the Littleton, Colo., school shootings on 35 years of liberalism.
In a speech Wednesday to about 500 Republican women, Gingrich said elimination
of prayer and "the Creator" from schools, lack of teaching about the Constitution and a
steady stream of violence in the movies and video games have produced teen-agers who
are morally adrift. He blamed an overtaxing government for forcing parents to spend
more time working, away from their children.
But Gingrich said attempts to make guns a scapegoat for last month's shootings at
Columbine High School were "banalities." "I want to say to the elite of this country—
the elite news media, the liberal academic elite, the liberal political elite," Gingrich told
the Republican Women's Leadership Forum, "I accuse you in Littleton, and I accuse you
in Kosovo, of being afraid to talk about the mess you have made, and being afraid to
take responsibility for things you have done, and instead foisting upon the rest of us
pathetic banalities because you don't have the courage to look at the world you have
created." . . .
Gingrich was harsh in placing the blame for the murder of 12 students and a teacher
in Littleton, Colo., by two classmates. He said the killers probably never realized they
were robbing the "inalienable rights of life, liberty and the pursuit of happiness" of their victims because the schools never taught them the constitutional meaning of the words.
"For 35 years, the political and intellectual elites (and) political correctness have
undermined the core values in American history, so the young people may not know
who George Washington is, or they may not know who Abraham Lincoln is—but they
know what MTV is," Gingrich said.
Gingrich said Republicans should lead a campaign to "expose" movie and video
game makers to liability lawsuits, and to challenge "Democrats to cut off the fund-
raising" from makers of violent movies.
Chuck Raasch, Gannett News Service, USA Today, May 13, 1999
C. How to Look for the Cause
I have a waterfall in my backyard in Cedar City. The pond has a thick rubberized
plastic pond liner, and I have a pump and hose that carry water from the pond along
the rock face of a small rise to where the water spills out and runs down more rocks
with concrete between them. Last summer I noticed that the pond kept getting low
every day and had to be refilled. You don't waste water in the desert, so I figured I'd
better find out what was causing the loss of water.
I thought of all the ways the pond could be leaking: The hose that carries the
water could have a leak, the valve connections could be leaking, the pond liner could
be ripped (the dogs get into the pond to cool off in the summer), there could be
cracks in the concrete, or it could be evaporation and spray from where the water
comes out at the top of the fountain.
I had to figure out which (if any) of these was the problem. First I got someone
318 CHAPTER 15 Cause and Effect
to come in and use a high pressure spray on the waterfall to clean it. We took the
rocks out and vacuumed out the pond. Then we patched every possible spot on the
pond liner where there might be a leak.
Then we patched all the concrete on the waterfall part and water-sealed it. We
checked the valve connections and tightened them. They didn't leak. And the hose
wasn't leaking because there weren't any wet spots along its path.
Then I refilled the pond. It kept losing water at about the same rate.
It wasn't the hose, it wasn't the connections, it wasn't the pond liner, it wasn't
the concrete watercourse. So it had to be the spray and evaporation.
I reduced the flow of water so there wouldn't be so much spray. There was a
lot less water loss. The rest I figured was probably evaporation, though there might
still be small leaks.
In trying to find the cause of the water leak I was using the method scientists
often use:
Conjecture possible causes, and then by experiment eliminate them
until there is only one. Check that one:
Does it make a difference? If the purported cause is eliminated,
is there still the effect? Could there be a common cause?
Not much spray, not much water loss. I couldn't be absolutely sure, but it seemed
very likely I had isolated the cause.
The best prophylactic against making common mistakes in reasoning about
causes is experiment. Often we can't do an experiment, but we can do an imaginary
experiment. That's what we've always done in checking for validity: Imagine the
possibilities. But note: This method will help you find the cause only if you've
guessed it among the ones you're testing.
Exercises for Section C
1. Come up with a method to determine whether there's cause and effect:
a. Pressing the "Door Close" button in the elevator causes the doors to close.
b. Zoe's belching caused Spot to run away.
c. Reducing the speed limit to 55 m.p.h. saves lives.
d. The red-headed lady walking by the classroom causes Professor Zzzyzzx to arrive at
class on time.
2. Flo: Isn't it amazing that of all the houses in this town, I was born in one where the
people look so much like me!
What is Flo overlooking?
3. Dick: {Bending over, sweating and cursing) There's something wrong with my bike.
Zoe: What?
Dick: Something's going "click, click, click" all the time.
EXERCISES for Section C 319
Zoe: Must be something that's moving.
Dick: Duh. Here, hold it up while I turn the pedals, {click, click, click, . . . )
Zoe: Yup, there it is.
Dick: It must be in the pedals or the wheels.
Zoe: Stop pedaling. . . . It's gone away.
Dick: It must be in the pedals, then.
Evaluate how Dick and Zoe have tried to isolate the cause here.
Tom was asked to bring in a causal claim he made recently and evaluate it. Here's his work.
The only time I've had a really bad backache is right after I went bicycling early in
the morning when it was so cold last week. Bicycling never bothered me before.
So it must be the cold weather that caused my back to hurt after cycling.
Causal claim: The cold weather caused my back to hurt after cycling.
Cause: It was cold when I went cycling.
Effect: I got a backache.
Cause and effect true! Yes.
Cause precedes the effect! Yes.
Valid or strong! I think so.
Cause makes a difference! Sure seems so.
Common cause! None.
Evaluation: The criteria seem to be satisfied. But now I'm wondering if I haven't
overlooked some other cause. I also had an upset stomach. So maybe it was
the flu. Or maybe it was tension, since I'd had a fight with Suzy the night
before. I guess I'l have to try cycling in the cold again to find out.
Good. But you're still looking for the cause, when it may be a cause. Another possible
cause: Did you warm up first? Another possibility: You'll never know for sure.
4. Write down a causal claim that you made recently and evaluate it. Have a classmate
critique your evaluation.
5. Make up three causal claims and trade with a classmate to analyze.
6. Judge: I find that Nancy sustained serious injuries in this accident. There is suff
icient
evidence that the defendant ran a red light and broadsided her car, causing the injuries.
But I hold that Nancy was partly responsible for the severity of her injuries in that she
was not wearing a seat belt. Therefore, Nancy shall collect only 50% of the costs
associated with this accident.
Explain the judge's decision in terms of normal conditions and foreseeable consequences.
7. Mickey has taken his four-wheel-drive Jeep out into the desert to explore on this hot
sunny Sunday. But his two cousins want to see him dead. Bertha has put poison in
Mickey's five-gallon canteen. Richard, not knowing of Bertha's plans, has put a very
small hole in the canteen.
320 CHAPTER 15 Cause and Effect
Mickey's car breaks down. He's getting hot and thirsty. His cellular phone doesn't
work because he forgot to recharge it. He goes to get some water and finds the canteen
empty. .. .
Overcome by guilt later in the year, both Bertha and Richard confess. Who should
be blamed for causing Mickey's death?
D. Cause and Effect in Populations
When we say "Smoking causes lung cancer," what do we mean? If you smoke a
cigarette, you'll get cancer? If you smoke a lot of cigarettes this week, you'll get
cancer? If you smoke 20 cigarettes a day for 40 years, you'll get cancer?
It can't be any of these, since we know smokers who did all that yet didn't get
lung cancer. And the cause always has to follow the effect. So what do we mean?
Cause in populations is usually explained as meaning that given the cause,
there's a higher probability that the effect will follow than if there were not the
cause. In this example, people who smoke have a much higher probability of
getting lung cancer than non-smokers.
That's how it's explained. But really we are talking about cause and effect just
as we did before. Smoking lots of cigarettes over a long period of time will cause
(inevitably) lung cancer. The problem is that we can't state, we have no idea how to
state, nor is it likely that we'll ever be able to state the normal conditions for smoking
to cause cancer. Among other factors, there's diet, where one lives, exposure to
pollution and other carcinogens, and one's genetic inheritance. But if we knew
exactly, we'd say: "Under the conditions , smoking (number
of) cigarettes every day for years will result in lung cancer."
Since we can't specify the normal conditions, the best we can do is point
to the evidence that convinces us that smoking is a cause of lung cancer and get an
argument with a statistical conclusion: "People who continue to smoke two packs
of cigarettes per day for ten years are % more likely (with a margin of error of
%) to get lung cancer."
What kind of evidence do we use?
1. Controlled experiment: cause-to-effect
This is our best evidence. We choose 10,000 people at random and ask 5,000 of
them never to smoke and 5,000 of them to smoke 25 cigarettes every day. We have
two samples, one composed of those who are administered the cause, and one of
those who are not, the latter called the control group. We come back 20 years later
to check how many in each group got lung cancer. If a lot more of the smokers got
lung cancer, and the groups were representative of the population as a whole, and we
can see no other common thread amongst those who got lung cancer, we'd be
SECTION D Cause and Effect in Populations 321
justified in saying that smoking causes lung cancer. The point of using a control
group is to show that, at least statistically, the cause makes a difference.
But we don't do such an experiment. It would be unethical. It's not acceptable
to do an experiment on humans that has a (major) potential for doing them harm.
So we use some animals sufficiently like humans that we feel are
"expendable," perhaps rats. We fit them with little masks and have them breathe the
equivalent of 25 cigarettes per day for a few years. Then if lots of them get lung
cancer, while the ones who don't smoke are still frisky, we can conclude with
reasonable certainty that smoking causes cancer in laboratory rats.
So? We then argue that since rats are sufficiently similar to humans in their
biological processes, we can extrapolate to say that smoking can cause cancer in
humans. We argue by analogy.
2. Uncontrolled experiment: cause-to-effect
Here we take two randomly chosen, representative samples of the general population
for which we have factored out other possible causes of lung cancer, such as working
in coal mines. One of the groups is composed of people who say they never smoke.
One group, comparable to the control group for controlled experiments, is composed
of people who say they smoke. We follow the groups and 15-20 years later check
whether those who smoked got lung cancer more often. Since we think we've
accounted for other common threads, smoking is the remaining common thread that
may account for why the second group got cancer more often.
This is a cause-to-effect experiment, since we start with the suspected cause
and see if the effect follows. But it is uncontrolled: Some people may stop smoking,
some may begin, people may have quite variable diets—there may be a lot we'll
have to factor out in trying to assess whether it's smoking that causes the extra cases
of lung cancer.
3. Uncontrolled experiment: effect-to-cause
Here we look at as many people as possible who have lung cancer to see if there is
some common thread that occurs in (almost all) their lives. We factor out those who
worked in coal mines, those who lived in high pollution areas, those who drank a lot,
. . . . If it turns out that a much higher proportion of the remaining people smoked
than in the general population, we have good evidence that smoking was the cause.
This is uncontrolled because how they got to the effect was unplanned, not
within our control. And it is an effect-to-cause experiment because we start with
the effect in the population and try to account for how it got there.
How do we "factor out" other possible causes? How do we determine whether
the sample of people we are looking at is large enough to draw conclusions about the
general population? How do we determine if the sample is representative? How do
we decide how many more cases of the effect—lung cancer—have to occur before it
322 CHAPTER 15 Cause and Effect
can be attributed to some cause rather than just to chance? These are the problems
that arise whenever we generalize (Chapter 14), and only a course on statistics will
make these issues clearer.
Until you do take such a course and have access to actual write-ups of the
experiments—not just the newspaper or magazine accounts—you'll have to rely on
"the experts." If the experiment was done by a reputable group, without bias, and
what we read passes the obvious tests for a strong generalization, a good analogy,
and a good causal argument, then we can assume that the researchers know statistics
well enough to conduct proper experiments—at least until some other reputable
group challenges their results.
Example 14 Reginald smoked two packs of cigarettes each day for thirty years.
Reginald n
ow has lung cancer. Reginald's smoking caused his lung cancer.
Analysis Is it possible for Reginald to have smoked two packs of cigarettes each
day for thirty years and not get lung cancer? We can't state the normal conditions.
So we invoke the statistical relation between smoking and lung cancer to say it is
unlikely for the cause to be true and effect false.
Does the cause make a difference? Could Reginald have gotten lung cancer
even if he had not smoked? Suppose we know that Reginald wasn't a coal miner,
didn't work in a textile factory, and didn't live in a city with a very polluted
atmosphere—all conditions that are associated with a higher probability of getting
lung cancer. Then it is possible for Reginald to have gotten lung cancer anyway,
since some people who have no other risks do get lung cancer. But it is very
unlikely, since very few of those people do.
We have no reason to believe that there is a common cause. It may be that
people with a certain biological make-up feel compelled to smoke, and that that
biological make-up also contributes to their getting lung cancer independently of
their smoking. But we have no evidence of such a biological factor.
So assuming a few normal conditions, "Reginald's smoking caused his lung
cancer" is as plausible as the strength of the statistical link between smoking and
lung cancer, and the strength of the link between not smoking and not getting lung
cancer. We must be careful, though, that we do not attribute the cause of the lung
cancer to smoking just because we haven't thought of any other cause, especially if
the statistical links aren't very strong.
Example 15 Zoe: I can't understand Melinda. She's pregnant and she's drinking.
Dick: That's all baloney. I asked my mom, and she said she drank
when she was pregnant with me. And I turned out fine.
Zoe: But think how much better you would have been if she hadn't.
Analysis Zoe doesn't say but alludes to the cause-in-population claim that drinking
during pregnancy causes birth defects or poor development of the child. That has
been demonstrated: Many cause-in-population studies have been done that show
EXERCISES for Section D 323
there is a higher incidence of birth defects and developmental problems in children
born to mothers who drink than to mothers who do not drink, and those defects and
Richard L Epstein Page 41
