Induction

June 7, 2000

1347

Induction and deduction are the two methods for forming logical conclusions based on premises. (Premises are statements believed to be true, and they include reasons, evidence, and observations.)

It is often said that the difference between the two methods is that induction involves the movement from particular to general while deduction moves from general to particular. This is a misleading oversimplification. Here is a better way to understand these two methods. Induction forms conclusions that reach beyond the premises (or evidence), beyond the current boundaries of knowledge, thus making inductive conclusions probable rather than certain. At the heart of inductive thinking is the “inductive leap,” the stretch of imagination that draws a reasonable inference from the available information. Because inductive conclusions are only probable, they can exist along a range of probability and can be made stronger or weaker, as we will see below.

Deduction, on the other hand, operates by discovering the necessary implications of established truths; that is, established generalizations are applied either to other generalizations or to specific cases in order to discover new conclusions that necessarily follow. In a deductive argument, if the premises are true and if the argument is valid (argued in accord with the rules), then the conclusion must also be true. The syllogism is a typical way that deductive arguments are structured.

Inductive arguments

Inductive arguments can be made in several ways. These include the following:

1. Inductive generalization. This involves forming a generalization based on a collection of evidence. There are two subtypes:

a. Sample to Group. Here an examination of a set of particular cases leads to the conclusion that what is true of a sample is true of the whole group. For example, after examining a sample of twenty kinds of apples, I discover that they all have seeds, so I conclude that all apples probably have seeds. Those who categorize information and who look for similarities, and who group things together use this type of thinking. As you can imagine, this is a principal kind of thinking in the sciences and social sciences.

b. Evidence to Conclusion. Here a collection of specific evidence points toward a single conclusion. For example, I walk into a parking garage and look at a particular car and discover that the hood of the car is warm, there is wet mud on the tires, and muddy tracks lead in from outside where it is now raining. These individual pieces of evidence, working together, allow me to conclude that this car was driven recently. Police detectives, accident investigators, literary scholars, historians, and all those looking for the meaning of miscellaneous data, effects, or causes, use this kind of thinking. Other examples might include the source of disease, the guilt or innocence of the accused, the influence of early music on a composer. In term papers and articles and many other arguments the inductive generalization is presented first, as the thesis, and then backed by the specific proofs: but this is simply a matter of form or order, and the piece is still an inductive argument. (For example, in the Declaration of Independence, Jefferson first accuses King George III of being a tyrant–the thesis–and then lists more than two dozen evidences in support of this thesis or conclusion.)

2. Pattern to Prediction. Several examples are analyzed to determine what they have in common so that a prediction about the next example can be made. For example, if chemical pollution begins to show up in water wells in a pattern that shows the pollution spreading, say south east, then a prediction about contamination of future wells in that direction can be made.

3. Correlation to Cause. Here a set of correlations is examined and the conclusion made that one thing causes another. For example, I know that blasting occurred at the rock quarry across the highway on six particular dates. When I returned home each evening on those dates, I noticed that my pictures were askew and needed to be straightened. On dates when there was no blasting, the pictures remained square on the walls (negative correlation). Conclusion: blasting in the rock quarry is jarring the pictures out of square. (Arguing from correlation to cause is one of the most difficult of all inductive leaps to make because correlations can also be caused by coincidence, or an underlying common cause for both correlates. There can even be a reverse relationship: instead of A causing B, B may be causing A.)

4. General to General. This often involves an argument based on consistency (consistency is one of the tests of truth). Example: The Hondota car company has been making good cars for a long time. So we argue thusly: All Hondota Circlet cars are high quality, and all Hondota Roundelle cars are high quality, so it is likely that all of the new Hondota Orbit cars will be high quality, too. (Note that this seems to resemble a syllogism, but it argues beyond the premises, making an inductive leap.)

5. General to Particular. Many times a particular conclusion is drawn from a sample or from knowledge that is generally true. Unlike a deductive conclusion, this conclusion, while likely, is not certain. For example, my family and friends have owned six Hondota Circlets in the past ten years and the cars have all been durable and high quality. It is probable, then, that this Circlet that I want to buy will also be durable and high quality.

This kind of induction is often done with statistical data. For example, we know that 70 percent of all laser printers in American offices were made by Hewlett Packard, so it is fairly probable that the laser printer in Ted Jones’ office was made by HP. And it is also probable that most of the laser printers at MegaCorp were made by HP.

6. Particular to Particular. This is sometimes called an analogical induction or reasoning a pari. For example, this muriatic acid cleaned the rust from this steel pipe so it will probably clean the rust from that pipe as well.

Inductive Conclusions
Inductive Generalization	Sample to Group
Evidence to Conclusion	Pattern to Prediction
General to Particular	General to General
Particular to Particular	Correlation to Cause

Since inductive conclusions possess probability, it follows that some are more probable than others and that the conclusions can be made more probable. The strength of an inductive argument can be increased in the following ways:

Add more premises, more evidence, more reasons, more examples.
Add or use stronger premises, premises that are more precise, more directly related, more convincing.
Claim a more modest conclusion. Reduce the level of generality claimed (for example, from “most” to “some”) or reduce the level of probability claimed (for example, from “highly probable” to “somewhat probable”). “The smaller the generalization, the stronger the argument.”

In sum, then, inductive arguments and their conclusions

reach beyond their premises
exist along a range of probability rather than being certain
can be strengthened or weakened by new evidence

Very persuasive and high quality inductions can be formulated by careful use of the assembled evidence, by avoiding the common pitfalls of improper procedure, and by drawing reasonable generalizations from the information available. Since many inductions are based on samples, it follows that the quality of the sample will bear a direct relationship to the quality of the induction. The sample (or group of evidence or facts) must be sufficient to support the generalization, be representative of the whole set, and contain no exceptions unless allowed for.

Sometimes it is difficult to tell whether the facts are representative or sufficient, and this is again why inductive conclusions must remain probable rather than necessary, and why we qualify them with expressions like “tends,” “probable,” “some,” or “generally,” and usually avoid such terms as “always,” “none,” “all,” and “never” (which are common in deductive argument). The law also recognizes the nature of induction by empaneling juries to weigh evidence and judge whether it does indeed establish a conclusion of guilt or innocence, or in the case of a civil suit, of legitimate claim.

Because inductive thinking is so common, the dangers of improper or careless induction are especially important to know. In fact, several of the material fallacies identify unwarranted inductive conclusions.

The risks of induction might be summarized as follows:

1. Too much generalization. The generalization can be no wider than the scope of the sample. Thus to examine ten thousand plants and conclude, “All living things have roots,” would be too broad a generalization. The circumstances of time, frequency, concentration, situation, and so forth can quickly alter the validity of conclusions derived under different conditions. Judgment must be exercised when drawing conclusions from evidence, to make sure that the generalizations are reasonable. And remember, “The bigger the generalization, the weaker the argument.”2. Flimsy or inaccurate evidence or evidence selected in a biased way; too little evidence or incomplete evidence. The evidence must be direct, and not merely circumstantial or partial; and it must be representative, whole, true, and systematic rather than chance or haphazard or prejudiced. It must not have been altered or distorted.

3. Misinterpreted evidence or evidence which has no relation to the conclusion. Each fact adduced to establish a conclusion must really be a proof or support of the conclusion and not merely a peripheral fact. For example, the fact that the defendant lives in a high crime area has nothing to do with his guilt or innocence.

4. An overlooked exception. There is always the chance that a sample will not include a statistically small but highly significant factor. So until time and experience have tested an inductive generalization (and even then), it must remain more or less “probable” instead of “absolute,” and more or less “usual” instead of “always.”

Fallacies of Induction

Several logical fallacies are associated with improperly formed inductive conclusions. These fallacies include the following:

The Fallacy of Accident (Dicto Simpliciter). The main form of this fallacy is committed when a general rule is applied to an exception. Almost all generalizations and many rules have exceptions or circumstances under which they do not apply. Generalizations should be thought of as just that–generally true–and not absolutely applicable to every case.

You would agree that taking what does not belong to you is an illegal act, wouldn’t you? Well, then, you committed an illegal act because you took a brochure from that rack marked “Take One” and the brochure didn’t belong to you.
Every person has a right to his own property, doesn’t he? Well, this enraged, drunken man wants the keys to his truck so he can “get even” with that bar owner.
C. [Rabshakeh said,] Beware lest Hezekiah misleads you, saying, “The Lord will deliver us.” Has anyone of the gods of the nations delivered his land from the hand of the king of Assyria? Where are the gods of Hamath and Arpad? Where are the gods of Sepharvaim? And when have they delivered Samaria from my hand? Who among all the gods of the lands have delivered their land from my hand, that the Lord should deliver Jerusalem from my hand? –Isaiah 36:18-20
The members of the committee are all reasonable men, so their proposal must be reasonable, too. How could it be otherwise?

Even generalizations we sometimes think of as fairly absolute have exceptions. We all have a Constitutional guarantee of free speech, but we are still prohibited from inciting a riot or yelling “fire” in a crowded theater. And we are legally required to stop at red lights, except when someone directing traffic waves us through.

Similarly, the fallacy of accident can be committed by insisting that a particular group member (belonging to the minority exceptions) must necessarily conform to the generalization applicable to the group. As much as we may be tempted to pigeonhole people, we must resist denying the diversity and variety of any group. Watch out for assertions that do this, especially those using “all,” “no,” “every,” or “any.”

Today’s freshmen cannot write very well. Joe is a freshman, so he must be a poor writer.
As we know, Baptists are Calvinists. Ted is a Baptist, so he must be a Calvinist.

Another form of the fallacy of accident takes the form of a sweeping or unqualified generalization. We usually refer to an unqualified generalization as suppressed quantification (see the discussion in Chapter 11), while a sweeping generalization we call a dicto simpliciter. In either case, the generalization may have some truth in it, but there are several or many exceptions. Often the assertion is that some characteristic shared by the majority of a group must be shared by all the members of the group–basically a shift from “most” to “all.”

The trust we put in the special people in our lives is always betrayed.
The classical music albums you can get for ninety-nine cents are just no good.
Every time there is an acquittal because of an illegal search or an improperly taken confession, some cop either didn’t know what he was doing or arrogantly thought that he could break the law himself. –Seymour Wishman

The problem in these examples lies in their attempt to be all inclusive. Is our trust always betrayed? Are no bargain classical albums any good? Certainly there are exceptions. And look at the last example, where we have the assertion of police culpability “every time” a search or confession is thrown out. Many times the legality of a search or the propriety of a confession will be unclear or even appear perfectly legitimate, only to be struck down by an appellate court. And the fact that appellate courts frequently disagree with each other as the case proceeds through the system shows that the problem is not always a simple one of the cop being either ignorant or criminal.

Hasty Generalization or Selected Instances (Secundum Quid). As its name implies, this fallacy arises when a generalization is formed on the basis of too few instances or too little evidence. A reliable general truth or rule cannot be formed after observing one or two or three samples, because they may be exceptions rather than representatives of the whole group.

This Venus fly trap and this pitcher plant both eat insects, so all plants must be carnivorous.
All Fords are junk. I know because I owned one once and it was junk.
Did you ever notice that all musicians have personality problems? All six of the ones I’ve met were really strange.
Generally speaking, most magazines cover movie star gossip, as you can see from the selection here in the store.

Hasty generalization occurs whenever a conclusion is drawn about a group or class on the basis of a sample that is either too small or non-representative; most of the time, as in the examples above, the fallacy is the result of sampling error or carelessness–someone generalizes from the first one or two random examples that come to his attention.

Sometimes, however, the examples are known to be non-representative but are selected on purpose because they back the arguer’s conclusion. This deliberate form of hasty generalization is called selected instances by the logicians and stacking the deck by the propaganda analysts. It is a widely used technique, usually involving a biased selection of facts for or against a position or person. In a word, it is a one-sided argument.

Well, of course snakes are dangerous. Look at the rattlesnake, the cobra, the water moccasin. Need I go on?
How can anyone vote for this bill? It will cost $100 million to implement; it will require more bureaucracy and a bigger state payroll to administer; it will interfere with the rights of the grocery industry; and it will be more trouble for the already hassled consumer. There’s nothing good about it.
I don’t think we want any Italians in our neighborhood. Italians are all crooks. Look at the Mafia and Al Capone. Even that famous Italian philosopher, Machiavelli, recommended killing one’s opponents.

A valid generalization must be drawn from a statistically significant number of accurately described, pertinent representatives of the class or group to be generalized about. Even then the generalization must be handled carefully in order to avoid the fallacy of accident.

A variety of this fallacy combines the hasty generalization with an equivocation between “some” and “all.” The argument exploits the fallacy of suppressed quantification (see the discussion under Ambiguity) to shift from an implied “some” to an implied “all.”

At the turn of the century businesses were guilty of putting sand in sugar, water in milk, and lard in butter. The purity of canned goods was always suspect. Such practices necessitated the passage of the Food and Drug Act. This proves that businesses cannot be trusted–they are crooked.
Rapes are committed by men, robberies are committed by men, murders are committed by men. Benedict Arnold and Adolf Hitler were men. Men are liars, seducers, and cheaters. Therefore, men ought to be caged up, because men are terrible.
College students in Southern California take dope, get drunk, sleep around, commit crimes, and don’t care about anybody but themselves. So don’t ask me for a donation; college students are just worthless punks.

We see the effects of this some-to-all shifting in many of the common prejudices we find: dumb blondes, women drivers, greedy doctors, reckless teenagers, wealthy snobs, absent-minded professors, computer nerds, and so on.

Some historians are occasionally guilty of this form of hasty generalization because in writing history they select both interesting details and details which they think epitomize a time or situation but do not always distinguish between “interesting but not typical” and “typical.” Thus, what the historian intended as a single isolated detail may appear to the reader to be a typical example of something common.

A Tale of Hasty Generalization

A farmer of Sung saw a rabbit dash into a tree trunk standing in the middle of his field. The rabbit broke its neck and died. From that day, the farmer left his plowing and kept watch by the tree trunk in hopes of getting another rabbit. The farmer never got another rabbit, but he did become the laughingstock of Sung. –Han Fei Tzu, tr. Moss Roberts

And a Quotation

I had . . . come to an entirely erroneous conclusion, which shows, my dear Watson, how dangerous it always is to reason from insufficient data. –Sherlock Holmes in “The Speckled Band”

False Cause. Those who commit this fallacy mistakenly assume that because two events are connected with each other, one must have caused the other. The most common form of false cause is post hoc, ergo propter hoc (“after this, therefore because of this”). The post hoc fallacy equates the time relationship with a cause and effect relationship, arguing that because one thing happened before another thing, the first thing must necessarily have caused the second thing.

I ate Wheaties cereal all my life and then I became an Olympic champion. See what Wheaties did for me?
Everyone who has ever gotten divorced has fallen in love and gotten married first, so love and marriage cause divorce.
Last night I drank a cup of hot milk and then couldn’t sleep. That proves that hot milk causes insomnia.

Obviously, a precursor event need have no relationship at all to a successor, and in fact, most events in the past do not have a causal relationship to most events following them. Nevertheless, causal events do precede their effects, and that is what makes this fallacy so attractive. It is reasonable to argue cause and effect:

I hit this ant with a hammer and then the ant died. Therefore, hitting the ant with the hammer must have caused its death.

The problem arises in cases where the cause and effect are arguable or not certain. In such cases careful thought and investigation are needed, coupled with a qualified opinion:

The governor signed the bill after Judy talked to him. So Judy’s influence must have been instrumental. [It may not be possible to resolve an issue like this one.]
The corporation improved its profitability after that infusion of new capital, so getting the extra money must have really helped. [More evidence is needed to argue beyond “might have” or “probably.”]

In these last two cases, it would be a post hoc to insist upon the conclusions without additional evidence or investigation. The conclusions are possible, and perhaps even probable, but the mere before/after relationship in each case is not strong enough by itself to establish the point.

A variety of the post hoc fallacy wrongly attributes to an earlier thing an influence on or imitation by a latter thing. But previous historical similarities do not automatically indicate influence, either in practices or in ideas. So, even though the influence of the earlier on the latter is a major factor in civilization, we still must exercise our judgments. Take, for example, the argument, “Harvey saws wood the same way old man Sanders did. Harvey must therefore have copied Sanders’ style.” This might be true, of course, but such a conclusion, based solely on the time relationship, is not logical because there are other equally plausible possibilities for explaining the facts:

1. Coincidence. Coincidences do happen, and are much more common than we might care to believe.2. Necessity. There may be only a certain number of ways to do something, and certain ways (or solutions or ideas) may have a greater tendency to suggest themselves than others.

3. Distant Common Source. Both the attributed influence and the alleged copy may really both be copies of an earlier source or influence.

What other possibilities for similarity do you see for the situations described in these arguments?

This tribe has a legend about a fisherman catching so many fish his boat sank. The tribe must have been influenced by the ideas of the earlier Musha tribe, which has a very similar legend.
In the two years following Pinetree’s election, the nation’s economy improved and crime went down. Pinetree must be a really effective congressman to have brought this about.

Post hoc can be combined with hasty generalization, too:

An owl flew by our house on three different occasions, and each time something bad happened later on. That proves that owls cause evil events.

Another variety of false cause is cum hoc, ergo propter hoc (“with this, therefore because of this”). This fallacy is committed by mistaking coincidence or simultaneity with cause–it is the claim that because two things correlate (are found together), that one of them must have caused the other. A parallel existence or relationship is not necessarily one of cause and effect. In fact, there are several possibilities for a relationship between two things found together. If A and B correlate, this may mean any of these:

A caused B
B caused A
A and B exist coincidentally and there is no relationship between them at all
A and B were both caused by a previous X
A and B are necessarily conjoined (and always occur together.)
A and B have some relationship of influence, but it is moderate

Some examples of this fallacy are these:

He drives a Volkswagen and has lots of girlfriends. Volkswagens must attract girls.
He died in bed, so being in bed must have killed him.
Successful men wear ties by Pierre.
There is an inverse correlation between a manufacturer’s alphabetical order and the number of cars it sells: Chrysler sells fewer than Ford which sells fewer than General Motors. That shows that people prefer to buy cars from companies further down the alphabet.

Whenever a correlation is presented as describing a cause-and-effect relationship, we must always apply our judgment of the situation to the claim to see whether the relationship is believable (probable) or whether a cum hoc is perhaps being committed. Take this example:

When the wind blows, sales of chain saws increase. In fact there is a direct correlation between the strength of the wind and sales of chain saws, so blowing wind must cause the sales.

This may actually be probable if we think that blowing wind tends to break branches off trees, and that the branches need to be cut up. But what about this one:

Sid Swift wore Gopher Skin Running Shoes. And he won the marathon. Get the winner; be a winner–with Gopher Skin Running Shoes.

Some relationship is possible, since bad shoes could hinder performance, but the major force of the appeal is a cum hoc. You cannot win the marathon just by buying a certain pair of shoes (sorry to have to tell you that). Any good running shoe would most likely permit you to achieve your best performance (though occasionally a marathoner will run barefoot). The evidence of mere correlation is not enough. Suppose the advertisement had claimed, “Sid Swift wore an Accutime wristwatch and won the marathon,” or “Sid Swift wore a bone in his nose and won the marathon.” Would you be persuaded to imitate him? (And what if the runner in last place also wore Gopher Skin Running Shoes?)

A major source of post hoc and cum hoc mistakes lies in the complex nature of cause itself. Many effects do not have a clear, single cause, but are the result of a combination of causes and influences. And many times we cannot say with certainty what caused an event. (For a fuller discussion on the different kinds of cause and the problems with discovering them, see the fallacy of causal reduction.)

Cause and Effect

An obviously disturbed gentleman came into a psychiatrist’s office, snapping his fingers and looking around him apprehensively.
“Calm down,” said the doctor. “What are you doing that for?”
“It keeps away the elephants,” the man replied.
“But there aren’t any elephants around here,” observed the doctor.
“You see,” the man said. “It works.”

Conversation between Friends

“Hey, what are you doing?”
“Dyeing my hair gray, of course.”
“But whatever for?”
“My computer shows that there is a direct correlation between gray hair and financial success. All the big tycoons have gray hair.”
“Aw, that’s crazy.”
“Hey, computers don’t lie man. Besides, you can’t argue with statistics.”
“A direct correlation, huh?”
“That’s right.”
“Say, can I borrow that stuff when you’re done?”

Logical Proof?

A tiger caught a fox. The fox said, “You wouldn’t dare eat me! The gods in Heaven have made me the leader of all animals. It would be a violation of the gods’ mandate for you to make a meal of me. If you doubt it, let me walk in front, and you follow to see if any animal dares stand his ground.” The tiger consented and went with the fox, nose to heels. Every animal that saw them fled. Amazed, and agreeing that the fox was leader of all the animals, the tiger went on his way. –Chan Kuo Ts’e, tr. Moss Roberts

Causal Reduction. This fallacy is committed by reducing a complex, interacting set of causes into a single, simple cause or into a few simple causes. The cause named for an effect may be the wrong cause altogether (that is, it may have had nothing at all to do with the effect); it may be a generalization of such proportions as to be practically meaningless (as in “Poor study habits are caused by non-positive reactions to learning”); or it may be a partial or aiding cause substituting for a sufficient and exclusive cause.

To clarify this last possibility, we must examine the three basic kinds of causes. A sufficient cause can effect a result all by itself, without help from other causes. Heavy rain is a sufficient cause for wet streets. But a sufficient cause may be accompanied by additional causes (which are then redundant): in addition to the rain, a leaking fire hydrant may be a redundant additional cause of wet streets locally. Being out of gas is a sufficient cause for a car not starting, but bad sparkplugs or a weak battery may be redundant additional causes. So even though we can identify with confidence a sufficient cause, we must still be careful about asserting it as the only cause.

A necessary cause, though not able alone to bring about an effect, must always be present for the effect to occur. A necessary cause is an essential one of several causes of an effect. Sunlight is a necessary cause for plant growth, but sunlight alone is not sufficient cause. Water and nutrients are also necessary causes. To point out a necessary cause and to assert that it is the cause for an event is thus inaccurate and oversimplified–and commits the fallacy of causal reduction.

A contributory cause is one of the several causes which together will produce an effect, but the contributory cause by itself is neither sufficient nor necessary. Seeing an advertisement for records may have contributed to your purchase of Music for the Royal Fireworks, but the advertisement was perhaps not sufficient (you had to have other reasons) nor necessary (in the sense that you would not buy the record until you saw the advertisement, because you may have already planned to buy the record, and the advertisement just jogged your memory). A contributory cause is sometimes a trigger cause–the final event that sets off an effect. But since the event could be set off by any one of several different trigger causes, a specific trigger cause cannot be said to be a necessary cause.

In the following examples of causal reduction, determine what kind of cause has been substituted for the real cause or causes. What might be the real cause or causes?

The adolescent flock urge caused the campus unrest of the 1960’s.
The assassination of Francis Ferdinand in 1914 caused World War I.
The presence of oxygen in the atmosphere caused my house to burn down.
Slurpo coffee outsells every other brand, obviously because it tastes best.

Most effects have several or even many causes. The usual problem in our thinking comes when we ignore this and begin a search for the cause of something. Perhaps part of the difficulty lies in our impatience. When we ask, “Why did that happen?” or “What caused that?” we are unwilling to sit still for a lengthy explanation of thirty contributing factors. We want a three-word answer so that we can file it away and rush off to our next task.

When the idea of causation is viewed as a search for a single agent (that is, only one cause is allowable), arguments are often made that not only fail to take into account the reality of multiple causation, but also fail to distinguish between actor and agent. Note these examples:

Guns don’t kill people; people kill people.
The cause of the crime is the person, not the weapon.

What really “causes” crime? A particular crime might be caused by a need for drugs, a need for money to buy the drugs, the pusher who originated the drug habit, the availability of a weapon or tool (such as a gun or crowbar) with which to facilitate or even make possible the crime, the availability of a victim, a lack of personal moral restraint by the criminal, a lack of fear of punishment by him, peer pressure (gang related or otherwise), and a lack of a value-rich upbringing. All of these can be at least contributory causes to a given crime. The absence of one of these causes may or may not have prevented that crime. Without a gun, for example, the criminal might have committed burglary instead of robbery, and if so, in that instance having a gun would have been a necessary, though not a sufficient cause of a robbery.

The assertion, “Guns don’t kill people; people kill people,” confuses the actor (or perhaps moral cause) with the agent (or efficient cause). While the guilt for a shooting murder certainly rests with the person using the weapon, the gun itself is clearly the agent of death–and a not entirely neutral one either, because it allows, or even tempts, the passions to accelerate to ultimate force; that is, the gun simplifies deadly power into a quick and easy squeeze. In the absence of a gun, a crime might still have been committed, but it might have been altered–for example, into a black eye or a stab wound rather than a fatal gunshot wound. Thus the tool or agent affects the character or quality of the crime, as well as its ease.

Notice the failure to distinguish between actor and agent in these arguments, which also insist on single-agent causation:

Autos don’t cause wrecks; people cause wrecks.
Whiskey doesn’t cause drunk driving; people cause drunk driving.
Wrenches don’t tighten bolts; mechanics tighten bolts.
Dynamite doesn’t blow up bridges; demolition engineers blow up bridges.

Non Sequitur (“It does not follow”). As its name implies, a non sequitur is a conclusion or statement which does not logically follow from a previously given reason, premise, or generalization. The conclusion cannot be drawn from the given statement either because there is no logical relationship between the statement and the conclusion or because too many steps in the reasoning process have been left out and the reader cannot perceive any relationship between premise and conclusion.

The statement, “I am six feet tall; therefore, tea is cheaper than coffee,” is a non sequitur of the first kind, where no logical relationship exists between the two statements. Some other examples are:

I don’t know why he would get cancer, what with all the money he’s got.
You like spinach? That shows what a fascist you are.

An example of the second kind would be, “All this rain we are now getting will make the drought worse,” where too many steps have been left out to allow the reader to understand the logical relationship. (The argument with its steps included is: The rain is tropical and therefore warm; warm rain melts the snowpack; melted snowpack gives diminished summer runoff; diminished summer runoff will worsen the drought.)

To avoid this latter kind of non sequitur a writer or arguer should take care to reconstruct his arguments to include all essential steps in the reasoning process. The reader does not know what a writer is thinking; only the stated connections between thoughts and the steps which appear on paper are obvious to the reader.

See whether you can discover the thinking behind these:

If God had made the world, there would not be any winter.
The more working mothers we have, the more crime we will have.
He just got a job so I guess he will be married soon.

Review

Terms and Concepts

induction
inductive argument
fallacy of accident (dicto simpliciter)
hasty generalization
false cause (post hoc and cum hoc)
causal reduction
non sequitur

Questions

1. What are six kinds of inductive conclusion?

2. Summarize the dangers of induction.

Test Yourself

For each question, choose the single best answer.

Frillith County California Pickle Consumption
Year	Pickles Consumed (Tons)	Crimes Committed
1960	40.2	2902
1970	63.7	5049
1980	84.0	7214
1990	111.4	10861

The evidence presented by this chart proves that eating pickles causes crime.