“Psychologists Say…” — But How Do They Actually Know?

A headline claims that social media makes teenagers unhappy. What evidence would convince you?

Imagine that two websites report the same fictional investigation:

“Heavy social-media users report lower wellbeing.”

“Social media causes unhappiness.”

Do those statements mean the same thing?

The first describes a relationship. The second claims a cause.

That is a substantial difference. Yet it can disappear in the few seconds between reading a headline and sharing it.

Perhaps social-media use contributes to unhappiness. Perhaps people who already feel unhappy spend longer online. Perhaps bullying, poor sleep or difficult circumstances influence both.

Before deciding which explanation to believe, we need to ask:

What did the researchers actually do?

That question takes us into research methods: how psychologists turn ideas into investigations, collect evidence and decide what their findings allow them to conclude.

The investigations and numerical results below are illustrative, rather than findings from published social-media or memory research.

A Pattern Is the Beginning of an Explanation

Suppose researchers recruit 200 teenagers. Each completes a wellbeing questionnaire, and researchers record their average daily social-media use over the previous week.

The results show that teenagers who spend longer on social media tend to have lower wellbeing scores.

Assuming higher questionnaire scores represent better wellbeing, this would be a negative correlation: as one measured variable increases, the other tends to decrease.

In correlational research, these measured variables are often called co-variables.

The researchers have identified a pattern. They have not allocated teenagers to different amounts of social-media use.

Three explanations for the same relationship

1. Social-media use affects wellbeing.

Some online experiences might contribute to distress.

2. Wellbeing affects social-media use.

Someone feeling isolated might spend more time online seeking company or distraction. This is the problem of direction of causation.

3. Another variable influences both.

Bullying, for example, might lower wellbeing and change someone’s online behaviour.

Sleep needs particularly careful thought. It might be a shared influence, or it might form part of a causal pathway: late-night scrolling could reduce sleep, which then affects wellbeing.

Those possibilities require different evidence.

A correlation alone cannot establish which causal explanation is correct.

That does not make correlations unhelpful. They reveal patterns, suggest questions and allow psychologists to investigate matters that would be difficult or unethical to manipulate experimentally.

Also remember that “positive” and “negative” describe the direction of a correlation. They do not mean beneficial and harmful.

What Makes an Experiment Different?

Consider a less sensitive question:

Does background music affect immediate word recall?

An experiment could compare volunteers who study a word list with instrumental music playing against volunteers who study the same list in silence.

The researcher changes one variable and measures another:

Independent variable, or IV: the sound condition during learning.
Dependent variable, or DV: the number of words correctly recalled under specified conditions.

Participants could be randomly allocated to the two conditions.

Random allocation helps reduce systematic differences between the groups. It does not guarantee that a small sample will contain equally strong memorisers in each condition, but it reduces deliberate or predictable allocation bias.

The researchers also standardise the task, timings, instructions and scoring.

If recall differs between the conditions, these design features make it easier to attribute the difference to the sound condition.

However, calling a study an experiment does not automatically make its conclusion convincing. Poor controls, unclear measurements or biased allocation can still undermine it.

A finding involving one instrumental recording and one word-list task also cannot establish that every kind of music affects every kind of learning in the same way.

Operationalisation: What Exactly Are We Measuring?

Psychologists investigate happiness, aggression, concentration, memory and many other concepts.

We cannot place “concentration” on a balance and read off its mass. We must decide what observable response or score will represent it.

This is operationalisation: defining precisely how a variable will be manipulated or measured.

Broad idea	Possible operational definition	Question to ask
Social-media use	Average daily minutes recorded by specified apps over seven days	Does this capture use on other devices?
Wellbeing	Total score on a specified questionnaire about the previous two weeks	Which aspects of wellbeing do its questions cover?
Aggression	Number of clearly defined insulting comments during a task	Does this miss other forms of aggression?
Concentration	Number of target symbols correctly identified in three minutes	How much does visual-search ability affect the score?
Memory	Number of words correctly recalled from a 20-word list in 60 seconds	Does this represent the memory we want to understand?

Specificity makes a procedure easier to understand and repeat. It also exposes its limitations.

“Measure memory” is an idea.

“Count the number of correct, distinct words recalled from a 20-word list during a 60-second recall period” is a procedure someone can carry out.

A questionnaire score is not the whole person

Suppose someone scores 18 on a wellbeing questionnaire.

We have measured their responses to particular questions, using particular scoring rules, at a particular time.

Whether that score provides a useful measure of wellbeing depends on the evidence supporting its interpretation.

Question wording, the period participants are asked to remember and their willingness to answer honestly can all matter.

Giving a concept a number does not automatically mean we have measured it well.

A reflection from teaching science

In my science teaching, measurement is often a useful starting point.

A thermometer gives a reading, but we still have to consider where it was placed, when the reading was taken and what question it answers.

Psychology deserves the same care.

When a student writes “measure concentration”, the useful follow-up is:

“What would I actually see you record?”

That question turns an abstract term into a practical decision.

Hypotheses: Predicting Something We Can Test

An aim states what an investigation intends to explore:

To investigate whether instrumental background music during learning affects immediate word recall.

A hypothesis predicts a testable outcome.

A non-directional experimental hypothesis could be:

There will be a difference in the number of words correctly recalled from a 20-word list during a 60-second silent recall period between participants who study the list for 60 seconds with a specified instrumental recording playing and those who study it in silence.

This predicts a difference without predicting which condition will produce higher scores.

A directional hypothesis would predict higher or lower recall in one condition. The direction should have a reason behind it, such as previous evidence or an existing explanation.

A corresponding null hypothesis predicts no difference beyond that expected by chance. In an exam answer, include enough operational detail to make clear which conditions and measurement you mean.

For a correlational study, predict a relationship between co-variables. For an experiment, predict a difference between conditions.

OCR’s 2024 Research Methods examiner report highlights the importance of operationalising variables in hypotheses and applying answers specifically to the scenario. Knowing the term is only the beginning; students must show how it works in the investigation. Summer 2024

What Else Could Explain the Result?

Suppose the music group recalls more words.

Before announcing that music improves memory, inspect the procedure.

Did the music group receive easier words? Were they given longer to learn them? Did they work in a quieter room?

Extraneous variables

An extraneous variable is a variable other than the IV that could affect the DV.

Examples include tiredness, distractions, familiarity with the words and differences in memory ability.

Confounding variables

A confounding variable varies systematically with the IV and offers an alternative explanation for a difference in the DV.

Imagine that the music group studies familiar words such as “chair”, “apple” and “river”, while the silence group studies unfamiliar technical terms.

Word difficulty now changes alongside sound condition.

If recall differs, we cannot tell whether the explanation is music or vocabulary difficulty.

This is why controls need a purpose.

“Use the same word list” matters because it prevents differences in word difficulty from explaining differences in recall.

That explanation is much more useful than simply saying, “It makes the study fair.”

Reliability: Is the Measurement Consistent?

Reliability concerns consistency.

If two people mark the same recall sheets, do they award similar scores?

Clear rules about spelling, repeated answers and acceptable responses can improve agreement between scorers. Having scorers independently mark the sheets allows us to assess inter-rater reliability.

For a questionnaire measuring a characteristic expected to remain stable, administering it again after a suitable interval can help assess test–retest reliability.

But repeating a memory test creates a complication: participants may remember the words or develop a better strategy.

A changed score might reflect practice rather than an unreliable measurement.

Standardised instructions and timings make procedures more consistent and easier to replicate. They support good research practice, but do not by themselves demonstrate that a measure is reliable.

Validity: Does the Evidence Answer Our Question?

Validity concerns whether a measurement or conclusion is appropriate for its intended purpose.

Three distinctions help.

Construct validity

Does the measure adequately represent the concept?

A word-recall task measures a particular aspect of memory. It does not measure every form of memory.

Remembering unrelated words immediately differs from remembering how to sail a boat, recognising a face or recalling a biology explanation next week.

Internal validity

Can we reasonably attribute the difference in recall to the sound condition?

If word difficulty or learning time also differs between conditions, the causal interpretation is weakened.

External validity

How far can we extend the findings to other people, settings or tasks?

A small classroom investigation may tell us something about those participants completing that task. Applying the result to all students revising all subjects requires further evidence.

Consistent does not necessarily mean valid

Imagine a clock that is always ten minutes fast.

Its readings can be consistent while still giving the wrong time.

Similarly, a task could consistently measure reading speed while being a poor measure of the concentration the researcher intended to investigate.

Reliability and validity are connected, but they are not interchangeable.

Try It: Does Background Music Affect Recall?

Before reading the design below, identify:

An independent variable.
A dependent variable.
Two controls, with reasons for using them.

A manageable classroom investigation

Aim: Investigate whether a specified instrumental recording during learning affects immediate word recall.

Participants: A small group of willing volunteers, randomly allocated to two conditions.

Each person completes only one condition. This is an independent-measures design.

Materials: A list of 20 common, unrelated words, a timer, answer sheets and an instrumental recording played at a fixed, comfortable volume.

Procedure

Explain the task, obtain informed consent and make clear that participation is voluntary.
Give everyone the same instructions. Ask participants not to write during learning.
Present the same word list in the same format for 60 seconds. Play the recording in the music condition; use silence in the other.
Remove the list and stop the music.
Give everyone 60 seconds to write down remembered words in silence.
Score correct, distinct words out of 20 using agreed rules.
Debrief participants and remove names from any shared results.

Music is manipulated during learning only. Testing everyone in silence helps keep the investigation focused on that learning condition.

Feature	Operationalised example
IV	Specified instrumental recording during the 60-second learning period versus silence
DV	Number of correct, distinct words recalled from the 20-word list during 60 seconds in silence
Control 1	Same word list, preventing vocabulary difficulty from differing between conditions
Control 2	Same learning and recall durations, preventing differences in time available
Further controls	Same presentation format, equivalent surroundings, standard instructions and scoring rules

What would the results mean?

Suppose the following invented results were obtained:

Silence condition: mean recall of 12 words.
Music condition: mean recall of 10 words.

Using standard text characters:

Mean words recalled = total words correctly recalled / number of participants

Difference between means = 12 - 10 = 2 words

We should also inspect the spread and individual scores.

The difference in means alone does not establish a statistically significant difference or justify advice for every learner.

A proportionate conclusion would be:

In this classroom investigation, participants in the silence condition recalled more words on average. Further research would be needed to assess how dependable and generalisable this pattern is.

Extension: let everyone try both conditions

A repeated-measures design allows each participant to complete both conditions. This reduces the problem of comparing people with different memory abilities.

However, it introduces practice, fatigue and word-list problems.

You would need matched lists and counterbalancing so that neither sound condition consistently receives the easier list or comes second.

Improving one feature of a design often creates another decision to manage.

Exam Writing: Replace Labels With Explanations

“The study was scientific” does not explain what makes its evidence convincing.

A useful pattern is:

Specific feature → how it works in this study → why it matters for the conclusion.

Vague statement	More useful explanation
“The study was controlled.”	“Both groups studied the same words for 60 seconds, reducing the chance that word difficulty or learning time explains a recall difference.”
“It was objective.”	“Agreed scoring rules reduce the influence of the scorer’s personal judgement when marking recalled words.”
“It was reliable.”	“Two scorers could independently mark the recall sheets; close agreement would provide evidence of inter-rater reliability.”
“It lacked validity.”	“Immediate recall of unrelated words may not represent remembering connected lesson material a week later, limiting generalisation to revision.”
“Random allocation was a strength.”	“Random allocation reduces systematic assignment of stronger memorisers to one condition, although chance differences remain possible.”

Adapt these explanations to the question and the actual procedure.

One clear, accurate explanation is more useful than several technical terms without a connection to the study.

Five Questions to Ask Before Sharing the Headline

Whenever you read “psychologists say”, ask:

What was measured? Look for the actual task, response or score behind the broad concept.
Was anything manipulated? Distinguish an experiment from a relationship between measured variables.
What else could explain the finding? Consider reverse causation, shared influences and confounds.
Who took part, and where? Consider how far the conclusion can reasonably extend.
How dependable is the evidence? Look at sample size, uncertainty, consistency and support from other research.

These questions matter when assessing learning apps, workplace training, advertising and productivity advice.

Research methods help us decide how much confidence a claim deserves.

Conclusion: Keep Being Curious After the Headline

Psychology becomes more interesting when we examine how its evidence was produced.

A claim about happiness becomes a discussion about measurement. A claim about learning becomes an investigation we can design. A surprising correlation becomes the beginning of several competing explanations.

In teaching, I want students to become comfortable asking why a conclusion follows from the procedure. That is valuable examination preparation and a useful habit outside the classroom.

The most productive response to “psychologists say” is thoughtful curiosity:

“How did they find that out — and what does their evidence allow us to conclude?”

25 October 2026