Checks for understanding
Knowing it is working

Checks for understanding: finding out whether it landed

A check is a question whose answers tell you what to do next. This chapter writes one against each objective, and decides in advance what you do at low, middle and high agreement.

Start here Essentials

A check is a question every person answers at once, whose answer changes what you do next. Write one, decide in advance what each result means, and put it on the run sheet with its minutes.

  1. Pick one objective and write a question whose wrong answers are the mistakes you actually see. Word for word (section 2).
  2. Decide how people answer at once. Hands, lettered cards, or a show-me (section 4).
  3. Write what you will do at low, middle and high agreement. Three sentences, before the day (section 3).
  4. Put it on the run sheet with its minutes. A check with no minutes is a check that gets skipped (section 3).

Minimum time: about 30 minutes for the first one. Done when: one row in the checks sheet of your session plan holds the question, the right answer, what each wrong answer tells you, and the three things you will do depending on how the room answers.

1. A check is a question that tells you what to do next Understand

"Any questions?" is not a check. It samples the people who are confident enough to speak, it produces silence most of the time, and the silence means nothing: it is equally consistent with everybody having understood and with nobody knowing what to ask.

A check has four properties. Everybody answers, at the same time. The answer is visible to you. You know in advance what the possible answers mean. And you have decided what you will do about each one.

The last property is the one that distinguishes a check from a quiz. A quiz measures. A check changes the next ten minutes. If a question's answer would not change anything you do, it is decoration, and it is costing you minutes you could have spent on practice.

What people usually doWho it samplesWhat you learn
"Any questions?"The confidentNothing from silence
"Does that make sense?"Everybody nodsNothing. Nodding is a politeness
Asking one personOne personAbout that person
A question answered by all hands at onceThe roomA count, and what to do next
A show-me: everybody holds up their workThe roomThe performance itself, not a sentence about it

Checks also do something to the room, which is worth knowing. A room that has answered once answers more easily the second time, and a room that has been asked for a count knows that being wrong is part of the arrangement. That is worth more in the first fifteen minutes than later, which is why Structuring a session: the run sheet asks for a check in the first quarter.

2. Writing a check whose wrong answers mean something

The craft in a check is in the wrong answers. A multiple-choice question with three obviously silly options measures nothing, because people pick the right answer by elimination. A question whose wrong answers are the mistakes you actually see tells you which mistake is in the room.

So build it from your error list. The errors sheet in your demonstration and practice sheets already holds the mistakes you expect; each one becomes an option, and each option has a meaning written against it.

An example, for a session on shaping dough:

OptionWhy somebody picks itWhat you do if many do
A: more flour on the benchThey think sticking is the problem to solveReteach the tension step: they are managing stickiness rather than building a skin
B: drag it toward you until the top feels tightCorrectMove on
C: press down from above to seal itThey saw the hand on top and copied the position, not the movementRepeat the demonstration from the side; this is a sightline failure, not a comprehension one
D: let it rest longer before shapingConfusing shaping with proofingOne sentence to separate the two; this is a vocabulary problem

Convention Notice that option C tells you about your demonstration rather than about the room, which is the most useful thing a check does for a new teacher. The mistakes people make are mostly manufactured by the teaching, and the check is where you find out which part manufactured them.

Three rules for writing one:

  • Write it word for word. A question improvised in the room is a question whose options you have not thought about, and it usually collapses into "does everybody follow?"
  • Aim at the decision, not the fact. "Which of these would you do next?" beats "what is this part called?"
  • Make every wrong answer attractive to somebody. If an option is there to be dismissed, cut it; three real options beat five with two jokes.

3. Reading the answer: the published band

Here is the one published number in this subject that most people teaching a workshop have not met, and it is useful.

The Science Education Resource Center at Carleton College publishes a protocol for ConcepTests, developed by David McConnell of North Carolina State University and citing Mazur's work. Its page on how to use them states that the optimal range of correct first responses is 35 to 70 percent. Within that band, students discuss their reasoning with neighbors in pairs or small groups for one to two minutes, and the class is polled again.

What the page says about the two ends: below 35 percent correct, students do not understand the topic well enough to discuss it, or the question is unclear or too hard. Above 70 percent, the question may have been too easy, and questions that consistently draw 90 percent or higher are described as a waste of class time.

So the band gives you three responses, which are the three columns in the checks sheet:

ResultWhat it suggestsWhat to do
Below 35 percent rightThey are not ready to discuss it, or the question is unclearReteach the step, differently rather than louder. Then ask again
35 to 70 percentThe room is genuinely divided and the disagreement is productivePairs discuss for one to two minutes, then poll again
Above 70 percentMost have it; the question may be easy for this groupName the right answer, say one sentence about the common wrong one, move on
Limits

The same page says the percentages are guidelines rather than requirements, and that they shift with the group. Treat the band as the best published starting point and watch what your own room does: if your checks keep coming back at 80 percent, your questions may be easy for this group rather than your teaching being excellent. The protocol comes from undergraduate science teaching with classes of a certain size, and a room of eight is not a lecture hall. The arithmetic still works; the confidence you can place in a count of eight is lower than in a count of eighty.

Put the minutes on the run sheet. The check cycle calculator prices one: a pause, thinking alone, pair discussion and a re-poll. Its defaults for thinking and discussion come from the protocol above; the re-poll time has no published figure and starts at zero, which means you supply it.

4. The show-me check, for a physical skill

If your objective is a performance, the strongest check is the performance itself. Everybody holds up what they made, or does the movement at the same time, and you look.

It has advantages a question does not. It checks the thing rather than a sentence about the thing. It is fast: a room of ten holds up their work in five seconds. And it is hard to fake, which matters because a person can select the right option in a quiz and still not be able to do it.

Three forms:

  • Hold it up. Everybody raises their work at a count of three. You scan and count how many have the feature you are looking for. Say what you are looking for first, which makes the count legible to them too.
  • Do it on three. Everybody performs the movement at once while you watch. Works for a short movement, and it lets you see the hands rather than the result.
  • Point at it. For something on a page or a screen: everybody points at the line, and you walk the room in fifteen seconds.

The same three-band reading applies: count how many have it, and respond at low, middle and high. Write the feature you are counting into the checks sheet, because "I will see if they look right" is not something you can count against.

5. The full ConcepTest protocol, with its timings Go deeper

The SERC pages set out the whole cycle, and the timings are worth having because they let you price a check honestly.

  1. A short presentation. Typically five to fifteen minutes.
  2. The question is posted. Visible, so people can reread it.
  3. Individuals consider it alone. Thirty seconds to one minute, choosing an answer.
  4. Answers are shown. By hands, cards, answer sheets or clickers.
  5. If the result is in the band, pairs discuss. One to two minutes with neighbors, on their reasons.
  6. Poll again. And then respond to the second count.

The page also recommends against grading the questions, which translates straightforwardly to a workshop: the moment a check feels like a test with consequences, people answer what they think you want and the count stops being information.

Two adaptations for a short practical session. The individual thinking time can be shorter for a question about something people just did with their hands, because they are retrieving an experience rather than reasoning from a concept. And with a group under about eight, the pair discussion can be a whole-room discussion, which costs the same minutes and loses less.

The three-second minimum pause after asking belongs here too, and it comes from a different source: the wait-time guidance in Asking, waiting and handling what comes back. It is the floor, not the thinking time.

6. The Minute Paper, the Muddiest Point and the rest Go deeper

The University of Connecticut's page listing 50 classroom assessment techniques, attributed on the page to Angelo and Cross, is the standard catalog. Two of them earn a place in a short practical session.

The Minute Paper, described on that page as perhaps the most frequently used technique, asks two questions in writing: what was the most important thing you learned during this class, and what important question remains unanswered? It belongs at the end, and the second question is the valuable one: it gives you a list of what to change, in the words of the people who were there. Did the training work? Finding out afterward, and changing the session uses it as an end-of-session measure.

The Muddiest Point asks one question: what was the muddiest point in this session? The page describes it as well suited to large lower-division courses, and notes it is less suited to courses that emphasize integration, synthesis and evaluation. In a workshop it is cheap, it takes ninety seconds, and it surfaces the thing that nobody would raise a hand about.

The same page lists the Background Knowledge Probe, which Finding out what the room already knows uses before the day, along with Focused Listing, Concept Maps, Chain Notes and about forty more. Teaching moves and checks, side by side puts the ones that fit a short session into a table with what each costs and what it cannot tell you.

One warning about written end-of-session checks: they tell you what people think they learned, which is not the same as what they can do. Both are worth knowing and they answer different questions, which is the distinction Did the training work? Finding out afterward, and changing the session is built around.

7. Checks on a call Go deeper

Online you lose the glance that tells you a room is lost, which makes a built-in check more important rather than less. What works, roughly in order of how much it tells you:

  • Everybody types an answer at once. Ask for the answer in the chat but tell people to hold it until you say go; otherwise the first answer anchors everyone else. On go, the answers arrive together and you have a count.
  • A poll. Most platforms have one. It gives a clean count and it takes about a minute, including the time for people to find the button.
  • Thumbs or reactions. Fast, and coarse. Fine for a two-option question.
  • Cameras-off show-me. Ask people to hold the thing up to the camera. With cameras off, ask them to type one word describing what they are looking at.

Whichever you use, say what the count was out loud. "Eleven of you went for B, eight for A" tells a room what sort of room it is in, which is information people get for free in person and lose on a call.

8. See it See it

Fictional example

Grace's check comes back at 30 percent. Grace Abara teaches a sixty-minute online session on reading a utility bill. Thirty people, cameras off, no practice possible in the room.

Her objective: point to the line on your own bill that changed most since last year, and say what it is.

Her check, written word for word: "Look at the sample bill on screen. Which line tells you how much electricity you actually used? A, the amount due. B, the kilowatt hours. C, the service charge. D, the rate per kilowatt hour. Type your letter in the chat, and hold it until I say go."

Known: thirty people, chat available, a sample bill on screen, two minutes budgeted for the check.

The result: nine of thirty pick B. That is 30 percent, below the band.

What the wrong answers tell her: twelve picked A, which means they are reading the bill as a demand for money rather than as a record of use. Six picked D, confusing the rate with the quantity. Three picked C.

What she does, from her own sheet: below 35 percent, she reteaches rather than sending them into discussion, and she reteaches differently: instead of explaining the structure again, she puts two bills side by side with the same amount due and different usage, and asks what differs. Then she asks the question again.

Assumed, and whose assumption it is: Grace assumes the twelve who chose A are reading the bill as a demand rather than misreading the question. That is her assumption, and the second poll is what tests it. Under that assumption the side-by-side comparison is the right reteach; if the second count stays low, the question itself may be unclear.

Open, and who settles it: whether a second reteach fits in sixty minutes. Her run sheet settles that, and she has one segment marked as the first thing to cut.

9. Practice Practice

Minimum: one check written word for word, with its three responses. Time: about 30 minutes. Materials: your objectives and errors sheets, and the checks sheet in your session plan.

  1. Pick the objective you are least sure will land. That is where a check earns most.
  2. List the three mistakes you actually see. From your errors sheet or from memory.
  3. Write the question word for word, with those mistakes as the options. Plus the right answer. Three or four options.
  4. Write what each wrong answer tells you. One line each. Some will point at your demonstration rather than at the room.
  5. Write the three responses and the minutes. Low, middle, high, and the cost from the check cycle calculator, onto the run sheet.
Go deeper

Write a second check for the same objective, as a show-me rather than a question, and decide which one you will use. Then write the feature you will count: not "looks right", but the specific thing, "the seam underneath and the top tight". Counting a named feature is what makes a show-me a check rather than a glance. If you have a helper, give them the feature too and compare counts afterwards; the difference between two counts of the same room is a useful thing to know about your own observation.

10. Your session Your session

11. Check your work Check your work

Criteria

  • Is the question written word for word, rather than described?
  • Is every wrong answer a mistake somebody would really make?
  • Does every person answer, at the same time, visibly?
  • Have you written what you will do at low, middle and high agreement?
  • Does the check have minutes on the run sheet?

Worked example

Grace's check is written out in full, including the instruction to hold the answer. Her options are three real misreadings plus the right answer. Everybody answers at once in the chat, on a count, which is the online equivalent of hands. Her three responses are written, and the low-agreement response is a different reteach rather than a louder one. Her check has two minutes on the run sheet, which the check cycle calculator says is tight for a group of thirty with a re-poll; she knows that and has marked the segment after it as the one to shorten.

Common mistakes

  • "Any questions?" in the slot where a check should be. It samples the confident and tells you nothing from silence.
  • Options nobody would pick. They turn a check into a guessing game that everybody wins.
  • Sequential answers. If the first answer is visible, the rest of the room agrees with it. Count of three, or hold it until go.
  • No planned response. Without the three sentences, a bad result produces the thing you were going to do anyway.
  • A check with no minutes. It gets cut, and it was the part that told you whether the session worked.
  • Reading 80 percent as success. It may mean the question was easy for this group.

What is still unresolved

This chapter does not cover the discussion a check opens, or how long to wait after asking: that is Asking, waiting and handling what comes back. It does not price the whole run sheet, which is Workshop timing: minutes on paper and minutes in the room. And a check tells you about the room on the day, which is a different question from whether anything survives the week, as Did the training work? Finding out afterward, and changing the session explains.

12. Knowledge check and scenario

What makes a check different from a quiz?
A quiz measures; a check changes what happens next. The test is whether you have written down what you will do at each possible result. If the answer would not change the following ten minutes, the question is decoration and it is spending minutes that practice could use.
Your check comes back at 25 percent right. What does the published guidance suggest, and why not discuss it in pairs?
Below 35 percent, the SERC protocol says students do not understand the topic well enough to discuss it, or the question is unclear or too hard. Pair discussion at that level mostly spreads the wrong answer and costs you two minutes. Reteach the step, differently rather than louder, then ask again. If the second count stays low, suspect the question.
Why is a wrong answer that points at your demonstration the most useful result?
Because it is the one you can fix cheaply and immediately. If people picked the option that corresponds to copying a hand position rather than a movement, the problem is a sightline or a merged step, and the remedy is to show it again from a different angle. Mistakes manufactured by the teaching are the mistakes most under your control.
(Go deeper) What are the published timings for one ConcepTest cycle?
A short presentation of typically five to fifteen minutes, the question posted, thirty seconds to a minute of individual thinking, answers shown by hands, cards, sheets or clickers, then one to two minutes of pair discussion when the result is in the 35 to 70 percent band, then a second poll. The page also recommends against grading the questions.
You planned three checks and the session is running fifteen minutes late at the halfway point. What do you cut?
This is the default and it removes the only thing telling you whether the session is working. You will finish on time having delivered content to a room whose state you cannot describe. If minutes have to come from somewhere, take them from a segment where the room is listening.
Teaching segments are the compressible part, because the material they carry can mostly go into the handout, and because a check is what tells you whether the remaining time is being spent well. Keep the check attached to the practice that matters most, and say out loud what you are skipping and where to read it.
Defensible only if the practice block is a repeat. Cutting practice to keep checks leaves you measuring a thing people did not get to do. If the last block is the second attempt at something, this is a real option, and the spacing you lose is a genuine cost.

Checklist

Records your learning only; it is not a qualification.