Writing a null hypothesis

This is the definition trap: the two hypotheses are opposites, so the wrong one loses the mark. A null predicts no difference or no relationship; an alternative predicts there will be one. The command word decides which — read and underline null or alternative before writing. A second slip is hedging: there may be no difference is not firm. A hypothesis says will, not may.

Three near-misses score nothing because none predicts an outcome. An aim (to find out if noise affects memory) states a purpose. A question ends in a question mark. A lifted line copies the procedure; examiners warn material should be used, not lifted, and repeating the stem earns nothing. Start with there will be no difference in.

A hypothesis is only as good as its variables. Naming the general topic, not the manipulation, is incomplete — the material in a memory study was ruled not enough. A null with that fault, no difference in memory between the groups, is not measurable: neither groups nor outcome is named. Build it as: no difference in [measured] between [condition 1] and [condition 2].

The two variables must sit on the right sides: the DV goes after no difference in (the measured thing), the two IV conditions after between (the groups). Swapping them reads as if the researcher measured the manipulation, which is untestable. Keep the order: no difference in [measured] between [condition 1] and [condition 2].

A subtle failure is a null that still carries a direction. Silence will not recall more words than music looks negative but keeps the word more, and rules out only one outcome — silence scoring higher — not silence scoring lower. A true null denies any difference either way. Build from there will be no difference in …

Asked to describe the difference between a null and an alternative hypothesis, candidates write two separate definitions and stop. That scores one mark. The second comes from a connective such as whereas or however: a null predicts no difference, whereas an alternative predicts there will be one. Define each, then join them into one comparative sentence.

Two beliefs mislead. First, the null is not what the researcher expects — usually the opposite: they set up the cautious null to test their idea against. Second, writing a null does not mean the study found no difference: it is written before the data, so it is a prediction to check, not a result. Afterwards the findings fit it or not — a conclusion, not a restatement.

Because the phrase drilled in class is no difference, students freeze when a study measures two variables from the same people. A correlational study has no conditions, so no difference to deny; its null denies a relationship — e.g. no relationship between sleep hours and test score. It carries no direction: no positive relationship would wrongly allow a negative one.