Science is based on observations which lead to a hypothesis. People do this all of the time - any explanation
or prediction based upon something that you've observed can be a hypothesis. However, good science
depends upon quality observations that can lead to a testable hypothesis.
1. What are various, fundamentally different ways that observations can be made? List at least three.
2. What are the features a proper scientific
hypothesis should have?
3. The class will come up with three spontaneous observations. For each, briefly describe a hypothesis linked
to each and a simple test for that hypothesis.
1. What is a variable?
2. Given the basic definition of an experimental variable, what difficulties
arise when designing an experiment for it?
Classically, a control test is a duplicate of your experiment with the variable removed; differences in results
between the two tests are assumed to come from the effect of the variable. However, only in fairly simple
experiments does it work like that, because exact duplication can be almost impossible to achieve when dealing
with living things. Controls in most real-world instances are set to account for what are known as confounding
factors - those aspects of unavoidable differences that can reasonably be expected to interfere with the clear
results of the experiment and should be duplicated between tests as much as possible. It may take several
different control tests to address these. For each of the examples given below, briefly explain what factors could
affect the results and need controls. We'll discuss possible control tests, but you don't have to write these all
1. You want to test a new type of eyedrops to relieve eyestrain. Your test runs in a large office building - you
get the word around that anyone with eyestrain should come to see you. The drops will be given in your office,
with the subjects' immediate reactions recorded and a follow-up call 30 minutes and 24 hours later to record
2. Your new birdseed blend, you're convinced, will draw more birds than other blends to backyard feeders.
You can set up standard feeders with your blend and record, over several days' time, how many total bird visits
and the numbers of different bird types are observed at the feeders.
3. It was recently noted that the U.S. crime rate dropped dramatically about two decades after the Supreme
Court's Roe v. Wade decision that legalized abortion. How could you determine if there is more than an
4. You heard someone say that the fastest way to freeze ice cubes is to use hot water, so hot water must freeze
faster than cold water. You are going to freeze identical pots of water, half with 100- degree temperatures and
half with 40-degree temperatures, and time how long they take to freeze to a predetermined expansion point.
5. A textile company has developed an ultra-smooth fabric that seems ideal for pillow-cases. They intend to
perform public preference tests with regular pillow-case materials to show that theirs is superior.
It's not a good experiment to set up circumstances and then "just see what happens." A scientist needs to
collect data, and before you run any type of test, you have to decide how you'll collect such data. Planning "to
take measurements" is not useful if you don't know (and state clearly, so your test will be reproducible - no one
can reproduce your results if they don't know how you measured them) how you're going to collect them. What
you'll do with the information - statistical analysis and comparisons, for instance - may be deferred until the data
is collected, although stronger science plans such processing ahead of time.
We will answer the questions individually and compare among the group. There are no wrong answers here,
unless you leave it blank, but you'll find on discussion that certain approaches and details are better than others.
1. There's a home remedy for flea infestations: "What you need to do is put a green light close to the floor over
a pan of soapy water. That's how my sister-in-law's cousin's girlfriend got rid of them." To test the attractive
powers of green versus other color lights, you set up various lights in enclosures containing a set number
of fleas. What data do you want to collect? This may involve extra apparatus.
2. If you wanted to compare the value of premium gasoline over regular, with a fleet of cars to act as both
experimental and control, what data could you collect to make the comparison?
3. When testing a headache remedy, what data addresses it effectiveness?
4. In a human preference test, how do you collect experimental and control (for false opinion) data, and how
do you fit them together at the end?
5. You are hired to rate the fishing among ten different Adirondack Lakes,
and you want to do it "scientifically." What data could lead you to a valid
6. In a study to compare "safety" between homes with guns and homes without them, what data would you
Results, or evidence of any sort, must be evaluated to decide whether it supports a hypothesis, and if so, how
strong is the support? For this part, read the passages and decide, on a scale of 1 to 10, how strongly the
evidence supports the hypothesis.
1. Smokers often cite friends and relatives who have smoked heavily through a long and healthy life to show
that smoking is not unhealthy.
2. For testing the hazards of drugs and other chemicals, mice are used as models for humans. Since mice only
live 1/25th as long as people, their daily dosage of a chemical is multiplied until it is made equivalent to almost
a month's exposure for a human, so that a lifetime burden can be built up. When cancers occur more in these
mice than controls (who get similar doses of some nontoxic material), it is assumed that long-term exposure
to that chemical will produce cancer in people.
3. Phone polls are used to track election races. Usually 1000 registered voters or more are used to calculate
what proportion of the overall voting public would support the candidates "if the election were held on that day."
4. Generally, infants look at new or unexpected things longer than old or expected things. Tests of early
number ability were run with infants. Toys were placed behind a screen one by one, or the screen would go
up and toys would be visibly removed, while out of sight, behind the screen, toys were added or subtracted.
Observations of how long infants looked at the toys after the screen went down indicated that they could do very
simple math - for instance, one toy added to one was a dull, "expected" two, but if instead one or three toys
were revealed, the infants seemed to regard this as unexpected.
5. A group of mothers in Connecticut thought that there was an unusual number of sick children in their county,
which had a long-running nuclear power plant almost in the middle. A researcher's check of medical records
found that the rate of childhood leukemia in that county was quite a bit higher than the national average rate.
These results were used to call for a shutdown of the power plant.
6. The researcher in Question 5 dug deeper. He found that for records going as far back as he could find,
including long before the plant was built, the local leukemia rate was significantly higher than the national
average. Should the plant be shut down?
7. It's common to use public opinion to support ideas. If the vast majority of people believe a particular thing,
how strong is that as evidence?
8. When scientists use computers to make predictions about the Earth's future climate, they use information
based upon what they understand and limits based upon what a computer can handle. When making
predictions about global warming, the models march forward with time jumps of about a month and weather
regions about 200 miles square - the computer assumes, in a way, that the New York weather is constant for
all of September. Any aspects of weather change not really understood, like water evaporation and cloud
formation, have to be ignored in these models. If a model predicts a 20-degree average temperature rise over
the next 20 years, how trustworthy is that?
9. It is thought that a giant asteroid-Earth collision pulverized the asteroid and some of Mexico, throwing up
huge clouds of dust and debris that changed the climate and wiped out the dinosaurs (and many others). The
first evidence was related to the fact that in the layer just above the very last dinosaur fossils, an unusual
amount of the element iridium is found, more than in other fossil layers. Iridium is uncommon in most Earth
rocks but found in certain types of asteroids.
10. An article in recent years noted that AIDS first appeared in Africa a few years after a widespread polio
vaccination program in the AIDS-infected areas. The article's writer claimed that an impure vaccine started the
11. A background check revealed that a single batch of polio vaccine went to both Africa and Poland. Polish
AIDS rates are very low, and do not date back as far as African AIDS.
12. AIDS virus, HIV-1, is thought to be a descendent of a similar virus found in apes, SIV. The genetic
sequences of the two viruses are about 95% identical.
13. Humans and chimpanzees are thought to be descended from the same ancestors. Human and
chimpanzee genetic sequences are about 95% identical.
14. A group claiming damages from silicon breast implants found that about 8% of a large group of
implant-receiving women had serious immune-system disease.
15. The silicone breast implant manufacturers pointed out that 8% of non-implant-receiving women had serious
immune-system disease, and so their implants did not cause these diseases.
Designing your own Experiments.
Now that you know the basic pieces and approaches to experimental design, you are going to apply these rules
to totally pretend experiments. We are not going to do these experiments; we are not going to actually gather
results. However, these should be experiments that could be done (money is no object) from the descriptions
you write. Once you have written your design down, your lab partner will look it over and offer criticisms - you
get to do it to theirs, too. What you are trying to include in your designs, and check for in your partner's, are...
Use the statements on the next page as a starting point. They are there to give you ideas, and also to
restrict how far afield these designs can get. You may use one starting hypothesis of your own, but make
sure that it's clear!!
State a clear, focussed-on-one-thing hypothesis. Your experiment will be very difficult to design if you try
to address multiple variables. Pick something specific for your hypothesis, using the statements as an
inspiration. Make sure that the meanings of your terms are clear.
Focus your experiment on your specific hypothesis. Your experiment should not answer more questions
than you are asking. Determine what will act here as your main variable.
Make sure your terms are clear. Someone else should be able to reproduce your experiment without it being
significantly different - if your terms are not clear, they might define them differently and do a different
experiment and get very different results.
Make sure you have included a control test. And remember that it should duplicate all aspects of the
experimental test except for the variable.
State how you intend to collect and process your results. Here, again, defining what you would measure
and how you would use that data to support your hypothesis is a very important part of any experiment's design.
A good scientist tries to keep track of everything as an experiment plays out, but they focus on those parts they
decided ahead of time would be important.
As the partner, check to see that an experimental design accomplishes all of these goals. You may note
whatever weaknesses you see, but the goals stated here are the ones you should pay closest attention to.
Use this list of items as starting points for forming your hypotheses...
It's widely "known" that dogs are color blind. What about
other animals? (How do you decide if what they are detecting is color and not
tone, or other senses are helping?)
What beneficial effects does taking Vitamin C have during the cold season? (Pick
something specific to test!)
How might Red Bull have determine that people would switch from Gator-Ade to
their product before they started producing it?
The Environmental Protection Agency wants to rip up the top three meters of soil
for just about all of Aspen, Colorado - they say the soil was contaminated with
lead from silver mining 100 years ago. The people there don't want them to do it
- how could they generate a case for saying that it's doing no harm?
Many people have opinions of how certain species of animals are more intelligent
than others. For two different types of animals, how do you design a test of
true intelligence, where physical, sensory, or behavioral differences can be
confounding factors? (It's got to be a "fair test.")
How accurate are psychics' predictions? (Critical factor: you must define
Tobacco use has steadily increased among young people for the last decade, after
being on the decline for years. What might explain this? (Beware of confounding
A study in Baltimore indicated that, contrary to widely-accepted opinion,
so-called "crack babies" had no significant, noticeable behavioral differences
from other babies. It is also widely-accepted that these babies go on to have
problems, socially and academically, later on. Do they? (How can you reliably
test for such things?)
What aspects of personal appearance have the greatest impact on a job interview?
(Remember, a good experiment is focused on a single variable!)
Is there really a connection between wearing a hat and how slowly you drive?
Is it really a bad idea to treat violent youthful offenders as youthful
offenders? (How do you define "bad" in this sort of question?)
How much of a person's intelligence is genetic and how much from upbringing?
(What would be your comparison groups?)
Some people claim that plants will grow better if you play certain types of
music for them.
Experiment Design Lab
Observations and Hypotheses.
|1. VARIABLE: Definition:
|Experiment Design challenges:|
Designing your own Experiments.
State your hypothesis or experimental question.
Briefly describe your experimental test.
Briefly describe your control test.
State how you intend to collect and process your results.
Partner's Critique: How have or haven't the requirements been met for...
The hypothesis / question.
Planning for results.