A scientific paper reads as a clean progression from question to result. And yet, research almost never happens that way. Experiments fail, conversations redirect projects, and none of it survives into print. When the founders of a field retire, their tacit knowledge goes with them, and nothing anywhere records that it existed.
Writing in Nature Physics , a team led by the University of Pittsburgh's Jeremy Levy and Chandralekha Singh describes the Inverse Seminar or "inverseminar," a format designed to recover some of the hidden knowledge and decisions that result in polished papers. In this format, a senior scientist is invited (usually by remote video conference) to participate without asking them to prepare. The scientist listens as their work is presented back to them by a junior scientist.
The format came out of the pandemic. When seminars moved online in 2020 and then stayed online, a talk that had merely been dull in a lecture hall became something worse. Cameras go dark, the audience becomes a grid of names, and people answer email while a colleague speaks into silence. Remote presence did not create the problem, it removed the last social pressure holding it back.
Inverting the seminar removes the possibility of drifting. The junior scientist is presenting a paper to the person who wrote it, which is as intimidating as it sounds and concentrates the mind accordingly. The audience has to follow closely enough to ask the next question. The senior scientist cannot fall back on a rehearsed talk, because nobody has told them which paper is coming or what will be asked about it.
What happens next, how the scientist reacts, is not a discussion the scientist has given 40+ times. Instead, the presentation becomes a trigger for episodic memory, a recollection of the events that preceded or took place during the struggle to produce new knowledge, rather than the organized version that repetition produces. Undocumented material is recovered or uncovered.
The senior scientists have been generous every time. Hearing your own research described by someone who had to reconstruct it from the outside turns out to be disarming rather than adversarial, and the reaction is almost always to supply what the presenter could not have known: the false start, the argument, the person whose name is on the paper for reasons the paper does not give.
"Every paper has a story that the paper itself does not tell," said Levy, Distinguished Professor of Physics at the University of Pittsburgh.
The research team revealed the accompanying https://inverseminars.org website they launched along with their "What physics papers don't tell you" paper, published July 31.
The team developed and tested the format in physics, where the example cited is Michael Crommie of the University of California, Berkeley. In 2010, his group published a result in Science on how graphene behaves under strain, and it started a subfield. The published account is orderly, each step following the last.
In the inverseminar, Crommie said the discovery happened "totally by accident." His group had been studying graphene for unrelated reasons when they saw something nobody could explain. Months of confusion followed, until the theorist Antonio Castro Neto "just happened to be in my office" and recognized what they were looking at. Castro Neto is a coauthor on the 2010 paper. Nothing in it indicates how he became one, or that the confusion occurred.
Crommie also described two separate projects, each running roughly two years, that did not work and were never published. No one can learn from them, because there is no record that they happened.
Nothing in the format depends on physics. It needs a body of published research, someone junior prepared to read it closely, and a senior scientist willing to sit and listen while it is described back to them. The research team ran seven inverseminars over 27 months, and the pattern held each time. Any field that publishes polished accounts of unpolished processes has the same distance between what its practitioners know and what its literature contains, and the incentives that produce the clean version are the same everywhere.
None of this is a perturbation of the existing seminar format. Moving the seminar to a different hour, or shortening it, or adding a discussion at the end, leaves it the thing it already was. Reversing who does the presenting changes what the hour is for, and reaching that arrangement means crossing a large intellectual energy barrier. It sounds wrong until you have sat through one, and everyone involved has to agree to be uncomfortable before anyone can find out that it works.
"Small adjustments to a format that is not working give you a slightly different version of the thing that is not working," Levy said. "It takes a leap of this kind to shake things up."
Alongside the paper, the team is launching the inverseminars.org site, open to researchers in any discipline. Each session receives its own citable DOI, so a recording becomes something that can be referenced the way a paper is. The Crommie session, 92 minutes, is available there.
Levy and Singh, also a Distinguished Professor of Physics and Astronomy at Pitt, struck upon inverse seminars amid meetings with students since 2023. Their coauthors are Aswini Ramankutty at Pitt, Jack Diab and Haiyue Huang at the University of California, Los Angeles, and Mamun Sarker, a chemist at the University of Nebraska-Lincoln.
