As we use the scientific method to study language, there are a few things we need to keep in mind about the process.

Falsifiable hypotheses

When we formulate a hypothesis, it is important that we make a falsifiable hypothesis, which is a hypothesis that is logically possible to be proven wrong. It is important that your hypothesis is falsifiable so that you can do the next step in the scientific method, test your hypothesis! If it is impossible to fail your test, then perfoming the test doesn’t help us figure anything out.

Let’s take an example: our hypothesis is that ghosts exist but are invisible, have no smell or taste, make no sounds, and cannot interact with the physical world whatsoever. How could we test this hypothesis? We cannot see, smell, touch, hear, or taste a ghost. We cannot trap a ghost as evidence. This hypothesis is unfalsifiable. There is no test we could do that would show that it is wrong. Therefore, it is not a good hypothesis.

Confirmation bias

Another thing to watch out for when doing scientific research, or in fact any kind of critical thinking, is confirmation bias. Confirmation bias is a human tendency to look for evidence and information that is consistent with what they already believe.

Here is an example. Let’s pretend you work at a bakery and you believe that the macarons are the best and most popular item sold at your bakery. Every time someone buys a macaron, you may think, “See! Macarons are the best! Everyone likes them!” But if someone buys something else, you may not notice it or dismiss it as an exception. This would be a simple example of confirmation bias at work. In order to counter against this confirmation bias, you could count how many times people buy each kind of treat. Sometimes, when you do this, you may be surprised at the numbers!

Negative evidence

Another thing you could do to counter confirmation bias is to consciously train yourself to watch for negative evidence. Negative evidence is evidence that something is not possible, did not occur, or is absent. If you wanted to look for negative evidence that macarons are the most popular treat, you might count every time someone does not buy a macaron.

Negative evidence is very important in linguistic analysis. In linguistics, negative evidence often takes the form of ungrammaticality. When we test a hypothesis about the structure of language, we should consider both what patterns we predict to be possible and what patterns we predict to be impossible according to our hypothesis, and we should check for both.

In the last section, we compared the use of we and us by looking at both positive and negative evidence. The positive evidence were the examples of the sentences we could say, such as those in (1).

The negative evidence were the examples of the sentences that are awkward or unusual, such as those in (2).

If we only had the examples in (1), we wouldn’t know for sure whether we and us were interchangeable or not. It could be just a coincidence that we was always used in subject position and us was always used in object position. It is the examples in (2) that show us that it matters what position they go in by showing us where they can’t go.

Necessary and sufficient evidence

When you are making a hypothesis, you should think about what kind of evidence would allow you to draw conclusions about the accuracy of your hypothesis. There are two kinds of evidence to consider: necessary evidence and sufficient evidence.

Necessary evidence is evidence that must be the case in order for your hypothesis to be true. For example, let’s say that you think your sister ate the last cookie from the cookie jar. In order for this hypothesis to be true, the cookie jar must be empty. If there are any cookies left in the cookie jar, then it is impossible that your sister ate the last cookie. An empty cookie jar is necessary evidence that your sister ate the last cookie from the cookie jar.

However, even if the cookie jar were empty, this would not be sufficient evidence that your sister ate the last cookie. Maybe someone else ate it! In order to prove that it was your sister who ate the last cookie, you’d have to catch her in the act, or at least with crumbs around her face.

But even so, in order to prove she ate the last cookie (and not the second-last one), you need to catch her taking the last cookie from the cookie jar and eating it. This would be sufficient evidence that your sister ate the last cookie from the cookie jar.

Let’s apply this to a linguistic example. In the last section, our initial hypothesis was that we and us are perfect synonyms. In this case, the fact that we and us have the same meaning is necessary evidence. It could not be true that they are perfect synonyms if they do not share the same meaning. It is not possible to get sufficient evidence, though. In order to prove that they are perfect synonyms, you would have to show that they are always interchangeable in every possible context. Since there are infinite possible contexts, it is not possible to test them all.

When we make our hypothesis, thinking about what constitutes necessary and sufficient evidence can help us decide what data to look for to test our hypothesis. It can also help us decide whether the hypothesis is falsifiable. If a hypothesis is falsifiable, it should be possible to identify the kind of evidence that would prove its opposite.

Check yourself!

References and further resources

For a general audience

🧠 Veritasium. 2014. The most common cognitive bias. YouTube. https://www.youtube.com/watch?v=vKA4w2O61Xo

Academic sources

Davis, Henry, Carrie Gillon, and Lisa Matthewson. 2014. How to investigate linguistic diversity: Lessons from the Pacific Northwest. Language 90(4): e180–e226. https://muse.jhu.edu/article/563094/summary