Lesson 8 of 8 · 18 min
Investigating social influence
Identify the designs, variables and controls of social psychology studies, analyse data from a conformity practical with descriptive statistics and a t-test, and judge the reliability, validity and reproducibility of social research.
What you will be able to do
- Identify the design, IV, DV and controls of Asch (1955), Milgram (1963), Haney, Banks and Zimbardo (1973) and Cialdini et al. (2006).
- Plan an ethical class investigation of conformity, with hypotheses and controls.
- Analyse conformity data with the mean, standard deviation and an unpaired t-test, and draw a conclusion with a p-value.
- Explain why social psychology results can be hard to reproduce, and how culture and history affect them.
You have now met four studies that are quoted in almost every psychology textbook. Exam questions about them go beyond "what happened": they ask what kind of study it was, what was manipulated and measured, how good the evidence is and whether it would come out the same today. This lesson pulls those research skills together, then has you analyse data from a conformity practical of the kind a class can run ethically.
| Study | Design | IV | DV |
|---|---|---|---|
| Asch (1955) | Laboratory experiment; independent groups (control participants judged alone) | A unanimous majority giving a wrong answer (present or absent); in variations, majority size or a partner | Percentage of critical trials answered with the majority |
| Milgram (1963) | Laboratory study with a single condition (no comparison group); the later variations compared separate groups (independent groups) | None in the 1963 study; in the variations, features of the situation such as the learner's closeness | Highest shock given (0–30 switches, 15–450 V); obedient or defiant |
| Haney, Banks and Zimbardo (1973) | Simulation with random allocation to two roles, studied by observation, recordings and questionnaires | The role assigned: guard or prisoner | Behaviour and mood, from observation, recordings, reports and questionnaires (not fixed in advance) |
| Cialdini et al. (2006) | Field experiment; 2 × 2 factorial; each visitor saw one sign (independent groups) | Type of norm (descriptive or injunctive) and wording (negative or positive) | Percentage of marked pieces of wood stolen |
Why is Milgram's (1963) study usually not described as a true experiment?
Running your own conformity study. You cannot repeat Milgram or the Stanford Prison Experiment, and even Asch's procedure needs careful thought (deception, embarrassment). A common class practical tests informational influence with an estimation task instead:
Year 11 students estimate the number of jelly beans in a jar. Twenty volunteers are randomly allocated to two groups. One group writes their estimate on a sheet that already shows five high estimates from "earlier students" (written by the researcher); the other group writes on a blank sheet.
- IV: whether earlier (high) estimates are visible. DV: the student's estimate, in jelly beans.
- H₀: there will be no difference in the mean estimate of students who see earlier estimates and students who do not. H₁: students who see earlier high estimates will give a different mean estimate from students who do not.
- Controls: the same jar, the same viewing time and distance, standardised instructions, estimates made privately, random allocation.
- Ethics: written informed consent (with a parent or carer's written permission for anyone under 16), the right to withdraw, estimates kept anonymous, and a debrief that explains the made-up earlier estimates, because the study uses mild deception.
Classify each variable, choose the operational definitions and pick the hypotheses for a class version of Asch's study.
Here are illustrative results from such a class (made up to behave like typical results, not from a published study). The jar held 350 jelly beans.
| Group | Estimates | Mean | SD |
|---|---|---|---|
| Saw high earlier estimates | 520, 480, 610, 450, 560, 590, 430, 500, 540, 470 | 515.00 | 59.86 |
| Blank sheet | 380, 420, 310, 450, 360, 400, 290, 470, 350, 330 | 376.00 | 59.29 |
Independent groups, so the unpaired t-test applies. Switch the error bars between SD, SE and the 95% confidence interval.
Worked example
Analyse a conformity practical with an unpaired t-test
In a class practical on informational influence, 20 students were randomly allocated to estimate the number of jelly beans in a jar after seeing five high "earlier estimates" (group A) or on a blank sheet (group B).
Group A: 520, 480, 610, 450, 560, 590, 430, 500, 540, 470 Group B: 380, 420, 310, 450, 360, 400, 290, 470, 350, 330
(a) Calculate the mean and standard deviation of each group, to two decimal places. (b) Identify the appropriate t-test, and use it to draw a conclusion at the 0.05 level.
Step 1: Means
Group A: . Group B: .
Step 2: Standard deviations
Using the sample SD () on a calculator: , .
Psychology uses the sample standard deviation, dividing by .
Step 3: Choose the test
Different students were in each group (independent groups), so use the unpaired two-sample t-test.
Step 4: Run the test
, , (two-tailed).
Step 5: Conclude
Since , reject H₀. Students who saw high earlier estimates gave a significantly higher mean estimate (515.00) than those who did not (376.00), which supports informational influence.
Try it yourself first, then check the working one step at a time.
Answer
(a) A: mean 515.00, SD 59.86; B: mean 376.00, SD 59.29. (b) Unpaired t-test: , , ; reject H₀; seeing high earlier estimates significantly raised estimates.
Method
Mean and sample SD for each group.
Independent groups → unpaired; repeated measures or matched → paired.
Compare with 0.05 and conclude in the study's own terms.
Watch out for
Using the paired test because the groups are the same size.
Writing "proves" instead of "supports".
An unpaired t-test on the jelly bean data gives , , . Which conclusion is correct?
Common mistake
Significant does not mean proven, or important
A p-value below 0.05 means a difference this large would be unlikely if H₀ were true. It does not prove H₁, and a small study can still be a type I error (a false positive). A non-significant result with 10 per group may be a type II error: a real effect missed because the sample was too small. Say "supports" or "provides evidence for", never "proves".
Reproducibility. A result is reproducible when other researchers who repeat the method get similar results. Social psychology finds this hard:
- Ethics now forbid the procedures of the most famous studies, so they cannot be repeated exactly. Researchers use partial replications (Burger's 2009 study stopped at 150 V) or other methods.
- Behaviour depends on culture and time. Asch-type conformity is lower in some cultures and has fallen in the United States since the 1950s (Bond and Smith, 1996). A study that is reproduced in a different place or decade may give a different answer because people have changed, not because the original was wrong.
- Samples are narrow. Most classic studies used young, male, white, middle-class Americans, and most psychology research has been done in Western, educated, industrialised, rich and democratic (WEIRD) societies.
- Data collection and reporting. Le Texier's (2019) study of the Stanford Prison Experiment shows how selective recording and reporting can produce a dramatic result that later analysis does not support. Peer review, publishing full methods and data, and pre-registering hypotheses are ways science guards against this.
When the Open Science Collaboration (2015) repeated 100 published psychology studies, only about 36% of the replications gave a significant result, compared with 97% of the originals, and social psychology studies replicated less often than cognitive ones.
Explain how the historical context of Milgram's (1963) research affected both the perspective of the researcher and the relevance of the research. (2 marks)
Worked example
Describe the data from a class conformity study
In an Asch-style class study, 12 students each faced 12 critical trials. The number of critical trials on which each student agreed with the wrong majority was:
0, 3, 5, 1, 0, 7, 2, 4, 0, 6, 2, 3
(a) Calculate the mean, median and standard deviation, to two decimal places. (b) Calculate the overall error rate on critical trials as a percentage, and compare it with Asch's 36.8%.
Step 1: Mean
Sum , so mean .
Step 2: Median
In order: 0, 0, 0, 1, 2, 2, 3, 3, 4, 5, 6, 7. The median is the mean of the 6th and 7th: .
Step 3: Standard deviation
Sample SD (calculator): .
Step 4: Error rate
Total critical trials . Errors . Error rate .
Like Asch, count errors over all critical trials of all participants, not over people.
Step 5: Compare
The class rate (22.92%) is lower than Asch's 36.8%, but like Asch, people differed widely: three students never conformed and one did on 7 of 12 trials.
Try it yourself first, then check the working one step at a time.
Answer
(a) Mean 2.75, median 2.50, SD 2.38. (b) 22.92% of critical trials, lower than Asch's 36.8%, with similarly wide individual differences.
Method
Order the data before finding the median.
Error rate = total conforming responses ÷ total critical trials × 100.
Watch out for
Dividing 33 by 12 and calling 2.75 a percentage.
Giving "the percentage of students who conformed" (9 of 12 = 75%) when the question asks for the error rate.
Summary
Asch and Cialdini et al. ran experiments with independent groups; Milgram's 1963 study had a single condition; the Stanford Prison Experiment randomly allocated roles but measured behaviour loosely.
A class conformity practical needs clear variables, H₀ and H₁, controls (random allocation, standardised instructions) and ethics (consent, withdrawal, anonymity, debrief).
Independent groups are analysed with the unpaired t-test; means reject H₀, with the risk of a type I error; a non-significant small study risks a type II error.
Social psychology results are hard to reproduce: ethics limit replication, behaviour changes across cultures and decades, samples are narrow, and data can be selectively collected.
Research is shaped by its historical context, which affects both what researchers ask and how far the results apply elsewhere.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.