← Teacher resources

How to assess speaking in Spanish class with five sections

Assess speaking with many small formative checks and a few scored tasks: an unrehearsed conversation for interpersonal skill and a prepared talk for presentational skill, each judged on the task, how understandable it is, the size of the language and how well the student keeps it going. Across five sections, combine partner rotations, sampling a few students a day, recorded assessments and exit-ticket recordings, knowing that each one gives up some evidence, some class time or some of your evenings.

Reed, co-founderGuide9 min read

On this page
  1. What’s the difference between formative and summative speaking checks?
  2. Interpersonal and presentational speaking need different tasks
  3. How often should you assess speaking?
  4. What to listen for when a student speaks
  5. Four ways to keep it manageable across five sections
  6. Where Avita Edu fits

The hard part of assessing speaking usually comes after you’ve decided what you want to hear: you have to hear it from 150 people. This guide covers which tasks to use, how often to use them, what to listen for, and four ways to keep the whole thing workable, along with what each one costs you.

We’re Reed and Ethan, two founders who learned Spanish as adults and build speaking software. We’ve never managed a Spanish 2 class of 32. The logistics below come from ACTFL, state guidance and teachers who’ve published how they run their rooms. The opinions on trade-offs are ours, so check them against your own classes.

What’s the difference between formative and summative speaking checks?

Ohio’s world language assessment guide reduces it to two words: formative means feedback, summative means summary. A formative check tells you and the student what to do next, and it can be as small as listening in on a think-pair-share. A summative task shows what a student can do at the end of a unit or a longer stretch.

The same guide splits summative evidence in two. A performance assessment uses the material you just taught. A proficiency assessment asks for language a student can use without rehearsal, usually over a semester or a year. A student can ace the restaurant skit and still stall when you ask what she did last night. Both results are accurate, and they answer different questions.

Formative Summative: unit performance Summative: proficiency
Question it answers What should we practice next? Can they use this unit’s language? What can they do unrehearsed?
Typical task Partner chat, question cards, exit recording Scored conversation or talk on the unit topic Conversation on familiar topics, questions not seen in advance
Preparation As much as helps Practice on the topic General practice only
What you record A note, a tally or nothing Rubric score Holistic rubric score

Interpersonal and presentational speaking need different tasks

The NCSSFL-ACTFL Can-Do Statements separate the modes this way: interpersonal communication means people interacting and negotiating meaning, and presentational means one speaker informing, explaining or narrating for an audience.

Scoring one tells you less about the other than you might hope. In their 1986 paper on language anxiety, Horwitz, Horwitz and Cope describe college students who felt fairly comfortable with drills and prepared speeches and froze in role-plays. That comes from counselors’ experience, not a measurement, but it’s a useful warning about reading a polished talk as evidence of conversation.

Two common tasks sit in a different mode than their labels suggest. A memorized skit is presentational, since nothing unexpected can happen in it. A student recording an answer alone is mostly presentational too, because nobody can say ¿Cómo? (What?) and make them try again. If you want to record interpersonal work, record two students talking.

Interpersonal Presentational
Try this Random pairs chat for two or three minutes: ¿Qué haces los fines de semana? (What do you do on weekends?) One minute on a prepared topic: Describe tu rutina de la mañana. (Describe your morning routine.)
Listen for Asking back, reacting, repairing when meaning breaks down Order, detail, staying on topic, clarity for a listener
Preparation General practice, not the exact questions Planning and rehearsal are expected
Main risk One partner carries the other, and pairs are hard to hear in a noisy room Reading a script, which shows rehearsal more than spoken ability

For the timed, one-sided format on the AP exam, see our AP Spanish speaking practice guide.

How often should you assess speaking?

We looked for a study that gives an ideal frequency and didn’t find one, so this schedule is a starting point and not a standard. The Can-Do guidance does give a principle: showing an ability once isn’t enough, and students should show it repeatedly across situations. That argues for many small checks and a few scored tasks, instead of one big oral exam.

How often What Stakes
Most days Listen in on partner work Feedback only
Weekly One sampled check or one exit recording None or light
Each unit One scored task, alternating interpersonal and presentational Gradebook
Each semester One unrehearsed conversation Gradebook, on a proficiency-style rubric

Horwitz and her colleagues also note that an oral test can set off test anxiety and speaking anxiety together. Allison Wienhold of Mis Clases Locas uses the same question cards for practice and assessment, so the format is never a surprise. Our guide on reluctant and anxious students has more on this.

What to listen for when a student speaks

ACTFL’s Proficiency Guidelines 2024 judge speaking on four criteria: functions and tasks, context and content, text type, and accuracy. That’s a lot to track during a two-minute chat, so we’d turn them into four questions, asked in this order:

  1. Did they do the task? If you asked for three weekend plans and a question back, did you get them?
  2. Could a listener follow it? CARLA’s generic rubric frames this around listeners used to language learners, which keeps accent out of it.
  3. What size is the language? ACTFL’s OPI guide places Novice speakers at words, lists and phrases, and Intermediate speakers at sentences. You can usually hear this within a few seconds.
  4. Do they keep it going? Listen for questions back, reactions and attempts to fix a misunderstanding.

Here are two invented answers to the same question:

¿Qué haces los fines de semana? (What do you do on weekends?)
Fútbol. Con amigos. Me gusta. (Soccer. With friends. I like it.)
¿Qué haces los fines de semana? (What do you do on weekends?)
Los sábados juego al fútbol con mis amigos porque es divertido. ¿Y tú? (On Saturdays I play soccer with my friends because it’s fun. And you?)

Neither has an error worth stopping for. The second has sentences, a reason and a question back, and you can hear the gap between them without counting a single mistake. Our free rubric has sample answers at every band.

Teachers disagree about how much accuracy should count. One view holds that students should be accountable for structures they were taught, and that a speaking grade should show it. The proficiency view holds that students grow by taking risks, and that heavy accuracy grading teaches them to say less. Both are reasonable. We’d keep accuracy as a rubric row that counts less in Spanish 1 and more each year, and we’d score taught structures more tightly on a unit task than on an unrehearsed one. Whether to correct those errors out loud is a separate question, and we’ve summarized what the research says about it. If you have heritage speakers in the room, the questions still work, but their targets will differ.

Four ways to keep it manageable across five sections

Say you have five sections of 30. That’s 150 students, and a two-minute recording from each is five hours of listening before you’ve written a comment. Each option below gives something up to avoid that.

Option Best for Time outside class Biggest weakness
Partner rotations Formative practice None You hear a slice of each student
Sampling a few a day Formative checks, light grades Almost none One short sample per student
Recorded assessments Summative tasks, retakes High Listening hours, and outside help at home
Exit-ticket recordings Deciding what to reteach Low if you sample Thin evidence on any one student

Partner rotations

Pairs talk, a timer goes off, one student moves. Wienhold’s rotating partner conversations ran four to five minutes a partner in Spanish 3, and she suggests one minute for novices. She walks with a clipboard of names and takes brief notes, which she says helps keep students on task in Spanish.

Everyone talks and you take nothing home. You also hear only part of each student, some students may switch to English the moment you walk past, and the noise makes fine judgments hard. We’d use rotations for practice and for formative notes, and we wouldn’t hang an individual grade on them.

Sampling a few students a day

While the class is in partner talk, pick four or five names and score each on a quick four-point check. At five a day, a section of 30 takes six class days. Wienhold describes a version of this with small groups chatting at once while she walks and assesses a few students or all of them.

Each student gets one short sample, on whatever day and prompt came up, so scores aren’t strictly comparable. Students who’ve been checked may also relax, so draw names at random and let some come up twice. A simple rubric helps. In a study of two classroom speaking tests at a Japanese high school, teachers using rubrics with three or four criteria and three levels, after about ten minutes of discussion, reached adequate reliability on overall scores, though agreement on single criteria varied widely. That was English class and group tasks, so take it as encouragement and no more.

Recorded assessments

Students record in class in random pairs, or alone at home. Wienhold’s guide to assessing speaking uses recording for large classes and for students who don’t want to be overheard, and she treats it as a last resort because grading it outside class takes so long.

You gain the ability to replay, to compare scores with a colleague and to offer retakes. Anxious students may benefit too: Poza’s 2011 study of an asynchronous voice tool found less fear of negative evaluation and more risk-taking, and linked that to losing the time pressure and being able to edit. Being able to edit is also the cost, since at home a student can script the answer, get help or re-record until it stops sounding like them. Recording in class, in random pairs, solves much of that. Cap the length, and check your school’s policy on student recordings first.

Exit-ticket recordings

In the last three minutes, students record 20 or 30 seconds answering one prompt: ¿Qué vas a hacer esta noche? (What are you going to do tonight?) Then listen to ten, not thirty, and look for patterns. Ohio lists exit slips among its formative examples, but we didn’t find research on the spoken version, so this one rests on common sense and teacher practice.

Room noise and device trouble are the daily problems, and 30 seconds won’t support a grade. It’s best at telling you what to reteach tomorrow.

Scored days: pairs at your desk

For the task that counts, the same Wienhold guide calls random pairs to the desk for two or three minutes each, with rubrics already labeled with names and the rest of the class working quietly from a projected list. She finishes in one or two class days. At upper levels she runs random groups of about four at once and scores on the spot with a clipboard. You lose the class time, and you need a room that will stay quiet. CARLA’s archived modes resources describe a similar idea, student-to-student conversations with a self-assessment beforehand, used to replace end-of-semester interviews.

Our take: Hear everyone at scored length a few times a semester, and hear a few students every day the rest of the time. We’d take 30 small, imperfect samples over one exhausting oral exam, though you know your room and we don’t.

Where Avita Edu fits

Avita Edu sits closest to the recorded option. You assign a short conversation for class or homework, each student talks through it out loud in the browser, and you get back a gradebook, transcripts, playback and mistakes grouped across the class, so you can see what to reteach. You can override any score. Library topics are matched to units in books such as Realidades 2 and Descubre 1. We did that alignment ourselves and aren’t affiliated with any publisher.

It doesn’t score pronunciation, it can’t stop a student from getting help outside the app, and the other speaker is software, so it doesn’t replace hearing two students talk. Teacher accounts are invite-only during the pilot, and we have no school results to show you. If a free voice recorder does what you need, use that.

Questions teachers ask

How do you grade speaking in a Spanish class?
Use a short rubric with a few rows: whether the student did the task, whether a listener used to learners could follow it, the size of the language from words to connected sentences, and how well they keep a conversation going. Weight accuracy lightly at novice levels and more each year. Keep scored tasks to about one a unit, and treat daily listening as feedback, not as a grade.
How often should students be assessed on speaking?
We found no study that sets an ideal frequency. A reasonable starting point is listening in on most days, one sampled check or exit recording a week, one scored task per unit, and one unrehearsed conversation per semester. ACTFL's Can-Do guidance supports the general shape, since it says students should show an ability repeatedly across situations, not once.
How do you assess speaking with a large class?
Mix formats. Use partner rotations for practice, score four or five sampled students a day during partner talk, and call random pairs to your desk for two or three minutes on scored days while the class works quietly. Recordings work too, but 150 two-minute files means five hours of listening, so cap them and sample where you can.
What is the difference between interpersonal and presentational speaking assessment?
Interpersonal assessment scores a two-way conversation, so you listen for asking back, reacting and repairing misunderstandings, and students shouldn't see the exact questions ahead of time. Presentational assessment scores one speaker delivering prepared information to an audience, so you listen for order, detail and clarity. A memorized skit or a solo recording counts as presentational, even when it looks like a conversation.
Should accent or pronunciation count in a speaking grade?
We'd score whether a listener used to language learners can understand the student, and leave accent out. An accent that doesn't block meaning shouldn't cost points, while pronunciation that does block meaning will already show up under comprehensibility. Treat other pronunciation work as formative practice, with modeling and specific feedback, separate from the rubric score.
Can students record speaking assessments at home?
They can, and it suits presentational tasks, retakes and students without a quiet moment in class. The weakness is that you can't see how the recording was made, since a student can script it, get help or re-record many times. Keep at-home recordings low in weight, use personal prompts, and ask a quick follow-up in class. Check your school's policy on student recordings too.

Hear every student in your class this week.

Avita Edu lets you assign speaking practice for class or homework. Students talk it through out loud in the browser, and you get it back graded. Free for teachers.

Invite-only while we open to our first teachers. Have a code? Create your account

See how it works