IB Psychology HLTopic 5 — Research DesignPaper 3 & IACore idea~11 min read
Bias in the Researcher and the Participant
Nobody in a psychology study is neutral. The researcher hopes for a particular result, and the participant knows they are being watched. Both of those facts bend the data before anyone runs a single calculation — and knowing the named biases turns a vague “the results might be affected” into a proper evaluation.
📚 What you need to know
Bias is any systematic influence that pushes results away from the truth, and it lowers validity.
Researcher biases come from the researcher’s presence, behaviour or expectations.
The main ones: investigator effects, confirmation bias, question order bias, leading questions, sampling bias.
Participant biases come from people behaving differently because they are being studied.
The main ones: demand characteristics, social desirability bias, dominant respondent bias, acquiescence bias.
Bias is usually unconscious on both sides. Nobody has to be dishonest for it to happen.
Most fixes are about removing information: neutral wording, anonymity, and researchers who do not know the condition.
Two sources, one victim
Learn the two lists as lists. Exam questions often ask for “one researcher bias and one participant bias”, and having a named pair ready saves you thinking time.
What the researcher brings
Investigator effects are the most physical of these. A researcher’s age, gender, ethnicity, tone of voice, body language and clothing all change how participants respond. A brisk or over-excited tone adds emotion to a task that was supposed to be neutral. Dramatic body language pushes people into the wrong frame of mind. Even a slogan on a t-shirt is a signal, and signals get read.
Confirmation bias is quieter and more dangerous. It is the pull towards noticing evidence that fits your hypothesis and skating past evidence that does not. Ainsworth may have overlooked secure behaviours in a child she had already judged as insecure — not through dishonesty, but because a researcher who expects to see something tends to see it.
Researcher bias
How it shows up
Standard fix
Investigator effects
Tone, dress and manner change participant behaviour
Standardised script, neutral clothing, same greeting for everyone
Confirmation bias
Evidence that fits the hypothesis gets noticed and recorded
A second researcher codes the data without knowing the hypothesis
Question order bias
Early emotive questions colour every answer after them
Start neutral and general, move to specific and sensitive later
Leading questions
The wording tells the participant which answer is wanted
Rewrite as neutral, and record interviews so wording can be checked
Sampling bias
The sample does not represent the target population
Random or stratified sampling instead of opportunity sampling
Bias can also leak in through the instructions rather than the questions. If a researcher expects one condition to score higher, they may unconsciously read those instructions with a little more energy. That is why scripts exist.
What the participant brings
Being in a study is not an everyday experience, so people stop behaving in everyday ways. Demand characteristics is the umbrella term: participants pick up cues about what the study is looking for, then either help by delivering it or quietly resist it. Either way, what you have measured is their guess about your aim.
Participant bias
What the person is doing
Standard fix
Demand characteristics
Guessing the aim and playing along with it, or against it
Disguise the aim; use a cover task or an independent measures design
Social desirability bias
Under-reporting the embarrassing, over-reporting the impressive
Anonymity, neutral wording, sensitive items placed last
Dominant respondent bias
One person in a group steering everyone else’s answers
Round-robin questioning, or written answers before discussion
Acquiescence bias
Agreeing with whatever is asked, out of politeness or disinterest
Open-ended questions, and reversing the wording of some items
Participant reactivity
Trying too hard, or not trying at all, because they are watched
Covert observation where ethical; longer settling-in period
The acquiescence fix is neat and worth remembering. Reverse-word some items so agreement means the opposite. If somebody ticks “strongly agree” on both “I sleep well” and “I sleep badly”, you know they were agreeing, not answering.
Telling the two apart
Social desirability = I want to look good.
Demand characteristics = I think I know what you want.
🧩 Designing bias out of a study
Write and read a standardised script. Removes tone and instruction differences at a stroke.
Keep the aim vague in the briefing, then explain it fully in the debrief.
Make responses anonymous and say so clearly before people start.
Get someone else to code your qualitative data without telling them what you predicted.
Reverse the wording of a few questionnaire items to catch automatic agreement.
Choose a sampling method deliberately, and justify it. Sampling bias is a researcher decision, not bad luck.
Worked examples
WORKED EXAMPLE
Name and explain one bias of each type
A researcher who firmly believes music aids concentration runs a study. She personally reads the instructions to each participant and tells them the study is about “whether music helps you focus”. Identify one researcher bias and one participant bias in this design.
Step 1: Researcher biasInvestigator effects combined with confirmation bias — she has a strong expectation and delivers the instructions herself, so her tone may differ between conditions.
Step 2: Participant biasDemand characteristics — being told the aim tells participants exactly what result is wanted.
Step 3: Explain the effect on the data
Participants may try harder in the music condition to be helpful, so any difference reflects the briefing rather than the music.
Step 4: Give one fix for each
Use a recorded standardised briefing that does not reveal the aim, and have an assistant who does not know the hypothesis run the sessions.
Investigator effects and confirmation bias; demand characteristics“one of each” questions need two distinct named biases, not two examples of one
WORKED EXAMPLE
Which bias is this?
An anonymous online survey asks students how many hours of homework they complete each week. The average reported is far higher than school records suggest. The survey is anonymous, so the researcher is puzzled. Explain what is happening.
Step 1: Identify the biasSocial desirability bias — homework hours have a socially “correct” answer, so people inflate them.
Step 2: Address the anonymity puzzle
Anonymity reduces the bias but does not remove it. People also shape answers to protect their own self-image, not only their reputation with the researcher.
Step 3: Note the second possibility
Self-report of a habit over a whole week is also inaccurate through poor recall, which inflates estimates independently of any bias.
Step 4: Suggest an improvement
Ask about yesterday rather than a typical week, or corroborate with an objective record such as a homework log.
Social desirability bias, worsened by inaccurate recalla strong answer notices that anonymity is a partial fix, not a cure
💡 Exam tip
Name the bias exactly. Mark schemes reward the term, and there are plenty to choose from.
Say bias lowers validity. It does not lower reliability, and getting that link right is worth a mark.
Every bias point should come with a mechanism: what the person did, and why the data changed.
Keep social desirability and demand characteristics separate. They come from different motives.
For the IA, one named bias plus one concrete control measure is usually enough for the design section.
Sampling bias belongs on the researcher side, because it results from a decision the researcher made.
⚠ Common mix-up
Treating bias as deliberate. Almost all of it is unconscious on both sides.
Merging social desirability and demand characteristics. One is about self-image, the other about guessing the aim.
Saying bias makes a study unreliable. Biased studies are often perfectly repeatable. Validity is what suffers.
Thinking anonymity removes social desirability bias. It reduces it. People still flatter themselves.
Confusing acquiescence with agreement. Acquiescence means agreeing regardless of content.
Ignoring dominant respondent bias. It is the specific term for focus groups, and it scores better than “one person talked a lot”.
Up next: Reflexivity: The Researcher in the Research — what qualitative researchers do about their own bias when removing it is not an option.
Want this explained one-to-one?
Book a free session with an experienced IB Psychology tutor and get your trickiest topics made simple.
