Advertisement
Research Article

A Virtual Reprise of the Stanley Milgram Obedience Experiments

  • Mel Slater mail,

    To whom correspondence should be addressed. E-mail: melslater@gmail.com

    Affiliations: Department of Computer Science, University College London, London, United Kingdom, Institució Catalana de la Recerca i Estudis Avançats ‐ Universitat Politècnica de Catalunya, Virtual Reality Centre of Barcelona, Barcelona, Spain

    X
  • Angus Antley,

    Affiliation: Department of Computer Science, University College London, London, United Kingdom

    X
  • Adam Davison,

    Affiliation: Department of Computer Science, University College London, London, United Kingdom

    X
  • David Swapp,

    Affiliation: Department of Computer Science, University College London, London, United Kingdom

    X
  • Christoph Guger,

    Affiliation: Guger Technologies OEG, Schiedlberg, Austria

    X
  • Chris Barker,

    Affiliation: SubDepartment of Clinical Health Psychology, University College London, London, United Kingdom

    X
  • Nancy Pistrang,

    Affiliation: SubDepartment of Clinical Health Psychology, University College London, London, United Kingdom

    X
  • Maria V. Sanchez-Vives

    Affiliation: Instituto de Neurociencias de Alicante, Universidad Miguel Hernandez - Consejo Superior de Investigaciones Científicas, Campus de San Juan, San Juan de Alicante, Spain

    X
  • Published: December 20, 2006
  • DOI: 10.1371/journal.pone.0000039

Reader Comments (8)

Post a new comment on this article

very interesting, but lots of questions remain

Posted by Eveaustria on 26 Dec 2006 at 19:30 GMT

So it does make an important difference if simulations seem to be real or are just "abstract". This confirms that games do have stronger effects on the players the more realistic they look/seem.

Still lots of important facts are missing yet:
* what kind of participants were chosen? How old? Of what "social status"? How many females/males? What IQ did they have?
* how much have the participants already been used or not used to virtual experiments - and of what kind? Did participants stop the experiment earlier if they were not used to playing 3D-computergames? Did participants show lesser signs of stress when "killing" the virtual person if they have already gotten used to killing virtual characters in 3D-games before?
* why only 23/11 participants? Isn't this a bit meager?
* what exactly were the participants told about the experiment? Did they know that they themselves were being tested? Were they told that only their personal "stress" was being measured so it was very important that they went through it all until the end?
* have the participants ever heard of the very well-known Milgram-experiment before? How many did not?
* if the participants did know about the Milgram-experiment, wouldn't they have wanted to prove their "humanity" and self-determination by stopping the experiment? Was this considered?
* if they were not informed but just "repeated" the Milgram experiment virtually, why did no participant claim that the experiment was useless because a virtual person wouldn't feel that punishment so there would be no data how punishing "it" could change learning-curves - or did any participant mention that?
* how did the "authority" act in this experiment? Did this make any difference? Was the stress lowered by an authority stating that the participant would not be held responsible for anything in this experiment, even so it was "only" virtual?
* what reasons did the participants give when asked why they completed the experiment - or why some would rather have wanted to stop it? What were their answers? Did they feel "guilty" somehow? Did they claim that it seems wrong to kill or induce pain even if it was done to a virtual being?


RE: very interesting, but lots of questions remain

melslater replied to Eveaustria on 29 Dec 2006 at 14:18 GMT

* So it does make an important difference if simulations seem to be real or are just "abstract". This confirms that games do have stronger effects on the players the more realistic they look/seem.

The issue of realism is an interesting one. To date the evidence in the presence research literature does not support the notion that greater (visual) realism results in greater presence. See the review in

Sanchez-Vives, M.V. and M. Slater, From Presence to Consciousness through Virtual Reality. Nature Reviews Neuroscience, 2005. 6(4): p. 332-339.


* what kind of participants were chosen? How old? Of what "social status"? How many females/males? What IQ did they have?

This is discussed in the first section of Methods. Of course we do not measure people’s IQ! This table gives the status of the participants:

Category Frequency
1. Undergraduate Student 11
2. Masters Student 5
3. PhD Student 3
4. Research Assistant/Fellow 4
5. Staff - systems, technical 0
6. Faculty 0
7. Administrative Staff 6
8. Other 5

* how much have the participants already been used or not used to virtual experiments - and of what kind?

On a scale of 1 to 7, where 1 = never experienced virtual reality before, to 7 = extensive experience, the results were:

1 19
2 7
3 4
4 3
5 1
6 0
7 0


* Did participants stop the experiment earlier if they were not used to playing 3D-computergames? Did participants show lesser signs of stress when "killing" the virtual person if they have already gotten used to killing virtual characters in 3D-games before?

No there is no evidence of this. In fact the regression analysis reported in the Section ‘Skin Conductance Reponses’ and the associated supporting information suggests the contrary, that there was a positive association between past computer game playing and the degree of arousal (just by looking at the skin conductance we cannot know that this was specifically stress, but it is likely to be so). It is important to note that this experience is very different from a computer game at several levels:

• The experience is life-size – unlike a computer game where the characters are finger sized on a display monitor, here the virtual character was life-sized, and could literally look the participant in the eye.

• In computer games the virtual characters who are killed do not plead for mercy, argue about the situation, or demand the termination of the game.

• As discussed in the section ‘Speculations on Obedience in Virtual Reality’ this was not set up like a computer game. Had the instructions to the participants been on the lines of: ‘This is a game where you win points by shocking the character as much as you can’, then the results might have been quite different.

* why only 23/11 participants? Isn't this a bit meager?

Yes this is true, more would have been preferable. It is a question of finite resources, and the huge effort financially, physically and emotionally that this experiment involved. In terms of the statistics most of the results are well into the range that would normally be considered as ‘highly significant’.

* what exactly were the participants told about the experiment? Did they know that they themselves were being tested? Were they told that only their personal "stress" was being measured so it was very important that they went through it all until the end?

The complete sets of instructions are available on

http://www.cs.ucl.ac.uk/s...
or
http://www.lsi.upc.edu/~m...

The instructions were given to the participants to read, and also read out to them, and repeated until they said that they were clear.


* have the participants ever heard of the very well-known Milgram-experiment before? How many did not?

This is discussed in the Section ‘Early Withdrawal’.

* if the participants did know about the Milgram-experiment, wouldn't they have wanted to prove their "humanity" and self-determination by stopping the experiment? Was this considered?

This was true for one participant who told us in the interview afterwards that she had decided to withdraw early before the experiment even started. Another participant withdrew after only 6 shocks, and we were suspicious about that. We eliminated both from the data used to obtain the results.

In general the evidence does not point in the direction that people wanted to prove their humanity. Our opinion is that even knowing intellectually that the situation was not real, and even knowing abstractly about the Milgram experiment, most people cannot help but get caught up in what they are seeing and hearing, and automatically having responses appropriate to that situation. Knowing that the situation was virtual, however, is what kept some people going until the end. Please see the discussion in ‘Speculations on Obedience in Virtual Reality’.

* if they were not informed but just "repeated" the Milgram experiment virtually, why did no participant claim that the experiment was useless because a virtual person wouldn't feel that punishment so there would be no data how punishing "it" could change learning-curves - or did any participant mention that?

It was surprising, as mentioned in the section ‘Speculations on Obedience in Virtual Reality’ that no participant queried the premise of the experiment (i.e., that a virtual character had learned word-pair associations or could improve by punishment). This was probably partly due to the serious overall context established by the experimental procedures. In addition since everything was presented seriously it is likely that many of the participants attributed rational meaning (to their own satisfaction) to the idea that the virtual character might improve its learning through this method. Very few of the participants were from the computer science department itself (4, 2 of whom were administrators). In the pre-questionnaire we had asked the participants about their level of programming knowledge (1=novice, 7=expert). The results were as follows:

Computer Programming Knowledge
1 = Novice 10
2 4
3 4
4 1
5 9
6 1
7 = Expert 5

It should be noted that some reported getting angry with the virtual character at her failing to get the correct answers on some word sequences even when the whole cycle started to repeat again and where she had heard the correct answer earlier.

* how did the "authority" act in this experiment? Did this make any difference? Was the stress lowered by an authority stating that the participant would not be held responsible for anything in this experiment, even so it was "only" virtual?

The behaviour of the experimenter is discussed briefly in the paper. Due to ethical requirements the experimenter could act nothing like the experimenter in the original experiment by Milgram, who was able to insist that the subject “had no choice” but to continue with the shocks. In our case participants were informed beforehand that they had the right to stop without giving reasons, and if the participants queried their continued participation during the experiment, the experimenter said: “You can stop whenever you like. It would be best for the experiment if you continue, but you can stop whenever you like.”

* what reasons did the participants give when asked why they completed the experiment - or why some would rather have wanted to stop it? What were their answers? Did they feel "guilty" somehow? Did they claim that it seems wrong to kill or induce pain even if it was done to a virtual being?

There are video recordings of the interviews with the participants. These will be discussed in detail in a later publication (yet to be written). However, we can say that participants who continued to the end said that they did so, in spite of how uncomfortable they felt, because they knew it was virtual. On the other hand those who withdrew early said that they did so because they found the situation too unpleasant to continue.

The Authors, 28th December, 2006.