Lesson 9 of 9 · 22 min
Stereotype priming and behaviour
Describe experiment 2 of Bargh, Chen & Burrows (1996), plan a modification of it, analyse walking-time data with means, standard deviations and an unpaired t-test, and judge the result's reliability in light of later replications.
What you will be able to do
- Describe the aim, design, procedure and results of experiment 2 in Bargh, Chen & Burrows (1996).
- Identify the IV, DV, hypotheses and controls of a modification of the experiment.
- Calculate means and standard deviations of walking times and interpret an unpaired t-test.
- Explain why later failures to replicate the result are a lesson in reliability.
- Identify the ethical issues in running a priming study.
You have probably had a song stuck in your head after hearing just its first few notes, or found yourself thinking of a word you saw on a billboard minutes earlier. Earlier experiences can shape what comes to mind next without your noticing. In 1996, John Bargh and his colleagues asked a bolder question: can activating a stereotype change what people do, without their knowing?
Definition
Priming
Priming is exposure to one stimulus that influences the response to a later stimulus or behaviour without the person being aware of the influence. Stereotype priming activates a stereotype, for example with words associated with it, so that it influences later judgements or behaviour.
Experiment 2 of Bargh, Chen & Burrows (1996). The researchers reasoned that if a stereotype is activated, behaviours linked to it in memory might become more likely, even when nothing in the situation calls for them. They chose the stereotype of older people, which in their participants' culture included being slow.
- Participants: undergraduate students at New York University, 30 in experiment 2a and 30 more in a repeat, experiment 2b.
- Cover story: participants were told the study was about language proficiency.
- IV (the prime): participants were randomly allocated to complete a scrambled-sentence task, making sentences from sets of jumbled words. In the elderly condition, the sets included words associated with the stereotype of older people, such as Florida, grey, wrinkle, bingo and retired. Words about slowness were deliberately left out, so any slowing could not be explained by the words themselves meaning slow. In the neutral condition, the sets contained words unrelated to age, such as thirsty, clean and private.
- DV (walking time): after the task, the experimenter thanked the participant and pointed them to the lift down the hall. A confederate, sitting in the corridor apparently waiting to see a professor, used a hidden stopwatch to time how long the participant took to walk from the door of the room to a strip of tape on the floor 9.75 m away.
Step through the design, predict the walking times, then see the results of experiments 2a and 2b.
| Condition | Mean walking time (s) |
|---|---|
| Elderly prime | 8.28 |
| Neutral prime | 7.30 |
Participants primed with the elderly stereotype took about a second longer to walk the corridor, and the difference was statistically significant, , . Experiment 2b, with a new sample of 30, found the same pattern. In a separate awareness check, only 1 of 19 participants showed any awareness of a link between the words and old age, and even that participant could not say how it might have affected their behaviour. Bargh and colleagues concluded that activating a stereotype can influence behaviour automatically, without awareness or intention.
In experiment 2 of Bargh, Chen & Burrows (1996), what was the dependent variable, as it was operationalised?
Classify each variable in a class version of the priming study, choose the operational definitions, and pick the null and alternative hypotheses.
Has it replicated? A single study, however neat, is not enough: a finding is reliable only if other researchers get the same result when they repeat it. This one has had a mixed record.
Doyen, Klein, Pichon and Cleeremans (2012) repeated the experiment with 120 participants, four times as many as the original, and timed them with infrared sensors 9.75 m apart instead of a person with a stopwatch. They found no difference in walking time between primed and unprimed participants. In a second experiment they told some of the experimenters running the sessions to expect primed participants to walk slowly, and told others to expect the reverse. Primed participants walked more slowly only when their experimenter expected them to. The experimenters' expectations seemed to be reaching the participants, perhaps through the experimenters' own pace or manner.
Other replications have produced mixed results, and behavioural priming is one of the findings at the centre of psychology's replication crisis, the discovery in the 2010s that many published results did not repeat.
Key idea
What the replications teach
- Measurement: a hand-held stopwatch operated by a person who may know the hypothesis is open to experimenter bias. Automated timing removes it.
- Blinding: everyone who has contact with participants or times them should be blind to the condition and, ideally, to the hypothesis.
- Sample size: small samples (15 per group) give unstable results; an effect found with a small sample may be a type I error.
- Replication: confidence in a finding comes from repeated, independent results, not from one significant -value.
See what happened when Doyen et al. (2012) timed the walk with infrared sensors, and when they changed what the experimenters expected.
Doyen et al. (2012) timed participants with infrared sensors and found no priming effect. What does this most strongly suggest about the original finding?
Note
Investigate it yourself (science inquiry)
The syllabus asks you to modify experiment 2 and investigate the effect of stereotype priming on behaviour. A school version usually keeps the logic and changes what is practical or what a better design needs:
- Participants: students in another class, randomly allocated to two conditions, rather than university students.
- The prime: a scrambled-sentence task with old-age words against neutral words, translated to Australian usage (pension rather than Florida).
- The DV: time in seconds to walk a measured distance (say 10 m) to the next room, or to complete a simple physical task.
- Controls: the same corridor, instructions read from a script (standardised procedure), and a timekeeper who does not know which condition each participant was in (double-blind), or automatic timing with a phone app triggered at two marks.
Ethics need care. The study depends on deception (participants cannot know the aim), so it needs approval, informed consent to take part in a study about language (with a parent or carer's written permission for students under 16), the right to withdraw, confidentiality of each participant's time, and a full debriefing that explains the deception and the hypothesis.
Worked example
Plan a modification of the priming experiment
A class plans to modify experiment 2 of Bargh, Chen & Burrows (1996). Twenty Year 11 volunteers will each unscramble 20 sentences: half of them (chosen at random) get sentences containing words associated with old age, and half get neutral words. A student who does not know which task each person did will then time them walking 10 m along the corridor with a stopwatch.
(a) Identify the IV and DV, with operational definitions. (b) Write the null and alternative hypotheses. (c) Identify two controls in the design and explain what each controls. (d) Identify an ethical issue and how to manage it.
Step 1: (a) IV and DV
IV: the type of prime, operationalised as unscrambling 20 sentences containing old-age words or 20 sentences of neutral words. DV: walking speed, operationalised as the time in seconds to walk 10 m along the corridor, measured with a stopwatch.
Step 2: (b) Hypotheses
Null: there will be no difference in the mean time to walk 10 m between students primed with old-age words and students primed with neutral words. Alternative: students primed with old-age words will take longer, on average, to walk 10 m than students primed with neutral words.
The alternative is directional because the original study predicted slower walking. Both name the IV and the DV as they were operationalised.
Step 3: (c) Controls
Random allocation to conditions spreads individual differences in walking speed (height, fitness) evenly across the groups. A timekeeper blind to the condition cannot unconsciously time one group more generously, controlling experimenter bias.
Step 4: (d) Ethics
The study uses deception: participants cannot be told the real aim. Manage it by getting approval and consent for a study "about language" (with a parent or carer's written permission for students under 16), then a full debriefing that explains the true aim and lets participants withdraw their data.
Try it yourself first, then check the working one step at a time.
Answer
(a) IV: old-age or neutral words in a 20-sentence scrambled-sentence task. DV: time (s) to walk 10 m. (b) H₀: no difference in mean walking time between the conditions. H₁: students primed with old-age words take longer on average to walk 10 m than those primed with neutral words. (c) Random allocation (participant differences); a blind timekeeper (experimenter bias). (d) Deception, managed by approval, consent and a full debriefing with the right to withdraw data.
Method
IV: what the researcher changes, with exactly how. DV: what is measured, with units.
Write both hypotheses about the population, naming the IV and DV.
Name each control and the threat it removes.
Name the ethical principle and the procedure that manages it.
Watch out for
Writing a hypothesis that does not mention the DV's measure ("old people words make people slow").
Calling 20 volunteers from one year level a random sample. It is a convenience (self-selected) sample; only the allocation is random.
Analysing class data. A class ran the modification above with 20 students: 10 primed with old-age words and 10 with neutral words. Each was timed walking 10 m.
| Condition | Walking times (s) |
|---|---|
| Old-age primes | 9.4, 8.7, 10.2, 9.1, 8.8, 9.9, 10.3, 8.9, 9.6, 9.5 |
| Neutral primes | 8.6, 9.3, 8.2, 9.4, 8.7, 8.1, 9.6, 8.8, 8.5, 9.0 |
Worked example
Analyse priming data with an unpaired t-test
A class modification of the priming experiment timed 20 students walking 10 m. Ten had been primed with old-age words and ten with neutral words.
| Condition | Walking times (s) |
|---|---|
| Old-age primes | 9.4, 8.7, 10.2, 9.1, 8.8, 9.9, 10.3, 8.9, 9.6, 9.5 |
| Neutral primes | 8.6, 9.3, 8.2, 9.4, 8.7, 8.1, 9.6, 8.8, 8.5, 9.0 |
(a) Calculate the mean and standard deviation of each condition, to two decimal places. (b) Identify the appropriate t-test and calculate and . (c) The test gives . Draw a conclusion.
Step 1: (a) Means
Old-age: . Neutral: .
Step 2: (a) Standard deviations
Squared deviations from each mean sum to (old-age) and (neutral). Divide by and take the square root: and . Your calculator's gives these directly.
Step 3: (b) Choose the test
Each student was in one condition only (independent groups), so use the unpaired t-test. .
Step 4: (b) Calculate t
Pooled SD: . Then .
With technology, enter both lists and run a 2-sample t-test with pooled variances, or use
T.TEST(range1, range2, 2, 2)in a spreadsheet for .Step 5: (c) Compare p with 0.05
, so the difference is statistically significant: reject the null hypothesis.
Step 6: (c) Conclude in words
Students primed with old-age words took significantly longer to walk 10 m (mean 9.44 s, SD 0.57) than students primed with neutral words (mean 8.82 s, SD 0.50). The result supports the alternative hypothesis, though with only 10 per group it should be replicated before it is relied on.
Try it yourself first, then check the working one step at a time.
Answer
(a) Old-age: mean 9.44 s, SD 0.57 s. Neutral: mean 8.82 s, SD 0.50 s. (b) Unpaired t-test (independent groups): , . (c) : reject the null hypothesis; the old-age primed group walked significantly more slowly.
Method
Mean = sum ÷ n; SD with (calculator ).
Independent groups → unpaired t-test, .
Compare with 0.05 and state the decision about .
Conclude in words with the IV, DV and the means.
Watch out for
Using the population SD (dividing by ): it gives 0.54 and 0.48.
Using the paired test. The groups are different people, so the scores are not paired.
Mean time to walk 10 m in each condition. Error bars show ±1 standard deviation. (Class data.)
Switch the error bars between SD, SE and the 95% CI, and see what happens to p when you switch to the (wrong) paired test.
The class's unpaired t-test gave , . Which conclusion is correct?
Worked example
Evaluate a priming study and its reliability
A student repeats the priming experiment with her netball team. The six players who arrive first are given the old-age sentences and the next six the neutral sentences. She times each player herself with her phone's stopwatch as they walk to the door. The old-age group takes a mean of 6.10 s (SD 0.42) and the neutral group 5.62 s (SD 0.35); an unpaired t-test gives , .
She concludes: "This proves that stereotype priming does not work."
Evaluate her method and her conclusion, with reference to validity, reliability and the replication history of Bargh, Chen & Burrows (1996).
Step 1: Allocation
Allocation was by arrival time, not random. Players who arrive first may differ systematically (keener, fitter, more punctual) from those who arrive later, a confounding variable that threatens internal validity. Random allocation would spread these differences.
Step 2: Timing and blinding
She timed the players herself and knew each player's condition, so her expectations could bias the times, the very problem Doyen et al. (2012) identified. A blind timekeeper or automatic timing would remove it.
Step 3: Sample
Twelve netballers is a small convenience sample from one sport, so the results generalise poorly (external validity), and with six per group the test has little power to detect a small effect.
Step 4: The statistics
, so she should fail to reject the null hypothesis. The difference is not statistically significant, but the means differ in the predicted direction, and with so few participants a real effect could have been missed (a possible type II error).
Step 5: The conclusion
A non-significant result does not prove that priming does not work. At most it fails to find evidence for it in this sample. Her conclusion overstates the data.
Step 6: Link to the wider evidence
The original study found a significant effect with 30 participants per experiment, but larger replications with automated timing (Doyen et al. 2012) did not, so the effect's reliability is genuinely uncertain. Her study cannot settle the question either way; a larger, randomly allocated, blind replication could add evidence.
Try it yourself first, then check the working one step at a time.
Answer
Her method has non-random allocation (a confound), an unblinded timekeeper (experimenter bias), and a small convenience sample (low power, poor generalisability). With she fails to reject the null hypothesis, but this does not prove priming has no effect: a type II error is possible. Given that the original finding has replicated poorly with tighter controls, the effect's reliability is uncertain; her study adds little evidence either way.
Method
Check allocation, blinding, sample size and sample type, and name the threat each creates.
Compare with 0.05 and use the right wording ("fail to reject", never "prove").
Consider type I and type II errors.
Place the result in the context of other replications.
Watch out for
Writing "accept the null hypothesis" or "proves". Statistics never prove a hypothesis.
Listing limitations without saying what each does to validity or reliability.
Common mistake
Choosing the wrong t-test
Each participant was in only one condition, so the two sets of times are independent: use the unpaired t-test, with . The paired test is for repeated measures or matched participants, where each score in one condition has a partner in the other.
Exam tip
Data questions on priming
- Give means and standard deviations to two decimal places, with units (s).
- Say which test you used and why (independent groups → unpaired).
- Report , and , compare with , and state the decision about the null hypothesis.
- Write the conclusion in words, naming the IV and the DV.
- When asked to evaluate, name a specific threat (an unblinded timekeeper, a small sample, a convenience sample) and its effect on validity or reliability, and mention the failed replications.
Summary
Bargh, Chen & Burrows (1996), experiment 2: students who unscrambled sentences with words linked to the elderly stereotype walked 9.75 m more slowly (8.28 s) than those with neutral words (7.30 s), , , apparently without awareness.
IV: elderly or neutral prime (independent groups, random allocation). DV: walking time over a fixed distance, timed covertly.
Doyen et al. (2012) found no effect with 120 participants and infrared timing, and found slowing only when experimenters expected it: the finding's reliability is in doubt.
Analyse a modification with means, standard deviations and an unpaired t-test (); reject the null hypothesis when .
A class modification needs random allocation, a blind timekeeper or automatic timing, standardised instructions, approval for deception, consent, the right to withdraw and a full debriefing.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.