The annual ritual of state testing in elementary and middle schools often comes within an unwelcome side effect: jittery, stressed-out kids.

Now, a first-of-its-kind study documents some of what’s actually happening to students.

It found that students in one New Orleans charter network saw modest spikes in cortisol, a hormone caused by stress, leading up to state exams. And the students whose cortisol spiked most or crashed furthest did worse than predicted — suggesting that the test scores reflect not just what students know, but how they perform under pressure.

The five researchers behind the study call that a “stress bias.” The paper finds some evidence that students living in higher-crime, higher-poverty neighborhoods are most affected.

That’s not surprising from a biological perspective, said Pamela Cantor, a psychiatrist and the founder of Turnaround for Children, a group that works to address the effects of trauma on children in schools.

“What we’re in effect doing to kids who are exposed to adversity on a chronic basis is actually putting them in a highly unfair situation, where their biology may overreact to the stress and not give them a very good opportunity to reveal the things they likely know,” she said.

The research looks at fewer than 100 students, and some of the findings are ambiguous. “I don’t want to make broad policy claims based on this one paper with a relatively small sample size in one setting,” said lead author Jennifer Heissel, a professor at the Naval Postgraduate School. But, she said, “if this is replicated in other settings for other students, we need to reconsider perhaps what we are using high-stakes tests for.”

The study, released earlier this month through the National Bureau of Economic Research, focuses on a network of three charter schools in New Orleans. The researchers analyzed an unusual data source: saliva samples of 93 elementary and middle students, obtained with parents’ consent, during the 2015-16 school year.

The research team compared cortisol levels of students at three points: during a regular week, a week when students took a low-stakes practice test, and the week students took the state test.

Cortisol is a hormone that generally increases after someone wakes up and declines from there. It jumps in response to stress or challenges and decreases due to boredom or disengagement.

The researchers focus on cortisol levels in the period right before the exam, when students were likely to be most be stressed about testing. Indeed, cortisol levels were about 15 percent higher at that time during testing week than they were during a regular week.

“That is in line with other stressors you might encounter through your day,” Heissel explained.

But some students responded more dramatically. “On average, there’s this increase, but individual kids are going in different directions,” she said.

Students whose cortisol noticeably spiked or dipped tended to perform worse than expected on the state test, controlling for past grades and test scores. Boys saw bigger changes than girls. and so did students from higher-poverty neighborhoods, though this difference was not statistically significant. (Keep in mind that the students in the study were almost all low-income, so there was limited room for comparison.)

Source: “Testing, Stress, and Performance: How Students Respond Physiologically to High-Stakes Testing”

It’s possible that some students’ jumps or dips in cortisol during testing week were due to other factors in their lives. But Cantor of Turnaround for Children said it’s not surprising that tests would induce stress, or that students more likely to have experienced trauma would respond differently.

“For children who face adversity in a chronic way, that system is pumped and primed much more so than other kids,” Cantor said. “If a child does overreact to a trigger, one of the ways that manifests itself is that they shut down, they freeze.”

Is that a “stress bias”? If tests unfairly penalize students who respond poorly to stress, that might suggest that “tests aren’t fully capturing what we want and perhaps what we are thinking that they capture,” said Heissel.

Another interpretation, she noted, is that the ability to perform well under pressure is part of what exams measure, and that it’s a skill “to be able to wrangle your stress response.”

The results raise a number of unanswered questions.

One is whether the results would hold for in-class exams administered by teachers. In many cases, those tests have higher stakes for students than do state exams, which in New Orleans have been used to grade and in some cases close schools.

Another is whether the way students are prepared for state exams might affect their stress. The paper offers limited information on the charter network being studied, which is anonymous. But a number of high-profile charter schools place substantial emphasis on preparation for state tests. It’s unclear whether this approach increases or reduces test-related stress.

Either way, the latest paper suggests one way to produce better scores is to ensure students stay calm during testing.

And a final question is, how else might the stress of testing affect students? Past national research offers mixed evidence on whether No Child Left Behind, the federal education law that led to the current state testing regimen, led to general increases in student anxiety.

Cantor said that the key to buffering against adversity is having warm, positive relationships, which can prompt the release of anti-stress hormones.

“A teacher who communicates belief and confidence and inspires trust in kids — that teacher is activating a hormonal system that opposes the effects of cortisol,” she said.