For an experiment that took place over 50 years ago, the infamous Milgram experiment has left a pretty long trail.
There are reasons for its longevity: It's fascinating; it took place during an era when psychology studies were allowed to go insanely overboard; it continually receives dramatic treatments (the latest being the Michael Almereyda film Experimenter). But the most substantial may be the perfect simplicity of its design.
The study consists of three parties: the experimenter, the teacher, and the learner. Before the experiment begins, the teacher and learner draw slips of paper to determine who plays which role. The teacher and learner are then separated into two rooms, the former accompanied by the experimenter. The stated goal is for the learner to memorize certain words; the penalty for failure is electric shocks. The teacher's role is to administer those shocks, increasing their power by whatever degree the experimenter tells them to.
If you've taken an Intro to Psychology class, you know the misdirect. There are no actual shocks. The study, in fact, is really about seeing how much the teacher is willing to harm another person if given instruction from an authority figure. (The premise was meant to somehow explain the mode of thinking that led to the Nazi atrocities of World War II.) The learner in the other room is just an actor, who screams bloody murder when given the fake shocks; the drawing slips of paper bit is a ruse.
There's a name for these "learner" roles in studies: confederates. In all cases, they're there to get a reaction out of the subject without revealing their hidden agenda. But are they always effective?
There's a long history of using confederates in psychology studies, and the reason is obvious. When you're in the field of social psychology—the study of how people interact with other people—it makes sense that you'd want to, well, investigate how people actually interact with other people. Having subjects do this with another actual person, then, is the most straightforward path to information gathering.
"When you use another human being as a manipulation, it quickly becomes complex. It's not just like dimming the lights in the room."
For the experiment to actually work, the subjects can't know what the experiment is trying to figure out. If they did know, their reactions wouldn't be accurate. That's where confederates come in. If researchers want to see how often people help strangers who fall, give a confederate some hidden kneepads. If you want to see how quickly a subject's cardiovascular system changes when interacting with someone who has a huge facial birthmark, stick some "port-wine stain" make-up on the confederate's face.
But using confederates may not always lead to the most accurate results.
"When you use another human being as a manipulation, it quickly becomes complex," says Anna Kuhlen, a neurocognitive psychologist at Berlin's Humboldt University. "It's not just like dimming the lights in the room."
In 2012, Kuhlen and Susan E. Brennan, a psychologist from Stony Brook University, published a paper looking into when confederates might be hazardous to your data. They broke down their concerns into four problematic confederate types: the biased, the covert, the know-it-all, and the scripted.
The first is a problem of too much information. Confederates will always have more information about the study than the subject—that's just a part of the process. But the worry in this instance is that the confederates also know what reaction the researcher is looking for. Trying to get genuine reactions when one party has an ulterior motive is not an easy task. "Even the most conscientious confederates are at risk of inadvertently shaping participants' behavior by giving verbal back-channels or nonverbal cues such as facial expressions, body posture, tone of voice, pauses, or eye gaze patterns," they write. The solution to this, then, is not telling confederates what action is being sought.
The flip-side to this is listed as "the covert" in the research paper, which is essentially the idea that the confederate is giving their role in the study away—through subconscious actions—and the subject is altering behavior because of it. Interestingly, the altered behavior usually sways toward the behavior that the researcher, and thusly the confederate, is looking for.)
Another big issue is simple repetition. In the "know-it-all" concern, the confederate's usually in a passive role, say, sitting in the room and listening to subjects repeat the same story. It'd seem that this activity is the easiest for a confederate to participate in—it's not as if the confederate has lines to remember—but, in fact, it can be more problematic. "If they already know in advance what the participant will be doing, it doesn't allow them to be truly engaged in their interaction," Kuhlen says. Imagine a friend telling you a story. Now, imagine that friend telling you the story a 40th time. Your reaction to the 40th time may not be nearly as enthusiastic. "Participants may pick up on that and also disengage, or treat the confederate as not a real partner but as part of the experiment," Kuhlen says.
The actual casting of confederates by researchers is another issue to consider: "You might be looking for a specific type of person, and the results wouldn't be generalized to the general population," Kuhlen says, "but only for confederates who show the one behavior you cast for."
The final concern listed in the paper is the confederates' actions being too scripted. In order to account for one, and only one, variable in the study, the experimenter will often give the confederate a tight script, coaching them to keep his or her behavior and word choice completely consistent. (In the Milgram experiment, the "experimenter" was given four scripted verbal cues in an attempt to urge the "teacher" to heighten the electric shock voltage.) "[Researchers should] make sure the confederate is not merely a puppet who performs a certain behavior,” Kuhlen says. "There's a genuine need for the confederate to react contingently to the participant."
Why? The reasons are twofold. First, unless you're dealing with a particularly talented actor, it's easy for the subject to know if the confederate is being disingenuous—which tends to elicit a suspicious reaction. But the second issue is perhaps more problematic. The entire point of social psychology is to study how people interact with one another in genuine circumstances. While the goal of simulating reality essentially goes out the window the moment a subject agrees to participate in a study, minimizing variables by heavily scripting the action mostly invalidates the study. It's like trying to study human interactions by watching an episode of Cheers. All you're really doing is studying how the writers, directors, and actors of the show believe humans interact.
In much the same way, unless researchers are extremely careful and judicious in their usage of confederates in studies, all they're doing is studying how they want humans to interact with one another. At that point, they might as well just publish a novel.