Bias is something systematically pushing a result away from what is really true.
Every part of that sentence is doing work, and the word carrying the most is systematically. Bias is not error. Error scatters. Bias pushes.
That distinction is what makes bias useful in an exam answer and what most answers miss.
Three things that are always true of it
It pushes. It does not scatter, which means more data will not fix it. A larger sample recruited the same way is a bigger biased sample, and its narrow margin of error can make it more misleading rather than less. Always name the direction it pushes.
It compares. Biased away from what? Bias lives in the gap between what was studied and what is being claimed. A study of one population is not invalid. It is a study of that population, and the bias appears only when the claim outgrows it.
It travels. It enters early and gets carried forward, through the measure, the analysis and the reporting, until it looks like an established fact.
The six places it enters
Most answers look for bias in the sample. It can enter at any of six stages, and knowing all six means you can find the one that actually matters.
| Stage | What can go wrong |
|---|---|
| Before the research | Historical, gender, cultural, publication and citation bias |
| The measure | Measurement, construct and method bias |
| Who is in | Sampling, selection, volunteer and attrition bias, and WEIRD samples |
| In the room | Demand characteristics, social desirability, recall, expectancy and observer effects |
| Reading it | Confirmation bias, interpretive bias, analysis choices |
| Using it | Diagnostic, institutional, algorithmic and communication bias |
The last row is the one students almost never reach, and it is often where the most interesting argument sits.
How to write about it
Six moves. You will rarely need all six in one paragraph, but a strong paragraph does at least the first three.
Name it. Not "there may be bias". The specific bias, with its actual name.
Give the direction. Which way does it push? This is the single move that turns a label into analysis.
Anchor it. To something specific in this study. If your sentence would fit any study ever conducted, it is not about this one.
Combine. Do two biases add up, or cancel out? They often add, and saying so is a genuinely sophisticated point.
Suggest a fix. What design change would test whether the bias is actually operating?
Narrow the claim. Which version of the conclusion survives?
That last move is what separates evaluation from complaint.
The sentence that earns nothing, and its replacement
There might be bias in this study.
Worth almost nothing. It is true of every study ever published.
Because participants volunteered in response to an advert about sleep, those most concerned about their own sleep were more likely to take part. That would inflate the reported association, so the true effect is probably smaller than this study suggests.
Named, directed, anchored, and it changes what the study can claim.
Choose one bias, do not list six
Listing six biases is not analysis. Choosing one is.
The question to ask is: which single bias most threatens this particular claim?
Answering that requires judgement, which is exactly what is being marked. Six named biases with no ranking between them shows recognition. One chosen and defended shows understanding.
If you have room, saying why you set the others aside is itself an analytical move.
Bias is not the same as perspective
These get conflated constantly, and keeping them apart is worth real marks.
A bias distorts a result away from something we could in principle get closer to. It has a direction, and reducing it makes the finding better.
A perspective is a standpoint from which questions get asked and categories get built. It has no direction to correct, because there is no view from nowhere to correct toward.
A researcher who samples only university students has a bias. A research tradition that treats individual self-report as the natural unit of analysis has a perspective. Different problems, different responses.
Finding bias does not mean the finding is wrong
This is the mature version of the concept and it is what top-band answers do.
Bias narrows a claim rather than destroying it. So the move is almost never to dismiss the study. It is to restate the conclusion at the size the evidence actually supports.
An answer that says "this study is unreliable because of sampling bias" has stopped too early. An answer that says what the study can still be trusted to show, and where exactly the claim stops, has done the harder thing.
Where this gets examined
Paper 1 Section C can name bias as the concept for a 15-mark question about any Area of Study.
Paper 2 Section B often makes bias one of the live concepts in an unseen study, usually alongside measurement or causality.
Paper 3 questions 1 and 3 both turn on it, question 3 explicitly, since one of its three tasks is how the researcher could reduce bias. Question 4 rewards it at the level of the whole evidence base rather than a single study.
And in your IA, the discussion asks you to discuss how researcher bias could affect your own investigation, which is the same concept turned on yourself.
Can you do this yet?
Tick only what you could do right now, on a study you have never seen.
- Explain why more data reduces noise but not bias
- Name a bias at each of the six stages
- Say which way a bias pushes, not just its name
- Explain what "biased compared to what" means, and use it on a claim
- Tell the difference between a bias and a perspective, with an example of each
- Explain why two biases might add up instead of cancelling out
- Choose which single bias most threatens a claim, and defend that choice
- Suggest a design change that would test whether a bias is operating
- Narrow an overstated claim down to the version the evidence supports
Fewer than seven? Do not reread this page. Take three studies from your class notes and run the questions on them. This becomes yours by using it, not by reading it again.
If you want the full concept worked through with labs and worked answers, that is what the PsychExcel Six Concepts Workbook is built for.