An oral language assessment checks how well a student speaks: how clearly, how fluently, with which words and grammar, and how well they respond to another person. It's how teachers track progress in speaking, a skill that leaves no written answer to mark.
Judging speaking from one end-of-term interview is like judging a student's maths from one question on one day.
Many English teachers already use a rubric, a scoring guide that describes each level of performance. The harder part is finding time to hear every student often enough for the score to mean something. This guide follows one teacher through five steps. It splits the job in two: short recorded tasks that collect evidence every week, and a few live tasks where the teacher's judgement matters most.
Illustrative scenario: Ms Anh teaches four Grade 8 English classes of 32 students each. Her department wants a speaking score for every student each term. Until now, she has run one short interview per student in the last two weeks of term.
What does an oral language assessment measure?
An oral language assessment measures the parts of speaking a listener notices: pronunciation, fluency, vocabulary, grammar and interaction.
- Pronunciation: how clearly the student says sounds, words and sentences, including word stress and intonation (the rise and fall of the voice).
- Fluency: how smoothly the student speaks, without long pauses or frequent restarts.
- Vocabulary and grammar: whether the student uses the right words and sentence patterns for the topic.
- Interaction: whether the student listens, responds and keeps a conversation going.
Each part needs a different kind of task. A read-aloud task shows pronunciation and fluency well. It shows very little about whether a student can answer a question they didn't expect.
Pronunciation is about being understood, not about copying one accent. Many schools describe language ability with the Common European Framework of Reference for Languages (CEFR), a six-level scale from the Council of Europe. Its updated pronunciation scale moved away from native-speaker models. It cites research showing that being easy to understand matters far more to communication than accent (Council of Europe, CEFR phonological competence).
| What to assess | What to listen for | Tasks that show it well |
|---|---|---|
| Pronunciation | Can a listener understand each word? Sounds, word stress, intonation | Recorded word and sentence practice |
| Fluency | Pace, pauses, restarts | Reading a passage aloud, a short prepared talk |
| Vocabulary | Range and accuracy of words for the topic | Prepared talk, conversation |
| Grammar | Accuracy of the sentence patterns taught in the unit | Conversation, prepared talk |
| Interaction | Responds, asks questions, keeps a conversation going | Live pair task or role-play |
Why is speaking so hard to assess in a full class?
Speaking is hard to assess in a full class because each score needs time to listen to every student and a consistent way of scoring.
Time first. In the illustrative scenario, Ms Anh gives each student 90 seconds to speak and takes 30 seconds to write down a score. That's 2 minutes per student and 64 minutes per class. Across her four classes, it adds up to about 4 hours 16 minutes for one score per student, and the rest of the class waits while each student speaks.
Consistency is the second problem. Two teachers can hear the same answer and give different scores. A 2025 study at Chulalongkorn University in Thailand asked two teachers to score 28 recorded student presentations using the same detailed scoring guide. The authors found that scores became more reliable when more than one person scored, and recommended detailed scoring guides and meetings where scorers agree how to use them (Limgomolvilas and Sukserm, LEARN Journal, 2025).
A teacher who scores 128 students alone, once a term, has no second scorer to check against and only one piece of evidence per student.
How do you set speaking criteria before the unit starts?
Choose three to five speaking criteria, describe three or four levels for each, and test the rubric on two sample recordings with a colleague.
- Pick criteria that match the unit. A unit on giving directions needs interaction. A unit built on a reading text needs pronunciation and fluency.
- Describe each level in words a student understands. For example, write "Most words are clear to a listener" rather than "Adequate phonological control".
- Score two recordings with a colleague. Where your scores differ by more than one level, reword the description until you both read it the same way.
- Share the rubric with students before the first task, so they know what each score means.
Illustrative scenario: Ms Anh picks pronunciation, fluency, unit vocabulary and interaction. She and a colleague score two recordings from last year. They disagree on fluency, so they add one line to the rubric: "Short pauses to think are fine; long pauses in the middle of a sentence are not."
How do you collect speaking evidence every week?
To collect speaking evidence every week, use short recorded tasks built from the unit's material, so every student speaks and the teacher reviews results later.
Recorded practice moves most of the listening out of lesson time. The whole class can practise during the same slot, each student on their own device, and the teacher reviews the results afterwards. ELSA's guide to speaking activities for ESL students explains why teachers rarely have time to hear every student speak in a lesson.
With ELSA Schools, teachers turn their own textbook material into speaking assignments. Students complete them in a web-based student app and get a score and feedback straight after each attempt. Teachers see results for each student and each class on a dashboard.
| Practice type | What the student does | What the score covers |
|---|---|---|
| Vocabulary and sentences | Says key words and sentences from the lesson | Pronunciation and delivery |
| Reading aloud | Reads a full passage aloud | Pronunciation, intonation and fluency |
| Real conversations | Holds a conversation with an AI partner, using the unit's words | Clarity, pronunciation and delivery |
Illustrative scenario: On Monday, Ms Anh assigns vocabulary and sentence practice with the unit's 15 key words. On Wednesday, students read the textbook passage aloud during a ten-minute slot she keeps free in the lesson. On Friday, they complete one conversation with an AI partner at home. By the end of the week, she has three pieces of speaking evidence for each student without listening to each one live.
When should the teacher still listen live?
Teachers should still score live tasks for what recordings judge less well: ideas, conversation and answers to questions a student didn't prepare for.
Automated scores suit some tasks better than others. A 2026 study in the United States compared an automated speech-scoring system with human scorers on early reading tasks for 429 kindergarten children. The system and the humans agreed fairly well on vocabulary tasks. They agreed poorly on tasks where children combined separate sounds into a word (Petscher et al., Frontiers in Education, 2026).
That study tested one system with much younger children, so its results don't carry over directly to a Grade 8 English class. The practical point does: before relying on an automated score, check which tasks it suits. For more on how AI scoring supports students, see how AI speech analysis helps students who struggle to speak in English.
Live tasks that work well:
- A pair interview where each student asks and answers three questions.
- A two-minute talk followed by one question the student hasn't seen.
- A small-group discussion, scored with a short checklist of what to look for.
Illustrative scenario: Ms Anh now runs one live pair task per unit instead of one interview per term. She uses the weekly results to decide whom to listen to first: students whose scores haven't moved, and students who score well on recordings but rarely speak in class.
How do you turn speaking scores into next steps?
Use each speaking score to decide what to reteach and which students to talk to, rather than as a final grade.
| If you see | Do | Owner |
|---|---|---|
| Many students score low on the same sound or word | Reteach it in the next lesson with a short drill, then reassign the vocabulary practice | Teacher |
| One student's scores haven't moved for three weeks | Talk to the student and check whether the task is at the right level | Teacher |
| High reading-aloud scores but weak answers in live tasks | Add pair tasks with follow-up questions | Teacher |
| Few students in one class complete their assignments | Keep five to ten minutes of each lesson free for practice | Teacher and head of department |
| Scores rising across the grade | Share two example recordings at the department meeting | Head of department |
How often should you review speaking results?
Review speaking results weekly to plan teaching, at the end of each unit to check the rubric, and each term for reporting.
| When | What to review | Who |
|---|---|---|
| Weekly | Class dashboard: how many students completed their assignments, and which words and sounds scored lowest | Teacher |
| End of each unit | Live task scores compared with weekly results; whether the rubric wording still works | Teacher with one colleague |
| End of term | Progress for each student and class; what to report to parents | Teacher and head of department |
What should teachers avoid?
- Scoring speaking from one task. One interview shows one moment. Combine weekly recordings with live tasks.
- Scoring accent. Score whether a listener can understand the student, in line with the CEFR's focus on being understood.
- Setting practice without lesson time. If no time is set aside in class, the students who need practice most are often the ones who skip it.
- Treating completion as progress. A finished assignment shows effort. The scores show what to reteach.
- Changing the rubric halfway through a term. Reword it at the end of a unit, so scores stay comparable.
Where does this approach fall short?
Recorded practice gives frequent evidence on how students speak, but it can't replace a teacher's judgement on meaning, ideas and real conversation.
- Students need a device and a fairly quiet place to record. Background noise affects recordings.
- Automated scores suit some tasks better than others, so teacher-scored live tasks stay in the plan.
- Agreeing on a rubric with a colleague takes time, usually one meeting per unit at the start.
- This approach supports classroom assessment. It doesn't replace an official speaking exam when a student needs a certificate.
Related reading
- Speaking activities for ESL students: a weekly plan
- How can AI speech analysis help students who struggle to speak in English?
- Blended English learning: how it builds speaking confidence
Recap
- An oral language assessment covers pronunciation, fluency, vocabulary, grammar and interaction.
- Set three to five criteria and test them with a colleague on sample recordings.
- Collect short recorded evidence every week from the unit's own material.
- Keep live tasks for ideas, conversation and unprepared questions.
- Review weekly to plan teaching, per unit to check the rubric, and per term to report.
To see how ELSA Schools fits weekly speaking practice into your lessons, book a demo.
