The Stanford Prison Experiment was wrong

The replication crisis in psychology has called into question some of the most famous studies in the field, including such stalwarts of undergraduate classes as the Stanford prison experiment and the Stanford marshmallow experiment (the connection with Stanford in both cases is a coincidence). In the case of the latter, a conceptual replication in 2018 failed to find many of the original longitudinal effects, undermining previous claims that a person's ability to inhibit the urge to eat a marshmallow as a child predicts various later life outcomes. The charges against the Stanford prison experiment (SPE) are more serious. Whereas the SPE has traditionally been claimed to show how ordinary people with no history of violence would come to act in brutal and dehumanizing ways when placed into particular situations, recent analysis of archival materials from the study by Thibault Le Texier suggests that this conclusion was reached fraudulently.  

A recent Psychology Today article describes Le Texier's conclusions:

"Zimbardo had actually decided in advance what conclusions he wanted to demonstrate. For example, on only the second day of the experiment, he put out a press release stating that prisons dehumanize their inmates and therefore need to be reformed. Moreover, contrary to his repeated claims that participants in the experiment assigned to the role of guards were not told how to treat the prisoners and were free to make up their own rules, the archival data clearly show that the guards were told in advance what was expected of them, how they were to mistreat the prisoners, and were given a detailed list of rules to follow to ensure that prisoners were humiliated and dehumanized."

Common sense would suggest that the discovery that results cannot be replicated or were fraudulent should have implications for psychological theory and teaching. We should not teach students conclusions that have been undermined or shown to be fraudulent. Nor should such findings be used to prop up theories.

In the case of the marshmallow task, I agree. Of course, it's a moot point because I haven't covered that study in recent years (I find it to exude self righteousness on behalf of the studious professional class whose achievements it ostensibly explains). But if I had, the largely failed replication would cause me to stop.

The case for dropping the SPE might seem to be even stronger. It's findings were fraudulent, after all. Yet, I don't see myself doing this, for reasons that I think are significant for understanding the underlying issues at stake with the replication crisis, and with the possibility of a science of psychology in the first place. 

While the replication crisis groups both experiments together as examples of (ostensibly) scientific psychology, the claims that they make are quite different. The longitudinal findings from the marshmallow experiment pertain to correlations amongst aggregated groupings of human behavior. They are essentially statistical facts, largely devoid of a narrative of human intentionality. The Stanford Prison experiment, on the other hand, lends itself to narration. It is a story, of particular people in a particular type of situation acting in particular ways. 

The recent revelation about the SPE mean that this story is fiction. The "John Wayne"-style guard who initiated the pattern of dehumanizing abuse did not, as the traditional story holds, do so spontaneously due to the situation, but at Zimbardo's behest. 

What is crucial here is that, in the case of the SPE, what is fake/fictional is a narrative, rather than a statistical or propositional conclusion. A proposition that is fake can be disregarded. The same cannot be said for a narrative. As Jerome Bruner explains, the value in a narrative is not a function of its veracity, but its versimilitude. A narrative that is false may nevertheless be lifelike and compelling, and as a result, worth telling. What matters in narrative is not that what is described did happen, but that it could happen. This is why the claims of fraud by Zimbardo do not make the study (at least in its general form) irrelevant.

This point has significant implications for psychology. In science, conclusions must be drawn from empirical evidence. A physicist could not derive a new theory from how they imagine the physical world works. History provides clear examples of why this would fail. If we didn't know better, we would likely imagine (as people did before Newton) that heavier objects necessarily fall faster. In the case of human behavior, things are different. We can narrate imaginary accounts of how people act, and these accounts may be compelling--in fact, they may even allow us to know people better than we did before--even when they are known to be fake.

There are other relevant issue here for the replication crisis. Given the strange fact that we can deepen our knowledge of the human mind from fictional accounts of the behavior of imagined people, it is somehow fitting that empirical records of actual behavior may function to undermine their own validity. So, the act of discovering and reporting that people in a given situation respond in a particular way can lead to people choosing, in response, to act in a different way.