ROBERT H.

E N N I S

IDENTIFYING IMPLICIT ASSUMPTIONS ~

Although we all make implicit assumptions, and frequently identify,

attack, defend, and use those made by others, the topic, implicit
assumptions, is a relatively neglected one. In this paper I treat this
topic in the hope that the distinctions made will be enlightening and
that the criteria suggested will be useful.
These distinctions and criteria can, if correct, be of use to everyone
who engages in argumentation (which is almost everybody), and in
particular to people who write for and teach courses devoted in whole
or part to informal logic. This includes not only courses of that name
in philosophy departments, but rhetoric and speech courses, courses
in law, and a wide variety of other courses and classes from primary
grades through advanced graduate school which aim to promote good
thinking.
My primary concern in this paper is with identifying implicit
propositions that are taken "for granted as a basis of argument or
action" (to crib from The Oxford English Dictionary). By "implicit
assumption identification" I mean the identification of such pro-
positions.

CONCLUSIONS, NOT-FULLY-ESTABLISHED PROPOSITIONS,

AND ADOPTIONS

Three common types of assumption are thereby not of primary

concern, but should be noted in order to avoid confusion: con-
clusions, less-than-fully established propositions, and adoptions. Here
is an example of a conclusion being called an assumption: "My
assumption is that you are going out, since you are wearing your
cap." Here is an example of a less-than-fully-established proposition
being called an assumption: "That's only an assumption. You don't
know it." (Calling something an assumption in this sense is an
accusation, though not necessarily a severe one, and the term is, in
this sense, pejorative to a degree.) And here is an example of an
adoption that is called an assumption: "His assumption of an air of

Synthese 51 (1982)61-86. 0039-7857/82/0511-0061 $02.60

Copyright © 1982by D. Reidel Publishing Co., Dordrecht, Holland, and Boston, U.S.A.
62 ROBERT H. ENNIS

humility was quite unbecoming." There is an element of less-than-

fully-establishedness alleged to attach to conclusions that are called
assumptions, so the first and second of the above types are asso-
ciated.
Some students, when requested to identify assumptions, often
interpret the request as one to list the less-than-fully-established
propositions involved, whether implicit or explicit. Listing such pro-
positions is in many situations an important thing to do, but it is not
the topic of this paper, which is also an important thing: identifying
implicit propositions that are taken for granted as a basis for
argument or action. So, if one wants students to produce such things,
instead of less-than-fully-established propositions, one should say so.
(Actually, in order that the activity be perceived to be significant, I
generally ask for implicit assumptions that in addition someone might
challenge.) In order to be able to make these practical suggestions and
other similar ones, and in order to avoid misunderstanding in the rest
of this essay, I mention here these three other kinds of assumptions:
conclusions, less-than-fully-established propositions, and adoptions.

UNSTATED PREMISES AND PRESUPPOSITIONS

One division of implicit propositions taken for granted as a basis of

argument and action is into unstated premises and presuppositions.
Roughly speaking, premises are propositions that are part of an
argument, the truth of which would be thought to contribute to the
support provided for the conclusion, and the falsity of which would
be thought to diminish or destroy the support provided for the
conclusion. On the other hand, a presupposition, roughly speaking, is
a proposition the falsity of which appears to make another pro-
position (that for which it is a presupposition) without point or truth
value. This notion of presupposition I take from Strawson
(1950, 1952) with amendments resulting in part from Donnellan's
(1966, 1968) criticisms. First a look at premise-type assumptions.

GAP-FILLERS AND BACK-UPS

Premise-type assumptions are at least of two types: those that pro-

vide backing for a proposition already thought to be part of an
IDENTIFYING IMPLICIT ASSUMPTIONS 63

argument ("back-ups"), and those that join with one or more other
premises in giving support to the conclusion ("gap-fillers").
To illustrate the difference, consider this brief argument:

EXAMPLE 1:
If Mike is a dog, then Mike is an animal.

Therefore Mike is not a dog.

In Example 1, the proposition, "Mike is not an animal" could be a

gap-filler, because it together with the given premise supports the
conclusion. It fills a gap in the given argument.
On the other hand, the proposition, "All dogs are animals", would
be a back-up because it backs up one of the premises. It does not join
with the given premise to support the conclusion.
The distinction between gap-filler and back-up premise-type
assumptions is helpful in informal logic instruction, because different
criteria apply to the identification of each, and they have different
places in the analysis of an argument, including tree-diagram
representations of an argument (Beardsley, 1975, pp. 16-20; Scriven,
1976, pp. 41-43; and Thomas, 1977, pp. 21-46, among others, offer
tree-diagramming methods of argument analysis).

NEEDED AND USED ASSUMPTIONS

Sometimes implicit assumptions are propositions that are needed to

support the conclusion, to make the argument a good one, to make a
position rational, etc. On the other hand, sometimes they are unstated
reasons that a person actually used consciously (or subconsciously, if
you believe in subconscious reasons) as a basis of argument or action.
These are different kinds of things. A person might need a particular
proposition, but not use it in a piece of reasoning; and a person might
use a particular proposition, but not need it in a piece of reasoning.
Furthermore, the same proposition might be both needed and used. I
mark the distinction with the terms "needed assumption" and "used
assumption." It is important to make the distinction because their
associated assumption-identification claims are significantly different,
resulting in the use of different criteria for identification of each type.
64 ROBERT H. E N N I S

USED ASSUMPTIONS

Used assumptions are unstated reasons. We look for them when we

are engaged in understanding someone. Teachers trying to understand
their students, students trying to "psych out" their teachers, psy-
chologists and lawyers trying to understand their clients, voters trying
to learn what a candidate's views are, and citizens trying to under-
stand the thinking of their country's bureaucracy ("understanding the
bureaucratic mind"), engage in identifying used assumptions. One
might want to secure this sort of understanding of a person in order to
judge motives, or in order to predict what the person will do in the
future, or to correct a mistaken view held by the assumer, or, as
counselors sometimes do, to reflect back to a person the nature of
that person's thinking to help the person understand himself or
herself.
Claims that a used assumption has been identified are empirical
mental-event-or-state claims about the thinking of a person. Roughly
in accord with the suggestion of Gilbert Harman (1973), I am content
to regard such unstated-reason-identification claims, when not
obvious, as best-explanation claims, that is, as explanatory hypo-
theses.
Although I have presented used assumptions as things that persons
have, one might also want to allow them to be things that institutions
have, if one will allow that institutions have reasons. People do try to
figure out the thinking of the Board of Trustees of the University of
Illinois, and to predict what the Board will decide. But I shall not
argue this point. If you think that institutions can be personified to
this extent, then I suppose you will allow this extension of the idea of
implicit reasons. If not, then not. Whatever can have a reason can
have an unstated reason, I shall assume, and is a potential subject for
used-assumption identification.

NEEDED ASSUMPTIONS

Needed assumptions are very different creatures. When identifying a

needed assumption we are searching for something that an argument
or action needs - for some purpose, usually unspecified. Keith Don-
nellan was doing this, I believe, in the following statement:
Strawson and Russell seem to make a c o m m o n assumption about the question of how
IDENTIFYING IMPLICIT ASSUMPTIONS 65

definite descriptions function: that we can ask how a definite description functions in
some sentence independently of a particular occasion upon which it is used. (1966, p.
282)
As I piece together Donnellan's argument, he is not saying that
Strawson and Russell actually thought this (although they might
have). Rather he is saying that their arguments need this alleged
assumption to be true in order that the arguments be good arguments,
or as good as can be.
If, contrary to my interpretation, Donnellan had been alleging that
Strawson and Russell actually believed and used the alleged assump-
tion, but did not need it, then, although he might have made them look
bad, he would not have been offering a strong challenge to their
results, because he would not then have been claiming that their
results (or arguments) needed the assumption (which he also held to
be false). Donnellan in this paper did, among other things, appear to
want to challenge the Strawson and Russell views, not just the men.
F. C. Copleston, in a debate with Bertrand Russell about the
existence of God, identified what he regarded as a needed assump-
tion:
I cannot see how science can be conducted on any other assumption than that of order
and intelligibility in nature. (Russell & Copleston, 1948, p. 124)
Copleston was not here claiming that scientists all think that nature is
orderly and intelligible, as I read Copleston. Instead he was claiming
that they need to hold this proposition in order that their activity
make sense - or be rational - or something to this effect.
John Stuart Mill claimed that an implicit assumption is necessary in
an inductive argument he discusses:
The induction, "John, Peter, &c., are mortal, therefore all mankind are mortal," may,
as he [Archbishop Whately] justly says, be thrown into a syllogism by prefixing as a
major premise (what is at any rate a necessary condition of the validity of the
argument) namely, that what is true of John, Peter, &c., is true of all mankind. (1879,
Ch. III, #2)
Here the claimed assumption was alleged to be necessary for the
validity of the argument. Mill was not claiming that the allegedly
assumed proposition is believed and consciously or unconsciously
employed by everyone making such an argument. Rather he was claim-
ing that the given generalization is needed in order to make a valid
argument out of the one given.
66 ROBERT H. E N N I S

Goals of Needed-Assumption Identification

There are primarily two different goals one might have in claiming to
have identified needed assumptions. One frequent one is evaluating a
conclusion or argument. The other is securing support.

The Goal of Evaluating a Conclusion or Argument. Donnellan

(quoted above) appeared to be using his attempted discrediting of the
alleged assumption as a way of discrediting the conclusions of
Strawson and Russell - or at least of discrediting their arguments
leading to the conclusions. A classic procedure is exemplified here: a
needed assumption is identified. It is judged for its acceptability,
which judgment reflects on the acceptability of either the conclusion
or the argument that assumes it. Usually people reporting the results
of this classic strategy allege to have identified a false assumption and
feel that they have thus discredited an argument or conclusion, but
the strategy can have as a result the alleged identification of a true
assumption, and the strengthening (or at least not weakening) of the
conclusion or argument.

Seeking Support. A second goal is to get support for some position

held by the assumption identifier. The strategy is to show that the
person or people doing the assuming are committed to some position
of the assumption identifier. Copleston (quoted above) seemed to be
doing this in alleging that scientists need to assume order and in-
telligibility in nature. Copleston's alleging identification of an
assumption was not performed in order to evaluate the work of any
particular scientist nor of all scientists. Rather, he wanted to show
that they were obligated to be on his side of the issue, "Is nature
orderly and intelligible?", the idea being that since most of us (in-
cluding Russell) accept the work of scientists, we are thereby obli-
gated to accept an assumption of this work. Copleston was seeking
support for his position.

CRITERIA FOR IDENTIFICATION OF PREMISE-TYPE

ASSUMPTIONS

As we might expect from the significant differences between used and

needed assumptions and between purposes for identifying needed
IDENTIFYING IMPLICIT ASSUMPTIONS 67

assumptions, criteria for their identification differ as well. First a look

at used-assumption criteria.

Used-Assumption Criteria
Since used assumptions are like beliefs, if not always beliefs, I regard
used-assumption identification claims as singular explanatory hypo-
theses, unless they are utterly obvious (thus making criteria un-
necessary anyway). Criteria for selecting correct singular explanatory
hypotheses have not been agreed upon by philosophers - or any other
group that is sensitive to the problems - but the following rules of
thumb (modeled after criteria for explanatory hypotheses that I have
offered elsewhere (1980, p. 11)) I suspect will be generally acceptable,
though vague. They require intelligent informed judgment in ap-
plication.
A used-assumption identification claim is justified roughly to the
extent that:

(a) It explains a bulk and variety of the data (consisting mostly of

facts about the assumer, especially the explicit parts of the argument).
("Jones' assuming X explains why he said he could infer Y from Z,
etc.")
(b) It fits in with what else is known about the person and situation.
("We could expect Jones, given the kind of life he has led, to believe
X.")
(c) It is not inconsistent with the facts of the situation.
(d) Its competitors do not explain, or are inconsistent with the data.
("Jones' assuming W would not explain why he said he could infer Y
from Z.")
(e) It is simpler than its competitors. Simplicity, a controversial
criterion, should be invoked only as a last resort, but at least this
criterion has explicit justification in unstated-reason identification:
people's beliefs, in my experience, tend to be fairly simple.
Let us apply these criteria to an example of identifying an unstated
reason as a gap-filler. Imagine the following short dialogue:

E X A M P L E 2:
Teacher: What type of word is "skiing" in the sentence, "Skiing is
dangerous?"
68 ROBERT H. ENNIS

Student: A gerund.
Teacher: Why?
Student: Because it ends in "ing".

In trying to understand the student and to decide what to do next the

teacher has a number of alternative possible gap-fillers, including:
1. All words ending in "ing" are gerunds.
2. All verbals ending in "ing" are gerunds.
3. All verbals ending in the suffix "ing" and serving as the subject
of a sentence are gerunds; "ing" is a suffix in "skiing".
4. All verbals ending in the suffix "ing" and functioning as a noun
are gerunds; "ing" is a suffix in "skiing".
5. All gerunds end in the suffix "ing"; "ing" is a suffix in "skiing".
The student's belief in any of the first four candidates for the
assumption could account for the student's concluding on the basis of
the word's ending in "ing" that it is a gerund. But if the student
customarily reasons fallaciously, a belief in the fifth could account for
the student's concluding that "skiing" is a gerund on the basis of its
ending in "ing". The fact that the fifth is a possibility should make it
clear that the process of identifying used assumptions is not neces-
sarily the rational reconstruction of an argument or line of reasoning.
Some people do regularly reason fallaciously, and this fact must be
taken into account in figuring out what they are thinking.
Suppose that, because this student never makes simple fallacy
errors, the teacher decides that the student was not reasoning fal-
laciously, and further suppose that for some reason or other, no other
candidates for the gap-filler are plausible candidates. On the basis of
these suppositions, let us limit ourselves to choosing among the first
four, a belief in any of which would explain (given the offered
"Because it ends in 'ing'") why the student gave the answer, "A
gerund."
The first two candidates are false. The next two are true, as I
understand things. At least let us so suppose. Which of the four do we
credit to the student?
If we decided simply on the grounds of simplicity, then we would
attribute one of the false ones to the student. But very important in
the attribution decision are other facts that are known about the
context and the student. For example, it might be that the student has
just identified an "ing" word as a participle, in which case, given that
IDENTIFYING IMPLICIT ASSUMPTIONS 69

we credit the student with consistency, we would reject the first and
second candidates. Belief in them would then be inconsistent with the
data (Criterion (c)).
One way to probe further is to ask the student, "What does that
have to do with it?", which calls for the student to show a connection
between the answer and the thing to be explained. If in response the
student says 4, then that is ordinarily good evidence that the student
thinks 4 (because the latter (the thinking) explains the former (the
saying)), and we should probably attribute 4 to the student.
In this very sketchy example I have tried to indicate an application
of the criteria for identifying unstated reasons. Two pieces of
different data were explained by the hypothesis that the student
believed 4; it fits with other knowledge about the student; it was not
shown inconsistent with anything known about the situation; two
competitors (that the student believes 1 and that the student believes
2) were inconsistent with some evidence; and simplicity was a back-
ground concern.

Needed-Assumptions Criteria: Evaluation

In judging whether a particular proposition is a needed assumption,
one must consider the purpose of the identification and whether one
is looking for the assumption(s) of a conclusion or of an argument.
One might be looking for the assumption(s) of a particular argument
when that argument is a classic one (as in one of Aquinas' proofs for
the existence of God), or when the arguer would be most likely
convinced that there is an error by an appraisal of the particular
argument that he or she offered.
Sometimes, on the other hand, one is primarily interested in the
conclusion, as I was when listening to the prosecutor's summary
speech at the end of a murder trial for which I was a juror. Although
important, since the State was supposed to prove its case, his sum-
mary argument did not for me have very special status. I was
interested in the conclusion (that the defendant was guilty of murder)
and what needed to be assumed in reaching it. Naturally I paid great
heed to the prosecutor's explicit premises and explicit steps, but I was
primarily interested in deciding whether the defendant was guilty, and
felt some freedom to revise and improve the prosecutor's summary
argument, based upon what had actually been shown in the case.
70 ROBERT H. ENNIS

In the following discussion of criteria for needed assumptions, I

shall first present criteria that apply to the purposes most frequently
pursued, the evaluation purpose. Then I shall consider adaptations for
use in pursuit of the second purpose, seeking support.

Fidelity - In Varying Degrees. A first criterion for needed assumption

identification is fidelity to the argument, if the argument is the focus -
with some revision and improvement in the original argument, if the
conclusion is the focus. If both are the focus, then varying degrees of
revision and improvement are appropriate, depending on the situa-
tion. Schwartz (1980, p. 152) offers strict fidelity as a first criterion in
assumption identification, presumably in part because he was focus-
ing only on arguments, not conclusions. It might also be that
Schwartz was combining the identification of used assumptions with
the identification of needed assumptions. The identification of used
assumptions would require leaving the arguer's argument as is,
because it is a primary datum to be explained by the unstated-reason-
identification claim.

Contributing to Gap-filling or Backing-up. A second criterion for a

needed assumption is the requirement that the alleged assumption
somehow help out - either by contributing to filling a gap (or gaps) in
the argument - or by providing support for some part of the
argument. It should make a unique - or at least nonredundant -
contribution if it is to be needed. A gap is filled when one can infer
without any question a conclusion (perhaps an intermediate con-
clusion) from its support. In stating this description of gap-filling, I
deliberately did not specify that the inference must be deductively
valid because I do not want to exclude here the view that a complete
reconstructed argument for purposes of argument criticism may not
be deductively valid. This is not my view but I believe it to be a
workable view, if one is vigilant in noting the nondeductive steps in
an argument completed in accord with this view. I prefer a deductive
view of argument reconstruction, as I shall presently explain, because
it forces us to make explicit every nondeductive step and thus makes
it less likely that something will slip past us. (My view is not that
every argument is really deductive; rather it is that for purposes of
evaluation, deductive argument reconstruction is an effective tool.
More about this shortly.)
IDENTIFYING IMPLICIT ASSUMPTIONS 71

To pass the gap-filling test, a candidate for an assumption must be

one of a set of propositions all of which satisfy the contribution-to-
gap-filling requirement and which jointly fill the gap(s). Unless there is
only one proposition that fills a gap by itself, gap-filling is a coopera-
tive enterprise.
There is much to be said on the question of how to tell whether a
proposed assumption actually provides back-up support (as con-
trasted with filling a gap), but in order to keep this essay within fairly
manageable bounds, I shall remain silent on the topic.

Plausibility. Scriven (1976, p. 166), Schwartz (1980, p. 152), and I (1969,

p. 271) have advocated a criterion of plausibility in identifying assump-
tions (though in none of these advocacies is the criterion applied only to
identifying needed assumptions, as I now think it should be). The
plausibility criterion calls for the assumption identifier to be as generous
as possible. One does this (selecting from the possible gap fillers) by
attributing to the argument or conclusion maximally-plausible pro-
positions. The reason for this is that is not fair to say that a less plausible
proposition is needed by a position, when a more plausible one is
available to do the job of gap filling. The idea is to give the argument or
conclusion its best chance to succeed (while at the same time exposing it
to full scrutiny, so that nothing slips by unnoticed). This is because our
ultimate concern is the acceptability of the conclusion.
Since there is often more than one assumption to attribute to an
argument, and sometimes a choice between attributing one broad one
and two more-narrow ones, the criterion, more fully stated, calls for
attributing assumptions with a maximal joint plausibility. Roughly
speaking, we should maximize the minimum plausibility of the mem-
bers of the set of identified propositions that jointly will complete the
argument, or, for back-up assumptions, will provide maximal support
for a premise. This is still only a rough guide, assuming as I do that
one cannot quantify plausibility. Intelligent, informed judgment is
required.
Plausibility, so formulated, does not apply to identifying used
assumptions. Here instead of seeking the maximally-plausible filling-
in of a position, one seeks accuracy in reflecting what the person
actually thought, which is sometimes not the maximally-plausible set.
We might know for other reasons that the person does not believe the
maximally-plausible set. If so, that is sufficient reason not to attribute
72 ROBERT H. ENNIS

that maximally-plausible set to the person. The idea is to be accurate

and fair, not generous, nor ungenerous, in identifying used unstated
assumptions.
There is a way in which plausibility does apply in identifying used
assumptions. That is, we sometimes reject a candidate for a used
assumption on the ground that it is too implausible for the assumer to
have believed it. This is an application of the criterion that the
hypothesis should not be inconsistent with the other data. If the other
data tell us that the assumer is a reasonable and well informed
person, then the assumer's using and believing a wildly implausible
proposition might well be inconsistent with those data.

Being Needed. Does satisfaction of these three criteria tell us that a

proposition is needed? I think so.
Satisfication of the fidelity criterion assures that we are evaluating
what we want to evaluate. Given its satisfication, the question reduces
to whether satisfaction of the other two criteria shows that pro-
position is needed. First let us consider gap-filling propositions.
It seems clear that a proposition is needed, if, and only if, it is a
contributing nonredundant member of a set that is needed. So the
question becomes one concerning the relationship between the set's
being needed and its satisfying the gap-filling and plausibility criteria.
Since the only sets under consideration are those that do fill the gap
- that do complete the argument - the question becomes that of
whether the maximally-plausible set is needed. Remember that it is
the maximally-plausible, not maximally-ambitious set that is under
consideration. The maximally-plausible set might not be very am-
bitious at all.
Assuming for the sake of simplicity that there is only one maxi-
mally-plausible set, then each other is less plausible as a whole. There
is then a basis for saying that the maximally-plausible set is needed. It
would not be fair to the argument or conclusion to say that any other
set is needed, since there is a more plausible set available (the
maximally-plausible one) to do the job. If at least one set is needed,
then by elimination, the maximally-plausible one is the one.
If there are two sets that are more plausible than all the others, but
equally plausible, then, unless they have propositions in common, we
cannot identify the needed assumption(s). If the sets have one or more
propositions in common, then it or they are the needed assumptions.
IDENTIFYING IMPLICIT ASSUMPTIONS 73

In a back-up context the situation is somewhat similar. For its

back-up support a premise needs the set of propositions that provides
it with maximum back-up, back-up strength coming from a com-
bination of joint plausibility of the propositions and the strength of
inference they generate. Given two sets that provide equally strong
inference support, we pick as necessary the one that provides maxi-
mal joint plausibility of its members. The premise does not need less
plausible support when more plausible support is available.
In the general back-up case where one set might have strong
inferential supporting power but less joint plausibility of its individual
members, then the plausibility criterion calls for a combined maximal-
support judgment (maximum available soundness, so to speak), each
member of the winning set being judged a needed assumption. Any-
thing less than maximal support is not needed when greater support is
available. (This general case does not arise in the argument recon-
struction 1 shall presently describe, because all inferences are deduc-
tive (and thus equally strong) in a reconstructed argument.)
In sum, in gap-filling contexts, the most plausible set of pro-
positions that jointly fill the gap is needed, as is each contributing
nonredundant member. In back-up contexts, the set providing maxi-
mal support (which is derived from a combination of plausibility of
members and inferential supporting strength) is needed, as is each of
its contributing nonredundant members.

Needed-Assumptions Criteria: Support Seeking

Since, in engaging in support seeking, the truth status of the proposed
assumption is at issue, the plausibility criterion cannot be applied,
because its application requires that we make a judgment about the
plausibility of the proposed assumption. In order to avoid circularity,
an assumption identifier engaged in support seeking must, without
assuming anything about its plausibility, show that the proposed
assumption is needed. If we cannot make a plausibility judgment
about one crucial member of a set of propositions, we cannot make a
joint plausibility judgment about the set.
How else to establish need? It seems then that the weakest
requirement that would make the assumption identification stick, and
this is a very demanding requirement, is to show that gap-filling or
backing-up with any set that excluded the alleged assumption would
74 ROBERT H. ENNIS

be irrational. By this criterion Copleston did not succeed. He did not

succeed because a scientist might rationally proceed on the basis that
there is only order in the particular phenomena he or she is in-
vestigating. Even weaker: that there might or not be order in the
particular phenomena under investigation; that we are looking to see
whether there might be order there - or (even weaker) imposed by us,
perhaps for purposes of simplicity. Russell's reply, incidentally, was
in this direction.
In sum, support-seeking needed-assumption identification is more
difficult to do successfully than evaluation-oriented needed-assump-
tion identification. Gap-filling or backing-up are still required, but
since a plausibility judgment cannot be made, generosity to the
assumer requires only attributing rationality to the assumer - which
requirement is easy on the assumer and hard on the assumption
identifier.
After a look at argument reconstruction and at presuppositions I
shall argue that being logically necessary, although a tempting cri-
terion for being needed, is not a satisfactory one.

ARGUMENT RECONSTRUCTION

Argument reconstruction is an extension of assumption identification

to the filling of all the gaps in an argument and rebuilding it. The
strategy that I shall describe consists of the deductive reconstruction
and elaboration of an argument in such a way that every part that is
not a deductively valid inference is an explicit premise. To be fair, the
reconstructed result should be as strong an argument as the recon-
structor can produce within the limits set by the context. The idea is
to expose those needed nondeductive claims (including a claim that a
particular step is legitimate) about which one can, without challenging
a deductively-valid step, ask "Why so?", thus making explicit those
points that might possibly require further defense.
Mill (quoted above) could be interpreted as being engaged in this
sort of argument reconstruction when he suggested a proposition that
provided deductive validity, "What is true of John, Peter, & c. is true
of all mankind." If so, his identification appears defective because
(going by the words he used to express it) the proposition he
produced is quite implausible. We do not know who is represented by
the "&c.", but whoever it is, there are no doubt things true of the
IDENTIFYING IMPLICIT ASSUMPTIONS 75

examined set of people that are not true of all mankind. One such
thing presumably is that all the examined people are conceived
already, a characteristic not possessed by all mankind. It would not
be fair to the inductive mortality argument to burden it with such a
far-fetched claim when a more plausible one would fill the gap.
Out of context as we are in viewing the argument Mill considered,
we cannot with confidence apply the maximum plausibility require-
ment. But I can note a minimal coverage maximally-plausible claim
that would serve as a piece in a reconstructed version of the Mill
argument: "That John, Peter, &c., are mortal is sufficient ground for
concluding that all mankind are mortal." That such an identification is
in a way trivial, as Scriven (1976, p. 163) would no doubt urge, I
readily grant. In argument reconstruction we are not required to make
only significant discoveries. We instead are engaged in discovering
those points about which it makes sense to ask the question, "Why
so?" And anyone who offered the unreconstructed version of the Mill
argument could sensibly be asked the "Why so?" question of at least
the suggested claim. Since we are not operating in a real context, I
cannot suggest whether the request should be to consider only the
specified claim or a more general one. But the identification of the
claim, though trivial in a way, can be a useful move in argument and
conclusion evaluation.
Thus, if one regards Mill as engaged in argument reconstruction,
then it appears that he is too demanding of the hypothetical arguer.
One who so argued need only be asked to defend a much lesser claim.
If on the other hand one regards Mill as here saying in effect that all
arguments are really deductive arguments ("deductive chauvinism"2),
then we would be engaged in a discussion of a different sort, one that
would go beyond the scope of the essay.
The deductive inference to which I refer in characterizing this
argument reconstruction function of needed-assumption identification
is not limited to that worked out in formal systems by logicians. I
have rather in mind a pretheoretic notion of deductive inference: a
deductively-valid inference is one such that the conclusion follows
necessarily from the premises. Since there is much that is not worked
out in this area, one will often need to rest one's decision on informed
intuition. In accord with mine, the step from the explicit premise in
Mill's argument together with the narrow claim I suggested containing
the words, "sufficient ground", is a deductively valid step.
76 ROBERT H. ENNIS

What I have just offered under the label, "argument reconstruc-

tion", is a suggested move in argument criticism, not a claim about
the true nature of arguments. It is not the claim that all arguments are
really deductive. It is not deductive chauvinism. 3

PRESUPPOSITIONS

Although the actual identification of a premise-type assumption,

whether it be an unstated reason or a needed assumption, is often
difficult and obviously open to controversy, the concept of a premise-
type assumption seems clear enough and I believe noncontroversial,
once the distinctions I suggested are made. The concept of a presup-
position on the other hand is controversial, but I think the notion is
still a useful one for the fields of argument criticism and rhetoric.
I here rely upon the notion of presupposition offered by Strawson
(1950, 1952) with changes based primarily upon Donnellan's criticisms
(1966, 1968), and with extensions to include more than the presup-
positions of definite descriptions in statements. Although there are
controversies concerning presupposition, I believe that I avoid (at
least) most of them for present purposes. So far as I can see, it does
not matter which way you decide on the controversies for the things
that seem to be most important in assumption identification.
Roughly speaking, a presupposition of statement X is a proposition
that, if false, would in a common (though imprecise) way of saying
things, make X neither true nor false. For example, a presupposition
of "Jones has stopped beating her husband" is the proposition,
"Jones at one time beat her husband." If Jones never beat her
husband, then, in a common way of saying things, it would be neither
true nor false that Jones has stopped beating her husband. That is, if
Jones has never beat her husband, then we cannot say that she has
stopped beating him. Nor can we say that she has not stopped beating
him. The question, "Has Jones stopped beating her husband?", does not
arise. It is a misleading and inappropriate question.
In discussions, debates, contentious arguments, etc., it is well to be
on guard for presuppositions, because it is so easy to accept them
without thinking about them. For example, I find myself less resistant
to believing the proposition that there is a missile gap when I am told,
"The missile gap will take five years to eliminate" than when I am
IDENTIFYING IMPLICIT ASSUMPTIONS 77

told, "There is a missile gap and it will take five years to eliminate"
It is for this reason that 1 introduce the concept of presupposition in the
face of some possible difficulties in attempts to give a precise for-
mulation.
I realize that in presenting the husband-beating example I moved a
negation from an internal to an external position and that, given that
Jones never beat her husband, we do have at least some inclination to
say that it would be false that Jones has stopped beating her husband.
That is (and this is the imprecise way of doing things), I have noted
that, given that Jones never beat her husband, it would at least on first
impression be incorrect to say that Jones has not stopped beating her
husband. From this I moved to the claim that it would therefore be
incorrect to say, "It is false that Jones has stopped beating her
husband." Whether this statement in quotes would in fact be incor-
rect is arguable.
However, I shall not attempt to argue the point either way here.
Whichever way the argument turns out, there is still at least a
practical failure to have expressed a true or false proposition, given
that Jones never beat her husband. The failure to have expressed a
true proposition is obvious. The practical failure to have expressed a
false proposition depends upon the widespread negation-moving (and
I think imprecise) habits of most people. Rarely - except among
philosophers and perhaps lawyers - do people use the locution, "It is
false t h a t . . . " Rather they generally move the negation inside to
produce, in this case, "Jones has not stopped beating her husband",
as a denial of "Jones has stopped beating her husband". And it is this
denial that appears not to follow from, and is apparently incompatible
with the proposition that Jones never beat her husband.
Furthermore, given that Jones never beat her husband, the ques-
tion, "Has Jones stopped beating her husband?", does not have a
"Yes" or " N o " answer. Either answer presupposes that she once did.
This observation seems to me to be not only practically acceptable
but literally and uncontroversially correct. It thus gives added support
for the distinction between premise-type assumptions and presup-
positions. Extending the notion of presupposition to questions, if the
presupposition of the question is false, then the question, if it is a
"Yes"-or-"No" type question, does not have "Yes" or " N o " as an
answer.
As with statements, people sometimes find themselves lulled into
78 ROBERT H. ENNIS

accepting the presupposition of a question, when they would not as

easily accept the same proposition in the form of a statement. For
example, an ignorant person (as all of us often are) is less likely to
ask "Why do you think that Jones has been beating her husband?"
after hearing the question "Has Jones stopped beating her husband?"
than after hearing the sequence, "Jones has been beating her hus-
band. Is she beating him these days?"
Particularly notable for having a presupposition is a "Why?" ques-
tion containing a proposition. For example, "Why does Jones beat her
husband?" would in many contexts presuppose that Jones does beat
her husband (though one might ask that question tongue in cheek
when told that Jones beats her husband).
I realize that my claims about the importance of presuppositions
for informal logic depend upon some pyschological claims about the
susceptibility of people. These claims are based on my experience in
dealing with and observing people in a variety of situations over a
number of years. There is no empirical research on the topic, though I
hope there will be some sometime. If I am wrong in my beliefs about
susceptibility, then the separate special importance of presuppositions
diminishes, but certainly does not disappear. Presuppositions are still
assumptions in the broad sense of the term "assumption": "taken for
granted as a basis of argument or action."
In sum then (and still roughly speaking), a presupposition of
statement X is a proposition that, if false, would make X not true and
(at least practically - or "carelessly" if you are so disposed) would
make X not false. Extended to "Yes"-or-"No" questions, a presup-
position of question Y is, roughly speaking, a proposition the falsity
of which would deprive a question of a "Yes" or " N o " answer. And a
presupposition of the question, "Why X ? " , is X (with subject-verb
order reversed).
But there are problems, noticed at least by Donnellan (1966).
Consider the standard example, "The King of France is wise". Under
the presupposition doctrine, as I have so far presented it, this state-
ment would be neither true nor false, on the ground that its presup-
position, "There is a King of France", is false. However, Donnellan
noted that a speaker might say, "The King of France is wise",
intending to refer to De Gaulle, and thus have said something about
him that might be true - or false. What was said about De Gaulle was
true if De Gaulle was wise, and false if he was not. The use of the
IDENTIFYING IMPLICIT ASSUMPTIONS 79

definite description, "the King of France", here is what Donnellan

called the referential use. The speaker used the description to refer to
a person (De Gaulle), perhaps because the speaker was being cynical
- or perhaps because the speaker was ignorant - but the speaker did
refer to De Gaulle and succeeded in saying something about him,
even though there was no King of France at the time. So the
presupposition doctrine must be applied especially cautiously when
definite descriptions are used referentially. The presupposition of a
referential definite description is not necessarily that there exists an
object that uniquely fits the description. Rather it is that there exists
an object to which the speaker thinks that he or she is referring.
The other use of definite descriptions mentioned by Donnellan is
the attributive use. In this use of a definite description, the speaker is
not referring to a person in particular, but rather to whoever fits the
description. When I say, believing that someone piled the rocks but
not knowing who, if anyone, did pile the rocks, "The person who
made the rock pile was aesthetically sensitive" then I am presuppos-
ing that someone did make the rock pile and am asserting that this
person (whoever it may be) was aesthetically sensitive. With this, the
attributive use of definite descriptions, a presupposition is that there
exists an object fitting the description. That is, if no one piled the
rocks (say, the pile was the result of an avalanche) then my statement
is neither true nor false in the practical way I indicated earlier.
Saul Kripke (1977) and John Searle (1979), among others, have
challenged this referential-attributive distinction, although both
Kripke and Searle agree that there is something different going on in
Donnellan's two kinds of use. Not only are their theories considerably
more complex, but I find that I need his distinction to explain their
explanations. So, although I shall not argue the point here, I shall
continue to use his language and distinction. In the present context,
where the concern is ultimately with argument analysis, they possibly
might not object to my using Donnellan's simpler intuitive notion
anyway, since they do agree that Donnellan has identified a significant
difference.
I am not here proposing that the distinction be taught to informal
logic students, nor am I proposing that it not be taught to them. That
would depend upon how sophisticated they are. But I do think that
informal logic teachers should have it in mind. Otherwise problems in
the classroom will arise that they will not be able to handle.
80 ROBERT H. ENNIS

To summarize then about presuppositions with a depiction that is at

least roughly accurate: a presupposition of a "Yes" or "No" question
is a proposition that, if false, would make both "Yes" and "No"
answers to a question inappropriate. A presupposition of a "Why
X ? " question is X (with subject-verb order reversed). A presup-
position of a statement is a proposition that if false, would make the
statement neither true nor false - in the practical way I indicated.
Because definite descriptions have at least both referring and attribu-
tive uses, the presupposed proposition when a definite description is
involved is not necessarily the one that asserts the existence of
something actually fitting the description.

CRITERIA FOR PRESUPPOSITION IDENTIFICATION

Since the concepts of referential and attributive presuppositions are

quite different, we can expect criteria for identification to differ.
Attributive presuppositions are intimately connected with word
meanings, and referential presuppositions with mental acts including
beliefs. But they share at least this similarity: both types have a
used-assumption aspect and a needed-assumption aspect. The needed-
assumption aspect is as follows: a presupposer is prima facie committed
to accepting some existence claim. The used-assumption aspect of each
I shall examine separately.

Referential Presupposition Identification

Referential-presupposition-identification claims are used-assumption
identification claims; they are claims about an operative belief of a
person, the belief of the form, "There is or was something to which I
refer [though it may not be picked out by my words] with my
expression and about which I am saying something." That the person
has such a belief and used it as a basis for reference is an empirical
claim about a matter of fact. If claims about mental events are to be
treated as singular explanatory hypotheses, then referential-presup-
position-identification claims should also be so treated. I earlier
suggested some rules of thumb for evaluating empirical explanatory
hypotheses.
In the De Gaulle example the explanatory hypothesis is that the
speaker has someone in mind, De Gaulle, and it is about him that the
IDENTIFYING IMPLICIT ASSUMPTIONS 81

speaker is saying that he is wise when he said, "The King of France is

wise." Another way of expressing this explanatory hypothesis is by
saying that the speaker is presupposing that De Gaulle exists and
referring to him in saying, "The King of France is wise."

Attributive Presuppositions

A part of an attributive presupposition also plays the role of an

assumption that was actually used and is part of the belief structure
of the presupposer. The total presupposition is that there is a unique
thing that fits the description given by the speaker interpreted in
accord with the meanings employed by the speaker (which are not
necessarily standard meanings). For example a speaker might say,
"The will of the group is unknown to us." An attributive (used)
presupposition of this statement might well indicate what the speaker
had in mind by the phrase, "will of the group." (Alternative pos-
sibilities might include (1) majority view, and (2) the view that the
group would hold if fully informed and fully rational.) This would be
attributive if the speaker was not referring to any particular group
opinion, but rather meant to be saying, "The will of the group
(whatever it may be) . . . " The needed-assumption part of the presup-
position is that there is something that fits the speaker's notion, will of
the group. The used-assumption part is the part that says what the
speaker meant by the words, "will of the group."

Two Clarifications

The Choice Between Referential and Attributive. The statement that a

speaker was using a definite description referentially - or attributively
- is itself an implicit-used-assumption claim. It is an empirical ques-
tion which way a speaker meant to be using a particular description.

The Prima Facie Existential Commitment. Another empirical ques-

tion is whether the person who is prima facie using a presupposition
actually is doing so. For example, a speaker might say, "The budget
deficit is held by many to be illusory." It could then be an empirical
question whether the person is committing himself or herself to the
existence of a budget deficit.
82 ROBERT H. ENNIS

Summary
Presupposition identification combines elements of needed-assump-
tion identification and of used-assumption identification. The used-
assumption part of the presupposition claim is to be judged as an
explanatory hypothesis, whether it be referential or attributive. The
needed-assumption part is to be judged in accord with whether the
falsity of the alleged presupposition would make the original state-
ment neither true nor false - in the practical way I have noted.

LOGICAL NECESSITY

Logical necessity is not the appropriate criterion for identification of

a needed assumption. That is, when the result is informative, (with
one possible type of exception for use in argument reconstruction) we
do not identify something that is logically necessary for the argument
or conclusion. I have argued this point elsewhere (Ennis, 1962), and
will summarize here (with adjustments made to accommodate the
referential-attributive distinction and my more-recently developed
approach to argument reconstruction).
The basic idea is that any incomplete argument can be completed in
a number of different ways. Consider the simple argument, "Since
Mike is a dog, Mike is an animal." To suggest the range of possible
gap-fillers, I shall note two, the one that first occurs to most people
and another one:
1. All dogs are animals.
2. All dogs whose name begins with "M" are animals.
A person is not logically prohibited from offering the argument,
denying 1, and using 2 as a gap-filler. This shows that 1 is not logically
necessary for the offering of the argument. So long as the offered
gap-filler is more general than a statement establishing connection
between the offered premise and conclusion, one can deny the offered
gap-filler and derive the conclusion from the offered premise by using
a narrower gap-filler. The limiting case is the simple statement that
the offered premise was assumed in the circumstances to be sufficient
to establish the conclusion. In this case, the limiting assumption
candidate would be, "The premise, 'Mike is a dog', is sufficient in
these circumstances to establish the conclusion that Mike is an
animal." Now even though that assumption candidate may be logic-
IDENTIFYING IMPLICIT ASSUMPTIONS 83

ally necessary for the argument, identifying it is not informative,

except for argument-reconstruction purposes. Schematically in the
non-pejorative sense of "assumes", we would generally not be
informed by the following assumption-identification move: Your
argument, 'Since p, q' assumes that in the circumstances your pre-
mise 'p', is sufficient to establish the conclusion 'q'." But in argument
reconstruction contexts, it might help to be reminded that an arguer is
committed to at least this minimal claim and can be asked by us, "Why
so?" without our making an error of (deductive) logical blindness.
My point about logical necessity can be applied to the parenthetical
remark I quoted earlier from Mill: "what is at any rate a necessary
condition of the validity of the argument." One might deny the
premise that Mill alleged to be a necessary condition of the validity of
the argument and achieve a valid argument with a set of premises that
include this narrower one: "Whatever is true of John, Peter, &c.,
except their having already been conceived, is true of all mankind."
I am not claiming that the substitute gap-filling premise is true. I am
only claiming that we can construct a valid argument without using,
and in fact denying, the supposed necessary condition for validity of
the argument. This shows it not to have been a logically necessary
condition.
If Mill did not mean to be supplying a logically-necessary condition,
then his statement is not a violation of my point about logically
necessary assumptions. However, it is still an instructive case
because it illustrates a strategy for dealing with allegations that a
proposition is a logically-necessary premise-type assumption.
Presuppositions also are not logically necessary, given just the
language of an argument or a statement. In the case of definite
descriptions this is in part because it is the speaker's intentions that
determine whether an expression is being used attributively or
referentially. Furthermore in all presupposition cases the speaker
means what he or she means, and we cannot say that the speaker's
statement necessarily presupposes something, when the speaker did
not mean the words in the sense we have in mind. We could of course
say that if the speaker is using the words in their standard sense, then
the statement necessarily presupposes something, if the words have a
standard sense.
For example we cannot say, of the statement, "The King of France
is wise", that it necessarily presupposes that there is a king of France
84 ROBERT H. ENNIS

(in our sense of "King of France"), for the speaker, as Donnellan

pointed out, might have been referring to De Gaulle. Presuppositions
are not logically necessary assumptions, given just the language of the
argument or statement. They assume logical necessity only as the
speaker's intentions permit.
What seems to be a logically-necessary assumption is the pro-
position that something exists or existed (possibly an event or state of
affairs) that either satisfied the speaker's description in the sense of
the words that the speaker had in mind, or to which the speaker
thought he or she was referring in an act of speaker reference. (I lean
on Donnellan's (1979) discussion of speaker reference.) But even this
is only prima facie logically necessary, as is shown by the budget-
deficit example.

SUMMARY

After distinguishing conclusion, pejorative, and adoption uses of the

word "assumption", I examined the nature of implicit assumption
identification, interpreted as the identification of implicit propositions
taken for granted as a basis of argument or action. Under this broad
notion I distinguished gap-filling from backing-up assumptions, used
from needed assumptions, and premises from presuppositions; and
suggested two purposes one might be pursuing in the identification of
needed assumptions - evaluating a conclusion or argument, and
seeking support.
On the assumption that used-assumption identification claims are
explanatory hypotheses, I adapted and offered as criteria some rules
of thumb for judging such claims. I also suggested criteria for iden-
tifying needed assumptions: (1) fidelity, more or less, depending on
the importance in its own right of the argument as presented; (2)
nonredundant contribution to the gap-filling ability of a set of pro-
positions; and (3) maximal plausibility of the set. Brief names for
these criteria are fidelity, gap-filling ability, and plausibility. When the
purpose of needed-assumption identification is the providing of sup-
port, the plausibility criterion must, in order to avoid circularity, be
altered to omit attention to plausibility of the allegedly needed
assumption. Instead, only rationality may be demanded, making it
much more difficult to show that an assumption is needed.
Argument criticism, I believe, can be facilitated by an argument
IDENTIFYING IMPLICIT ASSUMPTIONS 85

reconstruction process that requires each part of the reconstructed

argument to be either a deductively valid step or an explicitly stated
proposition (including such propositions as those that claim that the
inductive evidence is sufficient). This approach to argument criticism
does not hold that all arguments are really deductive. Rather it holds
that a useful step in argument criticism is the reconstruCtion of the
argument into a form that has as inference steps only deductively
valid ones. The other parts are then made explicit and are exposed to
the question "Why so?".
Employing Donnellan's distinction between referential and attribu-
tive uses of definite descriptions, I extended the discussion to
presuppositions, defined roughly as propositions the falsity of which
made the proposition of which they are presuppositions neither true
nor false, or which make a question without point in a certain way. I
counseled that we need to be alert to the possibility that propositions
that need to be challenged or explored might slip by unexamined
when playing the role of presuppositions.
Both referential and attributive presupposition have needed-
assumption aspects and used-assumption aspects. The primary
needed-assumption is the proposition that something (which might be
an occurrence or state of affairs) exists or existed that either fits the
description given by the speaker or that was picked out by the
speaker in an act of speaker reference.
In the interest of brevity, my treatment of this topic has been
heavily schematic. But I think that I have said enough to make my
points clear to an experienced teacher of informal logic, who I believe
will find the distinctions, claims and criteria useful. Much however
remains to be done before these ideas can be presented to students,
including simplification and the development of a range of illustrative
examples. I hope that what I have done constitutes a fruitful begin-
ning in the organization of this part of the subject matter of informal
logic.

University of Illinois

NOTES

I am indebted to David Bantz, Max Black, Nicholas Burbules, Beatrice Nelson,

Steven Norris, Allan Pearson, Emily Robertson, Fredrick Schmitt, Laura Spera, Frances
Wagner and Robert Wengert for suggestions.
86 R O B E R T H. E N N I S

2 Merrilee Salmon suggested this term, "deductive chauvinism", in a lively conference

discussion of the question whether this third function is a form of deductive chau-
vinism. The discussion took Place at the Carnegie-Mellon conference, 'Logic and
Liberal Learning', June 11-13, 1979, chaired by Preston Covey.
3 j. Anthony Blair has in a letter appropriately characterized the process as "heuristic".

REFERENCES
Beardsley, Monroe: 1975, Thinking straight, Prentice-Hall, Englewood Cliffs, N.J.
Donnellan, Keith, S.: 1966, 'Reference and definite descriptions', The Philosophical
Review 75, 281-304.
Donnellan, Keith, S.: 1968, Putting Humpty Dumpty together again', The Philosophical
Review 77, 203-215.
Donnellan, Keith, S.: 1979, 'Speaker reference, descriptions, and anaphora', in Peter A.
French, Theodore E. Uehling, and Howard K. Wettstein (eds.), Contemporary
Perspectives in the Philosophy of Language, University of Minnesota Press, Min-
neapolis, 28-44.
Ennis, Robert H.: 1962, 'Assumption finding', in B. Othanel Smith and Robert H. Ennis
(eds.), Language and Concepts in Education, Rand McNally, Chicago, 161-178.
Ennis, Robert H.: 1969, Logic in Teaching, Prentice-Hall, Englewood Cliffs, N.J.
Ennis, Robert H.: 1980, 'A conception of rational thinking', in Jerrold Coombs (ed.),
Philosophy of Education 1979, Philosophy of Education Society, Normal, Ill.
Harman, Gilbert: 1973, Thought, Princeton University Press, Princeton, N.J.
Kripke, Saul: 1979, 'Speaker's reference and semantic reference', in Peter A. French,
Theodore E. Uehling, and Howard K. Wittstein, Contemporary Perspectives in the
Philosophy of Language, University of Minnesota Press, Minneapolis, 6-27.
Mill, John Stuart: 1879, A System of Logic, Longmans, Green, and Co., London.
Russell, Bertrand and F. C. Copleston, S. J.: 1948, A discussion of the existence of God
(first broadcast on the Third Programme of the British Broadcasting Corporation,
January 28, 1948).
Schwartz, Thomas: 1980, The Art of Logical Reasoning, Random House, New York.
Scriven, Michael: 1976, Reasoning, McGraw-Hill, New York.
Searle, John R.: 1979, Expression and Meaning, Cambridge University Press, Cam-
bridge.
Strawson, Peter F.: 1950, 'On referring', Mind 59, 320--344.
Strawson, Peter F.: 1952, Introduction to Logical Theory, Methuen, London.
Thomas, Stephen N.: 1977, Practical Reasoning in Natural Language, Prentice-Hall,
Englewood Cliffs, N. J.