To prep for an upcoming course, Penn researcher David Lydon-Staley decided to conduct an experiment: Might melatonin gummies—supplements touted to improve sleep—help him, as an individual, fall asleep faster? For two weeks, he took two gummies on intervention nights and none on control nights. The point, however, wasn’t really to find out whether the gummies worked for him (which they didn’t), but rather to see how an experiment with a single participant played out, what’s known as an “n of 1.”

Randomized control experiments typically include hundreds or thousands of participants. Their aim is to show, on average, how the intervention being studied affects people in the treatment group. But often “there’s a failure to include women and members of minoritized racial and ethnic groups in those clinical trials,” said Dr. Lydon-Staley, an assistant professor at the Annenberg School for Communication. “The single-case approach said, instead of randomizing a lot of people, we’re going to take one person at a time and measure them intensively.”

In Dr. Lydon-Staley’s spring semester class, Diversity and the End of Average, seven graduate students conducted their own n-of-1 experiments—on themselves—testing whether dynamic stretching might improve basketball performance or whether yoga might decrease stress. One wanted to understand the effect of journaling on emotional clarity. They also learned about representation in science, plus which analytical approaches might best capture the nuance of a diverse population and individuals with many intersecting identities.

“It’s not just an ‘n of 1’ trying to do what the big studies are doing. It’s a different perspective,” said Dr. Lydon-Staley. “Though it’s just one person, you’re getting a much more thorough characterization of how they’re changing from moment to moment.”

Second-year doctoral student Adetobi Moses described the different options, including her choice of stream-of-consciousness journaling, then talked through her two-week experiment. In the end, her data showed that the writing helped only minimally with her emotional clarity. But the process itself? She found it empowering, a sentiment that others in the room echoed. Despite experimental results that may have lacked statistical significance, the grad students appreciated gaining deeper insight into an aspect of themselves.

Sometimes the results surprised them, too, like those of Darin Johnson, a third-year PhD student studying code-switching. For his experiment, he wanted to understand whether reducing social media use on his phone would drop his stress level. “I follow a lot of social justice–oriented pages, which include a lot about racism and police brutality. I would just sit there scrolling and be stressed out,” he said. He thought removing the input that caused these reactions might prevent the anxiety associated with them.

On intervention days, he would receive a notification when he reached the time limit that he’d set. On control days, his access remained unlimited. At the end of each day, he took a survey that he’d created. Before even analyzing his data, he realized that avoiding social media didn’t actually help him but instead made him feel isolated, cut off from his circle. The notion of n-of-1 experiments often raises eyebrows, said Dr. Lydon-Staley. “I’m on the fringe here, but I think it’s the way to go,” he said. “Randomized control trials give you a statistic for the ‘average person,’ but that’s a statistical artifact that doesn’t exist. I want to know what works for me.”

Dr. Lydon-Staley applies this framework to much of the research conducted in his Addiction, Health, & Adolescence Lab. For example, in a project about smoking cessation, he and his team are collecting reports from participants about their specific withdrawal symptoms—cravings, irritability—ten times a day for ten days, before and during a quit attempt.

Personalized medicine has already moved in this direction, using genetics and other biomarkers to guide treatment. “So many people deal with medical issues that may not have a one-size-fits-all solution,” said second-year PhD student Baird Howland, who is also in the class. “Anybody could, in theory, do this type of experiment to figure out what works and what doesn’t for them.”

Dr. Lydon-Staley sees great potential in the ability to scale up the single-case approach: Collect enough samples and patterns will emerge revealing natural rather than artificial clusters. “Often, you can’t take an intersectional approach with statistics,” Mr. Johnson said. “People might aggregate by race or by gender. Those are disparate categories, but I’m gay and Black. If I were to do a statistical analysis, I’d have to separate them out, and ‘n of 1’ allows us not to.”

Adapted from a Penn Today article by Michele W. Berger, May 3, 2022.