In one Pittsburgh elementary school classroom, students started the day in a circle, explaining how they were feeling as others listened intently. Some were happy, but others were sleepy or sad.

“Let’s remember those who said they’re tired or frustrated so we can help them out today,” the teacher said in closing.

A similar ritual for a group of sixth-graders in another class didn’t go as well. Asked to share their week’s high and low points, students talked over each other. When the teacher shared her own low point, a student yelled, “I thought your low point would be teaching us every day.”

Those experiences mirror the mixed results of two dozen Pittsburgh schools’ move to new discipline policies known as restorative justice, according to a comprehensive study released last week by the RAND Corporation. It appears to be the first randomized trial — the gold standard in social science research — of restorative justice in schools, a practice that has taken hold across the country.

The study finds that the move accomplished a key goal of school discipline reformers by reducing the number of days students were suspended, especially for black students.

Teachers also reported their school environments felt safer, rebutting critics who claim that reducing suspensions means chaos in the classroom. That was a key argument U.S. Secretary of Education Betsy DeVos used to back up her recent decision to revoke Obama-era rules aimed at curbing suspensions of students of color.

“I’m very much heartened by the finding that suggest that we could actually use this to address gaps, which has just been a challenge,” said Kristen Harper, an analyst with Child Trends and a former Obama education official.

But Pittsburgh’s changes came with notable downsides, raising questions about the tradeoffs that come with new ways of addressing misbehavior.

The policies appear especially unhelpful in middle school grades, where they didn’t reduce suspension rates but did hurt test scores. The shift did not boost student learning on the whole, and black students in particular actually saw significant reductions in test scores.

“I’m concerned, but not concerned enough to throw it away,” said Constance Lindsay, a researcher at the Urban Institute who has studied school discipline.

What happened: Ambitious changes across Pittsburgh

Pittsburgh Public Schools officials believed they had a problem.

Over one-third of students had said in a survey that they felt angry about how they were treated by adults in their school. Like many districts, Pittsburgh had high suspension rates meted out predominantly to black students. School climate needed to change, officials believed, and alternatives to suspensions should be a part of the fix.

Pittsburgh got a $3 million grant from the U.S. Department of Justice to try one approach. Twenty-two schools were randomly assigned to try restorative justice practices, and another 22 were not. The two groups of schools were compared over two school years, 2015-16 and 2016-17.

With two to four days of training, and ongoing help from coaches, the teachers were told to work with students rather than dole out punishments. Discussion circles were encouraged, both in response to misbehavior and on an ongoing basis. Disciplinary processes were supposed to become more transparent, and staff were told to express their personal feelings in response to positive or negative events.

None of that explicitly prohibited staff from suspending students, but teachers perceived the shifts — probably correctly — as discouraging them from doing so.

Nationally, this kind of move has been encouraged by civil rights groups and the Obama administration. They have pointed to evidence linking exclusionary discipline to worse academic outcomes, and argued that if schools resorted to suspension less often, students would be more engaged and avoid the academic issues that can result from missing school.

What did the study find?

As expected, suspensions dropped in schools that added restorative justice.

An interesting twist here is that, mirroring national trends, suspensions fell precipitously in both groups of schools — the ones with restorative justice and the ones without it. But suspensions dropped faster in restorative justice schools.

“It’s really encouraging and validating,” said Christine Cray, an official at Pittsburgh Public Schools who works on the restorative justice program.

In particular, 12.6 percent of students were suspended at least once in the restorative justice schools, compared to 14.6 percent in the other schools by the second year of the program.

Black students saw bigger declines in suspensions due to the program, though they were still much more likely to be suspended than white students. Restorative justice also all but eliminated the placement of students in “alternative” schools.

The academic results are somewhat grimmer. While restorative didn’t have a significant effect on reading scores, math scores scores for students in grades 3 through 8 did fall significantly.

The impact was not felt equally: It was black students, not white students, whose scores fell. A black student at the 50th percentile would have dropped to roughly the 44th percentile as a result of the initiative.

Lindsay of the Urban Institute said this could be because schools with more black students were more reliant suspensions, making the reforms more challenging to implement.

“It’s definitely a new way of doing business,” she said. “And in schools where there are more African-American students, it’s a dramatically different new way of doing business.” Indeed, the study found that black students in schools where most students were not black did not see their scores fall. White students in predominantly black schools did see declines.

In surveys, students in schools that implemented restorative justice were more likely to say their teachers struggled to manage their classes. But teachers themselves painted a brighter picture, reporting their schools had a better understanding of how to handle student behavior and a better culture for teaching and learning.

What does it all mean?

The results don’t offer clear answers to guide the polarized debate about school discipline.

One explanation for the uneven test results might simply be that teachers diverted time from academics, causing students to be less prepared for exams. That makes sense to Cray, the district official. The goal now, she said, is for “teachers to see how restorative practices can be used to complement academic instruction and not replace academic instruction.”

But it remains unclear why the shift hurt test scores in middle school grades, even though suspension rates didn’t change.

Meanwhile, the study does offer a roadmap to the challenges and benefits that can come with introducing restorative justice.

“I do feel like the kids are more willing and forthcoming with their problems and information to adults,” one teacher told researchers. “I feel like some of them do consider us to be more of an ally to them.”

The most basic challenge teachers said they faced was an implacable one: a lack of time, particularly for middle-school teachers with short class periods to spend with students.

“Teachers described the immense amount of curriculum they were obliged to cover and the assessments they had to prepare students for,” researchers wrote. “In light of those responsibilities, sparing 20-plus minutes for circles to build community or respond to conflict in the classroom seemed an insurmountable challenge to some.”

Not all students responded, either. “There are several students who have not benefited from the use of restorative practices at all,” one teacher said. “Rather, they disrespect it and scoff at it as a lenient form of discipline. Because of these students, it becomes harder to implement restorative practices on the whole.”

The researchers also suggest that the support offered to teachers in the program, while substantial, was not enough. Teachers described monthly team calls with a restorative justice coach as largely useless; more helpful were in-person visits, but they were infrequent.

Still, nearly two-thirds said the initiative had improved their relationships with students. Forty-five percent said it improved student behavior.

For now, the study is one part of a small body of research looking at whether these kinds of alternative discipline strategies work. But more can be expected soon — as part of the same federal grant program, a number of other studies have been funded elsewhere, Harper noted.

It’s also possible that results will change as teachers and schools adapt to the changes.

“Two years is a very short amount of time,” she said. “Much of the work involves a mind shift and a culture shift, not just training in different practices.”