Research hypotheses

In virtually all quantitative research, and in some qualitative research, the researcher begins by proposing a hypothesis. This is a simple and clear statement about the expected outcome of the experiment or study. It could be an “educated guess” or hunch, based on the researcher’s own work or others’ work, or a prediction that follows from a theoretical framework that the researcher has adopted. A hypothesis should also be specific, meaning* It is something that can be stated in a fairly straightforward sentence or statement (if not, one may need to break it up into separate hypotheses)

• It is something that is specific enough that it can be addressed with one experiment or study, or a small set of related experiments / studies; in other words, it is specific enough that it can be manageable and adequately addressed in, say, a single journal article or chapter. If it is not specific enough, it will be difficult to empirically verify in a manageable study, and should be broken up into smaller components (e.g., separate studies, separate hypotheses).

1 Null hypothesis

When one proposes and tests a hypothesis, one in theory is also testing it against a null hypothesis. For example, if your hypothesis says, “if X then Y” (or “X causes Y” or “X is related to Y), you test it in essence by comparing it with and testing it against a null hypothesis that says, “if X not Y” (or “X doesn’t necessarily lead to Y / doesn’t cause Y / is not related to Y)”. In contrast to the null hypothesis, the researcher’s real hypothesis is called the alternative hypothesis. This comparison of null and alternative hypotheses is the basis for how statistical tests are done, as the math behind the statistical tests are usually based conceptually on comparing the real hypothesis with the null. The null is the hypothesis of no difference, and the alternative is the hypothesis of difference. For example:

• Null hypothesis (H₀): There is no difference in the reading speed of native speakers of Chinese and Russian learners in reading English as an L2.
• Alternative hypothesis (H₁): Some difference exists in the reading speed of native speakers of Chinese and Russian learners in reading English as an L2.

The alternative hypothesis – that there is a difference between the two groups - is the real hypothesis that the researcher wants to investigate, and s/he does so in the experimental design and statistical analysis by comparing it with the null. Another example:

• Null hypothesis (H₀): There is no difference in English L2 proficiency levels for Korean males versus Korean females entering American universities as undergraduates.
• Alternative hypothesis (H₁): There exists dome difference in English L2 proficiency levels for Korean males versus Korean females entering American universities as undergraduates.

The hypothesis should be straightforward, clear, and fairly easy to understand. It should also clearly indicate the variables that will be studied (the independent and dependent variables – see the later readings on variables).

The hypotheses can be either directional or non-directional. A directional hypothesis actually predicts the direction of the results – which group or score will be higher or lower than the other. A non-directional hypothesis makes no such prediction. For example, in comparing reading speeds of Chinese and Russians, one might predict that there will be a difference, without predicting specifically which group will be faster – a non-directional hypothesis. Or one might have a reason to expect that one group will read faster, and thus propose a directional hypothesis.

• Non-directional: In reading texts in English as an L2, some difference exists in the reading speed of native speakers of Chinese and Russian.
  (That is, Russian ≠ Chinese, meaning that either Chinese will read faster than Russians, or Russians faster than Chinese, i.e., Russian > Chinese or Chinese > Russian.)
• Directional: In reading texts in English as an L2, some native speakers of Russian will read faster than native speakers of Chinese.
  (That is, Russian-English > Chinese-English.)

Researchers often prefer to propose directional hypotheses, because testing them is a bit more straightforward, and the research may be more interesting – if the researcher has some previous theoretical or empirical basis for making such an assumption. If the researcher has no basis for making a specific directional prediction, then a non-directional hypothesis will be proposed. The choice may also be made based on statistical criteria.

For simple research designs, researchers may use a t-test (a very basic kind of statistical test) in the statistical analysis. For t-tests, the researchers will describe the t-test as one-tailed if it is used to test a directional hypothesis, or two-tailed if it is used to test a non-directional hypothesis (for reasons to be seen later when we discuss statistical tests).

2 Operationalization

Finally, the variables or factors referred to in the hypothesis, or factors tested in the study, should be defined operationally, that is, defined in a way that is clear and specific, and in a way that it can be readily measured. For example, in the above example about reading speeds, reading speed would have to be quantified or operationalized, or defined in a specific, measurable, quantifiable way, such as words per minute, as measured by an eye tracking device. For the example of language proficiency above, one would have to quantify language proficiency in a credible and valid way, such as a specific area of language ability (like syntactic, pragmatic, or communicative competence) as measured by a specific test for that ability (like a grammatically judgment test). It is common to use more abstract constructs or concepts in research, but to empirically test them in quantitative research, they have to be operationalized in a specific way that can be measured and tested .For example, constructs like language proficiency, musical ability, reading speed, cultural attitudes toward Canadians, intrinsic motivation, and others need to be properly operationalized in order to good research to be done.

3 Exercises

Here are some possible research hypothesis for you to consider. For each one, evaluate whether it is a good hypothesis or a poor one (or somewhat good / poor); be sure that you can explain why. Some might not be valid hypotheses at all.

1. A person can contract the flu even if s/he has had a flu shot.
2. The Loch Ness monster exists.
3. Do girls run faster than boys, especially if they are on a low-fat diet or if they are lighter in weight?
4. Short people are nicer than tall people.
5. Short people have higher IQs than tall people.
6. Taking fish oil supplements makes a person smarter.
7. Each of us has a guardian angel.
8. Chinese-English bilinguals obtain significantly higher scores on the TOEFL than Hungarian-English bilinguals.
9. George W. Bush won the 2000 presidential election.
10. Those learning a tone language (like Chinese) as a second language will have significantly greater difficulty attaining native-like pronunciation of tones if they begin learning the language after the critical period.
11. Those learning a tone language (like Chinese) as a second language will have significantly greater difficulty attaining native-like pronunciation of tones if they have had musical training.
12. Fairies help humans to fall asleep at night.
13. Girls are better at learning languages than boys.
14. I was kidnapped and tortured by aliens.
15. Drinking large amounts of coffee leads to heart attacks.
16. Ice cream causes drownings, because in summer months, as people eat more ice cream, more people die from drowning.

Imagine you want to conduct your own research study. How might you operationalize the following variables?

1. Proficiency in English as a second language
2. English grammatical ability
3. English pronunciation accuracy
4. Type or amount of motivation for learning a language