REASONING*

K. P. Mohanan and Tara Mohanan

1. Introduction to Reasoning
1.1. What is “reasoning”?
1.2. Premises and conclusions
1.3. Spelling out the chain of reasoning
1.4. Implicit and explicit premises
2. Typology of Reasoning
2.1. Modus Ponens
2.1.1. Classical Modus Ponens
2.1. 2. Classical vs. Probabilistic Modus Ponens
2.1. 3. Classical vs. Defeasible Modus Ponens
2.2. Reverse Modus Ponens
2.2.1. Abduction
2.2.2. Speculative Deduction
2.3. Reasoning from Sample to Population
2.4. Analogical Reasoning

Exercises

*
This material has benefited from discussions with and feedback from Sunita Anne Abraham and Malavika Tara
Mohanan, and dozens of desperate questions from students over the last few years.
Include other rules of inference:
If P then Q, P therefore Q
If P then Q, not Q, therefore not P
If P then Q, If Q then R, therefore If P then R
If P then either Q or R, P and not R, therefore Q.

1 Introduction to Reasoning

1.1 What is reasoning?

If we see dark clouds in the sky, we infer that it is likely to rain. If we hear the sound of laughter
behind a closed door, we infer that behind that door there is either a person or a device like a
tape recorder or TV set. The experience of seeing the dark clouds and of hearing the sound of
laughter is an instance of the grounds that lead us to the conclusion about the likelihood of rain
and about a person or device behind the door. The process of inference connects the grounds to
the conclusion.
grounds
process of
inference

conclusion figure 1

In some cases, the grounds of inference and the steps leading from the grounds to the conclusion
can be expressed (either verbally or using some other symbolic system) such that someone else
can scrutinize the process and make an assessment its quality. For instance, given the
information that Zeno is a spider, we can infer that Zeno has eight legs. If we are asked to
express it verbally, we may do so as follows:
Zeno is a spider.
Therefore, it is reasonable to conclude that Zeno has eight legs.
If we are asked how exactly the conclusion of Zeno having eight legs follows from Zeno being a
spider, we might flesh out our reasoning as follows:
Zeno is a spider.
All spiders have eight legs.
Therefore, it is reasonable to conclude that Zeno has eight legs.
Likewise, given the information that Leda’s room is 10 ft wide and 15 ft long, it is reasonable to
conclude that the area of the room is 150 sq ft. We can articulate the process of inference in a
combination of words and numbers as follows:
Width of Leda’s room = 10 ft
Length of Leda’s room = 15 ft
Hence, the area of Leda’s room = 150 sq.ft
Reasoning is a form of inference that can be expressed in words or other symbols such that the
legitimacy of the process can be evaluated if needed.
Implicit in what we have said above is the idea that not all forms of inference can be overtly
expressed. Suppose you walk into an auditorium with a friend, look around at the seated people,
and tell your friend, “There seems to almost a hundred people here.” You didn’t actually count
the number of people and arrive at this conclusion, instead, you made in inference, based on
your visual experience. If your friend challenged you to express the steps of your inference in
words or numbers, you may not be in a position to do so.

2
Reasoning is an integral component of the various aspects of academic inquiry, including
i) looking for answers/solutions to questions/problems (methodology);
ii) justifying/proving our answers/solutions (justification); and
iii) evaluating the merit of the answers/solutions (critical thinking).
Given the centrality of reasoning in inquiry, it is important for us to understand the nature of
reasoning in academic inquiry. In what follows, we will explore forms of reasoning that can be
expressed in words (qualitative reasoning) or in numbers (quantitative reasoning).

1.2 Premises and conclusions

When formulated as statements expressed in words or numbers, the grounds that we appeal to in
our inferences are called premises. Take a look at the premises in our examples of Zeno’s legs
and area of Leda’s room:

(1) Zeno is a spider. Premise 1

All spiders have eight legs. Premise 2

Therefore, it is reasonable to conclude that Zeno has eight legs. Conclusion

Likewise, given the information that Leda’s room is 10 ft wide and 15 ft long, it is reasonable to
conclude that the area of the room is 150 sq ft. We can articulate the process of inference in a
combination of words and numbers as follows:
(2) The width of Leda’s room = 10 ft Premise 1
The length of Leda’s room = 15 ft Premise 2

Hence, the area of Leda’s room = 150 sq.ft Conclusion

Take another example to illustrate the concept of premises. Suppose you are told that all gleeps
are dovines, all dovines have green eyes, and Blimpsey is a gleep. You are now asked: “What is
the colour of Blimpsey’s eyes?” Even though you have no idea what gleeps and dovines are, and
who Blimpsey is, you will have no trouble inferring that Blimpsey’s eyes are green. The
reasoning in this example can be expressed as:
(3) All gleeps are dovines. Premise 1
All dovines have green eyes. Premise 2
Blimpsey is a gleep. Premise 3

Therefore, Blimpsey’s eyes are green. Conclusion

We may now see reasoning as the process of arriving at a conclusion from a given set of
premises:
Reasoning premises

steps of
reasoning figure 2

conclusion

1.3 Spelling out the chain of reasoning

Example (3) leaves out an intermediate step in reasoning. Given that all gleeps are dovines, and
that Blimpsey is a gleep, we conclude that Blimpsey is a dovine. Given that all dovines have
green eyes and that Blimpsey is a dovine, we can conclude that Blimpsey has green eyes. Here is
a version of (2) that spells out the implicit steps:

3
4
(4) All gleeps are dovines. Premise 1
Blimpsey is a gleep. Ppremise 3
Therefore: Blimpsey is a dovine. Conclusion (intermediate)

Blimpsey is a dovine.
All dovines have green eyes. Premise 2
Therefore: Blimpsey has green eyes. Conclusion (final)

1.4 Implicit and explicit premises

Intuitively, we recognize the reasoning in example (2) as quite legitimate. However, there are
two crucial premise that are taken for granted in this reasoning, namely, that Leda’s room is a
rectangle, and the area of a rectangle is the product of its length and width. To be explicit,
therefore, we would spell out these implicit premises as follows:
(5) Leda’s room is a rectangle. premise 1
The length of Leda’s room = 15 ft. premise 2
The width of Leda’s room = 10 ft. premise 3
The area of a rectangle = length x breadth. premise 4
Therefore: The area of Leda’s room = 150 sq ft. conclusion

In the actual practice of reasoning, we often leave out premises that are taken for granted by the
writer and the reader. This happens even in the most rigorous forms of reasoning such as that
involved in mathematical proofs.
Though the inexplicitness illustrated of missing steps and implicit premises is not a serious flaw
in the particular examples given above, we should watch out against the danger of the inexplicit
premises being false or untenable, and implicit steps involving an invalid derivation. For
instance, even though most rooms are rectangular in shape, there is a possibility that Leda’s
room is elliptical. If so, the conclusion in (2)/(5) would be false. When we evaluate a sample of
reasoning, therefore, it is important to unearth the hidden premises and check if any of them is
false or untenable.

2 Typology of Reasoning

2.1 Modus Ponens

2.1.1 Classical Modus Ponens
To repeat, reasoning is the process of arriving at inferences from a given body of information.
The discipline of logic is the systematization of reasoning. It studies the abstract patterns of
good reasoning and systematizes them, such that we can:
• further develop and strengthen our reasoning capacity, and
• evaluate the legitimacy of the reasoning that we are exposed to.
To do this, logic has to articulate explicit principles of good reasoning, so that we can
distinguish between good reasoning and bad reasoning. To this end, logicians formulate both the
information and the inferences involved in reasoning as propositions. The propositions carrying
the information are the premises, and the propositions carrying the inferences that we derive
from the premises are the conclusions.
Information Inferences
Premises steps of reasoning Conclusions

5
How do we distinguish between good and bad reasoning? Take the following examples:
Example A Example B
All humans are mammals. All humans are mammals.
Zenta is a human. Xilo is a mammal.
Therefore Zenta is a mammal. Therefore Xilo is a human.
Intuitively, we recognize A to be an instance of good reasoning and B to be an instance of bad
reasoning. We therefore expect logic to tell us why this is so.
To unearth what is bad about B, the first step is to express the propositions in such a way that
their logical structure becomes clear. The proposition expressed by the sentence All humans are
mammals is the same as “For any x, if x is a human, x is a mammal.” Let us restate examples A
and B as A’ and B’:
Example A’ Example B’
For any x, if x is a human, x is a mammal. For any x, if x is a human, x is a mammal.
Zenta is a human. Xilo is a mammal.
Therefore Zenta is a mammal. Therefore Xilo is a human.
Let us denote x is a human as P and x is a mammal as Q:
For any x, if x is a human, x is a mammal.
P Q
We now see that the sentence All humans are mammals expresses a complex proposition
consisting of two simple propositions P and Q, and has the form: If P, then Q.
Logicians use an arrow to denote the “if…then” relation, and write “If P, then Q” as: P ! Q. We
can now express the logical structure of examples A’ and B’ as A” and B”:
Example A” Example B”
Premise 1: P!Q P!Q
Premise 2: P Q
Conclusion: Therefore Q Therefore P
Why is the structure in B” bad, in contrast to that in A”? An answer calls for the concept of rules
of inference, the abstract patterns that sanction a conclusion from a given set of premises. One
1
such rule of inference in deductive logic is called Modus Ponens, which says:
Rule of inference: (Classical) Modus Ponens
Given that: If P is true, then Q is true
and P is true
it is reasonable to conclude that: Q is true.
Modus Ponens sanctions the reasoning in A”, but not in B”. The reasoning in B” is bad because
the rule of Modus Ponens does not sanction it, nor have we postulated any other rule of
inference that legitimizes it. (We will postulate more rules of inference as we proceed.)

1
Why we call it Classical Modus Ponens will become clear in the next section.

6
The assumption that we are making above is that each step in a chain of reasoning from
premises to conclusions must be mediated by a legitimate rule of inference:
Premises
Reasoning Rules of inference
Conclusion
2.1.2 Classical vs. Probabilistic Modus Ponens
Consider the following examples of reasoning:
(6) Every human being is mortal.
Socrates was a human being.
Therefore, Socrates was mortal.
(7) All human beings have their heart on the left side and their liver on the right side.
Plato was a human being.
Therefore, Plato had the heart on the leftt side and the liver on the right side.
The reasoning in these examples appeals to the rule of Modus Ponens discussed in the previous
section. Example (6) is found in most introductory books on logic. We will refer to this as
classical modus ponens.
Classical Modus Ponens has the following property:
If the premises are true, and
the rule is applied correctly,
the conclusion must necessarily be true.
It cannot be the case that the statements that every human being is mortal and that Socrates was
a human being are both true, and yet it was not true that Socrates was mortal. Thus:
If the premises are totally certain, and the application of the rule is correct, Classical
Modus Ponens yields a totally certain conclusion.
Now, we haven’t seen any evidence to believe that the statement that every human being is
mortal is incorrect: we have not come across any immortal human being yet. However, the first
premise in (7) is not accurate. There is a rare genetic condition in healthy human beings called
situs inversus in which the heart is on the right and the liver is on the left. The frequency of situs
inversus varies across populations, but apparently it is less than 1 out of 10, 000. But the point is
that if you take a large enough sample, you might come across a human being with heart on the
right and liver on the left. Hence, the reasoning in (7) should be revised as (8).
(8) 99.99% of human beings have the heart on the left side and the liver on the right side.
Plato was a human being.
Therefore, the probability of Plato having the heart on the left side and the liver on the
right side is 0.9999.
As it happens, there are also people with the heart on the left, and yet the liver is also on the left.
This condition is called situs inversus incompletes. It appears to be the case that 1 out of 22,000
cases of situs inversus are instances of situs inversus incompletes. Thus one out of 220, 000, 000
human beings have both their heart and liver on the left side of their body:
(9) 1 out of 220,000,000 human beings have both the heart and the liver on the left side of
their body.
Plato was a human being.
Therefore, the probability of Plato having both the heart and the liver on the left side of
their body is approximately 0.000000045

7
The reasoning illustrated in (8) and (9) is probabilistic. To see the distinction between (7) and
(8) more clearly, let us re-formulate (7) in numerical terms.
(7’) 100% of human beings have the heart on the left side and the liver on the right side.
Plato was a human being.
Therefore, the probability of Plato having the heart on the right side and the liver on the
right side is 1.
(7’) is an instance of classical deduction in which truth is either 1 (true) or 0 (false), i.e.,
conclusions are totally certain. (8) and (9) are examples of probabilistic reasoning that allow
degrees of certainty, truth and falsity ranging from 1 to 0.
What would a Probabilistic Modus Ponens look like? To answer this question, let us re-
formulate Classical Modus Ponens in quantitative terms (i.e., in terms of numbers):
(Classical) Modus Ponens (in quantitative terms)
Given that: If P is true, then the probability of Q is 1.
and P is true
it is reasonable to conclude that: the probability of Q is 1.
It is now easy to see that the probabilistic version of Modus Ponens can be formulated as
follows, replacing 1 with n, where n ranges from 1 to 0.
Probabilistic Modus Ponens (in quantitative terms)
Given that: If P is true, then the probability of Q is n.
and P is true
it is reasonable to conclude that: the probability of Q is n.
Probabilities are expressed qualitatively in ordinary language in terms of adjectives like likely,
most likely, almost certainly, and so on. So the qualitative version of the Probabilistic Modus
Ponens would be:
Probabilistic Modus Ponens (in qualitative terms)
Given that: If P is true, then Q is (most) likely to be true.
and P is true
it is reasonable to conclude that: Q is (most) likely to be true.

2.1.3 Classical vs. Defeasible Modus Ponens

Suppose we are told that Xena goes to her office every weekday, and that 2 January 2007 was a
weekday. We will conclude that Xena went to her office on 2 Jan. 2007.
Now let us suppose that we are given the additional information that Xena does not go to her
office when she is sick, and that she was sick on 2 Jan. 2007. We would most probably conclude
that Xena did not go to her office on 2 Jan. 2007, overriding the earlier conclusion that she went
to her office on that day.
Now we are told that there was an emergency in Xena’s office on 2 Jan. 2007, and whenever
there is an emergency, she goes to her office. Did she go to her office on 2 Jan. 2007? Our
answer would probably depend upon whether the emergency was strong enough to override the
sickness factor.
The shifts in the conclusions in this example illustrates an important property of a large class of
human reasoning, namely: when additional information becomes available, a conclusion that
was accepted earlier as correct can be rejected as incorrect.
A mode of reasoning that has the above property is called defeasible reasoning. Classical
Modus Ponens is non-defeasible. To see this clearly, let us derive the conclusions in the example
of Xena and her office using Classical Modus Ponens.

8
 (10) a. If x is a weekday, Xena goes to her office on day x.
b. If Xena is sick on day x, she does not go to her office
c. 2 Jan. 2007 was a weekday.

d. Therefore, by (a) and (c), Xena went to her office on 2 Jan. 2007.
(11) a. If x is a weekday, Xena goes to her office on day x.
b. If Xena is sick on day x, she does not go to her office
c. 2 Jan. 2007 was a weekday.
d. Xena was sick on 2 Jan. 2007.

e. By (a) and (c), Xena went to work on 2 Jan. 2007.

f. By (b) and (d), Xena did not go to her office on 2 Jan. 2007.
g. By (e) and (f), therefore, Xena went to her office and did not go to her office on 2
Jan. 2007.
The reasoning in (11), using Classical Modus Ponens, yields a logical contradiction. This is
because in this mode of reasoning, there is no way to cancel an earlier inference as long as the
premises remain. To model the mode of reasoning that blocks the conclusion that Xena went to
her office on Jan. 2, we need a rule of inference that is capable of suppressing an otherwise valid
conclusion. This is provided by Defeasible Modus Ponens:
Defeasible Modus Ponens
Given (i) if P is true, then Q is true
(ii) if R is true then not-Q is true
(iii) P is true.
(iv) R is true
and (v) (i) is stronger than (ii)
it is reasonable to conclude that Q is true and not-Q is false.

Thus, (v) cancels the inference from the combination of (ii) and (iv), namely, that Q is not true.
This is the crucial aspect of Defeasible Modus Ponens. For this mode to work, we need to survey
all the candidates that are potential conclusions, check if there is a logical contradiction, and if
there is, cancel the weaker conclusion:
(12) a. If x is a weekday, Xena goes to her office on day x.
b. If Xena is sick on day x, she does not go to her office
c. (b) is stronger than (a).
d. 2 Jan. 2007 was a weekday.

e. Candidate by (a) and (d): Xena went to her office on 2 Jan. 2007.
f. No rival candidate, so we conclude that Xena went to her office on 2 Jan. 2007.
(13) a. If x is a weekday, Xena goes to her office on day x.
b. If Xena is sick on day x, she does not go to her office
c. (b) is stronger than (a).
d. 2 Jan. 2007 was a week day.
e. Xena was sick on 2 Jan. 2007.

f. Candidate: by (a) and (d), Xena went to work on 2 Jan. 2007.

g. Candidate: by (b) and (e), Xena did not go to her office on 2 Jan. 2007.
h. By (c), candidate (g) wins, and hence Xena did not go to her office on 2 Jan.
2007.
Let us take another example. If we are told that Nixon is a Republican and Republicans are not
pacifists, we conclude by Classical Modus Ponens that Nixon is not a pacifist. If we are told that

9
Nixon is a Quaker, and that Quakers are pacifists, Classical Modus Ponens yields the conclusion
that Nixon is a pacifist. If we are told that Nixon is both a Quaker and Republican, classical
deductive reasoning leads to a conclusion that is logically contradictory, namely, that Nixon is
not a pacifist and Nixon is a pacifist.
(14) Nixon is a Quaker. Nixon is a Republican.
Quakers are pacifists. Republicans are not pacifists.
! Nixon is a pacifist. ! Nixon is not a pacifist.

Nixon is a pacifist and Nixon is not a pacifist.

In a system of defeasible reasoning, otherwise valid conclusions from one set of premises can be
defeated by stronger conclusions from other premises. If we are told that Nixon is both a Quaker
and Republican, we have to decide which of these conflicting pulls is stronger. If being a
Republican has priority over being a Quaker, for instance, we conclude that Nixon is not a
pacifist in a system of defeasible reasoning:
(15) Nixon is a Quaker. Nixon is a Republican.
Quakers are pacifists. Republicans are not pacifists.
! Nixon is a pacifist. ! Nixon is not a pacifist.

Nixon is not a pacifist.

Moral reasoning, that is, reasoning from moral principles to moral judgments, crucially needs
Defeasible Modus Ponens. Suppose we accept the following principles in our moral theory.
A. i. If an action brings about the death of a human being, given an option between
action and inaction, choosing action is immoral.
ii. If an action saves the life of a human being, given an option between action and
inaction, choosing inaction is immoral.
B. i. If an action causes suffering to a human being or increases the suffering, given an
option between action and inaction, choosing action is immoral.
ii. If an action prevents or alleviates the suffering of a human being, given an option
between action and inaction, choosing inaction is immoral.
C. i. Given an option between telling a lie and not telling a lie, choosing to tell a lie is
immoral.
ii. Given an option between speaking the truth and not speaking the truth, choosing
not to speak the truth is immoral.
Take the following scenario borrowed from philosopher Peter Singer as an illustration of moral
dilemmas. Imagine that you are a Christian living in Germany during Hitler’s time. In your
basement is a Jewish family hiding from the Nazis. One day, the Nazi secret police knock on
your door, and ask if you are harboring any Jews in your house. What is the right thing to do, tell
the truth and send the Jewish family to a prison camp, or tell a lie and save the family?
Most of us would opt for the latter. Given the moral theory that includes A-C above with the
specification that A and B are stronger than C, the reasoning that underlies this conclusion can
be given as follows.

10
 (16) a. If an action brings about the death of a human being, given an option between
action and inaction, choosing action is immoral. (Ai)
b. If an action causes suffering to a human being or increases the suffering, given an
option between action and inaction, choosing action is immoral. (Bi)
c. Given an option between telling a lie and not telling a lie, choosing to tell a lie is
immoral. (Ci)
d. (Ai) and (Bi) are stronger than (Ci)
e. If we tell the Gestapo about the Jews hidden in our house, it would cause
suffering and death for them.
f. If we tell the Gestapo that there are no Jews hiding in our house, we are telling a
lie.

g. Candidate: by (a), (b) and (e), it is immoral to tell the Gestapo that there are Jews
hiding in our house.
h. Candidate: by (c) and (f), it is immoral to tell the Gestapo that there are no Jews
hiding in our house.
i. By (d), the winner is (g). Hence, we conclude that it is immoral to tell the
Gestapo that there are Jews hiding in our house.

2.2 Reverse Modus Ponens

2.2.1 Abduction
Compare the reasoning in (17a) and (17b):
(17) a. Whenever it rains, the streets get wet. It was raining last night. Therefore, the streets
must have got wet.
b. Whenever it rains, the streets get wet. The streets are wet now. Therefore it must have
rained.
(17a) is an example of the familiar Modus Ponens. Given that (i) rain causes wetness of the
streets, and (ii) rain (the cause) is observed, we conclude that wetness of the streets (the effect)
must have happened. In other words, we infer the effect from the cause in (17a).
In contrast, we are inferring the cause from the effect in (17b), reversing the direction of
inference. It was philosopher Charles Sanders Peirce who, noticing the use of this form of
reasoning in everyday life and academic knowledge, called it abduction (to distinguish this
mode of reasoning from deduction.)
Abductive reasoning is common in the justification of interpretations, in almost all academic
fields outside of mathematics and logic. The form of reasoning that doctors use in diagnosing
diseases from symptoms, for instance, is abductive. Take the following example:
(18) When someone has a heart attack, (s)he tends to have a feeling of strangulation, pain in the
chest radiating to the left shoulder and arm, abnormal perspiration, shortness of breath, and
nausea.
Fanny Jenkins has just experienced a feeling of strangulation, pain in the chest radiating to
the left shoulder and arm, abnormal perspiration, shortness of breath, and nausea.
Therefore it is reasonable to conclude that Fanny Jenkins had a heart attack.
2
Similarly, a great deal of the reasoning used in a law court is also abductive.

2
For instance, consider this. If someone shattered a windowpane from outside in order to break into the apartment,
there would be broken glass inside near the window. In addition to the door being wide open and the missing stereo, I
see broken glass inside the apartment near the window. It is therefore reasonable to conclude that someone broke into
the house by shattering the windowpane.

11
It is important distinguish valid abduction from the form of illegitimate reasoning illustrated by
examples such as those in (19):
(19) a. Whenever Leda teases Zeno, Zeno cries.
Zeno is crying now. Therefore, Leda must have teased him.
b. All horses are mammals. A rabbit is a mammal. Therefore a rabbit is a horse.
The conclusions in (19a) and (19b) are not derivable through Modus Ponens. In other words, the
reasoning in (19a) and (19b) are not validated through Modus Ponens. But that by itself is not
sufficient to explain why they are examples of bad reasoning, because Modus Ponens does not
validate any of the examples of good reasoning in (17b) and (18) either.
To see why (17b) and (18) are examples of good reasoning while (19a) and (19b) are not, it is
necessary to spell out the rule of abduction:
Abduction
Given that: P is the best available explanation for Q
and Q is true
it is reasonable to conclude that: P is true
in the absence of (i) evidence to the contrary, and
(ii) a better or equally good alternative explanation for Q.

The crucial ingredient here is: in the absence of evidence to the contrary and a better or equally
good alternative explanation for Q. This specification is absent in Modus Ponens.
In the case of the streets being wet, we have not found any counterevidence (evidence to show
that it did not rain), and in the case of the doctors observations, we have not found any obvious
alternative causes. Hence the reasoning in both cases is legitimate. In the case of Zeno crying,
however, we are aware of other potential causes that could have led to the effect of Zeno crying:
he could have fallen down and hurt himself, Mary could have scolded him, his teacher may have
given him a poor grade, and so on. Likewise, there are many animals other than rabbits that are
mammals, and hence the rule of abduction does not legitimize (19b).
When it rains, the streets get wet.
The streets are wet.
Therefore, it must have rained. (valid abduction, alternative causes unavailable)

When Leda beats Zeno, Zeno cries.

Zeno is crying now.
Therefore, Leda must have beaten Zeno. (invalid abduction, alternative causes available)
We accepted the conclusion that it must have rained because we couldn’t think of a better or
equally good explanation. Suppose someone proposes the following explanation:
When the water truck goes down the streets watering the ground, the streets get wet.
The streets are wet.
Therefore, the water truck must have gone down the street.
We now have a competing explanation, and our previous conclusion is no longer valid. We must
therefore look for further evidence to choose between the two. Suppose we examine the
sidewalk and the lawns near by, and find that these are not wet. The water truck hypothesis
explains why only the streets are wet, while the rain hypothesis incorrectly predicts that the
sidewalk and the lawns also would be wet. Hence, we choose the water truck hypothesis. If, on
the other hand, we find that the sidewalk and the lawns are also wet, we choose the rain
hypothesis.

12
Which hypothesis would you choose if you found that the lawns are wet but not the sidewalk?
Why? We leave that for you to gnaw on.

2.2.2 Speculative deduction

Consider the following example of reasoning:
(20) During a total solar eclipse, the positions of stars near the sun as observed from the earth
appear farther away from the sun than their positions under normal conditions. This
puzzling phenomenon can be explained by assuming that the gravitational pull of the sun
causes the light from the stars to bend, as predicted by Einstein’s theory of gravity. Since
no other theory explains the observations, we accept Einstein’s theory as correct.
This is an instance of speculative deductive reasoning, a special type of reasoning that is
characteristic of the justification of scientific theories. Given an observation (e.g., The streets are
wet) and a “theory” that we take to be correct (e.g., Whenever it rains, the streets get wet.),
abductive reasoning provides justification for an interpretation (explanation/analysis) of the
observation within the theory. Given a set of observations, speculative deductive reasoning
provides justification for the theory itself on the basis of the observations.
Given below is another example of speculative deductive reasoning to support the theory of
continental drift in geography proposed by Alfred Wegener in 1912:
(21) The shape of the east coast of South America fits neatly with the shape of the west coast
of South Africa. Fossils of mesosaurus (small freshwater reptiles) are found only in two
different parts of the earth, namely, Brazil (South America) and Africa. Dinosaur fossils
are scattered in identical rock strata, again in the east coast of Brazil and west coast of
Africa. We can provide an explanation for the above facts if we assume that South
America and West Africa formed a single landmass at an earlier point in time, and drifted
apart subsequently (continental drift hypothesis). In the absence of a better or equally
good alternative explanation for the above observations, we accept the continental drift
hypothesis as correct.
Our last example is a speculative deductive argument to establish gravity and the law of
gravitation:
(22) Suppose we assume that
(a) there is such a thing as gravity, and
(b) it is governed by a law that gravitational attraction between any two bodies is
proportional to the product of their masses and indirectly proportional to the square of
the distance between them.
This law of gravity, combined with other laws, correctly explains the generalizations of
falling bodies on the earth as well as the generalizations of planetary motion. In the
absence of contrary evidence, and of an alternative theory, it is reasonable to conclude
that (a) and (b) are correct.
Notice that both abductive reasoning and speculative deductive reasoning are forms of evidence-
based reasoning: given what we experience or observe as the grounds (evidence), these forms of
reasoning allow us to justify a conclusion as an explanation for our experiences/observations. As
pointed out earlier, Charles Sanders Peirce who was responsible for pointing attention to the
distinction between abduction and deduction didn’t distinguish between abduction and
3
speculative deduction: he viewed both as “inference to the best explanation”.

3
Speculative deductive reasoning is used to justify scientific theories. The grounds for justification in such
justification are the data. In contrast, abductive reasoning is used to justify an interpretation within a theory. The
grounds in such justification are not only the data, but also the theory itself.

13
Notice also that the rule of abduction that we gave earlier equally fits speculative deduction. If
we wish to distinguish between the two in terms of a rule of inference, we would need to specify
that in abduction, the relevant explanatory principle we are appealing to is already established in
the particular domain, while in speculative deduction, the principle is a newly proposed one. The
rule for speculative deduction is given below, along with the revised rule of abduction:
Abduction (revised)
Given that: Theory T is true
P is the best available explanation for Q within T
and Q is true
it is reasonable to conclude that : P is true
in the absence of (i) evidence contrary to P, and
(ii) a better/equally good alternative explanation for Q within T.

Speculative Deduction
Given that: Theory T is the best available explanation for Q
and Q is true
it is reasonable to conclude that : T is true
in the absence of (i) evidence contrary to T, and
(ii) a better/equally good alternative explanation for Q.

2.3 Reasoning from Sample to Population

Most introductory textbooks on logic make a distinction between deductive and inductive
reasoning. The example in (5a) reproduced below is an example of deductive reasoning, in
contrast to (23), which is regarded as an example of inductive reasoning.
(5) a. Zeno is a spider. premise 1
All spiders have eight legs. premise 2
Therefore: Zeno has eight legs. conclusion
(23) All the spiders that we have observed so far have eight legs. premise
Therefore, in the absence of evidence to the contrary,
I conclude that all spiders have eight legs. conclusion
To see what is going on in the contrast between (5a) and (23), let us go through a thought
experiment. Imagine that you are shown a barrel of balls, and are told that all the balls in the
barrel are red. You are now instructed to close your eyes, and pick a hundred balls randomly
from the barrel, and are asked what colour the balls are. Your answer would be: red. Given that
the colour of the population of the balls in the barrel is red, it follows logically that the colour of
the balls in the sample drawn from the barrel is also red.
(24) Population to sample (P-to-S)
All the balls in this barrel are red. population

Therefore, the 100 balls in the sample that sample

I have picked from this barrel are red.
The reasoning you have used in this case is that of deduction, the relevant rule being that of
4
Classical Modus Ponens.

4
Classical Deductive Reasoning uses the rule of Classical Modus Ponens. Those who teach logic in the Departments
of philosophy and mathematics would probably object, but we may say that Probabilistic Deductive Reasoning uses

14
Now consider a different scenario. You are shown a barrel of balls, but are not told the colour of
the balls. You are instructed to pick a hundred balls randomly from the barrel. You see that the
colour of all the balls in the sample is red. You are now asked what colour the balls in the barrel
are. Your answer would be that they are red: given that all the balls in the sample drawn from
the barrel are red, it is most likely – though not completely certain – that the colour of the
population of balls in the barrel is red.
(25) Sample to population (S-to-P)
The 100 balls in the sample that sample
I have picked from this barrel are red.

Therefore, in the absence of evidence to the contrary,

I conclude that all the balls in this barrel are red. population
Most logic textbooks use the term induction to refer to the mode of reasoning illustrated in (25),
to contrast it with the deduction in (24).
Given the properties of a population (a set), P-to-S reasoning allows us to infer the properties of
a particular sample (a proper subset) of the population. Given the properties of a sample of a
population, S-to-P reasoning allows us to infer the properties the population:
P-to-S reasoning S-to-P reasoning
Population Sample

Sample Population
In terms of Venn diagrams, the distinction between the two would be as follows:
P-to-S reasoning S-to-P reasoning

We could also say that P-to-S reasoning allows us to infer the particular from the general, while
S-to-P reasoning allows us to infer the general from the particular.
P-to-S reasoning S-to-P reasoning
General Particular

Particular General
As in the case of reasoning involving Probabilistic Modus Ponens, Defeasible Modus Ponens,
and Reverse Modus Ponens, the sample to population reasoning illustrated here lacks total
certainty. Even if you take a sample of a thousand balls and discover all of them to be red, there
is still a chance that one of the balls in the barrel is white, and that you happened to not notice it.
In contrast, its deductive counterpart is totally certain: if it is true that all the balls in the barrel
are red, it cannot be the case that the balls in the sample are anything but red.

Probabilistic Modus Ponens, and Defeasible Deductive Reasoning employs Defeasible Modus Ponens. The
terminology is unimportant: what is important for us is the use of rules of inference appropriate to each context, and
an understanding of contrasts such as probabilistic vs. non-probabilistic, defeasible vs. non-defeasible, quantitative vs.
qualitative, population-to-sample vs. sample-to-population, and so on.

15
To see this more sharply, let us go back to the example of situs inversus incompletes. We know
that one in 220,000,000 human beings who have their heart on the left have their liver also on
the left. If you take two or three random samples of a thousand human beings each, chances are
that you will observe that every human who has the heart on the left has the liver on the right. If
you conclude on the basis of this evidence that every human being who has the heart on the left
has the liver on the right, you would be wrong. The risk of being wrong is unavoidable in this
mode of reasoning.
Until recently, the role model for reasoning in academia was the kind of Classical Deductive
Reasoning that the Greek mathematician Euclid introduced in his proofs for theorems, codified
by philosopher Aristotle. Inductive reasoning, lacking the certainty of deductive reasoning, was
somehow seen as inferior.5

Notice that in (24), if the premise is true, it cannot be the case that the conclusion is false. In
contrast, the conclusion in (25) (though reasonable) can be false even if the premise is true: there
may be one white ball somewhere in the barrel that might not have been included in the sample.
This possibility of error is acknowledged by the phrase in the absence of evidence to the
contrary. One way to play it safe is to modify the conclusion to reflect the uncertainty, as in
(26):
(26) All the balls in the sample of 1000 balls from this population are red.
Therefore, it is reasonable to conclude that all the 10,000 balls in this population are
most likely to be red.
The Non-probabilistic Sample-to-Population Reasoning in (25) involves the risk of the
conclusion being false. In contrast, the Probabilistic Sample-to-Population Reasoning in (26) is
cautious. From a different perspective, we could say that (25), unlike (26), employs “defeasible”
reasoning: when contrary evidence is discovered, the conclusion is “defeated” and hence
abandoned (or modified).
As in the case of Modus Ponens discussed earlier, S-to-P reasoning can be quantitative or
qualitative. Compare the quantitative reasoning in (27a) with qualitative reasoning in (27b) and
(27c):
27)
( a. 995 Martians in our representative random sample of 1,000 Martians have purple
eyes. Therefore, it is reasonable to conclude that 99.5% of Martians have purple
eyes.
b. Every Martian in our large representative random sample of Martians has purple
eyes. Therefore, it is reasonable to conclude that all Martians are likely to have
purple eyes.
c. Every Martian in our representative sample of Martians has purple eyes. Therefore,
until we find Martians with non-purple eyes, it is reasonable to conclude that all
Martians have purple eyes.
(27a) illustrates quantitative probabilistic S-to-P reasoning (statistical reasoning). (27b) is its
qualitative probabilistic counterpart. (24c) illustrates qualitative non-probabilistic S-to-P
reasoning.
The rules of probabilistic non-defeasible and non-probabilistic defeasible S-to-P reasoning can
be stated as follows:

5
Though the distinction between (24) and (25) is quite clear, an accurate definition of deductive and inductive
reasoning is far from trivial. One reason is that the term “inductive” has been used for a rag-bag of different kinds of
reasoning, equating “inductive” with everything that is outside of classical deductive, covering not only the Sample-
to-Population reasoning, but also what we have called Probabilistic Deduction, Defeasible Deduction, Abduction, and
Speculative Deduction. In this use, the concept of “inductive reasoning” is not particularly useful.

16
 Sample to population: probabilistic non-defeasible
Given that: Q is true of a large representative random sample of X
it is reasonable to conclude that Q is likely to be true of X.

Sample to population: non-probabilistic defeasible

Given that: Q is true of a sample of X

it is reasonable to conclude that Q is true of X
in the absence of evidence to the contrary.

Both P-to-S and S-to-P reasoning are used in various fields of academic inquiry. In university
courses, quantitative probabilistic P-to-S reasoning is typically taught under probability theory
in mathematics. Its qualitative counterpart is taught as probability logic in philosophy.
Quantitative probabilistic S-to-P reasoning is taught as statistics in mathematics, while its
qualitative counterpart is taught under the broad rubric of inductive reasoning in philosophy.

2.4 Analogical Reasoning

No survey of the modes of reasoning would be complete without mention of analogical
reasoning. Let us begin with an example. The English government has a law protecting animals
in experiments, requiring that animals belonging to a privileged class that feels pain should not
be operated on without anesthesia. Now, pain is an internal experience that cannot be directly
observed. How do we determine whether or not a given species of animal is able to feeling pain?
How do each of us know that other human beings are capable of experiencing pain? The strategy
is as follows. I know that under certain kinds of external stimuli (e.g., wounds, falls, burns) I feel
pain, and when I feel pain, I tend to behave in certain ways (e.g., tears, certain facial
expressions, body postures, noises). When other human beings behave in parallel ways in
response to parallel stimuli, I infer that they have the parallel internal experience of pain. This
inference is corroborated by what they say about their internal experience. It is therefore
justifiable to conclude that human beings are capable of feeling pain.
We extend the same strategy to animals, except that they cannot report their pain to us. If we
find that an animal responds to pain-inducing stimuli in ways parallel to the pain response of
human beings, we conclude that the animal is experiencing pain. The type of reasoning that
allows us to make such inferences is analogical reasoning.
Analogical reasoning involves making an inference about X based on (a) the analogy
(equivalence) between X and Y, and (b) what we know about Y. When we observe a dog
wounded in a traffic accident, for instance, we infer that the dog must be in pain, based on the
analogy between humans and dogs and what we know about humans. The structure of this piece
of reasoning can be unpacked as follows:
28)
( Dogs are analogous to humans. When humans are wounded, they experience pain.
Therefore, it is reasonable to conclude that when dogs are wounded, they experience pain.
Given below are additional examples of analogical reasoning:
29)
( a. The human heart has a structure and a function. Its structure is broadly that of a
complex piston. What it does is to pump the blood from the heart to the other parts of
the body. Now, we all agree that we cannot really understand the structure of the
heart without understanding its function. Likewise, we cannot understand the
structure of language unless we understand what it is used for.
b. We know that waves in water exhibit the property of interference. If we assume that
light travels in a vacuum analogous to waves in a medium, it would follow that light

17
 would also exhibit the property of interference. This prediction is borne out. Hence it
is reasonable to conclude that light travels as waves.
The reasoning in (29b) involves the analogy of light and waves, and hence is an example of
speculative deductive analogical reasoning. The example in (29a) involves analogical but not
speculative deductive reasoning. The central rule that sanctions analogical inferences are given
below:
Rule of analogical reasoning
Given that: x is analogous to y
and P is true of y
it is reasonable conclude that: P is true of x

To illustrate:
Given that: Dogs are analogous to humans.
and Humans experience pain when wounded.
it is reasonable to conclude that: Dogs experience pain when wounded.

Recall that the conclusions in many of the modes of reasoning discussed earlier need
caveats like “in the absence of evidence to the contrary” and “in the absence of a better or
equally good alternative”. Such caveats are characteristic of defeasible reasoning, those in
which the conclusion can be defeated by further evidence or by a better or equally good
explanation. Defeasible S-to-P reasoning, abductive reasoning, and speculative-deductive
reasoning have the property of defeasibility. So does analogical reasoning.

Exercises
1. Given below are additional examples of reasoning that contain implicit premises. Make the
reasoning explicit by supplying the missing premise(s) in each case. Check whether the
premises you have unearthed are acceptable.
a. The murder weapon has Zeno’s fingerprints on it.
Therefore, Zeno is the murderer.
b. Leda drinks whisky every day.
Therefore, Leda is immoral.
c. Dino is a prostitute.
Therefore Dino is immoral.
d. Blimpsey flunked all her exams.
Therefore Blimpsey is not intelligent.
e. On a scale of 1 to 10, Zeno has a score of 9.8 in his student feedback.
Therefore Zeno is an excellent teacher.
2. Convert the following statements into the “If… then…” format illustrated in section 2.1.1:
a. All slaidins are tall.
b. No mammal has wings.
c. Only birds have wings.
3. Represent the information in (2a) – (2c) in terms of Venn diagrams.
4. Consider the following promise from a teacher:
Anyone who wears shoes in this class will get an A for the assignment on reasoning.
You are told that Bill wears shoes but Susan doesn’t. Meena has an A, but Jacob has a B.
Your task is to determine whether the teacher’s promise has been kept with respect to these
students. You can interview the students to elicit the relevant information, but are allowed to
ask each of them only one question. Also, you have to select as few students as possible.

18
 Which students would you interview to check the truth of the proposition? What question
would you ask each one?
5. Consider the following scenarios:
You are asked to make an estimate of the density of trees (number of trees per square
kilometer) in North America.
You are asked to estimate the proportion of red and white balls in a barrel containing
ten thousand balls (without actually counting all the red and white balls in the barrel.)
What additional precautions do you need to take for the first task?
6. An argument is valid if its conclusion follows from its grounds.6 The argument below is
valid in one mode of reasoning but invalid in another. What are the two modes of reasoning?
Wherever there is fire, there is smoke. There is smoke coming from that apartment.
Therefore, in the absence of evidence to the contrary and an alternative explanation, it is
reasonable to conclude that there is fire in that apartment.
7. Specify the modes of reasoning you would use in the justification of each of these claims:
a. There is a strong correlation between IQ scores language competence.
b. A forensic specialist’s conclusion that the cause of death was strangling, on the basis
of the evidence of bruise marks around the neck.
c. The conclusion that the distinction between objects and adjuncts is legitimate, on the
basis of evidence from passives, agreement, case marking, and topicalization.
d. Jane’s three month old baby boy couldn’t possibly have told her that he loved Emily
Dickinson’s poetry.

6
Those who have taken an introductory course in philosophy would notice that our definition of validity
deviates from the definition given in logic textbooks. You will see at a later point why we choose not to
adopt the textbook definition.

19