A hypothesis is a precise and testable statement of what a researcher predicts will be the outcome of a study. This usually involves proposing a relationship between two or more variables.
Verifying a hypothesis, also sometimes referred to as a working statement, requires using the scientific method, usually by designing an experiment.
For instance, one common adage is 'an apple a day keeps the doctor away'. If we use this aphorism as our hypothesis then we can make a prediction that consuming at least one apple per day should result in fewer visits to the doctor than the general population that eats apples sparingly or never.
In 2015, researchers at Dartmouth College, the University of Michigan School of Nursing, and the Veteran Affairs Medical Center in White River actually investigated this hypothesis. They combed national nutrition data collected from nearly 8,400 men and women — 753 of whom ate an apple a day. The study found that “evidence does not support that an apple a day keeps the doctor away; however, the small fraction of US adults who eat an apple a day do appear to use fewer prescription medications.”
So perhaps there's a glimmer of truth to this hypothesis, but not necessarily because apples are some miracle foods. It could be that people who eat apples every day also consume other fresh produce and less processed foods than the general population, a diet that helps to prevent obesity, a huge risk factor for a myriad of illnesses such as hypertension and diabetes that require prescription medication. This is why hypotheses need to be defined as precisely and as narrowly as possible in order to isolate confounding effects.
Types of hypothesis
The 'apple a day' study is an example of an alternative hypothesis, which states that there is a relationship between two variables being studied, the daily apple consumption and visits to the GP. One variable, called the independent variable, has an effect on the other, known as the dependent variable. The independent variable is what you change and the dependent variable is what you measure. For example, if I am measuring how a plant grows with different fertilizers, the fertilizers are what I can change freely (independent) while the plant’s growth would be dependent on what it is given. In order for an alternative hypothesis to be validated, the results have to have statistical significance in order to rule out chance.
Examples of alternative hypotheses:
- Dogs wag their tails when they're happy.
- The accumulation of greenhouse gases in the atmosphere raises global average temperature.
- Wearing a seatbelt reduces traffic-related fatalities.
- Students who attend class earn higher scores than students who skip class.
- People exposed to higher levels of UV light have a higher incidence of skin cancer than the general population.
Another common type of hypothesis used in science is the null hypothesis, which states that there is no relationship between two variables. This means that controlling one variable has no effect on the other. Any results are due to chance and thus pursuing a cause-effect relationship between the two variables is futile.
The null hypothesis is the polar opposite of the alternative hypothesis since they contain opposing viewpoints. In fact, the latter is called this way because it is an alternative to the null hypothesis. An apple a day doesn't keep the doctor away, you could propose if you were designing a null hypothesis experiment.
Examples of null hypotheses:
- Taking an aspirin a day doesn't reduce the risk of a heart attack.
- Playing classical music doesn't help plants grow more biomass.
- Vaccines don't cause autism.
- Hyperactivity is unrelated to sugar consumption.
The acceptance of the alternative hypothesis, often denoted by H1, depends on the rejection of the null hypothesis (H0). A null hypothesis can never be proven, it can only be rejected. To test a null hypothesis and determine whether the observed data is not due to change or the manipulation of data, scientists employ a significance test.
Rejecting the null hypothesis does not necessarily imply that a study did not produce the required results. Instead, it sets the stage for further experimentation to see if a relationship between the two variables truly exists.
For instance, say a scientist proposes a null hypothesis stating that "the rate of plant growth is not affected by sunlight." One way to investigate this conjecture would be to monitor a random sample of plants grown with or without sunlight. You then measure the average mass of each group of plants and if there's a statistically significant difference in the observed change, then the null hypothesis is rejected. Consequently, the alternate hypothesis that "plant growth is affected by sunlight" is accepted, then scientists can perform further research into the effects of different wavelengths of light or intensities of light on plant growth.
At this point, you might be wondering why we need the null hypothesis. Why not propose and test an alternate hypothesis and see if it is true? One explanation is that science cannot provide absolute proofs, but rather approximations. The scientific method cannot explicitly "prove" propositions. We can never prove an alternative hypothesis with 100% confidence. What we can do instead is reject the null hypothesis, supporting the alternative hypothesis.
It just so happens that it is easier to disprove a hypothesis than to positively prove one. But the supposition that the null hypothesis is incorrect allows for a stable foundation on which scientists can build. You can view it this way: the results from testing the null hypothesis lay the groundwork for the alternate hypothesis, which explores multiple ideas that may or may not be correct.
The alternative and null hypotheses are the two main types you'll encounter in studies. But the alternative hypothesis can be further broken down into two categories: directional and nondirectional alternative hypotheses.
The directional alternative hypothesis predicts that the independent variable will have an effect on the dependent variable and the direction in which the change will take place. The nondirectional alternative hypothesis predicts the independent variable will have an effect but its direction is not specific, without stating the magnitude of the difference.
For instance, a non-directional hypothesis could be "there will be a difference in how many words children and adults can recall," while the directional hypothesis could predict that "adults will recall more words than children."
Hypotheses can be simple or complex. A simple hypothesis predicts a relationship between a single dependent variable and a single independent variable while a complex one predicts a relationship between two or more independent and dependent variables. An example of a complex hypothesis could be "Do age and weight affect the chances of getting diabetes and heart diseases?" There are two independent and two dependent variables in this statement whose relationship we seek to verify.
How to write a good hypothesis
The way you formulate a hypothesis can make or break your research because the validity of an experiment and its results rely heavily on a robust testable hypothesis. A good research hypothesis typically involves more effort than a simple guess or assumption.
Generally, a good hypothesis:
- is testable, meaning it must be possible to show that a hypothesis is true or false, and the results of this investigation have to be replicable;
- includes both an independent and dependent variable.
- allows for the manipulation of the variables ethically.
- has clear and focused language. Don't be vague.
- is related to other published research.
- is written, either explicitly or not, as an "if-then" statement because we can then make a prediction of the outcome of an experiment.
An example of a testable good hypothesis is a conjecture such as "Students recall more information during the afternoon than during the morning." The independent variable is the time of the lecture and the dependent variable is the recall of the information presented in the lecture, which can be verified with standardized tests.
A bad hypothesis could be something like "Goldfish make better pets than cats." Right off the bat, you can see a couple of problems with this statement. What constitutes a good pet? Is a good pet fluffy and interactive or one that is low maintenance? Can I predict whether a cat or goldfish will make for a good pet? This is more a matter of opinion that doesn't provide any meaningful results.
Often, the best hypotheses start from observation. For instance, everybody has witnessed that objects that are thrown into the air will fall toward the ground. Sir Isaac Newton formulated a hypothesis in the 17th-century that explains this observation, stating that 'objects with mass attract each other through a gravitational field.'
But despite Newton's hypothesis being very well written, in the sense that it is testable, simple, clear, and universal, we now know it was wrong. In the 20th-century, Albert Einstein showed that a hypothesis that more precisely explains the observed phenomenon is that 'objects with mass cause space to bend.' The lesson here is that all hypotheses are temporary and partial, they're never permanent and irrefutable. This is also a good example of why the null hypothesis is so paramount.
Hypothesis formulation and testing through statistical methods are integral parts of the scientific method, the systematic approach to assessing whether a statement is true or false. All the best stories in science start with a good hypothesis.