‘Hypothesis’ is a word I see being thrown around social media; sometimes it just doesn’t sit right. This inspired me to look into various technical resources to see if any types of hypothesis actually fit the use I saw so many insisting was “correct”. This article hopes to address the use of hypothesis in the colloquial setting, give more detail around the common misconception, and provide details about the various types of hypothesis.
The problem with the way hypothesis is being used is that it is a technical term, primarily used in science and statistics, so it has a contextually correct definition/set of parameters. What compounds the issue is, that most who are using the term think they are doing it in a technical way. This might lead them to assume there is a null hypothesis to the “hypothesis” they are discussing.
- The Common Misconception
- What is a Hypothesis?
- Types of Hypothesis
- Causal hypothesis
- Relational Hypothesis / correlative hypothesis / associative hypothesis
- The descriptive hypothesis/tentative law.
- Statistical hypothesis
- Null and alternative hypothesis.
- Simply Complex
- Quick Recap on the Types of Hypothesis
- Fundamental qualities of a hypothesis
- What would we call a Non-Falsifiable Hypothesis?
The Common Misconception
The most common misconception I see is that “a hypothesis is just an idea or explanation about the natural world” and you can forgive this misconception because if you only do a search for a definition and stop there, you don’t get much more information than that.
As is often the case with these simplistic definitions of technical language, they miss the actual context and application of these terms. You can see here that one of the synonyms is conjecture, an opinion formed on incomplete information. But that doesn’t really describe what a hypothesis is, and how it is used/applied.
What is a Hypothesis?
A hypothesis is an explanation of an observation of some feature of the natural world or dataset that suggests a relationship between two (or more) variables. The hypothesis, whilst not a prediction itself, has a predictive quality that, if the hypothesis is supported, we can expect to see.
What is a Variable?
A variable is an element, feature, or factor that has the ability to change.
How does a variable fit in the hypothesis framework?
It might not be clear what is meant by a relationship between variables. It basically means that if we manipulate one variable we would expect the other to change. It might be better to give an example:
Example: The Lawn Hypothesis
Let’s imagine for a moment we are in our garden and you notice that half of your grass is much longer than the other half. You note that the shorter half is in the shade. You take some time and notice that even through the day, the shorter half of your lawn is mostly in the shade, whilst the other side gets full light for a large portion of the day.
Your observation leads to the idea that the amount of sunlight directly affects the rate of growth of your grass.
This idea is an explanation, is it a hypothesis?
Well, we have our subject – the grass, and we have two variables, the amount of sunlight and the rate of growth. If we didn’t have these variables, or they were poorly defined like “stuff makes grass change” then at best we have conjecture. Therefore, if it is just conjecture, it is not really a hypothesis.
You could, however, regard conjecture as a “bad” or “baseless” hypothesis but I would suggest that this might be tied to valuing the use of technical/scientific language and forgetting the whole purpose of it.
Even in the simplistic definition at the start, it is mentioned that it is a “starting point of an investigation” – if you don’t have variables, or they are so vague you cannot identify or manipulate them, then how can you investigate?
With the lawn hypothesis, we also note its predictive quality. If the hypothesis is to be supported then we would expect to see the amount of sunlight affecting the growth.
We probably would have made our hypothesis more specific, more sunlight increases the rate of growth. So if this is supported then we would expect an increase in the amount of sunlight to increase the rate of growth.
Once we have these variables and predictions, the result of the prediction not coming true means our hypothesis has been falsified – it might be that we refine the hypothesis and set of predictions or that the hypothesis is outright proven false.
So, whilst we may not actually test our hypothesis, the hypothesis has the possibility of being falsified.
Falsification is considered a big part of a valid/credible scientific hypothesis. It is largely agreed that for a hypothesis to be considered valid/credible or even scientific, it must be falsifiable, at least in principle, especially when discussing a causal hypothesis (like sunlight increasing growth rate).
With these components, it is then possible to operationalisation the hypothesis. Essentially, design the test, the variables and so on. Operationalisation helps you turn some fuzzy or abstract concepts into more meaningful data.
Hypothesis – The Required Components
- An explanation for an observation
- Variables (at least 2)
- Ability to Manipulate the Variables
- Predictive quality
Strictly speaking, these are the components that make a hypothesis valid/credible and therefore worthy of investigation.
If the only component is the explanation then it shouldn’t really be regarded as a hypothesis. At best, it’s a disregarded/weak hypothesis, but really it is just conjecture, an idea, a proposition, or a prediction.
I will come to justify this further in a future article but I do not want to dissuade and divert further from the primary focus of this article which is the ‘Types of Hypothesis’.
Prediction vs Hypothesis
A prediction is an expected result if a hypothesis or theory is true. Whilst a Hypothesis ought to have a predictive quality, a prediction does not qualify as a hypothesis in and of itself. The prediction is something we draw from a hypothesis or theory.
Types of Hypothesis
Now that we have a clear understanding of what is meant by a hypothesis, and how it ought to be used, let us now take a look at the different types of hypotheses.
There’s overlap between types of hypothesis and they can be described slightly differently in different sciences, though largely they are described the same way, and the key features are still present.
The causal hypothesis is generally the “main” hypothesis spoken about within science. In fact, it is very rare to hear ‘hypothesis’ being used in any other way in the scientific domain, and this is mostly due to what science is used for; investigating the causal links between natural phenomena.
e.g Something is observed, an explanation is proposed and a causal link is looked for.
By manipulating an independent variable, we expect that to cause a change within a dependent variable.
Causal Hypothetical Process
- Observe some phenomenon
- Ask what caused this
- Advancing a causal hypothesis (defined as a proposed explanation) for what has been observed (e.g., “the grass grows better on this side because it is exposed to more sunlight on this side”).
- Planning a test of the hypothesis that incorporates the generation of a prediction from the
- Conducting the test and comparing the results with the prediction.
- Drawing a conclusion as to whether the results of the test support or contradict the
Examples of Causal Hypothesis
Assuming we’ve observed some phenomena:
If we increase the amount of caffeine taken we will increase the level of activity in a specific group in comparison to a group which has less/no caffeine.
If we increase the amount of vegetables a group eats then their weight loss will be improved due to having less room for snacking and better digestive operation.
The Lawn Hypothesis is also an example of a causal hypothesis.
The causal hypothesis could be split into 3 hypotheses, non-directional, positive direction, and negative direction.
A non-directional hypothesis does not state the change or relationship between the variables.
A non-directional hypothesis is proposed when you’re not sure the effect is going be from manipulating the variables.
Caffeine causes a change in activity level/heart rate.
Eating more vegetables will change how much weight you gain or lose.
On the flip side, the directional hypothesis does have a direction.
Caffeine causes and increase in activity level/heart rate.
Eating more vegetables will increase your weight loss.
Caffeine causes a decrease in the activity level/heart rate.
Eating more vegetables will increase your weight loss.
Relational Hypothesis / correlative hypothesis / associative hypothesis
A relational hypothesis looks at the degree of overlap between the variables. So, instead of looking for a causal link and comparing it against a control group, the relational hypothesis simply looks for correlative qualities, and the more the variables seem to overlap, the stronger the relationship is said to be.
The variables are interdependent, that is, manipulating (or seeing a change in) one affects the other and vice versa, and the closer the associated link the better the hypothesis. Consider things like the amount of funding a hospital gets vs wait times or mortality rates. Religious upbringing and having a good education and so on.
Of course, correlation is not causation. Whilst it can be indicative of a link, sometimes you end up with this sort of thing:
Even though the Nick Cage correlation seems ridiculous, we could still test this by not letting Cage be in any new roles for a few years.
In this time he could be recording many films and then release them all in the same year once he had at least double the current max, we would then have an example of extremes to see if the correlation still applied and if it didn’t, whilst not necessarily falsified it would show that the relationship is weaker than the graph makes it look. Enough examples of weakness could lead to the hypothesis being rejected.
Of course, getting Nick to agree to this test is a wholly different ball game.
The descriptive hypothesis/tentative law.
The descriptive hypothesis, better described as the tentative law, is an observation of a regular occurrence or pattern. Essentially, it’s a statement based on induction… E.g. all swans are white.
The descriptive hypothesis is better not considered a “real hypothesis” in science, because whilst it is falsifiable in theory, it’s not necessarily falsifiable in principle e.g. the black swan observation is only really “testable” by observing all species around the world, at least until you find a non-white swan, which isn’t really practical or even possible for an individual to do this in an instance.
Descriptive hypotheses can be built into law which can feed into causal hypotheses and then produce those proper predictive qualities.
There are still variables, e.g. in the example of the swan it’s the colour of the swan and the amount of the world that has been observed. Whilst the variables are defined and quantifiable, they are not in the same league as the sort of variables for a causal hypothesis.
The statistical hypothesis is the hypothesis from which the null hypothesis was born.
The statistical hypothesis is a statement about the nature of a population or dataset. It’s often stated in terms of a population parameter.
Statistical hypotheses are probabilistic and made falsifiable by setting up rejection parameters.
Like the causal hypothesis, it works in relation to a dependent and independent variable. These are set up in the form of null and alternative hypothesis.
Null and alternative hypothesis.
The alternative hypothesis (or hypotheses) Ha/h1-n proposes a significance in the relationship between the variables whilst the null hypothesis suggests there’s no significant relationship between the variables.
If we consider our coffee example, the non-directional, positive directional, and negative directional are all alternative hypotheses as they all describe a relationship by manipulating the independent variable (quantity of caffeine) it causes and effect to the dependent variable, heart rate/level of activity.
The null, on the other hand, would be based on if there is no relationship between variables.. e.g. no matter how much caffeine someone had, there’s no significant difference in the heart rate/level of activity.
Hypotheses can also be split into groups like simple and complex.
A simple hypothesis is one with a single dependent and a single independent variable.
A complete Hypothesis is one with two or more dependent variables and two or more independent variables.
To lean into the coffee again, the amount of caffeine and the quantity of sugar causes and increase in the heart rate and level of perspiration.
Quick Recap on the Types of Hypothesis
The types of hypothesis overlap and are sometimes subsets.
E.g. the directional (or non-directional) hypotheses relate to the causal hypothesis and what sort of prediction we expect.
Simple and complex relate to the quantity of variables in a hypotheses.
The null and alternative hypothesis relate to the relationship between variables, e.g. the directional and non-directional hypotheses are alternative hypotheses and the null is when there is no significant relationship between the variables.
The statistical hypothesis doesn’t really need to be mentioned in the realm of science as it’s pretty much taken care of within the causal hypothesis.
The associative hypothesis is looking for an overlap between two variables
Essentially, if we consider all of the above types of hypothesis, they are all different sides of the same die and fit within a paradigm of alternative and null hypothesis.
That said, in science, when someone mentions a hypothesis it is almost guaranteed they are speaking of a causal hypothesis.
Fundamental qualities of a hypothesis
A hypothesis is supposed to provide an explanation for an observation of some natural phenomenon.
All of the types of hypothesis described propose some relationship between variables. A hypothesis requires at least one independent variable and one dependent variable (or 2 interdependent variables).
The variables are quantifiable and measurable. Without these variables there can be no null hypothesis, as the null hypothesis is not merely a negation of the alternative hypothesis but the result of there being no signification relationship between the variables.
Whilst a hypothesis isn’t strictly a prediction in and of itself, through explanation and operationalisation there should be clear predictions. If the hypothesis is true then we would expect to see this result.
Predictions can extend past hypothesis testing and be additional expectations one has if a hypothesis or theory is true. Consider something like the prediction of gravitational waves long before we ever had the ability to detect them.
Operationalisation & Falsification
“A fundamental requirement of a hypothesis is that is can be tested against reality, and can then be supported or rejected.”Research Hypothesis: Definition, Types, & Examples – Simply Psychology
When we operationalise a hypothesis we are essentially designing the experiment we could perform, variables and all, and this would include our acceptance/rejection criteria. Essentially if can operationalise a hypothesis, we ought to be able to falsify it too. It isn’t really considered operationalised if part of the experiment includes technology that hasn’t been invented yet.
The doctrine of falsification states that a scientific hypothesis is only considered credible if it is falsifiable, and that means having the ability to be tested and proven wrong.
This doctrine is considered a core part of the scientific method, especially in relation to the statistical hypothesis. If I proposed a green monkey that gives aids to all those who don’t pray to it, even though it was an “explanation of a natural phenomenon” it has neither the variables nor the ability to be falsified.
The doctrine doesn’t necessarily work when we’ve got to the limit of our ability to test. Consider string theory, we have no way of actually detecting these tiny strings but the math checks out. The enquiry has been a scientific one… It doesn’t suddenly become pseudoscience because it’s not testable or falsifiable… But it is as a bit of a dead end… At least with current technology. It’s why the string theory isnt always given the label hypothesis, and considered a strong hypothetical idea (that could one day be a theory of everything).
So, falsification isn’t a strict demarcation of science Vs pseudoscience in the way Popper seemingly wanted but it is indicative of whether something is a credible hypothesis or not. If there isn’t a body of supporting scientific information to make deductions and predictions from, the lack of falsifiability would indicate the statement is mere conjecture.
What would we call a Non-Falsifiable Hypothesis?
If a hypothesis is a supposition or proposed explanation made on the basis of limited evidence that proposes a relationship between 2 or more variables as a starting point for further investigation and is usually tentative and subject to testing and revision based on new data, what do we call an explanation that does not meet these criteria?
If a proposed explanation does not meet these criteria it might be regarded as a non-falsifiable hypothesis, as it cannot be tested or proven false by any observation or physical experiment. Therefore, it does not count as a scientific hypothesis, but rather as a conjecture, assumption or belief.
If a “hypothesis” is not falsifiable and/or doesn’t have at least 2 quantifiable variables with a proposed relationship, then it also doesn’t fit the paradigm of null and alternative hypothesis.
Examples of non-falsifiable hypotheses are:
Chocolate is always better than vanilla. [subjective]
There is a teapot orbiting the Sun between Earth and Mars. [technically untestable]
God exists. [metaphysical]
Therefore if anyone talks about these “hypotheses” and suggests there is a null hypothesis in relation to them, they are either wrong or are also changing the definition of null hypothesis.
The question then comes, why would we call an explanation that doesn’t have the variables, predictive qualities, testability or falsifiability a hypothesis at all?
Like there is a difference between a scientific theory and the colloquial use of theory, which just means an idea, there is a difference between all the types of hypothesis mentioned, especially the scientific(causal) hypothesis, and the non-falsifiable hypothesis which is mere conjecture, assumption, or belief.
The main issue here isn’t so much the colloquialism/broad use, as it is the use of technical language in a vague way as if it is the same as a scientific hypothesis, or even one of the other types.
There’s a variety of types of hypothesis, they overlap in places or describe concepts within a hypothesis/hypothesis testing.
When it comes to science, the causal hypothesis (inclusive of its directions and nulls) is what is meant by hypothesis as we are looking for the explanations, causes, how and why from our observations.
The descriptive hypothesis is used within science but is likely better described as “tentative law” as that more accurately describes what it is.
I’ll address why variables, operationalization, predictive qualities etc. are considered fundamentals of a hypothesis and why if a hypothesis is not falsifiable it is not really a hypothesis in my next article.
For now, consider that all the types of hypothesis outside of the colloqualism/broad use contain variables and operate in a way that allows them to have an alternative and null within. If you propose a hypothesis without those clearly defined variables then you don’t have an alternative and null. If you refer to this as a hypothesis, you’re not really using it in the way hypotheses are usually thought of.
Hypothesising something isn’t the same as formulating a hypothesis, though it is what leads there. Much in the same way thinking about what to have for dinner tonight isn’t the same as making dinner.
Hopefully, it’s helped explain hypotheses and we’re left with a better understanding, but failing that I’ve included my references below.
- Association versus Causation (bu.edu)
- Operationalization – Wikipedia
- Operationalisation | A Guide with Examples, Pros & Cons (scribbr.co.uk)
I’m Joe. I write under the name Davidian, not only because it is a Machine Head song I enjoy but because it was a game character I used to role-play that was always looking to better himself.
This is one of many things I hope to do with Answers In Reason.
I run our Twitter and IG accounts, as well as share responsibility for our FB group and page, and maintain the site, whilst writing articles, DJing, Podcasting (and producing), keeping fit and more.
Feel free to read a more detailed bio here: https://www.answers-in-reason.com/about/authors/4/
You can find my main social links here:
TikTok (AiR): https://www.tiktok.com/@answersinreason
TikTok (ADHD): https://www.tiktok.com/@adultadhdjourney
Ask me a question on Wisdom: https://app.wisdom.audio/ask/0be23c32-0fac-4d8f-bf68-671d9c8a3b95