What is formative assessment, and does it work?
Assessment is formative when the evidence is used to adapt teaching while learning is still happening. A quiz you grade and file is summative. The same quiz, used to decide tomorrow's lesson, is formative.
The research case was set out by Paul Black and Dylan Wiliam of King's College London. Their 1998 review of about 250 studies found that formative assessment experiments produced typical effect sizes of 0.4 to 0.7, "larger than most of those found for educational interventions," and that many studies showed it helps lower attainers more than the rest (Black and Wiliam, Phi Delta Kappan, 1998; ERIC record).
In England it is a professional requirement: Teachers' Standard 6 says teachers must "make use of formative and summative assessment to secure pupils' progress" (DfE Teachers' Standards). In the US, the five strategies below were set out for teachers by Leahy, Lyon, Thompson and Wiliam in ASCD's Educational Leadership magazine (Educational Leadership, 2005):
- Clarify and share learning intentions and criteria for success.
- Engineer effective classroom discussions, questions and tasks.
- Provide feedback that moves learners forward.
- Activate students as owners of their own learning.
- Activate students as instructional resources for one another.
How do you make success criteria clear?
1. Good and weak examples side by side. Hand out four or five anonymous samples of last year's work, at clearly different quality levels. Students rank them and say what makes the best one better. Leahy and colleagues describe teachers doing this with lab reports, then moving to samples of similar quality as students get sharper.
2. Rewrite the objective in student language. Leahy and colleagues note that posting the standard or objective at the start "is rarely successful because the standards are not written in student-friendly language." Write the one-line version a 13-year-old would say: "I can explain why the plant's mass goes up."
3. Build the rubric with the class. After idea 1, turn the students' reasons into three or four criteria. Post them where students can see them while they work.
Which questions give you the best evidence?
4. A range-finding question at the start. Ask one question that shows where the class already is, so you know where to begin. Leahy and colleagues give the example "How many fractions can you find between 1/6 and 1/7?"
5. A pre-question. Ask about the new topic before you teach it. The US Institute of Education Sciences recommends pre-questions to introduce a topic, though it rates the evidence for this use as low (IES practice guide, 2007). Keep it short and tell students wrong answers are expected.
6. A hinge question. One multiple-choice question at the point where the lesson can go two ways. England's teacher training framework names hinge questions as a way "to pinpoint knowledge gaps" and suggests building "common misconceptions within multiple-choice questions" (DfE ITTECF, 2024).
7. Everyone answers at once. ABCD cards, mini whiteboards or clickers, so you see 30 answers and not one. Leahy and colleagues point out that with hands up, "the teacher gets to hear only one student's thinking."
8. No hands up, except to ask a question. Pick names at random after think time. Every student has to prepare an answer.
9. A retrieval starter. Open with five questions: two from last lesson, two from last month, one from last term or semester. The IES guide rates spacing review over weeks to months as moderately supported. In a US sixth grade social studies study, students scored 94% on chapter exam items they had been quizzed on in class, against 81% on items that were not quizzed (Roediger et al., 2011).
10. An exit ticket. One question at the end, read before the next lesson. Sort answers into three piles: got it, nearly, not yet. See exit ticket ideas for formats.
How do you give feedback that moves learning forward?
11. Comments without a grade. Leahy and colleagues summarize research showing that "when students receive a grade and a comment, they ignore the comment." On drafts, give the comment alone, framed as the next step.
12. "Five of these are wrong." Instead of marking every error, write "Five of these 20 answers are incorrect. Find them and fix them!" (an example from Leahy and colleagues). The student does the thinking, and you mark less.
13. Whole-class feedback from the chart. After a class vote, show the spread, name the most popular wrong answer and explain why it is tempting. It takes about a minute.
How do students take ownership of their learning?
14. Traffic lights. Green, amber or red cards on the desk during practice. Go to the reds first. Leahy and colleagues use green and red cards; amber adds a "not sure" option.
15. Answer plus confidence. Ask a question, then "How sure are you? 1 to 3." A wrong answer with high confidence is a misconception to deal with now. A right answer with low confidence needs more practice, not reteaching.
16. Self-check before handing in. Students mark their own work against the class rubric (idea 3) and highlight the criterion they are least sure of. You read that line first.
How do students help each other learn?
17. Vote, discuss, revote. Everyone answers alone, discusses with a neighbor, then answers again. In a study of an introductory genetics course, scores on a new question about the same concept averaged 21 percentage points higher after discussion than on the first vote (Smith et al., Science, 2009).
18. Peer feedback on one criterion. Partners swap work and comment on one line of the rubric only. Leahy and colleagues add a caution: students "should not be giving another student a grade that will be reported."
How do you do formative assessment with large groups and adults?
Cards and whiteboards work in a high school or secondary school classroom. With 150 first-year students or a room of new hires, you need answers that are counted for you.
19. Split results by group. Start with a background question (course, department, years of experience) and compare how each group answered later questions. If one cohort keeps missing the same idea, look at what their prior course covered before you change today's session.
20. Track the same questions over time. Reuse three key questions at the start, middle and end of a course or training program. If the percentage correct does not move, the teaching on that point needs to change.
OMBEA Response does ideas 7, 13, 19 and 20 inside PowerPoint: ResponsePad clickers for everyone, a grouping view that splits any result by a background question, and question and participant reports with trends over sessions (see features). Trainers and course providers can see typical setups on the training providers page.
How do you act on what you find out?
Decide before you ask. A simple rule many teachers use:
- Most of the class correct: move on, and pick up the few who missed it during practice.
- A split: use vote, discuss, revote (idea 17).
- Most wrong: reteach with a different example, and check that the question itself was clear.
Our guide on how to check for understanding puts numbers on these cut-offs (about 80% to move on, 35% to 70% for discussion) and where they come from.
How do you keep formative assessment from adding to your workload?
England's teacher training framework warns that "assessment can become onerous and have a disproportionate impact on workload." Four ways to keep it small:
- Plan one hinge question per lesson, not a worksheet.
- Read, don't mark. Exit tickets go into three piles. No comments needed.
- Let the chart do the counting. A class vote takes 30 seconds to read. Thirty notebooks take an evening.
- Use peers. Leahy and colleagues found that once students did self and peer assessment, teachers could be "more selective" about which work they looked at.
Start with one idea
Pick idea 6, the hinge question. Write one for tomorrow's lesson with three wrong answers that match real misconceptions, and decide what you will do for each result. To have every answer counted and saved without phones, see the features or ask us for a quote.
