Chapter 4: Entailment
Learning Objectives
After reading this chapter, you should understand:
● L.O. 4.1: what entailment is, and how to evaluate arguments for deductive validity;
● L.O. 4.2: the nature and limits of logical form, as well as how to recognize certain valid
sentential and predicate forms; and
● L.O. 4.3: why we often overlook validity and invalidity in assessing inferences.
Introduction
As we have seen, an argument has the highest degree of suppositional strength when we know that its
premises entail its conclusion. If we learn that the premises not only entail the conclusion but also are
true, then the conclusion is inescapable.
However, it's not always obvious whether an argument's premises entail its conclusion. In this chapter, we
will consider some methods and tests that can help us identify entailment, and investigate some common
reasons why entailment gets misdiagnosed.
4.1 Deductive Validity
As we saw in the last chapter, we say that some premises entail a conclusion when they meet this
condition: if the premises were true, then the conclusion would also have to be true.
Also, we saw that an argument in which the premises are presented as entailing the conclusion is called a
deductive argument. And an argument in which the premises actually do entail the conclusion (rather than
just being presented as if they do) is called deductively valid.
So, in a deductively valid argument, the truth of the premises guarantees the truth of the conclusion. It's
important to stress, though, that the kind of guarantee we are talking about is an absolute guarantee, not
just strong support. For example, consider these two arguments:
Every raven ever encountered by anyone has been black.
So, the next raven we see will be black.
Only 1 in a million ravens is not black.
So, the next raven we see will be black.
Neither argument is deductively valid. These arguments may have great suppositional strength, but in
neither case does the premise absolutely guarantee the truth of the conclusion. In both cases, it could
happen that the premise is true and the conclusion false. That may be very unlikely, but it's still possible.
When premises entail a conclusion, there is no possible situation in which the premises are true and the
conclusion false. So, in evaluating whether premises entail a conclusion, we don't take for granted
anything else we happen to know about the world. Instead, we pretend that all we know about the world is
that the premises are true. For example, "Some cars are red" does not entail "Some people drive red cars"
because that would mean assuming that people drive red cars. It would be a very strange world if there
were red cars but nobody ever drove any of them—but that situation is not impossible.
Often, of course, when someone presents an argument, they do want us to take for granted something they
take to be an obvious truth about the world. In that case, it's not helpful to point out that the premises they
actually spoke out loud don't entail the conclusion. Instead, we should treat them as having intended an
additional implicit premise; the resulting argument may be deductively valid after all. And this outcome is
important to realize, because we may really disagree about the truth of the implicit premise, not whether
the explicit premises entail the conclusion.
Step by Step
If an argument's premises entail its conclusion and its premises are true, then the conclusion is
inescapable. But some hard thinking may be needed to realize that the conclusion is inescapable, because
it is sometimes hard to tell whether the premises are true, or whether they entail the conclusion.
One kind of case where it's difficult to see that the premises entail the conclusion is when the argument is
very complex. Many "logic puzzles," including those on some standardized tests, are like this. For
example, consider this argument:
Figure 4.1 [1]
Some monkeys are cute.
Only friendly animals are eating veggies.
No friendly animals are cute.
So, some monkeys are not eating veggies.
For most of us, some effort is required to work out that this example is a case of entailment. We can do
this by actively imagining that all the premises are true, which requires holding several different claims in
mind at the same time, and then seeing if they guarantee the conclusion.
With a complex argument, it's often too hard to hold all the premises in mind at once in order to suppose
that they are true. So, it's often easier to take the argument in several steps. In the monkey argument, for
example, we can take the premises two at a time and see if any pair guarantees an interim conclusion. In
particular, take the first and third premises.
Some monkeys are cute.
No friendly animals are cute.
From these two premises, we can conclude that some monkeys are not friendly animals. We can now try
using this interim conclusion along with the second premise, and seeing what they entail together:
Some monkeys are not friendly animals.
Only friendly animals are eating veggies.
It's now easier to see that these premises entail the conclusion, namely that some monkeys are not eating
veggies. Even if it took us multiple steps to get here, we now know that the premises of the original
argument do entail its conclusion.
Flipping the Argument
In addition to taking an argument step-by-step, it sometimes helps to try flipping the [Link] means
starting with the conclusion instead of the premises.
When we consider an argument in the forward direction, we suppose all of its premises to be true, and ask
whether the conclusion must then be true. When we flip the argument, we suppose its conclusion to be
false, and ask whether it's possible for all the premises to be true. If they can still be true even though the
conclusion is false, they don't really entail the conclusion.
The key point to understand here is that the following two features of an argument amount to the same
thing:
● If all the premises were true, the conclusion would have to be true.
● If the conclusion were false, the premises could not all be true.
If you don't see why these amount to the same thing, it's worth taking a minute to think about them until
you do.
Figure 4.2 [2]
Imagine a single premise that entails a conclusion—for example, This horse is a mare entails This horse is
female. We know that if a horse is a mare, then that horse must be female. But that also means that if the
horse is not female, the horse is not a mare. If the conclusion is false, the premises can't all be true—and
in this case, there's just one premise.
Flipping the argument can be useful when the argument is hard to evaluate in the forward direction. It can
also be useful as a way of double-checking for entailment. Let's consider a simple example:
All mosses are plants that Bob loves.
Some plants that Bob loves are ferns.
So, some ferns are mosses.
Let's try flipping the argument. Suppose the conclusion is false. That means no ferns are mosses. Could
both premises still be true? The premises say that all mosses—and some ferns—are plants that Bob loves.
But that could be true even if none of the ferns are mosses. (After all, the premises don't say that all the
plants Bob loves are mosses!) So the premises could be true, even if the conclusion were false. But that
means the premises don't entail the conclusion.
4.2 Logical Form
It would be nice if there were an easy way to test for entailment, so we wouldn't have to think hard about
whether the premises could be true and the conclusion false. Can't we just identify entailment by the
general shape of an argument?
Well, I have some good news and some bad news. The good news is there are some easy tests that can
sometimes identify entailment. These involve checking whether an argument has a certain kind of
form—if so, we can be certain that any argument that fits the form has premises that entail its conclusion.
The bad news is that this won't help us identify every case of entailment. In fact, there is no easy test for
entailment that will identify every case.
Argument Recipes
Let's start by focusing on the good news. When we say that two arguments have the same form, we mean
they both follow a kind of general recipe for constructing arguments. For example, consider these two
arguments:
Figure 4.3 [3]
If mice can talk, then I'm a wizard.
Mice can talk.
So, I'm a wizard.
If the moon is up, the birds are singing.
The moon is up.
So, the birds are singing.
In both arguments, the premises entail the conclusion. But they have something else important in
common—in both, the entailment has to do with the sentential connective if...then.... A sentential
connective combines two sentences to form a larger one; and in the case of if...then..., we call the larger
sentence that is formed a conditional. In a conditional, the sentence that immediately follows if is called
the antecedent, and the sentence that immediately follows then is called the consequent.
The mice argument and the moon argument both contain a conditional premise, but they also fit a
particular recipe for constructing arguments, which we can describe as follows:
● Components:
○ Any two sentences
○ The sentential connective if... then...
● Method:
○ For the first premise, link the two sentences with the connective to form a conditional.
○ For the second premise, use the antecedent of the conditional.
○ For the conclusion, use the consequent of the conditional.
This recipe allows for any two sentences to be combined this way to form an argument. And any
argument that follows this particular recipe will share certain general features, which we can illustrate by
using the variables P and Q to stand for the two sentences:
If P then Q
P
_________
Q
This schema of an argument illustrates its logical form, and it is really just a simple way of presenting the
recipe given above. Any argument we can create by substituting two sentences for P and Q has this
logical form.
This particular logical form is called modus ponens. And it has an important feature: regardless of which
two sentences we substitute for P and Q, the premises will entail the conclusion. We can see this just by
knowing the positions of those two sentences and understanding what the words if and then mean. So, we
can tell that any argument that has the form of modus ponens will be deductively valid.
Consider a case where we don't even understand the sentences that substitute for P and Q. Here's an
argument with the form of modus ponens using real but seldom-known English words:
Figure 4.4 [4]
If the road is anfractuous, I shall divagate.
The road is anfractuous.
So, I shall divagate.
We don't need to know what the words anfractuous and divagate mean in order to see that the premises
entail the conclusion—as long as the premises aren't nonsense. All we really need to know is that the
argument is made from two meaningful sentences combined in the right way using if, then, and so. If it
has the form of modus ponens, it's deductively valid.
When we have a logical form like modus ponens—one that guarantees deductive validity—we call it a
deductively valid form. As we will see, there are many deductively valid forms, all of which can help us
identify entailment.
Before looking at other deductively valid logical forms, though, let's return briefly to the bad news. Even
though logical forms can help us identify many deductively valid arguments, they can't help us identify all
of them. And that's because some arguments are deductively valid even though they lack a deductively
valid form. That is, sometimes premises entail their conclusion, even when they don't have any form that
is a general and reliable recipe for entailment. We'll look at some examples once we've become more
familiar with various logical forms. The point for now is that logical forms can often be used to identify a
deductively valid argument, but can never be used to prove that an argument is not deductively valid.
Some Deductively Valid Sentential Forms
Now let's look at some other deductively valid forms. The next one uses two conditionals in a row, linked
by a single sentence that is the consequent of one premise and the antecedent of the other. This one is
called hypothetical syllogism.
If P then Q
If Q then R
_________
If P then R
This is pretty straightforward, and we can see why it has to be a deductively valid form. If both premises
are true, then we can see that if P were also true, we would have to conclude R—simply by applying
modus ponens twice. The conclusion captures this fact about the two premises: knowing that they are true
would allow us to conclude R from P.
Here's an argument with the form of a hypothetical syllogism:
If Bob studies hard, he'll do well on the test.
If Bob does well on the test, he'll be smiling.
So, if Bob studies hard, he'll be smiling.
Note that the conditionals we're using in this chapter must be unqualified in order for their forms to be
deductively valid. They can't be hedged with probably, as in, for example, If P then probably Q, or even If
P then almost certainly Q. This is especially clear in the case of hypothetical syllogism. It could be that if
P, there is a greater than 50 percent chance of Q; and if Q, there is a greater than 50 percent chance of R;
and yet it's not true that if P, there is a greater than 50 percent chance that R. (Maybe P is 'it's raining
here,' Q is 'it's raining one town over,' and R is 'it's raining two towns over.') As we'll see in a later chapter,
this is due to the way that probabilities combine. Despite this, if an argument genuinely has the form of a
hypothetical syllogism—which requires that its conditionals are not hedged—we can be sure that if all the
premises are true, the conclusion is also true.
Next, here is a deductively valid logical form called modus tollens:
If P then Q
Not Q
_________
Not P
Like the recipe for modus ponens, this one calls for any two sentences and the sentential connective if...
then.... But it also calls for not, which is shorthand in this logical form for any way of forming the
negation of another sentence.
Figure 4.5 [5]
The negation of a sentence is true when the original sentence is false, and false when the original sentence
is true. Sometimes the negation of a sentence can simply be formed by putting the word notin front. For
example, the negation of the sentence All poppies are red can be formed simply by putting not in front:
Not all poppies are red.
But sometimes negating a sentence doesn't work that way. For example, to form the negation of some
poppies are red, we can't just put not in front. Not some poppies are redis not even grammatical! Instead,
we have No poppies are red or There are no poppies that are red, or All poppies are not red. (But be very
careful about that last construction, because in informal English, it is syntactically ambiguous between
meaning that no poppies are red and meaning that not all poppies are red—i.e., that some poppies are not
red.)
When the logical form of modus tollens calls for Not P, it's really calling for any way of negating P. So,
for example, here's an argument with the form of modus tollens:
If the poppies are in bloom, Spring is over.
Spring is not over.
So, the poppies are not in bloom.
In this example, not is used to form the negation of P and Q, but not just by putting it in front of the whole
sentence. And if you give it enough thought, you'll see that the argument's premises entail its conclusion.
Indeed, any argument with the form of modus tollens will entail its conclusion: so, modus tollens is a
deductively valid logical form.
So far, we've seen logical forms involving conditionals, but our next form involves disjunctions. The
disjunction of two sentences P and Q is a sentence that is true as long as either P is true or Q is true, and
false only when both P and Q are false. Sometimes we form the disjunction of two sentences by joining
them with either... or..., but disjunctions can be formed in other ways. For example, to form a disjunction
from Bob danced and Alice danced, we could say Either Bob danced or Alice danced, but we could also
use the simpler sentence Bob or Alice danced. Either way, the disjunction is true as long as at least one of
them danced, and false only if neither danced. (Depending on how you say it, this disjunction can also
suggest that they didn't both dance; but let's ignore that and assume that P or Q is true when P and Q are
both true.)1
A common deductively valid logical form using disjunction is illustrated by the following argument:
Figure 4.6 [6]
The sun is either rising or setting.
The sun is not rising.
So, the sun is setting.
You can tell that the argument above entails its conclusion.
But another way to identify the entailment is to notice that it has the deductively valid form known as
disjunctive [Link] form of disjunctive syllogism is this:
Either P or Q.
Not P.
_________
Q
Note that this also works if the negated premise is Q and the conclusion is P: if either component sentence
in the disjunction is false, you can conclude that the other is true.
Some Deductively Valid Predicate Forms
So far, we've only looked at deductively valid logical forms containing variables that take the place of
entire sentences. But consider this argument:
All dogs are mammals.
Fido is a dog.
So, Fido is a mammal.
The premises entail the conclusion. In fact, any argument formed by replacing the words dog, mammal,
and Fido with other words in the same grammatical category would also entail its conclusion. For
example:
All toasters are time-travel devices.
Fred is a toaster.
So, Fred is a time-travel device.
Figure 4.7 [7]
Of course, this argument is a weird one: the first premise is obviously false, and who names their toasters?
Still, if the premises were true, the conclusion would have to be true.
The two arguments we just considered entail their conclusions for the same reason: they seem to share a
form. But if we replace all the non-compound sentences in these arguments with variables, the form we
get is:
P
Q
______
R
This form is absolutely not deductively valid. So, to illustrate the structural feature that the two arguments
above share, we can replace parts of each sentence with variables. In particular, we'll hold fixed the words
they share and replace the common nouns with variables, using F, G, H... rather than P, Q, R... in order to
clarify that they are not substituting for whole sentences. And for any proper name, we'll use a lowercase
n. This gives us the following deductively valid logical form:
All F are G
n is an F
_________
n is a G
Any argument formed by replacing F and G with common nouns, and n with a proper name, will be one
in which the premises entail the conclusion.
But wait: what happens when we're not just using common nouns? For example, All dogs are mammals
contains two nouns, but what if the first premise had been All dogs bark or All dogs are happy? Logicians
solve this by recasting those sentences slightly to fit the mold:
All dogs bark. ⇒ All dogs are barking things.
All dogs are happy. ⇒ All dogs are happy things.
Figure 4.8 [8]
This way, barking thing and happy thingcount as substitutions for G in the deductively valid form given
above. In other words, we can treat them as predicates in our logical form. (This is why logical forms that
use variables for predicates and names are called predicate forms, and logical forms that use variables for
entire sentences are called sentential forms.)
There are many deductively valid predicate forms, and it's worth looking at some of them and making
sure you understand why they're deductively valid. (Two important points for how these forms are used:
as used in these forms, some F are G means at least one F is G, and does not guarantee that some F are
also not G. Also, every name is assumed to refer to a real object.)
Figure 4.9
Long Text Description
These are all deductively valid logical forms, and there are many others. We won't be giving them names
in this text, because it's unrealistic to memorize a long list of deductively valid logical forms in the hopes
of recognizing arguments with those forms in everyday life. If we do encounter arguments with these
forms, it's easier just to assess whether the premises guarantee the conclusion!
However, it is useful to acquire the skill of quickly identifying entailment. And one aspect of that skill is
being able to notice when an argument has a deductively valid form. The best way to acquire this ability
is not to memorize a list of deductively valid forms: it's to repeatedly engage with arguments and logical
forms until you begin noticing more easily which forms are deductively valid—that is, to practice.
The forms above are listed so that you can spend some time trying to understand why each is deductively
valid. (In some cases, it may be easier to understand why a form is deductively valid if you flip it: assume
the conclusion is false, then see why the premises could not then be true.) It's good practice.
Counterexamples
If we are evaluating the claim All dogs bark, all that we need to show that the claim is false is to find a
dog that doesn't bark. More generally, when faced with a universal claim—one that says about a class of
things that they all have some feature—all it takes for that claim to be false is for one of the things in that
class not to have that feature. We call this a counterexample. If Fido is a dog that doesn't bark, then Fido
is a counterexample to the claim that all dogs bark.
So, how is this relevant to logical forms? If a logical form is deductively valid, then every instance of it
with true premises has a true conclusion. So, all it takes to show that a logical form is notdeductively
valid is a single instance in which the premises are true and the conclusion false. That would be a
counterexample to the claim that the logical form is deductively valid. For example, consider this logical
form:
All F are G.
Some G are H.
_________
Some F are H.
Is this form deductively valid? One way to assess this question is to see if we can come up with a
counterexample to it. If so, then it's not deductively valid. If we can't—well, we might be missing
something. Our inability to come up with a counterexample doesn't prove the absence of
counterexamples. Luckily, in this case, our job is easy. We can find a simple example of an argument that
is an instance of this form in which the premises are true and the conclusion false:
Figure 4.10 [9]
All dogs are mammals.
Some mammals are wombats.
_________
Some dogs are wombats.
Clearly, the premises are true and the conclusion false. So this argument is not deductively valid. And it's
an instance of the logical form above. Any time we can provide a counterexample like this, it
demonstrates that the logical form is not deductively valid.
We can do this with sentential forms as well. Consider:
If P and Q, then R.
Not R
_________
Not P
Is this a valid deductive form? Well, let's try to come up with an instance in which the premises are true
and the conclusion false. It's not very difficult to do:
If Fido is a dog and Fido is good, then Fido gets treats.
Fido does not get treats.
_________
Fido is not a dog.
Something has gone badly wrong here: the premises can be true even if the conclusion is false (e.g., in the
case where Fido is a bad dog and doesn't get treats). So, the argument is not deductively valid. But it's an
instance of the form above, so it's a counterexample to the deductive validity of the form.
The Limits of Logical Form
Let's recap this section. While seeking an easy way to identify entailment, we noticed that some
arguments share structural features that guarantee entailment—namely, deductively valid logical forms.
We then considered two types of deductively valid logical forms: some that replace entire sentences with
variables, and some that replace only predicates and names with variables.
We can now be clearer about the "good news" and the "bad news" mentioned at the start of this section.
The good news is that deductively valid forms can often help us identify entailment. And it's possible,
with practice, to get good at recognizing deductively valid logical forms. The bad news is that not every
case of entailment involves an argument with a deductively valid form. Consider, for example, these two
arguments:
Bob was killed.
So, Bob died.
Alice is older than I am.
So, I am younger than Alice is.
In both cases, the premise entails the conclusion, but neither argument seems to be an instance of a
general reliable recipe for deductive validity. In other words, they don't have deductively valid forms—at
least not forms that can be constructed using the two methods of generating logical forms that we've
encountered.2
So, if you are sure that an argument has a deductively valid form, you can be sure that the premises entail
the conclusion. But that test will miss any deductively valid arguments that don't have deductively valid
forms.
4.3 Pitfalls in Evaluation
In this section, we'll mainly be considering cases where we think an argument's premises entail its
conclusion, even though they don't. But first, let's consider the opposite kind of case, in which we think an
argument's premises do not entail its conclusion even though they do.
Overlooking Deductive Validity
Let's consider three reasons why entailment might get overlooked. First, we might check to see if the
argument has a deductively valid logical form, and if it doesn't, conclude that its premises don't entail its
conclusion. As we just saw at the end of the previous section, though, that would be a mistake.
Second, if an argument has false or poorly supported premises, we might be tempted to assume that it
can't be good in any other respect either. The key here is to remember that there are two separate elements
of a good argument: how well the premises link to the conclusion, and how well-supported the premises
are. Entailment only concerns the first of these, so even obviously false premises can entail a conclusion.
Third, sometimes arguments are just complex or tricky. It can even be hard to recognize an instance of
one of the simple deductive forms we've considered. (For example, people have a much easier time with
modus ponens than with modus tollens.)3 It helps a lot to simply slow down. If an argument is complex,
take it step by step. If it's unclear whether the premises guarantee the conclusion, try flipping it, as
described above. And if you have a hard time recognizing whether a particular type of argument has a
deductively valid form, try repeatedly thinking through arguments of that kind until the pattern sinks in.
Biased Evaluation
Let's turn now to cases where we mistakenly take an argument to entail its conclusion. One reason for this
is that our old friend biased evaluation can apply to deductive arguments.
Figure 4.11 [10]
For example, is the form of this argument deductively valid?
All blueberries are brightly colored.
Some healthy things are brightly colored.
So, some blueberries are healthy.
It's easy to be fooled. We already accept the conclusion, which in turn makes us inclined to find every
aspect of the argument good, even though we're consciously aware that entailment is unrelated to the truth
of the conclusion.
If we look closely, though, it should be clear that the premises don't actually guarantee the conclusion.
(They allow for the possibility that none of the healthy and brightly colored things are blueberries.)
Interestingly, when the same form of argument is presented with nonsense predicates, people are far less
likely to think the premises entail the conclusion:4
All snorgs are approbines.
Some jamtops are approbines.
So, some snorgs are jamtops.
With no subconscious tug telling us that the argument is good because its conclusion is true, we can now
assess whether the premises actually entail the conclusion. Using nonsense predicates forces us to
decouple our evaluation of the argument from our assessment of its conclusion—and the same is true for
the evaluation of logical forms, since they abstract away from any actual conclusions.
Of course, it becomes more obvious that this form of argument is not deductively valid when we consider
this example:
All blueberries are brightly colored.
Some oranges are brightly colored.
So, some blueberries are oranges.
Not only is it possible for the premises to be true and the conclusion false, it's actually the case here! So
this is a clear counterexample to the deductive validity of the form.
Some Deductively Invalid Forms
Despite this, logical forms can sometimes seem deductively valid when they're not. Many of the errors in
reasoning that are traditionally called "fallacies" involve cases where we get fooled into thinking that a
logical form is deductively valid. Let's consider a few well-known examples.
The logical form known as affirming the consequent can be deceiving, because it looks a lot like modus
ponens: in fact, it's modus ponens in reverse, which is not a deductively valid form:
Figure 4.12
Figure 4.13 [11]
Here is an example that might seem good at first:
If this plant is a fern, it has spores.
It has spores.
So, this plant is a fern.
But this argument actually has true premises and a false conclusion. As it happens, ferns are not the only
plants that have spores. Mosses do too, for example! So, just from these premises, we can't conclude that
the plant is a fern.
Another well-known deductively invalid form is denying the antecedent, which is like modus tollens in
reverse:
Figure 4.14
Long Text Description
An example would be:
If this plant is a fern, it has spores.
It's not a fern.
So, it doesn't have spores.
It's true that ferns have spores, but (for example) mosses have spores too. So, if the plant at hand turns out
to be a moss, the premises will be true and the conclusion false. The argument form is not deductively
valid.