Analyzing Plato’s Arguments: Plato and Platonism

S. Marc Cohen and David Keyt

1. Introduction
Does Plato have a philosophy? If so, what is it and how does he argue for it? Simple
questions. But there are well-known obstacles standing in the way of their answer. First
of all, Plato writes dialogues; and it is often unclear which character, if any, in a given
dialogue speaks for Plato. Secondly, when a character in a dialogue advances a thesis, it
is often unclear what the thesis is. And, finally, when a thesis is backed up by an
argument, crucial premises are often missing.
In this paper we wish to focus on this last obstacle and consider some of the issues
it raises, though the other two obstacles will not go unnoticed. Suppose, then, that one
has at least surmounted the first obstacle and is dealing with an argument that can
reasonably be attributed to Plato himself and not just to a character in a dialogue.
Suppose, further, that the argument is missing a crucial premise. The basic issue that
we shall address is the proper goal of an interpretation that supplies the missing
premise. Is the goal to divine Plato’s thought or to extend it? When an interpreter
supplies the missing premise, what is he really doing? Is he expounding Plato or
platonizing? This issue arises because of the paradoxical consequences of a major
principle of interpretation.
In attempting to understand a passage from a major philosopher such as Plato,
Aristotle, Kant, or Wittgenstein, interpreters often seem to be guided by a principle of
charity that directs them to put as favorable a construction as possible on the passage
under consideration. And this is wise strategy. For an interpreter who ignores this
principle risks missing a good, perhaps even a profound, point. If the passage under
consideration contains an argument, the principle of charity says that, other things
being equal, one interpretation is better than another just to the extent that the one
produces a better argument than the other. Suppose, now, that the principle is applied
to a philosophical text that contains an argument whose conclusion follows from its
explicit premises only by the addition of a “tacit” or “suppressed” premise1—the sort of
argument that traditional logic calls an enthymeme. Paradoxical consequences follow in
six stages.

2. A Paradox of Interpretation

First Stage
Following the principle of charity, an interpreter, when faced with a passage from a
major philosopher that contains an enthymeme, will search for a suppressed premise
rather than charge the philosopher with a non sequitur. Ordinarily an interpreter will
have an indefinite number of possible premises to choose among. Still pursuing the
principle of charity, the interpreter tries to make the most sympathetic choice among
the possible premises. Often it will be possible for an interpreter to judge that one
premise is a better choice than a second by appealing to other passages in the same
work or, failing this, in other writings of the same philosopher. If there is a passage in
which the one premise is asserted by the philosopher but no passage in which the
second is asserted, then in selecting a tacit premise the one is a better choice than the
other. In this case the interpreter fills the hole in the argument by extending the context
of the argument. We shall call such an argument an apparent enthymeme. An argument
 2

with a suppressed premise is a real enthymeme, on the other hand, if, and only if, the
suppressed premise cannot be supplied by extending the context.
In the case of a real enthymeme how does an interpreter choose among the various
possibilities? He continues to invoke the principle of charity. If the proposition
expressed by one possible premise seems to the interpreter to be more reasonable than
the proposition expressed by another, the interpreter, led by the principle of charity, will
supply the one rather than the other. But reasonable in what sense, and to whom? One
possibility is that it is reasonable for the interpreter to believe that the author of the text
believed the proposition, or would have believed it had he entertained it. Such a
proposition is reasonable for the interpreter to attribute to the author. Another
possibility is that it is reasonable for the interpreter himself to believe the proposition
according to his own lights. Such a proposition is reasonable for the interpreter to hold.
Thus the proposition that the earth is shaped like a drum is a reasonable one for us to
attribute to Anaximander, but not, of course, a reasonable one for us to hold.
Which sort of reasonableness is demanded in the interpretation of a real
enthymeme? At first sight, it may appear that it is only the first sort—reasonableness to
attribute—that is at issue. For historians are expected, aren’t they, to ferret out the
beliefs of the historical figures they study, not necessarily to share them? But we
contend that a tacit premise for a real enthymeme must be reasonable for the
interpreter to hold. For any proposition that it is reasonable to attribute to an author is
reasonable on the basis of the text or the context. But the gap in a real enthymeme is
precisely the sort of gap that cannot be filled by an appeal to the text or context however
much the context is broadened. Consequently, an interpreter here has no basis for
supplying the missing premise beyond his own sense of what it is reasonable for a
rational person to hold.
Suppose two interpreters differ on the question of reasonableness. They propose
competing and inconsistent interpretations of a real enthymeme.2 Now, if the
correctness of an interpretation were an objective fact, at most one of the interpreters
could be right. But since we are dealing with a real enthymeme, there is no principle
beyond the principle of charity to adjudicate between the two interpretations. And by
hypothesis the principle of charity is unable to decide between them; each interpreter
applies the principle correctly, and each prefers his own interpretation. Of course one
interpreter may be an idiot and the other a Princeton professor, but the situation
envisaged often arises when both interpreters are good philosophers and equally
rational. This brings us to our first conclusion:

Either there is no objectively correct interpretation of any real

enthymeme found in the text of a major philosopher or else it is
inaccessible to us.

Second Stage
The situation envisaged so far is this. An interpreter is analyzing an argument with
a suppressed premise and considering which premise among various competing
candidates to supply. In choosing among the candidates, the principle of charity directs
him to choose the most reasonable. But the interpreter may regard none of the possible
premises as reasonable to hold. This is an unstable situation in scholarship and leads
to the second stage. An interpreter who is guided by the principle of charity always
seeks a premise that is reasonable to hold, and he will not regard the enthymeme as
adequately interpreted until such a premise has been found. Hence our second
conclusion:
 3

No adequately interpreted real enthymeme of a major philosopher has a

conclusion that an interpreter will judge to be false unless it also has at
least one explicit premise that he also judges to be false.

Third Stage
In attempting to understand a passage from a major philosopher that contains a
real enthymeme, it is usually necessary not only to supply a suppressed premise but
also to interpret the enthymeme’s conclusion and its explicit premises. By the principle
of charity an interpreter puts as favorable a construction on a given sentence as
possible. He tries to find an interpretation under which it expresses a proposition that
in his judgment is close to the truth. Indeed, if a given sentence appears to express a
false proposition, he will try to find an interpretation under which it expresses a true
proposition.3 Consequently, an interpreter strives for an interpretation under which all
the premises of a real enthymeme, explicit as well as implicit, express propositions that
are close to the truth. Thus we arrive at our third conclusion:

An interpreter will not regard a real enthymeme of a major philosopher as

adequately interpreted until he has found a way of reading it that makes
it into a good argument, that is to say, an argument that is valid and all
of whose premises, explicit and implicit, express propositions that in his
judgment are close to the truth.

Fourth Stage
An interpreter’s judgments about truth and reasonableness are affected, if not
determined, by the philosophy that is current in his day. Thus an interpreter will always
strive for a reading of a passage from a major philosopher by which the passage
expresses something that would be reasonable for a contemporary philosopher to hold.
Hence our fourth conclusion:

All philosophy, including that written 2400 years ago, is contemporary

philosophy. Or, phrased another way, all interpretation is anachronistic.

Fifth Stage
The foregoing conclusion can be strengthened. A charitable interpreter will
undoubtedly rank Plato higher than any contemporary philosopher. Thus in
interpreting Plato he will set his sights higher than the philosophy of his own day. His
standard of reasonableness will be perfect reasonableness rather than that of
contemporary philosophy. But such reasonableness belongs only to a god. Hence our
fifth conclusion:

To a charitable interpreter every classical text is a sacred text and every

classical philosopher infallible and omniscient.

Sixth Stage
Furthermore, there can be no discord in heaven. Charitable interpretation cannot
allow for the possibility that two major philosophers might disagree. For the contrary
hypothesis leads to a contradiction. Suppose that Plato and Aristotle disagree over some
issue. Suppose, for example, that Aristotle claims that an idea of Plato’s is false and
that Plato’s argument for it is invalid. Since Plato and Aristotle are both major
philosophers, both must be interpreted charitably. On a charitable interpretation of
Aristotle, Aristotle reads Plato correctly, and his criticism of Plato is well-taken. On the
 4

other hand, on a charitable interpretation of Plato, Plato’s idea is reasonable and his
argument for it valid. Thus an interpreter who is interpreting Aristotle interpreting Plato
and who is charitable to both Plato and Aristotle must find the very same argument
both valid and invalid. This is impossible. So the original hypothesis, that it is possible
for two major philosophers to disagree, must be false. Hence our sixth and final
inference from the principle of charity:

Major philosophers never disagree.

This conclusion seems preposterous. But the venerable philosophical tradition of
syncretism reflects the propensity of interpreters to reach the sixth stage.

3. An Illustration: The TMA

Modern Platonic scholarship might seem far removed from syncretism. But it is not
difficult to find in recent scholarship a series of discussions of a single Platonic text
guided by the principle of charity that maps most of the stages of the paradox of
interpretation. The recent history of the interpretation of the Third Man Argument (TMA)
in Plato’s Parmenides is one example.
The TMA, as all students of the Parmenides will recall, threatens the Theory of
Forms with an infinite regress:4

This, I suppose, is what leads you to believe that each form is one.
Whenever many things seem to you to be large, some one form probably
seems to you to be the same when you look at them all. So you think that
largeness is one. . . . But what about largeness itself and the other large
things? If you look at them all in your mind in the same way, won’t some
one largeness appear once again, by virtue of which they all appear
large? . . . So another form of largeness will have made an appearance,
besides largeness itself and its participants. And there will be yet another
over all these, by virtue of which they will all be large. So each of your
forms will no longer be one, but an infinite multitude.
The argument is plainly an enthymeme. Its only explicit premise is a One-Over-
Many assumption:5

(OM) If a number of things are all F, it follows that there is a Form in

virtue of which they are all F.
The reasoning proceeds as follows. OM is applied to an initial collection of F things
and generates a Form of F-ness. The Form is then added to the initial collection, and
OM is applied to this new collection, generating another Form. The new Form is added
to the previous collection, and OM is applied again. The process can be repeated, ad
infinitum.
In Gregory Vlastos’s seminal article,6 OM appears in the following form:

(OMv) If a number of things . . . are all F, there must be a single Form F-

ness, in virtue of which we apprehend [them] as all F.
And as Vlastos observes, OMv alone does not entail an infinite regress of Forms. For (a)
no provision has been made for the second application of OMv, and (b) nothing explicitly
stated entitles us to infer that the second application of OMv generates a second Form,
distinct from the one generated by the first application of OMv.
 5

Did Vlastos dismiss the TMA as invalid? Not at all. Declaring that “there must have
been something more in Plato’s mind” (p. 236) than what appears in OMv, he proposed
two additional tacit premises to fill in the gaps in the reasoning. To justify the second
application of OMv, Vlastos proposed a Self-Predication assumption:

(SPv) F-ness is F.
Now when F-ness is collected together with the F things that participate in it, SPv
guarantees that they are, all of them, F. Hence, the second application of OMv is
provided for.
To justify the inference to a new Form, Vlastos proposed a Non-Identity
assumption:

(NIv) If x is F, then x cannot be identical with F-ness.

The idea is simple enough. Something which is F by virtue of participating in a Form
cannot be identical to that Form. Hence the Form generated by the first application of
OMv cannot be identical to the Form generated by the second application; otherwise, it
would be identical to the Form in virtue of which it is F.
A striking feature of this reconstruction, as Vlastos himself noted, is that his two
tacit premises are mutually inconsistent. (The contrapositive of NIv says that if x is
identical to F-ness, then x is not F. More simply put, this says that F-ness is not F,
which is the contradictory of SPv.) Still, he was convinced that these two assumptions
had to be made for the argument to go through. So, believing that the premises
necessary to generate the regress are inconsistent, Vlastos maintained that Plato must
have been unaware of them. He concluded that there is a buried inconsistency in the
Theory of Forms of which Plato was only dimly aware, and that the TMA is a “record of
[Plato’s] honest perplexity” (p. 254).
Not surprisingly, Vlastos’s article unleashed a torrent of critical response. From the
first, Vlastos’s critics have urged that there need be no inconsistency in the TMA’s
premises. Wilfrid Sellars7 advanced a reconstruction of the TMA which not only
reconciled its two tacit premises (SP and NI) but also resolved the inconsistency between
them and OM. The first reconciliation was carried out by reformulating NI as a principle
that denies self-participation rather than self-predication. To assert that nothing
participates in itself (and, in particular, that F-ness does not participate in F-ness) does
not contradict SP. It merely entails, when conjoined with SP, that at least one F thing—
namely, F-ness itself—does not participate in F-ness.
But this last proposition conflicts with OMv, which entails that every F thing
(including F-ness itself) participates in F-ness. So further repairs were needed to remove
the inconsistency from the entire premise-set. Sellars observed that Vlastos had used
the expression “F-ness” as if it represented a proper name of a Form, and proposed
instead that “F-ness” be taken to represent a quantifiable variable.8 This simple
syntactic maneuver removes the remaining inconsistency. Here are the TMA’s premises
as Sellars formulated them:

(OMs) If a number of things are all F, it follows that there is an F-ness in

virtue of which they are all F.

(SPs) All F-nesses are F.

(NIs) If x is F, then x is not identical with the F-ness by virtue of which it

is F.9
 6

These three premises are mutually consistent; they do not entail a contradiction. But
they do entail that if there are any F things at all, there is an infinite regress of F-
nesses.
In response,10 Vlastos conceded that his formulation of the Non-Identity
assumption had been defective, and that Sellars had succeeded in “deriv[ing] the
regress by an internally consistent argument” (p. 353). But Vlastos denied that this
could have been the argument that Plato intended. For where Plato’s version of OM
posits a unique Form (“whenever many things seem to you to be large, some one form
probably seems to you to be the same when you look at them all,” 132a2-3), Sellars’
OMs has an ordinary existential quantifier (“there is an F-ness”). But throughout this
context, and in numerous others, Vlastos insists, it is clear that by ‘one Form’ Plato
means ‘exactly one Form’. Concluding that OM cannot be correctly formulated without a
uniqueness quantifier, Vlastos offered this revised version of the TMA’s premise-set:

(OMv2) If any set of things share a given character, then there exists a
unique Form corresponding to that character; and each of these things
has that character by participating in that Form (p. 348).

(SPv2) The Form corresponding to a given character itself has that

character (p. 351).

(NIv2) If anything has a given character by participating in a Form, it is

not identical with that Form (p. 351).
On this revision SP and NI are compatible, but the three axioms together are not.11
For suppose there are F things. Then there is a unique Form corresponding to the
character F (OMv2); but this Form is an F thing (SPv2), and so shares the character F
with its own participants, and so participates in the unique Form corresponding to F
(OMv2), that is, participates in itself; but this contradicts NIv2. Vlastos thus continued to
maintain that the TMA’s premise-set is inconsistent.
Cohen12 agreed with Vlastos that the correct formulation of OM will have a
uniqueness quantifier but did not concede that this must render the premise-set
inconsistent. Just as a more careful formulation of SP and NI showed them to be
consistent with one another, so a more sophisticated approach to the formulation of OM
was required.
Cohen’s analysis of the TMA exploits an analogy with number theory: Plato’s infinite
regress of Forms is analogous to the generation of the infinite sequence of natural
numbers (NNs). Consider how the NNs are defined by Peano’s Postulates:13
1. 0 is a NN.
2. The successor of any NN is a NN.
3. No two NNs have the same successor.
4. 0 is not the successor of any NN.
5. Any property that belongs to 0, and also to the successor of every NN
that has the property, belongs to all NNs.

In the usual (von Neumann) set-theoretic construction in which the NNs are
represented by sets, the successor of a set is defined as the union of that set with its
own unit set:
α′ = α ∪ { α }.
The NNs are thus represented as the members of the following infinite sequence of sets:
 7

0=Λ
1=Λ ∪{0}
2 = Λ ∪ { 0, 1 }
3 = Λ ∪ { 0, 1, 2 }
etc.14
Zero is identified with the empty set; the successor of zero (i.e., 1) is identified with the
set whose only member is the empty set; etc. The sequence clearly satisfies the five
Peano Postulates. Note that it also has the curious feature that every NN “belongs” to
every one of its “descendants”:

0∈1∈2∈3∈...
This is a harmless side-effect of the construction in number theory, but it captures the
central idea in Plato’s regress.
The symbols of the foregoing sequence of equations are, of course, subject to
reinterpretation. Suppose we take ∈ to represent participation, rather than set-
membership, Λ to denote the set of F particulars, rather than the empty set, and = to
denote a one-one relation pairing a form with the set of its participants, rather than
identity. So interpreted, our formulas represent a different (but structurally identical)
sequence of objects:
F0 = Λ
F1 = Λ ∪ { F0 }
F2 = Λ ∪ { F0 , F1 }
F3 = Λ ∪ { F0 , F1 , F2 }
etc.
F0 is the Form which has all and only F particulars as its participants; F1 is the Form
whose participants are F0 and all the participants in F0; F2 is the Form whose
participants are F1 and all the participants in F1; etc. Suppose finally that ‘NN’ denotes
the set { F0, F1, F2, . . .}, instead of the set of natural numbers.
Each of the Peano Postulates now has a familiar Form-theoretic analogue: (1)
asserts that there is a Form which has (all and only) the F particulars as its
participants. (2) amounts to SP (each Form by virtue of which an F thing is F is itself F).
(3) has NI as a consequence.15 (4) asserts that F0, the first Form in the sequence, has
only particulars as participants. (5) guarantees (although this is not obvious) that, for
every n, Fn is a member of the sequence.
The only premise of the TMA that has no counterpart among the Peano Postulates
is OM. Since OM’s role is to generate a new Form at each stage of the regress, its
number-theoretic counterpart is the successor function, which generates the members
of the infinite sequence of NNs. If OM is to have a uniqueness quantifier, then, it will
need to be based on something stronger than Plato’s over relation, which is not a
function. But we can use Plato’s over relation to define an immediately-over function
that corresponds to the successor function:16

y is immediately over x =df y is over x, and there is no z such that y is

over z and z is over x.
 8

(That is, one Form is “immediately over” another if no third Form intervenes between
the two.17) Cohen’s One-Immediately-Over-Many axiom thus guarantees that every
Form has a unique “successor”:

(IOM-axiom) For any set of F’s, there is exactly one Form immediately
over that set.
This axiom blocks self-participation, since it entails that Forms do not belong to the sets
they are over. NI is thus built into IOM-axiom. SP is presupposed as well, for the values
of the variables in the definition of the immediately-over function have been restricted to
things that are F.
The regress develops as we would expect. We are given a set of F’s; IOM-axiom
generates a Form they all participate in (and which is the unique Form immediately over
that set). That Form is itself F, and we may thus obtain a new set of F’s (in the usual
way) by adding the Form to the previous set. IOM-axiom is applied to the new set,
generating a new Form, and so on. Assuming that the Peano Postulates are consistent,
the TMA’s premise-set is thus capable of a consistent formulation, even with a
“uniqueness” quantifier in the One-Over-Many premise.
Vlastos was clearly correct in conceding that SP and NI, properly formulated, are
not incompatible; but he was mistaken in supposing that any version of OM with a
uniqueness quantifier would reintroduce the inconsistency. But notice what was
required to show this. Vlastos had been working in first-order logic with quantifiers
ranging over particulars and Forms, whereas Cohen’s reconstruction required
quantifying over sets, as well.18 To demonstrate the consistency of the TMA’s premises it
had been necessary to employ more sophisticated logical machinery.
Let us pause for a moment to review the history of the use of the principle of
charity in interpreting the TMA. In his 1954 article, Vlastos used the principle of charity
to convert a non sequitur into a valid argument, but two of the premises in his
reconstruction contradicted one another. His 1969 revision again illustrates the same
use of the principle of charity, but now at a second level: he replaced this blatant
inconsistency with a more subtle one—an inconsistent dyad with an inconsistent triad.
Sellars, Geach,19 Strang,20 Cohen and others too numerous to mention illustrate a
further use of the principle of charity; they sought to remove inconsistency altogether.
But consistency (though much preferable to inconsistency) is still a weak
requirement. After all, the members of a consistent set might all be false. Our
hypothesis about the use of the principle of charity would predict that scholars would
next turn their attention to the truth or reasonableness of the various premises of the
TMA, and this, in fact, is what we find. In the face of a long tradition of ridiculing the
Self-Predication assumption as too absurd to attribute to Plato,21 Sandra Peterson
published a paper entitled “A Reasonable Self-Predication Premise for the Third Man
Argument.”22 In it she proposed a version of SP both plausible enough to be attributed
to Plato and powerful enough to fill the gap in the TMA. The use of the principle of
charity was raised to another level.
The key to Peterson’s interpretation is to treat instances of SP as being akin to what
she calls “Pauline” predications. One of her standard examples of this kind of
predication, appropriately enough, is the sentence ‘Charity is kind.’ People normally
take it to express something true, she notes, even though its subject (‘Charity’) names
an abstraction (what Plato would call a Form), and its predicate (‘ . . . is kind’) seems
inappropriate to such an entity. So we should not reject self-predications as absurd
merely because their predicates seem inappropriate to their subjects.
 9

What, then, do self-predications mean? Consider Pauline predications again. In

saying that charity is kind, we may mean something like this: charity is a virtue that
causes those who have it to be kind. Similarly, the sentence ‘Justice is just’ may be
construed as asserting (plausibly enough) that Justice is a Form which causes its
participants to be just. The Form may be just in a different way from the way in which
its participants are just; but this does not mean that we equivocate on ‘is just’ when we
say that each of them is just. Peterson puts the point as follows:23

The difference between the way the F is F and the way many of its
participants are F is best brought out by saying that the F is a form
which is F perhaps solely by bringing about that its participants are F.
Many of its participants may be F without being able to have
participants. Such categorial differences, however, do not make a
difference to what it is to be F . . . .
Hence not only can one say, without absurdity, that F-ness is F; one can also say that
F-ness and its participants are all F in the same sense. That is, SP is both reasonable to
hold and powerful enough to help generate the TMA.
There are, of course, degrees of reasonableness. The highest degree (at least among
mortals) is that of a first-rate contemporary philosopher. So our hypothesis about the
use of the principle of charity would now predict that some contemporary scholars
would find such a high degree of plausibility in Plato’s metaphysics. One scholar who
makes a strong case for reading Plato’s dialogues as a contemporary text is Terry
Penner.24 Indeed, Penner treats Plato himself as a participant in a dialogue with Frege
over contemporary issues in philosophy of language.25 Consequently, historians of
philosophy must also be thoroughly versed in contemporary theorizing. Penner writes
(p. 288):

You can’t expect to do good exegesis of passages on Plato’s Forms—or

indeed on almost any other metaphysical topics in Plato—without making
up your own mind on matters as fundamental as the nature and
metaphysical presuppositions of logic, and the sources of the paradoxes
of logic, semantics and set theory.
Since Plato is on the cutting edge of contemporary thought, so, too, must his
interpreters be, if they are going to understand him. Penner’s work thus carries us to
the fourth stage of the paradox of interpretation where the distinction between historical
and contemporary philosophizing breaks down altogether.

4. The Principle of Parsimony

An interpreter who is guided solely by the principle of charity is doomed, it seems,
to commit the twin sins of overinterpretation and anachronism. A second principle
seems demanded to curb the excesses of the first. Charity needs to be limited by some
principle of economy. One plausible candidate is the traditional principle of parsimony
or simplicity. In following such a principle an interpreter seeks the simplest explanation
for the text before him. Other things being equal, the simplest explanation for a writer’s
use of a particular sentence is that it provided the most straightforward way of
expressing what he wanted to say. The simplest explanation for a missing premise in an
argument is that its author, being human, failed to notice that his argument is invalid
without it. And the simplest explanation for an apparent inconsistency in a
philosophical work is that the philosopher’s thought actually is inconsistent.
Most interpretation involves balancing charity and parsimony. The stage is set for a
use of the principle of charity whenever the evidence supporting an interpretative
 10

hypothesis underdetermines the hypothesis—that is to say, whenever the evidence is

compatible with the falsity of the hypothesis. The principle of charity is used at every
step from establishing a text to supplying the tacit premises in an argument. A modern
edition of the Greek text of a Platonic dialogue with its accentuation, punctuation, and
separation of words is far removed from the unaccented, unpunctuated, unbroken
string of capital letters actually written by Plato. Since there is often more than one way
to divide an unbroken string of capitals into individual words and the words into
sentences, the principle of charity must be invoked to choose among competing
hypotheses. A grammatical sentence, for example, will be preferred to one that is
ungrammatical; a grammatical sentence that makes sense to one that does not; and a
true sentence to a false one. If the text contains an argument, the identification of its
explicit premises and its conclusion is often a matter of dispute; and once identified,
these sentences are still subject to conflicting interpretations. The stage is thus set for
further uses of the principle of charity.
If the paradoxical results outlined earlier are to be avoided, the principle of charity
must be reined in at some point by an opposing principle of parsimony. There is, as far
as we can see, no third principle for determining the point where charity and parsimony
are properly balanced. This is why one cannot hope for consensus among interpreters.
The situation is similar to that which arises in an ethical system that tries to balance a
maximizing principle (say, of utility) and a fairness principle. In both situations the
balancing of the two principles involves judgment and common sense (nebulous as
these are).
There is, however, a special problem when it comes to supplying the missing
premise in a real enthymeme. The simplest explanation for the missing premise is that
the author of the argument was unaware that it was needed. Once one moves beyond
this explanation, the principle of parsimony loses its foothold and seems unable to
provide any restraint on the principle of charity. For in interesting cases there will
always be an indefinitely large number of interpretative possibilities; and since the
enthymeme is real and not apparent, there is no broader context to provide a basis for
choosing among them. Although an interpretation of a text is always underdetermined
by the evidence, in normal cases an interpreter has at least some evidence he can
appeal to in ranking one interpretative hypothesis ahead of another. In the case of a real
enthymeme such evidence does not exist. The various ways of filling the gap in a real
enthymeme are not simply underdetermined by the evidence—they are hyper-
underdetermined.

5. An Example from Euclid

It may be helpful in dealing with this problem of hyper-underdetermination to
contrast the interpretation of the Third Man Argument with the interpretation of a text
from the closely related area of the history of mathematics. Euclid’s Elements have been
studied as closely as Plato’s dialogues and present many of the same problems of
interpretation. One of the major shortcomings of the Elements from a modern
perspective is their constant use of tacit premises. Furthermore, the Euclidean scholar
is likely to be as well versed in modern mathematics as the Platonic scholar in modern
philosophy. So we might expect the one to be as prone to anachronism and
overinterpretation as the other. But this turns out not to be the case. Thus it may be
enlightening to consider how the tools of modern mathematics are used in the
interpretation of an ancient mathematical text.
The very first proof in Book I, which shows how to construct an equilateral triangle
on a given finite straight line, is a good example. The proof, divided into discrete steps,
runs as follows:26
1. Let AB be the given finite straight line.
 11

2. With center A and distance AB, let the circle BCD be described. (Postulate 3)
3. With center B and distance BA, let the circle ACE be described. (Postulate 3)
4. The circles cut one another at point C.
5. From the point C to the points A and B let the straight lines CA and CB be
joined. (Postulate 1)
6. Since the point A is the center of the circle CDB, AC is equal to AB. (Definition
1527)
7. Since the point B is the center of the circle CAF, BC is equal to BA. (Definition
15)
8. Therefore, AC is equal to BC. (From 6 and 7 by axiom 128)
9. Hence the three straight lines AC, AB, and BC are equal to each other. (From 6,
7, and 8)
10. Therefore, the triangle ABC is equilateral. (Definition 2029)
An obvious flaw in this proof and one that has often been pointed out30 is that no
justification is given for step (4). What guarantees that the two circles intersect at a
point and do not slip through each other without touching? Step (4) tacitly assumes
that Euclidean circles are continuous. Can this be proved? Is there anything in Euclid’s
postulates that ensures that Euclidean lines have no gaps?
Here are Euclid’s five postulates:
“1. Let it be postulated to draw a straight line from any point to any point,
2. and to produce a limited straight line in a straight line,
3. and to describe a circle with any center and distance,
4. and that all right angles are equal to one another,
5. and that, if one straight line falling on two straight lines makes the interior
angles in the same direction less than two right angles, the two straight lines, if
produced ad infinitum, meet one another in that direction in which the angles
less than two right angles are.”31
One way to settle the question of continuity is to see if there is an interpretation of
the concepts that enter into these five postulates and into the above proof under which
the postulates are true but under which step (4) is false. Such an interpretation can be
found by exploiting the techniques of analytic geometry and set theory. The basic idea is
to give the relevant terms an arithmetic interpretation in the domain of rational
numbers. The domain of rational numbers is chosen since a line whose points
correspond to rational numbers, though everywhere dense (between any two points
there is a third), is not continuous. (There is a gap, for example, between all the points
of such a line greater than the square root of two and all the points less than the square
root of two.) Following this strategy each of the relevant terms is assigned an arithmetic
meaning that corresponds by way of a Cartesian (or rectangular) coordinate system to
the intended geometric meaning of the term.
Under this arithmetic interpretation the word ‘point’ means ‘ordered pair of rational
numbers’; ‘straight line’ means ‘set of points that satisfy an equation of the form ax + by
+ c = 0’, and ‘circle’ means ‘set of points that satisfy an equation of the form x2 + y2 + ax
+ by + c = 0’. (In these equations and in those that follow ‘x’ and ‘y’ are variables and all
other letters are constants.) These two equations are chosen since the graph of the first,
using a Cartesian coordinate system, is a geometric straight line and that of the second
 12

is a geometric circle. Finally, ‘line AB intersects line CD’ means ‘the intersection of the
set of points identified with line AB and the set of points identified with line CD is not
the null set’.
Consider now Euclid’s five postulates. Postulate Four is provable,32 so it can be set
aside. The other four are all true under the foregoing arithmetic interpretation when it is
elaborated in an obvious way.
If the points mentioned in Postulate One are the two ordered pairs <h1, k1> and <
h2, k2>, the following set is the straight line through these points (where Ra is the set of
rational numbers):

{<x, y> | x,y ∈ Ra & (k1 - k2)x + (h2 - h1)y + h1k2 - h2k1 = 0 }.
Under the arithmetic interpretation Postulate One makes the true assertion that this set
has members. Postulate Two makes the true assertion that if h1 ≠ h2, then for any given
rational number, n, the set contains an ordered pair whose first element is larger than n
and a second ordered pair whose first element is smaller than n, and it makes a similar
true assertion for the second element if k1 ≠ k2. (If h1 = h2, the line is parallel to the y-
axis; and if k1 = k2, it is parallel to the x-axis.)
In order to interpret Postulate Three it is necessary to specify that by ‘any center’ is
meant ‘any point as center’ and by ‘distance’ ‘distance between two points’. If this is laid
down, then under the arithmetic interpretation Postulate Three makes the true
assertion that the following set, which is determined by the equation for a circle with
the center <h,k> and the radius r, is not null:
2 2 2
{ <x, y> | x,y ∈ Ra & (x - h) + (y - k) = r }
(In the equation determining this set r2 must always be a rational number though r can
be irrational.)
The fifth, or parallel, postulate is the most complex. Consider the two following sets:

A = { <x, y> | x,y ∈ Ra & a1x + b1y + c1 = 0 }

B = { <x, y> | x,y ∈ Ra & a2x + b2y + c2 = 0 }

Postulate Five asserts that if A and B each have at least two members and if a1/a2 ≠
b1/b2, then A ∩ B ≠ 0. This amounts to the claim that the equations that determine A
and B have a simultaneous solution in the domain of rational numbers. It is established
by showing that a linear equation that is satisfied by two distinct ordered pairs of
rational numbers has rational coefficients33 and that the simultaneous solution of two
linear equations with rational coefficients must be a pair of rational numbers.34
All of Euclid’s postulates are true under the proposed arithmetic interpretation, but
under this interpretation the key step in Euclid’s proof of his first theorem and the
theorem itself are false. For suppose that <-1,0> and <1,0> are the end points of the
finite straight line upon which the equilateral triangle is to be constructed. Then under
the arithmetic interpretation the two circles are:
2 2 2
{ <x, y> | x,y ∈ Ra & (x - 1) + y = 2 }
2 2 2
{ <x, y> | x,y ∈ Ra & (x + 1) + y = 2 }
But these sets do not intersect. (In the domain of real numbers the points of
intersection are { <0, √3>, <0, -√3> }.) Thus Euclid’s first proof cannot be repaired, and
 13

the reason is that his postulates do not guarantee that the lines of his geometry are
continuous.
Euclid’s geometry is in need of a continuity postulate. What form should it take?
There are various possibilities of increasing levels of generality. The following postulate
is sufficient to deal with the gap in Euclid’s first proof:
1. If A is the center of one circle and B the center of another and if the straight line
joining A and B is a radius of both circles, then the two circles cut one another
at a point C.
The difficulty with this postulate is that it is too specific. Unlike Euclid’s other
postulates, it speaks to one case only: it guarantees that two circles have a point in
common only when the circles are the same size and share a radius. It seems unlikely
that Euclid would have considered it sufficiently general to deserve a place among his
other postulates.
To increase generality one might try something like the following:
2. If one point of a circle lies inside and another point lies outside another circle,
then the two circles have exactly two points in common.
This postulate may still be too specific. Does it, for example, entail that a straight
line that joins a point inside a circle with a point outside the circle intersects the circle
at a point? A further question is whether this postulate expresses the assumption about
continuity that Euclid was tacitly assuming in his first proof. Although a charitable
interpretation fuses these two questions, parsimony keeps them apart. A parsimonious
interpreter will point out that proposition (2) is not sufficient by itself to bridge the gap
in Euclid’s first proof. Before Euclid can appeal to proposition (2) he will need to
establish its antecedent, a process requiring several additional steps. A charitable
interpreter will need to supply these steps as well as proposition (2). Charity is getting
out of hand. A simpler, more parsimonious interpretation is clearly preferable: Euclid
did not notice the hole in his first proof and did not realize that his postulates do not
guarantee the continuity of his lines. Although a continuity postulate and a revised
proof are needed to establish Euclid’s first theorem, both of these belong, not to Euclid’s
geometry, but to Euclidean geometry.35
By pursuing this example in such detail we have tried to show, among other things,
how modern techniques can deepen one’s understanding of an historical text. A scholar
using the tools of modern mathematics can understand Euclid’s Elements better than
Euclid understood them himself. Thus Euclid could have discovered that his first proof
is defective by noticing that no justification is offered for step (4). But an unproved
proposition might still be provable. The tools of analytic geometry and model theory are
needed to show that Euclid’s first theorem is not provable from his postulates and why.
Such insights were beyond Euclid’s scope. The main point of the example, however, is
that in the history of mathematics a sharp line is drawn between Euclid’s geometry and
Euclidean geometry, even though the latter rises on the basis of the former. The
principle of parsimony seems to play a larger role in the interpretation of mathematical
texts than it does in the interpretation of philosophical ones.

6. Two Models of Interpretation

When an interpreter of Plato supplies a suppressed premise in a real enthymeme,
what is his aim? There are two models. One is retrospective; the other, prospective. On
the retrospective model the interpreter supposes either that Plato consciously
entertained the suppressed premise but, for one reason or another, did not write it
down, or that there is a premise he would have supplied if he had been queried about
the gap in his argument. The interpreter’s job, on this model, is to discover the idea that
 14

was in Plato’s head or the answer he would have given if the question had been put to
him. Thus his interpretation is correct or incorrect depending upon whether or not the
premise he supplies corresponds to an actual or latent premise of Plato’s. On the
prospective model, on the other hand, the interpreter supposes that the gap in Plato’s
argument reflects a gap in Plato’s thinking. When the interpreter fills the gap, he
considers it a free act of creation on his part. His goal is not to recapture Plato’s thought
(since there is no thought to recapture) but to construct as good an argument as
possible on the foundation that Plato lays.
An analogy may be helpful. On the retrospective model the interpreter is like a
scholar who is attempting to establish a text from a sole surviving manuscript that
happens to be worm-eaten. Due to the worm holes one of the important words in the
manuscript is missing. The scholar makes a conjecture about this word. His conjecture
is either true or false depending upon the word the author actually wrote. There may be
a serious epistemic problem about recovering this word, but nevertheless one word is
correct and all others wrong. On the prospective model, on the other hand, the
interpreter is like a poet whose help is sought by a colleague having difficulty finding
the right word to end a stanza of a poem he is writing. In this case the right word is the
one that makes the best poem. There is no question of truth and falsity.
Both models have their place in the interpretation of Euclid. The historian of
mathematics who is intent on recapturing Euclid’s thought adopts the retrospective
model, whereas the mathematician who repairs Euclid’s proofs by supplying missing
premises adopts the prospective model. (The same person might, and often will, wear
both hats.) This does not mean that the historian eschews the principle of charity, and
the mathematician, the principle of parsimony. As we have already pointed out, one
cannot even establish a text without using the principle of charity. Thus a retrospective
interpretation will be guided by charity as well as parsimony. Similarly, parsimony as
well as charity has a role in prospective interpretation. The principle of charity exhorts
the interpreter to maximize validity, truth, and content. Since there is always more than
one way to do this—since there is always more than a single true (or reasonable)
proposition that will restore the validity of a real enthymeme—a second principle is
needed to guide the choice among the candidates nominated by the principle of charity.
The principle of parsimony, in exhorting the interpreter to choose the simplest, is a
principle of elegance. Although the principle of parsimony plays a role in both models of
interpretation, it functions differently in the two. In the prospective interpretation of a
real enthymeme, the principle of parsimony guides the interpreter to the simplest, most
elegant supplement of the text; in the retrospective interpretation of such an
enthymeme, it guides him to the simplest explanation for the text.
It is our contention that the interpretation of Euclid provides a guide for the
interpretation of Plato. The missing premises in the real enthymemes in Plato’s
dialogues reflect gaps in Plato’s thought just as the missing premises in Euclid’s proofs
reflect gaps in his thought. And in both cases, when an interpreter supples a missing
premise, he is extending his author’s thought rather than expounding it. As the
distinction between retrospective and prospective interpretation leads in the one case to
the distinction between Euclid’s geometry and Euclidean geometry, it leads in the other
to that between Plato and Platonism.

7. Two Objections
One objection that might be made to the prospective model of interpretation is that
it presupposes that there are real gaps in Plato’s arguments, that is, gaps that cannot
be filled by scouring the dialogues and by increasing one’s background knowledge of
Greek culture. But, the objection goes, there are no such gaps. Plato did not write in a
vacuum, and research will always permit a diligent scholar to isolate the premise Plato
 15

intended his reader to supply. According to this objection, all the enthymemes in the
dialogues are apparent; none are real.
But how are we to understand the claim that there are no real enthymemes in
Plato? Is it a maxim of interpretation or is it a factual generalization over all the
arguments in the dialogues? If it is a maxim, the objector is begging the question. For to
assert as a maxim that there are no real enthymemes in Plato is simply to use the
principle of charity retrospectively, and the issue is whether there are cases where such
retrospective use is inappropriate. If, on the other hand, the claim that there are no real
enthymemes in Plato is a generalization, it will be difficult to establish, as negative
existential propositions typically are. For in each case the evidence one brings forward
to fill the gap is bound to be controversial.
Suppose, for example, one attempts to support the claim that the TMA is not a real
enthymeme by producing external evidence that Plato indeed subscribed to literal Self-
Predication, one of the argument’s implicit premises. One might point to such a passage
as Symposium 210e-211b where Plato clearly asserts that the Form of Beauty is
perfectly beautiful. Or, appealing to Plato’s characterization of the Forms as
paradeigmata (Rep. 472c4, 484c8, 500e3, 540a9, Parm. 132d2, Tim. 29b4, 31a4, 39e7,
48e5, and elsewhere), an intrepid gap-filler might even refer beyond the dialogues to the
nature of the paradeigmata used by Greek craftsmen. In Greek architecture they were
three-dimensional full-scale models. This is how they are described in a recent book on
Greek architects:

The paradeigma . . . was used for elements like triglyphs or capitals

which required a design in three dimensions, and in cases where carved
or painted decoration had to be shown. These specimens were often made
of wood, stucco, or clay, even where they were to be copied in more
permanent materials; but in at least one case the material was stone, for
the specimen capital was to be set in place in the actual building, along
with the others. In this case the specimen must obviously have been full
size, and that is likely to have been normal, for the use of full-size
specimens would be the easiest way of ensuring the uniformity which is
so important to the effect of a Greek building.36
This practice is noteworthy because it shows that the paradeigma of a capital could
itself be used as a capital. In some cases, at least, paradeigmata were literally self-
predicational.
Evidence of this sort is extremely important and should be avidly sought by anyone
attempting to understand Plato’s dialogues. The particular evidence just cited lends a
good deal of support to the claim that Plato subscribed to literal Self-Predication. But a
scholar who finds the idea of literal Self-Predication absurd will not find it compelling.
Such a scholar can avoid the implications of the Symposium passage either by denying
the relevance of an earlier dialogue in interpreting the Parmenides (Plato changed his
mind in the interval between the two dialogues) or by denying the legitimacy of inferring
that Plato held that all Forms are literally self-predicational from the fact that he held
that one Form, the Form of Beauty, is literally self-predicational.37 The external evidence
from Greek architecture is even easier for an opponent of literal Self-Predication to
discount; for one can hold that Plato’s Forms, even though described as paradeigmata,
are only analogous to the literally self-predicational paradeigmata used by Greek
craftsmen. All analogies break down; this analogy breaks down, one can argue, on
precisely the matter of literal self-predication. So the best external evidence that we
have for ascribing this tacit premise to Plato falls short of being conclusive. Thus it will
be virtually impossible to establish on the basis of either internal or external evidence
that no enthymeme in Plato’s dialogues is real.
 16

A second objection that might be made to the prospective model of interpretation is

that it presupposes that the arguments in Plato’s dialogues are assertoric: it
presupposes, that is, that Plato is attempting to establish certain theses by advancing
arguments for them. But, the objection goes, this presupposition is obviously false since
Plato never advances any thesis in his own name. Certain arguments are presented by
the characters in his dialogues, but Plato never tells the reader how these arguments
are to be taken. And there are various possibilities. Plato may intend the reader to take
a given argument as an example of sophistry, or as an interesting intellectual
experiment, or as a piece of rhetoric, or as raising an interesting problem, or as an ad
hominem, or simply as a report of an argument current in fourth or fifth century
Athens. The prospective model of interpretation supposes, so the objection goes, that
Plato is a forerunner of Spinoza; but this is simply to misunderstand the point of writing
dialogues.
The interpretation of the TMA advanced by R.E. Allen and by Harold Cherniss
provides a good example of a maneuver that illustrates this last objection.38 According
to Allen and Cherniss, Plato deliberately presented an argument he knew to be invalid.39
They hold that the TMA tacitly assumes that Largeness is large, but they also hold that
this self-predicational language expresses nothing more than a trivial claim of self-
identity. When Plato asserts that the Form of justice is just, for example, he is simply
making the true (and trivial) assertion that justice is justice. Hence all that the tacit
assumption that Largeness is large can mean for Plato (and all he is committed to) is
that Largeness is Largeness. Since this premise is too weak to generate the regress, the
TMA poses no threat to Plato’s theory of Forms. According to this interpretation, the
TMA is an aporia, or perplexity, to which Plato knew the answer, not an assertoric
argument indicating a flaw in the theory of Forms.40
Although strategies of this sort can be used to evade the putative conclusion of any
given argument in the dialogues, it seems to us to be methodologically unsound to use
them to evade the putative conclusion of every argument in the dialogues. To do so
would be to sacrifice the idea that Plato is ever committed to any thesis that he is
prepared to argue for. If Plato’s philosophy is not to be thus cut off from his own
expression of it, it must be conceded that at least some of Plato’s arguments are
assertoric. And among these it should not be too difficult to find a real enthymeme. So
the prospective model is back in business.

8. Conclusion
Given that the prospective model of interpretation leads to Platonism rather than to
Plato, the question arises whether prospective interpretation has a place in Platonic
scholarship. We believe it does. For without it the interpretation of Plato is dry and
barren, lacking in intelligence and imagination. Consider what a purely retrospective
interpretation of the TMA would amount to. It could establish no more than that the
argument is a non sequitur. The aridity of interpretation that is purely retrospective
explains, no doubt, the need for both models of interpretation in accounting for the
actual practice of historians of philosophy. And this need partially explains in turn why
the study of the history of philosophy is such a peculiarly philosophical enterprise. To
do good work in the history of philosophy, it has often been observed, one must be a
good philosopher, not just a good historian. Why this should be so is not a simple
question,41 but one that is much easier to answer if philosophical interpretation has a
prospective as well as a retrospective component. For while he is engaged in prospective
interpretation, the historian of philosophy is augmenting the philosophical work of his
subject: that is, he is doing philosophy.
If a Platonic scholar needs to employ both models of interpretation, he also needs to
maintain the distinction between them. Otherwise he will end up attributing his own
 17

contribution to Plato. He will end up conflating Platonism and Plato. He will be tempted,
for example, in searching for the true point, meaning, or moral of a text, to discover one
bestowed on the text by his own augmentation of it. Interpreters of the TMA have often
succumbed to this temptation. There is no such thing as the moral of the TMA if, as we
contend, retrospective interpretation is unable to advance beyond the observation that
the TMA is a non sequitur.42
We suggest, finally, that Plato himself might find our view of interpretation
congenial. For Plato’s complaint about written words—that they do not respond to
questions but repeat the same thing endlessly (Protagoras 329A, Phaedrus 275D)—
corresponds to our complaint about the sterility of retrospective interpretation. On the
other hand, since Plato devoted so much of his life to putting words on paper, he must
have hoped that this defect of writing could sometimes be alleviated. He must have
hoped that his words would occasionally kindle a philosophical dialogue in the mind of
an attentive reader, albeit a dialogue that the reader would have to conduct on his own
or with another prospective interpreter.43
University of Washington

NOTES

1 Since two premises can always be reduced to one by means of conjunction, no

enthymeme stands in need of more than a single additional premise.
2 The premise that the one supplies need not itself be inconsistent with the premise
supplied by the other. But so long as the premises are distinct, they cannot both be
the most reasonable one to supply.
3 One finds an instance of this in the famous Simonides episode of Plato’s Protagoras
(339a-347a). The text that Socrates is constrained to interpret contains these lines
(345d):
But all who do no evil
Voluntarily I praise and love.
Charitably refusing to attribute a false belief to the poet, Socrates construes
‘voluntarily’ with the words that follow rather than with those that precede.
4 Parm. 132a1-b2. The translation is our own.
5 It is almost universally assumed that Plato intended the TMA to hold for any
predicate for which there is a Form; hence the schematic letter ‘F ‘ is typically used
in place of ‘large’ to express this generality. For an interesting alternative view, see
William E. Mann, “The Third Man = The Man Who Never Was,” American
Philosophical Quarterly 16 (1979), pp. 167-76.
6 “The Third Man Argument in the Parmenides,” Philosophical Review 63 (1954), pp.
319-49; reprinted with an addendum in Studies in Plato’s Metaphysics, ed. by R.E.
Allen (London, 1965), pp. 231-63. Subsequent references will be to the reprinted
version.
7 “Vlastos and ‘The Third Man,’“ Philosophical Review 64 (1955), pp. 405-37;
reprinted in Philosophical Perspectives (Springfield, IL, 1967), pp. 23-54.
Subsequent references will be to the reprinted version.
 18

8 For further details and a more precise characterization of this revision, see S. Marc
Cohen, “The Logic of the Third Man” (cited henceforth as LTM), Philosophical Review
80 (1971), pp. 452-53.
9 Sellars’ formulation of NI is still not quite right, as was pointed out in LTM, p. 453,
n. 14. The problem is that NIs entails (or presupposes) that there is such a thing as
the (unique) F-ness by virtue of participating in which a given F thing is F. This
conflicts with Plato’s idea that particular F’s are F in virtue of participating in the
first Form in the regress and also (along with that Form) in virtue of participating in
the second Form in the regress, etc. Hence, for Plato, there will not be such a thing
as the (unique) F-ness by virtue of which a given F thing is F. The correct Sellarsian
formulation of NI would be this: If x is F, then x is not identical with any of the F-
nesses by virtue of which it is F.
10 “Plato’s ‘Third Man’ Argument [Parm. 132a1-b2]: Text and Logic,” Philosophical
Quarterly 19 (1969), pp. 289-301. Reprinted with revisions in his Platonic Studies,
Second Edition (Princeton, 1981), pp. 342-65. Page references will be to the revised
version.
11 Strictly speaking, the three axioms are consistent but entail that nothing has the
given character. But clearly all participants in this dialogue agree that some things
are large.
12 LTM, pp. 448-75. Cf. note 8.
13 Our formulation of the Peano Postulates is due to Bertrand Russell, Introduction to
Mathematical Philosophy (London, 1919), pp. 5-6.
14 The occurrences of ‘Λ ∪’ are redundant under the usual interpretation of Λ as the
empty set but not under the reinterpretation to follow.
15 Recall that on the von Neumann construction, each NN is a member of all its
“descendants”; this, together with postulate 3, entails that no NN is its own
successor. The Form-theoretic analogue of this is that no Form in the sequence
participates in itself.
16 One might object that two distinct Forms may be immediately over the same object
since Socrates, for example, participates in both the Form man and the Form
philosopher. But these two Forms belong to different sequences, the Third Man
sequence and the Third Philosopher sequence. The immediately-over relation is a
function only with respect to a single sequence.
17 In LTM, “immediately over” was defined differently (but equivalently) in terms of the
over relation and the notion of the level of an object: x is immediately over y iff the
level of x is one greater than the level of y (whereas x is over y iff the level of x is
greater than the level of y). The notion of level is defined recursively in terms of
participation: F things in which nothing participates (i.e., F particulars) are on level
0; F things which have as participants all and only the things on level 0 are on level
1; F things which have as participants all and only the things on level n or lower
are on level n + 1. The possibility of simplifying the definition of immediate overhood
as we have in the text was also discovered (independently) by Richard Patterson.
See his Image and Reality in Plato’s Metaphysics (Indianapolis, 1985) p. 54.
18 Vlastos’s 1981 version of the TMA (Platonic Studies, Second Edition, p. 363)
contains a statement of OM which appears to quantify over sets of F’s, but this
appearance is misleading. His English version begins promisingly: “If any set of
things share a given character, say, large, then there is a unique corresponding
 19

Form, Largeness . . . .” But what is it to which this Form uniquely corresponds? His
wording seems to suggest: corresponding to whatever set the initial quantifier picks
out. That, we maintain, would be the right idea. But formalizing that idea requires
a quantifier ranging over sets, and no such quantifier is to be found in Vlastos’s
formalization. (In English, what his formalization says is this: if a, b, and c are all
large, then this is in virtue of their participating in a Form, Largeness, the one and
only Form in virtue of participating in which things are large.) And his glossary of
logical symbols confirms that he intends a unique Form corresponding to a given
character, as he did in OMv2, where the quantification over sets was purely
adventitious.
19 “The Third Man Again,” Philosophical Review 65 (1956), pp. 72-82.
20 “Plato and the Third Man,” Proceedings of the Aristotelian Society, supplementary
volume 37 (1963), pp. 147-63.
21 R.E. Allen and Harold Cherniss are two prominent members of this tradition. Their
view will be discussed below, p. 17.
22 Philosophical Review 82 (1973), pp. 451-70.
23 Ibid., p. 470. We have made minor revisions in Peterson’s notation to make it
conform to our own.
24 The Ascent From Nominalism (Dordrecht, 1987).
25 Ibid., pp. 57 ff.
26 Thomas L. Heath, The Thirteen Books of Euclid’s Elements, 2nd ed. (Cambridge,
England, 1926), vol. I, pp. 241-2.
27 “A circle is a plane figure contained by one line such that all the straight lines
falling upon it from one particular point among those lying within the figure are
equal.”
28 “Things which are equal to the same thing are also equal to one another.”
29 “Of the three trilateral figures, an equilateral triangle is the one which has its three
sides equal . . .”
30 Heath, op. cit., pp. 235, 242; Felix Klein, Elementary Mathematics from an Advanced
Standpoint: Geometry trans. E.R. Hedrick and C.A. Noble (New York, 1939), p. 197;
Howard Eves, A Survey of Geometry revised ed. (Boston, 1972), p. 321; Ian Mueller,
Philosophy of Mathematics and Deductive Structure in Euclid’s Elements (Cambridge,
MA, 1981), p. 28.
31 Mueller’s translation, op. cit., pp. 318-9.
32 The proof, which goes back at least to Proclus, is given in Mueller, op. cit., p. 22.
33 The equation for a straight line through the two points <h1, k1> and <h2, k2> is:

(k1 - k2)x + (h2 - h1)y + h1k2 - h2k1 = 0

Since h1, h2, k1, and k2 are all rational, the coefficients in the equation are rational.
34 A system of two linear equations can be solved by successive uses of addition and
multiplication. Hence if all of the coefficients of each equation are rational, the
solution must also be rational.
 20

35 In modern Euclidean geometry the continuity postulate that is favored is

Dedekind’s Postulate: “If all the points of a straight line fall into two classes, such
that every point of the first class lies to the left of every point of the second class,
there exists one and only one point of the line which produces this division of all
points into two classes, this division of the straight line into two parts.” On the
basis of this postulate the proof that the two circles of Euclid’s first proof share a
point C is quite elaborate. For details see Heath, op. cit., vol I, pp. 234-40.
36 J.J. Coulton, Ancient Greek Architects at Work (Ithaca, NY, 1977), p. 55.
37 This second alternative is a relatively weak response since the paradox of the Third
Man breaks out again with the Form of Beauty. The Third Man Argument will
simply be replaced by the Third Beauty Argument. Thus this response would at
most localize the paradox without eliminating it.
38 R.E. Allen, “Participation and Predication in Plato’s Middle Dialogues,” in Studies in
Plato’s Metaphysics edited by R.E. Allen (London, 1965), pp. 46, 59. Harold
Cherniss, “The Relation of the Timaeus to Plato’s later Dialogues,” in Allen, op. cit.,
pp. 370-4. See also Cherniss, Aristotle’s Criticism of Plato and the Academy (New
York, 1944), pp. 294-8.
39 Thus Cherniss writes: “I take it therefore as proved not only that both versions of
the regress are invalid arguments but also that when Plato put them into
Parmenides’ mouth he believed them to be invalid and invalid for reasons which he
felt himself to have indicated satisfactorily for anyone who would compare the
assumptions of these arguments with what he had already said concerning the
nature of the ideas” (in Allen, op. cit., p. 374).
40 See in this regard R.E. Allen, Plato’s Parmenides (Minneapolis, 1983), pp. 96-8.
41 For a useful examination of this unjustly neglected question, see David M.
Rosenthal, “Philosophy and its History,” in The Institution of Philosophy: a Discipline
in Crisis?, ed. Avner Cohen and Marcelo Dascal (LaSalle, IL, 1989).
42 Unless of course, as Cherniss and Allen maintain, the point is to present an invalid
argument.
43 Earlier versions of this paper were presented to the Philosophy Department at
Rutgers University and at the March, 1989 meeting of the Pacific Division of the
American Philosophical Association in Berkeley. We would like especially to thank
our two commentators on the latter occasion, Gerasimos Santas and Martin
Tweedale, for their helpful comments and criticisms. We also gratefully
acknowledge the help of Christine Keyt, Gareth Matthews, Nicholas Smith, Charles
Young, and the participants in the discussions at Rutgers and Berkeley.