ISSN 1807-9792 Volume 8, Number 1, pp.

31–39, 2014

The Theory of the Formal Discipline and

the Possible Interpretations of Conditionals:
Material Versus Defective Conditionals1
Miguel López Astorga
Universidad de Talca
Instituto de Estudios Humanísticos
milopez@utalca.cl

Keywords: conditional, defective, material, mathematics, mental models

Abstract
Attridge and Inglis try to check whether or not the ‘Theory of Formal Discipline’ is correct.
This theory states that learning mathematics improves logical reasoning, and Attridge and
Inglis review it by means of an experiment. Their conclusion is that, indeed, learning
mathematics improves conditional inferences causing that conditionals are interpreted as
defective. In this paper, I analyze Attridge and Inglis’s experiment and hold that its results
can be interpreted in a different way and that hence does not really prove that learning
mathematics lead to defective interpretations of conditionals. Equally, the paper includes
a brief reflection on how the mental models theory can explain the results achieved by
Attridge and Inglis.

1 Introduction
Attridge and Inglis (2013) focus on the discussion, open by Plato, about whether mathemat-
ics improves reasoning. Thus they carry out an experiment and interpret that their results
certainly demonstrate that idea, which is known as the ‘Theory of the Formal Discipline’.
However, they hold that, in addition, their results lead to other interesting findings. The im-
provement made by mathematics consists of understanding conditionals as defective, and not
as material.
Nevertheless, Attridge and Inglis’s (2013) results do not necessarily invalidate other alter-
native hypotheses and other interpretations of them (based on a material interpretation) are
also possible. In my opinion, the problem is that Attridge and Inglis do not consider other
interpretative possibilities and it is what causes that they consider that studying mathematics
leads one to interpret conditionals as defective. This does not mean that the task used by them
is not appropriate. In fact, it is a task that gave Evans, Clibbens, and Rood (1995) interesting
results. The real problem is how Attridge and Inglis (2013) interpret the results that they obtain
by means of that task. Their results certainly show that learning mathematics to some extent
improves conditional reasoning (the trend to biconditionality decreases and fallacies tend to
1 Thispaper is a result of the Project N. I003011, “Algoritmos adaptativos e inferencias lógicas con enunciados
condicionales”, supported by the Directorate for Research of Talca University, Chile (Dirección de Investigación de
la Universidad de Talca, Chile). The author and lead researcher of this Project would like to thank the mentioned
institutions for their help in funding this paper, and the anonymous expert that reviewed it for his (or her) comments,
which, undoubtedly, improved this text.

http://abstracta.oa.hhu.de
32 Miguel López Astorga

disappear) but there are also crucial aspects involved in conditional reasoning that seem to be
worse. In particular, those aspects refer to the use of the rule of Modus Tollens (from now on,
MT), which appears not to be appropriately applied by the participants studying mathematics
in Attridge and Inglis’s experiment.
Maybe the problem is that Attridge and Inglis’s (2013) participants are teenagers that have
only had one year of advanced mathematics (post-compulsory level in England). In particular,
they compare the results obtained by those participants in conditional reasoning tasks before
and after completing such studies. In fact, they also compare these students’ results to the
results achieved by students that have not studied advanced mathematics. However, for this
paper, as it can be noted below, only mathematics students’ results are relevant.
I say that the participants can be the problem because it is obvious that their characteristics
limit the scope of Attridge and Inglis’s (2013) results. Such results are not useful for deter-
mining whether or not learning mathematics improves logical reasoning, but only for showing
whether or not the mathematics post-compulsory level in England improves logical reasoning.
Therefore, their conclusions and the conclusions that I will expose in this paper must be con-
sidered under this limitation. In this way, it should not be forgotten that those conclusions do
not refer to learning mathematics in general, but to a concrete type of learning mathematics.
On the other hand, Attridge and Inglis (2013) also state that their results cannot ex-
plained by the mental model theory (Byrne & Johnson-Laird, 2009; Johnson-Laird, 1983, 2001,
2006, 2012; Johnson-Laird & Byrne, 2002; Johnson-Laird, Byrne, & Girotto, 2009; Orenes &
Johnson-Laird, 2012). >From a semantic perspective, this theory (from now on, MMT) claims
that human reasoning works by means of explicit and implicit models, but, according to At-
tridge and Inglis (2013), their findings are hard to explain based on it. Nevertheless, in my
view, what happens is that Attridge and Inglis do not take certain essential aspects of MMT
into account.
In this way, next pages will show the interpretation problem that I see in Attridge and
Inglis’s (2013) research, explain why their results do not demonstrate that learning mathematics
is linked to a defective interpretation of conditionals, and, finally, expose the reasons why it can
also be said that the semantic framework of MMT is coherent with such results. However, first
of all, it seems opportune to describe in more details Attridge and Inglis’s (2013) experiment.

2 Logical rules, fallacies, and interpretations of conditionals

Attridge and Inglis (2013) think that conditional can be interpreted in four ways. It can
be taken as material, biconditional, defective, and conjunction. I briefly explain what these
interpretations are (from now on, I assume that ‘1’ stands for truth and ‘0’ stands for falsehood):
• Material: This is the classical interpretation from logic and, as it is well-known, it
establishes that v(p → q) = 0 only if v(p) = 1 and v(q) = 0. Otherwise v(p → q) = 1.
• Biconditional: The causes that make conditionals are considered as biconditionals
have been studied in great details in the literature, and the conditional perfection
phenomenon (e. g. van der Auwera, 1997a, 1997b; Geis & Zwicky, 1971; Horn, 2000;
Moldovan, 2009) is especially interesting in this way. Nonetheless, what is important
here is that v(p ↔ q) = 0 when either v(p) = 0 and v(q) = 1 or v(p) = 1 and v(q)
= 0. Otherwise v(p ↔ q) = 1.
• Defective: This interpretation is related to probabilistic logic (e. g., Adams, 1998;
Adams & Levine, 1975), but, in this paper, the basic aspect of it that needs to be
considered is that, under a defective interpretation, v(p → q) = 1 if v(p) = 1 and v(q)
 Theory of the formal discipline and the possible interpretations of conditionals 33

= 1, and v(p → q) =0 if v(p) = 1 and v(q) = 0. If p is not present (i. e., ¬p happens), p
→ q is not a relevant relation.
• Conjunction: As it can be easily understood, under this interpretation, p → q is
considered as p ∧ q, and, as it is also well-known, v(p ∧ q) = 1 when v(p) = 1 and
v(q) = 1. Otherwise v(p ∧ q) = 0.
The case is that Attridge and Inglis (2013) provide equivalences between these four interpreta-
tions and two logical rules and two fallacies related to conditional. The two rules are the rule of
Modus Ponens (from now on, MP) and MT, and the two fallacies are affirming the consequent
(from now on, AC) and denying the antecedent (from now on, DA). Their equivalences are
the following:
• The material interpretation only allows using MP and MT.
• The biconditional interpretation allows using MP, MT, AC, and DA.
• The defective interpretation only allows using MP.
• The conjunctive interpretation only allows using MP and AC.
Thus, taking this equivalences into account, Attridge and Inglis (2013) carried out an exper-
iment in which, among other tasks and exercises, their participants had to solve reasoning
tasks related to MP, MT, AC, and DA. In those tasks abstract conditional propositions (which
referred to numbers and letters) were used as first premise. The second premise was the an-
tecedent of the rule (MP), the denial of its consequent (MT), its consequent (AC), or the denial
of its antecedent (DA). In this way, obviously, the conclusion was the consequent of the rule
(MP), the denial of its antecedent (MT), its antecedent (AC), or the denial of its consequent
(DA), and participants’ task was to indicate, responding ‘yes’ or ‘no’, whether or not, in each of
those four cases, the conclusion follows from the two premises.
As mentioned, Attridge and Inglis’s (2013) participants solved those tasks twice, before
and after completing a year of advanced mathematics, and the results relevant for this paper
obtained by them were these: before completing the level, the participants tended to accept in
large numbers the conclusions corresponding to the four inferences (a proportion between 0.7
and 0.8 in a scale from 0.2 to 0.8). Nevertheless, after completing it, they only tended to admit
MP (a proportion between 0.5 and 0.6 in a scale from 0.2 to 0.8). This fact was interpreted
by Attridge and Inglis (2013) as clear evidence that learning mathematics improves logical
reasoning. However, in their view, the most important finding was that the improvement
was related to the defective interpretation (which was that linked only to MP), and not to the
material interpretation.
Nonetheless, as said above, I think that Attridge and Inglis’s (2013) results do not show a
global or general improvement of their participants’ logical abilities. Their interpretation of
their results is problematic, since it is also possible to interpret such results from an approach
based on the material interpretation and to state that mathematics post-compulsory level in
England only improves certain aspect of logical reasoning (it limits the tend to interpret condi-
tional as biconditional). In my opinion, a material interpretation of such results can lead one to
say that other aspects (those related to MT) are not improved by that mathematics level, and
that, in a sense, they even worsen. This is explained in the next part.

3 The material interpretation and the rejection of MT

The main problem in the analysis of their results offered by Attridge and Inglis (2013) is that
the link between the defective interpretation and the use of only MP is not obvious. When, in
a conditional reasoning task, an individual responds ‘yes’ (that is, that the conclusion follows
34 Miguel López Astorga

from the premises), his (or her) answer means that he (or she) thinks that the denial of the
conclusion is not possible. The difficulty appears when he (or she) responds ‘no’ (that is,
that the conclusion does not follow from the premises). This situation is a difficulty because
we have no information on his (or her) thoughts, and hence we do not know the causes of
his (or her) response. Certainly, the defective interpretation can be correct and, when the
scenario refers to ¬p, he (or she) can respond ‘no’ because he (or she) considers that scenario
to be irrelevant. However, we cannot be sure of that. It is also possible that the participant
responds ‘no’ because he (or she) thinks that the conclusion is false or that the denial of the
conclusion is possible. Thus, the acceptance of only MP is not necessarily linked to the defective
interpretation. Therefore, it seems legitimate to assume other perspective and to interpret the
answer ‘no’ in a different way in order to give an alternative explanation of Attridge and Inglis’s
(2013) results.
In my view, beyond the defective interpretation, the best alternative explanation to assume
is that the participant responds ‘no’ because he (or she) thinks that the denial of the conclusion
is possible. The answer ‘no’ does not imply that the individual thinks that the conclusion is
necessarily false. Simply, it means that a scenario in which the denial of the conclusion is true
is possible. Of course, the participant can respond ‘no’ because, in his (or her) opinion, the
conclusion is false, but, given that we only know that he (or she) chose ‘no’, all we can state is
that he (or she) admits the possibility that the denial of the conclusion is true, since we cannot
know for sure whether or not he (or she) considers the conclusion to be false.
>From this perspective, in which the defective interpretation is not considered, Attridge
and Inglis’s (2013) results can have other meaning. This different meaning can be clear if we
think about the logical structure of the tasks used by Attridge and Inglis (2013) and the semantic
possibilities linked to both the answer ‘yes’ and the answer ‘no’.
As far as MP is concerned, its premises are p → q and p, and the participant must indicate
whether or not q follows. If he (or she) responds ‘yes’, he (or she) is saying that v(p ∧ ¬q)
= 0. However, if he (or she) responds ‘no’, he (or she) is not necessarily saying that v (p ∧
q) = 0. We can only be sure that he (or she) thinks that a scenario in which v(p ∧ ¬q) = 1
is possible.
On the other hand, the premises of MT are p → q and ¬q, and, in this case, the participant
must decide whether or not ¬p follows. If he (or she) responds ‘yes’, he (or she) is stating that
v(p ∧ ¬q) = 0. However, if he (or she) responds ‘no’, he (or she) is not necessarily stating that
v( ¬p ∧ ¬q) = 0. We can only be sure that he (or she) thinks that a scenario in which v(p ∧
¬q) = 1 is possible.
The case of DA is similar. The premises are now p → q and q, and participants’ task is to
decide whether or not p follows. If he (or she) responds ‘yes’, he (or she) is claiming that v
( ¬p ∧ q) = 0. However, if he (or she) responds ‘no’, he (or she) is not necessarily claiming
that v(p ∧ q) = 0. We can only be sure that he (or she) thinks that a scenario in which v( ¬p
∧ q) = 1 is possible.
Finally, in DA the premises are p → q and ¬p, and the participants must answer whether or
not ¬q follows. If he (or she) responds ‘yes’, he (or she) is saying that v( ¬p ∧ q) = 0. However,
if he (or she) responds ‘no’, he (or she) is not necessarily saying that he (or she) thinks that
v( ¬p ∧ ¬q) = 0. We can only be sure that he (or she) thinks that a scenario in which v( ¬p
∧ q) = 1 is possible.
Based on these arguments, it is correct to link biconditional to MP, MT, AC, and DA. The
answer ‘yes’ means the following:
 Theory of the formal discipline and the possible interpretations of conditionals 35

• In the cases of MP and MT: v(p ∧ ¬q) = 0.

• In the cases of AC and DA: v( ¬p ∧ q) = 0.
Therefore, given that v(p ↔ q) = 1 when v(p ∧ q)= 1 or v( ¬p ∧ ¬q) = 1, and that, by accepting
MP, MT, AC, and DA, these last possibilities are not rejected, it can be said that, certainly, an
individual that tends to consider those four inferences as valid is an individual that tends to
interpret conditional as biconditional.
Equally, it is also appropriate to relate the material interpretation to only MP and MT. The
answer ‘yes’ means the following:
• In the cases of MP and MT: v(p ∧ ¬q) = 0
But the answer ‘no’ means the following:
• In the cases of AC and DA: It is possible that v( ¬p ∧ q) = 1.
Therefore, given that v(p → q) = 0 only if v(p ∧ ¬q) = 0, and that this possibility is precisely
the only possibility that is explicitly rejected (that is in the cases of MP and MT), it can be said
that an individual that tends to consider only MP and MT as valid is an individual that tends to
the material interpretation of conditional.
Likewise, it is also right to link the conjunctive interpretation to only MP and AC. The
answer ‘yes’ means:
• In the case of MP: v(p ∧ ¬q) = 0.
• In the case of AC: v( ¬p ∧ q) = 0.
And the answer ‘no’ means:
• In the case of MT: In principle, it is possible that v(p ∧ ¬q) = 1. However, given that
this possibility is forbidden by MP, the answer ‘no’ can only mean here that v( ¬p ∧
¬q) = 0.
• In the case of DA: In principle, it is possible that v( ¬p ∧ q) = 1. However, given that
this possibility is forbidden by AC, the answer ‘no’ can only mean here that v( ¬p ∧
¬q) = 0 as well.
Therefore, given that v(p ∧ q) = 1 only if v(p) = 1 and v(q) = 1, and that all other possibilities
are rejected, it can be said that an individual that tends to accept only MP and AC is an
individual that tends to interpret conditional as conjunction.
The problem is, as indicated, the relation of the defective interpretation with only MP. Of
course, it can be thought that an individual that only accepts MP is an individual that rejects
MT, AC, and DA because they are irrelevant (i. e., they refer to scenarios in which p is not).
Nevertheless, given that, as commented, we do not know participants’ thoughts, it can also
be argued that the acceptance of only MP can also be linked to the material interpretation. I
explain this idea below.
As said, in the case of MP, the answer ‘yes’ means that the individual thinks that v(p ∧
¬q) = 0. However, in the other cases, the answer ‘no’ allows certain possibilities. As also
mentioned, in the cases of AC and DA, v( ¬p ∧ q) = 1 is allowed, and this fact, evidently, is
not a problem. The difficulty is given by MT. In its case, the answer ‘no’ means, apparently,
that the participant accepts as possible a scenario in which v(p ∧ ¬q) = 1. Obviously, this is an
inconvenience, since the combination p ∧ ¬q is forbidden by MP. Of course, this inconsistency
can lead one to think that the material interpretation cannot be linked to the participants that
only accepted MP. The material interpretation not only requires the acceptance of MP and the
rejection of AC and DA, but also the acceptance of MT. Therefore, this is the point that needs
to be explained.
36 Miguel López Astorga

MT is very different from MP. MP is a simple and basic rule, but MT is complex and it
cannot be used without other rule: Reductio ad Absurdum. If one wants to apply MT, he (or she)
needs to adopt p as assumption to apply MP (considering the premise p → q and the assumption
p), which leads to obtain q, and to note that the premise ¬q is inconsistent with q, and that
hence p is not possible. This process is far more complex than that of MP (which only needs one
step) and this additional difficulty has been reported in the literature (e. g., Byrne & Johnson-
Laird, 2009; López Astorga, 2013). In fact, Attridge and Inglis (2013) also refer to this issue.
They state that this is the explanation of the difficulty of MT that the authors that support the
idea of a mental logic often raise. They even seem to acknowledge that this explanation can
be valid, even though the defective interpretation is assumed. Nonetheless, what is important
is that, if this explanation is considered valid, Attridge and Inglis’s (2013) results do not prove
that mathematical study (at least, the kind of mathematical study corresponding to the post-
compulsory level in England) causes a trend towards the defective interpretation. It is possible
to interpret conditionals materially and, however, not to use MT. And this is because this last
rule is hard to apply.
Thus, from this perspective, it can be said that mathematics only improves logical reasoning
in a certain sense. Before the post-compulsory level, students tend to interpret conditionals
as biconditionals and, according to Attridge and Inglis’s (2013) results, learning mathematics
corrects this problem. Nevertheless, logical reasoning ability tends to be worse in other sense
after that same level, since students have difficulties to use MT. Such difficulties are not observed
before the post-compulsory level because, under the biconditional interpretation, v( ¬p ∧ ¬q)
= 1, which, apparently, must be noted to apply MT. Maybe it would be interesting to review
the syllabi corresponding to the mathematics post-compulsory level. It is possible that they
do not include the resolution of problems with processes similar to the application of MT,
and that this fact is the cause that mathematics students do not often use this rule after the
post-compulsory level.

4 MMT and learning mathematics

The main aim of this paper is not to propose arguments in favor of MMT. My more impor-
tant goal is only show that Attridge and Inglis’s (2013) results do not necessarily mean that
mathematical study lead to a defective interpretation of conditionals. However, given that At-
tridge and Inglis (2013) state that their results are difficult to understand from the basic theses of
MMT and that I think that that idea is not correct, it can be opportune to explain briefly, before
concluding, why, in my view, Attridge and Inglis’s (2013) research does not cause difficulties
to MMT.
MMT is a wide theory on human reasoning. Nevertheless, what is relevant for this paper is
only its explanation about conditionals. According to MMT, when an individual reasons about
conditional propositions, he (or she) considers the possibilities, or models, corresponding to
each conditional. Thus, a proposition such as p → q has three models:
A.– p & q
B.– ¬p & q
C.– ¬p & ¬q
The key point is that only A is an immediate and explicit model. Both B and C need certain
cognitive effort and the action of memory for becoming explicit models. In this way, faced to a
conditional, individuals firstly take only one model (A) into account. B and C, as indicated, are
models more difficult to consider.
 Theory of the formal discipline and the possible interpretations of conditionals 37

Attridge and Inglis (2013) think that these theses are not coherent with their results. If their
participants interpreted conditionals as biconditionals before mathematical advanced level, it
means that they considered two models in that moment. In particular, the considered models
were A (explicit) and C (implicit). A allowed the application of MP and AC, and C allowed
the application of MT and DA. The problem is that, as I understand Attridge and Inglis’s
arguments, it is not easy for MMT to explain what happens at the later point in time, i. e.,
the acceptance of only MP. Certainly, it could be argued that C is eliminated after learning
mathematics, but, if A is the only model that is considered, MP should not be the only rule
that is accepted. AC should be accepted as well. This is, at least in my view, what Attridge
and Inglis (2013) seem to mean as regards MMT.
However, I think that arguments such as these are not correct. A relevant datum that must
be considered is that, after studying mathematics, a greater tendency to accept MP and AC (i. e.,
to the conjunctive interpretation) was also observed, which means that mathematical advanced
study is somehow linked to the rejection of the biconditional interpretation and the acceptance
of the conjunctive interpretation, that is, to the rejection of the model C and the acceptance
of only A (the explicit model). Nevertheless, because what was most striking was that the
mathematics post-compulsory level leaded the participants to admit only MP, it is obvious that
MMP needs to clarify this last fact. In this way, I can say that, in my opinion, it is true that
A and C are the models that must be taken into account if the interpretation is biconditional,
but the acceptance of only MP can be explained, from the perspective of MMT, as a change of
models. Indeed, if the considered models are not A and C, but A and B, MP must be accepted
(by virtue of A), MT must be rejected (C is not considered), AC must also be rejected (by
virtue of B), and DA must be rejected as well (by virtue of B too). Thus, based on MMT,
it can be stated that learning mathematics also improves logical reasoning in a sense, since it
causes that conditional relations are not interpreted as biconditional relations (the model B is
added). Nonetheless, it is worse in another sense, because the model C is lost and hence it is
not possible to accept MT.
Of course, this explanation proposed by me can be questioned by referring to certain theses
of MMT. For example, Johnson-Laird and Byrne (2002) state that individuals only have two
options: they can consider the explicit model (A) or all the models (A, B, and C). Thus, it
does not seem possible that the participants only consider A and B (without C). However,
as far as this difficulty is concerned, I can say that my arguments are only an interpretation
of MMT that respects their main and basic theses and that, at the same time, is consistent
with Attridge and Inglis’s (2013) results. After all, MMT allows some models to be blocked
in certain circumstances and, according to it, individuals only represent what they believe to be
true. In this way, it can be thought that my explanation develops ideas that, although they are
not clearly held by the adherents of MMT, do not contradict the core of this theory.
Thus, if my interpretation of MMT is right, it can be said that it continues to be clear that
mathematics post-compulsory level in England corrects the mistake of interpreting condition-
als as biconditionals. Equally, it can be stated that, from the perspective of MMT, the problem
is also MT, since it seems that proofs related to the opposite of what must be proved (which are
linked to C) are not practiced by students in post-compulsory level.

5 Conclusions
It is hence obvious that, if we do not assume the defective interpretation, it can be said that
mathematical study, or mathematics post-compulsory level in England, only improves logical
reasoning in a sense: p → q is not interpreted as p ↔ q. The ability of use MT is not improved.
38 Miguel López Astorga

This fact does not mean that the theory of the formal discipline can only be considered true
if the defective interpretation is accepted. Maybe the problem is that the kind of mathematics
taught in the post-compulsory level is not the type required for improving the use of some
logical rules (especially MT). In addition, other possibility is that the post-compulsory level
is not enough and that more mathematics levels are necessary for a better use of such rules.
In any case, there is no doubt that Attridge and Inglis’s (2013) results do not conclusively
prove that learning mathematics lead one to a defective interpretation of conditionals. As
shown above, such results are coherent with the idea that their participants continue to inter-
pret conditionals materially after the post-compulsory level. Furthermore, as also commented,
MMT is also compatible with their results and can explain them.
Obviously, I cannot deny that the defective interpretation continues to be a possibility that
must be taken into account. Undoubtedly, although Attridge and Inglis’s (2013) results are
consistent with a material interpretation and MMT, it is evident that they are also consistent
with the defective interpretation proposed by Attridge and Inglis (2013). In this way, it can
only be said that, if mathematical study leads to a defective interpretation of conditionals, that
fact needs to be proved by means of further research. So far, we have no proofs that, certainly,
that is the case. The discussion is open and other interpretations are also possible.

References
Adams, E. W. (1998), A Primer of Probability Logic, Center for the Study of Language and
Information Publications, Stanford.
Adams, E. W. & Levine, H. P. (1975), ‘On the uncertainties transmitted from premises to
conclusions in deductive inferences’, Synthese 30, 429–460.
Attridge, N. & Inglis, M. (2013), ‘Advanced mathematical study and the development of condi-
tional reasoning skills’ PLoS ONE 8 (7), 269–399. Retrieved from http://www.plosone.org/article/
info%3Adoi%2F10.1371%2Fjournal.pone.0069399 (March 11, 2014).
Auwera, J. van der (1997a), ‘Pragmatics in the last quarter century: The case of conditional
perfection’, Journal of Pragmatics 27, 261–274.
Auwera, J. van der (1997b), Conditional perfection, in A. Athanasiadou & R. Dirven, eds.,
‘On Conditional Again’, John Benjamins, Amsterdam, pp. 169–190).
Byrne, R. M. J. & Johnson-Laird, P. N. (2009), ‘“If” and the problem of conditional reasoning’,
Trends in Cognitive Science 13, 282–287.
Evans, J. St. B. T., Clibbens, J., & Rood, B. (1995), ‘Bias in conditionals inference. Implications
for mental models and mental logic’, The Quarterly Journal of Experimental Psychology 48,
644–670.
Geis, M. L. & Zwicky, A. M. (1971), ‘On invited inferences’, Linguistic Inquiry 2, 561–566.
Horn, L. R. (2000), ‘From if to iff: Conditional perfection as pragmatics strengthening’, Journal
of Pragmatics 32, 289–326.
Johnson-Laird, P. N. (1983), Mental Models. Towards a Cognitive Science on Language, Inference
and Consciousness, Harvard University Press, Cambridge, MA.
Johnson-Laird, P. N. (2001), ‘Mental models and deduction’, Trends in Cognitive Science 5,
434–442.
Johnson-Laird, P. N. (2006), How We Reason, Oxford University Press, Oxford.
Johnson-Laird, P. N. (2012), Inference with mental models, in K. J. Holyoak & R. G. Morrison,
eds., ‘The Oxford Handbook of Thinking and Reasoning’, Oxford University Press, New
York, pp. 134–145.
 Theory of the formal discipline and the possible interpretations of conditionals 39

Johnson-Laird, P. N. & Byrne, R. M. J. (2002), ‘Conditionals: A theory of meaning, pragmat-

ics, and inference’, Psychological Review 109, 646–678.
Johnson-Laird, P. N., Byrne, R. M. J., & Girotto, V. (2009), ‘The mental models theory of
conditionals: A reply to Guy Politzer’, Topoi 28, 75–80.
López Astorga, M. (2013), ‘Are conditional and disjunction really comparable?’, Universum 28
(2), 229–242.
Moldovan, A. (2009), ‘Pragmatic considerations in the interpretation of denying the an-
tecedent’, Informal Logic 29 (3), 309–326.
Orenes, I. & Johnson-Laird, P. N. (2012), ‘Logic, models and paradoxical inferences’, Mind &
Language 27 (4), 357–377.