This article first ran in The Chronicle of Higher Education and appears here under an agreement with The Chronicle and with the author’s approval.
Earlier this year, The Chronicle’s Teaching newsletter described distracted students sitting in the back of a 600-person lecture class. What strategies, we asked readers, could professors use to hold students’ attention in such a setup? While some readers offered suggestions, many responses were similar to one from Autar Kaw, a professor of mechanical engineering at the University of South Florida, who suggested that students in such a large, in-person section aren’t likely to learn much. “Why,” Kaw asked, “do we have a class of 600 in the first place?”
The idea that small classes are best is intuitive. With fewer students in the room, a professor should be able to devote more time to each one, giving, for instance, more in-depth feedback on their writing. But comments like Kaw’s piqued my curiosity. What is the connection between class size and quality, anyhow? And if small classes really are better, how do you define what counts as small?
Studies of the relationship between class size and teaching effectiveness “have not reached consensus,” wrote the authors of a 2013 IDEA Center study in their summary of previous work. Still, they add, it’s reasonable to conclude that teaching effectiveness doesn’t improve, at least, as classes get larger. Other research has considered the connection between class size and a related but distinct variable, student achievement. It too, the authors write, has been “equivocal.”
For its study, the IDEA Center, a nonprofit group that researches and develops assessments, examined class size and students’ self-reported learning outcomes from the more than 400 colleges that use its evaluations. It found that students in small classes—which it defines as those with 10 to 14 students—said they’d made the most progress and put in the most effort.
“There is an inverse relationship between class size and student self-reported learning,” said Steve Benton, senior research officer at IDEA and one of the study’s authors. But that relationship, he added, is neither linear nor very strong.
And while IDEA found a relationship between small classes and learning, the study could not determine causality. Neither students nor professors are randomly assigned to courses. It’s possible that professors with a certain teaching style choose to teach small classes, or that more motivated students decide to take them.
IDEA can pinpoint other variables that have stronger links to learning, Benton said: students’ background preparation, their work habits, and their desire to take the course.
An Arbitrary Number?
The idea that small classes are preferable has been enshrined in the influential U.S. News & World Report “Best Colleges” rankings. The measure was added in the 1990s because of research linking small classes to higher student engagement, said Bob Morse, the company’s chief data strategist. It rewards colleges if a lot of their classes have fewer than 20 students.
Until relatively recently, U.S. News essentially credited colleges for having small classes and penalized them for having large ones. A college where a large share of classes had under 20 students, and a small fraction had 50 or more, would score better than a college that had fewer small classes and more large ones.
A couple of years ago, U.S. News revised the methodology on this measure, Morse said, in part to make it more difficult for colleges to game. Colleges still get the highest marks for the share of classes with fewer than 20 students, but they now get partial credit for medium-size classes, and no credit for those with 50 students or more.
Under either system, class size accounts for 8 percent of a college’s ranking, Morse said.
As Morse’s mention of gaming suggests, colleges have been known to make decisions with an eye to how they might affect their position in the rankings. “Over time, these formulations have produced a powerful mythology across many postsecondary institutions,” wrote Mark Salisbury, assistant dean and director of institutional research and assessment at Augustana College, on his blog, Delicious Ambiguity, late last year. “Classes with 19 or fewer students are better than classes with 20 or more.” Are they?
To investigate, Salisbury, a longtime rankings critic, analyzed Augustana’s IDEA data. What he found: For his college, at least, there’s no big difference in outcomes until classes cross the 30-student threshold. At a big research university, Salisbury suspects, that might not happen until classes become even larger. The U.S. News cut points, he thinks, are arbitrary.
Even Morse acknowledges that the cutoffs aren’t research-based. “It’s true there wasn’t social-science research that went into measuring the size of these groupings,” he said. The class-size groupings are pulled from the Common Data Set, a standardized set of information developed by colleges and publishers, but U.S. News independently determines how to use them, Morse said.
But most analyses of class size come at the question all wrong, says Dan Chambliss, a professor of sociology at Hamilton College and a co-author of How College Works. That’s because they reflect what’s happening at an institution, not what’s happening to individual students.
Here’s what he means. The U.S. News data show the proportion of a college’s classes that are small. They don’t tell you the proportion of students enrolled in a small class.
An extreme example can help illustrate this difference, Chambliss said. Imagine a college with 1,000 students, each taking one class. Say the college offers 99 single-student tutorial courses and sends the other 901 students to the football stadium to watch a movie. That college could claim that 99 percent of its classes are small. But just 9.9 percent of its students are enrolled in a small class. If you’re an administrator, you might highlight that first statistic on your website. But if you’re a student, it’s the second one that matters.
The IDEA data, too, is limited, Chambliss said. Students can only assess what they learned in the classes they got to take. But small classes, definitionally, are ones that only so many students are in. Students can’t report on the great learning experiences they missed out on by being boxed out of a good, small class that they were dying to take. If a college wants to give all students the best experience possible, Chambliss argues, the trade-off between offering students small classes and giving them access to the courses they want is important.
Hamilton, where Chambliss works, is the kind of college where students probably expect to reap the benefits of taking small classes. Currently, 75 percent of the college’s classes have fewer than 20 students, according to U.S. News. It boasts one professor for every nine students.
But in the course of researching his 2014 book, Chambliss consulted students’ transcripts and was surprised to learn that the typical class Hamilton students were taking had 26 students. How does that happen? It’s a less-dramatic version of the tutorials and the football stadium. If resources are fixed, that means that “every time my class gets smaller, someone else’s gets bigger,” he says, “and we don’t see that.” And of course, more students are taking that bigger class than the smaller one it is making room for.
Even when students manage to enroll in one, Chambliss isn’t sure small classes are all they’re cracked up to be. A student in a bad, large class can zone out or multitask. A student in a bad, small class is simply miserable. And such classes do exist. “Some classes are small,” Chambliss said, “because no one wants to take them.”
Besides, Chambliss thinks, large classes can offer something of value. On some campuses, so many students take and talk about a particular big class that it becomes a shared experience, uniting students in the way that sporting events and parties—but less frequently academic offerings—can do.
In the end, class size is meant to signify something important, something connected to learning and the student experience. At best, it’s a proxy measure for something else. Perhaps the real question is what that underlying quality is. Perhaps one day there will even be a more direct way to measure it.