Gary Neuhoff outside his home on Prospect Street in Ypsilanti, Mich., on Wednesday, holding the ninety-one-word summary of his political position that he has carried in his wallet since March. Credit: Lowell Barrineau/The New York Time5
Gary Neuhoff outside his home on Prospect Street in Ypsilanti, Mich., on Wednesday, holding the ninety-one-word summary of his political position that he has carried in his wallet since March. Credit: Lowell Barrineau/The New York Time5

YPSILANTI, Mich. — The federal study that arrived on Capitol Hill last month was authorized under Section 9(d) of the 2022 reauthorization of the National Science Foundation’s Directorate for Social, Behavioral and Economic Sciences, a provision that received no floor debate, was inserted in conference at the request of a subcommittee staff director who has since left government, and appropriated $14.1 million over four years for what the statutory language describes only as “research into impediments to civic exchange.” The completed report was transmitted to the Senate Committee on Rules and Administration on Aug. 11 and referred, the same afternoon, to its Subcommittee on Legislative Branch Operations, where it has sat pending a markup that has been noticed twice and held once. It runs to 611 pages. Its central finding is that the American political conversation is not being avoided. It is being queued.

The report concerns, in substantial part, a 58-year-old commercial heating and cooling estimator named Gary Neuhoff, who lives on Prospect Street here, seven miles east of the laboratory that studied him, and who has not completed a political sentence in the presence of another adult since October 2011. Mr. Neuhoff is one of 4,117 subjects enrolled by the Center for Discursive Throughput at the University of Michigan in a procedure the paper calls full elicitation: repeated written sessions, conducted over a mean of six hours and forty minutes, in which a subject sets down every proposition he holds about American political life, together with the relations among those propositions, until additional sessions produce no new material. The resulting document, which the paper terms the subject’s reference position, averaged 2,914 discrete propositions and 13,800 dependency relations among them. Having obtained it, the researchers asked each subject to say it out loud.

None of them could. At a measured mean delivery rate of 138 words per minute, the complete serial transmission of an average reference position would require nineteen hours and forty minutes of uninterrupted speech. The median political conversation observed in the study’s field arm permitted forty-one seconds. Listeners, asked afterward to restate what the speaker believed, produced reconstructions matching the reference document at a median fidelity of 9 percent; in 71 percent of trials the position the listener came away with was one the speaker had explicitly rejected, in writing, during elicitation. The reconstructions correlated at r = 0.81 with the speaker’s first independent clause and at r = 0.04 with everything he said after it.

That last figure is the one the center’s investigators consider the finding. A political position, the paper argues, is held all at once — the report uses the term “the bolus” for the undifferentiated mass of belief awaiting transmission — while the mouth is a serial device that emits one word at a time, in an order, and the order is nowhere in the belief. The speaker must invent it at the threshold, under time pressure, in front of a listener who has already begun concluding. Ninety-one percent of subjects identified the selection of a first clause as the most aversive element of the entire protocol, ranking it above the disclosure of income. Seventy-nine percent, offered the opportunity to begin, declined.

“There is no first thing,” said Dr. Terrence Kolbeck, the center’s director and the study’s principal investigator, in an interview at his office in Ann Arbor. “The bolus has no front. Whatever you put in front becomes, to the listener, the thing you believe, and the remaining eighteen hours are heard as qualification of it, if they are heard at all. Our subjects understand this. They have understood it for years without a vocabulary for it. What the literature has been calling reticence, or polarization avoidance, or self-censorship, is on our data none of those things. It is a bandwidth judgment, and it is correct.” Asked whether the finding might be summarized for a general audience, Dr. Kolbeck said that it could not, and then attempted it four times.

Mr. Neuhoff’s own account is more specific and, in the center’s assessment, typical. He estimates commercial mechanical systems for a contractor in Van Buren Township, work that consists of converting a building — which exists all at once — into a sequential document that a stranger will read from the top. He is good at it. He said the October 2011 conversation, which took place in a driveway on this street and concerned a school millage, ran approximately ninety seconds and ended with a neighbor telling a second neighbor what Mr. Neuhoff thought. “He wasn’t lying,” Mr. Neuhoff said. “He was working from what I gave him, in the order I gave it. I gave him the wrong order.” He has since declined to discuss politics with anyone in Ypsilanti, a decision he characterized as logistical rather than fearful, and noted, without being asked, that the study of him had been conducted in Ann Arbor, where he said the rent is higher.

The finding has been received with some vindication at the Center for Nonparticipation Studies in Washington, whose Civic Silence Initiative has tracked the phenomenon since 2019 without a mechanism to explain it. Dr. Cassandra Murch, a senior fellow there, read Mr. Neuhoff’s 2,900-proposition reference position at this newspaper’s request and reduced it, in a single afternoon, to ninety-one words. Mr. Neuhoff has had the ninety-one words printed on a laminated card, which he carries in his wallet and has handed to eleven people since March. “The subject’s position was never unclear,” Dr. Murch said. “It was unspeakable, in the narrow sense — it could not be produced by a mouth in the time a mouth is given. I did not clarify anything. I put it in a medium that does not have a first clause, in the manner of a map, which you may enter anywhere. On the data currently available, that appears to be the whole of the intervention.”

In the Senate, the report has produced a legislative vehicle. Senator Marion Aldridge, Democrat of Oregon, introduced S. 3120, the Deliberative Throughput and Civic Record Act, on Aug. 26, which would establish within the Government Publishing Office a registry permitting any citizen to file a political position of unlimited length and to receive a filing number, and would provide that a citizen in conversation may state that number in lieu of an opinion. The bill creates no obligation on any listener to retrieve the filing. Ms. Aldridge, Democrat of Oregon, who in March described a previous measure as the product of years of private conversations among colleagues who lacked a procedural vehicle, said the principle was unchanged and the constituency merely larger. The bill was referred jointly to the Committee on Rules and Administration, which asserts jurisdiction on the ground that the registry is a record, and to the Committee on Commerce, Science and Transportation, which asserts jurisdiction on the ground that a conversation is a transmission. Senator Gerald Frisch, Republican of Ohio, called the measure “constitutionally unserious,” a phrase he has now applied to four bills.

Mr. Neuhoff has not filed, on the grounds that the registry does not yet exist. He produced the card in his driveway on Wednesday and, asked what it said, handed it over rather than answering. Read aloud back to him, in his own presence, at 138 words per minute, the ninety-one words took thirty-nine seconds. He listened to all of them. He was quiet for a moment afterward and said that was not quite it.

“It’s fine on the card,” he said.