The research comes from a closely watched group led by Carl Wieman, a Nobel laureate in physics at the University of British Columbia who leads a $12 million initiative to improve science instruction using research-backed methods for both testing students’ understanding and improving how science is taught.

In one of the initiative’s most visible studies, Dr. Wieman’s team reports that students in an introductory college physics course did especially well on an exam after attending experimental, collaborative classes during the 12th week of the course. By contrast, students taking the same course from another instructor — who did not use the experimental approach and continued with lectures as usual — scored much lower on the same exam.

In teleconference last week, Dr. Wieman and his co-authors said that some instructors at the university were already eager to adopt the new approach and that it should improve classroom learning broadly, in other sciences and at many levels.

Yet experts who reviewed the new report cautioned that it was not convincing enough to change teaching. The study has a variety of limitations, they said, some because of the difficulty of doing research in the dude-I-slept-through-class world of the freshman year of college, and others because of the study’s design. “The whole issue of how to draw on basic science and apply it in classrooms is a whole lot more complicated than they’re letting on,” said Daniel Willingham, a psychology professor at the University of Virginia.

Dr. Willingham said that, among other concerns, the study was not controlled enough to tell which of the changes in teaching might have accounted for the difference in students’ scores.

In the study, Dr. Wieman had two teaching assistants take over one of the two introductory physics classes during the 12th week of the term, teaching the material in a radically different way from the usual lectures. Both this class and the comparison one were large lecture-hall courses, each with more than 260 students enrolled. Instead of delivering lectures, the new instructors conducted collaborative classes, in which students worked in teams to answer questions about electromagnetic waves. The new teachers circulated among the students, picking up on common questions and points of confusion, and gave immediate feedback on study teams’ answers.

The techniques are rooted in an approach to learning known as deliberate practice, which previous research suggests is what leads to the acquisition of real expertise.

“As opposed to the traditional lecture, in which students are passive, this class actively engages students and allows them time to synthesize new information and incorporate it into a mental model,” said Louis Deslauriers, a postdoctoral researcher who, with Ellen Schelew, a graduate student, taught the experimental classes. “When they can incorporate things into a mental model, we find much better retention.”

At the end of the study, students in the experimental class who took a test on the material scored 74 percent, on average, more than twice the average of students in the comparison course who took the test. On midterm exams the two classes had scored almost exactly the same.

Yet this being college — and the end of the term, at that — not everyone showed up with their calculators. More than 150 of the students were absent from the test, most of them from the comparison class. The researchers had no way to know how those students, if they’d come, would have changed the overall findings.

Experts said, too, that it was problematic for authors of a study to also be delivering the intervention — in this case, as enthusiastic teachers. “This is not a good idea, since they know exactly what the hypotheses are that guide the study, and, more importantly, exactly what the measures are that will be used to evaluate the effects,” James W. Stigler, a professor of psychology at the University of California, Los Angeles, said in an e-mail. “They might, therefore, be tailoring their instruction to the assessment — i.e., teaching to the test.”

Dr. Wieman said he strongly doubted that the new instructors had this kind of effect on the students. As a rule, he said in an e-mail, students in such large classes “are remarkably removed from any sense of personal connection with the instructor.  That does change with a more interactive class, but not enough and not fast enough to have any significant impact on learning in a week.”

Either way, Dr. Stigler said, the study is an important step in a journey that is long overdue, given the vast shortcomings of education as usual. “I think that the authors are pioneers in exploring and testing ways we can improve undergraduate teaching and learning,” he said. “As a psychologist, I’m ashamed that it is physicists who are leading this effort, and not learning scientists.”