Characteristics of excluded studies
Bloom (1992)
Reason for exclusion	Compares students from one district to students in another district (unit bias)
Dykstra (2013)
Reason for exclusion	This study investigates issues related to instructional group size, not class size.
Furno (1967)
Reason for exclusion	Class size is measured as: the pupil's median class size over a period of 4 years, or, in particular, the school years 1959–1960, 1960–1961, 1961–1962, and 1962–1963. Only outcomes averaged over the six school years:1959–1960, 1960–1961, 1961–1962, 1962–1963, 1963–1964, and 1964–1965 are analysed.
Hart (2011)
Reason for exclusion	Not about class size: A total of 33 children with ADHD were randomly assigned within days to either small‐group instruction, whole‐group instruction, or independent seatwork. The effects of instructional contexts on on‐task behaviour during instruction and on‐task behaviour and work productivity during testing were examined.
Keith (1993b)
Reason for exclusion	No numbers reported.
Patterson (2016)
Reason for exclusion	Compares self‐contained classrooms to inclusion and mainstream. Also has a class size component specifically by placement type, but there is no variation in self‐contained classroom sizes (only 1–10), see Table 6.
Snart (1985)
Reason for exclusion	Investigates student/teacher ratio, not class size. Also, the outcome in this study is not a validated measure of student classroom behaviour. Furthermore, authors state the following on p. 293: ‘Limited research access to the classrooms discussed within this study resulted in a confounding of condition with classroom, since we had agreed to spend only one full day per classroom’ (unit bias).
Steinbrenner (2015)
Reason for exclusion	The classrooms analysed served between six and ten students (i.e., a varying number), but the analysis is not about class size, but instructional group size.
	Large groups is the whole class: The classrooms all used some large group instruction (e.g., morning group, academic instruction).
	Small group: A few of the classrooms also had small group times, in which the classroom staff worked with dyads or triads on academic tasks such as worksheets or book reading. The observations were planned to be conducted during two one‐to‐one sessions, two small group sessions (i.e., 2–3 students) and two large group sessions (i.e., 4 or more students) when possible. However, many classrooms did not have regularly scheduled small group sessions; therefore, additional large group sessions were observed for students who did not participate in small group sessions
Thurlow (1988)
Reason for exclusion	Does not investigate class size, but how many teachers pr. student in instructional group, where the same students can be in more than one grouping.
Thurlow (1993)
Reason for exclusion	Does not investigate class size, but how many teachers pr. student in instructional group, where the same students can be in more than one grouping (see p. 310 and Table 2).