IB Psychology HLTopic 5 — Research MethodsPaper 3 & IACore skill~11 min read
Correlational Studies and Their Limits
A correlation tells you two things move together. That is genuinely useful — it lets you predict, and it points researchers towards questions worth testing properly. What it never tells you is which one is doing the pushing, or whether something else is pushing both.
📚 What you need to know
A correlation is not really a method, it is an analysis of a relationship between two co-variables.
There is no IV and no DV, because nothing is manipulated. Both variables are simply measured.
Each participant gives two scores, one for each co-variable, and these are plotted on a scattergraph.
Relationships come in three shapes: positive, negative and none.
The correlation coefficient runs from −1 to +1. The sign gives direction; the number gives strength.
Correlation cannot show cause and effect, because of the third variable problem and the direction problem.
Correlations only work well for linear relationships. Curved ones are badly described by a single coefficient.
Co-variables, not variables
In an experiment you would call them the IV and the DV. Here you must not, and examiners will penalise it. Neither variable has been changed by anybody, so calling one of them “independent” claims a control you never had. The correct word is co-variables: two things measured on the same people at the same time.
The two hypothesis words
Experiments predict a difference.
Correlations predict a relationship.
That single word swap is worth learning properly. “There will be a difference in stress scores between students who sleep more and students who sleep less” is an experimental hypothesis. “There will be a relationship between hours of sleep per night and self-reported stress score” is a correlational one. Use the wrong word and the hypothesis mark disappears.
Reading a scattergraph
Which co-variable goes on which axis makes no difference to the result. The pattern is the same either way round, which is another clue that neither one is causing the other.
The correlation coefficient
Eyeballing a scattergraph is fine for a first look, but psychologists want a number. That number is the correlation coefficient, usually written as r. It has two parts and you must read both.
Students lose easy marks by treating negative coefficients as weak ones. A coefficient of −0.9 describes an extremely tight relationship.
Two words, two jobs. Direction comes from the sign, strength comes from the size. Write both in every answer: “a strong negative correlation of −0.78” scores where “a correlation of −0.78” does not.
Why it can never prove a cause
This is the part examiners care about most, and there are two separate reasons. Learn them as two, not one.
The direction problem: even if one does cause the other, the data cannot say which way round it goes. Does poor sleep cause stress, or does stress cause poor sleep? A scattergraph looks identical either way.
The third variable problem: some other factor may be driving both. This is the one worth drawing.
Hot weather sends people to the ice cream van and into the water. The link between the two co-variables is real, and completely uninformative about cause.
One more limit worth a mark: the coefficient only describes a straight-line relationship. Stress and performance form an upside-down U, so a coefficient near zero would wrongly suggest “no relationship” when a very clear one exists.
Worked examples
WORKED EXAMPLE
Interpret the coefficient
A study measures hours of screen time before bed and quality of sleep in 60 students. The correlation coefficient is −0.71. Describe the relationship and state what conclusion may and may not be drawn.
Step 1: Read the sign
Negative, so as screen time rises, sleep quality falls.
Step 2: Read the size0.71 is close to 1, so this is a strong relationship.
Step 3: State the conclusion you may draw
There is a strong negative relationship between the two co-variables, and screen time can be used to predict sleep quality.
Step 4: State what you may not conclude
You cannot say screen time causes poor sleep. A third variable such as anxiety could raise screen use and lower sleep quality at the same time.
Strong negative correlation; prediction yes, causation noalways give a named third variable, not “some other factor”
WORKED EXAMPLE
Write the hypotheses
A researcher plans to investigate whether the number of close friends a person reports is related to their self-esteem score. Write a suitable non-directional alternative hypothesis and a null hypothesis.
Step 1: Pick the right word
This is correlational, so use relationship, never “difference”.
Step 2: Operationalise both co-variables
Number of close friends reported on a questionnaire; self-esteem score out of 40 on a rating scale.
Step 3: Write the alternative hypothesis
There will be a relationship between the number of close friends reported and self-esteem score out of 40.
Step 4: Write the null hypothesis
There will be no relationship between the number of close friends reported and self-esteem score out of 40.
Non-directional relationship hypothesis + matching nullnon-directional means you do not say positive or negative
💡 Exam tip
Say co-variables, not IV and DV. It is a small word that signals you understand the method.
Every evaluation of a correlation should include the third variable problem with an example.
Correlations are ideal when manipulation would be unethical — you cannot randomly assign people to be bullied or to smoke.
Mention that correlations can use existing data, which makes very large samples possible.
If asked what a correlation is “useful for”, the safe answer is prediction and identifying links worth testing experimentally.
Watch for curved relationships. A near-zero coefficient does not always mean nothing is going on.
⚠ Common mix-up
Reading a negative correlation as a weak one. The sign is direction only. −0.9 is very strong.
Writing “difference” in a correlational hypothesis. The word must be relationship or correlation.
Saying correlation is “not scientific”. It is a legitimate analysis; it simply answers a different question from an experiment.
Confusing the direction problem with the third variable problem. One is about which way the arrow points, the other is about a hidden cause behind both.
Assuming a big sample proves causation. A larger sample makes the coefficient more trustworthy but never turns it into a cause.
Calling correlation a research method. Strictly, it is an analysis applied to data gathered by another method.
Up next: Case Studies: Depth Over Breadth — what you can learn from one person that a thousand participants would never show you.
Want this explained one-to-one?
Book a free session with an experienced IB Psychology tutor and get your trickiest topics made simple.
