Testing a Hypothesis

By Michael Edmonds 26/01/2014 2


Human beings are extremely good at spotting patterns in our environment. This has served us well in terms of survival. For example, being able to identify which colours and shapes of berry are edible or toxic, or which fresh animal poo is that of a predator or of prey would have helped our ancestors make appropriate decisions regarding their survival.

However, our brains seem to be so primed to see patterns, sometimes we perceive them when they aren’t really there. For example, “seeing faces” in geological features, toasted sandwiches or a starry night sky. Or where patterns do exist, we may attribute them to the wrong thing.

And One of the things science help us to do is decide when an apparent pattern is due to a cause and effect, a correlation* or whether there is actually no pattern at all.

In science if we think we have observed a pattern, we might form a hypothesis and then test it, to see if it is true (or at least supported by the evidence).

Say for example, your cat is always waiting for you on the door step when you get home. A workmate suggests that this is because cats are telepathic and can sense when you are about to arrive home. How might you test this claim (or hypothesis in scientific terms)?

First, you would check that your observation is actually correct – is your cat always there when you get home? Often we unconsciously “cherry pick” data to fit an idea. So a first step might be to record if your cat is really there every day when you get home.

Assuming you do this for several weeks and find it to be the case then you could start thinking about alternative hypotheses that might also explain the cat being there when you get home.

How many alternative explanations/hypotheses can you think of?

.

.

.

.

.

.

.

.

.

.

.

.

 

Here are three that I came up with

1) You have a lazy cat who sits there all day waiting for you

2) The cat knows what time you arrive home and turns up just before you get home

3) The cat recognises the sound of you car when you turn into your street and turns up just before you get home

So how would you test these?

You might put a video camera near where you cat sits and see if she does sit there all day. Or maybe you could go high tech and get a collar which tracks where the cat goes each day. This could confirm or eliminate hypothesis 1

To test hypothesis 2 you could turn up at different times of the day to see if the cat turns up.

Hypothesis 3 could be tested by using a different car.

If the evidence supports any of these three hypotheses then you now have a more likely explanation than a telepathic cat. However, if the evidence does not support them, this does not automatically mean you have a telepathic cat – perhaps there is another explanation you have not yet thought of? Forming more hypotheses and testing them, is the only way to see  if the original “telepathic cat” is the best explanation.

Many people think that science is about gathering together all of the information that fits your “favourite” hypothesis, but this is only part of what scientists do. As you can see from the example above, science is also about looking for alternative explanations (hypotheses) and seeing if these are better supported by the evidence. In essence, the only way to “prove” a hypothesis is by trying to disprove it. If , after trying to disprove it, it still matches the data then it is a stronger hypothesis. If it does not then the hypothesis needs to be rethought.

 

*A correlation is where there is a connection between two events but not a cause and effect.

For example, there is a correlation between ice cream consumption and sunburn because both increase in summer. This does not mean that ice cream causes sunburn (i.e it is not a cause and effect relationship).


2 Responses to “Testing a Hypothesis”

  • You seem to be confusing patterns with their causes. You start by saying that we are good at seeing patterns, but sometimes we see them when they aren’t there! Then you seem to say that a pattern is only really there if there is a cause and effect relationship. I would say that this article is really about attribution of the wrong causes to patterns (e.g. telepathic cats), not about seeing patterns that aren’t really there.

  • Stephen Thorpe

    “Then you seem to say that a pattern is only really there if there is a cause and effect relationship.”

    Ah, no I didn’t.

    Though I do take your point about overlapping two different points about perceiving patterns when they aren’t there and
    attributing the cause of a pattern incorrectly.

    And a couple of points might have been clearer if I had used the word perceive instead of see, I’ll adjust that and highlight it so readers know what I have changed – altered sections shown in italics.