Assessment that actually tells you something

Most assessment in language classrooms produces a number and very little else. The number goes into a spreadsheet, the students look at it for four seconds, and nothing about next week's teaching changes.

That is a lot of work for very little information. Here is how to get more out of the same hours.

The question assessment should answer

Not "how good is this student?" but "what should I teach next, and to whom?"

If a task cannot change what you do next week, it is administration, not assessment. That is not always avoidable — institutions require grades — but it does tell you where to spend your design effort.

Separate the two jobs

Assessment does two different jobs and they conflict.

Certifying — producing a defensible grade for a report card. Needs to be fair, consistent and documented.

Informing — finding out who has understood what. Needs to be frequent, cheap and low-stakes.

Trying to do both with the same instrument is why so much assessment is exhausting and uninformative. Certify a few times a term, formally. Inform constantly, informally, without grading most of it.

Cheap ways to inform

The two-minute oral check. While the class works on a task, call three students in turn and ask them to say two sentences about something. Thirty seconds each. Over two weeks you have heard everyone. Nothing beats this for finding out who can actually produce language.

Exit tickets. One sentence on a slip of paper at the end of class: the answer to a specific question. Five minutes to read thirty of them, and you know whether to move on.

Whiteboard responses. Everyone answers at once, everyone shows at once. You see the whole class's understanding in one glance instead of the one student who volunteered.

The same task twice. Give a short writing prompt in week one and the identical prompt in week eight. Do not grade the first. The comparison is the most motivating evidence a language student can see, because progress in a foreign language is invisible from the inside.

Rubrics people can actually use

A rubric with five criteria and five levels is twenty-five cells. Nobody reads that, and applying it consistently across thirty texts is impossible.

Better: three criteria, three levels, and each descriptor written as something observable.

Bad descriptor: Uses grammar appropriately.

Good descriptor: Past-tense verbs are correct in most sentences; errors do not block meaning.

The test for a descriptor is whether two teachers reading the same text would put it in the same box. If not, it is too vague.

Give students the rubric before they write, not with the grade. A rubric that arrives after the work is a justification for a mark; a rubric that arrives before is a description of the target.

Feedback that produces a change

Three rules that cost nothing:

Fewer comments, acted upon. Fifteen corrections produce zero revisions. Two comments plus ten minutes of class time to act on them produce two changes. Choose two.

Comment on the pattern, not the instance. Not "wrong verb" in eleven places, but "your past tense is solid in short sentences and breaks down in long ones — look at sentences 3, 7 and 12."

Ask, do not fix. "What is the subject of this sentence?" teaches. Writing the correct sentence for them does not.

And build in the time to respond. Feedback with no revision opportunity is a report, not teaching.

Assessing speaking without losing your mind

Speaking is the skill most avoided in assessment because it seems to require one-to-one time you do not have. Two ways around it:

Recorded tasks. Students record sixty seconds on their phones and send the file. You listen while commuting. No class time consumed, and shy students often perform better.

Paired tasks with a simple observation grid. Students work in pairs on a task; you circulate with a grid of three criteria and tick as you go. You will not capture everything. You will capture enough.

What to do about the test-prep pressure

Standardised tests are a real constraint in Colombian schools, and pretending otherwise does not help anyone. But test performance is mostly a by-product of general proficiency plus familiarity with the format. Familiarity takes a few sessions. Proficiency takes the whole year.

Teaching exclusively to the format for months tends to produce students who recognise question types and cannot use the language. Do the format work deliberately and briefly, and spend the rest of the time on the thing the test is trying to measure.

A place to start

Pick one class. For two weeks, add a two-minute oral check and an exit ticket, and grade neither. Then look at what you learned about who is where — and change one thing in your plan because of it.

That is the whole loop. Everything else is refinement.

Keep going

• The webinar library has recorded sessions on assessment, including digital assessment practices and task-based assessment.
• Assessment is a recurring topic in HOW Journal — see the current and past issues.
• Join a Special Interest Group to keep the conversation going through the year.