science

‘Hypothesis’ is a scientific term that, whilst it may have some colloquial application, is defined as:

a supposition or proposed explanation made on the basis of limited evidence as a starting point for further investigation.

Oxford Languages

Or

an idea that proposes a tentative explanation about a phenomenon or a narrow set of phenomena observed in the natural world

https://www.britannica.com/science/scientific-hypothesis

However, this is quite a limited definition and doesn’t fully explain what a hypothesis is, the prerequisites or how it is applied.

A hypothesis is part of the scientific method, and one of the prerequisites of a hypothesis to be considered more than a simple idea is that it has to be falsifiable. For something to be falsifiable it usually needs to be testable, though that is not always the case. It doesn’t always have to be through experimentation e.g. a logical argument could falsify something, though we would generally be talking about some form of empirical testing.

This ties into another prerequisite for a scientific hypothesis. It needs the possibility of being operationalised – otherwise, it is simply conjecture.

When we operationalise something, it is essentially describing what we are looking at, how you measure the cause, how you measure the effect etc.

So how do we form this hypothesis?

We start with observations, questions and ideas. If available we would also do some background research and lean on relatable knowledge and experience.

Hypothetical

Let’s propose a world exactly like ours except it’s never had cars or similar road vehicles.

Let’s assume you’re one of those people from a ‘carless’ world and you’ve been transported to this ‘carful’ world.

You hear a noise, you ask “what’s that noise?” And see this big metal thing behind you with a person inside.

gray mercedez benz coupe
Photo by Mike B
on Pexels.com

“What is that thing?”
“How did it make that noise?”

Let’s say you moved closer to one of these things.

You might notice that the man inside seemed to have something he twists and pulls out that makes the noise stop.

It might not be too much of a stretch of the imagination to then consider putting that thing in and twisting it might make it start making that noise again.

This is something that is falsifiable so it could be considered a hypothesis, even though you can’t actually test this yourself.

car keys on white surface
Photo by Brett Jordan on Pexels.com

In fact, let’s say the man gets out of the car and drops these long things, you walk over and recognise them as keys. From your knowledge of locks and keys, you understand how this works and your suspicions are confirmed.

There are two keys on the ring which means it could be either-or. This means you can form at least 2 hypotheses.

H1. Key one when turned (in the ignition) will cause the noise to happen.
H2. Key two, when turned, will cause the noise to happen.
H3. Both keys, when turned, will cause the noise to happen.

There is also something called the null hypothesis which is when there isn’t any significant statistical difference between variables.

In this instance, the H0 would be that neither key causes the noise to happen.

A null hypothesis is never actually accepted only rejected. It exists in the realm of rejected and not-rejected. Some might refer to this as tentatively accepted, but it’s more a case of we haven’t got a working hypothesis yet rather than that being the answer. The idea needs refining and modifying and put through more tests. In the case of the car, it could be that we needed different keys, for example, perhaps you picked up house keys rather than car keys, or perhaps the car battery had died etc.

There are a few that might argue in an incredibly limited set of things where enough work has been done only the null remains, and therefore can be accepted, but in general, this is thought that it should only be tentatively done.

Watch this video on YouTube.

Just to further explain the null hypothesis slightly, H0 does not stand alone and belongs to H1-n. Without H1-n there is no H0 and it only belongs to the variables set out as measurable. Making a jump from H0 being the base or default state of an untestable H1 requires quite a bit of something that just isn’t there.

Consider a hypothesis about drinking water making you happier. They came to this idea from an observation that people who drank a lot of water seemed happier.  You could split this into a number of hypotheses about drinking 1 pint, 2 pints, 3 pints etc of water per day and seeing the effect on someone’s happiness. The H0 would be that there was no significant difference between how many pints of water someone drank and their happiness.

From here it would be a case of refining things, perhaps those that tend to drink more water also exercise more and have a better diet, so rather than water being the cause, it is just something that is done by healthier people and healthier people tend to be happier.

We could then form a new set of hypotheses.

So, a hypothesis, including the null hypothesis can’t actually be whether something exists or not because that isn’t properly formed. What we could do is perhaps have a hypothesis around expectations we might have if something exists, but even if those expectations are not met, it wouldn’t necessarily mean that thing doesn’t exist. You could just be wrong with your assumptions.

Is there such a thing as a non-falsifiable hypothesis?

Within science, if a hypothesis is not falsifiable or cannot be operationalised it is not considered a hypothesis, however, not all hypotheses are scientific.

Much like there is a more colloquial use of theory there is also a colloquial use of hypothesis which is equivocal. It can mean idea, opinion, rudimentary explanation, conjecture etc.

Of course, with a non-falsifiable hypothesis, you also can’t have a null hypothesis, at least, not without changing what the null hypothesis also means.

At best, this form of “null hypothesis” simply means “I don’t agree with your idea”.

At this point, you have to ask why you’re even using scientific terminology as you’re only confusing things and causing a miscommunication. You’ve broadened the terminology to a point where it loses all impact, much like I discussed in ‘The English Language Barrier‘.

Summary

A hypothesis isn’t simply an idea, it is a falsifiable idea that proposes a tentative explanation about a phenomenon or a narrow set of phenomena observed in the natural world.

Not only that, but this explanation can be operationalised so it isn’t mere conjecture.

If something is not falsifiable it is not a hypothesis, at least as far as science is concerned.

If you’re someone that usually uses scientific language in a scientific way but to discuss a hypothesis or the null hypothesis you switch to colloquial definitions you are going to be causing confusion and talk past people.

Special Thanks

Special thanks to Luke (MSc Forensic Psychology) and Steen (PhD Evolutionary Biology) for giving this the once over and suggesting a couple of additions.