complexity 25
shared between different sciences – that is, at the level of ideational

overlaps. For as one acute observer has written:
But this very over-specialization has provoked an inverse or rather a comple-

mentary process, that of interdisciplinary synthesis; thus, from physics and
chemistry there has grown up a new discipline of physical chemistry, which is
influenced by both these sciences. This process has given rise to a whole series
of new sciences with double or even triple names-astrophysics, biochemistry,
mathematical chemistry, physico-chemical biology, etc. Thus, the diverging
lines of the subjects of scientific research are connected by cross-links which
restore unity to the whole.14
All the same, for every conceptual commonality and shared element
a dozen differentiations and distinctions will emerge.
The increasing complexity of our world picture is a striking phe-
nomenon throughout the development of modern science. Whatever
information the sciences achieve is bought dearly through the prolif-
eration of complexity. It is, of course, possible that the development of
physics may eventually carry us to theoretical unification where every-
thing that we class among the “laws of nature” belongs to one grand
unified theory – one all-encompassing deductive systematization inte-
grated even more tightly than that Newton’s Principia Mathematica.15
But on all discernible indications the covers of this elegantly contrived
“book of nature” will have to encompass a mass of ever more elaborate
diversity and variety. And this integration, on the principle of a pyra-
mid, will cover further down an endlessly expansive range of the most
variegated components.
The lesson of such considerations is clear. In the course of scientific
progress our knowledge grows not just in extent but also in complexity.
The history of science tells an ongoing story of taxonomic complexifi-
cation, and in modern science extensive specialization and division of
labor continue inexorably. The years of apprenticeship that separate
master from novice increase. A science that moves continually from
an over-simple picture of the world to one that is more complex must
have more elaborate processes for its effective cultivation. And as the
14 Pierre Auger, Current Trends in Scientific Research (Paris: UNESCO Publications, 1961),
pp. 15–16.
15 See Steven Weinberg, Dreams of a Final Theory (New York: Pantheon, 1992). See
also Edoardo Amaldi, “The Unity of Physics,” Physics Today, 261 (September, 1973),
p. 23–29. Compare also C. F. von Weizsäcker, “The Unity of Physics,” in Ted Bastin
(ed.), Quantum Theory and Beyond (Cambridge: Cambridge University Press, 1971).
P1: KAE
scientific enterprise itself grows more extensive, the greater elaborate-

ness of its productions requires an increasingly intricate intellectual
structure for its accommodation. The complexification of scientific
process and product escalate hand in hand. The regulative ideal of
science is to integrate our knowledge of the world’s modus operandi
into a coherent and cohesive unifying system. But the world’s com-
plexity makes this an aspiration rather than an accomplished fact: it
represents a goal toward which we may be able to make progress but
which we will never attain.16
Yet complexity is not unqualified negative. It is an unavoidable con-
comitant of progress in this domain, for we could not extend our
cognitive or our practical grasp of the world without taking its com-
plexification in stride. Throughout the realm of human artifice –
cognitive artifice included – further complexity is part and parcel
of extending the frontiers of progress. The struggle with complex-
ity that we encounter throughout our cognitive efforts is an inherent
and unavoidable aspect of the human condition’s progressive impetus
to doing more and doing it better.
Spencer’s law of cognitive development through taxonomic com-
plexification accordingly carries a mixed bag of news in its wake. The
good news is that cognitive progress through the enticement of infor-
mation brings conceptual progress and thereby deeper understanding
in its wake. But the bad news is that this progress can be realized only at
great cost in regard both to its production and its utilization. For its pro-
duction calls for even more information and thereby an ever-growing
investment of resources and of effort. Operating at the frontier of any
branch of natural science becomes increasingly more demanding and
difficult.17
A Quantitative Perspective
As we examine the cognitive complexity reflected in the heterogeneity
and diversification of relevant information, at what pace does it grow
as inquiry proceeds? Here it seems sensible to go back to basics.
16 For variations on this theme see the author’s The Limits of Science (Berkeley: University
of California Press, 1984).
17 On further issues relevant to these ideas see also the author’s Scientific Progress
(Oxford: Basil Blackwell, 1978).
P1: KAE
A Quantitative Perspective 27
Spencer’s Law of Cognitive Development means in effect that the

extent of our actual knowledge about the world and its ways will stand
correlative with the complexity of its systemic articulation. Now com-
plexity is not a feature of information as such but rather of its structure.
And this structural complexity is indicated by and indeed encompassed
in the taxonomy by whose means we organize this information. It is a
matter of the distinctions and classifications we use in giving a coherent
exposition of the body of information at issue. The systemic architec-
ture of our information is the crux here.
And seeing that knowledge is a matter of organized and sys-
tematized information, informational complexity becomes our best
standard for assessing the extent of its attainment. Knowledge
grows in line with the extent to which our increasingly extensive
and thereby complex body of information expands its hierarchial
structure.
Returning to the text-geared themes of Chapter 2, let us consider
the hierarchy of informational structure represented by the sequence:
sentence, paragraph, section, chapter, volumes, shelf, cabinet, room,
library, library system, and so on. Along just these lines, the hierar-
chical structure of a body of information comes into view. And it can
now be assessed, in principle, along quantitative lines. For illustration,
(and a simple example), let us hitch our wagon to the star of orders
of magnitude and assume a factor of 10 all along the way: 10 sen-
tences to be a paragraph, 10 paragraphs to a section, 10 sections to a
chapter, and so on. Assuming a superlibrary of 100 million volumes
(the Library of Congress nowadays has some 20 million) we contem-
plate the following hierarchy:
sentences 10,000,000,000
paragraphs 1,000,000,000
sections 100,000,000
chapters 10,000,000
volumes 1,000,000
shelves 100,000
cabinets 10,000
rooms 1,000
libraries 100
library complexes 10
library systems 1
P1: KAE
Thus the hierarchic depth of body of our million volumes is 10, that
is to say, log 10,000,000,000. And more generally, we can say that the
hierarchic depth of a body of information can be measured by the
logarithm of its textual volume.
Now on the basis of Spencer’s Law we would expect that the cog-
nitive sophistication that reflects actual knowledge would be faithfully
reflected in the taxonomic complexity of the information at issue.
Spencer’s Law is thus the pathway to determining the relation of
knowledge to mere information, enabling us at the same time to quan-
tify the assessment of knowledge. And on this basis, the complex and
sophisticated information – the premium information that constitutes
actual knowledge – will increase in proportion not with the mere bulk
of our body of information as such, but only with its logarithm.
A look at the process of knowledge development indicates the plau-
sibility of this perspective. By all indications, the taxonomy needed
for the cognitive accommodation of a body of information (I ) – and
correspondingly the complexity of that information – grows in inverse
proportion to the volume of that body of information (#I ); the larger
a body of information already is, the more sedately its taxonomy
expands. Accordingly we have the relation
d 1
compl (I ) ≈ .
dI #I
And on this basis we have

d#I
compl (I ) ≈ ∝ log #I.
#I
For on its basis the complexity of a body of information as reflected
in its taxonomic depth stands not as its volume but merely as the
logarithm thereof. This relationship gives a quantitative expression
to Spencer’s Law, that with the quantitative growth of information
there indeed is a natural increase in heterogeneity (in taxonomic
diversification) – albeit at a decelerating pace. With the expansion
of information, its complexity does increase, but only at a decreasing
(logarithmic) rate. And in this quantified form, Spencer’s Law of Loga-
rithmic Development has profound implications for the development
of knowledge.
P1: KAE
Gibbon’s Law of Logarithmic Returns
(1) Cognitive progress is subject to Kant’s Principle of Question propagation

to the effect that new knowledge always brings new questions in its wake.
(2) However, the increase of mere information does not yield a corresponding
increase in knowledge: knowledge is not proportional to the volume of informa-
tion, but only to its logarithm. This key epistemological principle traces back at
least to Edward Gibbon. (3) Gibbon’s Law of Logarithmic Returns as a princi-
ple of the realm of conception parallels the Weber-Fechner Law in the epistemics
of perception. (4) The Law of Logarithmic Returns accounts for Max Planck’s
Thesis that scientific progress becomes ever more difficult, so that diminishing
returns on effort are an unavoidable facet of inquiry.
Kant’s Principle of Questions Propagation

New knowledge that emerges from the progress of inquiry can bear
very differently on the matter of questions. Specifically, we can dis-
cover
(1) New (that is, different) answers to old questions.

(2) New questions.
(3) The inappropriateness or illegitimacy of our old questions.
With (1) we learn that the wrong answer has been given to an old
question: we uncover an error of commission in our previous question-
answering endeavors. With (2) we discover that there are certain ques-
tions that have not heretofore been posed at all: we uncover an error of
omission in our former question-asking endeavors. Finally, with (3) we
29
P1: KAE
30 Gibbon’s Law of Logarithmic Returns
find that we have asked the wrong question altogether: we uncover an

error of commission in our former question-asking endeavors, which
are now seen to rest on incorrect presuppositions (and are thus gener-
ally bound up with type [1] discoveries). Three different sorts of cogni-
tive progress are thus involved here – different from one another and
from the traditional view of cognitive progress in terms of a straight-
forward “accretion of further knowledge.”
The coming to be and passing away of questions is a phenomenon
that can be mooted on this basis. A question arises at the time t if it
then can meaningfully be posed because all its presuppositions are
then taken to be true. And a question dissolves at t if one or another
of its previously accepted presuppositions is no longer accepted. Any
state of science will remove certain questions from the agenda and dis-
miss them as inappropriate. Newtonian dynamics dismissed the ques-
tion “What cause is operative to keep a body in movement (with a
uniform velocity in a straight line) once an impressed force has set it
into motion?” Modern quantum theory does not allow us to ask “What
caused this atom on californium to disintegrate after exactly 32.53
days, rather than, say, a day or two later?” Scientific questions should
thus be regarded as arising in a historical setting. They arise at some
juncture and not at others; they can be born and then die away.
A change of mind about the appropriate answer to some question
will unravel the entire fabric of questions that presupposed this ear-
lier answer. For if we change our mind regarding the correct answer to
one member of a chain of questions, then the whole of a subsequent
course of questioning may well collapse. If we abandon the luminifer-
ous aether as a vehicle for electromagnetic radiation, then we lose at
one stroke the whole host of questions about its composition, struc-
ture, mode of operation, origin, and so on. The course of erotetic
change is no less dramatic than that of cognitive change.
Epistemic change over time thus relates not only to what is “known”
but also to what can be asked. Newly secured information opens up
new questions. And when the epistemic status of a presupposition
changes from acceptance to abandonment or rejection, we witness
the disappearance of various old presuppositions through dissolution.
Questions regarding the modus operandi of phlogiston, the behavior of
caloric fluid, the structure of the luminiferous aether, and the charac-
ter of faster-than-light transmissions are all questions that have become
P1: KAE
Kant’s Principle of Questions Propagation 31
lost to modern science because they involve presuppositions that have

been abandoned.
The second of those aforementioned modes of erotetic discovery is
particularly significant. The phenomenon of the “birth” of new ques-
tions was first emphasized by Immanuel Kant, who in his classic Critique
of Pure Reason depicted the development of natural science in terms
of a continually evolving cycle of questions and answers, where “every
answer given on principles of experience begets a fresh question, which likewise
requires its answer and thereby clearly shows the insufficiency of all sci-
entific modes of explanation to satisfy reason.”1 This claim suggests
the following Principle of Question Propagation – Kant’s Principle,
as we shall call it: “The answering of our factual (scientific) questions
always paves the way to further as yet unanswered questions.”
Note, however, that Kant’s Principle can be construed in two rather
different ways:
1. A universalized mode: EACH specific (particular) question Q that can be

raised on a basis of a state-of-knowledge K engenders a (Q-correlative) line of
questioning that leads ultimately to a question Q’ whose answer lies outside of
K – a question that forces an eventual shift from K to some suitably augmented
or revised modification thereof.
2. A particular mode that arises when the capitalized EACH of the preceding
formula is replaced by SOME.
On the first construction, science is an essentially divergent process,

with questions leading to more questions in such a way that the erotetic
agenda of successive stages of science continually grows in scope and
size. This view was endorsed by W. Stanley Jevons, who wrote: “As it
appears to me, the supply of new and unexplained facts is divergent
in extent, so that the more we have explained, the more there is to
explain.”2 The second construction is, however, a far more modest
proposition, which merely sees science as self-perpetuating with some
new questions arising at every stage, thereby opening a window of
opportunity for the investigation of new issues. However, the question
agenda of science is not necessarily a growing one, since questions may
well die off by dissolution at a rate roughly equal to the birth of new
questions.
1 Immanuel Kant, Prolegomena to Any Future Metaphysic, sect. 57.

2 W. S. Jevons, Principles of Science, 2nd ed. (London: Macmillan 1876), p. 753.
P1: KAE
Kant himself undoubtedly intended the principle in the first (uni-

versalized) sense. But it would actually seem more plausible and real-
istic to adopt it in the second, more modest (particularized) sense,
which yields a thesis amply supported by historical experience: that
every state-of-the-art condition of questioning ultimately yields, some-
where along the road, a line of questioning that engenders the transi-
tion. The states of science are unstable: the natural course of inquiry
provides an impetus by which a given state is ultimately led to give way
to its successor. At bottom, Kant’s Principle rests on the insight that no
matter what answers are in hand, we can proceed to dig deeper into
the how and why of things by raising yet further questions about the
matters involved in these answers themselves. Accordingly, whenever
we obtain new and different answers, interest is at once deflected to
the issues they pose. When physicists postulate a new phenomenon
they naturally want to know its character and modus operandi. When
chemists synthesize a new substance they naturally want to know how
it interacts with the old ones.
This Kantian principle of question-propagation in empirical inquiry
indicates a fact of importance for the theory of cognitive progress. One
need not claim longevity – let alone immortality – for any of the current
problems to assure that there will be no problems ten or one hundred
generations hence. (As immortal individuals are not needed to assure
the immortality of the race, so immortal problems are not needed to
assure the immortality of the scientific frontier.)
But what sort of reliance do we get on the development of new
information with the cognitive progress of question resolution?
Edward Gibbon and Logarithmic Returns

As Spencer’s Law of Cognitive Development would suggest, the cogni-
tive depth of a body of information can be represented by its correlative
taxonomy – by the elaborateness of the taxonomic structure that it
engenders.3 And on this basis, the preceding deliberations suggest
that actual knowledge does not coordinate with the mere volume of
3 What concerns us here is emphatically not the “information” at issue in the theory
of communication, which has nothing to do with the informative content of a mes-
sage but only addresses its presentational complexity irrespective of the scope of its
substantive content.
P1: KAE
information but only with the systemic complexity it exhibits – and

thereby with its logarithm. Accordingly, letting K[I ] represent the
quantity of knowledge inherent in a body of information I of size #I,
we have: K[I ] = log #I.
And the purport of such a Law of Logarithmic Returns on expand-
ing information is clear enough. To increase knowledge additively we
must increase information multiplicatively: to advance knowledge by
a single step we must enlarge the base of its information pyramid by a
constant multiplicative factor.4
After all, not every insignificant smidgeon of information consti-
tutes knowledge, and the person whose body of information consists of
utter trivia really knows virtually nothing. To provide a simple illustra-
tion for this matter of significance, let us suppose an object-descriptive
color taxonomy – for the sake of example, a very simple one based
merely on Blue, Red, and Other. Then that single item of knowledge rep-
resented by “knowing the color” of an object – that it is red – is bound
up with many different items of (correct) information on the subject
(that it is not Blue, is rather similar to some shades of Other, etc.). As
such information proliferates, we confront a situation of redundancy
and diminished productiveness. Any knowable fact is always potentially
surrounded by a vast prenumbral cloud of relevant information. And
as our information becomes more extensive, those really significant
facts become more difficult to discern.
Cognitive progress brings growing complexity in its wake. Its ongo-
ing development involves us in ever-more free grained informative
detail. Even as “little fleas have smaller fleas to bite ’em,” so with issues
settled at one level of generality there are deeply subordinated and
thus smaller-scaled questions that are parasitically dependent on ear-
lier ones. More information in greater detail is forthcoming, at an
ever-accelerating pace, albeit at ever smaller orders of informative
4 In information theory, entropy is the measure of the information conveyed by a mes-

sage; it is measured by k log M, where M is the number of structurally equivalent
messages available via the available sorts of symbols. By extension, the present log I
measure of the knowledge contained in a given body of information might accordingly
be designated as the enentropy. Either way, the concept at issue measures informative
actuality in relation to informative possibility. For there are two types of informative
possibilities: (1) structural/syntactical as dealt with in classical information theory,
and (2) hermeneutic/semantical (i.e., genuinely meaning-oriented) as dealt with in
the present theory.
P1: KAE
I II III IV
Proportionality Asymptotic Accelerating Decelerating
Growth Growth Growth
K K K K
I I I I
display 4.1. The growth of knowledge in the wake of expanding information.
magnitude. But those increasingly fine-geared distinctions and sub-

tler considerations provide a diminishing informative yield. Thus a
vast amount of information can be exfoliated about any sort of thing
(be it a tree, a tool, or a person), but the really salient and signif-
icant knowledge of the item is generally something that can be set
out on a far more modest and small-scale basis. Our knowledge cer-
tainly increases with the addition of information but at a far less than
proportional rate.
A helpful perspective on this situation comes through the idea of
“noise”: consider that expanding bodies of information encompass
so much unproductive redundancy and unhelpful irrelevancy that it
takes successive increases in information to effect successive fixed-size
increases in actual knowledge.
In theory, three sorts of relationships can be contemplated in the
development between information and knowledge – along the lines
depicted in Display 4.1. We have already seen that alternative I is unten-
able. So are II and III. The Law of Logarithmic Returns stipulates that
it is IV that actually obtains. It is not the most pessimistic alternative;
III clearly is. And it not the most optimistic; II clearly is – although IV,
like II, envisions a situation of ever-diminishing returns.
The Law of Logarithmic Returns is in line with – and is interest-
ingly illustrated by – the idea of a cognitive life span as expounded by
the sagacious Edward Gibbon (1737–94), the great historian of the
Roman republic. In his Memoirs of My Life he wrote: “The proportion
of a part to the whole is the only standard by which we can measure
the length of our existence. At the age of twenty, one year is a tenth
perhaps of the time which has elapsed within our consciousness and
memory; at the age of fifty it is no more than a fortieth, and this relative
P1: KAE
value continues to decrease till the last sands are shaken out [of the
hour-glass measure of our lifespan] by the hand of death.”5 On this
basis, knowledge development is a matter of adding a given percentage
increment to what has gone before. Thus fresh experience superadds its
additional increment E to the preexisting total E in such a way that its
effective import is measured by the proportion that movement bears to
the total: E /E . And cumulatively we of course have it that this comes

to the logarithm of E: E /E = log E . On such an approach, an incre-
ment to one’s lifetime has a cognitive value determined on strict analogy
with Daniel Bernoulli’s famous proposal to measure the utility value of
incremental economic resources by means of a logarithmic yardstick.
Derek Price designated a “scholar’s solidness” – his stature as a con-
tributor to his field – as the logarithm of his life’s total of publications.6
But, as he himself suggests in passing, we might just as well expand this
from the individual to the community, and use it to measure not just the
productivity of a person but also the aggregate of total contributions
in a field. And in taking this step, we once again reach a logarithmic
measure of “cognitive solidarity” that would – as already noted – qualify
as a measure of the overall knowledge that the field encompasses.
To be sure, a devotee of Thomas Kuhn’s The Structure of Scientific
Revolutions7 might suggest that logarithmic retardation occurs only
within the routine (“normal”) phases of scientific development
whereas the revolutions that periodically lift science to a new level
see a sprint forward with new findings rushing in. But this view is over-
optimistic in ignoring the increasingly vast labor need to lift the sci-
ence from one level to the next. For a realistic picture see Display 4.2,
which conveys some idea of how such a process would proceed. And
it shows graphically that a Kuhnian view of spirit-penetrated scientific
progress is nowise at odds with the logarithmic retardation inherent
in Gibbon’s Law. The Kuhnian picture of new paradigms of scientific
5 Edward Gibbon, Memoirs of My Life (Harmondworth, 1984), p. 63. Gibbon’s “law of

learning” thus means that a body of experience that grows linearly over time yields a
merely logarithmic growth in cognitive age. Thus a youngster of 10 years has attained
only one-eighth of his or her expected chronological life span but has already passed
the halfway mark of his or her expected cognitive life span.
6 Derek J. Price, Little Science, Big Science (New York: Columbia University Press, 1963),
p. 50.
7 Thomas Kuhn, The Structure of Scientific Revolutions (Chicago: University of Chicago
Press, 1962).
P1: KAE
K(I)
(linear)
I
(logarithmic)
display 4.2. A Kuhnian Perspective.
understanding opening up the prospect of periodic leaps in progress

followed by a logarithmic phase of “moral science” is perfectly com-
patible with the present account.8
Gibbon’s Law means that an immense mass of information stands
coordinate with a modest body of knowledge. From the standpoint
of knowledge, information is highly redundant. Consider an exam-
ple. The yield of knowledge from information afforded by legibility
impaired manuscripts and papyri and inscriptions in classical pale-
ography provides an instructive instance. If we can decipher 70 per-
cent of its letters we can generally identify the text at issue in such a
manuscript. If we can make out 70 percent of its phases we can pretty
well figure out a sentence. If we can read 70 percent of its sentences we
can understand the message of the whole text. Some one-third of the
letters suffice to carry the whole message. So the mass of information at
hand boils down to a far smaller amount of knowledge. And this sort of
situation has far-reaching implications for the progress of knowledge.
One instructive way to measure the volume of available informa-
tion is through the opportunities for placement in a framework for
describing or classifying the features of things. With n concepts, you
can make n2 two-concept combinations. With m facts, you can project
m2 fact-connecting juxtapositions, in each of which some sort of char-
acteristic relationship is at issue. A single step in the advancement of
knowledge is here accompanied by a massive increase in the prolif-
eration of information. Extending the previous example, let us also
contemplate shapes in addition to colors, again supposing only three of
8 For a fuller treatment of the relevant issues see the author’s Scientific Progress (Oxford:
Blackwell, 1978), especially chapters X–XIII.
P1: KAE
them: rectangular, circular, other. Now when we combine color and

shape there will be 9 = 3 × 3 possibilities in the resultant (cross-) classi-
fication. So with that complex, dual-aspect piece of knowledge (color +
shape) we also launch into a vastly amplified (i.e., multiplied) infor-
mation spectrum over that increased classification-space. In moving
cognitively from n to n + 1 cognitive parameters we enlarge our knowl-
edge additively but expand our information field multiplicatively.
If a control mechanism can handle three items – be it in physical or
in cognitive management – then we can lift ourselves to higher levels of
capacity by hierarchical layering. First we can, by hypothesis, manage
three base-level items; next we can, by grouping three of these into
a first-level complex, manage nine base-level items; and thereupon
by grouping three of these first-level complexes into a second-level
complex we can manage 27 base-level items. Hierarchical grouping is
thus clearly a pathway to enhanced managerial capacity. But in cogni-
tive contexts of information management it is clear that a few items
of high-level information (=knowledge) can and will correspond to
a vast range of low-level information (=mere truths). Knowing (in
our illustration) where we stand within each of those three levels –
that is, having three pieces of knowledge – will position us in a vastly
greater information space (one of 27 compartments). And this situ-
ation is typical. The relational structure of the domain means that a
small range of knowledge (by way of specifically high-grade informa-
tion) can always serve to position one cognitively within a vastly greater
range of low-level information.
It is also instructive to view this idea from a different point of view.
Knowledge commonly develops via distinctions (A vs. non-A) that are
introduced with ever-greater elaboration to address the problems and
difficulties that one encounters with less sophisticated approaches. A
situation obtains that is analogous to the “Game of 20 Questions” with
an exponentially exfoliating possibility space being traced out stepwise
(2, 4, 8, 16, . . . , 2n , . . . ). With n descriptors one can specify for 2n
potential descriptions that specify exactly how, over all, a given object
may be characterized. When we add a new descriptor we increase by
one additional unit the amount of knowledge but double the amount
of available information. The information at hand grows with 2n , but
the knowledge acquired merely with n. The cognitive exploitation of
information is a matter of dramatically diminishing returns.
P1: KAE
#1 c1 (with c = .1) log c1 (=K[I]) (K[I]) K[I] as % of #I

100 10 1 – 1.0
1,000 100 2 1 .2
10,000 1000 3 1 .03
100,000 10,000 4 1 .004
display 4.3. The structure of the knowledge/information relation.
And so, as Display 4.3 illustrates, there is a substantial imbal-

ance between the magnitation of K[I] and #I itself, and with ever-
increasing information, the corresponding increase in knowledge
shrinks markedly.9 To increase knowledge by equal steps we must
amplify information by successive orders of magnitude. Each time we
double our mere information we increase our knowledge by only one
single step. The cognitive exploitation of information is a matter of dra-
matically diminishing returns. The difference at issue is that between
merely linear and drastically exponential growth. For linear growth
increases what has gone before by a fixed amount per period; expo-
nential growth increases what has gone before by a fixed percentage per
period. And the difference is of course drastic.
The Law of Logarithmic Returns thus presents us with an epistemo-
logical analogue of the old Weber-Fechner law of psychophysics, assert-
ing that inputs of geometrically increasing magnitude are required to
yield perceptual outputs of arithmetically increasing intensity, for the
presently contemplated law envisions a parallelism of perception and
conception in this regard. It stipulates that (informational) inputs of
geometrically increasing magnitude are needed to provide for (cog-
nitive and thus) conceptual outputs of arithmetically increasing magni-
tude. And so Gibbon’s Law of Logarithmic Returns as a principle of
the realm of conception parallels the Weber-Fechner Law in the epis-
temics of perception. In searching for meaningful patterns, the ongoing
proliferation of data-points makes contributions of rapidly diminish-
ing value.10
It is not too difficult to come by a plausible explanation for the
sort of information/knowledge relationship that is represented by
9 Throughout this book the symbol # will represent a numerical content measure con-
trast functions in such a way that, given any set S, #S will represent the conditionality
of the membership of S.
10 On the relevance of the Weber-Fechner law to scientometrics see Derek Price, Little
Science, Big Science, pp. 50–51.
P1: KAE
K[I ] = log #I. One good reason for such a K[I]/#I imbalance may
lie in the efficiency of intelligence in securing a view of the modus
operandi of a world whose law-structure is sufficiently simple to be
learner friendly – at least in part. For then one can extract a dispro-
portionate amount of general fact from a modest amount of infor-
mation. (Note that whenever an infinite series of 0’s and 1’s, as per
01010101 . . . , is generated – as this sample series indeed is – by a rel-
atively simple law, then this circumstance can be gleaned from a com-
paratively short initial segment of this series.) In rational inquiry we
try the simple solutions first, and when they cease to work – when they
are ruled out by further findings (by some further influx of coordi-
nating information) – we move on to the more complex. Things go
along smoothly until an over-simple solution becomes destabilized by
enlarged experience. We get by with the comparatively simpler options
until we have to do otherwise by the expanding information about the
world’s modus operandi made possible through enhanced new means
of observation and experimentation. But with the expansion of knowl-
edge, new accessions mean new demands. At bottom, then, there are
two closely interrelated reasons for that disparity between K[I ] and #I.
For one thing, where order exists in the world, intelligence is rather
efficient in finding it. And for another, if the world were not orderly –
were not amenable to the probes of intelligence – then intelligent
beings would not and could not have emerged in it through evolu-
tionary mechanisms.
The circumstance that ample information yields only modest knowl-
edge must not be invoked as an argument for skepticism as this doc-
trine is usually understood. For, of course, skepticism as usually con-
ceived maintains that the realization of knowledge is infeasible, whereas
the line of thought that is presently at issue hold that it is difficult and
challenging. What is at issue here is that the obstacles to knowledge are
substantial, not that they are insurmountable.
Diminishing Returns and Logarithmic Retardation

Consider now the result of combining two ideas:
(1) Knowledge is distinguished from mere information as such by its signifi-

cance. In fact, knowledge is simply particularly significant information – informa-
tion whose significance exceeds some threshold level (say q). (In principle
P1: KAE
there is room for variation here according as one sets the quality level of entry
qualification and the domain higher or lower.)
(2) The significance of additional information is determined by its impact
on preexisting information. Significance in this sense is a matter of the relative
(percentage-wise) increase that the new acquisitions effect upon the body of
preexisting information (I), which may – to reiterate – be estimated in the
first instance by the sheer volume of the relevant body of information: the
literature of the subject.
Accordingly, the significance of incrementally new information can be mea-

sured in terms of how much it adds, and thus by the ratio of the increment of
new information to the volume of information already in hand: #I/#I.
Thus knowledge constituting significant information is determined
through the proportional extent of the change effected by a new item
in the preexisting situation (independently of what that preexisting
situation is). In milking additional information for cognitively signifi-
cant insights it is generally the proportion of the increase that matters:
its percentage rather than its brute amount. So here too we arrive at
the Law of Logarithmic Returns governing the extraction of significant
knowledge from bodies of mere information.
On this basis of this fundamental relationship, the knowledge of a
two-sector domain increases additively notwithstanding a multiplica-
tive explosion in the amount of information that is now on the scene.
For if the field (F) under consideration consists of two sub-fields (F1
and F2), then because of the cross-connections obtaining within the
information-domains at issue the overall information complex will take
on a multiplicative size:
#I = inf (F ) = inf (F 1 ) × inf (F 2 ) = #I1 ×#I2 .
With compilation, information is multiplied. But in view of the indi-

cated logarithmic relationship, the knowledge associated with the body
of compound information I will stand at
K [I ] = log #I = log(#I1 ×#I2 ) = log #I1 + log #I2 = K [I1 ] + K [I2 ].
The knowledge obtained by compiling two information-domains (sub-

fields) into an overall aggregate will (as one would expect) consist sim-
ply in adding the two bodies of knowledge at issue. Where compilation
increases information by multiplicative leaps and bounds, the increase
in knowledge is merely additive.
P1: KAE
It is not too difficult to come by a plausible explanation for the sort

of information/knowledge relationship that is represented by K = log I.
The principal reason for such a K/I imbalance may lie in the efficiency
of intelligence in securing a view of the modus operandi of a world
whose law-structure is comparatively simple. For here one can learn
a disproportionate amount of general fact from a modest amount of
information. (Note that whenever an infinite series of 0’s and 1’s, as
per 01010101 . . . , is generated – as this series indeed is – by a relatively
simple law, then this circumstance can be gleaned from a compara-
tively short initial segment of the series.) In rational inquiry we try the
simple solutions first, and only when they cease to work – when they
are ruled out by further findings (by some further influx of coordi-
nating information) – do we move on to the more complex. Things
go along smoothly until an over simple solution becomes destabilized
by the enlarged experience afforded us by enhanced new means of
observation and experimentation.
The implications for cognitive progress of this disparity between
knowledge and mere information are not difficult to discern. Nature
imposes increasing resistance barriers to intellectual as to physical
penetration. Consider the analogy of extracting air for creating a vac-
uum. The first 90 percent comes out rather easily. The next 9 per-
cent is as difficult to extract as all that went before. The next .9 is
proportionally just as difficult. And so on. Each successive order-of-
magnitude step involves a massive cost for lesser progress; each suc-
cessive fixed-size investment of effort yields a substantially diminished
return.
The circumstance that the increase of information carries with it a
merely logarithmic return in point of increased knowledge suggests
that nature imposes a resistance barrier to intellectual as well as to
physical penetration. Intellectual progress is exactly the same: when
we extract actual knowledge (i.e., high-grade, descriptively significant
information) from mere information of the routine, common “gar-
den variety,” the same sort of quantity/quality relationship obtains.
Initially a sizable proportion of the available is high grade – but as
we press further this proportion of what is cognitively significant gets
ever smaller. To double knowledge we must quadruple information.
As science progresses, the important discoveries that represent real
increases in knowledge are surrounded by a growing penumbra of
P1: KAE
mere items of information. (The mathematical literature of the day

yields an annual crop of over 200,000 new theorems.)11
Planck’s Principle
In the ongoing course of scientific progress, the earlier investigations
in the various departments of inquiry are able to skim the cream,
so to speak: they take the “easy pickings,” and later achievements of
comparable significance require deeper forays into complexity and
call for much larger bodies of information. (It is important to realize
that this cost-increase is not because latter-day workers are doing better
science, but simply because it is harder to achieve the same level of
science: one must dig deeper or search wider to achieve results of
the same significance as before.) This situation is reflected in Max
Planck’s appraisal of the problems of scientific progress. He wrote
that “with every advance [in science] the difficulty of the task is increased; ever
larger demands are made on the achievements of researchers, and the need
for a suitable division of labor becomes more pressing.”12 The Law
of Logarithmic Returns would at once characterize and explain this
circumstance of what can be termed Planck’s Principle of Increasing
Effort to the effect that substantial findings are easier to come by in
the earlier phase of a new discipline and become more difficult in the
natural course of progress.
A great deal of impressionistic and anecdotal evidence certainly
points toward the increasing costs of high-level science. Scientists fre-
quently complain that “all the easy research has been done.”13 The
need for increasing specialization and division of labor is but one indi-
cation of this. A devotee of scientific biography cannot help noting
the disparity between the immense output and diversified fertility in
11 See Stanislaw M. Ulam, Adventures of a Mathematician (New York: Scribner’s, 1976).

12 Max Planck, Vorträge und Erinnerungen, 5th ed. (Stuttgart, 1949), p. 376, italics added.
Shrewd insights seldom go unanticipated, so it is not surprising that other theorists
should be able to contest claims to Planck’s priority here. C. S. Peirce is particularly
noteworthy in this connection.
13 See William George, The Scientist in Action (New York: Arno Press, 1938), p. 307.
The sentiment is not new. George Gore vainly lambasted it 100 years ago: “Nothing
can be more puerile than the complaints sometimes made by certain cultivators
of a science, that it is very difficult to make discoveries now that the soil has been
exhausted, whereas they were so easily made when the ground was first broken.”
George Gore, The Art of Scientific Discovery (London: Macmillan, 1878), p. 21.
P1: KAE
Planck’s Principle 43
the productive careers of the scientific collosi of earlier days and the
more modest scope of the achievements of their successors. As science
progresses within any of its established branches, there is a marked
increase in the overall resource-cost of realizing scientific findings of a
given level of intrinsic significance (by essentially absolutistic standards
of importance).14 At first one can skim the cream, so to speak: take
the “easy pickings.” But later achievements of comparable significance
require deeper forays into complexity and an ever-greater investment
of effort and material resources.15
The idea that science is not only subject to a principle of esca-
lating costs but also to a law of diminishing returns is due to the
nineteenth-century American philosopher of science Charles Sanders
Peirce (1839–1914). In his pioneering 1878 essay on “Economy of
Research” Peirce put the issue in the following terms:
We thus see that when an investigation is commenced, after the initial expenses
are once paid, at little cost we improve our knowledge, and improvement
then is especially valuable; but as the investigation goes on, additions to our
knowledge cost more and more, and, at the same time, are of less and less
worth. All the sciences exhibit the same phenomenon, and so does the course
of life. At first we learn very easily, and the interest of experience is very great;
but it becomes harder and harder, and less and less worthwhile.
(Collected Papers, Vol. VII [Cambridge, MA, 1958], sect. 7.144)
More information will always yield more knowledge – but simply

less of it. The increase of knowledge over time stands to the increase
14 The following passage offers a clear token of the operation of this principle specif-
ically with respect to chemistry: Over the past ten years the expenditures for basic
chemical research in universities have increased at a rate of about 15 per cent per
annum; much of the increase has gone for superior instrumentation, [and] for the
staff needed to service such instruments. . . . Because of the expansion in research
opportunities, the increased cost of the instrumentation required to capitalize on
these opportunities, and the more highly skilled supporting personnel needed for
the solution of more difficult problems, the cost of each individual research prob-
lem in chemistry is rising rapidly. (F. H. Wertheimer et al., Chemistry: Opportunities
and Needs [Washington, DC, 1965; National Academy of Sciences/National Research
Council], p. 17.)
15 The talented amateur has virtually been driven out of science. In 1881 the Royal
Society included many fellows in this category (with Darwin, Joule, and Spottis-
woode among the more distinguished of them). Today there are no amateurs. See
D. S. C. Cardwell, “The Professional Society” in Norman Kaplan (ed.), Science and
Society (Chicago: University of Chicago Press, 1965), pp. 86–91 (see p. 87).
P1: KAE
of information in a proportion fixed by the inverse of the volume of

already available information:
d d 1 d
K[I ] ≈ log #I ≈ #I.
dt dt #I dt
The more knowledge we already have in hand, the slower (by very rapid
decline) will be the rate at which knowledge grows with newly acquired
information. And the larger the body of available information, the
smaller will be the proportion of this information that represents real
knowledge.
The Law of Logarithmic Returns thus has substantial implications
for the rate of scientific progress.16 We cannot hope to predict the
content of future science, but the knowledge/information relationship
does put us in a position to make plausible estimates about its vol-
ume. There is no inherent limit to the possibility of future progress in
scientific knowledge, but the exploitation of this theoretical prospect
becomes more difficult, expensive, and demanding in terms of effort
and ingenuity. New findings of equal significance require greater
aggregate efforts. Accordingly, the historical situation has been one
of a constant progress of science as a cognitive discipline notwithstanding
its exponential growth as a productive enterprise (as measured in terms
of resources, money, manpower, publications, etc.).17 If we look at the
cognitive situation of science in its quantitative aspect, the Law of Loga-
rithmic Returns pretty much says it all. In this perspective, the struggle
to achieve cognitive mastery over nature presents a succession of esca-
lating demands, with the exponential growth in the enterprise associated
with a merely linear growth in the discipline. But do the facts bear this
out?
16 One might ask: “Why should a mere accretion in scientific ‘information’ – in mere
belief – be taken to constitute progress, seeing that those later beliefs are not necessarily
true (even as the earlier ones were not)?” The answer is that they are in any case better
substantiated – that they are “improvements” on the earlier ones by the elimination
of shortcomings. For a more detailed consideration of the relevant issues, see the
author’s Scientific Realism (Dordrecht: D. Reidel, 1987).
17 We are caught up here in the usual cyclic pattern of all hypothetico-deductive rea-
soning, in addition to explaining the various phenomena we have been canvassing
that projected K/I relationship is in turn substantiated by them. This is not a vicious
circularity but simply a matter of the systemic coherence that lies at the basis of
inductive reasonings. Of course the crux is that there should also be some predictive
power, which is exactly what our discussion of deceleration is designed to exhibit.
P1: KAE
Adams’s Thesis on Exponential Growth
(1) There should, of course, be sure empirical confirmation for Gibbon’s Law.
(2) And this can be found in the observation of Henry Adams that recent his-
tory has seen an exponential growth in scientific activity – and correlatively
in information as well. (3) This phenomenon is substantiated on many sides.
(4) However, there is also good reason to think that the expansion of scientific
knowledge has proceeded throughout this period at a merely linear (rather than
exponential) rate. (5) It emerges on this basis that knowledge stands not as the
volume of information but merely as its logarithm. So viewed in this light, the
development of modern science amply substantiates Gibbon’s Law of Logarithmic
Returns. (Appendix) A conspectus of relevant literature.
Scientific Progress
It is instructive also to consider the information/knowledge relation-
ship from the angle of history. In developmental perspective, there is
good reason to suppose that our body of bare information will increase
more or less in proportion with our resource investment in informa-
tion gathering. Accordingly, since this investment grows exponentially
over time (as has historically been the case in the recent period), we
have it that
d
#I(t) ∝ c t and correspondingly also #I (t) ≈ c t .
dt
(Note that here x ≈ y means that x = c y for some constant c, and
x ∝ c y + y for constants c and k.) Given Gibbon’s Law of Logarithmic
45
P1: KAE
46 Adams’s Thesis on Exponential Growth
Returns, this just-indicated relationship means
K [I (t)] = log #I(t) ≈ log c t ≈ t
and consequently
d
K [I (t)] = constant.
dt
It follows on this basis that, since exponential growth in I is coordinated
with a merely linear growth in K[I ], the rate of scientific progress in
point of knowledge during the information-exploding past has actually
remained essentially constant.
But do the observed facts bear this out? The preceding discussion
has argued for Gibbon’s Law K [I ] = log #I on the basis of general
principles. But will the historical data provide empirical substantiation
for it? Let us examine the issue.
Adams’s Thesis
Knowledge is an unusual resource; it is one of those very rare
commodities that are not diminished but rather amplified through
consumption. This makes the growth of knowledge a potentially
explosive process, and by the end of the nineteenth century no
clear-eyed observer could fail to remark the striking pace of scien-
tific advance. Among the first to detail this phenomenon in exact
terms by endeavoring to give it mathematical formulation was Henry
Brooks Adams (1838–1918), the American scholar, historian, and stu-
dent of cultural affairs. (He was a grandson of John Adams, George
Washington’s successor as president.) Noting that scientific work
increased at a rate fixed by a constant doubling time – so that sci-
ence has an exponential growth-rate – Adams characterized this cir-
cumstance as a “law of acceleration” governing the progress of science.
He wrote:
Laplace would have found it child’s play to fix a ratio in the progres-
sion in mathematical science between Descartes, Leibnitz, Newton and him-
self. . . . Pending agreement between. . . . authorities, theory may assume what
P1: KAE
Adams’s Thesis 47
it likes – say a fifty or even a five-and-twenty year period of reduplication. . . . for

the period matters little once the acceleration itself is admitted.1
Roughly speaking, Adams envisaged science as doubling in size and

content with every succeeding half-century, producing a situation of
exponential growth. And Adams did not stand alone here, as an antic-
ipation of his idea occurred in an obiter dictum in the 1901 Presidential
Address to the British Association for the Advancement of Science by
William Thomson (Lord Kelvin):
Scientific wealth tends to accumulate according to the law of compound inter-

est. Every addition to knowledge of the properties of matter supplies the
naturalist with new instrumental means for discovering and interpreting
phenomena of nature, which in their turn afford foundations for fresh
generalizations.2
Anticipations notwithstanding, the thesis that science develops

under compound-interest conditions with the effect of a continually
accelerating – indeed an exponential – growth rate may properly be
designated as Adams’s Thesis of Exponential Growth. It amounts to the
claim that the percentage rate of the per annum increase in the levels of scien-
tific effort and quantitative productivity remains constant with the passage of
time.3 Adams characterized his principle as a “law of acceleration, defi-
nite and constant as any law of mechanics [which] cannot be supposed
to relax its energy to suit the convenience of man.”4
1 The Education of Henry Adams (Boston, 1918; privately printed in 1907), chapter 34
(see p. 491). This chapter was written in 1904.
2 See Kelvin’s address in G. Basalla, William Coleman, and R. H. Kargon (eds.), Victorian
Science: A Self-Portrait through the Presidential Addresses of the British Association for the
Advancement of Science (New York: New York Academy of Science, 1970), pp. 101–28
(see p. 114, and compare p. 488). The idea also occurs in embryonic form in A. Conan
Doyle’s 1894 short story The Great Keinplatz Experiment, which claims that “knowledge
begets knowledge as money bears interest.”
3 The exponential growth of a quantity – its growth at “compound interest” – is governed
by the principle that at any time the rate of growth is proportional to the (then extant)
volume. It produces fixed doubling times (or tripling times, etc.), and sequential
fixed-period augmentations that increase in a geometric ratio.
4 The Education of Henry Adams, p. 493. For an interesting discussion of Adams’s numer-
ous, complex, and changing ideas regarding scientific and intellectual progress see
Ernest Samuels, Henry Adams: The Major Phase (Cambridge, MA: Harvard University
Press, 1964).
P1: KAE
The historical facts bear Adams out in this regard. For irrespective of
whether S represents scientific literature or manpower or information,
the growth of the enterprise has been exponential. Throughout recent
history it transpires that5
S ∝ e t or equivalently logS ∝ t.
And this, in effect, is the purport of Adams’s Thesis.
Dimensions of the Phenomenon

When Henry Adams enunciated his “law of acceleration,” he did so on
the basis of shrewd perception and hunch based on informal insight
into the historical course of things. Only a few scattered fragments
of statistical information needed to substantiate the historical reality
of this phenomenon were available in his day. However, this situation
is nowadays greatly improved. The continued data-gathering effort of
statisticians, sociologists, historians of science, and science administra-
tors provide impressive documentation of the phenomenon discerned
by Adams.6 Let us examine some of its dimensions.
Manpower
In 1875 the American Association for the Advancement of Science
had 807 members. Since then its membership has increased by leaps
and bounds into the tens of thousands. The register of American Men
of Science commenced publication in 1903 and exhibits a comparable
history of exponential growth. During most of the present century the
number of doctrinal scientists in the United States has been increasing
at around 6 percent yearly to yield an exponential growth rate with
a doubling time of roughly 12 years.7 (In the U.K. the comparable
figures stand at some 4 percent and 18 years, respectively.)8 A startling
consideration that is often but deservedly repeated is that well over
80 percent of ever-existing scientists (in even the oldest specialties,
5 We here use the notation x ≈ y for x = cy and x ∝ y for x = cy + k. The former represents
a relation of strictly linear proportionality; the latter represents the same apart from
the positioning of the origin.
6 For details regarding the literature see the Bibliographic Appendix to this chapter.
7 See, e.g., Derek J. Price, Little Science, Big Science, p. 11. This holds in virtually every
field of exact natural science.
8 Statistics of Science and Technology (London: HMSO, 1970), p. 115.
P1: KAE
Dimensions of the Phenomenon 49
e.g., mathematics, physics, and medicine) are alive and active at the
present time.
Literature and Information

First, a bit of hermeneutics. Scientific knowledge is coordinated, inter-
connected, and systematized. It is predicated on providing information
about how things are interrelated and interconnected. Now consider
any system of n elements. There will, in principle, obtain between these
items n × (n − 1) interconnections. But let a new element be brought
on the scene. In its wake there will now be (n + 1) × n interconnections
in the new state of things that has come to obtain. Accordingly, the shift
from n to n + 1 that has been engendered by the progress represented
by the introduction of that one single new item has increased our cog-
nitive system by 2n. A single step forward has thus seen an increase in
the field of information by an amount proportionate to the size of the
body of knowledge that was already at hand. In other words, we have
exponential growth.
Deliberations along these lines indicate that as further elements are
introduced into the range of our cognitive horizons the amount of raw
information with which we have to cope will grow exponentially. Little
wonder, then, that exactly this sort of thing is what has been happen-
ing in the realm of reference works, of scientific and scholarly jour-
nals, of libraries, and of the information domain of electronic search
engines.
In regard to the literature of science, it is readily documented that –
as Display 5.1 indicated – the number of books, of journals, and of jour-
nal papers has been increasing exponentially over the recent period,9
with scientific information growing at an average of some 5 percent
annually throughout the last two centuries. The result is exponential
growth with a doubling time of ca. 15 years – an order-of-magnitude
increase roughly every half century. By 1960, some 300,000 different
book titles were being published in the world, and the two decades
from 1955 and 1975 saw the doubling of titles published in Europe
9 Cf. Derek J. Price, Science since Babylon, 2nd ed. (New Haven, CT: Yale University
Press, 1975), and also Characteristics of Doctrinal Scientists and Engineers in the Univer-
sity System, 1991 (Arlington, VA: National Science Foundation, 1994), Document
No. 94–307.
P1: KAE
1,000,000
100,000
number of journals
10,000
Scientific Journals
1,000
(200) (200)
100
10
Abstract Journals
(1665)
1700 1800 1900 2000
date
display 5.1. The number of scientific journals and abstract journals founded,
as a function of date. Source : Derek J. Solla Price, Science since Babylon (New
Haven: Yale University Press, 1961).
from around 130,000 to over 270,000,10 and science has had its full
share of this literature explosion. The result is a veritable flood of
scientific literature. It is reliably estimated that, from the start, about
10 million scientific papers have been published, and currently some
30,000 journals publish some 600,000 new papers each year.
Facilities and Expenditure

Research and development expenditures in the United States grew
exponentially after World War II, increasing at a rate of some 10 per-
cent per annum. By the mid-1960s, America was spending more on
scientific research and development than the entire federal budget
before Pearl Harbor.
And America has not been alone here. Throughout the world, the
proliferation of scientific facilities has proceeded at an impressive pace
over the past hundred years. (In the early 1870s there were only 11
10 Data from An International Survey of Book Production during the Last Decades (Paris:
UNESCO, 1985).
P1: KAE
physics laboratories in the British Isles; by the mid 1930s there were
more than 300; today there are several thousand.)11 And, of course,
the scale of activities in these laboratories has also expanded vastly. The
immense cost in resources of the research equipment of contemporary
science is well known; even large organizations can hardly continue to
keep pace with the rising levels of research expenditures.12 Radiotele-
scopic observatories, low-temperature physics, research hospitals, and
lunar geology all involve outlays on a scale that requires the funding
support of national governments – sometimes even consortia thereof.
Science has become a vastly expensive undertaking. In a prophetic
voice, Alvin M. Weinberg (then Director of the Oak Ridge National
Laboratory) wrote:
When history looks at the 20th century, she will see science and technology as
its theme; she will find in the monuments of Big Science – the huge rockets,
the high-energy accelerators, the high flux research reactors – symbols of our
time just as surely as she finds in Notre Dame a symbol of the Middle Ages.13
The expansion of scientific activity – and cost! – throughout the 20th

century has been one of the characteristic phenomena of the era.
The Linear Growth of Knowledge

But let us now turn attention from scientific effort to scientific progress.
The picture that confronts us here is not quite so expansive. There is
good reason that the substantive level of scientific innovation has remained
roughly constant over the last few generations. This contention – that while
11 Data from William George, The Scientist in Action (London: Williams & Norgate,
1938).
12 Du Pont’s outlays for research stood at $1 million per annum during World War I
(1915–1918), $6 million in 1930, $38 million in 1950, and $96 million in 1960.
(Data from Fritz Machlup, The Production and Distribution of Knowledge in the United
States [Princeton: Princeton University Press, 1962], pp. 158–59; and see pp. 159–60
for the relevant data on a larger scale.) Overall expenditures for scientific research
and its technological development (R&D) in the United States stood at $.11×109 in
1920, $.13×109 in 1930, $.38×109 in 1940, $2.9×109 in 1950, $5.1×109 in 1953–54,
$10.0×109 in 1957–58, $11.1×109 in 1958–59, and ca. $14.0×109 in 1960–61 (ibid.,
pp. 155 and 187). Machlup thinks it a not unreasonable conjecture that no other
industry or economic activity in the United States has grown as fast as R&D (ibid.,
p. 155).
13 “The Impact of Large-Scale Science on the United States,” Science, 134 (1961; 21 July
issue), 161–64 (see p. 161).
P1: KAE
scientific efforts have grown exponentially the production of really

high-level scientific findings has remained constant – admits of vari-
ous substantiating considerations.
The number of scientists and the number of scientific publications
has increased roughly 10-fold every 50 years since the 17th century;
on that basis, one would say that the sum total of scientific knowledge
has seen a one-unit increase during every half century of recent his-
tory.14 One indicator of this constancy in high-quality science is the
relative stability of honors (medals, prizes, honorary degrees, mem-
bership of scientific academics, etc.). In some instances these reflect
a fixed-number situation (e.g., Nobel prizes in natural science). But if
the volume of clearly first-rate scientific work were expanding drasti-
cally, there would be mounting pressure for the enlargement of such
honorific awards and mounting discontent with the inequity of the
present reward-system. There are no signs of this.
Another significant indication of a merely linear increase in sci-
entific progress relates to references to specific findings – references
not in “the literature” at large, but in the synoptic handbooks, mono-
graphs, and treatises that endeavor to give a rounded picture of “the
state of the discipline.” The indices to such works have increased only
linearly.
Yet another sign of a merely linear growth of knowledge in spite of
an exponential increase in information emerges from the study of cita-
tions. The scientific literature has been growing at an exponential rate
in the modern era, subject to a doubling time of roughly 15 years; nev-
ertheless, the proportion of contributions that are of sufficient impor-
tance to be cited in the later literature declined greatly.15 Specifically,
publications among a given annual crop that are cited at least once
within a fixed interval (the persisting citations, so to speak) has also
been shrinking by one-half every 15 years. Correspondingly – albeit
not coincidentally – the number of persistents among the given year’s
14 Compare Price, Little Science, Big Science, pp. 55–56.

15 Citations are doubtless problematic as a measure of merit. All the same, citation
indices can surely be exploited to yield a fairly good indication of statistical trends.
Thus a reasonable indication of a paper’s utility would be its citation in a reasonable
proportion of papers in its field within a suitable time after its publication. Of course,
this utility might be practical/experimental rather than cognitive/theoretical. But
ways can always be devised to make allowances for such cases.
P1: KAE
crop is constant. In other words, while the literature grows exponen-

tially, the added number of significant results (as estimated by those
citation persistents) remains constant over time.16 And the citation
evidence also indicates that science as a body of knowledge has grown at
a merely linear rate.17
In the end, it should not be all that surprising that the pace of
authentically cognitive progress should remain constant even during
a time of dramatic information growth. After all, if an important, really
outstanding result in some field of inquiry is one (as is usually the case)
that is cited in X percent of the literature, then the number of important
contributions can (and generally will) remain constant over an expo-
nential growth of the literature simply because the bar of qualification
is here being raised in proportion with the literature itself.
Specifically, consider the following:
The citation-prominent members of a literature L – let this be prom(L) – are

those cited by at least some fixed fraction thereof (say 1%).
Now in this regard, citation studies of the scientific literature show

that a linear relationship obtains between #L and #prom(L) on log-log
graphing:18
log #L ≈ log # prom(L)
This means that

log(k × #L) = log prom(k × #L).
But we then have

log(k × #L) = log k + log #L ∝ log #L ≈ log # prom(L).
In other words, as L increases multiplicatively, prom (L) increases addi-

tively.
A host of relevant considerations thus conspire to indicate that when
science grows exponentially as a productive enterprise, its growth as an
intellectual discipline proceeds at a merely constant and linear pace.
16 For the data that substantiate this line of reasoning, see Price, Little Science, Big Science,
pp. 79–80.
17 Recall that in speaking of accumulation here, we have in mind enlargement in the
number of “significant results” whose contents may be in no sense cumulative.
18 For further detail see Price, Little Science, Big Science.
P1: KAE
The Lesson
At this point a significant conclusion comes into view. Let us put the
pieces together. The situation represented by Adams’s Thesis indicates
an exponential growth of scientific information in recent times. On
the other hand, a closer look at scientific understanding in terms of
the knowledge that represents our depth of understanding indicates a
merely constant growth throughout this recent period. And this com-
bination tells the story. If an exponential growth in information stands
propositional to a merely linear growth of the correlative knowledge,
then it indeed follows that #K = log #I, just as specified by Gibbon’s
Law. This of course means that as an information domain expands
exponentially – as is the case during an Adams’s Law era exponen-
tial scientific expansion – those top-quality findings that represent the
growth of knowledge will increase merely linearly. Thus there is, as we
have seen, good reason to think that this is actually the case, and the
available data do indeed substantiate the Gibbon’s Law relationship
between information and knowledge.19
Bibliographic Appendix
The first to take up Adams’s Thesis in a serious way on the basis of sta-
tistical data was F. K. Richtmyer, who, on analyzing the review literature
in physics, concluded that “since its beginning physics has increased,
with each generation, in a geometrical ratio.”20 In due course, the idea
became commonplace:
The time scale of human progress is certainly not linear. Technical progress
grows more rapid as time goes on, and perhaps the best chronological scale
for the history of science and technology would be one in which the divisions
of the scale were proportional to the logarithms of their distance from the
present time.21
19 What of the future? If it requires an exponential increase in scientific work and

information to achieve a merely linear growth in scientific knowledge, what can be
expected in a future in which there is at best a steady-state constancy in resource
investment in science? Here Gibbon’s law K = log I implies what might be called
a logarithmic retardation in the price of scientific growth with ever-larger entailment
between the achievement of significant breakthroughs. The case for this view of
things has been argued in the author’s Scientific Progress (Oxford: Blackwell, 1978).
20 “The Romance of the Next Decimal Place,” Science, 75 (1932), p. 1–5 (see pp. 1–2).
21 C. E. K. Mees, The Path of Science (New York: Wiley & Sons, 1946), p. 21.
P1: KAE
Bibliographic Appendix 55
However, the documentation of Adams’s Law improved greatly in

the early 1950s. On the side of outputs (i.e., scientific papers) a pioneer
study was Derek J. Price’s “Quantitative Measures of the Development
of Science,” Archives internationales d’Histoire des Sciences, vol. 14 (1951),
pp. 85–93, which maintained that the literature of science has grown
with a doubling time of some 12 years (a period that seems a bit too
small in the light of later studies). On the side of inputs the pioneer
study was by Raymond H. Ewell of the National Science Foundation,
who examined research and development (R&D) investments made in
the United States during the period from the foundation’s inception
in 1776 to 1954 (a space of 178 years). Ewell showed in detail that R&D
expenditures in the United States have grown at 10 percent per annum
at least since the first decade of this century, and probably longer. (He
also concluded that half of the total R&D expenditure of some 40 bil-
lion dollars to 1954 had been incurred during the six years from 1948
to 1954.) For a brief description of this work see Raymond H. Ewell,
“The Role of Research in Economic Growth,” Chemical and Engineering
News, vol. 33 (1955), pp. 2980–2985. The exponential growth of sci-
ence (in terms of publications, manpower, findings) also provides one
of the basic premises of Pierre Auger’s 1960 UNESCO study on Current
Trends in Scientific Research (Paris and New York: UNESCO Publications,
1961), see especially p. 15.
Overall, Derek J. Price has devoted much effort to the compilation
and depiction of the relevant facts, and this chapter has drawn on
his discussion. His initial publication in the area was extended in a
later paper titled “The Exponential Curve of Science,” Discovery, 17
(1956), pp. 240–243, and amplified in his books Science since Babylon
(New Haven: Yale University Press, 1961), and Little Science, Big Science
(New York: Columbia University Press,1963). See also Bentley Glass,
The Timely and the Timeless (New York: Putnam, 1970) as well as Rescher,
Scientific Progress (Oxford: Blackwell, 1978).
P1: KAE
56
P1: KAE
Quality Retardation
(1) The issue of importance is obviously a crucial factor for the utilization of
information. (2) This, however, is something subject to decidedly different degrees,
ranging from the pedestrianly routine to the uncontestably first-rate. (3) The
phenomenon of quality retardation so functions that as our information grows,
the rate of growth at increasingly higher quality is ever diminished. (4) Higher
quality elites are not just increasingly exclusive but slower growing as well. (5)
The quality of information is reflected in the structural constitution of texts and
in the taxonomies that such structural divisions reflect. (6) In approximation,
at least, importance can also be assessed in terms of citations that reflect the
utilization of texts in the literature of their subject.
The Centrality of Importance

The previous discussion has conceived of knowledge in terms of top
quality, first-rate information. But this is something of an oversimplifi-
cation. For in cognitive matters, our questions and answers, problems
and findings, come in various sizes – some virtually trivial, others por-
tentous, and many somewhere in between. And the difference matters
a great deal here. For it is the biggies that figure prominently in text-
books and histories, while the smallies get a footnote at best and blank
omission for the most part. Prizes, recognition, and career advance-
ment reward the big, while indifference befalls the small.
In much of the recent period, the pool of information workers
has grown rapidly (some 5 percent annually), and their output even
more so (an increase of around 15 percent yearly during the twentieth
57
P1: KAE
58 Quality Retardation
century) indicating an impressively growing efficiency in the produc-

tion of information.1 However, what matters in scientific inquiry is
progress, and this is determined not through the merely numerical
proliferation of findings but through their size – not their mere quan-
tity but the magnitude of their importance in the larger scheme of
things. Without the distinction between the important and the unim-
portant, mankind could neither adequately understand, successfully
teach, or effectively practice science.
Importance is a comparative conception: one thing is more impor-
tant than another. And perhaps the most critical fact about cognitive
importance is that it is an index of quality – of comparative signif-
icance in the context of understanding. It is precisely because one
finding is more important than another that it claims and deserves a
larger amount of attention and respect. Accordingly, importance is an
inherently elitist conception: there is nothing democratic about it.
One key consideration is that being cognitively important is differ-
ent from being interesting. Interest is subjective; it is dependent on
what people happen to be interested in, on what they find interest-
ing. By contrast, cognitive, and in particular scientific, importance is a
matter of how prominent a role a fact or finding deserves and thereby
demands in an adequate exposition of an area of inquiry. Importance
does – or should – function in a way that is objective and impersonal.
Importance in general is a teleological conception that connects peo-
ple with purposes. Something is important to someone for something,
even as eating a good diet is important to people-in-general for the
maintenance of their health. Now what will concern us here is specifi-
cally cognitive and in particular scientific importance in the sense not
of the importance of science for something else (such as human well-
being) but rather importance in science. What is at issue here is thus
the importance to serious inquirers for the proper understanding of
nature’s ways.
People have comparatively little difficulty in telling what is impor-
tant, but saying what importance is is another matter. Importance is like
1 See M. V. Parat, The Information Economy: Definition and Measurement (Washington, DC:
U.S. Department of Commerce, May 1977; OT Special Publication 77–12 (1) p. 133.
See also Derek J. Price, Little Science, Big Science (New York: Columbia University Press,
1963).
P1: KAE
pornography – we can generally spot it when we see it all right; it’s

the matter of adequate definitions and standards that is the difficulty.
And so, given the inherent significance of the issue, it is surprising
how little literature there is on the topic. I have been able to find only
one philosophical handbook or encyclopedia that has an entry under
the heading of importance – namely, the Spanish encyclopedia of José
Ferrater Mora. Its entry is brief and I cite it in full:
Importance : see relevance.
But this is clearly not very helpful.

One of the few contemporary philosophers of science who has
discussed the topic is Larry Laudan in Progress and Its Problems.2 He
observes, “The literature of the methodology of science offers us nei-
ther a taxonomy of its types of scientific problems, nor any acceptable
method of grounding their relative importance” (p. 13). But Laudan’s
own discussion is better at diagnosis than therapy: it offers us various
examples of important problems, but no effective criteria for what
constitutes this virtue. In essence, Laudan sees problems as important
to the extent that currently fashionable theories disagree about them.
Unfortunately, this idea shipwrecks on the circumstance that theories
can disagree about smaller issues as well as large ones.
Perhaps the most promising approach is to see importance as hing-
ing on the answer to the following question: “How large a loss by way of
emptiness and unknowing would be created for our grasp of a certain
domain if we lost our grip on the information at issue?” So regarded,
cognitive importance consists in making a difference for adequate
understanding. It is a matter of how large a gap would be left in the
body of our presumed knowledge by losing the item at issue. And
on this basis it seems plausible at least in the first instance to assess
importance in volumetric terms of space allocation in the relevant
expositions. Let us consider what can be done along these lines.
Levels of Cognitive Importance

The deliberations of the preceding chapters have been predicated on
construing knowledge as elite information, viewing this as subject to
2 Larry Laudan, Progress and Its Problems (Berkeley: University of California Press, 1971).
P1: KAE
The quality level of the

When λ equals THEN information at issue is
1 routine (= I )
1/ significant
4
1/ important
2
display 6.1. Levels of informative quality. Note: Here quality and quantity
are related by the equation: #K λ [I ] = (#I )λ .
a standard which, at the quantitative level, measures the amount of

actual knowledge (#K ) encompassed in a body of information (I )
as standing quantitatively merely at the logarithm thereof: #K [I ] =
log #I . But how does this situation alter if we lower the standard for
“knowledge” from the austere condition of first-rate, premium infor-
mation to something less demanding? How is one to deal in quantita-
tive terms with the lesser levels of “knowledge”?
The obvious response is that “knowledge” then becomes something
easier to come by and that with this relaxation a great deal more infor-
mation will qualify for this honorific designation. Again, a promising
way of viewing this matter is to approach it from a quantitative per-
spective. On this basis, we can now contemplate a spectrum of quality
levels whose makeup has the character depicted in Display 6.1 and
whose quantitative content is such that
#K λ [I ] = (#I )λ where 0 < λ ≤ 1.
Here K λ [I] is the information – or as we can now say, lower-grade

“knowledge” – at the quality level λ that is available within the overall
body of information I. And this is to be subject to the just-presented
stipulation of #K λ(I) (with the further specification that #K 0[I ] =
log #I).
Thus as conceived here, importance or significance is decidedly not
a matter of mere proportionality as per “the top 10 percent,” or “the
leading quarter,” or the like. The relationship at issue is not #S λ(I) =
λ × 100I. This sort of percent elitism is not at issue. For while percent
elitism is indeed a concept in the economics of money, the economics
of information require a more rigorous standard to do justice to the
way the concept of importance works in this domain where quality
does not keep step with quantity. With cognitive importance we have
P1: KAE
#(I )
#Kλ[I ]
(logarithmic)
0
∈ λ→ 1
Quality level
display 6.2. The fundamental principle of quality.
to operate the more rigorous standard of a root elitism where #K λ (I)

stands as the λ-root of #I: #K λ(I) = (#I) λ .
This of course means that as I grows (exponentially) over time our
“knowledge” at all of the lesser levels of generality will also grow (expo-
nentially), yet doing so at a rate that constantly decreases as the quality
level is raised by decreasing λ.
On this basis we arrive at what may be termed the Fundamental
Principle of Quality, namely, that since log #K λ[I] = λ log (#I), with a
fixed body of information I, the number of items at a given quality level
is simply proportionate to the logarithm of the number of items at that
quality level. Display 6.2 presents this linear relationship graphically.
Three quality levels are particularly important here. First there is
the λ = 1 level of routine information whose magnitude is that of I itself,
namely #I. This may be taken to be measured by the bulk volume of
the literature of the field.
A second characteristic level of importance is that afforded by our
now familiar standard of paramount quality, “first-rate” information
with λ = 1. Here we have Gibbon’s Law: #K(I) = log (#I).
A third level of importance is rooted in the suggestion of Jean
Jacques Rousseau3 that the number of really important items within
a population stands as its square root. Here we have λ = 1/2 with the
following result:
√
Number of important items = total number of items.
3 “Rousseau curiously enough, argued that the size of the [political] elite (i.e., gov-
ernment) varied with the square root of the population, if I understand correctly.
This statement that is so frequently imputed to J. J. Rousseau seems to evade spe-
cific reference although its sense is apparent in his Contrat Social ” (George K. Zipf,
Human Behaviour and the Principle of Least Effort [Boston: Addison-Wesley, 1949],
p. 452 & n.).
P1: KAE
And there is good reason to think that this is how matters stand in
many contexts.4 Thus, for example, Francis Galton found that at a
time when the adult male population of England was some 9,000,000,
there were about 3,000 really eminent individuals (persons qualifying
as noteworthy by a variety of criteria on the order of featuring in the
obituary notices of the Times). Also, the square root of the population
of a country – or of a profession or other category – yields (to within a
reasonable approximation) the number of listings in a head count of
the top elite for the constituency at issue.5 Or, there are in the United
States roughly 1,600 educational institutions that grant a baccalaureate
√
degree, but only some 40(= 1,600) of them comprise the really
“major universities,” which produce three-quarters of the Ph.D.s and
the vast majority of scholarly books and papers.
These observations situate three particularly significant levels
of informative quality within the overall spectrum presented in
Display 6.1.
On the basis of these considerations, quality can be assessed straight-
forwardly in terms of quantity: more is less. But of course what is at
issue here is not so much a discovery about quality as a definition of
those K λ -quality levels. In putting a number to the volume of high-
quality findings, the question remains: which is the dependent and
which the independent variable? Do we have standards first and then
numbers or the other way round? And here our present approach is
essentially that of the second alternative.
Quality Retardation
It is obvious that information of high quality will always be less in
quantity than information at large. But what is far from obvious – and
perhaps surprising – is that the growth of high-quality information is not
proportional to that of information at large, but is distinctly slower; moreover,
this rate of growth decelerates as the bar of qualification at issue with “high
quality” is raised. This quantitative slowing of growth in the wake of
4 Derek J. Price, Little Science, Big Science (New York: Columbia University Press, 1963),
pp. 33–36, provides a sketch of Galton’s findings.
5 However, the 2003–04 Who’s Who in America lists 79,000 individuals at a time when
the population of the country stands at some 280 million (with a square root of ca.
5,300). This exaggeration by a factor of 15 the Rousseau’s law standard is typical of
national Who’s Whos. No doubt, the editors of such works are inclined for commercial
reasons to exaggerate who is a who.
P1: KAE
Quality Retardation 63
Cumulative number of
Quality level (K λ ) items of at least Doubling time Per annum
that level growth rate
λ = 1: “routine” I = I 15 years 5%
.75
λ = .75: “significant” I 20 years 2.75%
.5
√
λ = .5: “important” I = I 30 years 2.5%
display 6.3. The principle of quality retardation: The decreasing rate growth
of scientific information at increasing quality levels. Note: This relation between
quantity and quality is predicated on taking I as the scientific literature at large
which has been growing at a doubling time of some 15 years throughout the
modern era.
qualitative escalation may be characterized as quality retardation. The

structure of its phenomenology is exhibited in Display 6.3.6
There are many illustrations of this phenomenon. For instance,
during the 20th century, memberships in the American Physical Soci-
ety grew from some 100 in 1900 to over 40,000 in 2000, doubling
roughly every 12 years. The fellows of the society who represent its
more distinguished, elite members, while also increasing in numbers,
represented 100 percent of its membership in the early years but only
one-eighth thereof at the end of the century. Then too, while the lit-
erature in genetics has increased exponentially during the twentieth
century with a doubling time of some 15 years, the “significant discov-
eries” in the field as inventoried by Bentley Glass has merely increased,
with a doubling time of some 20 years.7
On this basis, it is a significant lesson of the Principle of Quality
Retardation that the increase of significant information lags substan-
tially behind the increase of information at large. Given an information
manifold I, the basic relation #K λ [I] = (#I) λ leads straightforwardly to
#K λ [2 × I ] = #(2 × I )2 = K λ [I ] × 2λ .
6 On the general situation see the author’s Scientific Progress (Oxford: Basil Blackwell,
1978), especially Chapter 6, “The Quantity of Quantity,” pp. 95–111. For the empir-
ical substantiation of this effect see Roland Wagner–Döbler, “Rescher’s Principle
of Decreasing Marginal Returns for Scientific Research,” Scientometrics, 50 (2001),
pp. 419–36.
7 Bentley Glass, “Milestones and Rates of Growth in the Development of Biology,”
Quarterly Review of Biology, 54 (March 1979), pp. 31–53.
P1: KAE
Thus, with information at the λ = 1/2 level of “importance, we have

√
K 1/ [2 × I ] = 2 × K 1/ [I ]. In doubling the information we have
2 2
increased its informativity important component by a factor of only
√ ∼
2 = 1.41, so that in doubling information we have amplified our
“knowledge” not by another unit of 1 but merely by 0.41. And the fur-
ther step to 3I would carry us further by only 0.32. We thus encounter
an instance of the generic situation that the amplification of infor-
mation is subject to diminishing cognitive returns. Within any given
course of congnitive growth, the rate of progress decreases drastically
as one raises the standard of quality: the higher one sets the bar, the
slower the advance.
Elites
All of the standards of the “knowledge” at issue with quality informa-
tion, with λ > 0 are elitist in a manner suggestive of “lion’s share”
possession. We often hear that the richest 5 percent of the American
population owns 70 percent of the nation’s privately held wealth; or
again, that the top 25 percent of university libraries account for 50 per-
cent of all university library holdings in the United States. In this vein,
consider the eliteness function defined by the following condition:
E(x) = y iff x% of the overall population at issue accounts for y% of the whole
of some parameter.
Let us define the “haves” subgroup of a population as the one that

accounts for 50 percent of the whole “ownership” of the paramet-
ric quantity at issue. With a “fair” deliberation, the haves will be just as
numerous as the have-nots. However, with what we shall call an “elitist”
distribution, this relationship functions like the curve of Display 6.4.
As this representation indicates, with an elitist distribution a compar-
atively small proportion of the population will have a lion’s share of
the whole.
A vivid illustration of this phenomenon is “Lotka’s Law” to the effect
that if k is the number of scientists who publish just one paper, then
the number publishing n papers is k/n 2 . In many branches of science
this works out to some 5–6 percent of scientists producing half of all
papers in their discipline, putting the subgroup of the haves at about
one-twentieth of the whole.8
8 On Lotka’s Law see Derek Price, Science since Babylon (New Haven: Yale University
Press, 1961; 2nd ed. 1975), pp. 174–75. For further information see Paul A. David,
P1: KAE
Elites 65
an elitist distribution
100
% of the 50 an egalitarian
whole distribution
0 50 100
% of P
display 6.4. An elitist distribution. Note: The straight line represents a perfect
egalitarianism, while the curve represents an elitist distribution of sorts.
E0 = P, the overall population

E1 = 10% of E0 = .1 × P
E2 = 10% of E1 = .01 × P
Ei = 10% of E2−1 = (.1)i × P
display 6.5. Nested n-th order proportionality elites.
In this regard, the idea of rank importance paves the way to what
may be called the Proportional Law of Elites. The guiding idea here is
that the comparative size of an elite (at a certain level of exclusivity)
is to be determined by a specified proportion or percentage of the
population at issue. Thus with an overall population of size P and
an elitism proportion of 10 percent we would have the sizes for the
nested sequence of successively higher, “more exclusive” elites given in
Display 6.5. Here every step onward in the eliteness sequence reduces
the group at issue by an order of magnitude. And this is only to be
expected since diminished quantity and increased quality go hand in
hand.
The situation depicted in Display 6.5 leads to the Proportional Law
of Elites (as we shall call it), which specifies the overall size of succes-
sively “exclusive” elites. In general, with a population of size P there
will be a series of logP one-tenth proportionality eliteness levels, with
each such level E i containing (.1) i × P members. Overall, elitist hier-
archies are by nature reflective of an exponential decay. (An army
will always have far fewer generals than privates.) And with any elitist
“Positive Feedback and Research Productivity in Science,” in O. Granstrand (ed.), The

Economics of Technology (Amsterdam: Elsevier, 1994), pp. 65–89.
P1: KAE
distribution of a good in the manner of Display 6.4, an Ei -elite rather

far down the line in Display 6.5 will manage to comprise all the “haves”
there are. Thus in scientific publications, for example, some 10 percent
of the articles will engross 50 percent of citation references, so that
the E 1 elite effectively constitutes the “haves” in point of professional
recognition.9
Cognitive Importance as Reflected in Space Allocation

What form will the idealized systematization of a body of knowledge
take? A promising answer is that it is a matter of “perfected textbooks” –
ideally perfected systemic accounts. After all, information gets orga-
nized through textual exposition – for simplicity let us say by getting
put into print in books. And cognitive systems so construed are encom-
passed in (usually) well-formed and cogently organized symbolically
presented expositions which, in their turn, yield libraries, library sec-
tions, books, and then out to chapters, sections, and so on.
To illustrate this approach to systemic structuring of cognition, let’s
begin with a relevantly structured unit such as a scientific/technical
book or monograph along by now thoroughly familiar lines. Such a
treatise will of course be divided into chapters, sections, paragraphs,
and sentences, and for the sake of discussion, we may suppose a situ-
ation of the type sketched in Display 6.6. The result will be a more or
less standard modest-size book of some 105 = 100,000 words or 250
pages. And of course a complex topic will require many such volumes.
On this basis, it is instructive to look at the matter of importance
from the angle of the allocation of space in a well-designed articula-
tion of the issues. For an ideal exposition space allocation in expository
matters would of course reflect the importance of the matters involved.
And on this basis the importance of a cognitive unit such as an idea or
thesis or line of thought can be measured by the magnitude (book or
chapter or section) of the space devoted to it in an adequate exposition
of its domain. And, of course, our book analogy could be replicated
on a larger scale in such a sequence as cognitive domain, discipline,
substantive, specialty, problem area, problem, problem component.
Accordingly, if space – and thus, effectively, attention – is seen as
9 See H. W. Holub, G. Tappeiner, and V. Eberharter, “The Iron Law of Important
Articles,” Southern Economic Journal, 58 (1991), pp. 317–28.
P1: KAE
Cognitive Importance as Reflected in Space Allocation 67
1 book ∼ 100% (level 1)

1 chapter ∼ 10% (level 2)
1 section ∼ 1% (level 3)
1 paragraph ∼ .1% (level 4)
1 sentence ∼ .01% (level 5)
display 6.6. Space allocation for a hypothetical treatise. Note: In general, one
level n unit merits a space allocation proportional to 1000%
10n
= 10n−3 %. On this
basis we have log Sn ∝ n: the logarithm of the space allocated is an effective
index of importance levels.
Q0 1 big (first importance) idea for the book-as-a-whole.

Q1 10 sizeable (second importance) idea: one for each chapter.
Q2 100 moderately (third importance) idea: one for each section.
Q3 1,000 smallish (fourth importance) idea: one for each paragraph.
display 6.7. Quality levels. Note: In general,Q i = 10i or equivalently log Q i ∝ i.
importance-reflective, then we will have the upshot that our illustrative

book affords us ideas or findings at the level of magnitude indicated
in Display 6.7. As this relationship of quality and quantity suggests,
quality varies with the inverse of quantity, or put differently, eliteness
and quantity stand in a relationship of complimentarity.
Let’s suppose the organizational taxonomy structure of a book is
indicated by its table of contents – and the structure of a body of texts
by their taxonomic inventory. The comparative importance is repre-
sented by taxonomic depth (or rather, shallowness). So here structure
reflects significance. The further down an item figures on the list of
cumulating chapters, sections, or paragraphs, the lower the grade of
its importance becomes.
We thus arrive at what might be characterized as the ideal space-
allocation standard of importance. A scientific idea, concept, principle,
thesis, theory, finding, or fact is important exactly to the comparative
extent to which it merits space allocation in a perfected exposition of
its field.
So importance is plausibly viewed in the light of the idea of a perfected
textbook for the domain at issue. And importance will here be reflected
P1: KAE
in space allocation. To reemphasize: the crucial determinative factor

for increasing importance is the extent of seismic disturbance of the
cognitive terrain. Without the idea or concept of issue, would we have
to abandon and/or rewrite the entire textbook, or a whole chapter, or
a section, or a paragraph, or a sentence, or a mere footnote?
This, of course, puts the issue in the light of idealization. But we
can also look to the real and thus implement the distinction between
putative and actual importance. And here as elsewhere the actual may
be seen as affording at least an approximation of the ideal.
Importance is thus seen as reflected in a Perfected Scientific Library
when each domain of factual knowledge would be given its canonically
definitive systematization – its perfected account in terms of correct-
ness and completeness.10 However to implement this idea we must
deal with the reality of actual science libraries in place of that idealiza-
tion. We must, in short, take the Hegelian line that the real is rational,
and that the reality of things stands for that otherwise unattainable
idealization.
This being so, it should be clear that when we ourselves actually
engage in the business of attributing importance to facts and findings
we are providing estimates of importance. Importance for science as we
have it here-and-now is one sort of thing – namely, putative or estimated
and thus subjective importance – while real, objective importance is a
matter of how matters stand in ideal or perfected science. But given
that science-as-best-we-can-devise-it is more or less by definition our best
available estimate of science as it would ideally be developed, the two can be
viewed in practice as representing different sides of the same coin.
Cognitive Importance as Reflected in Citation

In relation to particular members of a population we can adopt the
following principle:
The importance ranking of a particular item is to be assessed by the index of

the smallest (most restrictive) elite to which it belongs.
10 This model of a Perfected Library is – to be sure – something very different from

the Borges Library contemplated by the Argentinean polymath Jorge Luis Borges.
For the Borges Library is universal: it deals not only with actuality but seeks to map
out the realm of possibility as well. Accordingly, the vast bulk of its holdings will be
works of fiction rather than of science. Our presently envisioned Perfected Scientific
Library, by contrast, concerns itself with fact alone, and leaves fiction aside.
P1: KAE
P
% of
articles
at least n citations of the articles

display 6.8. The citation distribution of articles. Note: In effect, we have a
pattern of exponential decay.
On this basis, the importance of rank-ordered items of any sort is pro-

portional to the logarithm of the sum total of at least equally important
items – a principle we have already seen at work with respect to knowl-
edge. The key to the quality-quantity relationship at issue here is pro-
vided by the work of George K. Zipf. His classic book, Human Behaviour
and the Principle of Least Effort,11 insisted that in a wide variety of circum-
stances the situation of Display 6.8 prevails. The relationship set out
here is linked to an equilateral hyperbola of the format s × log (#s) =
const. On this basis we can look at the situation in light of the Funda-
mental Principle of Quality discussed in relation to Display 6.2 above,
which in effect says measure quality by the logarithm of quantity.
Scientific importance is not a qualitative but a relational feature,
a function of how one item (fact or idea, etc.) relates to the others.
Specifically, when something is important, a lot else depends on its
being the way it is, and this is bound to be reflected in how much occa-
sion there is to have recourse to it in an adequate systematization of
the domain at issue. This approach inflects a fundamentally pragmatic
perspective. It views cognitive objects such as concepts, ideas, and the-
ories as tools. And with any sort of production process – be it physical
or cognitive – the importance of a tool lies in how much occasion one
has to make use of it.
And here issues of citation frequency and space allocation come in.
For starters, throughout scientific publication (as numerous structures
indicate), the logarithm of the number of papers cited n times graphs
11 Boston: Addison-Wesley, 1949.

P1: KAE
Quality level λ-value n-value Percentage of papers

(Kλ ) (for En ) of at least this level
(at a 50% reduction perlevel)
Routine 1 0 100
Significant .75 1 50
Important .5 2 25
Very important .25 3 12.5
display 6.9. The quantity of quality.
linearly against n, so it transpires that we can also use citation frequency

as a measure of quality. As a rough approximation, the fraction of the
number of papers in a scientific field, the number that will be cited n
times, is given by 12 , 14 , 18 , · · · ( 12 )n . . . respectively as n = 1, 2, 3, 4, . . . This
means that the percentage of papers that will be cited exactly n times
will stand at 50, 25, 12.5, 6.2, . . . respectively as n = 1, 2, 3, 4, . . . The
statistics of the situation invite us to associate citation frequencies with
the quality levels of Display 6.1 above, and on this basis we obtain the
situation of Display 6.9.12
These considerations provide another convenient bridge to the
qualification of knowledge. Setting out from the idea of a citation
network, we can assess the size of an item by the extent of its enmesh-
ment – the number of citation linkages with others. Now in view of
Zipf ’s findings,
The importance-reflective size of an item (in terms of the number of its link-
ages) is inversely proportional to the number of items of at least that quality.
But from the analysis of elites it emerges that

The quality of an item is inversely proportional to the logarithm of a number
of items of at least that quality.
On this basis we can conclude that
The quality of an item is proportionate to (and can thus be measured in terms

of) the number of its citation linkages.
12 For the empirical substantiation of this general pattern see the already cited publica-
tions of Derek Price as well as William Shockley, “On the Statistics of Productivity in
Research,” Proceedings of the Institute of Radio Engineeers, 45 (1957), pp. 279–90, and
William Goffin and Kenneth S. Warren, Scientific Information Systems and the Principle
of Solidarity (New York: Praeger, 1981).
P1: KAE
a straight line with

log #(s)
Log s + log #(s) = const
log s
display 6.10. Zipf’s law of rank order. Note: s = the size of an object #(s) =
the object’s place in the rank-order of size.
We thus arrive at the conclusion that the analysis of citation linkages

affords a suitable way of assessing the cognitive quality of individual
items of information. And given that “knowledge” (in its various
degrees) is a matter of the quality of the information at issue, this serves
to indicate the utility of citation analysis in measuring knowledge.
But now another line of consideration enters in. Many studies of cita-
tion statistics confirm the picture of Display 6.10 as holding through-
out science and its branches.13 The same drastic picture of exponential
decline in point of quantity obtains on both sides. And so the lesson
is clear: importance can just as effectively be estimated in terms of
prominence in citation space as by prominence in discussion space.
And so, to whatever extent we are prepared to see the actual space
allocations as an approximation of the ideal situation, we can under-
take the shift from seeming to actual importance.
13 A starting point is provided by the books of Derek J. Price, Science since Babylon (New
Haven: Yale University Press, 1975), and Little Science, Big Science (New York: Columbia
University Press, 1963). An updated source is H. W. Holub, Gottfried Tappeiner, and
Veronica Eberharter, “The Iron Law of Important Articles,” Southern Economic Journal,
58 (1991), pp. 317–28. The journal Scientometrics has published a great deal of relevant
material in recent years.
P1: KAE
72
P1: KAE
How Much Can Be Known?

A Leibnizian Perspective on the Quantitative
Discrepancy Between Linguistic Truth
and Objective Fact
(1) Propositional knowledge is a matter of textualization that hinges on linguistic

realizability. (2) As Leibniz realized, this imposes limits. (3)–(6) While statements
(and thus truths and realizable knowledge) are enumerable, actual objective
facts are not, for there is good reason to think that facts are inexhaustible and
nondenumerable. (7) Accordingly, there are more facts than truths: reality bursts
the bounds of textualization. (8)–(9) There are various facts that we finite beings
cannot manage to know. (10)–(11) There is no warrant for the view that reality
as such answers to what we can know of it.
How Much Can a Person Know? Leibniz

on Language Combinatorics
Gibbon’s Law holds that the extraction of knowledge from mere infor-
mation becomes exponentially more demanding in the course of cog-
nitive progress. However, while finite resources will doubtless impose
limits in practice, the process is one which in theory and principle goes
endlessly on and on. Does this mean that there really are no theoretical
limits to the enlargement of knowledge?
To this point, the book’s deliberations have dealt with knowledge
in relation to what people know. But above and beyond what is known
there is also the issue of what is knowable.
How much can someone possibly know? What could reasonably be
viewed as an upper limit to an individual’s knowledge – supposing
that factually informative knowledge rather than performative how-to
knowledge or subliminally tacit knowledge is at issue?
73
P1: KAE
74 Linguistic Truth and Objective Fact
In pursuing this question, it is convenient to continue along the

lines of our textual approach. Accordingly, let us suppose someone
with perfect recall who devotes a long life span to the acquisition of
information. For 70 years this individual spends 365 days per annum
reading for 12 hours a day at the rate of 60 pages an hour (with 400
words per page). That yields a lifetime reading quota of some 7.4 × 109
words. Optimistically supposing that, on average, a truth regarding
some matter of fact or other takes only some seven words to state, this
means a lifetime access to some 109 truths, around a billion of them:
1,000,000,000. No doubt most of us are a great deal less well informed
than this. But it seems pretty well acceptable as an upper limit for the
information that a human individual could probably not reach and
certainly not exceed.
After all, with an average of 400 pages per book, the previously indi-
cated lifetime reading quota would come to some 46,000 books. The
world’s largest libraries – the Library of Congress, for example – cur-
rently have somewhere around 20 million books (book-length assem-
blages of monographs and pamphlets included). And it would take a
very Hercules of reading to make his way through even one-quarter
of 1 percent of so vast a collection (= 50,000), which is roughly what
our aforementioned reading prodigy manages. So while a given indi-
vidual can read any book (there are no inherently unreadable books),
the individual cannot possibly read every book (for any one of us there
are bound to be very many unread books indeed). If mastery of Library
of Congress-encompassed material is to be the measure, then few of
us would be able to hold our heads up very high.1
All this, of course, still only addresses the question of how much
knowledge a given person – one particular individual – can manage to
acquire. There yet remains the question of how much is in principle
knowable – that is, can be known. And here it is instructive to begin
1 To be sure, there lurks in the background the question of whether having mere
information is to count as having knowledge. With regard to this quantitative issue it
has been argued here that authentic knowledge does not increase propositionally
with the amount of information as such, but only proportionally with its logarithm.
This would suggest that the actual knowledge within the Library of Congress’s many
volumes could be encompassed telegraphically in some far more modest collection,
so that our Herculean reader could access about half of the actual knowledge afforded
by the LC’s vast collection.
P1: KAE
Leibniz on Language Combinatorics 75
with the perspective of the great 17th-century polymath G. W. Leibniz

(1646–1717).
Leibniz took his inspiration from The Sand Reckoner of Archimedes,
who sought to determine the astronomically large number of sand
grains that could be contained within the universe defined by the
sphere of the fixed stars of Aristotelian cosmology – a number
Archimedes effectively estimated at 1050 . Thus even as Archimedes
addressed the issue of the scope of the physical universe, so Leibniz
sought to address the issue of the scope of the universe of thought.2 For
just this is what he proceeded to do in a fascinating 1693 tract On the
Horizon of Human Knowledge (De l’horizon de la doctrine humaine).3
Leibniz pursued this project very much in the spirit of the preceding
ideas. He wrote:
All items of human knowledge can be expressed by the letters of the

alphabet . . . so that it follows that one can calculate the number of truths of
which humans are capable and thus compute the size of a work that would
contain all possible human knowledge, and which would contain all that could
ever be known, written, or invented, and more besides. For it would contain
not only the truths, but also all the falsehoods that men can assert, and mean-
ingless expressions as well.4
Thus if one could set an upper limit to the volume of printed matter
accessible to inquiring humans, then one could map out by combina-
torial means the whole manifold of accessible verbal material – true,
false, or gibberish – in just the manner that Leibniz contemplated.
Any alphabet devisable by man will have only a limited number of
letters (Leibniz here supposes the Latin alphabet of 24 which takes
w and W ). So even if we allow a word to become very long indeed
2 On Archimedes’ estimate see T. C. Heath, The Works of Archimedes (Cambridge: Cam-

bridge University Press, 1897).
3 See G. W. Leibniz, De l’horizon de la doctrine humaine, ed. by Michael Fichant (Paris:
Vrin, 1991). There is a partial translation of Leibniz’s text in “Leibniz on the Limits
of Human Knowledge,” by Philip Beeley, The Leibniz Review, 13 (December 2003),
pp. 93–97. (Note that in old French “doctrine” means knowledge.) It is well known
that Leibniz invented entire branches of science, among the differential and integral
calculatus, the calculus of variations, topology (analysis situs), symbolic logic, and
computers. But he deserves to be seen as a pioneer of epistemetrics as well. The rel-
evant issues are analyzed in Nicholas Rescher, “Leibniz’s Quantitative Epistemology,”
Studia Leibnitiana, (in press).
4 Leibniz, De l’horizon, pp. 37–38.
P1: KAE
(Leibniz overgenerously supposes 32 letters)5 there will be only a lim-

ited number of words that can possibly be formed (namely 24 exp
32). And so, if we suppose a maximum to the number of words that a
single run-on, just barely intelligible sentence can contain (say 100),
then there will be a limit to the number of potential “statements” that
can possibly be made, namely 100 exp (24 exp 32).6 This number is
huge indeed – far bigger than Archimedes’ sand grains. Nevertheless,
it is still finite, limited. Moreover, with an array of basic symbols dif-
ferent from those of the Latin alphabet, the situation is changed in
detail but not in structure. (And this remains the case even if one adds
the symbols at work in mathematics, and here Leibniz was thinking of
Descartes’s translation of geometrically pictorial propositions into an
algebraically articulated format as well as his own project of a universal
language and a calculus ratiocinator.)7
The crux of Leibniz’s discussion is that any fact that can be framed
in a proposition can in principle be spelled out in print. And there
is only so much, so finitely much, that can be stated in sentences of
intelligible length – and therefore only so much that can explicitly be
thought by beings who conduct their thinking in language. Moreover,
because this speculation encompasses fiction as well, our knowledge
of possibility is also finite, and fiction is for us just as much language-
limited as is the domain of truth.
5 The longest word I have seen in actual use is the 34-letter absurdity supercalifragilistic-
expialidocious from the musical Mary Poppins.
6 G. W. Leibniz, De l’horizon, p. 11. This of course long antedates the (possibly apoc-
ryphal) story about the Huxley-Wilberforce debate, which has Huxley arguing that
sensible meaning could result from chance process because a team of monkeys typ-
ing at random would eventually produce the works of Shakespeare – or (on some
account) all the books in the British Library, including not only Shakespeare’s works
but the Bible as well. (The story – also found in Sir Arthur Eddington’s The Nature of
the Physical World [London: McMillan, 1929], pp. 72–73 – is doubtless fictitious since
the Huxley-Wilberforce debate of 1860 antedated the emergence of the typewriter.)
However, the basic idea goes back at least to Cicero: “If a countless number of the
twenty-one letters of the alphabet . . . were mixed together it is possible that when cast
on the ground they should make up the Annals of Ennius, able to be read in good
order” (De natura deorum, II, 27). The story launched an immense discussion that
continues actively on the contemporary scene as is readily attested by a Google or
Yahoo search for “typing monkeys.” It has also had significant literary repercussions
as is exemplified by Jorge Luis Borges’s well-known story of “The Library of Babel,”
which contains all possible books.
7 Louis Couturat, La logique de Leibniz (Paris: Alcan, 1901) is still the best overall account
of this Leibnizian project.
P1: KAE
The Leibnizian Perspective

The moment one sets a realistic limit to the length of practicably mean-
ingful sentences one has to realize that the volume of the sayable is
finite – vast though it will be. And this means that as long as people
transact their thinking in language – broadly understood to encom-
pass the whole diversity of symbolic devices – the thoughts they can
have, and thereby the things they possibly can know, will be limited in
number.
Moving further along these lines, the cognitive (in contrast to the
affective) thought-life of people consists of the language-framed propo-
sitions they consider. And let us suppose that people can consider tex-
tualized propositions at about the same speed at which they can read –
optimistically, some 60 pages per hour with each page consisting of 20
sentences. Assuming a thought-span of 16 waking hours on average, in
the course of a year a person can entertain a number of propositional
thoughts equal to
365 × 16 × 60 × 20 ∼
= 7 × 106 .
So, subject to the hypotheses at issue, this is how much material one
would need if he were to replicate in print the stream of consciousness
thought–life of a person for an entire year. Once again, this number
of seven million, though not small, is nevertheless limited. And these
limits will again finitize the combinatorial possibilities. A person can
manage only so much thinking. And in the context of a finite species,
these limits of language mean that there are only so many thoughts
to go around – only so many manageable sentences to be formulated.
Once again we are in the grip of finitude.
Now as Leibniz saw it, matters can be carried much further. For
the finitude at issue here has highly significant implications. Con-
sider an analogy. Only a finite number of hairs will fit on a person’s
head – say 1,000. So when there are enough individuals in a group
(say 1,001 of them) then two of them must have exactly the same num-
ber of hairs on their heads. And so also with thoughts. Even as laws
have to fit within the available dermatology, so they will have to fit
within the available textuality. If there are sufficiently many thinking
intelligences in the aeons of cosmic history while yet the number of
thoughts – and thus also thought-days and thought-lives – are finite,
P1: KAE
then there will inevitably be several people in a sufficiently large linguis-

tic community whose thoughts are precisely the same throughout their
lives.
It also becomes a real prospect that language imposes limits on
our grasp of people and their doings. Thus suppose that the Detailed
Biography of a person is a minute-by-minute account of his or her
doings, allocating (say) 10 printed lines to each minute, and so roughly
15,000 lines per day to make up a hefty volume of 300 fifty-line pages.
If a paradigmatic individual lives 100 years we will need 365 × 100
or roughly 36,500 such substantial tomes to provide a comprehensive
step-by-step account of his or her life.
But, as we have seen, the number of such tomes, though vast, is
limited. In consequence, there are only so many Detailed Biographies
to go around, so the number of Detailed Biographies that are available
is also finite. This, of course, means that if the duration of the species were
long enough – or if the vastness of space provided sufficiently many
thinkers – then there would have to be some people with exactly the same
Detailed Biography. Given enough agents, eventual repetitions in point
of their doings become inevitable.
And now, moving on from biographies (or diaries) to public annals,
Leibniz thought to encounter much the same general situation once
again. Suppose that (as Leibniz has it) the world’s population is one
hundred million (that is 108 ) and that each generation lives (on aver-
age) for 50 years. Then in the 6,000 years during which civilized man
may be supposed to have existed, there have lived some 1.2 × 1010
people – or some 1010 of them if we assume smaller generations in
earlier times.8
Recall now the above-mentioned idea of 36,500 hefty tomes needed
to characterize in detail the life of an individual. It then follows that
we would need some 36.5 × 1013 of them for a complete history of the
species. To be sure, we thus obtain an astronomically vast number of
possible overall annals for mankind as a whole. But though vast, this
number will nevertheless be finite. And so, if the history of the race
is sufficiently long, then some part of its extensive history will have
to repeat itself in full with a parfaite repetition mot pour mot, since there
are only so many possible accounts of a given day (or week or year).
8 Leibniz, De l’horizon, p. 112.

P1: KAE
For, once again, there are only a finite number of possibilities to go

around, and somewhere along the line total repetitions will transpire
and life stories will occasionally recur in toto (ut homines novi eadem ad
sensum penitus tota vita agerent, quae alii jam egerunt).9
As Leibniz thus saw it, the finitude of language and its users carries
in its wake the finitude of possible diaries, biographies, histories –
you name it, including even possible thought-lives in the sense of
propositionalized streams of consciousness as well. Even as Einstein
with his general relativity (initially) saw himself as finitizing the size of
the physical universe, so Leibniz’s treatise saw the size of mankind’s
cognitive universe as a manifold of limited horizons – boundless but
finite.
It was accordingly a key aspect of Leibniz’s thought that human
understanding cannot keep up with reality. For Leibniz, the proposi-
tional thought of finite creatures is linguistic and thereby finite and
limited. But he also held that reality – as captured in the thought of
God, if you will – is infinitely detailed. Only God’s thought can encom-
pass it, not ours. Reality’s infinite detail thus carries both costs and
benefits in its wake. Its cost is the unavoidability of imperfect compre-
hension by finite intelligences. Its benefit is the prospect of endless
variability and averted repetition. And the result is a cognitively insu-
perable gap between epistemology and metaphysics. Everything that
humans can say or think by linguistic means can be comprehended in
one vast but finite Universal Library.10
But what do these Leibnizian ruminations mean in the larger
scheme of things?
Twentieth-century philosophers of otherwise the most radically dif-
ferent orientation have agreed on prioritizing the role of language.
“The limits of my language set the limits of my world” (“Die Grenzen
meiner Spache bedeuten die Grenzen meiner Welt”) says the Wittgen-
stein of the Tractatus at 5.6. “There is nothing outside text” (“Il n’y a
pas de hors de texte”) say the devotees of French deconstructionism.
But already centuries earlier Leibniz had taken the measure of this
sort of textualism. He looked at it up close and saw that it could not
be sustained.
9 Ibid., p. 54.
10 This of course leads back to the Borges Library of footnote 10 of Chapter 6.
P1: KAE
Statements Are Enumerable, As Are Truths

The preceding deliberations have unfolded on the basis of the emphat-
ically contingent supposition that there are certain limits to human
capabilities – and, in particular, to the length of the words and sen-
tences with which our discourse can effectively operate. But let us
now also waive this (otherwise surely realistic) restriction and break
through the limits of finitude in the interests of getting a grip on the
general principles of the matter.
Even if one construes the idea of an “alphabet” sufficiently broadly
to include not only letters but symbols of various sorts, it still holds that
everything stateable in a language can be spelled out in print through
the combinational concatenation of some sequential register of sym-
bols.11 And with a “language” construed as calling for development
in the usual recursive manner, it transpires that the statements of a
language can be enumerated in a vast and indeed infinite but never-
theless ultimately countable listing.12 But since the world’s languages,
even if not finite in number, are nevertheless at most enumerable, it
follows that the set of all statements – including every linguistically
formable proposition – will be enumerably infinite (and thus have the
transfinite cardinality that mathematicians designate as alef-zero).
As a matter of principle, then, we obtain
Thesis 1: The Enumerability of Statements. Statements (linguistically formulated

propositions) are enumerable and thus (at most) denumerably infinite.
Our linguistic resources for describing concrete states of affairs are

thus subject to quantitative limitation. And insofar as our thought
about things proceeds by recursively developed linguistic means, it
is inherently limited in its reach within the confines of countability.
And so the upshot is that the limits of textuality impose quantitative
limitations on propositionalized thought – albeit not limits of finitude.
Being inherently linguistic in character, truths are indissolubly
bound to textuality, seeing that any language-framed declaration can
be generated recursively from a sequential string of symbols – that is,
11 Compare Philip Hugly and Charles Sayward, “Can a Language Have Indenumerably
Many Expressions?” History and Philosophy of Logic, 4, 1983, pp. 98–117.
12 This supposes an upper limit to the length of intelligible statements. And even if this
restriction were waived, the number of statements will still be no more than countably
infinite.
P1: KAE
The Inexhaustibility of Fact 81
that all spoken language can in principle be reduced to writing. Since

they correspond to statements, it follows that truths cannot be more
than countably infinite. And on this basis we have
Thesis 2: The Denumerability of Truth. While the manifold of the truth cannot
be finitely inventoried, nevertheless, truths are no more than denumerably
infinite in number.
Truths versus Facts

We serve the interests of clarity to introduce a distinction at this stage –
that between truths and facts. Truths are linguistically stated facts,
correct statements, in sum, which, as such, must be formulated in
language (broadly understood to include symbols systems of various
sorts). A “truth” is something that has to be framed in linguistic/symbolic
terms – the representation of a fact through its statement in some
language, so that any correct statement represents a truth.
A “fact,” on the other hand, is not a linguistic item at all, but an
actual aspect of the world’s state of affairs and is thereby a feature
of reality.13 Facts correspond to potential truths whose actualization as
such waits upon their appropriate linguistic embodiment. Truths are
statements and thus language bound, but facts outrun linguistic limits.
Once stated, a fact yields a truth, but with facts at large there need in
principle be no linguistic route to get from here to there.
The Inexhaustibility of Fact

Accordingly, facts need not be exhausted by truths. It is a key facet of
our epistemic stance toward the real world that its furnishings possess
a complexity and diversity of detail so elaborate that there is always
more to be said than we have so far managed. Every part and parcel of
reality has features beyond the range of our current cognitive reach –
at any juncture whatsoever.
Moreover, any adequate account of inquiry must recognize that the
process of information acquisition at issue in science is a process of
13 Our position thus takes no issue with P. F. Strawson’s precept that “facts are what
statements (when true) state.” (“Truth,” Proceedings of the Aristotelian Society, Supple-
mentary Vol. 24, 1950, pp. 129–56; see p. 136.) Difficulty would ensue with Strawson’s
thesis only if an “only” were added.
P1: KAE
conceptual innovation. In consequence, the ongoing progress of scien-

tific inquiry always leaves various facts about the things of this world
wholly outside the conceptual realm of the inquirers of any particular
period. Caesar did not know – and in the state of the cognitive art of
his time could not have known – that his sword contained tungsten
and carbon. There will always be facts about a thing that we do not
know because we cannot even express them in the prevailing concep-
tual order of things. To grasp such a fact means taking a perspective
of consideration that as yet we simply do not have, because the state
of knowledge (or purported knowledge) has not reached a point at
which such a consideration is feasible. And so, the facts about any actual
physical object – are in theory inexhaustible.
Its susceptibility to further elaborate detail – and to changes of mind
regarding this further detail – is built into our very conception of a
“real thing.” The range of fact about anything real is thus effectively
inexhaustible. There is, as best we can tell, no limit to the world’s
ever-increasing complexity that comes to view with our ever-increasing
grasp of its detail. The realm of fact and reality is endlessly variegated
and complex. And so we also arrive at
Thesis 3: The Inexhaustibility of Fact. Facts are infinite in number. The domain
of fact is inexhaustible: there is no limit to facts about the real.
In this regard, however, real things differ in an interesting and

important way from fictive ones. For a key about fictional particulars
is that they are of finite cognitive depth. In characterizing them we
shall ultimately run out of steam as regards their nongeneric features.
A point will always be reached when one cannot say anything fur-
ther that is characteristically new about them – presenting nongeneric
information that is not inferentially implicit in what has already been
said.14 New generic information can, of course, always be forthcoming
through the progress of science: when we learn more about coal in
general, then we also know more about the coal in Sherlock Holmes’s
grate. But the finiteness of its cognitive depth means that the prospect
14 To deny inferentially implicit information the title of authentic novelty, is not, of

course, to say that it cannot surprise us in view of the limitations of our own deductive
powers.
P1: KAE
of ampliatively novel nongeneric information must by the very nature

of the case come to a stop when fictional things are at issue.
With real things, on the other hand, there is no reason of principle
why the elaboration of nongenerically idiosyncratic information need
ever end. On the contrary, we have every reason to presume real things
to be cognitively inexhaustible. The prospect of discovery is open-
ended here. A precommitment to description-transcending features –
no matter how far description is pushed – is essential to our conception
of a real thing.
The detail of the real world is inexhaustible: obtaining fuller infor-
mation about its constituents is always possible in principle – though
not of course in practice, since only a finite number of things have
actually been said up to now – or indeed up to any actually realized
moment of world history. Something whose character was exhaustible
by linguistic characterization would thereby be marked as fictional
rather than real.15
And so we have it that facts regarding reality are infinite in number.
But just how infinite?
Facts Are Transdenumerable

While statements in general – and therefore true statements in partic-
ular – can be enumerated, and truths are consequently denumerable
in number, there is good reason to suppose that this will not hold for
facts. On the contrary, there is every reason to think that, reality being
what it is, there will be an uncountably large manifold of facts.
The reality is that facts, unlike truths, cannot be enumerated: no
listing of fact-presenting truths – not even one of infinite length – can pos-
sibly manage to constitute a complete register of facts. Any attempt to reg-
ister fact-as-a-whole will founder: the list is bound to be incomplete
because there are facts about the list-as-a-whole which no single entry
can encompass.
15 This also explains why the dispute over mathematical realism (Platonism) has little
bearing on the issue of physical realism. Mathematical entities are akin to fictional
entities in this – that we can only say about them what we can extract by deductive
means from what we have explicitly put into their defining characterization. These
abstract entities do not have nongeneric properties since each is a “lowest species”
unto itself.
P1: KAE
We thus arrive at the next principal thesis of these deliberations:
Thesis 4: The Transdenumerability of Facts. The manifold of fact is transdenu-

merably infinite.
The idea of a complete listing of all facts is manifestly impracticable.

Consider the following statement: “The list F of stated facts fails to have
this statement on it.” But now suppose this statement is on the list. It
clearly does not state a fact, so the list is after all not a list of the facts
(contrary to hypothesis). And so it must be left off the list. But then
that list will not be complete because the statement is true. Facts, that
is to say, can never be listed in toto because there will always be further
facts – facts about the entire list itself – that a supposedly complete list
could not manage to register.
This conclusion can be rendered more graphic by the following
considerations. Suppose that the list F
F : f1 , f2 , f3 , . . .
were to constitute a complete enumeration of all facts. And now consider

the statement
(Z )the list F takes the form f 1 , f 2 , f 3 , . . .
By hypothesis, this statement will present a fact. So if F is indeed a

complete listing of all facts, then there will be an integer k such that
Z = fk .
Accordingly, Z itself will occupy the k-th place on the F listing, so that
f k = the list L takes the form f 1 , f 2 , f 3 , . . . f k , . . .
But this would require fk to be an expanded version of itself, which is

absurd. With the k-th position of the F listing already occupied by fk we
cannot also squeeze that complex fk -involving thesis into it.
The crux here is simply that any supposedly complete listing of facts
f1 , f2 , f3 . . .
will itself exhibit, as a whole, certain features that none of its individual
members can encompass. Once those individual entries are fixed and
P1: KAE
the series is defined, there will be further facts about that series-as-a-
whole that its members themselves cannot articulate.
Moreover, the point at issue can also be made via an analogue of the
diagonal argument that is standardly used to show that no list of real
numbers can manage to include all of them, thereby establishing the
transdenumerability of the reals. Let us begin by imagining a suppos-
edly complete inventory of independent facts, using logic to streamline
the purported fact inventory into a condition of greater informative
tidiness through the elimination of inferential redundancies, so that
every remaining item adds some information to what has gone before.
The argument for the transdenumerability of fact can now be devel-
oped as follows. Let us suppose (for the sake of reductio ad absurdum
argumentation) that the inventory
f1 , f2 , f3 , . . .
represents our (nonredundant but yet purportedly complete) listing of

facts. Then by the supposition of factuality we have (∀i)fi. And further
by the supposition of completeness we have it that
(∀ p)( p → (∃i)[ f i → p]).
Moreover, by the aforementioned supposition of non-redundancy, each

member of the sequence adds something quite new to what has gone
before.
(∀i)(∀ j )[i < j →∼ [( f 1& f 2 & . . . & f i) → f j)].
Consider now the following course of reasoning.
(1) (∀i) f i by “factuality”

(2) (∀ j )f j → (∃i)( f i → (∀ j ) f j )from (1) by “completeness” via the
substitution of (∀j)fj for p
(3) (∃i)( f i → (∀ j ) f j ) from (1), (2)
But (3) contradicts non-redundancy. This reductio ad absurdum of our

hypothesis indicates that facts are necessarily too numerous for com-
plete enumeration.
P1: KAE
More Facts Than Truths

In such circumstances, no purportedly comprehensive listing of truths
can actually manage to encompass all facts. The long and short of it is
that the domain of reality-characterizing fact inevitably transcends the
limits of our capacity to express it, and a fortiori those of our capacity to
canvass completely. The realm of fact is endlessly complex, detailed,
and diversified in its makeup. And the limitedness of our recursively
constituted linguistic resources thus means that our characterizations
of the real will always fall short.16 We arrive at
Thesis 5. There are quantitatively more facts than truths seeing that the facts are
too numerous for enumerabilty.
The basic reason the domain of fact is ampler than that of truth is
that language cannot capture the entirety of fact. It is not only possi-
ble but (apparently) likely that we live in a world that is not digital but
analogue and whose manifold of states of affairs is simply too rich to be
fully comprehended by our linguistically digital means. To all visible
appearances the domain of fact transcends the limits of our capacity to
express it, and a fortiori those of our capacity to canvass it. In confronting
any landscape in nature, our representation of it in propositional dis-
course or thought – our description-scape, so to speak – is invariably
far less complex than the real scene and inevitably suppresses a vast
amount of detail. (Even the physics of discrete quanta requires con-
tinuous – and thus nondiscrete – parameters for its characterization.)
Truth is to fact what film is to reality – a merely discretized approx-
imation. Cognition, being bound to language, is digital and sequen-
tially linear. Reality, by contrast, is analogue and replete with feedback
loops and nonsequentially systemic interrelations. It should thus not
be seen as all that surprising that the two cannot be brought into
smooth alignment. The comparative limitedness of truth that can
16 Even in matters of actual linguistic practice we find an embarrassing shortcoming

of words. The difficulty in adapting a compact vocabulary to the complexities of
a diversified world are betokened by the pervasive phenomenon of polysemy – the
contextualized pluralism of varied senses and differentiated uses of the same words in
different semantical and grammatical categories. On this phenomenon see Hubert
Cuyckens and Britta Zawada (eds.), Polysemy in Cognitive Linguistics (Amsterdam: John
Benjamins, 2003).
P1: KAE
be encapsulated in language points to an inevitable limitedness of

knowledge.
Noninstantiable Properties and Vagrant Predicates

In this context of cognitive limits the cognitive intractability of certain
temporalized issues has significant implications. Thus consider such
questions as those already mentioned:
r What is an example of a problem that will never be considered by
any human being?
r What is an example of an idea that will never occur to any human
being?
There are sound reasons of general principle (the potential infini-

tude of problems and ideas; the inherent finitude of human intel-
ligence) to hold that the items at issue in these questions do actu-
ally exist as problems that will never be considered or ideas that will
never occur. But it seems altogether plausible that other (nonhuman)
hypothetically envisionable intelligences could answer these questions
correctly.
However, we can also contemplate the prospect of globally intrac-
table questions – questions that, despite having correct answers, are
nevertheless such that nobody (among finite intelligences at least) can
possibly be in a position to answer them (in the strict sense described
at the outset). With issues of absolutely unrestricted generality there
will be many examples of this:
r an idea that has never occurred to anybody
r an occurrence that no one ever mentions
r a person who has passed into total oblivion
r a never-formulated question
r an idea no one any longer mentions
Yet while there undoubtedly are such items, they of course cannot
possibly be instantiated. Accordingly, the corresponding example-
demanding questions are inherently unanswerable insolubilia.
In all such cases, the particular items that would render a contention
of the format (∃u)Fu true are referentially inaccessible : to indicate any of
them individually and specifically as instantiations of the predicate at
P1: KAE
issue is ipso facto to unravel them as so-characterized items. And so, non-
instantiability itself is certainly not something that is noninstantiable:
many instances are available along the following lines:
r a never-stated contention (truth, theory, etc.)

r a never-mentioned topic (idea, object, etc.)
r a truth (a fact) no one has ever realized (learned, stated)
r someone whom everyone has forgotten
r a never-identified culprit
r an issue no one has thought about since the 16th century
In the light of these considerations, the positivist dogma that a ques-

tion can qualify as empirically meaningful only if it is in principle pos-
sible for a finite knower to answer it (correctly) becomes problematic.
Cognitive positivism is not a plausible prospect.
The conception of an applicable but nevertheless noninstantiable
predicate comes to the fore at this point. Such predicates are “vagrant”
in the sense of having no known address or fixed abode. Though they
indeed have applications, these cannot be specifically instanced – they
cannot be pinned down and located in a particular spot. Accordingly,
F is a vagrant predicate if (∃u)Fu is true while nevertheless Fu0 is false for every
specifically identified u0 .
For the sake of contrast, consider such further uninstitutable pred-

icates as these:
r a book no one has ever read

r a sunset never witnessed by any member of homo sapiens
Such items may be difficult to instantiate – but certainly not impossible.

The former could be instantiated by author and title; the latter by
place and date. In neither case will an instantiation unravel that item
as described. Being read is not indispensably essential to books, nor
being seen to sunsets: being an unread book or being an unwitnessed
sunset involves no contradiction in terms. By contrast, however, with
actual vagrancy, epistemic inaccessibility is built into the specification
at issue. Here we have predicates of such a sort that one can show on
the basis of general principles that there must be items to which they
P1: KAE
apply, while nevertheless one can also establish that no such items can
ever be concretely identified.17
The existence of vagrant predicates shows that applicability and
instantiability do not amount to the same thing. By definition, vagrant
predicates will be applicable: there indeed are items to which they
apply. However, this circumstance will always have to be something
that must be claimed on the basis of general principles; doing
so by means of concretely identified instances is, by hypothesis,
infeasible.
While vagrant predicates are by nature noninstantiable, we can
nevertheless use them to individuate items that we can never identify.
For instance, with regard to “the youngest unknown (i.e., never-to-be-
identified) victim of the eruption of Krakatoa,” one can make various
true claims about the so-individuated person – for example that he
or she was alive at the time of Krakatoa’s eruption. One can certainly
discuss that individual as such – that is, as an individual – even though,
by hypothesis, we cannot manage to identify him or her. Predicative
vagrancy thus reinforces the distinction between mere individuation
and actual identification.
If F is a variant predicate, then Fix has no true substitution instances:
Fa is in fact false for every specifiable value of a. But nevertheless (∃x)Fx
is going to be true. (The logic of vagrant predicates is emphatically not
“intuitionistic,” and the substitutional interpretation of the quantifiers
is certainly not going to work.)
It will not have escaped the acute reader that all of the demonstrably
unanswerable questions and unknowable facts at issue in the present
discussion relate to issues of cognition and knowledge themselves.
Clearly if this is the best that can be done in this direction an interesting
conclusion follows – namely, that unknowability automatically arises
with the emergence of knowers. In a universe without them there
could possibly be a total absence of in-principle unknowable facts. But
once finitely intelligent beings emerge in the world, unanswerable
questions and unknowable facts emerge inexorably in their wake. All
of those vagrant predicates involve some limitation of knowledge. For
17 Reference, to be sure, does not require identification. A uniquely characterizing

description on the order of “the tallest person in the room” will single out a particular
individual without specifically identifying him.
P1: KAE
an omniscient God, the phenomenon of vagrancy vanishes. But we, of

course, cannot claim to be in this position.
Musical Chairs Once More

It is instructive at this point to consider once more the analogy of
Musical Chairs. Of course, any individual player can/might be seated.
And the same goes for any team or group with one exception: the
whole lot. But since the manifold of knowable truth is denumerable
and the manifold of fact in toto is not, then (as in our Musical Chairs
example) the range of the practicable will not, cannot encompass the
whole. (And note then while a team of individuals is not an individual,
a complex of facts will nevertheless constitute a fact.)
With regard to language too we once again confront a Musical
Chairs situation. Conceivably, language-at-large might, in the abstract,
manage to encompass nondenumerably many instances – particu-
larly so if we indulge the prospect of idealization and resort to
Bolzano’s Saetze an sich, Frege’s denkerlose Gedanken, and the like. But
given the granular structure of a universe pervaded by atoms and
molecules, only a denumerable number of language-using creatures
can ever be squeezed into the fabric of the cosmos. And so the
realistically practicable possibilities of available languages are at best
denumerable.
When reality and language play their game of Musical Chairs, some
facts are bound to be left in the lurch when the music of language
stops. The discrepancy manifests itself in the difference between any
and every. Any candidate can possibly be accommodated. (We have
(∀x)♦(∃y)Syx.) But it is not possible to accommodate every candidate.
(We do not have ♦(∀x)(∃y)Syx.) The limits of knowledge are thus in the
final analysis quantitative. The crux of the problem is a discrepancy of
numbers. They root in the Musical Chairs Perplex – in the fact that the
realm of fact is too vast for the restrictive confines of propositionalized
language.
And this situation has important cognitive ramifications that are
brought to view by the following line of thought:
1. Everything there is – (and indeed even presumably everything

there possibly can be) – has an idiosyncratic property, some
P1: KAE
Musical Chairs Once More 91
feature, no doubt complex and perhaps composite, that holds

for it and it alone. (Metaphysical principle)
2. The possession of such a unique characteristic property cannot
obtain in virtue of the fact that the item at issue is of a certain
natural kind or generic type. It can only obtain in virtue of some-
thing appertaining to this item individually and specifically.
3. Accordingly, for anything whatsoever, there is a fact – that that
thing has that particular idiosyncratic property – that you can
know only if you can individuate and specify that particular
thing.
4. The inherent limitations of language mean that there are more
things that it is possible to individuate and specify.
The inevitability of unknown facts emerges at once from these consid-

erations of general principle.
The reality is that the domain of fact is ampler than the domain of
truth so that language cannot capture the entirety of fact. We live in a
world that is not digital but analogue and so the manifold of its states of
affairs is simply too rich to be fully comprehended by our linguistically
digital means.18 The domain of fact inevitably transcends the limits of
our capacity to express it, and a fortiori those of our capacity to canvass
it in overt detail. Truth is to fact what moving pictures are to reality –
a merely discretized approximation.
To be sure, the numerical discrepancy at issue with the Musical
Chairs Perplex does no more than establish the existence of unknown
facts. It does not got so far as to establish the existence of facts that
are unknowable, facts that cannot, as a matter of principle, possibly be
known. To see what can be done in this line we shall have to look at
matters in a different light.
There clearly is, however, one fact that is unstatable to language
and thereby unknowable by creatures whose knowledge is confined to
the linguistically formulatable. This is the grand mega-fact consisting
18 Wittgenstein writes “logic is not a body of doctrine, but a mirror-image of the world”
(Tractatus, 6.13). This surely gets it wrong: logic is one instrumentality (among others)
for organizing our thought about the world, and this thought is (at best and at most)
a venture in describing or conceiving the world and its modus operandi in a way that –
life being what it is – will inevitably be imperfect, and incomplete. And so any talk of
mirroring is a totally unrealistic exaggeration here.
P1: KAE
of the amalgamation of all facts whatever. For language-dependent

knowers can at most and at best have cognitive access to a denumerable
number of facts, whereas factuality itself in principle encompasses a
nondenumerable quantity.
And a very important point is at issue here. With Musical Chairs we
know that someone will be unseated, but we cannot (given the ordi-
nary contingencies) manage to say who this will be. And with facts,
which from a cognitive point of view reduplicate the Musical Chairs
situation, we also cannot manage to say which facts will be unknown.
For here too there is a lot of room for contingency. But there is one
very big difference. With Musical Chairs the totality of individuals,
while of course not reliable, does not combine to form a single unseat-
able mega-individual. But the totality of facts – which cannot possibly
be known – does indeed combine to form one grand unknowable
megafact.
So here indeed we have managed to individuate a particular un-
knowable fact, namely, the all-encompassing megafact. But of course
while we know that it is unknowable, we do not know what it is. We
have individuated but not identified it. So here, as elsewhere, the details
of our ignorance are hidden from our sight.
Just what does this mean in the larger scheme of things?
Coda: Against Cognitive Nominalism

If we are going to be realistic about it – in both the ordinary and the
philosophical sense of this term – then we still have to resist the temp-
tation of a nominalistic textualism that identifies the nature of reality
with what we can manage to know about it. However, we are also well
advised to avoid succumbing to the siren call of mysticism. Certainly
the present deliberations regarding unknowable fact do not lead to
this destination. Granted, given the limitations of language, there will
of course be things that have to go unsaid – facts with respect to what
words literally fail us. But there is no reason to join with Wittgen-
stein in characterizing this as “the mystical” (Tractatus, 6.522). These
considerations afford no automatic reason for thinking that what is
unsaid is going to be significantly different in kind from the things one
can say, no constraining reason to see them as somehow strange and
different – any more than those individuals who go unseated in Musical
P1: KAE
Appendix: Further Implications 93
Chairs need be strange and different. What makes for the unsayability
of these things is not their inherent ineffability but merely that there
just are too many of them. No doubt reality is stranger than we think.
But the ground of this circumstance will ultimately lie in the nature of
reality and not in the limitedness of language.
Appendix: Further Implications

It is worthwhile to note that the numerical discrepancy between truths
and facts that textuality imposes recurs time and again in other con-
texts, and in particular between these:
r names and entities

r statements and possibilities
r descriptions and objects
r novels and plots
r instructions and actions
r explanations and phenomena
The same disproportion between the verbal and the ontological realm
occurs throughout. While in each case the former is a verbalized place-
holder for the latter, there just are not enough of the former to go
around.
In particular, consider names. Of course everything is capable
of being named. Nothing is name resistant. We could (as someone
has quipped) simply name everything Charlie. The real question is
whether everything could have a unique name characteristic of itself
alone: an identifying name.
Now everything that has actually been identified could be named
via the specification: “the item identified in such and such a way.”
Or at least this would work if the identification process answered to
some verbalized formula. But even supposing this to be the case, the
question remains: Are there enough verbal/textual identifiers to go
around? Can everything that has an identity be individuated by verbal-
ized formulas?
And the answer is categorically negative. Select any language you
please – take your pick. As long as it – like any other human language –
is produced recursively it will only have countably many expressions
P1: KAE
(words, sentences, texts). But we know full well that the number of
objects is transdenumerable: uncountably infinite. (Think of the real
numbers, for example.) So there just are not enough names to encom-
pass everything. In musical chairs not everybody gets to be seated. In
reality not everything gets to be named.
Of course, things will stand differently if we radically revise the
concept of language. Thus if we are prepared to countenance a thing
language (rather than a word language) we could adopt the rule that
everything names itself. And then of course everything is at once nam-
able and named. But this sort of thing is clearly cheating.
And so while nothing is textually name-resistant and everything
is namable in the sense of being able, in principle, to bear a verbal
name, the possibility of realizing this prospect across the board – with
everything whatsoever actually bearing a name – is precluded by the
general principles of the situation.19
19 This Appendix has benefited from exchanges with C. Anthony Anderson.

P1: KAE
On the Limits of Knowledge

A Kantian Perspective on Cognitive Finitude
(1) It is in principle impossible ever to give an example of an unknowable fact.

(2) While universalizations do enable us to assert truths about nonsurveyable
totalities, such totalities nevertheless serve to demarcate us as cognitively finite
beings. (3) For general facts regarding an open-ended group will – when con-
tingent and not law-inherent – open the door to facts that are beyond the cogni-
tive grasp of finite knowledge. (4) Nevertheless, as Kant insightfully saw it, the
realm of knowledge – of ascertainable fact – while indeed limited, is nevertheless
unbounded.
Limits of Knowledge
The cognitive beings that will concern us here are language-dependent
finite intelligences. These by their very nature are bound to be imper-
fect knowers. For the factual information at their disposal by way of
propositional knowledge that something or other is the case will –
unlike practical how-to knowledge – have to be verbally formulated.
And language-encompassed textuality is – as we have seen – outdis-
tanced by the facts themselves. Just what is one to make of the numer-
ical disparity between facts and truths, between what is knowable in
theory and what our finite intelligences can actually manage to know?
Just what does this disproportion portend?
It means that our knowledge of fact is incomplete – and inevitably
so! – because we finite intelligences lack the means for reality’s compre-
hensive characterization. Reality in all its blooming buzzing complexity
is too rich for faithful representation by the recursive and enumerable
95
P1: KAE
96 A Kantian Perspective on Cognitive Finitude
resources of our language. We do and must recognize the limitations

of our cognition, acknowledging that we cannot justifiably equate fac-
ticity with what can explicitly be known by us through the resources of
language. And what transpires here for the circumstantial situation of
us humans obtains equally for any other sort of finite intelligence as
well. Any physically realizable sort of cognizing being can articulate –
and thus can know – only a part or aspect of the real.
The situation as regards knowing facts is akin to that of counting
integers – specifically in the following regards:1
1. The manifold of integers is inexhaustible. We can never come

to grips with all of them as particular individuals. Nevertheless –
2. Further progress is always possible: in principle we can always
go beyond whatever point we have so far managed to reach.
Nevertheless –
3. Further progress gets ever more cumbersome. In moving
onward we must be always more prolix and make use of increas-
ingly elaborate symbol complexes so that greater demands in
time, effort, and resources are unavoidable. Accordingly –
4. In actual practice there will be only so much that we can effec-
tively manage to do. The possibilities that obtain in principle
can never be fully realized in practice. However –
5. Such limitations nowise hamper the prospects of establishing
various correct generalizations about the manifold of integers
in its abstract entirety.
And a parallel situation characterizes the cognitive condition of all

finite intelligences whose cognitive operations have to proceed by a
symbolic process that functions by language. Inductive inquiry, like
counting, never achieves completeness. There is always more to be
done, and in both cases we can always do better by doing more. But it
also means that much will always remain undone – that we can never
do it all.
But are those unknown facts actually unknowable? The answer is
neither yes nor no. As already foreshadowed above, it all depends
on exactly how one construes the possibilistic matter of “knowability.”
1 We here take “counting” to be a matter of indicating integers by name – for example,

as “thirteen” or “13” – rather than descriptively, as per “the first prime after eleven.”
P1: KAE
Cognitive Finitude 97
Using Kxf to abbreviate “the individual x knows the fact f,” there will
clearly be two rather different ways in which the existence of an inher-
ently unknowable fact can be claimed, namely “Some fact is necessarily
unknown.”
(∃ f ) (∀x) ∼K x f or equivalently ∼(∀ f )♦(∃x)K x f
and “Necessarily, some fact is unknown”
(∃ f )(∀x) ∼K x f or equivalently ∼ ♦(∀ f )(∃x)K x f.
The difference in the quantifier placement in the preceding two for-

mulas is crucial when one contemplates the idea that all facts are know-
able. Now the first of these two contentions logically entails the second.
And this second is in the circumstances inevitable, there being more
facts than finite humans ever will or can know. However, the first,
stronger contention is clearly false. For as long as the nonexistence of
an omniscient God is not a necessary circumstance there can be no fact
that is of necessity unknown.
But of course even though there are – or may well be – unknow-
able facts (in the indicated sense of this term so that for some fact f
we have (∀x) ∼K x f ), such a fact can never be identified as such,
namely as a fact, since doing so is effectively to claim knowledge of
it. It is thus in principle impossible for us ever to give an example
of one of the unidentifiable facts even though there must be some.
And nothing more strikingly betokens the imperfection and limited-
ness of our knowledge than the ironic circumstance of the unelim-
inable incompleteness of our knowledge regarding our knowledge
itself, which arises because we do not – and cannot – know the details
of our own ignorance (i.e., cannot possibly know what it is that we do
not know).2
Cognitive Finitude
First the good news. Generalizations can of course refer to every-
thing. Bishop Butler’s “Everything is what it is and not another thing”
holds with unrestricted universality. And once continuous quantities
2 This discussion has profited from the constructive comments of several Pittsburgh
colleagues, including Jason Dickinson, Mickey Perloff, and Laura Ruetsche.
P1: KAE
are introduced, the range of inferentially available statements becomes

uncountable. “The length of the table exceeds x inches.” Once known,
this straightaway opens the door to uncountably knowable conse-
quences. And fortunately, a case-by-case determination is not gener-
ally needed to validate generalizations. We can establish claims about
groups larger than we can ever hope to inventory. Recourse to arbi-
trary instances, the process of indirect proof by reductio ad absurdum,
and induction (mathematical and scientific) all afford procedures for
achieving generality knowledge beyond the reach of an exhaustive
case-by-case check.
But will this always be so? Or are there also general truths whose
determination would require the exhaustive surveying of all specific
instances of a totality too large for our range of vision?
At this point our cognitive finitude becomes a crucial considera-
tion and the difference between finite and infinite knowers becomes
of fundamental importance and requires closer elucidation. For an
“infinite knower” need not and should not be construed as an omni-
scient knower – one from whom nothing knowable is concealed (and so
who knows, for example, who will be elected U.S. president in the year
2200). Rather, what is at issue is a knower who can manage to know in
individualized detail an infinite number of independent facts. (Such a
knower might, for example, be able to answer such a question as, “Will
the decimal expansion of π always continue to agree at some future
√
point with that of 2 for 100 decimal places?”) Finite knowers cannot
manage this sort of thing.
Finite knowers can, of course, know universal truths. After all, we
must acknowledge the prospect of inductive knowledge of general
laws; we will have it that a knower can unproblematically know – for
example – that “All dogs eat meat.”3 But what finite knowers can-
not manage is to know this sort of thing in detail rather than at the
level of generality. They cannot know specifically of every u in that
potentially infinite range that Fu obtains – that is, while they can know
collectively that all individuals have F, they cannot know distributively
3 To be sure, the prospect of inductively secured knowledge of laws is a philosophically

controversial issue. But this is not the place to pursue it. (For the author’s position
see his Induction (Oxford: Blackwell, 1980).)
P1: KAE
Surd Facts and Unknowability 99
of every individual that it has F – something they could not do without

knowing who they individually are.
So the issue now before us is that of the question of general truths
that can be known only by assessing the situation of an untractable
manifold of individual cases.
Surd Facts and Unknowability

One cannot, of course, provide concrete examples of facts that are
unknowable to finite knowers, seeing that a claim to factuality auto-
matically carries a claim to knowledge in its wake. However while we
cannot know specifically what is such a fact, we can certainly substanti-
ate the claim generally that there are such things. Let us consider this
situation more closely.
Given any collection of items there are two importantly different
kinds of general properties: those that all members of the collection
DO have in common, and those that all members of the collection
MUST have in common. The latter are the necessitated general features
of the collection, the former its contingently geared features. Thus that
all prime numbers greater than 2 are odd is a necessity-geared feature
of this set of primes. Or consider the set of all U.S. presidents. That
all of them are native born and that all of them are over 35 years
of age is a necessitated general feature of the collection in view of
our Constitution’s stipulations. However, that all were born outside
Hawaii will (if indeed true) be a contingently geared feature of the
collection that is nowise necessitated by the general principles of the
Constitution.
Now the crucial consideration for present purposes is that while
the necessary features of a collection must inhere in (and be derivable
from) the generalities that govern the collection at issue as a matter
of principle, its contingent features will be surd in that they cannot be
established on the basis of general principles.
Accordingly, a general fact F(C) regarding a collection C is surd if
whenever a set of premises entails this fact, then every item belong-
ing to C must secure specific individual mention in one or another
of these premises. We thus have it that F(C) represents a surd fact
about C if
P1: KAE
Whenever P1 , P2 , . . . Pn F (C ) then every member of C must receive specific

mention in one or another of P i.
And of course since F (C ) F (C ) this means that F(C) must itself be

so constructed as to involve a specific mention of all C-members. The
salient point, at any rate, is that such a feature can only be ascribed to
a collection on the basis of a comprehensive one-by-one examination
of its members.
For the sake of an example, let the object under consideration be
the material numbers (1, 2, 3, etc.) and make the set
S = {1, 3, 17}.
The feature of S represented by “All S-members are either odd or

even” is a non-surd feature of this set. (Given that integers are at issue,
we need not inspect its members to assure ourselves that this gen-
eralization obtains): that “all” can be construed collectively. But the
feature of S represented by “All S-members are < 20” is indeed a surd
feature of this set. Its characterizing S is a fact that cannot be assured
without inspecting each one of its members. Here, that “all” must be
construed distributively.
So when and if contingent generalities actually hold for a collec-
tion, this can be ascertained only through a case-by-case check of its
entire membership. And this means that finite knowers can never decisively
establish a surd/contingent general feature of an infinite collection. For when-
ever a generality holds for a collection on a merely contingent basis,
this is something that we finite intelligences can never determine with
categorical assurance because determination of such kind-pervasively
surd would thus require an item-by-item check, which is by hypothesis
impracticable for us.
It must also be presumed to be a fact that as long as the paper
exists, every issue of the New York Times will be such that the word
THE occurs more than five times on its front page. This is almost
certainly a fact. But since this assertion cannot be settled by general
principles (laws) but requires a case-by-case check it is generally char-
acterized as surd. And such a property of something is contingent:
it cannot be accounted for on the basis of the general principles at
issue.
P1: KAE
Consider now a set of objects of a certain sort S that is infinite or

interminably open-ended (lions, say, or sunrises at Acapulco). And
let P be a surd/contingent property of some S -item X which, while in
principle is applicable to S -members, is nevertheless unique to X – that
is, is such that no other S -member actually has P. Note that now this
uniqueness could only be determined on a case-by-case check across
the whole range of S. Accordingly, that X is unique within S in point of
P-possession is (by hypothesis) a truth that no finite intelligence could
ascertain, seeing that an item-by-item canvass of an infinite/indefinite
range is beyond its capacity. Such claims illustrate the prospect of truths
beyond the cognitive grasp of finite knowers.
Of course, “unknowably true” is a vagrant predicate in the sense of
the preceding chapter – one that has no determinate address in that
it admits of no identifiable instance. Instantiating this sort of thing
can be done only at the level of schematic generality and not that of
concrete instantiation. For while we can convince ourselves – for good
reason – that there indeed are such things, it nevertheless remains in
principle impracticable to provide examples of them.
Larger Lessons: Isaiah’s Law

Something about which we cannot possibly be mistaken is our belief
that we are beings who make mistakes. And analogously, one regard
in which our knowledge cannot be incomplete – something that it is
effectively impossible for us to overlook – is the fact of our cognitive
finitude itself, the realization that we do not actually “know it all.”
However, the situation regarding our cognitive limits is not quite as
bleak as it may seem. For even though the thought and knowledge of
finite beings is destined to be ever incomplete, it nevertheless has no
fixed and determinate limits. Return to our analogy. As to our counting
integers, there is a limit beyond which we never will get. But there is
no limit beyond which we never can get.
The line of thought operative in these deliberations was already
mooted by Kant:
[I]n natural philosophy, human reason admits of limits (“excluding limits,”
Schranken) but not of boundaries (“terminating limits,” Grenzen), namely, it
admits that something indeed lies without it, at which it can never arrive,
P1: KAE
but not that it will at any point find completion in its internal progress. . . .
[T]he possibility of new discoveries is infinite: and the same is the case with the
discovery of new properties of nature, of new powers and laws by continued
experience and its rational combination.4
Given the inevitable discrepancy – and numerical disproportion –

between our propositionally encodable information about the real
and the factual complexity of reality itself, we have to be cautious
regarding the kind of scientific realism we endorse. Any claims that
reality in its details is exactly as the science of the day claims it to
be is (irrespective of how the calendar reads) extremely questionable.
This is something that already became apparent early in the discussion
of the security/definiteness trade-off. And everything that the subse-
quent discussion has brought to light only serves to substantiate this
conclusion in greater detail.
Some writers analogize the cognitive exploration of the realm of
fact to the geographic exploration of the earth. But this analogy is
profoundly misleading. For the earth has a finite and measurable sur-
face, and so even when some part of it is unexplored terra incognita its
magnitude and limits can be assessed in advance. Nothing of the kind
obtains in the cognitive domain. The ratio and relationship of known
truth to knowable fact is subject to no fixed and determinable pro-
portion. Geographic exploration can expect eventual completeness;
cognitive exploration cannot.
All the same, there can be no doubt that ignorance exacts its price
in incomprehension. And here it helps to consider the matter in a
somewhat theological light. The world we live in is a manifold that is
not of our making but of Reality’s – of God’s, if you will.
4 Prolegomena to any Future Metaphysics, sect. 57. Compare the following passage from
Charles Sanders Peirce: “For my part, I cannot admit the proposition of Kant – that
there are certain impassable bounds to human knowledge. . . . The history of science
affords illustrations enough of the folly of saying that this, that, or the other can
never be found out. Auguste Comte said that it was clearly impossible for man ever to
learn anything of the chemical constitution of the fixed stars, but before his book had
reached its readers the discovery which he had announced as impossible had been
made. Legendre said of a certain proposition in the theory of numbers that, while
it appeared to be true, it was most likely beyond the powers of the human mind to
prove it; yet the next writer on the subject gave six independent demonstrations of
the theorem.” (Collected Papers [Cambridge, MA: Harvard University Press, 1931–58],
2nd ed., vol. VI, sect. 6.556.)
P1: KAE
What is at issue might be called Isaiah’s Law on the basis of the verse:
“For as the heavens are higher than the earth, so are my ways higher
than your ways, and my thoughts than your thoughts.”5 A fundamental
law of epistemology is at work here – to wit, that a mind of lesser power
is for this very reason unable to understand adequately the workings of a mind
of greater power. To be sure, the weaker mind can doubtless realize that
the stronger can solve problems that the lesser cannot. But it cannot
understand how it is able to do so. An intellect that can only just manage
to do well at tic-tac-toe cannot possibly comprehend the ways of one
that is expert at chess.
Consider in this light the vast disparity of computational power
between a mathematical tyro like most of us and a mathematical
prodigy like Ramanujan. Not only cannot our tyro manage to answer
the number-theoretic question that such a genius resolves in the blink
of an eye, but the tyro cannot even begin to understand the processes
and procedures that the Indian genius employs. As far as the tyro is
concerned, it is all sheer wizardry. No doubt once an answer is given
he can check its correctness. But actually finding the answer is some-
thing that lesser intellect cannot manage – the how of the business
lies beyond its grasp. And for much the same sort of reason, a mind
of lesser power cannot discover what the question-resolving limits of a
mind of greater power are. It can never say with warranted assurance
where the limits of question-resolving power lie. (In some instances it
may be able to say what’s in and what’s out, but never map the dividing
boundary.)
It is not simply that a more powerful mind will know more facts
than a less powerful one, but that its conceptual machinery is ampler
in encompassing ideas and issues that lie altogether outside the con-
ceptual horizon of its less powerful compeers.
Now insofar as the relation of a lesser toward a higher intelligence
is substantially analogous to the relation between an earlier state of
science and a later state, some instructive lessons emerge. It is not
that Aristotle could not have comprehended quantum theory; he was
a very smart fellow and could certainly have learned. But what he
could not have done was to formulate quantum theory within his own
conceptual framework, his own familiar terms of reference. The very
5 Isaiah, 58:9.
P1: KAE
ideas at issue lay outside the conceptual horizon of Aristotle’s science,

and like present-day students he would have had to master them from
the ground up. Just this sort of thing is at issue with the relation of
a less powerful intelligence to a more powerful one. It has been said
insightfully that from the vantage point of a less developed technology,
another substantially advanced technology is indistinguishable from
magic. And exactly the same holds for a more advanced conceptual
(rather than physical) technology.
It is instructive to contemplate in this light the hopeless difficul-
ties encountered nowadays in the popularization of physics – of trying
to characterize the implications of quantum theory and relativity the-
ory for cosmology into the subscientific language of everyday life. A
classic obiter dictum of Niels Bohr is relevant: “We must be clear that,
when it comes to atoms, language can be used only as in poetry.” Alas,
we have to recognize that in philosophy, too, we are in something
of the same position. In the history of culture, Homo sapiens began
his quest for knowledge in the realm of poetry. And in the end it
seems that we are destined to remain at this starting point in some
respects.
All the same, our cognitive limitedness as finite beings notwithstand-
ing, there nevertheless are no boundaries – no determinate limits – to
the manifold of discoverable fact. And here Kant was right – even
on the Leibnizian principles considered earlier in the discussion. For
while the cognitive range of finite beings is indeed limited, it is also
boundless because it is not limited in a way that blocks the prospect of
cognitive access to ever new and ongoingly more informative facts that
afford us an ever ampler and ever more adequate account of reality.
P1: KAE
0521861209con CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:21
Conclusion
In concluding, a brief survey of the principal theses may be in order.

They stand as follows:
Duhem’s Law of Security/Detail Complimentarity. The security and detail

of our knowledge stand in a relation of inverse proportionality.
(Chapter 1)
Kant’s Principle of Cognitive Systematization. Knowledge, in the quali-
tative and honorific sense of the term, is a matter of the extent
to which information is coherently systematized. (Chapter 2)
Spencer’s Law of Cognitive Development. Cognitive progress is accompa-
nied by complexification and can be assessed in terms of the taxo-
nomic complexity of the information manifold at hand. However,
this complexity is not proportional of the amount of information,
but to its logarithm. And this yields –
Kant’s Principle of Question Propagation. The progress of knowledge-
development in the course of resolving our questions always
brings new questions to light. (Chapter 4)
Gibbon’s Law of Logarithmic Returns. The quantity of knowledge inter-
est in a body of information is proportional not to the size of this
body, but merely to the logarithm thereof. (Chapter 4)
Adams’s Thesis of Exponential Growth. Throughout recent history the
body of scientific information has been growing exponentially.
But, in view of Gibbon’s Law, this means that cognitive progress
105
P1: KAE
0521861209con CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:21
106 Conclusion
in terms of actual knowledge has been growing at a rate that is

merely linear and thereby stable. (Chapter 5)
Quality/Quantity Alignment. The lower levels of informative quality
information that define the lesser degrees of “knowledge” are in
volumetric alignment with the λ-power (0 < λ ≤ 1) of the total
amount of information at hand. (Chapter 6)
Zipf’s Law. With objects rank-ordered by measurable size, the prod-
uct of this size with the number of objects of at least that size is
constant. And with cognitive importance as size this puts quality
and quantity into a relationship of complementarity. (Chapter 6)
Quality Retardation. Cognitive progress at lesser levels of what is not
so much knowledge as high-quality information proceeds at a
pace that is ever slower as the bar of quality is raised. (Chapter 6)
Leibniz’s Thesis of Cognitive Limitation. Mere volumetric considera-
tions indicate that knowledge cannot keep up with fact. See-
ing that our knowledge is language-bound, it is inherently
unable to accommodate reality’s unending complexity of details.
(Chapter 7)
Isaiah’s Law. A mind of a lesser power cannot fathom the workings
of one of greater power: some of its operations are bound to
appear as “magic.” (Chapter 8)
Kant’s Perspective on Cognitive Finitude. While the realm of knowl-
edge – of ascertainable fact – is indeed limited, it is nevertheless
unbounded. (Chapter 8)
Taken together these various theses and ideas provide a quantita-

tive perspective on knowledge and its limits. They thus combine to
clarify the situation that confronts a finite intelligence in its effort to
gain a cognitive grip on an endlessly complex and in some respects
unavoidably inscrutable world.
P1: KAE
0521861209ref CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:22
Bibliography
Adams, Henry, The Education of Henry Adams (Boston, 1918; privately printed
already in 1907).
Amaldi, Edoardo, “The Unity of Physics,” Physics Today, 261 (September, 1973),
pp. 23–29.
Auger, Pierre, Current Trends in Scientific Research (Paris, 1961; UNESCO
Publications), pp. 15–16.
Basalla, G., William Coleman, and R. H. Kargon [eds.], Victorian Science:
A Self-Portrait through the Presidential Addresses of the British Association for
the Advancement of Science (New York: New York Academy of Sciences,
1970).
Beeley, Philip, “Leibniz on the Limits of Human Knowledge,” Leibniz Review,
13 (December 2003), pp. 93–97.
Borges, Jorge Luis, “The Library of Babel.”
Brandon, Robert N., Adaptation and Environment (Princeton: Princeton Uni-
versity Press, 1990).
Cardwell, D. S. C., “The Professional Society” in Norman Kaplan (ed.), Science
and Society (Chicago, 1965), pp. 86–91.
Cicero, De natura deorum.
Couturat, Louis, La logique de Leibniz (Paris: Alcan, 1901).
Cuyckens, Hubert, and Britta Zawada (eds.), Polysemy in Cognitive Linguistics
(Amsterdam: John Benjamins, 2003).
David, Paul A., “Positive Feedback and Research Productivity in Science,” in O.
Granstrand (ed.), The Economics of Technology (Amsterdam: Elsevier, 1994),
pp. 65–89.
Dolby, R. G. A., “Classification of the Sciences: The Nineteenth-Century
Tradition” (unpublished study issued for its author by the University of
Kent at Canterbury ca. 1975).
Doyle, A. Conan, The Great Keinplatz Experiment (1894).
107
P1: KAE
108 Bibliography
Duhem, La théorie physique: son objet, et sa structure (Paris: Chevalier and Rivière,
1906); tr. by Philip P. Wiener, The Aim and Structure of Physical Theory (Prince-
ton: Princeton University Press, 1954).
Dupré, John, The Disorder of Things: Metaphysical Foundations of the Disunity of
Science (Cambridge, MA: Harvard University Press, 1993).
Eddington, Sir Arthur, The Nature of the Physical World (London: Macmillan,
1929), pp. 72–73.
Euclid, Elements.
Flint, Robert, Philosophy as Scientia Scientiarun: A History of Classifications of the
Sciences (Edinburgh: W. Blackwood and Sons, 1904).
George, William, The Scientist in Action (London: Williams & Norgate,
1938).
Gibbon, Edward, Memoirs of My Life (Harmondworth, 1984).
Glass, Bentley, “Milestones and Rates of Growth in the Development of
Biology,” Quarterly Review of Biology, 54 (March 1979), pp. 31–53.
Goffin William, and Kenneth S. Warren, Scientific Information Systems and the
Principle of Solidarity (New York: Praeger, 1981).
Gore, George, The Art of Scientific Discovery (London: Longmans, Green, 1878).
Heath, T. C., The Works of Archimedes (Cambridge: Cambridge University Press,
1897).
Holland, John H., Hidden Order: How Adaptation Builds Complexity (Reading,
MA: Addison Wesley, 1995).
Holub, H. W., Gottfried Tappeiner, and Veronica Eberharter, “The Iron Law
of Important Articles,” Southern Economic Journal, 58 (1991), pp. 317–28.
Hugly, Philip, and Charles Sayward, “Can a Language Have Indenumerably
Many Expressions?” History and Philosophy of Logic, 4, 1983.
International Survey of Book Production during the Last Decades (Paris: UNESCO,
1985).
Isaiah, 58:9.
Jevons, W. S., The Principles of Science, 2nd ed. (London: MacMillan, 1876).
Kant, Immanuel, Critique of Pure Reason.
Kant, Immanuel, Prolegomena to any Future Metaphysics.
Kaufmann, Stuart, At Home in the Universe: To Search for the Laws of Self-
Organization and Complexity (New York: Oxford University Press, 1995).
Kuhn, Thomas S., The Structure of Scientific Revolutions (Chicago: University of
Chicago Press, 1962).
Laudan, Larry, Progress and Its Problems (Berkeley: University of California Press,
1971).
Leibniz, G. W., De l’horizon de la doctrine humaine, ed. by Michael Fichant (Paris:
Vrin, 1991).
Machlup, Fritz, The Production and Distribution of Knowledge in the United States
(Princeton: Princeton University Press, 1962).
Mariétan Joseph, Probléme de la classification des sciences d’Aristote à St. Thomas
(St. Maurice and Paris: Fribourg, 1901).
Mees, C. E. K., The Path of Science (New York: Wiley, 1946).
Newton, Principia.
P1: KAE
Bibliography 109
Parat, M. V., The Information Economy: Definition and Measurement (Washington,

DC: U.S. Department of Commerce, May 1977; OT Special Publication
77–12).
Peirce, Charles Sanders, Collected Papers, 8 vols. (Cambridge, MA: Harvard
University Press, 1931–58).
Pelzholdt, Julius, Bibliotheca Bibliographica (Leipzig: W. Engelmann, 1866).
Petley, B. W., The Fundamental Physical Constants and the Frontiers of Measurement
(Bristol: Hilger, 1985).
Planck, Max, Vorträge und Erinnerungen, 5th ed. (Stuttgart: Hirzel, 1949).
Price, Derek J., Little Science, Big Science (New York: Columbia University Press,
1963).
Price, Derek J., Science since Babylon, 2nd ed. (New Haven, CT: Yale University
Press, 1975).
Price, Derek J., Characteristics of Doctrinal Scientists and Engineers in the University
System, 1991 (Arlington, VA.: National Science Foundation, 1994); Docu-
ment No. 94–307.
Rescher, Nicholas, Dialectics (Albany: State University of New York Press, 1977).
Rescher, Nicholas, Scientific Progress (Oxford: Blackwell, 1978).
Rescher, Nicholas, Cognitive Systematization (Oxford: Basil Blackwell, 1979).
Rescher, Nicholas, Induction (Oxford: Blackwell, 1980).
Rescher, Nicholas, “Leibniz and the Concept of a System” in his Leibniz’s Phi-
losophy of Nature (Dordrecht: D. Reidel, 1981), pp. 29–41.
Rescher, Nicholas, The Limits of Science (Berkeley and Los Angeles: University
of California Press, 1984).
Rescher, Nicholas, The Strife of Systems (Pittsburgh: University of Pittsburgh
Press, 1985).
Rescher, Nicholas, Scientific Realism (Dordrecht: D. Reidel, 1987).
Richardson, Ernest Cushing, Classification: Theoretical and Practical, 3rd. ed.
(New York: H. W. Wilson, 1930).
Richtmyer, F. K., “The Romance of the Next Decimal Place,” Science, 75 (1932),
pp. 1–5.
Rousseau, J. J., Contrat Social.
Samuels, Ernest, Henry Adams: The Major Phase (Cambridge, MA, 1964).
Shields, Charles W., Philosophia Ultima, vol. II, The History of the Sciences and the
Logic of the Sciences or the Science of the Sciences (New York: Scribner’s 1888–
1905), 3 vols.
Shockley, William, “On the Statistics of Productivity in Research,” Proceedings
of the Institute of Radio Engineeers, 45 (1957), pp. 279–90.
Spencer, Herbert, First Principles, 7th ed. (London: Appleton’s, 1889); see sects.
14–17 of Part II, “The Law of Evolution.” Statistics of Science and Technology
(London: HMSO, 1970).
Strawson, P. F., “Truth,” Proceedings of the Aristotelian Society, Supplementary
Vol. 24, 1950, pp. 129–56.
Ulam, Stanislaw M., Adventures of a Mathematician (New York: Scribner’s 1976).
Vollmer, H. M, and D. L. Mills, eds., Professionalization (Englewood Cliffs, NJ:
Prentice Hall, 1966).
P1: KAE
110 Bibliography
Wagner-Döbler, Roland, “Rescher’s Principle of Decreasing Marginal Returns

for Scientific Research,” Scientometrics, 50 (2001), pp. 419–36.
Weinberg, Alvin M., “The Impact of Large-Scale Science on the United States,”
Science, 134 (1961; 21 July issue), pp. 161–64.
Weinberg, Steven, Dreams of a Final Theory (New York: Pantheon Books, 1992).
Weizsäcker, C. F. von., “The Unity of Physics” in Ted Bastin (ed.), Quantum
Theory and Beyond (Cambridge: Cambridge University Press, 1971).
Wertheimer, F. H. et al., Chemistry: Opportunities and Needs (Washington, DC:
National Academy of Sciences/National Research Council, 1965).
Wible, James R., “Rescher’s Economic Philosophy of Science,” Journal of Eco-
nomic Methodology, vol. 1 (1994), pp. 314–29.
Wible, James R., The Economic of Science (London & New York: Routledge,
1998).
Wittgenstein, Ludwig, Tractatus.
Zipf, George K., Human Behaviour and the Principle of Least Effort (Boston:
Addison-Wesley, 1949).
P1: KAE
0521861209ind CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:31
Index of Names
Adams, Henry, 45–55, 47n4, 107, Dolby, R. G. A., 24n12, 107

109 Doyle, A. Conan, 47n2, 107
Amaldi, Edoardo, 25n15, 110 Duhem, Pierre Maurice, 1, 2, 3n3,
Anderson, C. Anthony, 94n19 105, 108
Aquinas, St. Thomas, 13n2, 20 Dupré, John, 24n13, 108
Archimedes, 11, 75, 75n2
Aristotle, 20, 104 Eberharter, Veronica, 66n9, 71n13,
Auger, Pierre, 20, 20n8, 25n14, 55, 108
107 Eddington, Arthur, 76n6, 108
Avogadro, Armadeo de, 21 Einstein, Albert, 79
Euclid, 11, 13n2, 108
Baer, Karl Ernst von, 15 Ewell, Raymond H., 55
Beeley, Philip, 75n3, 107
Bernoulli, Daniel, 35 Fechner, Gustav T., 29, 38, 38n10
Bohr, Neils, 2n2, 104 Ferrater-Mora, José, 59
Boltzmann, Ludwig, 21 Flint, Robert, 20n7, 108, 23n11
Bolzano, Bernhard, 90 Frege, Gottlob, 90
Borges, Jorge Luis, 68n10, 76n6, 107
Brandon, Robert N., 16n2, 107 Galton, Francis, 61, 62n4
George, William, 42n12, 51n11, 108
Cardwell, D. S. C., 43n15, 107 Gibbon, Edward, 29–46, 35n5, 54,
Cicero, 76n6, 107 54n19, 61, 73, 105, 108
Comte, Auguste, 20, 22, 102n4 Glass, Bentley, 55, 63, 63n7, 108
Couturat, Louis, 76n7, 107 Goffin, William, 70n12, 108
Gore, George, 42n12, 108
Darwin, Charles, 15, 43n15
David, Paul A., 64n8, 107 Heath, T. C., 75n2, 108
Descartes, René, 11, 46, 76 Holland, John H., 15n2, 108
Dickinson, Jason, 97n2 Holub, H. W., 66n9, 71n13, 108
111
P1: KAE
0521861209ind CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:31
112 Index of Names
Hugly, Philip, 80n11, 108 Ramanujan, Srinivasa, 105

Huxley, T. H., 76n6 Rescher, Nicholas, 55, 75n3, 109
Richardson, Ernest Cushing, 20n7,
Isaiah, 105, 108 23n11, 109
Richtmyer, F. K. R., 54, 109
Jevons, W. Stanley, 31, 31n2, 108 Rousseau, Jacques, 61, 61n3, 62n5,
Joule, J. P., 43n15 109
Ruetsche, Laura, 99n2
Kant, Immanuel, 9–28, 29, 31, 31n1,
32, 97, 103, 104n4, 106, 108 Samuels, Ernest, 47n4, 109
Kaufmann, Stuart, 16n3, 108 Sayward, Charles, 80n11, 109
Kelvin, Lord (William Thomson), Scotus, John Duns, 13n2
47, 47n2 Shakespeare, William, 76n6
Kuhn, Thomas, 35, 35n7, 108 Shields, Charles W., 23n11, 109
Shockley, William, 70n12, 109
Laplace, P. S. de, 46 Spencer, Herbert, 15–28, 15n1, 23,
Latka, Frantisek, 64, 64n8 32, 107, 109
Laudan, Larry, 59, 59n2, 108 Spinoza, Baruch, 11
Legendre, A. M., 102n4 Strawson, P. F., 81n13, 109
Leibniz, G. W., 11, 11n1, 46, 73–93, Suarez, Francesco, 13n2
75n3, 75n4, 76n6, 78n8, 106, 108
Tappeiner, Gottfried, 66n9, 71n13,
Maclup, Fritz, 108 108
Mariétan, Joseph, 20n9, 108 Thomson, William, 47
Mees, C. E. K., 54n21, 108
Mendeleev, D. I., 20 Ulam, Stanislaw M., 42n11, 109
Mill, J. S., 23
Wagner-Döbler, Roland, 63n6,
Newton, Isaac, 11, 13n2, 25, 46, 108 110
Warren, Kenneth S., 70n12, 108
Ockham, William of, 17 Washington, George, 46
Weber, E. H., 29, 38, 38n10
Parat, M. V., 58n1, 109 Weinberg, Alvin M., 51, 110
Peirce, C. S., 23, 42n12, 43, 102n4, Weinberg, Steven, 2n2, 25n15, 110
109 Weizsäcker, C. F. von, 25n15, 110
Pelzholdt, Julius, 23n11, 109 Wertheimer, F. H., 43n14, 110
Perloff, Mickey, 97n2 Whewell, William, 23
Petley, B. W., 21n10, 109 Wible, James R., 110
Planck, Max, 21, 29, 42–44, 42n12, Wiberforce, William, 76n6
109 Wittgenstein, Ludwig, 79, 91n18, 92,
Price, Derek J., 35, 35n6, 38n10, 110
48n7, 49n9, 50, 52n14, 53n16, Wolff, Christian, 11, 13n2
53n18, 55, 58n1, 62n4, 64n8,
70n12, 71n12, 109 Zipf, George K., 61n3, 69–71, 106,
Ptolemy, 20 110

Nicholas Rescher - Epistemetrics (2006)

Uploaded by

Copyright:

Available Formats

Nicholas Rescher - Epistemetrics (2006)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nicholas Rescher - Epistemetrics (2006)

Uploaded by

Copyright:

Available Formats

What is the main topic discussed in the document?

What is the main topic discussed in the document?

What disciplines are mentioned in relation to knowledge?

What disciplines are mentioned in relation to knowledge?

P1: JZZ

0521861209pre CUNY338B/Rescher 0 521 83164 2 December 7, 2005 13:42

This page intentionally left blank

Epistemetrics is a scholarly discipline waiting to be born. With regard

Nicholas Rescher is University Professor of Philosophy at the Uni-

Cambridge University Press

© Nicholas Rescher 2006

This publication is in copyright. Subject to statutory exception and to the provision of

First published in print format

isbn-13 978-0-511-14650-3 eBook (EBL)

isbn-13 978-0-521-86120-5 hardback

1 Asking for More Than Truth: Duhem’s Law of Cognitive

5 Adams’s Thesis on Exponential Growth 45

This book develops the theory of knowledge from a quantitative per-

the foreground, various of its key principles – for example Duhem’s

1 My personal effort in the domain of scientometrics is represented by my Scientific

present discussion will endeavor to synthesize and coordinate some of

Asking for More Than Truth

(1) Duhem’s Law of cognitive complementarity holds that inquiry is subject to

The Security/Definiteness Trade-off and the Contrast

2 Duhem’s Law of Cognitive Complementarity

there is always a characteristic trade-off relationship between the evi-

Science versus Common Sense 3

In effect, these two factors – security and detail – stand in a relation of

Science versus Common Sense

3 Duhem, La théorie physique, pp. 178–79. Italics supplied.

4 Duhem’s Law of Cognitive Complementarity

A law of physics is always provisional and relative. It is provisional also in that

Science decidedly prioritizes accuracy and detail over security. So

Consider the situation in which you correctly accept P-or-Q. But –

4 Compare H. M. Vollmer and D. L. Mills, eds., Professionalization (Englewood Cliffs, NJ:

6 Duhem’s Law of Cognitive Complementarity

believe that Smith bought some furniture because he bought a table.

THESIS 2: By constraining us to make vaguer judgments, ignorance enhances our

For example, if I have forgotten that Seattle is in Washington State,

will become. And so if we want our scientific claims to have realistic

Kant’s Conception of Knowledge as

Distinguishing Knowledge and Information

10 Knowledge as Systematized Information

else it is, knowledge is information of substantial quality. And this leads

Kant on the Systematicity of Knowledge

Kant on the Systematicity of Knowledge 11

not a mere contingent aggregate, but a system connected according to nec-

1 Leibniz’s theory of cognitive systematization is detailed in the author’s essay entitled

12 Knowledge as Systematized Information

The nature of systematization was accordingly explained by Kant

In accordance with reason’s legislative prescriptions, our diverse modes of

The Hierarchical Textualization of Knowledge 13

10 chapters per book

display 2.1. A Hypothetical Treatise.

The Hierarchical Textualization of Knowledge

14 Knowledge as Systematized Information

smaller scale, worthy of a sentence – or a paragraph at most. And this

Spencer’s Law of Cognitive Development

(1) Herbert Spencer’s “Law of Cognitive Development” through taxonomic exfo-

Spencer’s Law: the Dynamics of Cognitive Complexity