Doubting the power of prestige: obedience to authority beyond institutional and research justifications

Abstract

Stanley Milgram’s groundbreaking research on obedience to authority remains a foundational study in social psychology. His work revealed that a significant majority of individuals when instructed by an authoritative figure, were willing to administer what they believed were severe electric shocks to another person, all in the context of a supposed learning experiment. Given the high levels of stress Milgram imposed on participants, ethical concerns now make a direct replication of his study untenable. In 2009, however, Burger introduced a modified version, often referred to as “obedience lite,” which halts the procedure after participants administer a 150-volt shock. Both Burger and subsequent researchers have replicated Milgram’s experiment using this modified approach, recruiting participants via newspaper advertisements and utilizing a large shock-emitting device similar to Milgram’s original setup. However, questions remain as to whether such replication studies should account for the technological advancements that have occurred over the decades. This article presents a study in which participants were recruited online, and a smaller, computer-connected device was used to simulate the electric shocks. The research also explored the impact of different environments (university vs. non-university settings) and the framing of the study (scientific vs. marketing) on participants’ behavior. The findings revealed that the levels of obedience closely mirrored those observed by Milgram in the 1960s, with neither the study’s location nor the framing significantly influencing participants’ actions.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-10331-4.

Introduction

A series of experiments conducted by Stanley Milgram^1,2, which revealed that the majority of participants, under the command of an experimenter, were willing to administer electric shocks of up to 450 volts to another person, has become one of the most well-known findings in social^3–5. Milgram’s research paradigm, on one hand, provided shocking insights into human obedience, yet on the other hand, it caused significant distress for the participants. Milgram⁶ himself acknowledged that the participants experienced intense negative emotions and extreme stress during the experiment. This was evident both through observations of their behavior and their statements during debriefing. Unsurprisingly, soon after Milgram published his initial findings on obedience to authority^1,2 serious ethical concerns about his experimental methods emerged^7–10. Critics not only pointed to the suffering participants endured during the experiments but also raised concerns about potential long-term psychological harm.

It is, therefore, unsurprising that the global community of academic social psychologists agreed not to pursue further research using the original Milgram paradigm due to ethical concerns. The validity of this decision is further supported by contemporary studies conducted in virtual reality settings, where detailed psychophysiological and neuropsychological measurements have been taken. These studies consistently reveal the intense negative emotional responses experienced by participants^11,18.

After years of reflection, Burger¹⁹ identified a critical turning point in Milgram’s original procedure: while participants in Milgram’s study could administer shocks of up to 450 volts in 15-volt increments, it was at 150 volts that nearly all participants hesitated and expressed reluctance to continue. This observation led Burger to modify the experiment, capping the voltage at 150 volts (the 10th switch) in his replication. While Burger focused on replicating Milgram’s Experiment No. 5⁶, which involved participants confronting the dilemma of whether to obey the experimenter or respond to the learner’s distress, European researchers such as Dolinski and Grzyb^20,21, and Dolinski et al.²², opted for a different approach. Rather than putting participants in a situation where they had to decide whether to continue administering shocks despite the learner’s pleas, they replicated Experiment No. 2 from Milgram’s 1974 studies, which does not include this element of explicit protest from the learner. By doing so, they sought to reduce the psychological burden on participants while still exploring obedience dynamics.

Conducting replication studies many years after the original research presents a range of challenges, particularly due to technological advancements over time. To begin with, Milgram recruited participants for his study through advertisements placed in the printed press, which was a common medium at the time. In today’s world, however, where only a small percentage of the population regularly reads printed newspapers, it raises the question of whether similar recruitment methods should be employed, or if modern platforms, such as online advertisements, would be more appropriate.

Another issue pertains to the equipment used in the study. In Milgram’s experiments, the participants administered shocks using a large, imposing box equipped with 30 parallel buttons—an apparatus that resembled mid-20th-century electrical devices. This visual and tactile aspect of the machine likely contributed to the participants’ perception of its authenticity. Decades later, one might question whether the same device should be used in replication studies or whether it should be updated to reflect contemporary electronic technology, such as a computer-based interface that mimics the original’s function while still simulating the capability of delivering electric shocks.

Both Burger¹⁹ and Dolinski & Grzyb^20,21 opted for a conservative approach in their replications: they recruited participants through press advertisements (although Burger also used some online recruitment and leaflets distributed in libraries, farmers’ markets, coffee shops, and community centers)¹ and retained the use of a bulky, 20th-century-style shock machine, an exact replica of the one used by Milgram. In contrast, the replication presented in this article takes a different approach, one that acknowledges the need to adapt the experimental conditions to the realities of modern times. In this study, participants were recruited exclusively online, and the shock device was housed in a small box connected to a laptop, reflecting contemporary electronic aesthetics while still maintaining the illusion of delivering electric shocks.

The role of a university’s prestige in inducing obedience

Although Milgram’s obedience studies are often referred to as a single “experiment,” it is more accurate to describe them in the plural, as Milgram conducted a series of experiments, each with slight procedural modifications^6,23. One of the more well-known variations involved changing the location where the experiment was conducted. In Study 10⁶ Milgram hypothesized that participants’ obedience might be heavily influenced by the prestigious setting of the experiment, which in his case was the Department of Psychology at Yale University. Participants likely assumed that the responsibility for what transpired in the laboratory rested not with them, but with the authority figures conducting the experiment within such a respected academic institution.

This sense of transferred responsibility becomes harder to maintain in settings that lack the prestige and institutional authority of a renowned university. To test this, Milgram moved the experiment to the industrial city of Bridgeport, where he rented office space for the study. Participants were told that the research was being conducted by the now-defunct organization “Research Associates of Bridgeport.” In this less prestigious, more anonymous setting, Milgram aimed to determine whether obedience levels would differ when the participants could not rely on the implicit authority and reputation of a prestigious institution like Yale to absolve them of personal responsibility for their actions.

The obedience observed among individuals tested in Bridgeport was found to be slightly lower (48%) than that of participants tested in the Department of Psychology at Yale University (65%). Notably, even under these conditions, nearly half of the individuals exhibited complete obedience. In our study, we aimed to replicate this finding to determine whether relocating the research from a university setting to a space not associated with academic research would reduce participant obedience. It is also important to highlight that in Milgram’s experiments, despite the variation in locations (Yale University vs. Bridgeport), participants were informed that the research was of a scientific nature. This raises the question of how crucial this aspect of experimental instruction is. It is possible that if participants were told that the experiment was not intended to advance scientific knowledge, their obedience would significantly decrease.

To investigate these issues, the ideal design would have been a 2 (setting: prestigious vs. non-prestigious) × 2 (goal: scientific vs. commercial) factorial experiment. However, in Poland—the country where the study was conducted—universities do not typically engage in commercial research in the social sciences. As a result, a prestigious setting paired with a commercial goal might have appeared unusual or even suspicious to participants. This could have introduced an unwanted confounding variable into the design. Therefore, we decided to exclude this condition from the study.

Thus, we established three experimental conditions for our study:

1. The research is conducted at the university and framed as scientific in nature;

2. The research is conducted outside the university premises but is still presented as scientific;

3. The research is conducted outside the university premises and framed as non-scientific, with a commercial purpose.

Study

Participants

Participants were recruited through an advertisement posted on a popular online portal frequented by individuals seeking additional employment opportunities. The ad specified that candidates were needed for a study on memory, provided a contact phone number, and detailed the compensation (100 PLN, approximately 25 USD). Those interested were asked to call the provided number, where they underwent a screening process. Potential participants who were students or graduates in social sciences, as well as individuals with prior knowledge of psychological experiments, were excluded from the study.

Simmons, Nelson, and Simonsohn²⁴ recommend a minimum of 20 participants per condition in psychological experiments, while also emphasizing that larger sample sizes help guard against questionable research practices. For several reasons, we chose to meet only this minimum threshold, with 20 participants assigned to each experimental condition. This decision was influenced not only by the significant time and logistical challenges involved in conducting research based on the Milgram paradigm, but more importantly, by ethical considerations. Despite implementing various measures to minimize participant discomfort—including terminating the study once the “teacher” pressed the 10th button, using Milgram’s⁶ Experiment No. 2 instead of the more stressful No. 5, conducting thorough debriefings, and offering access to psychological support even long after participation—it remains clear that participants experience stress and emotional discomfort. Limiting the sample size to the recommended minimum helped reduce the overall number of individuals exposed to this distress.

Although we did not formulate specific hypotheses regarding gender differences, we aimed for an equal distribution of men and women, recruiting 30 women and 30 men for the study.

Method

The study was conducted using the “obedience lite” procedure developed by Burger¹⁹, which offers significant ethical improvements over the original Milgram procedure by substantially reducing participant discomfort and stress. First, the maximum “shock” administered was limited to 150 volts (corresponding to button 10 on the apparatus, which is why our device only included 10 buttons). Second, the procedure significantly shortened the experiment’s duration, thereby limiting participants’ exposure to stressful stimuli. Importantly, the “obedience lite” method eliminates the potential for serious health risks to the “learner,” a concern raised in Milgram’s original study.

To further minimize ethical concerns, we adopted the design of Milgram’s Experiment 2, rather than Experiment 5, as Burger did. In Experiment 5, the learner requests to leave the laboratory at a certain point, which introduces an additional stressor for participants. We avoided this element to reduce participant discomfort.

Participants were randomly assigned to one of three experimental conditions. Individuals allocated to Condition 1 were invited to a university facility, whereas those assigned to Conditions 2 and 3 were invited to a private apartment located on the third floor of a townhouse in the city center. In all cases, participants were provided with the appropriate address and scheduled for a specific appointment at a mutually convenient date and time.

Randomization was based on a pre-determined allocation list, ensuring balanced assignment across gender and conditions. For example, the first eligible woman was assigned to Condition 2, the first eligible man to Condition 1, the second woman to Condition 2, the second man to Condition 3, the third woman to Condition 1, and so on.

In the first and second groups (E1 and E2), participants were told that the experiment was of a scientific nature and part of the university’s academic research activities. In the third group (E3), participants were informed that the study was commissioned by a marketing company, designed to examine how consumers remember advertising content.

In all conditions, the experiment began with the experimenter (a white male 41 years old) greeting the participants and providing them with PLN 100 as compensation for their participation. It was made clear that this payment was a reward for attending and that they could withdraw from the study at any time without forfeiting the payment. Participants were then asked to read and sign an informed consent form, which reiterated their right to discontinue participation at any point during the experiment. All participants had signed the consent form.

The experimenter then explained that the study, which aimed to explore the determinants of learning, required one person to take on the role of a “learner” and the other to act as a “teacher.” To facilitate this, participants were presented with two tickets, with the explanation that one ticket was labeled “teacher” and the other “learner.” However, in reality, both tickets were marked “teacher,” ensuring that the participant was always assigned the teacher role. Meanwhile, a confederate of the experimenter (a white male, 23 years old) pretended to have drawn the ticket marked “learner.”

At this stage, the instructions provided by the experimenter varied depending on the experimental condition. In the first and second groups (E1 and E2), the experimenter emphasized the scientific nature of the study, reinforcing that it was part of university research. In the third group (E3), the experimenter explained that the study was commissioned by a marketing firm and aimed at investigating how consumers remember advertising messages.

In the groups where the study was commissioned by the University, participants heard:

‘We are interested in how psychologists want to understand how people memorize associations between product and its property, such as location or prestige. The idea is that when we know this, our knowledge about human beings and human memory expands. This way, science advances and moves forward.

When the study was commissioned by the Marketing Agency, participants heard:

‘We are interested in how advertising agencies need to know how people memorize associations between product and its property, such as location or prestige. The point is, when they know this, they can create more effective advertisements and better reach their audience.

In all variations, the experimenter’s assistant was supposed to learn the following associations:

• Margarine - Plant-Based.

• Bread - Crunchy.

• Car - Economic.

• Smartphone - Multimedia.

• Television - Panoramic.

• Coffee - Aromatic.

• Waffle - Crunchy.

• Vegetables – Fresh.

Our choice of a different set of words from those used by Milgram⁶ stemmed from the desire to employ verbal stimuli that could be equally applicable in both scientific and commercial (marketing) research contexts.

While the learner was purportedly memorizing the aforementioned word pairs, the participant (the ‘teacher’) was introduced to a device designed to administer punishment for any mistakes made by the learner. It was explained to the participant that the researchers were particularly interested in examining how punishment influences the speed and effectiveness of learning. To test this, a specialized electrical device would be used, which was then demonstrated to the participant. The experimenter proceeded to attach electrodes to the participant’s wrist and pressed the third button on the electric current generator (labeled with a 45-volt symbol). Although the resulting shock was unpleasant, it was not excessively painful.

In all three conditions, the same equipment was used - a modern version of an electric shock device. It was a small black plastic box the size of a book (17 × 12 × 5 cm), equipped with inputs and outputs for electrical cables and electrodes designed to be attached to the wrist of the person receiving the shock. The device was controlled through an application installed on a laptop (see: Fig. 1). Both the device and the application were created with the aim of increasing experimental realism.

Fig. 1.

The control panel for the electric shock generator, as seen on the screen of the laptop.

After approximately six minutes into the learning session, the experimenter and the participant (acting as the ‘teacher’) entered the room where the learner was seated and attached electrodes to the learner’s wrist. This marked the beginning of the main phase of the study.

Similar to Burger’s original experiments, at pre-determined questions (numbers 8, 12, 16, 18, 23, 28, 31, 35, 39, and 41), the ‘learner’ made a mistake (for example, instead of answering “vegetables” to the question, he responded with “healthy” instead of “fresh”). After each mistake, the experimenter instructed the teacher to increase the intensity of the shock displayed on the laptop screen and press the ‘shock’ key. Each subsequent mistake required an increase in the shock voltage (from 15 to 150 volts). After each shock above 45 volts, the ‘learner’ would start screaming, indicating clear signs of pain. If the ‘teacher’ refused to press further buttons or engaged in a discussion about the ethics of their actions, the experimenter applied 4 prompts (analogous to those used by Milgram):

• “Please continue.”

• “The experiment requires that you continue.”

• “It is absolutely essential that you continue.”

• “You have no other choice; you must go on.”

In the event of four refusals to follow the command or if the button was pressed for the 10th time (demonstrating obedience to ten consecutive shock commands), the experiment was terminated.

Following the conclusion of the study, a very thorough debriefing was conducted. Firstly, it was ensured whether the participants believed that they were indeed administering electric shocks to the person sitting behind the wall (none of the participants verbally expressed any doubts that this was the case, but researchers watching the recordings after the experiment had the impression that two people said it hesitantly). Subsequently, the details of the experimental procedure were explained, and apologies were offered for the deception. Participants were informed that in such studies, most people go as far as pressing the final button. However, they were also told that this experience could serve as a life lesson, highlighting that, at times, individuals may succumb too easily to the pressure of others. Any questions from the participants were addressed, and they were encouraged to contact the experimenters by phone or email if they had any further questions or concerns (a specific phone number and email address were provided). The entire debriefing procedure was carried out by a trained clinical psychologist.

The procedure was prepared in accordance with the Declaration of Helsinki. The full experimental procedure had been previously consulted with the Ethics Committee of SWPS University and received its approval (decision 12/E/02/2018).

Results

A total of 60 participants took part in the study, 30 of whom were women. However, upon reviewing the recordings in detail, we found indications that two participants (one woman and one man) may have questioned whether the learner was truly receiving electric shocks. We reached out to these individuals for clarification, but their inability to provide a definitive statement on the matter led us to exclude them from the analysis. As a result, the final analysis was based on the reactions of 58 participants. It is worth noting, that to reconcile methodological rigour with ethical obligations, the institutional ethics committee required the planned Milgram-paradigm replication to reduce its sample from the originally proposed 150 to 60 participants (20 per experimental condition). The committee emphasised that, even with all available safeguards, each session inevitably exposes participants to a transient but meaningful degree of emotional strain. Consequently, lowering the sample served to curb the prevalence of this discomfort across the population while still preserving a minimally sufficient—though not optimal—statistical power for hypothesis testing.

Participants were assigned to one of three experimental conditions, which differed based on the study’s location and, in cases where it took place outside the university building, the stated purpose—either scientific or commercial. In a separate study using an independent sample, we confirmed that these variations elicited distinct and systematically aligned evaluations of each situation regarding perceived prestige, importance, and significance of the research. The detailed results of this study are presented in the Supplementary Material 1 – Tables S1 and S2.

The exact number of individuals who agreed to reach the 10th button (indicating complete obedience to the experimenter’s instructions) was recorded in the Table 1.

Table 1.

Number of participants refusing/showing total obedience in each experimental group.

Group	Obedience		Total
	Yes	No
E1 (University/University)	18	2	20
E2 (Private flat/University)	16	2	18
E3 (Private flat/Marketing company)	16	4	20
Total	50	8	58

The differences between the groups turned out to be statistically insignificant (χ2 = 0.99, df = 2, p = .607, Cramer’s V = 0.131). Given that one of our groups had fewer participants than the others, we also considered alternative ways of presenting the results – see Supplementary Material 2. We also examined the differences in obedience between women and men, and these differences also proved to be insignificant (χ2 = 0.58, df = 2, p = .446, Cramer’s V = 0.1). In the study, we also recorded the last switch at which participants stopped, and here too, no differences were observed (F = 0.390, df = 2,33, p = .68). The average switch, along with a 95% confidence interval for each group, has been presented in Fig. 2.

Fig. 2.

The last switch pressed by participants in each experimental group.

We also conducted a detailed analysis of more subtle indicators of resistance to authority observed among participants across the experimental conditions. Specifically, when participants expressed verbal hesitation about continuing their participation or responded passively—by failing to press the button—the experimenter issued the first standardized prompt, as outlined in the procedure: “Please continue.” If the participant again failed to comply with the command to press the button, the experimenter proceeded to deliver the next prompt in the predetermined sequence, and so on. The experiment was terminated only when a participant failed to obey the command even after the final prompt had been delivered (“You have no other choice; you must go on.”).

Because the number of prompts delivered by the experimenter reflects the participant’s resistance to complying with authority, it serves as a proxy for disobedience—or at least a tendency toward disobedience. Therefore, we compared the number of prompts required in each of the three experimental conditions. A Welch’s one-way ANOVA indicated no significant difference in the number of prompts between groups, F(2, 36.5) = 0.51, p = .60. The details (mean and SD for each group) are in the Table 2.

Table 2.

Average number of prompts in each experimental group.

Group	Mean	SD
E1 (University / University)	0.550	1.28
E2 (Private flat / University)	0.556	1.29
E3 (Private flat / Marketing company)	0.950	1.47

Discussion

Milgram’s series of landmark experiments^1,2,6 revealed the remarkable tendency of individuals to obediently follow commands from authorities, even when those authorities hold no formal power over them. Rigorous replications of these studies, employing similar equipment to administer electric shocks to others, have consistently validated these findings—both in the years immediately following Milgram’s original work (see Blass, 1991 for review) and more recently^19,25. In the present study, we demonstrated that this pattern of high obedience persists when both the recruitment methods and technical aspects of the experiment—such as the device used to deliver electric shocks—are updated to reflect 21st-century conditions, rather than those from more than 50 years ago. Therefore, this “modern” replication further confirms the enduring and universal nature of the phenomena first observed by Milgram⁶.

However, our study had an additional objective. We aimed to explore the potential relationship between the information provided about the study’s purpose and its location, and the participants’ level of obedience. Participants were divided into three groups that differed in both the stated purpose of the research (scientific versus marketing) and the location of the study (a private apartment rented by a marketing agency versus a university building). In all conditions, participants were asked to administer electric shocks to other ‘participants’ (in reality, a confederate) as a form of punishment. The results revealed no significant differences in obedience across the different experimental conditions. Whether the purpose of the research was framed as scientific or related to effective advertising, and whether the study took place in a private apartment or a university setting, participants displayed a similar level of obedience to that observed in Milgram’s classic experiments and their later replications. This similarity applied both to the percentage of participants who pressed all ten buttons and to the average voltage applied to the supposed learner. Similarly, the number of prompts the experimenter had to issue to the participant did not vary across experimental conditions.

It is important to acknowledge that the absence of differences in participant behavior across the three conditions may be due to the ‘obedience-lite’ procedure we employed¹⁹. In Milgram’s original research⁶, as well as in various studies using this paradigm from the same era²⁶, the vast majority of participants across different experimental conditions agreed to press the 10th button on the electric current generator, labeled with the 150-volt symbol. This same pattern was observed in Burger’s¹⁹ study conducted in the USA and in more recent European studies²⁵. Similarly, in the current study, the vast majority of participants displayed this behavior. Of the 58 individuals whose responses we analyzed, only eight refused to press all ten buttons on the electric generator. In Milgram’s⁶ experiments, participants who refused to continue most often did so in the later stages of the study. For instance, in his Experiment #10 seven out of 40 people refused to press button number 11, while only three people refused to press button number 10 (with two of these participants declining to press even the first button). It is possible that if our participants had been required to press an additional 20 buttons, with the voltage increasing to 450 volts, differences in obedience might have emerged across the three conditions. However, for ethical reasons, this cannot be tested.

An important point that warrants separate attention is the absence of significant behavioral differences among participants across the three experimental conditions. We wish to clearly acknowledge that a failure to detect differences is not equivalent to evidence of their absence. It is possible that, had a much larger sample size been used, some differences between the conditions might have emerged. However, even if such effects were to appear with a larger sample, their magnitude would necessarily be small. Therefore, our findings suggest that either the location and stated purpose of the study have no meaningful impact on participants’ obedience, or—if such an effect does exist—its strength is minimal.

This raises the important question of what other factors might reduce the level of obedience observed in such paradigms. Milgram⁶ already demonstrated that the spatial arrangement of the experimental setting plays a critical role: obedience tends to decrease when there is no physical barrier separating the participant from the “learner.” Dolinski and Grzyb²⁷ argued that this finding aligns closely with the concept of an avoidance–avoidance conflict, as introduced by Lewin²⁸. In this framework, participants are caught between two highly aversive options: either inflicting pain on the learner or disrupting the experiment being conducted by the authority figure who recruited them. In such a dilemma, the participant is motivated to avoid both outcomes. However, under the standard conditions of Milgram’s experiment—where the learner is positioned behind a wall and thus not directly visible—the harm done to the learner remains abstract and psychologically distant. In contrast, disobeying the experimenter, who is physically present and invested in the study, feels more immediate and personally uncomfortable. As a result, participants tend to resolve the conflict by following the experimenter’s instructions. When the learner is physically present in the same room and his suffering is directly observable, this balance shifts. The aversiveness of harming a visibly distressed individual becomes more salient than the discomfort associated with defying the experimenter. In such settings, participants are more likely to prioritize the well-being of the learner and refuse to continue the procedure. Dolinski and Grzyb²⁷ not only replicated Milgram’s finding in this regard but extended it by introducing a condition in which the learner and teacher were located in the same room, while the experimenter was situated in a different one. Under these circumstances, obedience declined even further, suggesting that the relative proximity and visibility of the victim and authority figure significantly shape the participant’s behavioral choices.

Complementing these results, Grzyb and Dolinski²⁹ showed that inducing strong emotional empathy toward the learner also reduces obedience. In such cases, the avoidance–avoidance conflict is more frequently resolved by choosing to disobey the experimenter, since the alternative—inflicting harm on someone with whom one empathizes deeply—becomes particularly aversive. These findings reinforce the notion that both spatial dynamics and emotional factors can critically influence the likelihood of obedience within authority-based paradigms.

The assumption that participants in Milgram’s paradigm experience an avoidance–avoidance conflict also opens the door to generating additional hypotheses about factors that may reduce obedience. For example, if the experimenter were to behave in a rude or impolite manner, this might make it psychologically easier for participants to choose the option of making the experimenter uncomfortable—thereby increasing their likelihood of refusing to administer electric shocks to the “learner.” Whether such manipulations would indeed lead to decreased obedience remains an open question that warrants future empirical investigation.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

TG and DD were responsible for the experimental design and wrote the main manuscript text. MSM, GK, and ŁM were responsible for conducting the experiment, gathering the data, and preparing the data for analysis.

Data availability

Data availability statement: Documents associated with the project (including data) are available at: https://osf.io/kumwy/.

Declarations

Competing interests

The authors declare no competing interests.