A student scores ninety percent on a grammar test, then struggles to ask a simple question in conversation. This does not make the test worthless. It means that the test and the conversation required different kinds of performance.
Assessments measure particular skills
Multiple-choice items can show whether a learner recognizes a correct form among options. An open writing task requires producing that form independently. Speaking adds time pressure, listening and interaction. Success in one task cannot automatically stand in for all the others.
Begin by defining the skill you want evidence of. Then choose an assessment that actually requires it.
Prompts can create an illusion
A worksheet titled “Past Simple” tells learners which grammar area to use. Real communication does not provide that signal. A student may do well on a whole page of one construction but struggle when several forms could fit.
Useful checks include mixed tasks, fresh contexts and questions in which the learner must decide what matters.
Look for transfer and retention
Ask whether the student can apply an idea to a case that looks different from the lesson. Return after several days or weeks rather than relying only on the immediate post-lesson result.
In spoken language, notice whether the learner can ask for clarification, adjust a reply or recover from a misunderstanding. Counting errors alone may miss important aspects of communication.
What does a high test score actually establish?
A learner receives 90 per cent on a multiple-choice grammar test. The score accurately reports performance on that test. It does not automatically tell us how well the learner can write a difficult email, respond to an unexpected question or retain the grammar several months later.
Every assessment samples a particular kind of behaviour. Choosing an answer from options is not the same activity as producing one without cues. Correcting an error is not the same as speaking fluently while attending to another person’s reply. A thoughtful assessment therefore starts with the intended skill: what should the learner be able to do?
This distinction is especially important when a school uses scores to communicate progress. A result can be useful and motivating without supporting unlimited claims about competence.
Recognition, recall and application
Imagine three tasks involving the same grammatical contrast. First, the learner chooses the right tense from four options. Second, they complete a sentence without a list of forms. Third, they explain a personal experience in a conversation with a new partner.
The first measures recognition of a form in a supplied context. The second requires retrieval and production under more controlled conditions. The third adds an independent communicative decision while the learner manages meaning and social interaction. Performance may differ substantially across the tasks.
| Assessment | What it directly measures |
|---|---|
| Multiple choice | Choosing from visible alternatives |
| Completion without options | Retrieving a particular form |
| Free written response | Constructing a message independently |
| Live conversation | Managing language during interaction |
| Delayed new task | Some degree of retention and transfer |
None of these formats is inherently worthless. Problems arise when a measure is treated as evidence for skills it was not designed to test.
Why examples and cues sometimes create illusions
During a lesson, students see grammar tables, model dialogues and teacher prompts. These supports are useful for learning but can make immediate performance look stronger than independent competence. A learner may complete a paragraph accurately because every example contains the same distinctive time expression.
A more revealing check removes some cues. Ask students to tell an unfamiliar story or write a short message where the target form could be useful but has not been explicitly named. Watch whether they choose it appropriately.
If the result is weaker, the earlier exercise was not necessarily pointless. It may have taught a component skill. The next step is to build a bridge from supported choice to self-directed use, not to dismiss everything already learned.
What transfer research finds
Pan and Rickard’s 2018 meta-analytic review drew on 122 experiments examining whether learning through retrieval can transfer to different kinds of tasks. The authors reported positive average transfer relative to re-exposure controls, while emphasising variation across types of test and learning condition.
The analysis does not show that a high quiz score automatically predicts real-world conversation ability. It shows that transfer can occur under certain circumstances and that the details matter. Read the study abstract and bibliographic record for the scope and measurement.
A separate classroom or tutoring exercise should not be presented as a replication of these experiments. The result of an individual lesson helps plan instruction; controlled causal conclusions require a different research design.
Designing an assessment that matches the goal
Suppose a student is preparing to communicate at work. The learning objective might be ‘Can clarify a project deadline and propose an alternative.’ A multiple-choice vocabulary quiz can check whether useful words are recognised. But an assessment of the actual objective should include a conversation or written exchange requiring the student to ask questions and propose a workable solution.
The task should be clear enough to measure the intended skill. Introducing deliberately confusing instructions or an unrelated specialist subject would add noise. A meaningful transfer task is new in the decision it requires, not arbitrary in every detail.
Similarly, a student preparing for the Goethe B1 exam needs practice with the actual official format because exam performance is a legitimate goal. It should be complemented by communication outside that format if everyday language use is also the objective.
Feedback should be diagnostic
A raw score tells you how many items were correct. It rarely tells you why the incorrect responses happened. A student may have misunderstood the task, forgotten a word or used the right construction with an inaccurate ending. Those problems call for different next steps.
After a test, identify a small number of recurring difficulties and choose a practice task for each. If students confuse two forms, use contrasts. If they can recognise a word but cannot produce it, practise retrieval. If they can write a form accurately but not use it during speech, train a short interactive exchange.
This approach makes the test part of learning rather than a final label attached to the student.
A simple two-stage progress check
At the end of a study unit, ask the learner to solve one familiar task without prompts. A week later, give a related but different task and compare. Record not only accuracy but whether the learner could begin without help, select a suitable strategy and recover after an error.
Avoid claiming that an improvement in two personal attempts proves the superiority of a teaching method. It is evidence of that person’s performance on those particular tasks. For a formal study, we would need explicit questions, controls, ethical data collection and appropriate analysis.
A useful score is one that helps answer a useful question. If a learner needs to speak with clients, include real speaking decisions. If they need to read academic texts, ask them to evaluate claims and evidence, not just recognise isolated terminology.
Explore desirable difficulty and grammar that works for ways to design practice after an assessment.
Evidence: Pan & Rickard (2018), transfer of test-enhanced learning and Dunlosky and colleagues on learning techniques.
A fictional assessment that reveals two kinds of progress
A student achieves a high score on a quiz about English past tenses. The teacher then asks them to describe an unexpected event at work. The student remembers the vocabulary but uses the same tense for every sentence. The multiple-choice result was genuine; it simply measured recognition under specific conditions rather than the full set of decisions required for a story.
The next lesson separates the difficulties. The learner first contrasts background events and interruptions in short examples, then tells a new story without a tense label in the prompt. A later check changes the topic again. That sequence helps the teacher see whether supported recognition has begun to turn into independent production.
This is a fictional assessment example, not an observation of a LangiaFlow class or evidence that one lesson guarantees transfer.
Are multiple-choice tests useless?
No. They can efficiently assess recognition and particular distinctions. They are less informative about abilities that require generating a response, interacting with someone or making choices in unfamiliar contexts. Use them for the question they can answer.
What is transfer in learning?
Transfer involves applying what was learned in one setting to another task or situation. The new task must preserve enough relevant structure to make transfer meaningful while requiring a genuine independent choice. Changing every feature arbitrarily can make a test noisy rather than informative.
Should the final check happen immediately or later?
Both can be useful. An immediate check reveals what the learner can currently do; a delayed check provides information about retention. If real communication is the goal, at least some assessment should require independent writing or speaking rather than recognition alone.
A two-part assessment for language learners
A student finishes a multiple-choice exercise on the past tense with an excellent score. The teacher has evidence that the student can recognise the required forms among the provided options. The result says less about how reliably the same person will choose those forms during an unscripted account of an event.
To assess the communicative skill, add a second task: ask the student to describe an unexpected event from yesterday to someone who does not know the story. A partner should ask one question that requires clarification. The teacher can now observe both the accuracy of the forms and whether the student can continue communicating.
These tasks are not rivals. A well-designed recognition test can reveal a precise grammatical weakness; a communicative task shows whether that knowledge is accessible in a different context. The useful question is what evidence each assessment actually provides.
Why transfer is not automatic
The meta-analysis by Pan and Rickard examined transfer from testing across 122 experiments and found that the effect varied according to the kind of new task and the learning conditions. This is a stronger and more cautious conclusion than claiming that every quiz will improve any future skill.
A word-pair exercise may make vocabulary easier to retrieve, yet a new conversation also requires selecting the right register, understanding the listener and managing sentence structure. The transfer is possible, not guaranteed.
Build a small evidence record
| Evidence | What you may conclude | What remains untested |
|---|---|---|
| Correct answer with four options | Recognition in that format | Independent production |
| Explanation of a grammar rule | Explicit understanding | Quick use under conversational pressure |
| New spoken or written task | Application in a related context | Retention after a longer delay |
| Delayed task on another topic | Some retention and transfer | Performance in every real-life situation |
Avoid announcing a student’s whole competence from one successful or unsuccessful task. Language ability is made of several interdependent skills that may develop at different rates.
For a personal experiment, record one short task before practice, repeat the same kind of task several days later, then test a related but not identical situation. Analyse the errors rather than merely comparing totals. If your original mistake has disappeared but a new one has emerged, the practice may be working at one level while another requires attention.
A good score is worth recognising. It simply becomes more informative when you know precisely what the test was designed to measure.
Use results to plan the next step
A number is most helpful when paired with information about what happened. Did the learner misunderstand the task, forget a form, choose the wrong method or run out of time? Each pattern suggests a different next lesson.
Testing should provide an honest picture of a skill and help improve it. A high score can be worth celebrating while still leaving room for practice in the situations that matter outside the classroom.
