*This guest post comes from **Patrick Bijsmans** (Maastricht University) and **Arjan Schakel** (University of Bergen)*

In one of his recent contributions to this blog, Chad asks why students should attend class. In his experience

[C]lass attendance and academic performance are positively correlated for the undergraduate population that I teach. But I can’t say that the former causes the latter given all of the confounding variables.

The question whether attendance matters often pops up, reflected in blog posts, such as those by Chad and by Patrick’s colleague Merijn Chamon, and in recent research articles on the appropriateness of mandatory attendance and on student drop-out. In our own research we present strong evidence that attendance in a Problem-Based Learning (PBL) environment matters, also for the best students, and that attending or not attending class also has an influence on whether international classroom exchanges benefit student learning.

Last year we reported on an accidental experiment in one of Patrick’s courses that allowed us to compare the impact of attendance and the submissions of tasks in online and on-campus groups in Maastricht University’s Bachelor in European Studies. We observed that that attendance appeared to matter more for the on-campus students, whereas handing in tasks was important for the online students.

This year the same course was fully taught on-campus again, although students were allowed to join online when they displayed symptoms of or had tested positive for Covid-19 (this ad-hoc online participation was, unfortunately, not tracked). We did the same research again and there are some notable conclusions to be drawn.

In the first-year BA course that we looked at, students learn how to write a research proposal (see here). The course is set up as a PBL course, so it does not come as a big surprise that attendance once again significantly impacted students’ chances of passing the course.

Figure 1 displays the impact of the number of attended meetings on the probability that a student will pass for the course. Not surprisingly, the impact of attendance is large, a student who attends only one meeting is quite certain to *fail* (35% to pass) whereas a student who attends all meetings is quite certain to *pass* (70%).

*Notes: Shown are the predicted probabilities and their 95% confidence intervals. The results are based on a logit model whereby 175 students are clustered by 18 tutor groups and that includes the attended number of meetings and the number of tasks that were handed-in and their interaction. All the differences between the predicted probabilities are statistically significantly different from each (p < 0.01).*

Figure 2 displays the impact of the number of tasks that are handed-in on the probability to pass for the course. The impact of the number of handed-in tasks is also large, a student who hands in only one task is quite certain to *fail* (34% to pass) whereas a student who hands-in all tasks is quite certain to *pass* (76%).

Comparing the impacts of attendance and handing in assignments we observe that attendance matters as much as handing in assignments, but a significant interaction effect signals that both strengthen each other. In Table 1 we display the impact of attendance and handing-in tasks on the probability to pass for the course. Most students (112/175 = 64%) attended 4 to 6 meetings and handed-in 5 to 7 tasks. Hence, we zoom in on these students to disentangle the separate impact of attendance and tasks handed-in.

*Notes: Shown are the predicted probabilities and their 95% confidence intervals. The results are based on a logit model that includes an interaction effect between the attended number of meetings and the number of tasks that were handed-in and whereby 175 students are clustered by 18 tutor groups. All the differences between the predicted probabilities are statistically significantly different from each (p < 0.05; except for when the number of attended meetings is 4: p < 0.10).*

The differences between predicted probabilities for 5 and 7 handed-in tasks ranges between 8% when a student attended 4 meetings to 15% when a student attended 6 meetings. This impact is significant but also a bit smaller than the impact of attendance. The differences between predicted probabilities for 4 and 6 attended meetings ranges between 13% when a student handed-in 5 tasks to 20% when a student handed-in 7 tasks.

An important take-away message from Table 1 is that attendance and handing-in tasks reinforce each other. That is, the impact of attendance is larger when a student hands-in more tasks (i.e. from 8% to 15% is 7% increase), and the impact of handed-in tasks is larger for students who attend more meetings (i.e. from 13% to 20% is 7% increase).

*Notes: Shown are predicted probabilities and their 95% confidence intervals. The results are based on a logit model whereby 175 students are clustered by 18 tutor groups. The model includes the attended number of meetings (att) and the number of tasks type I and tasks type II and their interactions. All the differences between the predicted probabilities are statistically significantly different from each other for tasks type-I when a student attends 5 or 6 meetings (p < 0.01). None of the differences between the predicted probabilities are statistically significant for tasks type II.*

We further explore the impact of handing-in tasks by looking at the impact of the type of tasks (Figure 3). The first group concerns general writing tasks that were specifically discussed in class, but students didn’t receive written feedback from tutors (tasks type I). The second group concerns writing tasks that directly prepared for the final course research proposal. These tasks were not specifically discussed in class, but students receive extensive written feedback from tutors (tasks type II).

Whereas one may expect that tasks type II mattered most given that they prepare for the final exam, we actually find that their effect was negligible. At the same time, handing in task type I assignments – those discussed in class, without written feedback – did have a positive effect on chances of passing the course. We explain this striking result by one of the core elements of PBL, namely effective learning occurs through collaboration. While discussing a wide range of students’ assignments in class (tasks type I) students do not only learn and reflect on their own assignment but also from those of their fellow students. This increases their understanding of what is good academic writing and what is not.

These striking results also raise interesting questions regarding writing assignments, staff feedback and workload and how these issues should be dealt with in an active learning environment such as PBL. Perhaps writing assignments – in different forms – can be integrated more into class discussions, decreasing the workload that normally comes with giving feedback on individual writing assignments?