What would be useful to know about class size.

by Harry on May 16, 2019

This is as much a request for information as anything else, and I imagine that it is information that other academics want too!

Most of the courses my department teaches fall into one of four categories. There are large and small courses, and there are courses aimed at majors and courses not aimed at majors. Plenty of small courses are not aimed at majors; most large courses are not aimed at majors (the exception, sort of, is 101, which is aimed, partly at least, at attracting majors). In our context, large means 80-100 (with, occasionally, 160); small means 15-30.

I suspect, as do most of my colleagues, that more learning-per-student occurs in the small than in the large classes. This is far from certain, because we lack high quality measures of learning. We also assume that some teachers are better than others in large courses, and some are better than others in small courses, but we don’t know a great deal about who, because…. we lack high quality measures of learning.

We are not going to engage in a wholesale reform of our curriculum (that’s a prediction, not an insistence). Like most departments, though, we have some latitude in deciding how large our courses will be. The key choice that we can make is in how big to make the big courses, and how small to make the small courses. So, for example, we could keep our large courses as low as 80, and pay the cost in terms of allowing caps on our smaller courses to rise to (say) 30. [Of course we could decide to teach all, or nearly all, of our classes, as mid-sized, and if you could find me research that convinced me that was optimal I would share it with colleagues and try to persuade them, but I am putting that possibility aside to simplify things]. Because we lack high quality measures of learning, and because sample sizes are anyway small, we can’t make those decisions on the basis of what we know about our own situation, so it would be handy to have research that could tell us something useful.

But the research I have seen doesn’t tell me anything useful.

I’ve read a smallish number of studies and what they all suggest, unsurprisingly, is that smaller classes are, generally, better than larger classes (typically in these studies ‘small’ means somewhere around 15, and ‘large’ means something like 80 and more). I suppose that is helpful in one way. But what I want to know is this: does the gain from reducing class size from, say, 25 to 20, outweigh the loss from increasing class size from, say, 80 to 100?

My conjecture is that it does. Here’s how I experience it a teacher. In a class with 20 students I can know every student, make sure all of them are engaged in the class, get them all to know each other, and the costs of making them come to class are virtually zero because they nearly all come to nearly every session, and they all feel accountable for telling me beforehand if they are going to miss. Because nearly everyone will speak in nearly every class session they almost all arrive prepared to almost every class. In a class of more than 25 none of that is true, and a gap opens between the committed students (most of them, but fewer than 20 of them) and the others. There isn’t enough time in each class session for me to ensure that they are all accountable for being prepared and, at some level, they know that, and, again, fewer than 20 are fully prepared every time. I suspect that if I had better skills I could manage a class of 25, but for a class of 35 my skills would have to be much, much better than they are.

Whereas. When my large lectures have 100 students, they are pretty much indiscernible from when they have 80. I get to know more students, and more students talk in class, but a lot of work is done in discussion sections which are capped around 20 where, in each section, students get the same attention as they do in a class of 80. The largest lecture I’ve taught is 180, and that does seem substantially different from 100 or 80 — more absenteeism, and actually fewer students talk in lecture than when there are only 100, and hardly any come to my office hours (far fewer than when there are 100). This semester I taught 170, and all that was true, though the quality of the work they produced was very high: in particular the online discussions were much better than they have ever been in my lecture classes, though I suspect that is partly because I experimented with making them sit in their discussion sections during lectures, and changed the settings for online discussions so they were only reading and commenting on the posts of their fellow discussion section members.

So, my guess is that it is worth teaching significantly larger lectures for the sake of teaching more, and slightly smaller, small classes, at least for someone with my limited skills. Of course, my colleague have different arrays of skills than I do, and no doubt some are better with small classes, and some better with large lectures, than I. But… what I want is not evidence about whether more learning per student happens in small than in large classes, but whether there are thresholds for each kind of class that we should be attentive to in making decisions about class size.



Scott P. 05.16.19 at 2:47 pm

An important variable: it sounds like you have TAs/discussion sections for your larger classes. We do not, at least in my department.


Jen Morton 05.16.19 at 3:00 pm

Thanks for this Harry. I also want to know what the research shows about the different class sizes vis-a-vis different learning goals and student populations. I have had a similar experience to yours, although in my context I’ve taught 80 person classes without a TA, which leads to virtually no learning happening, in my opinion. I have wondered whether a 40-60 person class with only committed majors would actually work perfectly fine. It seems like it is the most vulnerable students who need the 20 person class. So perhaps the optimal sizes would be a 20 person first-year/intro class and a 40-60 person higher-level majors class. This is how some of the undergrad/grad classes at Stanford were and my sense is that they worked quite well. But I have no evidence for this.


Z 05.16.19 at 3:32 pm

For a couple of years, I and two colleagues taught 90 students divided in three 30 student groups. Then we decided one of us would teach the bulk of the course material to the 90 of them and another team of 4 would discuss exercises, difficulties and problems in smaller groups of 22/23 (the way our teaching duties were set up, the two solutions entailed an equal number of hours taught, so were financially equivalent). This was generally considered by all (teachers and students) to be an improvement. It’s not exactly the question you ask, but nevertheless, I think it gives some support to your conjecture that reducing the class size in the 20-30 range yields improvements.

Another element in support is that the high-school teachers I know report a huge difference between classes with fewer than 25 pupils and classes with more than that number, especially in terms of class management and in the possibility of having student speak (my wife teaches English – as a foreign language – and she once remarked that she felt that a class in a foreign language in which a student speaks less than two minutes has probably been a waste of time for that student, or at the very least has not been an opportunity to improve – she then on to point out that because we typically schedule one-hour lectures, this meant that the maximum number of students should probably never exceed 22/23).


Neville Morley 05.16.19 at 3:59 pm

Entirely agree at the lower end – I find that I have to change my approach if numbers get over 20, let alone 25. At the upper end, I haven’t found a substantial difference between lectures of 80 and of 150 – attendance isn’t great in either (and the university’s insistence on lecture recording, even though I don’t record anything involving discussion, counteracts the measures they put in place to try to encourage attendance), and not is participation, but I don’t have any sense of 80-100 being less bad. What I would add is that in groups of 20-40 I can, mostly using small group work, set up a more engaged and interactive session than is possible in larger classes, albeit never as good as the smaller seminars, which then has positive effects on attendance and engagement.


M Caswell 05.16.19 at 8:35 pm

Agreed there is an effectively qualitative threshold between 20 and 25. (And probably not one between 80 and 100.)

Re: “small classes”- There’s another threshold lower down, I think, somewhere in the mid-teens. Though one can feel the addition of any added member at these numbers, I find there’s a kind of learning available to a 14-member class that is almost impossible at, say, 18. Since these numbers all read as “small” to the market, one has to be careful entrusting class sizes to enrollment administrators.


Isaiah O'Rear 05.16.19 at 8:50 pm

As you say, we lack high-quality measures of learning in higher education. Most of the existing research on higher education class size effects uses the l0w-quality measure of student evaluations. That being said, there does seem to be evidence of a non-linear relationship:

“…there are large reductions in mean evaluation scores as class size rises from 1–19 to 20–39 and from 20–39 to 40–59. At this point, scores become quite flat until class size jumps over 150 students and then there is another 0.3 drop in the mean evaluation score.”

-Where class size really matters: Class size and student ratings of instructor effectiveness, Bedard and Kuhn, 2008.


LizardBreath 05.16.19 at 8:51 pm

From my limited, long-ago experience teaching secondary school, I’d agree with you that there’s a real break between a class of 20 and one of 25. I found that with 20 or fewer I was personally engaging with each student, and more than that I was kind of lecturing to a mass of faces.


Nicholas John Munn 05.16.19 at 9:01 pm

I found that my upper level course became far more difficult to manage when it went from 18 students (2017) to 30 (2019). 30 was too many to wrangle into staying on-topic and engaged throughout, at least in the manner I had designed the course to be taught.

But in lower level classes, where we get 150-220 students (and then 20-25 in tutorial groups), the numbers don’t matter much at all.

Incidentally, we have been having a lot of success replacing the *lectures* in large courses with online lessons, and using the *tutorials* as the only face-to-face contact the students get. The number of students completing lessons in a 200 person class sits above 90% all through the semester, whereas lecture attendance drops significantly below that (60-70% on bad days, ie, when it rains or every other class has assessments due). We have been collecting data on completion rates, student experience, and learning outcomes for lesson based courses, and so far it seems as though they just work better than lectures do for large courses.


Fake Dave 05.17.19 at 12:47 am

I definitely think it’s worth considering the social skills of the teacher in all of this. Some teachers can memorize 100+ names in a couple weeks while others will never learn your name even in smaller classes. I’ve been in plenty of classrooms with 30-50 people where it was still possible to have a rapport with the instructor and a sense of classroom community, but I’ve also seen it breakdown a lot. It’s much harder to feel anonymous when there are fewer people to hide behind in class. In a room with only ten or twelve students, even a fairly oblivious instructor is going to get to know them pretty well. I think it’s still possible to connect with a larger group, but an instructor has to have a fair bit of personal energy and perhaps even a touch of genuine charisma.

Instructor skill is going to be tricky to quantify or control for, but if you don’t nail it down somehow it could screw up the whole analysis. For instance, if we make the reasonable assumption that instructors with more experience or better training can handle larger class sizes, then the corollary is that the least skilled/qualified teachers are best left teaching the smallest classes. Most departments at most universities take the opposite approach and give the small upper-level courses to the most qualified teachers, so that has to be accounted for somehow.


Alan White 05.17.19 at 1:39 am

“But… what I want is not evidence about whether more learning per student happens in small than in large classes, but whether there are thresholds for each kind of class that we should be attentive to in making decisions about class size.”

Thank you for this post, and its focus. I have no knowledge of actual data to determine the kind of thresholds you’re searching for, but I think the following observations are germane, Harry.

First, at many institutions class enrollment maxima are determined administratively to make the most efficient bottom-line use of enrollment versus anticipated tuition factored against expenditure for faculty. This was true of my institution (UW Colleges) before its demise by absorption into the 4-year comprehensives; the present 4-year that now governs my campus has slashed lecturer salaries by about 20% while retaining the usual class maxima at 33. There seems to be little concern at all about learning efficiencies when administration makes these determinations without faculty input, which now is the norm at many institutions.

Second, clearly the data we do have both formally and anecdatally favors small classes. It seems that your engagement of students in discussion sections of large classes mimics that of small classes, set in the context of large lecture classes. While I taught for my career in an institution that allowed first maxima of 35, then 40, then 33 (except for certain science and English classes), I often taught sections of logic that enrolled less than 20 just due to lack of interest. But not surprisingly those were often my most successful classes in terms of holistic accomplishment and for the very reasons you give.

Third, huge lecture sections for introductory courses, even with TA discussion sections, need to recognize that they serve the vast majority of students who might never take another course in that particular area. I was reminded of this today while reviewing the comments of a philosophy professor at UW–Madison, Cornelius Golightly (a pioneering African-American professor–someone Harry and I know well from some collaborative work), who in 1953 (!) mused over this very point for his introductory course, concluding that perhaps classical literature was not the best option for exposing students to the problems of philosophy. So the goals of curriculum, I’d argue, can’t be separated from questions of faculty-student instructional efficiency.

As usual, thanks again for very penetrating inquiry about undergrad instruction.


navarro 05.17.19 at 2:35 am

for what it’s worth i’d like to approach this from two directions: my experiences as a student taking classes in college and my experiences as a teacher in public secondary schools.

for the most part, in undergraduate work, whether the class had 40 or 60 didn’t seem to make a lot of difference in how well other students and i performed in the class. the exceptional situations of either much larger classes or very small classes had opposite results. during one school year the lecture halls in the chemistry building were being renovated and so freshmen chemistry classes were given in one of the four science auditoriums, more frequently used for scholarly symposia or paid lectures from well known scientists. i was in a class of 280 freshman taking the first semester of chemistry. that was a nightmare. in addition to the mob of students the chemistry department had the professors take on the courses in teams of 4 so they could all get paid. this meant that roughly every 4 weeks a new teacher would take over where the previous one had left off. the exam grades were laughable, routinely a 55 would get an “a” while a 30 would get a “c” and at least half the group failed each exam. when i took freshman chemistry in a standard lecture group of 40 i made an “a” each semester. for that case for sure, that was too large of a group.

in some of my later coursework i took a graduate level class in old english literature for undergraduate credit. there were 20 of us in the class and that was almost too many to handle all of the presentations and discussions the professor gave us. it was an enjoyable class though. when i took the second half of advanced calculus my class had 8 people and we found that we could engage in actual dialogue with the professor and really get some depth in the subject.

as a teacher in sixth grade i’ve had classes that ranged in size from 14 to 26. i always found 26 a bit hard to manage simply because at 11 and 12 my students truly needed a larger portion of my attention than i was able to distribute with that number of students. 21-24 seemed to be my upper bound on that score and the extra one or two i would sometimes have reduced my ability to give attention disproportionately to the number added. as a high school teacher of juniors and seniors i have had classes of 2 to 22. with this age group i found that groups from 16-18 worked best for me but i would rather have 22 than just 2. with two students i found it almost impossible to make the kind of progress i could make in a larger class.


David Y. 05.17.19 at 2:39 am

I’ve taught classes as small as 5 and as large as 400. For 10+ years I’ve been at a SLAC with a cap of 25 on all classes and more typically 10-20 in my department. For a seminar style, discussion-oriented class I think 12-15 is the sweet spot. Any smaller and you run the risk of too high a proportion of students being naturally quiet or frequently unprepared. Any larger and you can’t involve everybody regularly enough. 8-11 or 16-20 might work for an upper division class with motivated students. 21+ is too much.


M Caswell 05.17.19 at 12:14 pm

Instructor facility is one factor; more important at the small end of the spectrum might be the capacity for the students to get to know each other as partners in discussion.


Marc 05.17.19 at 3:21 pm

I see a definite drop in engagement, attendance, and class performance for lecture classes larger than 100. I don’t see a lot of differences in the 70 – 100 range, although the quality of in-class discussions rises as the size drops.


Moz of Yarramulla 05.18.19 at 2:36 am

When I was studying engineering my year cohort was about 120, and we did the first two years largely in one group. I recall that some lecturers who we quite good later with smaller classes really struggled with 120, but for the most part the ones who sucked with 120 students also sucked with 10. Attendance was high because if you weren’t there to work hard at learning you probably didn’t make it through the “intermediate year” of general university courses to even start engineering (our required course load was one over the maximum allowed for science students, even though we were taking the same courses).

Partly it’s easier to deal with engineering courses because 80% of the assignments are correct, and there is a correct answer. Likewise, 80% of the students are grinding away busily to get the work done. The other 20% is where they put the serious teaching resources.

For my BA papers it was all over the place. But for the most part mass lectures were over 100 students, with weekly tutorials of less than 20, led by postgrads (and in some cases effectively co-led by students). I had a couple of papers at 3rd year that were less than 20 students and I have no idea why the lecturer even bothered turning up – they read their notes at us and left. Others… I was forcibly* enrolled in a Greg Newbold course for having to temerity to claim that not being so was sufficient excuse for not handing in the first essay. I did it, for the same reason I was there without being enrolled – the guy is a great teacher.

* he made it clear that if I wanted to attend I had to do the work, and that he would help persuade my other commitments to let me enrol (restrictions on postgrads)


Chris W (merian) 05.18.19 at 6:31 pm

I wish we had measuring instruments of student learning that narratively take the whole experience into account while at the same time having a quantitative component. I’m a little pessimistic (uncharacteristically) as I’ve seen some of the most preposterous “results” cited that appeared to show that students basically learn nothing in college, when it is clear that this is not the case. [*]

And I wish that we could have metrics whose usefulness isn’t right away undermined by being institutionalized and gamed.

I also wish that there was a markedly low-on-ideology school of pedagogy research.

[*] Some errors are obvious — there’s a difference between a student who gets 0 on a calculus test because s/he has never taken calculus and a working engineer whose career is based on once having mastered calculus but who, 10 years later, doesn’t remember how to solve a problem. But there are also more subtle learning scenarios.


Miranda 05.18.19 at 7:32 pm

The national literature on composition pedagogy suggests that writing classes be capped at 15 students, and this was born out by better passing rates in iur composing classes at my institution when we dropped our cap from 22 to 16 in comp classes. It depends on what you’re teaching sometimes


Moz of Yarramulla 05.19.19 at 6:55 am

I wish that we could have metrics whose usefulness isn’t right away undermined by being institutionalized and gamed.

Any metric that is useful will be gamed. In a way that is the point – you’re measuring something you care about so people try to improve their performance as measured by the metric.


Mr Punch 05.20.19 at 5:09 pm

The “Harkness table” used a some American prep schools for discussion-based classes holds 12 students plus a teacher. My (limited) experience suggests that participation degrades rapidly beyond the 12-15 range; I’d attribute this partly to group dynamics among the students, and partly to a greater challenge for the instructor.

Comments on this entry are closed.