Scientific Method

Scientific Method
Aka: (Hypothetico-Deductive Method)
Unless otherwise noted the artwork and photographs in this slide show are original and © by Burt Carter.
Permission is granted to use them for non-commercial, non-profit educational purposes provided that credit is given for their origin.
Permission is not granted for any commercial or for-profit use, including use at for-profit educational facilities.
Other copyrighted material is used by permission or under the fair use clause of the copyright law of the United States.
"Hypothetico-Deductive Method" is just a more general name for what science textbooks like to call the "scientific
method". Practicing scientists tend to think of it as a lifestyle rather than a method because it becomes second
nature when you've used it for a while and seen how well it works.
Presumably you’ve heard of this before now. Here I want to give you a familiar example of its application to show
you how the various parts work and also to remind you that it’s a good way to think even if what you’re doing isn’t
science. We will continue to a more detailed example later.
Remember -- the steps of the "method" are:
1) Observe,
2) Hypothesize,
3) Test the hypotheses, and
4) Repeat as many times as possible.
…
5) (And then, after you’ve done this many times, maybe your idea will be declared a theory and not just a
hypothesis.)
Some possible observations and hypotheses we might explore:
Perhaps you live in Genoa in 1400 and have noticed that as ships sail away from you their hulls disappear from view first , then their sails, and
finally their flags. You might have an interesting hypothesis to explain that curious fact. Testing the hypothesis would involve determining the
shape of Earth, which, of course, you cannot see from Earth. How do you proceed?
Perhaps you have accidentally dropped a brick and a piece of 2x4 on your feet and noticed that though one hurt a lot worse than the other they
hit your feet at the same time. Through the pain you might wonder how they fell at the same rate but with different force.
Or maybe you noticed that your kitty looks so much like a tiger that you’ve named it “Tiger”. You might wonder if there’s some relationship
between that fact and the fact that all your relatives say you have your mother’s eyes. You might hypothesize that kitties and tigers are related in
the same sense as you and your mother, only not nearly so closely.
Lets look at a very simple (and probably very familiar) example of the application of this
method. Consider the upper left corner of a crossword puzzle like the one
below. Remember that the actual puzzle would continue across and down.
Just as a heads-up, tests of hypotheses usually work, at least in part, by ruling out
possibilities – by proving that something is not true rather than that it is true. Without even
examining any clues we can rule out lots of possibilities for the correct words in this
puzzle. For example, "rutabaga" will not work in any of these slots because it has too
many letters. Neither will "up", for essentially the same reason.
In fact, my Random House Collegiate Dictionary has over 170,000 entries and the vast
majority of them cannot be the solution to 1-across because they have the wrong number
of letters in them.
1
2
3
4
9
11
(etc.)
17
(etc.)
Of course we don’t have to guess what the words are in a vacuum. We could just stick a random 4letter word in 1-across and then see if we could make others fit it, but that’s not really the point.
Somebody (and we’re going to assume a male somebody because it was me) wrote this puzzle and
had specific words in mind, to which he has provided clues.
You observe the clue for 1 across:
1. Common device for closing off access through a hole in a wall.
You also observe that there are four spaces for letters in the answer.
Immediately two possibilities come to mind. What are they?
1
2
3
4
9
11
17
(etc.)
23
(etc.)
We begin with multiple working hypotheses. In this case there are two possibilities we can
think of. There may be others (we don’t have the best of vocabularies, perhaps), but these are
the best we can come up with.
Door may be the better guess since the clue mentions “wall” rather than “fence”, but we
remember that saloons in western movies have gates rather than doors, that gardens at the
homes of some well-to-do people are walled, and that crossword puzzle authors like to
obfuscate …
1
9
2
3
4
D O O R
1
9
11
11
17
17
23
2
3
4
G A T E
23
How do we decide which word to choose? In other words, how do we test our
hypotheses?
Hypotheses are tested in one general way: you must consider what other things must be true if the
hypothesis is correct and then find (or predict) new observations that will either verify that those
other things are true or show that they are not true.
If your predictions or new observations turn out to be true then the hypothesis is supported, but
not proven. If they turn out wrong then the hypothesis has been falsified.
The notion of falsifiability is a powerful one, and really stands at the base of scientific thought. As
one of the earliest geological writers put it:
"The state of the phenomena [bending of layers of rock] is a matter of perfect indifference to such a theory as this
[crystallization causing the bending]: all things are explained by it with the fame facility; the straight and the
crooked, the square and the round, the moveable and the immovable. [...] It should never be forgotten, that a
theory which accounts for any thing, and a theory which accounts for nothing, stand precisely on the same
footing, and ought to be banished from all parts of philosophy ..." (John Playfair, 1802, p. 233.)
In our example the tests, the predictions, in other words, are very straightforward because the
writer has insured that they exist. In the real world it is not so easy, and is often impossible, at
least until some new technique for "seeing" things has been invented. Even if we can propose tests
we might not be able to actually make the necessary observations.
So, what is the test that we propose in solving a crossword puzzle?
The test, of course, is to predict that words exist 1) that mean what the words that connect to our hypothetical
ones are supposed to mean, 2) that start with the correct letters to join our hypothetical words, and 3) that all
have the correct number of letters to fit into the available spaces.
Notice that the requirement that the words connect will cascade as we go along. The words don't just have to
connect to 1-across, but also to 9-across, 11-across, and so on. I have gone back many steps later and
corrected hypothetical words because of the lack of interconnectedness. In other words, I have held
hypotheses for quite some time before falsifying and rejecting them.
In this case we begin by seeing if 1-down will let us choose between our two hypothetical words. The clue is:
DOWN:
1. Sad.
Unfortunately this doesn't help. What can I say? That’s life.
1
9
2
3
4
D O O R
O
W
N
1
9
2
3
4
D O O R
O
U
R
1
9
11
11
11
17
17
17
23
23
2
3
4
G A T E
L
U
M
23
1
9
2
3
4
D O O R
O
U
R
1
9
2
3
4
D O O R
O
U
R
1
9
11
11
11
17
17
17
23
23
2
3
4
G A T E
L
U
M
23
So we move on. There are plenty of words which we can try to fit to our hypothesis. Let's try 9
across first:
9. To beat.
I'm stumped. Anybody? O suggests nothing at all to me. L could be the beginning of "lash", which
is promising, but "lash what"? Also, if lash is correct then 3-down begins "ts ... ", which is not
impossible, but certainly obscure. Ok, there's more than one way to skin a cat. Let's try 2down. The clue is:
2. Hank
Right away we might remember that "hank" means a coil of rope, but that doesn't suggest anything
to me. You? It could also be a nickname for someone named "Henry". Any
ideas? Georgians? Baseball fans?
There is an obvious alternative for someone who lives in Georgia – a very famous name for “Hank” that starts
with “Aa”. So let’s try that and see where it goes.
We are now favoring one of our hypotheses for 1 across, but we will not throw the other one out until we get a
little farther along.
Preferred hypothesis
so far:
G A T E
L A
U R
M O
N
1
2
3
4
Not yet fully rejected hypothesis:
9
11
1
17
23
9
2
3
4
D O O R
O
W
N
1
9
3
4
D O O R
O
U
R
11
11
17
17
23
2
23
Let’s try 11 across. It’s short and we have half the letters – the first half, which is usually the best part to have
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
1
9
11
17
DOWN
1. Sad
2. Hank
3. Stomach
4. _____ Dagi (Extinct Turkish Volcano)
G
L
U
M
2
3
4
A T E
A
R
O
N
23
Assuming we’re on the right track so far we need an acid that begins “ur__”. There’s an obvious one that
you make every time if you eat dried beans or drink beer, though you probably don’t notice having done it.
Any ideas?
And here it is. Now let’s see if we can make any sense of 9 across by solving the downs,
starting with 3.
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach (in the grocery store)
4. _____ Dagi (Extinct Turkish Volcano)
Who knows?
1
9
11
17
G
L
U
M
2
3
4
A T E
A
R I C
O
N
23
Have any of you ever looked at this stuff in the store and wondered who eats enough to
make it marketable?
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach (in the grocery store)
4. _____ Dagi (Extinct Turkish Volcano)
1
9
11
17
G
L
U
M
2
A
A
R
O
N
23
3
4
T E
R
I C
P
E
Let's slow down a bit and realize where we have gotten. At this point there is only a tiny probability that "gate" is wrong. It simply
fits too well with too many other words that also fit the puzzle. Maybe there are other words that mean the right things, fit
together the right way, and have the right number of letters, but the probability is what is generally called “vanishingly small”. Now
we can all but reject the alternative hypothesis "door“ for 1-across. Wise thinkers don't ever really forget a hypothesis, but this one
we’ve built is the clear favorite. Notice that the exactly same argument can be made for all the other words we’ve already placed -they all fit too well, never mind in what order we filled them in.
This looks more and more like a sound theory about the structure of this puzzle, and not just a guess or hypothesis. With each new
word we put in we become more confident we have the correct theory of this puzzle.
Moving on we find our second stumper. Any Turks in the class? Geographers? Volcanologists?
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach (in the grocery store)
4. _____ Dagi (Extinct Turkish Volcano)
1
9
11
17
G
L
U
M
2
A
A
R
O
N
23
3
4
T E
R
I C
P
E
Let’s try 17 across.
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach
4. _____ Dagi (Extinct Turkish Volcano)
1
9
11
17
G
L
U
M
2
A
A
R
O
N
23
3
4
T E
R
I C
P
E
We might immediately think of “moaning” but it has the wrong number of letters. Any other hypotheses?
Note that the clue ends in “ing”. Maybe the word does too. If so, it’s a piece of cake because there are only
three more letters.
That works. It makes a perfectly good word that means what the clue asks for.
Let’s move on to 23 across. This fellow was very famous. He was known in France as “the
bravest of the brave” and was very close to Napoleon early on. Then Napoleon had him
guillotined.
History majors?
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach
4. _____ Dagi (Extinct Turkish Volcano)
1
9
11
17
G
L
U
M
2
A
A
R
O
N
23
3
4
T E
R
I C
P I N G
E
His name was “Ney”. Seven done, two to go. Lets go back to 9 across. We now have the first three
letters, which should help. But I’m still at a loss.
When doing a crossword, if I’m stumped but have three or more letters in order in a word (and
particularly if they’re the first three) I sometimes go to the dictionary and try to scout them out. In
my Random House College Dictionary there is less than a page of words starting “lar___”, so I scan
until I find:
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach
4. _____ Dagi (Extinct Turkish Volcano)
1
9
11
17
G
L
U
M
2
A
A
R
O
N
23
3
T
R
I
P
E
4
E
C
I N G
Y
“larrup” – to beat or thrash.
(Ironically, I suppose, I don't ever remember seeing this word in a puzzle until a few days after I wrote
this one, and then there it was ... And then a few puzzles later who did I find but Marechal Ney?)
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
1
9
11
17
DOWN
1. Sad
2. Hank
3. Stomach
4. _____ Dagi (Extinct Turkish Volcano)
G
L
U
M
2
A
A
R
O
N
23
3
T
R
I
P
E
4
E
R U P
C
I N G
Y
At this point we can be virtually certain that we have the right words in the right places. Our testing has given us great confidence
in our hypothesis as the correct one. Each word, each observation, is not now a hypothesis but a piece of a theory. The only thing
than can screw it up, and it’s almost certain that nothing at all can, is to find a blank into which we cannot fit a word. So before we
finish this off, how many of you have even a smigeon of a doubt that there is a volcano in Turkey named "Erciyxx Dagi"? I think
you can be absolutely sure there is.
Now we just need that Turkish volcano, but we actually have quite a lot of the start of it, so our cheat might work
again. We really shouldn’t expect any English word to start with “Erciy__” (and my spell-checker/correct-as-yougo just supported my contention).
The Random House Dictionary has no word at all beginning like this, but the American Heritage Dictionary does.
And so we fill in the last of this part of the puzzle.
Another approach to finding the word is to start typing in google and let it recommend words. It finds Erciyes
after only “Erci”, and informs me that it is a stratovolcano in Turkey. Voila. (Except for the rest of the puzzle.)
ACROSS
1. Device for closing off access through a hole in a wall.
9. To beat or thrash.
11. ____ acid
17. Displaying sadness or self-pity
23. Michel ___ (First Marechal of France under Napoleon)
DOWN
1. Sad
2. Hank
3. Stomach
4. _____ Dagi (Extinct Turkish Volcano)
1
9
11
17
G
L
U
M
2
A
A
R
O
N
23
3
T
R
I
P
E
4
E
R U P
C
I N G
Y
E
S
A purist might suggest that we go back and try to find words that would go with "door" (and "down" and "dour"). What do you think
the probability is that we'd find them? Zero, but go ahead if you want.
Once a theory has been tested beyond a certain point, the probability that it is incorrect is small, and each new test makes it
exponentially smaller. At the same time the probability that competing theories are correct gets closer and closer to zero as well.
So, the probability that we’ll find two things in the universe that do not exert a calculable
gravitational force on each other (theory of universal gravitation) , or that Earth’s
continents are not moving (theory of continental drift), or that we should routinely find
quartz and pyroxene together in the same igneous rock (can anybody say “Bowen’s Theory
of Reaction Series”?), or that species have not evolved over time (theory of organic
evolution) is similarly small.
The theories of universal gravitation, of continental drift, of magmatic behavior, and the
most basic version of organic evolution (that things actually have evolved) are scientific
theories, not cockamamie theories. All the clues – the observable facts connected to the
phenomena the theories address – are just the right size, spelled the right way, and
connect together perfectly.
The probabilities that these theories (and others) are wrong is so close to zero that we can
assume it is zero. We “know” these things as surely as it is possible to know a thing, and in
the true sense of the word “know”.
Chan and McGarrett “knew” that the man whose death they investigated had committed
suicide in the same way. If some future observation were to call that conclusion into
question, then the theory could certainly be revisited, but nobody is holding their breaths
waiting for such an observation.
RETURN TO INTRODUCTORY MATERIAL
RETURN TO DR. CARTER’S HOMEPAGE