The Null Hypothesis… Why?
(Trilogy: Part One)
“Null hypotheses may be boring, but they're very useful.”
After class, Colum walked straight up to his psychology teacher, Helen, with a puzzled look on his face.
“Sorry, Helen… I still don’t understand,” he said. “Why do psychologists test the null hypothesis instead of the actual hypothesis they care about?”
Helen smiled gently, as if she had heard this question many times before. “That’s a very good question,” she replied. “The funny thing is that we do care about the real hypothesis — the alternative — but we test it indirectly.”
“Indirectly?” Colum asked, frowning. “How?”
“Well,” Helen said, “the null hypothesis gives us a very precise and clear prediction because it describes what usually happens. For example, if it rains and we’re outside, we expect to get wet. That’s the normal situation — nothing unusual, nothing unexpected.”
Colum crossed his arms. “Okay… but why can’t we just measure the hypothesis itself?”
Helen chuckled softly. “Because the alternative hypothesis — the one we actually care about — is usually far too vague. Let me give you a simple example.”
She picked up a pen from her desk, as if acting out a small scene. “Imagine the claim I just mentioned: if it rains, people get wet. That’s a clean, testable prediction. We can easily measure two things: whether it rained, and how wet people got. Simple.”
“Right,” Colum said, nodding. “That’s straightforward.”
“Now imagine the opposite claim,” Helen continued, “which would be our hypothesis: if it rains, people won’t get wet. Okay… but why wouldn’t they get wet? Maybe they used an umbrella, or wore a waterproof coat, or stood under a roof, or ran very fast," she couldn't hide an ironic smile about her own words, "...or the rain was light, or the wind blew the drops away. You see? There are endless possibilities.”
Colum raised an eyebrow. “So… what would we even measure then?”
“That's the point,” Helen replied. “So, the alternative doesn’t give us one clear thing to test. But the null hypothesis does.”
She paused briefly, then went on. “So then, here’s what we do: we test the precise prediction made by the null hypothesis. If the data don’t fit that prediction — if it rains and people somehow don’t get wet — then the null hypothesis becomes unlikely. And when the null fails, our hypothesis succeeds.”
Colum’s expression finally brightened. “Ah… so when the null hypothesis fails, the real hypothesis becomes the more believable explanation!”
“Exactly,” Helen said. “We support the alternative hypothesis indirectly, by showing that the null hypothesis’s prediction doesn’t match reality.”
Colum sighed with relief and smiled. “Okay, now I get why psychologists focus so much on the null hypothesis. Thanks, Helen.”
“Any time,” Helen said with a wink. “Null hypotheses may be boring, but they’re very useful.”