Small group tutoring: a quiz that runs the session

Small group tutoring changes one thing about the job. With one student you hear every answer. With five you hear one at a time, and an hour planned for the middle of the group fits nobody in it.

A short quiz gives that hearing back. Every student answers the same six questions at the start, alone. The answers, read by topic, decide what the hour is spent on, and four new questions at the end show what moved.

EdProcto, which publishes this guide, runs both quizzes on the students' own devices and counts the answers by question, and on the Educator plan by topic: see its page for tutors.

Does small group tutoring work as well as one-to-one?

In most comparisons a small group did about as well as one-to-one, and a group of five as well as a group of two; in one randomised trial one-to-one came out ahead. We read every study in this guide as its abstract in the ERIC database, except Wiliam's article, which we read in full.

• Reading, pooled. Elbaum et al. (2000) pooled one-to-one reading programmes taught by adults to 1,539 primary pupils at risk of reading failure. The studies that set one-to-one against small groups showed no advantage for one-to-one.
• One, three or ten. Vaughn et al. (2003) gave struggling second-grade readers the same programme for the same number of sessions, alone, in threes or in tens. One-to-one and threes were both highly effective, and one-to-one was ahead on no measure.
• Two or five. Clarke et al. (2017) randomised kindergarten children to a 50-lesson mathematics programme in groups of two, in groups of five, or not at all. The differences between the two sizes were negligible and not significant.
• The dissent. Schwartz et al. (2012) had 85 Reading Recovery teachers each teach at-risk first-graders one-to-one and in a group of two, three or five: 170 children in all. One-to-one scored significantly higher on eight of nine measures. The group sizes did not differ significantly from one another, though results trended down as groups grew.

All four concern young children and reading or early mathematics; none is about a private tutor's group of teenagers. Our reading: group size matters less than what the tutor knows about each member, which is the first thing a group takes away.

Why question every student at once?

Because one student's answer tells you about one student, and the plan for the hour needs all of them.

Wiliam (2014) sets out the problem with a question put to the room. Hands up makes taking part voluntary. Picking a name at random is fairer, yet it still assesses one or two students, and the whole group's next step rests on them. His remedy is an all-student response: everyone answers the same question at the same moment, on a card, a small whiteboard or their fingers. He asks for one at least every 20 to 30 minutes of group teaching, and prefers multiple choice because the answers can be scanned.

Going round a table of five is no substitute, since the second student has heard the first: our point, and the reason the starter quiz is answered alone.

The measured evidence is on response cards. Randolph (2007) pooled 18 studies that set cards held up by everyone against hand raising. Test scores, quiz scores and participation rose by large, statistically significant margins, and disruptive behaviour fell. Marsh et al. (2023), adding 15 later studies to 14 of the originals, reported the same direction in special and general education.

The objection to a record of every answer

Wiliam prefers cards and boards to electronic voting, partly because a system that keeps every response by every student sits badly with a classroom meant to be safe for mistakes.

Our answer is in the set-up. Tell the group the starter counts for nothing, and make it true: in EdProcto, set "Show the score" to "Never". The record then serves your plan, and no student is handed a number.

Does acting on the answers improve learning?

Yes, though pooled estimates of how much run from 0.29 to 0.70, and the hard part is choosing the next step, not reading the answers.

Researchers call the practice formative assessment. Fuchs and Fuchs (1986), in a special education journal, pooled 21 controlled studies of systematic formative evaluation: the average weighted effect size was 0.70, and it varied with how the data were evaluated and displayed. Lee et al. (2020) reviewed 33 studies in American schools and found a smaller overall effect of 0.29. It was larger where the evidence was formal, such as written feedback on quizzes (0.40), and where the cycle ran within or between units of teaching (0.52).

Heritage et al. (2009) found the weak link. In their study of teachers' knowledge for teaching mathematics, teachers were better at inferring from assessment information what students understood than at deciding what to teach next.

Smith et al. (2011) tested one next step. In two genetics courses, students answered a question alone; then talked it over with peers, heard the instructor explain, or did both; then answered a similar question alone. Discussion followed by explanation raised scores substantially more than either alone, for weak, middling and strong students. The plan below borrows that order.

How do you plan a small group tutoring session?

Plan the hour as a starter quiz, three minutes of reading, two teaching blocks aimed at what was missed, and a closing check. The timings are ours.

This starter looks forward, at today's three topics. The tutor's weekly check looks back at what survived the week, and makes the case for tutoring itself.

Decide beforehand what each result will mean

Heritage's finding is the reason to write the rules down first. Ours:

• A topic at 85% or more: skip it, and ask one question on it next week.
• Most miss it, on the same wrong answer: a shared wrong idea. Teach it first, to everyone, starting from that answer, as in fixing misconceptions.
• Most miss it, wrong answers scattered: not learned yet. Teach it from the start.
• One or two miss it: their gap, not the group's. Split the group, as Wiliam suggests.
• Nearly everything wrong: something underneath is missing. Drop a level, as in the first-session diagnostic.

Why two questions a topic

One miss can be a slip and one hit luck; two that agree are a finding. With five students, each answer moves a topic's figure by ten points, so read 60% and 70% as one result. How many questions a quiz needs has the arithmetic.

What does the plan look like with five students?

It looks like one topic skipped, one taught to everyone and one taught to three. The group below is an illustration, not a real one; its marks and the quoted lines came from running these answers through EdProcto's own marking and analysis code.

Five students revising algebra sat six multiple-choice questions, every wrong option a known error (how to write distractors).

These are the lines of the full analysis, which comes with the Educator plan. By topic, the results page gave 20% for two brackets, 60% for one bracket and 90% for like terms. Under "What it says about the teaching" it printed three lines:

• "Reteach before moving on: Expanding two brackets (20%)."
• "Secure, and safe to build on: Collecting like terms."
• "A shared wrong answer on Q5, Q6: the class was taught something, and it was not this. Worth a minute in the next lesson."

Beside Q5 it added "Most of the class missed Q5. Check the key first, then whether this was taught." The key was right: (x + 3)² is x² + 6x + 9. Then the hour:

1. Like terms: skipped.
2. Two brackets: everyone, first. Q5 went on the table with two answers, x² + 9 and x² + 6x + 9, and no names. Pairs tried x = 1 in each: (1 + 3)² is 16, the first answer gives 10 and the second 16. The tutor drew the square as four parts and set (x + 5)² on paper.
3. One bracket: a split. The three students who had missed Q3 or Q4 sat with the tutor on the sign in −2(x − 5). The other two expanded (2x + 3)(x − 4).
4. Closing check. Four new questions, hosted live in EdProcto: (x + 4)², (x + 1)(x + 6), 4(3x − 2) and −3(x − 4).

The closing results read 80% for each topic. One student still chose x² + 16 and x² + 6: the old rule, twice. That student went home with a study quiz on two brackets (homework quizzes that mark themselves), and next week's starter opens with one.

Read the 80% modestly. It is what the group could do with the explanation ten minutes old; next week's first question says what lasted. Neither quiz is a grade.

How does EdProcto run a group tutoring session?

EdProcto runs it as two quizzes: the starter on each student's own device with the answers held back, and the closing check hosted live with each answer checked at once. It is the best tool for a tutor's group because it helps with the step Heritage found hardest while the group is still at the table: the answers arrive sorted by your own topics, with a line on what to reteach.

"Exam" only means handed in once and marked at the end. Set up both like this:

1. Draft and label. Paste the worksheet into the quiz maker on the home page; Procto AI drafts, and you check each question against it. Type each question's topic in "Topic (for the results by topic)", worded identically in both quizzes.
2. Names only. Switch off "An account is needed to take it" and ask for the same first name every week: the "Students" page then keeps one row for each student across quizzes.
3. The starter. Switch off "Shuffle the questions for each student", so Q5 is the same question for everyone. Set "Show the correct answer", "Show the explanation" and "Show the score" to "Never".
4. The closing check. Write an explanation for every question. The scoreboard ranks, so keep the host's screen on your own laptop and say so first: each student's last page reads "You are in. Watch the host’s screen for the scoreboard." (Hosting a quiz night has the research.) Press "End the quiz night" once "Finished" equals "Players".

On every plan, with three or four students the results open with "Fewer than five submissions: the numbers below describe these students, not the class or the teaching yet." The rest is the Educator plan's full analysis. The "Secure" line is printed from five, and the shared-wrong-answer line needs three students on one option. Click a question for "Chose it" beside each option. "Separates" fills in from six submissions, and "Reliability" earns its own sentence from eight.

A group of eight fits every plan: Free holds thirty students a quiz, Tutor 100 and Educator 500. Free carries ten quizzes a month, enough for one weekly group at two a session; Tutor removes that limit and opens all 111 templates. Every plan gives the results question by question, where "Got it" shows Q5 at 20%; the full analysis, with the score by topic, the option counts and what to reteach, comes with Educator. See pricing.

Quick questions

How many students should a tutoring group have?

The trials that compared sizes stopped at five, apart from one arm of ten. Our working range is three to eight, with more structure from six up.

Should the starter quiz be scored?

Mark it, but show no score and count it for nothing. Its job is to plan the hour.

What if one student is far ahead of the group?

Bring harder questions on every topic, and keep that student in the paired discussion. In Smith et al. (2011), explanation alone did not help the strongest students.

Can students join a group quiz without accounts?

Yes. In EdProcto, with "An account is needed to take it" switched off, a student opens the link and gives a name. See the page for tutors.

References

• Clarke, B., Doabler, C. T., Kosty, D., Kurtz Nelson, E., Smolkowski, K., Fien, H., & Turtura, J. (2017). Testing the efficacy of a kindergarten mathematics intervention by small group size. AERA Open, 3(2). https://doi.org/10.1177/2332858417706899
• Elbaum, B., Vaughn, S., Hughes, M. T., & Moody, S. W. (2000). How effective are one-to-one tutoring programs in reading for elementary students at risk for reading failure? A meta-analysis of the intervention research. Journal of Educational Psychology, 92(4), 605–619. https://doi.org/10.1037/0022-0663.92.4.605
• Fuchs, L. S., & Fuchs, D. (1986). Effects of systematic formative evaluation: A meta-analysis. Exceptional Children, 53(3), 199–208. https://doi.org/10.1177/001440298605300301
• Heritage, M., Kim, J., Vendlinski, T., & Herman, J. (2009). From evidence to action: A seamless process in formative assessment? Educational Measurement: Issues and Practice, 28(3), 24–31. https://doi.org/10.1111/j.1745-3992.2009.00151.x
• Lee, H., Chung, H. Q., Zhang, Y., Abedi, J., & Warschauer, M. (2020). The effectiveness and features of formative assessment in US K-12 education: A systematic review. Applied Measurement in Education, 33(2), 124–140. https://doi.org/10.1080/08957347.2020.1732383
• Marsh, R. J., Cumming, T. M., Randolph, J. J., & Michaels, S. (2023). Updated meta-analysis of the research on response cards. Journal of Behavioral Education, 32(3), 450–473. https://doi.org/10.1007/s10864-021-09463-0
• Randolph, J. J. (2007). Meta-analysis of the research on response cards: Effects on test achievement, quiz achievement, participation, and off-task behavior. Journal of Positive Behavior Interventions, 9(2), 113–128. https://doi.org/10.1177/10983007070090020201
• Schwartz, R. M., Schmitt, M. C., & Lose, M. K. (2012). Effects of teacher-student ratio in response to intervention approaches. The Elementary School Journal, 112(4), 547–567. https://doi.org/10.1086/664490
• Smith, M. K., Wood, W. B., Krauter, K., & Knight, J. K. (2011). Combining peer discussion with instructor explanation increases student learning from in-class concept questions. CBE Life Sciences Education, 10(1), 55–63. https://doi.org/10.1187/cbe.10-08-0101
• Vaughn, S., Linan-Thompson, S., Kouzekanani, K., Bryant, D. P., Dickson, S., & Blozis, S. A. (2003). Reading instruction grouping for students with reading difficulties. Remedial and Special Education, 24(5), 301–315. https://doi.org/10.1177/07419325030240050501
• Wiliam, D. (2014, March 1). The right questions, the right way. Educational Leadership, 71(6). https://www.ascd.org/el/articles/the-right-questions-the-right-way