# Logical Constructions

*First published Wed Nov 20, 1996; substantive revision Wed Sep 10, 2014*

The term “logical construction” was used by Bertrand
Russell to describe a series of similar philosophical theories
beginning with the 1901 “Frege-Russell” definition of
numbers as classes and continuing through his
“construction” of the notions of space, time and matter
after 1914. Philosophers since the 1920s have argued about the
significance of “logical construction” as a method in
analytic philosophy and proposed various ways of interpreting
Russell's notion. Some were inspired to develop their own projects by
examples of constructions. Russell's notion of logical construction
influenced both Carnap's project of constructing the physical world
from experience and Quine's notion of *explication*, and was a
model for the use of set theoretic reconstructions in formal
philosophy later in the twentieth century.

It was only when looking back on his work, in the programmatic 1924 essay “Logical Atomism”, that Russell first described various logical definitions and philosophical analyses as “logical constructions”. He listed as examples the Frege-Russell definition of numbers as classes, the theory of definite descriptions, the construction of matter from sense data and then series, ordinal numbers and real numbers. Because of the particular nature of Russell's use of “contextual” definitions of expressions for classes, and the distinctive character of the theory of definite descriptions, he regularly called the expressions for such entities “incomplete symbols” and the entities themselves “logical fictions”.

Logical constructions differ in whether they involve
explicit definitions or contextual definitions, and in the extent to which their result should be described as showing that the constructed object is a mere “fiction”. Russell's 1901
definition of numbers as classes of equinumerous classes is
straightforwardly a case of constructing one
sort of entity as a class of others with an explicit definition. This was followed by the theory
of definite descriptions in 1905 and the “no-classes”
theory for defining classes in *Principia Mathematica* in 1910, both of which involved the distinctive technique of
contextual definition. In a contextual definition apparent singular
terms (either definite descriptions or class terms) are eliminated
through rules for defining the entire sentences in which they occur. Constructions which are like those using contextual definitions are generally called
“incomplete symbols”, while those like the theory of classes are called “fictions.” Russell
included the construction of matter, space and time as classes of sense data at
the end of his 1924 list. The main problem for interpreting the
notion of logical construction is to understand what these various examples have in common, and how the construction
of matter is comparable to either of the early constructions of
numbers as classes or the theory of definite descriptions and
“no-classes” theory of classes. None of the expressions
“fiction”,
“incomplete symbol” or even “constructed from” seems appropriate for an analysis of the
fundamental features of the familiar physical world and the material
objects that occupy it.

- 1. Honest Toil
- 2. Logical Analysis and Logical Construction
- 3. Natural Numbers
- 4. Definite Descriptions
- 5. Classes
- 6. Series, Ordinal Numbers and Real Numbers
- 7. Mathematical Functions
- 8. Propositions and Propositional Functions
- 9. The Construction of Matter
- 10. Successors to Logical Construction
- Bibliography
- Academic Tools
- Other Internet Resources
- Related Entries

## 1. Honest Toil

The earliest construction on Russell's 1924 list is the famous “Frege/Russell definition” of numbers as classes of equinumerous classes from 1901 (Russell 1993, 320). The definition follows the example of the definitions of the notions of limit and continuity that were proposed for the calculus in the preceding century. Russell did not rest content with adopting the Peano axioms as the basis for the theory of the natural numbers and then showing how the properties of the numbers could be logically deduced from those axioms. Instead, he defined the basic notions of “number” , “successor” and “0” and proposed to show, with carefully chosen definitions of their basic notions in terms of logical notions, that those axioms could be derived from principles of logic alone.

Russell defined natural numbers as classes of equinumerous
classes. Any * pair *, a class with two members, can be put into a one to
one correspondence with any other, hence all pairs are
* equinumerous*. The number *two* is then identified
with the class of all pairs. The relation between equinumerous classes
when there is such a one to one mapping relating them is called
“similarity”. Similarity is defined solely in terms of
logical notions of quantifiers and identity. With the natural numbers so
defined, Peano axioms can be derived by logical means
alone. After natural numbers, Russell adds “series, ordinal numbers and real
numbers” (1924, 166) to his list of constructions, and then
concludes with the construction of matter.

Russell credits A. N. Whitehead with the solution to the problem of the relation of sense data to physics that he adopted in 1914:

I have been made aware of the importance of this problem by my friend and collaborator Dr Whitehead, to whom are due almost all the differences between the views advocated here and those suggested in

The Problems of Philosophy. I owe to him the definition of points, and the suggestion for the treatment of instants and “things,” and the whole conception of the world of physics as aconstructionrather than aninference. (Russell 1914b,vi)

It is only later, in an essay in which Russell reflected on his philosophy that he also described his earlier logical proposals as “logical constructions.” The first specific formulation of this method of replacing inference with construction as a general method in philosophy is in the essay “Logical Atomism”:

One very important heuristic maxim which Dr. Whitehead and I found, by experience, to be applicable in mathematical logic, and have since applied to various other fields, is a form of Occam's Razor. When some set of supposed entities has neat logical properties, it turns out, in a great many instances, that the supposed entities can be replaced by purely logical structures composed of entities which have not such neat properties. In that case, in interpreting a body of propositions hitherto believed to be about the supposed entities, we can substitute the logical structures without altering any of the detail of the body of propositions in question. This is an economy, because entities with neat logical properties are always inferred, and if the propositions in which they occur can be interpreted without making this inference, the ground for the inference fails, and our body of propositions is secured against the need of a doubtful step. The principle may be stated in the form: ‘Whenever possible, substitute constructions out of known entities for inferences to unknown entities’. (Russell 1924, 160)

Russell was referring to logical constructions in this frequently
quoted passage from his *Introduction to Mathematical
Philosophy*. He objects to introducing entities with implicit
definitions, that is, as being those things that obey certain axioms or
“postulates”:

The method of ‘postulating’ what we want has many advantages; they are the same as the advantages of theft over honest toil. Let us leave them to others and proceed with our honest toil. (Russell 1919, 71)

He charges that we need a demonstration that there are any objects which satisfy those axioms.The “toil” here is the work of formulating definitions of the numbers so that they can be shown to satisfy the axioms using logical inference alone.

The description of logical constructions as “incomplete
symbols” derives from the use of contextual definitions that
provide an analysis or substitute for each sentence in which a defined
symbol may occur. The definition does not give an explicit definition,
such as an equation with the defined expression on one side that is
identified with a *definiendum* on the other, or a universal
statement giving necessary and sufficient conditions for the
application of the term in isolation. The connection between
being a fiction and expressed by an “incomplete symbol” can
be seen in Russell's constructions of finite cardinal and ordinal
numbers by means of the theory of classes. That
“no-classes” theory, via the contextual definitions for
class terms, makes all the numbers “incomplete symbols”,
and so numbers can be seen as “logical fictions”.

The notions of construction and logical fiction appear together in this account from Russell's “Philosophy of Logical Atomism” lectures:

You find that a certain thing which has been set up as a metaphysical entity can either be assumed dogmatically to be real, and then you will have no possible argument either for its reality or against its reality; or, instead of doing that, you can construct a logical fiction having the same formal properties, or rather having formally analogous formal properties to those of the supposed metaphysical entity and itself composed of empirically given things, and the logical fiction can be substituted for your supposed metaphysical entity and will fulfill all the scientific purposes that anyone can desire. (Russell 1918, 144)

Incomplete symbols, descriptions, classes and logical fictions are identified with each other and then with the “familiar objects of daily life” in the following passage from earlier in the lectures:

There are a great many other sorts of incomplete symbols besides descriptions. There are classes… and relations taken in extension, and so on. Such aggregations of symbols are really the same as what I call “logical fictions”, and they embrace practically all the familiar objects of daily life: tables, chairs, Piccadilly, Socrates, and so on. Most of them are either classes, or series, or series of classes. In any case they are all incomplete symbols, i.e. they are aggregations that only have a meaning in use and do not have any meaning in themselves. (Russell 1918, 122)

In what follows these various features of logical constructions will be disentangled. The result appears to be a connected series of analyses sharing at least a family resemblance with each other. The common feature is that in each case some formal or “neat” properties of objects that had to be postulated in axioms before could now be derived as logical consequences of definitions. The replaced entities are variously “fictions”, “incomplete symbols” or simply “constructions” depending on the form that the definitions take.

## 2. Logical Analysis and Logical Construction

It would be a mistake to see Russell's logical constructions as the product of the second stage of a method that begins with logical analysis. Analysis was indeed the distinctive method of Russell's realist and atomistic philosophy with the method of construction appearing only later. Russell's new philosophy was self-consciously in opposition to the Hegelianism prevailing in philosophy at Cambridge at the end of the nineteenth century (Russell 1956, 11–13). Russell first needed to defend the process of analysis, and to argue against the view of the idealists that complex entities are in fact “organic unities” and that any analysis of these unities loses something, as the slogan was “analysis is falsification”. (1903, §439) The subject of our analysis is reality, rather than merely our own ideas:

All complexity is conceptual in the sense that it is due to a whole capable of logical analysis, but is real in the sense that it has no dependence on the mind, but only upon the nature of the object. Where the mind can distinguish elements, there must be different elements to distinguish; though, alas! there are often different elements which the mind does not distinguish. (1903, §439)

As ultimate constituents of reality are what is discovered by logical analysis, logical construction cannot be the converse operation, for undoing the analysis by putting things back together only returns us to the complex entities with which we began. What then is the point of constructing what has already been analyzed?

Russell's methodology is best understood by analogy with the logical approach to scientific theories. On this model the result of “analysis” will be the definitions and primitive propositions or axioms from which the laws of a formalized scientfic theory can be derived by logical inference. The reduction of one theory to another consists of rewriting the axioms of the target theory using the language of the reducing theory, and then proving them as theorems of that reducing theory. Construction, then, is best seen as the process of choosing definitions so that previously primitive statements can be derived as theorems. (See Hager 1994 and Russell 1924.)

Although this picture fits well with this linguistically
oriented notion of “theory construction”, it does follow the notion of
construction in the tradition of mathematics. Euclid prefaces each
demonstration with a “construction” of a figure that
features in the following proof. Gottlob Frege begins every proof
in his *Basic Laws of Arithmetic* (1893) with an
“Analysis”, which informally explains the notions used in
the theorems and the strategy of the derivation, followed by the
actual, gapless proof, which is called the “Construction”.
Historically, then, there is no notion of a construction as a synthetic
stage following an analytic stage as two processes of a comparable
nature, but leading in opposite directions.

Even when described in terms of stages of theory construction, analysis and logical construction are not simply converse operations. Russell stresses that the objects discovered and distinguished in analysis are “real” as are their differences. Thus there is a constraint on the “choice” of definitions and primitive propositions with which to begin. The relationships between a deductive system and a realistic ontology differ among the various cases that Russell lists as examples of logical constructions. Propositions and “complexes” such as facts are analyzed in order to find the real objects and relations of which they are composed. A logical construction, on the other hand, results in a theory from which truths follow by logical inferences. The truths that are part of a deductive system resulting from logical construction are only “reconstructions” of some of the “pre-theoretic” truths that are to be analyzed. It is only their deductive relations, in particular their deducibility from the axioms of the theory, that are relevant to the success of a construction. Logical constructions do not capture all of the features of the pre-theoretic entities with which one begins.

Much of the attention to logical construction has focused on whether it is in fact a unified methodology for philosophy that will introduce a “scientific method in philosophy” as Russell says in the subtitle of (Russell 1914b). Commentators from Fritz (1952) through Sainsbury (1979) have denied that Russell's various constructions fit into a unified methodology, as well as questioning the applicability of the language of “fiction” and “incomplete symbol” to all examples. Below it will be shown how, nevertheless, constructions do fall into several natural families that are described by various of these terms with a considerable degree of accuracy.

## 3. Natural Numbers

Russell's definition of natural numbers as classes
of *similar*, or equinumerous, classes, first published in
(Russell 1901), was his first logical construction, and was the model
for those that followed. Similar classes are those that can be mapped
one to one onto each other by some relation. The notion of a
“one-to-one relation” is defined with logical notions: R
is one-one when for every \(x\) there is a unique \(y\) such that \(x \rR
y\), and for every such \(y\) in the range of \(\rR\) there is a unique
such \(x\). These notions of existence and uniqueness come from logic,
and so the notion of number is thus defined solely in terms of classes
and of logical notions. Russell announced the goal of his logicist
program in *The Principles of Mathematics*: “the proof
that all pure mathematics deals exclusively with concepts definable in
terms of a very small number of fundamental logical concepts, and that
all its propositions are deducible from a very small number of
fundamental logical principles…” (Russell
1903, *xv*). If * class * is also shown to be a logical
notion, then this definition would complete the logicist program for
the mathematics of natural numbers.

Giuseppe Peano (Peano 1889, 94) had stated axioms for elementary arithmetic, which were later formulated by Russell (1919, 8) as:

- 0 is a number.
- The successor of any number is a number.
- No two numbers have the same successor.
- 0 is not the successor of any number.
- If a property belongs to 0 and belongs to the successor of \(x\) whenever it belongs to \(x\), then it belongs to every number.

For Peano these were the axioms of number, which, along with axioms of classes and propositions, describing the properties of these entities and lead to the derivation of theorems that express the other important properties of those entities.

Richard Dedekind (Dedekind 1887) had also listed the properties of
numbers with similar looking axioms, using the notion of
*chain*, an infinite sequence of sets, each a subset of the next, that is well
ordered and has the structure of the natural numbers. Dedekind then
proves that the principle of induction (Axiom 5 above) holds for
chains. (See entry on Dedekind). Although Russell finds it “most remarkable that
Dedekind's previous assumptions suffice to demonstrate this
theorem” (Russell 1903, §236), he compares the two
approaches, of Peano and Dedekind, with respect to simplicity and their
differing ways of treating mathematical induction, and concludes
that:

But from a purely logical point of view, the two methods seem equally sound; and it is to be remembered that, with the logical theory of cardinals, both Peano's and Dedekind's axioms become demonstrable. (Russell 1903, §241)

It was Peano and Dedekind that Russell had in mind when he later speaks of “the method of ‘postulating’” when he compares the “advantages” of their method over construction as those of theft over honest toil.

To complete his project Russell needed to find definitions and some
“very small number of fundamental logical principles”
(Russell 1903, *xv*) and then produce the required derivations.
Finding an adequate definition of classes with the “no-classes
theory” and the principles of logic needed to derive the
properties of numbers and classes was only completed with *Principia
Mathematica* (Whitehead and Russell 1910–13). This construction of
numbers was a clear example of defining entities as classes of others
so as to be able to prove certain properties as theorems of logic
rather than having to rest with the theft of hypotheses. With the
device of contextual definition from the theory of descriptions Russell
then eliminated classes too, taking as fundamental the logical notion
of a propositional function and so showing that the principles of
classes where a part of logic.

## 4. Definite Descriptions

Definite descriptions are the logical constructions that Russell has in mind when when he describes them as “incomplete symbols”. The notion of a “logical fiction”, on the other hand, applies most straightforwardly to classes. Other constructions, such as the notions of the domain and range of a relation, and of one to one mappings that are crucial to the development of arithmetic, are only “incomplete” in an indirect sense due to their being defined as classes of a certain sort, which are in turn constructions.

Russell's *theory of descriptions* was introduced in his paper
“On Denoting” (Russell 1905) published in the
journal *Mind*. Russell's theory provides an analysis of
sentences of the form ‘The \(F\) is \(G\)’ where
‘The \(F\)’ is called a *definite description* in
contrast with ‘An F’ which is an *indefinite
description*. The analysis proposes that ‘The \(F\) is
\(G\)’ is equivalent to ‘There is one and only one \(F\)
and it is \(G\)’. Given this analysis, the logical properties of
descriptions can be deduced using just the logic of quantifiers and
identity. Among the theorems in *14 of *Principia Mathematica*
are those showing that, (1) if there is just one \(F\) then ‘The
\(F\) is \(F\)’ is true, and if there is not, then ‘The
\(F\) is \(G\)’ is always false and then, (2) if the \(F =
\text{the } G\), and the \(F\) is \(H\), then the \(G\) is
\(H\). These theorems show that proper (uniquely referring)
descriptions behave like proper names, the “singular
terms” of logic. Some of these results have been controversial
—Strawson (1950) claimed that an utterance of ‘The present
King of France is bald’ should be truth valueless since there is
no present king of France, rather than “plainly” false, as
Russell's theory predicts. Russell's reply to Strawson in (Russell
1959, 239–45) is helpful for understanding Russell's
philosophical methodology of which logical construction is just a
part. It is, however, by assessing the logical consequences of a
construction that it is to be judged, and so Strawson challenged
Russell in an appropriate way.

The theory of descriptions introduces Russell's notion of
*incomplete symbol*. This arises because no definitional
equivalent of ‘The F’ appears in the formal analysis
of sentences in which the description occurs. The sentence ‘The
\(F\) is \(H\)’ becomes:

of which no subformula, or even a contiguous segment, can be identified as the analysis of ‘The F’. Similarly, talk about “the average family” as in “The average family has 2.2 children” becomes “The number of children in families divided by the number of families = 2.2”. There is no segment of that analysis that corresponds to “the average family”. Instead we are given a procedure for eliminating such expressions from contexts in which they occur, hence this is another example of an “incomplete symbol” and the definition of an average is an example of a “contextual definition.”

It is arguable that Russell's definition of definite descriptions was the most prominent early example of the philosophical distinction between surface grammatical form and logical form, and thus marks the beginnings of linguistic analysis as a method in philosophy. Linguistic analysis begins by looking past superficial linguistic form to see an underlying philosophical analysis. Frank Ramsey described the theory of descriptions as a “paradigm of philosophy” (Ramsey 1929, 1). While in itself surely not a model for all philosophy, it was at least a paradigm for the other examples of logical constructions that Russell listed when looking back on the development of his philosophy in 1924. The theory of descriptions has been criticized by some linguists and philosophers who see descriptions and other noun phrases as full-fledged linguistic constituents of sentences, and who see the sharp distinction between grammatical and logical form as a mistake. (See the entry on descriptions.)

Following Gilbert Ryle's (1931) influential criticisms of Meinong's theory of non-existent objects, the theory of descriptions has been taken as a model for avoiding ontological commitment to objects, and so logical constructions in general are often seen as being chiefly used to eliminate purported entities. In fact, that goal is at most peripheral to many constructions. The principal goal of these constructions is to allow the proof of propositions that would otherwise have to be assumed as axioms or hypotheses. Nor need the introduction of constructions always result in the elimination of problematic entities. Yet other constructions should be seen more as reductions of one class of entity to another, or replacements of one notion by a more precise, mathematical, substitute.

## 5. Classes

Russell's “No-Class” theory of classes from *20 of
*Principia Mathematica* provides a contextual definition like
that of the theory of definite descriptions. One of Russell's early
diagnoses of the paradox of the class of all classes that are not
members of themselves was that it showed that classes could not be
individuals. Indeed Russell seems to have come across his paradox by
applying Cantor's famous diagonal argument to show that there are more
classes of individuals than individuals. Hence, he concluded, classes
could not be individuals, and expressions for classes such as
‘\(\{x: Fx \}\)’ cannot be the singular terms they appear to be. Inspired by the theory of descriptions,
Russell proposed that to say something \(G\) of the class of
\(F\)s, \(G\) \(\{x: Fx \}\), is to say that there
is some (predicative) property \(H\) coextensive with (true of the
same things as) \(F\) such that \(H\) is \(G\). The
restriction to predicative properties, or those which are not defined
in terms of quantification over other properties, was a consequence of
the *ramification* of the theory of types to avoid intensional
or “epistemic” paradoxes which motivated the theory of
types in addition to the set theoretical “Russell's
Paradox” (see Whitehead and Russell 1910–13, Introduction,
Chapter II). These predicative properties are intensional, however, in
the sense that two distinct properties might hold of the same objects.
(See the entry on the
notation in Principia Mathematica.)
That classes
so defined have the feature of extensionality is thus derivable, rather
than postulated. If \(F\) and \(H\) are coextensive then
anything true of \(\{x:Fx \}\) will be true of
\(\{x:Hx \}\). Features of classes then follow from the
features of the logic of properties.

Because classes would at first seem to be individuals of some sort, but on analysis are found not to be, Russell speaks of them as “logical fictions,” an expression which echoes Jeremy Bentham's notion of “legal fictions.” (Hart 1994, 84) (See entry on law and language). That a corporation is a “person” at law was for Bentham merely a fiction that could be cashed out in terms of the notion of legal standing and of limits to the financial liability of real persons. Thus any language about such “legal fictions” could be translated in other terms to be about real individuals and their legal relationships. Because statements attributing a property to particular classes are analyzed by existential sentences saying that there is some propositional function having that property, this construction also can be characterized as showing that class expressions, such as ‘\(\{x:Fx \}\)’, are incomplete symbols. They are not replaced by some longer formula expressing a term. On the other hand, the definition should not be seen as avoiding ontological commitment entirely, as showing that something is literally a “fiction”. Rather it shows how to reduce classes to propositional functions. The properties of classes are really properties of propositional functions and for every class said to have a property there really is some propositional function having that property.

## 6. Series, Ordinal Numbers and Real Numbers

Whitehead and Russell define a *series* in volume II of *Principia
Mathematica* at *204.01 as the class
**Ser** of all relations which is transitive, connected
and irreflexive. A relation \(R\) is
*transitive* when, if \(xRy\) and
\(yRz\) then \(xRz\). It is
*connected* when for any \(x\) and \(y\) for which it
is defined, either \(xRy\) or \(yRx\). Finally,
an *irreflexive* relation is one such that for all \(x\), it is not
the case that \(xRx\). Any
relation that has those properties forms a
*series* of the things that it relates. Such relations are now called
“linear orderings” or simply, “orderings”. Here the “logical
construction” simply consists of an implicit definition of a
certain property of relations. There is certainly no thought that series
are merely invented “fictions”, and the symbol ‘**Ser**’ for them is
“incomplete” only in that it can be explicitly defined as
the intersection of other classes (a class of classes) and classes are themselves “incomplete”.

Russell's definitions of ordinal numbers and real numbers resemble the
definitions of natural numbers. Ordinal numbers are a special case of
* relation numbers*. Just as a cardinal
number can be defined as a class of similar classes where the
similarity is simply equinumerosity, the existence of a one to one
mapping between the two classes, a relation number is a class of
similar classes which are ordered by some relation. Ordinal numbers are the relation
numbers of well-ordered classes. “Relation-Arithmetic” is the subject of
Part IV of Volume II of *Principia Mathematica*, chapters
*150 to *186. All of the properties of the arithmetic of ordinal
numbers are derived from the more general arithmetic of relation
numbers. Thus, for example, the addition of ordinal numbers is not
commutative. The first infinite ordinal \(\omega\) is the relation number
of the well-ordered classes similar to \(1, 2, 3, \ldots\) etc. The sum
\(1 + \omega\) will be the relation number of ordered classes which
result from adding one element at the beginning of the ordering, say
\(0, 1, 2, 3, \ldots\) etc., which has the same ordinal number
\(\omega\). Thus \(1 + \omega = \omega\). On the other hand, adding an
element at the “end” of such a well ordered class will
give an ordering that is not similar: \(1, 2, 3, \ldots \text{etc.}, 0\).
Consequently, \(1 + \omega \ne \omega + 1\). On the other hand
addition of ordinals, and indeed relation numbers in general, is
associative, that is,
\((\alpha + \beta) + \gamma = \alpha + (\beta + \gamma)\),
which is proved with certain restrictions in *174. Ordinal
numbers are thus defined exactly as natural numbers, as classes of
similar classes, in such a way that all the desired theorems can be
proved. Describing ordinal numbers as “fictions”,
“incomplete symbols” and “constructions”
applies in the same way as in the case of natural numbers.

The class of real numbers, Θ, is defined in Volume III of
*Principia Mathematica* at *310.01 as consisting of “Dedekindian
series” of rational numbers, which are in turn relation numbers
of “ratios” of natural numbers. Whitehead and Russell
follow the account of real numbers as *Dedekind cuts* of the
rational numbers, and only differ from more standard developments of
the numbers in contemporary set theory by treating rational numbers as
relation numbers of a certain sort, rather than ordered pairs of and
integers (the “numerator” and
“denominator”). Like the construction of relation numbers
as classes of similar classes, the “logical construction”
of real numbers differs from the theory of definite descriptions and
classes in general in not defining “incomplete symbols” or
by showing that these numbers are really “fictions”. They
are best characterized as definitions that allow for the proof of
theorems about these numbers that would otherwise have to be
postulated as axioms. They are the product of the “honest
toil” that Russell prefers.

## 7. Mathematical Functions

Mathematical functions are not mentioned by Russell in the 1924 list
of “logical constructions” although the analysis of mathematical
functions is the principal application of the theory of
definite descriptions in PM. The basic “functions” of PM are
propositional functions. The Greek letters \(\phi, \psi, \theta, \ldots\)
are variables for propositional functions, and, with
individual variables
\(x, y, z, \ldots\) go together to form open
sentences \(\phi(x), \psi(x,y)\), etc. This
is the familiar syntax of modern predicate logic. Mathematical
functions, such as the
*sine* function and addition, are represented as * term
forming operators* such as \(\sin x\), or \(x + y\). In contemporary
logic they are symbolized by function letters that are followed by the
appropriate number of arguments, \(f(x),g(x,y)\), etc. In chapter *30
Whitehead and Russell propose a direct analysis of such expressions
for mathematical functions in terms of definite descriptions, which
they call “descriptive functions”. Consider the relation
between a number and its sine, the relation which obtains
between \(x\) and \(y\) when
\(y = \sin x\). Call this relation
“\(\text{Sine}(x,y)\)” or more simply,
“\(\bS(x,y)\)”, as a two-place
relation. The mathematical function can then be expressed with a
definite description, interpreting our expression “the sine of
\(x\)” not as “\(\sin(x)\)”, but
literally as “*the* Sine of \(x\)”, with a
definite description, or “the \(y\) such that
\(\text{Sine}(x,y)\)”. Using the notation of the theory of
definite descriptions, this is
‘\((\iota x)\bS(x,y)\)’.
The effect of this analysis is that Whitehead and Russell can replace all expressions
for mathematical functions with definite descriptions based on
relations. This definition involves
relations *in extension*, which are represented with upper case Roman
letters and with the relation symbol between the variables. The
definition in PM is: *30.01. \(R`y = (\iota x)xRy\), with the notation
\(R`y\) to be read as “the \(R\)
of \(y\).” As with the theory of descriptions, the result
of this definition is to facilitate the proofs of theorems which capture the logical
properties of mathematical functions that will be needed in the
further work of PM.

The logical analysis of function expressions in PM presents them as a special case of definite descriptions, “the \(R\) of \(x\)”. In the Summary of *30 we find:

Descriptive functions, like descriptions in general, have no meaning in isolation, but only as constituents of propositions. (Whitehead and Russell 1910–13, 232)

Mathematical or descriptive functions are thus explicitly included among the
incomplete symbols of *Principia Mathematica*.

## 8. Propositions and Propositional Functions

In *Principia Mathematica* Russell's *multiple relation
theory* of judgment is introduced by presenting an ontological
vision:

The universe consists of objects having various qualities and standing in various relations. (Whitehead and Russell 1910–13, 43)

Russell goes on to explain the multiple relation theory of judgment, which finds the place of propositions in this world of objects and qualities standing in relations. (See the entry on propositions.)

Russell's multiple relation theory, that he held from 1910 to around 1919, argued that the constituents of propositions, say ‘Desdemona loves Cassio’, are unified in a way that does not make it the case that they constitute a fact by themselves. Those constituents occur only in the context of beliefs, say, ‘Othello judges that Desdemona loves Cassio’. The real fact consists of a relation of Belief holding between the constituents Othello, Desdemona, and Cassio; \(B(o,d,L,c)\). Because one might also have believed propositions of other structures, such as \(B(o,F,a)\) there need to be many such relations \(B\), of different “arities”, or number of arguments, hence the name “multiple relation” theory. Like the construction of numbers, this construction abstracts from what a number of occurrences of a belief have in common, namely, a relation between a believer and various objects in a certain order. The analysis also makes the proposition an incomplete symbol because there is no constituent in the analysis of ‘\(x\) believes that \(p\)’ that corresponds to ‘\(p\)’. As a result Russell concludes that:

It will be seen that, according to the above account, a judgment does not have a single object, namely a proposition, but has several interrelated objects. That is to say, the relation which constitutes judgment is not a relation of two terms, namely the judging mind and the proposition, but is a relation of several terms, namely the mind and what we call the constituents of the proposition…

Owing to the plurality of the objects of a single judgment, it follows that what we call a “proposition” (in which it is to be distinguished from the phrase expressing it) is not a single entity at all. That is to say, the phrase which expresses a proposition is what we call an “incomplete” symbol; it does not have meaning in itself, but requires some supplementation in order to acquire a complete meaning. (Whitehead and Russell 1910–13, 43–44)

Although bound variables ranging over propositions hardly occur in
*Principia Mathematica* (with a prominent exception in *14.3),
it would seem that the whole theory of types is a theory of
propositional functions. Yet following on the claim that propositions
are “not single entities at all”, Russell says the same
for propositional functions. In the *Introduction to Mathematical
Philosophy*, Russell says that propositional functions are really
“nothing”, but “nonetheless important for
that” (Russell 1919, 96). This comment makes best sense if we
think of propositional functions as somehow constructed by abstracting them from their values,
which are propositions. The propositional function “\(x\) is
human” is abstracted from its values “Socrates is
human”, “Plato is human”, etc. Viewing propositional
functions as constructions from propositions, that are in turn
constructions by the multiple relation theory, helps to make sense of
certain features of the theory of types of propositional functions
in *Principia Mathematica*. We can understand how propositional
functions seem to depend on their values, namely propositions, and how
propositions in turn can themselves be logical constructions. The
relation of this dependence to the theory of types is explained in the
Introduction to *Principia Mathematica* in terms of the notion
of ”presupposing“:

It would seem, however, that the essential characteristic of a function is

ambiguity… We may express this by saying that “\(\phi x\)” ambiguously denotes \(\phi a, \phi b, \phi c,\) etc., where \(\phi a, \phi b, \phi c,\) etc. are the various values of “\(\phi x\).” … It will be seen that, according to the above account, the values of a function are presupposed by that function, not vice versa. It is sufficiently obvious, in any particular case, that a value of a function does not presuppose the function. Thus for example the proposition “Socrates is human” can be perfectly apprehended without regarding it as a value of the function “\(x\) is human.” It is true that, conversely, a function can be apprehended without its being necessary to apprehend its values severally and individually. If this were not the case, no function could be apprehended at all, since the number of values (true and false) of a function is necessarily indefinite and there are necessarily possible arguments with which we are not acquainted. (Russell 1910–13, 39–40)

The notion of “incomplete symbol” seems less appropriate than “construction” in the case of propositional functions and propositions. To classify propositions and even propositional functions as instances of the same logical phenomenon as definite descriptions requires a considerable broadening of the notion.

The ontological status of propositions and propositional functions
within Russell's logic, and in particular, in *Principia
Mathematica*, is currently the subject of considerable debate. One
interpretation, which we might call “realist,” is
summarized in this footnote by Alonzo Church in his 1976 study of the
ramified theory of types:

Thus we take propositions as values of the propositional variables, on the ground that this is what is clearly demanded by the background and purpose of Russell's logic, and in spite of what seems to be an explicit denial by Whitehead and Russell in PM, pp. 43–44.

In fact, Whitehead and Russell make the claim: “that what we call a ‘proposition’ (in the sense in which this is distinguished from the phrase expressing it) is not a single entity at all. That is to say, the phrase which expresses a proposition is what we call an ‘incomplete symbol’ …” They seem to be aware that this fragmenting of propositions requires a similar fragmenting of propositional functions. But the contextual definition or definitions that are implicitly promised by the “incomplete symbol” characterization are never fully supplied, and it is in particular how they would explain away the use of bound propositional and functional variables. If some things that are said by Russell in IV and V of his Introduction to the second edition may be taken as an indication of what is intended, it is probable that the contextual definitions would not stand scrutiny.

Many passages in [(Russell 1908)] and [(Whitehead and Russell 1910–13)] may be understood as saying or as having the consequence that the values of propositional functions are sentences. But a coherent semantics of Russell's formalized language can hardly be provided on this basis (notice in particular, that, since sentences are also substituted for propositional variables, it would be necessary to take sentences as names of sentences.) And since the passages in question seem to involve confusions of use and mention or kindred confusions that may be merely careless, it is not certain that they are to be regarded as precise statements of a semantics. (Church 1976, n.4)

Gregory Landini (1998) has proposed that there is indeed a coherent semantics for propositions and propositional functions in PM, which treats functions and propositions as linguistic entities. Landini proposes that this “nominalist semantics” is the intended interpretation of PM and is what remains of Russell's earlier “substitutional theory.” He argues that Russell was led to this nominalism after first rejecting the reality of classes,then of propositional functions, and finally the reality of propositions. This rejection, according to Landini, leaves us with only a nominalist metaphysics of individuals and expressions as the interpretation of Russell's logic. See also Cocchiarella (1980), who describes a “nominalist semantics” for ramified type theory, but rejects it as Russell's intended interpretation. Sainsbury (1979) describes a “substitutional” interpretation of the quantifiers over propositional functions, but combines this with a truth-conditional semantics that does not require the ramification of the theory of types that is central to Russell's interpretation in PM.

Propositions and propositional functions are unlike definite descriptions and classes in that there are no explicit definitions of them in PM. It is unclear what it means to say that a symbol for a proposition, such as a variable \(p\) or \(q\), has “no meaning in isolation”, and that, however, the meaning can be given “in context”, as it would seem that there no definition is possible, it would seem, in a logic in which propositions and propositional functions appear as primitive notions in the statement of the axioms and definitions of logic.

## 9. The Construction of Matter

Whether or not they are provided with contextual definitions by Whitehead and Russell, logical constructions do not appear as the referents of logically proper names, and so by that account constructions are not a part of the fundamental “furniture” of the world. Early critical discussions of constructions, such as Wisdom (1931), stressed the contrast between logically proper names, which do refer, and constructions, which were thus seen as ontologically innocent.

Beginning with *The Problems of Philosophy* in 1912, Russell
turned repeatedly to the problem of matter. As has been described by
Omar Nasim (2008), Russell was stepping into an ongoing discussion of
the relation of sense data to matter that was being carried on by T.P.
Nunn (1910), Samuel Alexander (1910), G.F. Stout (1914), and G.E. Moore
(1914), among others. The participants of this “Edwardian
controversy”, as Nasim terms it, shared a belief that direct
objects of perception, with their sensory qualities, were nonetheless
extra-mental. The concept of matter, then, was the result of a loosely
described social or psychological “construction”, going
beyond what was directly perceived. A project shared by the
participants in the controversy was the search for a refutation of
George Berkeley's idealism, which would show how the existence
and real nature of matter can be discovered. In The *Problems of
Philosophy* (Russell 1912) Russell argues that the belief in the
existence of matter is a well supported hypothesis that explains our
experiences. Matter is known only indirectly, “by
description”, as the cause, whatever it may be, of our sense
data, which we directly know by “by acquaintance”. This is
and example the sort of hypothesis that Russell contrasts with
construction in the famous passage about “theft” and
“honest toil”. Russell saw an analogy between the case of
simply hypothesizing the existence of numbers with certain properties,
those described by axioms, and hypothesizing the existence of
matter.

The need for some sort of account of the logical features of matter,
what he called “the problem of matter”, had already
occupied Russell much earlier. While we distinguish the certain
knowledge we may have of mathematical entities from the contingent
knowledge of material objects, Russell says that there are certain
“neat” features of matter that are just too tidy to have
turned out by accident. Examples include the most general
spatiotemporal properties of objects, that no two can occupy the same
place at the same time, which he calls “impenetrability”,
and so on. In *The Principles of Mathematics* (Russell 1903,
§453) there is a list of these features of matter including
“indestructibility”, “ingenerability” and
“impenetrability”, which were all characteristic of the
atomic theory of the day. Russell followed the progression through the
exact sciences from logic through arithmetic, and then real numbers and
then to infinite cardinals. There followed a discussion of space and
time, with the book ending with a last part (VII) on Matter and Motion,
chapters §53 to §59. In them Russell discusses what he calls
“rational Dynamics as a branch of pure mathematics”
(Russell 1903, §437). This *rational Dynamics*, would
involve justifying many of the fundamental principles of physics with
pure mathematics alone, from definitions that yield the geometry of
space and time and the formal properties of its occupants, quantities of matter and
energy. In this respect the construction of matter most resembles the construction of numbers as classes as an effort to replace the “theft” of postulating axioms with the “honest toil” of devising definitions that will validate those postulates.

In the later project of constructing matter, from 1914 on, beginning
with *Our Knowledge of the External World* (Russell 1914b),
material objects come to be seen as collections of sense data, than of
“sensibilia”. Sensibilia are potential objects of
sensation, which, when perceived become “sense data” for
the perceiver. Influenced by William James, Russell came to defend a
*neutral monism* by which matter and minds were both to be
constructed from sensibilia, but in different ways. Intuitively, the
sense data occurring as they do “in” a mind, are material
to construct that mind, the sense data derived from an object from
different points of view to construct that object. Russell saw some
support for this in the theory of relativity, and the fundamental
importance of frames of reference in the new physics.

## 10. Successors to Logical Construction

In the 1930s Susan Stebbing and John Wisdom, founding what has come
to be called the “Cambridge School of Analysis,” paid
considerable attention to the notion of logical construction (see
Beaney 2003). Stebbing (1933) was concerned with the unclarity over
whether it was expressions or entities that are logical constructions,
and with how to understand a claim such as “this table is a
logical construction” and indeed what it could even mean to
contrast logical constructions with inferred entities. Russell had been
motivated by the logicist project of finding definitions and elementary
premises from which mathematical statements could be proved. Stebbing
and Wisdom were concerned, rather, with relating the notion of
construction to philosophical analysis of ordinary language. Wisdom's
(1931) series of papers in *Mind* interpreted logical
constructions in terms of ideas from Wittgenstein's *Tractatus*
(1921).

Demopoulos and Friedman (1985) find an anticipation of the recent
“structural realist” view of scientific theories in
(Russell 1927), *The Analysis of Matter*. They argue that the
logical constructions of sense data in Russell's earlier thinking on
the “problem of matter” were replaced by inferences to the
structural properties of space and matter from patterns of sense data.
We may sense patches of color next to each other in our visual field,
but what that tells us about the causes of those sense data, about
matter, is only revealed by the structure of those relationships. Thus
the color of a patch in our visual field tells us nothing about the
intrinsic properties of the table that causes that experience. Instead
it is the structural properties of our experiences, such as their
relative order in time, and which are between which others in the
visual field, that gives us a clue as to the structural relationships
of time and space within the material world that causes the experience.
The contemporary version of this account, called “structural
realism”, holds that it is only the structural properties and
relations that a scientific theory attributes to the world about which
we should be scientific realists. (See the entry on
structural realism.)

According to this account, Russell's initial project of replacing
inference with logical construction was to find for each pattern of
sense data some logical construction that bears a pattern of isomorphic
structural relations. That project was transformed, Demopoulos and
Friedman argue, by replacing inference from the given in experience to
the cause of that experience with an inference to the rather
impoverished, structural, reality of the causes of those experiences.
Russell's matter project was interpreted in this way by others, and
led, in 1928, to G.H. Newman's apparently devastating objection. Newman
(1928) pointed out that there is always a structure of arbitrarily
“constructed” relations with any given structure if only
the number of basic entities, in this case sense data, is large enough.
According to Demopoulos and Friedman, Newman shows that there must be
more to scientific theories than trivial statements to the effect that
matter has some structural properties isomorphic to those of our sense
data. The project of *The Analysis of Matter* does indeed face a
serious difficulty with “Newman's problem”, whether
or not those difficulties arise for the earlier project of logical
construction (see Linsky 2013).

The notion of logical construction had a great impact on the future course of analytic philosophy. One line of influence was via the notion of a contextual definition, or paraphrase, intended to minimize ontological commitment and to be a model of philosophical analysis. The distinction between the surface appearance of definite descriptions, as singular terms, and the fully analyzed sentences from which they seem to disappear was seen as a model for making problematic notions disappear upon analysis. Wisdom (1931) proposed this application of logical construction in the spirit of Wittgenstein. In this way the theory of descriptions has been viewed as a paradigm of philosophical analysis of this “therapeutic” sort that seeks to dissolve logical problems.

A more technical strand in analytic philosophy was influenced by the
construction of matter. Rudolf Carnap quotes (Russell 1914a, 11) as the
motto for his “*Aufbau*”, the *Logical Structure
of the World* (1967):

The supreme maxim in scientific philosophizing is this: Whenever possible, logical constructions are to be substituted for inferred entities. (Carnap 1967, 6)

In the *Aufbau* the construction of matter from
“elementary experiences”, and later Nelson Goodman (1951)
continued the project. Michael Friedman (1999) and Alan Richardson
(1998) have argued that Carnap's project of construction owed
much more to his background in neo-Kantian issues about the
“constitution” of empirical objects than with
Russell's project. See, however, Pincock (2002) for a response
that argues for the importance of Russell's project of
reconstructing scientific knowledge in (Carnap 1967). More
generally, the use of set theoretic constructions became widespread
among philosophers, and continues in the construction of set theoretic
models, both in the sense of logic where they model formal theories
and to provide descriptions of truth conditions for sentences about
entities.

Willard van Orman Quine saw his notion of “explication”
as a development of logical construction. Quine presents his
methodology in *Word and Object* (1960) beginning with an
allusion to Ramsey's remark in the title of section 53:
“The Ordered Pair as Philosophical Paradigm”. The problem
of apparently referring expressions that motivates Russell's
theory of descriptions is presented as a general problem:

A pattern repeatedly illustrated in recent sections is that of the defective noun that proves undeserving of objects and is dismissed as an irreferential fragment of a few containing phrases. But some times the defective noun fares oppositely: its utility is found to turn on the admission of denoted objects as values of the variables of quantification. In such a case our job is to devise interpretations for it in the term positions where, in its defectiveness, it had not used to occur. (Quine 1960, 257)

The notion of a “defective noun” that is to be “dismissed as an irreferential fragment” clearly echoes the description of constructions as logical fictions and their expressions as mere incomplete symbols that so aptly describe the contextual definitions for definite descriptions and classes. The task of “devising interpretations” is more like the positive aspect suggested by the term “construction” and illustrated in the cases of the construction of numbers and matter. After concluding that the expression “ordered pair” was such a “defective noun”, Quine says that the notion of an ordered pair \(\langle x,y \rangle\) of two entities \(x\) and \(y\) does have “utility” and is limited only in having to fulfill one “postulate”:

- (1) If \(\langle x,y \rangle = \langle z,w \rangle\) then \(x = z\) and \(y = w\).

In other words, that ordered pairs are distinguished by having unique first and second elements. Quine then continues:

The problem of suitably ekeing out the use of these defective nouns can be solved once for all by systematically fixing upon some suitable already-recognized object, for each \(x\) and \(y\), with which to identify \(\langle x,y \rangle\). The problem is a neat one, for we have in (1) a single explicit standard by which to judge whether a version is suitable. (Quine 1960, 258)

Again Quine echoes Russell's language with his mention of a “neat” property that calls out for a “construction” from known entities. Quine distinguishes his project, which he calls “explication”, by the fact that there are alternative possible ways to fix the notion. Although Whitehead and Russell give an analysis in PM *55, where they are called “ordinal couples”, the first proposal to treat ordered pairs as classes of their members is from Norbert Wiener (1914) who identifies \(\langle x,y \rangle\) with \(\{\{ x \}, \{ y, \Lambda \}\}\), where \(\Lambda\) is the empty class. From this definition it is easy to recover the first and second elements of the pair, and so Quine's (1) is an elementary theorem. Later, Kuratowski proposed the definition \(\{\{ x \}, \{x,y\}\}\), from which (1) also follows. For Quine it is a matter of choice which definition to use, as the points on which they differ are “don't-cares”, issues which give a precise answer to questions about which our pre-theoretic account is mute. An explication thus differs considerably from an “analysis” of ordinary, or pre-theoretic language, both in giving a precise meaning to the expression where it might have been obscure, or perhaps simply silent and in possibly differing from pre-theoretic use, as suggested by the name. This fits well with the asymmetries we have noted between analysis and construction, with analysis aimed at the discovery of the constituents and structure of propositions which are given to us, and construction which is more a matter of choice, with the goal being the recovery of particular “neat” features of the construction in a formal theory. The ordered pair is thus a “philosophical paradigm” for Quine just as Russell's theory of descriptions was a paradigm of philosophy for Ramsey, and each is a “logical construction”.

## Bibliography

### Primary Literature: Works by Russell

- 1901, “The Logic of Relations”, (in French)
*Rivista di Matematica*, Vol. VII, 115–48. English translation in Russell 1956, 3–38 and Russell 1993, 310–49. - 1903,
*The Principles of Mathematics*, Cambridge: Cambridge University Press. 2^{nd}ed. 1937, London: Allen & Unwin. - 1905, “On Denoting”,
*Mind*14 (Oct.), 479–93. In Russell 1956, 39–56 and Russell 1994, 414–27. - 1908, “Mathematical Logic as Based on the Theory of
Types”,
*American Journal of Mathematics*30, 222–62. In van Heijenoort 1967, 150–82 and Russell 2014, 585–625. - 1910–13, A.N. Whitehead and B.A. Russell,
*Principia Mathematica*, Cambridge: Cambridge University Press. 2^{nd}ed. 1925–27. - 1912,
*The Problems of Philosophy*, London: Williams and Norgate. Reprinted 1967 Oxford: Oxford University Press. - 1914a, “The Relation of Sense Data to Physics”,
*Scientia*, 16, 1–27. In*Mysticism and Logic*, Longmans, Green and Co. 1925, 145–179 and Russell 1986, 3–26. - 1914b,
*Our Knowledge of the External World: As a Field for Scientific Method in Philosophy*, Chicago and London: Open Court. - 1918, “The Philosophy of Logical Atomism” in
*The Monist*, 28 (Oct. 1918): 495-527, 29 (Jan., April, July 1919): 32-63, 190-222, 345-80. Page references to*The Philosophy of Logical Atomism*, D.F. Pears (ed.), La Salle: Open Court, 1985, 35–155. Also in Russell 1986, 157–244 and Russell 1956, 175–281. - 1919,
*Introduction to Mathematical Philosophy*, London: Routledge. - 1924, “Logical Atomism”, in
*The Philosophy of Logical Atomism*, D. F. Pears (ed.), La Salle: Open Court, 1985, 157–181. Russell 2001, 160–179. - 1927,
*The Analysis of Matter*, London: Kegan Paul, Trench, Trubner & Co. - 1956, Logic and Knowledge: Essays 1901–1950, ed. R. C. Marsh, London: Allen & Unwin.
- 1959,
*My Philosophical Development*, London: George Allen & Unwin. - 1973,
*Essays in Analysis*, ed. D. Lackey. London: Allen & Unwin. - 1986,
*The Collected Papers of Bertrand Russell*, vol. 8,*The Philosophy of Logical Atomism and Other Essays: 1914–1919*, ed. J. G. Slater, London: Allen & Unwin. - 1993,
*The Collected Papers of Bertrand Russell*, vol. 3,*Towards the “Principles of Mathematics”,*ed. Gregory H. Moore, London and New York: Routledge. - 1994,
*The Collected Papers of Bertrand Russell*, vol. 4,*Foundations of Logic: 1903–1905*, ed. A. Urquhart, London and New York: Routledge. - 2001,
*The Collected Papers of Bertrand Russell*, vol. 9,*Essays on Language, Mind and Matter: 1919–1926*, ed. J.G. Slater, London and New York. - 2014,
*The Collected Papers of Bertrand Russell*, vol. 5,*Towards Principia Mathematica, 1905–1908*, ed. G. H. Moore, London and New York: Routledge.

### Secondary Literature

- Alexander, S., 1910, “On Sensations and Images”,
*Proceedings of the Aristotelian Society*, X: 156–78. - Beaney, M., 2003, “Susan Stebbing on Cambridge and Vienna
Analysis”,
*The Vienna Circle and Logical Empiricism*, F. Stadler (ed.), Dordrecht: Kluwer, 339–50. - Carnap, R., 1967,
*The Logical Structure of the World & Pseudo Problems in Philosophy*, trans. R. George, Berkeley: University of California Press. Originally*Der Logische Aufbau der Welt*, Berlin: Welt-Kreis, 1928. - Church, A., 1976, “Comparison of Russell's Resolution
of the Semantical Antinomies with That of Tarski”,
*Journal of Symbolic Logic*, 41: 747–760. - Cocchiarrella, N., 1980, “Nominalism and Conceptualism as
Predicative Second-Order Theories of Predication”,
*Notre Dame Journal of Formal Logic*, 21(3): 481–500. - Dedekind, R., 1887. Was sind und was sollen die Zahlen?, translated
as “The Nature and Meaning of Numbers” in
*Essays on the Theory of Numbers*, New York: Dover, 1963. - Demopolous, W. and Friedman, M., 1985, “Bertrand Russell's
*The Analysis of Matter*: Its Historical Context and Contemporary Interest”,*Philosophy of Science*, 52(4): 621–639. - Friedman, M., 1999,
*Reconsidering Logical Positivism*, Cambridge: Cambridge University Press. - Frege, G., 1893/1903,
*Basic Laws of Arithmetic*, Jena: Pohle, 2 volumes, trans. P. Ebert & M. Rossberg, Oxford: Oxford University Press, 2013. - Fritz, Jr., C. A., 1952,
*Bertrand Russell's Construction of the External World*, London: Routledge & Kegan Paul. - Goodman, N., 1951,
*The Structure of Appearance*, Cambridge Mass: Harvard University Press. - Hager, P., 1994,
*Continuity and Change in the Development of Russell's Philosophy*, Dordrecht: Kluwer. - Hart, H.L.A., 1994,
*The Concept of Law*, 2^{nd}edition, Oxford: Clarendon Press. - Landini, G., 1998,
*Russell's Hidden Substitutional Theory*, Oxford: Oxford University Press. - Linsky, B., 1999,
*Russell's Metaphysical Logic*, Stanford: CSLI. - –––, 2004, “Russell's Notes on Frege for
Appendix A of
*The Principles of Mathematics*”,*Russell: The Journal of Bertrand Russell Studies*, 24: 133–72. - –––, 2007, “Logical Analysis and Logical
Construction”,
*The Analytic Turn*, M. Beaney (ed.), New York: Routledge, 107–122. - –––, 2013, “Russell's Theory of Definite
Descriptions and the Idea of Logical Construction”,
*The Oxford Handbook of the History of Analytic Philosophy*, M. Beaney, ed., Oxford: Oxford University Press, 407–429. - Moore, G. E., 1914, “Symposium: The Status of
Sense-Data”,
*Proceedings of the Aristotelian Society*, XIV: 335–380. - Nasim, O. W., 2008,
*Bertrand Russell and the Edwardian Philosophers: Constructing the World*, Houndsmill, Basingstoke: Palgrave Macmillan. - Newman, H. A., 1928, “Mr. Russell's ‘Causal Theory of
Perception‘”,
*Mind*, 37: 137–148. - Nunn, T. P., 1910, “Symposium: Are Secondary Qualities
Independent of Perception? ”,
*Proceedings of the Aristotelian Society*, X: 191–218. - Peano, G., 1889, “The Principles of Arithmetic; Presented by
a New Method”, translated by J. van Heijenoort,
*From Frege to Gödel*, Cambridge, Mass: Harvard University Press, 1967, 81–97. - Pincock, C., 2002, “Russell's Influence on
Carnap's
*Aufbau*”,*Synthese*, 131(1): 1–37. - Richardson, A., 1998,
*Carnap's Construction of the World: The Aufbau and the Emergence of Logical Empiricism*, Cambridge: Cambridge University Press. - Quine, W. V. O., 1960,
*Word and Object*, Cambridge Mass: The MIT Press. - Ramsey, Frank, 1929, “Philosophy”, in F. P. Ramsey,
*Philosophical Papers*, ed. D. H. Mellor, Cambridge: Cambridge University Press, 1990, 1–7. - Ryle, G., 1931, “Systematically Leading Expressions”,
*Proceedings of the Aristotelian Society*, 32: 139–70; reprinted in*The Linguistic Turn: Essays in Philosophical Method*, R. M. Rorty (ed.), Chicago: University of Chicago Press, 1992, 85–100. - Sainsbury, M., 1979,
*Russell*, London: Routledge & Kegan Paul. - Stebbing, S., 1933,
*A Modern Introduction to Logic*, London: Methuen and Company, 2nd edition. - Stout, G. F., 1914, “Symposium: The Status of Sense-Data
”,
*Proceedings of the Aristotelian Society*, XIV: 381–406. - Strawson, P. F., 1950, “On Referring”,
*Mind*LIX (235): 320–344. - van Heijenoort, J., ed., 1967,
*Frege to Gödel: A Source Book in Mathematical Logic, 1879–1931*, Cambridge, Mass: Harvard University Press. - Wiener, N., 1914, “A Simplification of the Logic of
Relations”,
*Proceedings of the Cambridge Philosophical Society*, 17: 387–390; reprinted in van Heijenoort 1967, 224–227. - Wisdom, J., 1931, “Logical Constructions (I.).”,
*Mind*, 40: 188–216. - Wittgenstein, L., 1921,
*Tractatus Logico-Philosophicus*, 1961, trans. Pears and McGuinness, London: Routledge and Kegan Paul.

## Academic Tools

How to cite this entry. Preview the PDF version of this entry at the Friends of the SEP Society. Look up this entry topic at the Indiana Philosophy Ontology Project (InPhO). Enhanced bibliography for this entry at PhilPapers, with links to its database.

## Other Internet Resources

- Russell, B.,
*Introduction to Mathematical Philosophy*, available in PDF.

### Acknowledgments

I am grateful to Allen Hazen for explaining the significance of Quine's chapter on ordered pairs.