1.3.2 - Burger’s partial replication

Why stop at 150 volts?

Burger’s study addressed a historical comparison while deliberately limiting how far the obedience procedure could continue.

The research was conducted in 2006 and published in 2009. Its base procedure partially replicated Milgram’s Experiment 5, in which the learner mentioned a heart condition and strongly protested after the 150-volt shock. Burger wanted to see whether contemporary adults would continue after this protest, whether seeing someone refuse would reduce obedience, and whether gender or personality was associated with responses.

The 150-volt point was chosen because it was a clear moment of decision in the earlier procedure. In the comparable Milgram condition, 33 of 40 participants continued beyond it, and 26 of those 33 reached 450 volts: about 79% of those who passed that decision point. This suggested that the response at 150 volts could inform a cautious prediction about further obedience. It is a conditional percentage, not 79% of all Milgram participants.

In Burger’s study, no participant pressed a switch above 150 volts. After administering the 150-volt shock and hearing the learner’s protest, a participant counted as continuing if they began reading the next test item. The experimenter immediately stopped the session. A participant who refused after the prescribed prods, or had already stopped earlier, counted as a stopper. Thus “continued after 150 volts” means an observed decision to continue the task, not delivery of a 165-volt or 450-volt shock.

Stopping at this point avoided the later, more intense part of Milgram’s procedure. However, the later stages were also the very behaviour Burger wished to predict. Ethical restraint therefore created an inference limit: willingness to take the next step was observed; eventual obedience to the maximum voltage was not.

Participants and protection

The recruitment and screening decisions protected participants but also shaped the population to which the results apply.

Adults volunteered in response to newspaper and online advertisements and local flyers. They were offered $50 for two sessions. The final sample comprised 70 adults: 29 men and 41 women, aged 20–81, with a mean age of 42.9 years. This was a volunteer sample around Santa Clara, California, not a random sample of the United States.

Burger used two stages of screening. An initial telephone screen excluded people whose reported medical or psychological circumstances suggested greater risk, and people who had taken more than two psychology classes, to reduce familiarity with Milgram’s study. At the second stage, participants completed inventories and underwent an interview with a licensed clinical psychologist. The clinician used anxiety and depression information and interview responses to judge whether participation might cause a negative reaction, with instructions to be cautious. People who revealed knowledge of Milgram during the experimental session were also excluded from the final sample.

Several safeguards operated during the procedure. Participants were told repeatedly that they could withdraw and keep the $50. A sample shock, given only with consent, was 15 volts rather than Milgram’s 45 volts. The experimenter was a clinical psychologist who could stop the study if excessive stress appeared. Immediately after the session, participants learned that the learner had received no shocks; the learner entered to reassure them, followed by a fuller debriefing. The university’s institutional review board approved the procedure.

These precautions reduced avoidable risk rather than guaranteeing an absence of discomfort. Participants were still deceived about the true purpose, the rigged roles and the learner’s apparent suffering. The instruction that they had “no other choice” could also conflict with their earlier withdrawal assurance. Approval and debriefing are relevant protections, but they do not themselves answer every ethical objection.

Screening introduces a scientific trade-off. Excluding people judged vulnerable makes the study safer, but results concern a screened volunteer group. Their reactions may differ from those of adults excluded for health, stress or prior knowledge. One cannot simply infer that the sample represents all adults.

The base and modelled-refusal conditions

Burger changed the presence of a refusing peer while keeping the main authority-and-learning situation similar.

Participants were randomly allocated to two conditions, with an attempt to keep gender proportions similar. The base condition contained 40 participants; the modelled-refusal condition contained 30. Comparing these independent groups tested whether witnessing one peer’s refusal changed subsequent obedience. Personality and gender were measured characteristics, not variables randomly assigned by the researcher.

In the base condition, an apparent draw always made the real participant the teacher and a male confederate the learner. A confederate is someone working with the researcher while appearing to be another participant. The teacher read paired words and response choices and was told to administer an increasing shock after each wrong answer, starting at 15 volts. A generator displayed switches up to 450 volts in 15-volt steps, but only those up to 150 volts could actually be used. The learner received no real shocks; recorded reactions created the impression of pain.

The learner mentioned a heart condition. After the 150-volt switch, he demanded release and said his heart was troubling him. If the teacher hesitated, the experimenter gave a standard sequence of four prods, moving from a request to continue to a statement that the participant had no choice. Standardising the task, reactions and prods made responses more comparable. It also preserved an authority situation, so the measure concerned obedience under particular pressures, not a person’s general willingness to harm someone.

In the modelled-refusal condition, another confederate of the participant’s gender initially acted as Teacher 1. After pressing 90 volts and hearing the learner, the model expressed doubt and refused to continue. The experimenter asked the real participant, Teacher 2, to take over; the model remained silently seated. The manipulation was therefore one comparatively restrained refusal before the participant delivered shocks, rather than Milgram’s variation involving two dramatic refusals later in the procedure.

If one visible refusal supplied a norm that disobedience was acceptable, the real participant might be less likely to continue. But an observed difference would need to be evaluated alongside sampling variation. A lower percentage alone does not establish an effective intervention.

Findings and justified inferences

The principal outcome was the proportion who began the next item after the 150-volt protest, not the highest voltage reached.

The base percentage was 12.5 percentage points below Milgram’s corresponding decision-point rate, but the difference was not statistically significant in Burger’s reported analysis. Nor was the difference between Burger’s two conditions significant. The data therefore did not support the predicted reduction from the particular modelled refusal used here. They do not prove that the conditions are identical, that no peer refusal can ever help, or that obedience is unchanged in every population.

No significant difference in continuation rates between men and women was found. Personality results were more qualified. Greater empathic concern was associated with expressing reluctance earlier in some analyses, but did not reliably distinguish people who stopped from those who continued. In the base condition, stoppers had higher desire-for-control scores than continuers; this was not found in the modelled-refusal condition. These measured associations leave the relationship between personality and obedience uncertain and do not show that a trait caused an individual’s decision.

Burger interpreted the broad similarity with Milgram as evidence that strong situational pressures still operated decades later. A perceived legitimate expert, an unfamiliar task and gradually increasing demands could encourage continuation despite concern for the learner. That interpretation is compatible with evidence of early hesitation: discomfort and refusal are not the same response.

The paper’s counts, endpoint and qualifications can be checked in Burger’s original report, pp. 7–10. Distinguish the observed findings from the author’s prediction that many continuers might have obeyed further.

Evaluating the partial replication

The strength of this replication depends on whether a particular conclusion requires the features retained or the features changed.

Reproducing the generator, word task, learner reactions and much of the script improved comparability with Milgram up to the selected endpoint. Standard procedures and random allocation strengthened the comparison of the base and modelled-refusal groups. The study measured a consequential decision rather than asking people to predict whether they would obey, reducing reliance on socially desirable self-description.

However, the earlier decision point is an imperfect substitute for the full procedure. The 79% conditional continuation in Milgram’s sample makes a prediction plausible; it cannot guarantee the same transition from 150 to 450 volts in Burger’s participants. They might have reacted differently to later protests. Treating 70% as an observed maximum-voltage obedience rate would change the dependent variable and overstate the evidence.

Historical comparison also bundles several differences. Burger’s sample included women and older adults, was screened and volunteered in a different place and period, and received explicit withdrawal assurances and a weaker sample shock. These changes were ethically purposeful, but mean that “year” was not isolated in a controlled experiment. Similar results are informative without demonstrating that social attitudes or all individual differences had remained unchanged.

Finally, an unusual laboratory task involving scientific authority differs from real employment, military or political situations. Actual harmful orders can involve ideology, personal reward, continuing relationships or punishment for refusal. The study helps identify a mechanism to investigate; it does not provide a complete explanation of complex atrocities. Conversely, its artificial setting does not mean the observed decision was fabricated: participants acted within the situation they believed they were in.

An evidence-based judgement is that Burger found substantial willingness to continue at an ethically limited decision point and no detected benefit of his particular peer-refusal manipulation. The safeguards show how scientific questions and participant welfare can reshape research together. The costs are narrower behavioural conclusions and qualified comparison with Milgram, not permission to recreate the procedure in a student investigation.