Separate explanations can hide the real difficulty
Imagine learning two constructions on Monday and Thursday. Both pages make perfect sense. The problem appears on Friday when a real sentence could tempt you toward either one. If the materials never put the alternatives next to each other, the learner has not practised the point where the error actually happens.
A useful Contrast is therefore selective. It does not build a giant graph of everything that looks vaguely similar. It states the distinction that changes a language decision: meaning, register, grammatical constraint, speaker stance or typical context.
Choice turns explanation into retrieval
Metkagram’s Choice object presents a bounded situation and asks the learner to select or produce a Frame before seeing feedback. That small commitment matters because rereading a contrast and making a decision are different cognitive events. The first can feel fluent while the second exposes what is actually retrievable.
Retrieval research is one reason to take this distinction seriously, although a result from memory research does not automatically validate every language exercise. The design implication is modest: if future use requires choosing without an answer key, some practice should also require choosing without an answer key.
Good contrasts are about boundaries, not trivia
A contrast should answer a question a learner could realistically ask: Which phrase is too formal here? Which construction changes the implied certainty? Which one requires a completed event? Why does the direct translation from Russian sound possible but unnatural in this context? These distinctions produce decisions worth practising.
By contrast, obscure terminology that never changes production is mostly metadata. Metkagram can store technical detail, but the learner-facing comparison should prioritise the boundary that helps the next sentence come out better.
The contrast belongs back in a sentence
After choosing, the learner should see the Frame in context again. Otherwise contrastive practice becomes another multiple-choice game detached from language. A strong loop is: inspect two real examples, state the decisive difference, choose for a new situation, receive feedback, then produce one more example of your own.
This is also where mistakes become useful data. A repeated wrong choice can point to a weak distinction, an unclear explanation or a transfer problem that deserves its own study set.