The Fallacy of 'Ad Ignorantiam'

Author(s): John WOODS and Douglas WALTON

Source: Dialectica , 1978, Vol. 32, No. 2 (1978), pp. 87-99
Published by: Wiley

Stable URL: https://www.jstor.org/stable/42968398

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Wiley is collaborating with JSTOR to digitize, preserve and extend access to Dialectica

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 The Fallacy of 'Ad Ignorantiam'
by John WOODS * and Douglas WALTON **

Summary
This paper outlines a three-part analysis of the traditional informal fallacy of
ad ignorantiam. As initially characterized, the fallacy consists in arguing that failure
to prove falsity (truth) implies the truth (falsity) of a proposition.
First, the fallacy is located within confirmation theory as a confusion between
the categories of "lack of confirming evidence" and "presence of disconfirming
evidence". Second, the structure of the fallacy can be seen as an illicit negation shift
in Hintikka-style epistemic logic. Third, the fallacy can be studied as an attempt to
unfairly shift the burden of proof in a dialectical game. We suggest that research on
ad ignorantiam needs a broadening of the scope of philosophical logic to encompass
concepts of correct argument in these three contexts.

Résumé
Cet article propose une analyse en trois parties du sophisme informel traditionnel
ad ignorantiam. Sous sa forme originale, celui-ci consiste à argumenter que l'absence
de preuve de la fausseté (vérité) implique la vérité (fausseté) d'une proposition.
1. Dans le cadre de la théorie de la confirmation, le sophisme apparaît comme
une confusion entre les catégories « absence de confirmation expérimentale » et « pré-
sence d'infirmation expérimentale ».
2. Dans le cadre d'une logique épistémique style Hintikka, la structure du sophisme
peut être considérée comme un passage illicite à la négation.
3. Dans un jeu dialectique, le sophisme peut être présenté comme une tentative
déloyale pour rejeter sur l'autre la charge de la preuve.
Nous pensons que cette recherche sur le sophisme ad ignorantiam exige un élar-
gissement du domaine de la logique philosophique de manière à ce qu'elle puisse assu-
rer une argumentation correcte dans ces trois contextes.

Zusammenfassung
Es wird eine dreiteilige Analyse des traditionellen Fehlschlusses ad ignorantiam
vorgelegt. Wie eingangs gezeigt wird, besteht der Fehlschluss darin, dass vom Miss-
lingen, die Falschheit eines Satzes zu beweisen, auf seine Wahrheit geschlossen wird.
Erstens wird der Fehlschluss im Rahmen der Bewährungstehorie als Folge einer
Verwechslung zwischen den beiden Kategorien des «Mangels an Bewährungsinstanzen»

* Department of Philosophy
Victoria University of Wellington, P. O. B. 196, Wellington/New Zealand
** University of Winnipeg
515 Portage Ave, Manitoba/Canada R3b 2E9

Dialéctica Vol. 32, No 2 (1978)

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
88 John Woods and Douglas Walton

und des «Vorhandenseins von Bewährungsinstanzen» dargestellt. Zweitens kann die

Struktur des Fehlschlusses als eine unzulässige Negationsverschiebung im Sinne von
Hintikkas epistemischer Logik betrachtet werden. Drittens kann der Fehlschluss als
Versuch gedeutet werden, die Beweislast in einem dialektischen Spiel in unfairer Weise
auf den Gegner abzuschieben.
Es wird nahegelegt, dass die Erforschung dieses Problems eine Erweiterung des
Bereichs der philosophischen Logik erfordert, so dass diese auch die Begriffe des
korrekten Argumentierens in den drei genannten Kontexten enthält.

John Locke wrote in 1690 that men commit the fallacy of ad ignorantiam
when requiring an adversary in debate to admit what they allege as a proof,
or assign a better K If, however, the proof is a good one, it is not easy to
see how there is anything fallacious in what Locke describes. Irving Copi
wrote in 1972 that men commit the fallacy of ad ignorantiam when they
argue that a proposition is true simply on the basis that it has not been
proved false 2. Obviously, however, sometimes failure to prove falsity does
constitute some legitimate evidence for the truth of a proposition. The ease
with which we can overturn both accounts of the fallacy suggests that
ignorance or arguing from ignorance is a relic of our heritage that has been
preserved intact. Still more disheartening, even these two clearest and most
helpful articulations of ad ignorantiam appear to lack any real agreement
or unity of direction.
Incoherence and lack of direction is the style we are accustomed to in
the long and dismal history of the study of informal fallacies. Modern logic
texts continue the tradition, sequestering the study of fallacies from the
growing, better defined branches of logic. While the crucial pedagogical
importance of the fallacies is acknowledged by their continued appearance
on university curricula, teachers do not feel that they can commend the study
of fallacies as a decisive strategy in the adjudication of argument. The
bereftness of theoretical resources, poorly compensated for by the prolifera-
tion of hackneyed examples, does not encourage a systematic treatment of
the basic indeterminacies which students quickly perceive in The Standard
Treatment'3. We would here like to address ourselves to a long-overdue
initial attempt to locate a framework for the further analysis of the ad
ignorantiam . We do not think we have the complete analysis of this fallacy,

1 John Locke, An Essay Concerning Human Understanding , ed. John Yolton,

London, Dent, 1961, p. 278.
2 Irving Copi, Introduction to Logic, New York, MacMillan, 1972, p. 76.
3 C. L. Hamblin, Fallacies, London, Methuen, 1970, ch. I.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 The Fallacy of 'Ad Ignorantiam' 89

for, as we hope to show, this task partly awaits upon d

largely unresolved, areas of related study.

1. Find the Fallacy

It is part of the lore of experimental science that the

evidence for a hypothesis is not to be confused with di
and stern warnings against this confusion are found in
scientific method. Conversely, so we are toldt, the wa
evidence does not confirm a hypothesis either. Argume
of one or the other of the following pair are said, ther
fallacy of ad ignorantiam.
(Aj) There is no disconfirming evidence for H.
Therefore, H is confirmed.
(A2) There is no confirming evidence for H.
Therefore, H is disconfirmed.
No doubt there is truth in these admonitions. The fact is however that in
some contexts, arguments having the forms (Ax) and (A2) are not fallacious.
Consider the case where there are high antecedent probabilities in favour of
H; then (AJ may not be fallacious. Similarly, low antecedent probabilities
might vindicate (A2). Or if it seems to beg the question to bring in antecedent
probabilities, consider the case where Pr(H) = . 5 (exactly). Even here (Aļ)
and (A2) need not be fallacious. We might imagine a case where a series of
ingenious experiments has been run, where any negative result would have
occasioned the downfall of H , yet in which no negative results were in fact
reported. Ceteris paribus , we might well suppose H to have been, to a certain
degree, confirmed; and (AJ would be regarded as inductively correct.
Similarly, if deliberate, well-formulated attempts to confirm H were repeat-
edly to give rise to nil results, all else being equal, our confidence in H would
quite reasonably be diminished, provided the antecedent epistemic proba-
bilities are regarded as fixed. Either way, whether a priori probabilities are
counted or not, (Aj) and (A2) do not seem to deserve their reputation as
fallacies. Can it be that ad ignorantiam is not a fallacy? Or is it rather that
the schemata (Aļ) and (A2) do not manage to specify it?

2. Confirmation: Dis and Un

The fallacy might more perspicuously be located if we reminded ourselves

that any testable hypothesis has a class of statements (deductively or

t Though not by Professor Popper.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
90 John Woods and Douglas Walton

inductively) deducible from it (plus some auxiliary hy

call the test class4. For a hypothesis to be of any leg
interest, the test class must exclude certain stateme
are neutral for the purpose of confirmation or discon
include certain other statements, although we may n
statements it includes are true, or are likely to be tru
false. Thus for any hypothesis we require the general
to be divisible into two disjoint, non-empty classes,
neutral class.

For any hypothesis, H, any statement my be said to be either //-neutral

or //-testable, but not both. If we were to imagine the class of all statements
as a line, a hypothesis H would cut off a segment of that line representing
the //-testable statements.

/ 1
//-neutral I //-tes

Let us also agree that hyp

ing degrees, and that some
ed: call these unconfirmed
hypotheses.

unconfirmed disconfirmed confirmed

That there is some such tripartite partitioning essential to confirma

is a relatively uncontroversial assumption. How the distinc
implemented in actual experimental practise is of course a very
substantive problem of confirmation theory, and we will not try
here.

Three further minimal assumptions are needed that seem esse

reasonably standard notion of confirmation. First, we assum
//-testable statement is confirmed, then H is either confirme

4 For some background to the notions developed in this section, v

Carnap, Logical Foundations of Probability, University of Chicago Press, 1962,
Ch. II, also 'Preface to the Second Edition.' Discussion of some applications of
inductive logic to the domain of informal fallacies is included in John Woods and
Douglas Walton, Towards a Theory of Argument,' Metaphilosophy, forthcoming.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 The Fallacy of 'Ad Ignorantiam' 91

firmed. Second, we assume that if an ¿/-testable state

then H is either confirmed or disconfirmed. That is, w
//-testable confirmation always leads to //-confirmation
firmation to //-disconfirmation. Nor will we indicate the manner in which
the degree of confirmation of //-testable statements statistically escalates or
lowers the degree of confirmation of hypotheses. But third, we do assume
that if an //-testable statement is unconfirmed, then H is also unconfirmed,
prior probabilities aside. Thus our three assumptions are these: unconfirma-
tion is carried over from //-testable statements to H, and the union of discon-
firmation and confirmation is likewise carried over.

3. A Fallacy of Confirmation
We see from the assumptions of the previous section that the following
equivalences obtain.
(Ej) H is confirmed ^ ^ ~ (// is unconfirmed v H is disconfirmed)
(E2) H is disconfirmed < - ^ ~ (// is confirmed v H is unconfirmed)
(Eo) H is unconfirmed ^ ^ ~ (// is confirmed v H is disconfirmed)

We have it, therefore, that the pair of argument forms below are valid.

(Vļ) H is not disconfirmed

Therefore, H is confirmed v H is unconfirmed
(V2) H is not confirmed
Therefore, H is disconfirmed v H is unconfirmed

But the following pair are incorrect.

(Fj) H is not disconfirmed (F2) H is not confirmed

Therefore, H is confirmed Therefore, H is disconfirmed

The latter pair are not valid forms of argument, since the conclusion fails
to exhaust all (in this case both) possible alternatives. Comparison of (Vļ)
with (FJ and (V2) with (F2) shows that the fallacy in both cases consists
in suppression of the possibility that H may be unconfirmed, i. e., the live
possibility that there are no known data for H is omitted. And in so saying,
we have the suggestion ignorantiam has an epistemic aspect.

4. An Epistemic Fallacy

Typically the " proved " in 'Nobody has ever proved that 0 is false
(true), therefore 0 is true (false)' carries an epistemic load; the ad ignoran-
tiam therefore can be significantly viewed as an epistemic fallacy. The

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
92 John Woods and Douglas Walton

argument here may exhibit one of the followin

structures 5.

(F3) ~(3x)Kxp (F4) -(Hx)Kx ~p

Therefore, <vp Therefore, p
Simply because nobody knows that p is true, it does not f
Simply because nobody knows that p is false, it does not
Thus (F3) and (F4) are fallacies. Moreover, it is easy
(F4) are invalid argument forms in Hintikka's system
system, of epistemic logic. For in such systems, for any
rKa~p z> ~p~l are theorems, but r ~Kap 3 -p"1 and r ~Ka~p z) p"1
are not theorems. The fallacy can thus be exhibited as confusion between
the pair,
(F5) -Kap (V5) Ka-p
Therefore, /vp Therefore, /-v/p
or between the pair.
(F6) - Ka-p (V6) Ka~~p
Therefore p Therefore p
The forms on the left are incorrect and the one
standard systems. The resemblance within each pa
scope of the negation operator is made explici
effected by an unlicenced shift of the negation
operator, a manœuvre not difficult to obscure in ord

5. A Dialectical Fallacy
An important class of fallacies involve constru
the onus of proof 6. In general, if a person assert
of argument, he may reasonably be obliged t

5 For background to the notions developed in this se

Knowledge and Belief , Ithaca, Cornell University Press
Wu, 'A New Approach to Formalization of a Logic of Knowledge and Belief,'
Logique et Analyse, 63-64, 1973, 513-525. See also the review of the latter by Douglas
Walton, Mathematical Reviews , MR 12670, 50, 1975. For ohter applications of epistemic
logics to the informal fallacies, see John Woods and Douglas Walton, ť Petitio
Principii,' Synthese, 31, 1975, 107-127, and Douglas Walton, 'Mill and DeMorgan on
whether the Syllogism is a Petitio / Historia Mathematica , 2, 1975, 336-337.
6 Richard Whately, Elements of Rhetoric, ed. Douglas Ehnnger, Carbondale,
Southern Illinois University Press, 1963, Part I, Ch. Ill, §2. This is a reprint of the
seventh edition of 1828. See also the comments in Hamblin, op. cit., p. 170 ff.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 The Fallacy of 'Ad Ignorantiam' 93

independent evidence. This is plain enough, yet o

complex or convoluted debate, considerations of onu
blurred. The simplest scenario is where we might h
Mr. X and Mr. Y, and Mr. X at some point in the
Then later in the debate, perhaps when the issue
clouded, Mr. X may aggressively demand that Mr. Y
the negation of p, in a case where Mr. Y has expres
that p obtains. Now it is altogether reasonable for
no definite or articulatable positive evidence to pro
of p, but yet on the other hand, that it is not o
evidence sufficiently warrants the acceptance of p
Mr. Y may maintain that no relevant evidence suff
acceptance or rejection is available. In this case it ma
even fallacious, for X to insist that Y produce evide
Since Mr. X originally asserted that p, it would see
his to marshall evidence in p' s behalf. Alternatively,
ate that X retreat into a more neutral modality concern
None of these requirements can be warranted by a
logic, however powerful; standard logics do not
proclaim upon, the dialectical patterns of propositio
Only a dialectical logic, of the kind studied by Ham
of games theory than standard logic 7, can chart the
skeptical challenges and refutations. A dialectical
dialogue or family of dialogues with two or more p
in turn in accord with a set of rules or conventions. Formal dialectic consists

in the setting up of simple systems of logically well-formulated but not neces-

sarily very realistic rules, and studying dialogues that can be run in accor-
dance with the rules. Hamblin has developed several interesting formal
dialectical systems.
A dialectical game that begins to capture the ad ignorantiam suggested
by the simple scenario above requires a rule allowing the questioner to put
forward r Why p? "I without requiring that his question be open to the same
sorts of justification procedures covering the case of the respondent who opts
to assert one of the pair { p, ~p } in response to a question. For this

7 See Hamblin, Ch. 8, and John Woods and Douglas Walton, 'On Fallacies.' The
Journal of Critical Analysis , IV, 3, 1972, 103-112.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
94 John Woods and Douglas Walton

simplest case, the kind of mechanism required is illus

rule in Hamblin's 'Why-Because-System-with-Question
S 3. 'Why S?' must be followed by
(a) 'Statement ~S'
or (b) 'No commitment S'
or (c) 'Statement T' where T is equivalent to S
by primitive definition,
or (d) 'Statements T,TdS' for any T.
Transparent attempts to shift the burden of proof unfai
we considered earlier, can now be classed as violation
the following specimen of dialogue between a and
rule S 3.

b : WhyS?
a : Why ~S?
It should also be added that not all questions of burden of proof are so easily
dealt with. We need to distinguish kinds and degrees of assertion, commit-
ment and concession, if the formal system is to reflect actual argumentation
more adequately to the more refined violations of burden of proof. But the
simplest cases of ad ignorantiam can be usefully understood as syntactical
aberrations of dialectic.

6. The Analysis

We suggest that an adequate understanding of the fallacy of ad ignoran-

tiam is to be achieved only through recognition of its three principal features.
First, it is a fallacy of confirmation; it is an argument from the unconfirmed
to the disconfirmed. Second, the argument from ignorance has an epistemic
aspect; it is the fallacy of arguing from the not-known to the known-to-be-
false. Third, the modus operandi of the fallacy in actual argumentation can
only be understood as a dialectical deviancy. The ad ignorantiam is com-
mitted when at least one of these three breaches is committed. Sometimes
ad ignorantiam will specifically consist in exactly one of these errors. But
more often, two or more of them will be present, and in varying degrees.
Each of the above three aspects of the ad ignorantiam outruns standard
first-order logic. Accordingly, further research on ad ignorantiam requires
an enlarged notion of correct argument for such contexts.

7. Presumptions
It is often observed that there is one special context where ad ignorantiam
is not a fallacious mode of reasoning, namely in the courts. Does not the

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 The Fallacy of 'Ad Ignorantiam' 95

law rule, for example, that a person is presumed innocent

Thus, writes Copi, "[t]he defense can legitimately claim
tion has not proved guilt, this warrants a verdict of no
no fallacy here, however, according to our analysis of ad
is no fallacy to presume a statement is false, unless pre
imply knowledge of the falsity or disconfirmation of the
The legal requirement is not that innocence be confir
only presumed. Moreover, proof of innocence in law i
achievement. Proof of innocence is merely a presump
presumptions are not dialectically fallacious either, fo
porated into dialectical games as legitimate responses t
utilization of the device of common commitment-stores
pants. In general, the set of commitment-stores for a n
may be disjoint - that is, have no statements in comm
be some statements that occur universally, in the comm
participants; these may be regarded as common or uni
of all parties. In general, as Hamblin shows, any dialec
certain metalogical agreements, concerning procedures
statement presumptions or common commitments are
there exists a commitment, C, universal to the stores of al
effect becomes a legitimate substitution instance for T
operates here like a presumption in actual argumentat
within syntactically sanctioned procedures for respon
thus evading the charge of ad ignorantiam , without
claim or confirmation-assertion concerning the substan
Presumptions have a genuine logical role to play, no
where they are extremely prominent, but also in scien
often been observed that well-established hypothes
certain inertia, or presumption of correctness, that a revol
must strive to overcome by a strong preponderance o
The more well-entrenched the current hypothesis, the
ist would-be competitors be supported by evidence. Of c
use of such presumptions is a source of fallacies - the
petitio principii may occur here as well as the ad ignora
presumptions are possible in dialectical contexts, and t
involve the ad ignorantiam . The structure of dialectic
them, and in fact they reflect their legitimacy.

8 Copi, op. cit., p. 77.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
96 John Woods and Douglas Walton

8. Testability

A complication often introduced into the discussion of ad ignorantiam

in many texts, however incidental to the structure of the fallacy as we see it,
is the question of verifiability. Typical examples of ad ignorantiam have to do
with ghosts, telepathy or other psychic phenomena, or religious argumenta-
tion, all contexts where questions regarding the verifiability, the testability-
in- principle of hypotheses, naturally arise. Ad ignorantiam here takes the
form: nobody has been able to devise an experimentally adequate means of
testing H , therefore H is false, or disconfirmed. Rather, however, the sugges-
tion is often implicit in such arguments that the hypothesis in question has
never been confirmed because it is not a testable hypothesis at all - in our
terms, the class of H-testable statements is either null or maximal, excluding
no statement, and consequently what we have is merely a pseudo-hypothesis.
Perhaps the fallacy suggested here is confusion of a testable hypothesis with
a pseudo-hypothesis of one kind or another - but this is not the fallacy of
ad ignorantiam as we conceive it.
However, a genuine case of ad ignorantiam can occur in these contexts
(if fallacies are ever genuine) as follows: nobody has ever been able to
formulate a testable hypothesis that concerns the existence of ghosts, there-
fore ghosts do not exist. Even granting the premiss, this argument proceeds
from the unknown to the known-to-be-false, from the unconfirmed (though
not through lack of data, as such, but through lack of theory) to the
disconfirmed. It is thus a true case of the classical ad ignorantiam. We have
here a variant of the fallacy, but one that conforms to our analysis.

Addendum to 'Ad Ignorantiam'

In some recent work1 we have found that the Kripke semantics for
intuitionistic logic can be very useful in providing models for forms of
argument appropriate to the study of informal fallacies that have an
epistemic character. Above we have emphasized that ad ignorantiam should
be thought of as an epistemic fallacy, and we have suggested that the fallacy

1 John Woods and Douglas Walton, ' Petitio and Relevant Many-Premissed Argu-
ments,' Logique et Analyse , 20, 1977, 97-110, 'Arresting Circles in Formal Dialogues,'
Journal of Philosophical Logic , 1978, 73-90, and 'Laws of Thought and Epistemic
Proofs,' Idealistic Studies , to appear.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 Addendum to 'Ad Ignorantiam' 97

can be modelled in a Hintikka-style syntax of epistem

seems reasonable to think that ad ignorantiam might be
semantics 3 too. Indeed this possibility is stimulated eve
known mappings between the modal system S4, wh
Hintikka's epistemic logic, and the intuitionistic ca
outline the gist of the Kripke semantics below and b
notion of ad ignorantiam reasoning can be modelled in i
The model Kripke uses is that of the tree , such
just below 5, where G is the origin, the unique botto
H4 are ascending nodes of the tree. A sentence lette
a point if written above that node; if omitted, it is
Thus in the example below, A is verified at G but B

A, B, C

ri4

Ho '

'^A,B
H:ļ

XA, C

Hi

2 Jaakko Hintikka, Knowledge and Belief , Ithaca, Cornell University Press, 1962.
3 Saul Kripke, 'Semantical Analysis of Intuitionistic Logic,' Formal Systems and
Recursive Functions , ed. J. N. Crossley and M. A. E. Dummett, Amsterdam, North-
Holland, 1965, 92-130.
4 J. C. C. McKinsey and A. Tarski, 'Some Theorems about the Sentential Calculi
of Lewis and Heyting,' Journal of Symbolic Logic , 13, 1948, 1-15. See also G. E.
Hughes and M. J. Cresswell, An Introduction to Modal Logic , London, Methuen,
1968, 305 ff.
5 See Kripke (note 3), p. 97 ff.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
98 J°hn Woods and Douglas Walton

The nodes Hi represent points of time or "evidential s

times we may have various pieces of information. If we
mation to ground a particular proposition A at a p
verified at H ; if we lack enough information to ground
we say A is not verified at H.
The Kripke tree structure as a semantic representat
advancing states of knowledge may be illustrated w
example above 6. One possibility is that we may remain
gaining any new information. But it is possible that we
information to "jump" to point Ht where we have a ver
as of A, or likewise to H2, where we will have veri
Similarly, we could be "stuck" at H2 an arbitrarily lon
to H3 or H4. Note that H2 is significantly different from
A and B are verified at both points. As long as we rem
bility of advancing to H4 remains open, but if we are a
enough information to exclude ever verifying C.
What sorts of connectives do we have in a Kripke m
and conjunction are exactly analogous to the classical tr
tion and implication however are quite different7. T
point H, we need to know at H not only that A has no
but that A cannot possibly be verified at any later time, n
new information is gained. Accordingly, ' 1 A ~i is sai
if, and only if, A is not verified at every point H' acce
rA d B "I at H, we need to know that at any later situa
a verification of A, we also have a verification of B. Th
to be verified at H if, and only if, for every point H' ac
verified at H or B is verified at H'.
Formally, the model theory is as follows. An (intuitionistic) model
structure is an ordered triple < G, K, R >, where K is a set, G is an element
of K and R is a reflexive and transitive relation on R. An (intuitionistic)
model on a model structure is a binary function 0 (P, H), where P ranges
over arbitrary proposition letters, and H ranges over elements of K, whose
range is the set {T, F}, and which satisfies the condition: if 0 (P, H) =
T and HRH' (H, H' e K), then 0 (P, H' = T. We get from atomic proposi-
tions to formulae A, B, C, . . . of the propositional calculus by the following
clauses.

(a) 0 (A A B, H) = T iff 0 (A, H) = 0 (B, H) = T; otherwise

0 (A A B, H) = F.

6 See Kripke, p. 99.

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms
 Addendum to 'Ad Ignorantiam' 99

(b) 0 (A VB, H) = T iff 0 (A, H) = T or 0 (B, H

0 (AVB) = F.
(c) 0 (A 3 B, H) = T iff for all H' s K such that HRH', 0 (A, H') =
F or 0 (B, H') = T; otherwise 0 (A z> B, H) = F.
(d) 0 O A, H) = T iff for ail H' s K such that HRH', 0 (A, H7) - F;
otherwise 0 (~i A, H) = F.
A formula A of propositional calculus is called valid iff 0 (A, G) = T for
every model on a model structure <C G, K, R >.
We should notice that it is possible to distinguish between strong and
weak negation in the Kripke model. For strong negation , ~l A, it is required
that A be not-verified at every point accessible to a given point, whereas
for weak negation , ~A, A is thought of as simply having the opposite value
of A exactly at the given point. ~ is thus in effect classical negation, and
can be defined exclusively with reference to a single point. ~i is intuitionistic
negation, and characteristically requires essential reference to a spread of
points in its definition. For "~1 A to obtain at H, ~ A must obtain at every H¡
accessible to H. Thus the inference form '~A, therefore ~l A' is in general
fallacious, and indeed clearly represents a form of ad ignorantiam ; merely
because A is unverified at a given point in time it need not follow that A
cannot be verified at any future time. In terms more reminiscent of our
Hintikka-style analysis: merely because A is not known to be true it does
not follow that A is known to be false, i. e., can never be known to be true.
Obviously much more remains to be said about the various interrela-
tionships between the Hintikka and Kripke logics in this connection. However
we must leave the elaboration of our conjecture to another paper. For the
moment we merely wish to present the basic idea of how intuitionistic logic
can provide us with a method for approaching the epistemic aspect of the
basic structure of ad ignorantiam by means of an exact formal procedure.
Ad ignorantiam begins to appear somewhat less obscure than informal
fallacies are commonly thought to be.

7 See Kripke, p. 94 and p. 99.

Dialéctica Vol. 32, No 2 (1978)

This content downloaded from

180.190.74.82 on Mon, 05 Sep 2022 05:18:55 UTC
All use subject to https://about.jstor.org/terms