Somewhere after the first year a lot of people notice they are getting more things wrong than they used to, and conclude they have gone backwards.
The measurement says otherwise, and it is oddly specific about it: at B1, a Finnish learner gets the object case wrong about a third of the time.
That is not a complaint about learners. It is a measurement, from a project at the University of Jyväskylä that rated 669 texts written by adult learners of Finnish and then went looking for what changes between one level and the next.1 The texts sit in the Language Bank of Finland alongside the other L2 learner corpora, which is why the figures can be checked rather than taken.2 The number that matters is not the 70 per cent. It is where the 70 per cent sits: at B1, after the biggest jump in the whole scale, and one full level before accuracy arrives.

The two lines move at different times
The Cefling project measured three things about each structure: how often learners used it, how often they got it right, and what kinds of sentence they used it in. The transitive construction, a verb with an object, is the clearest case, because in Finnish the object case is genuinely hard and the rest of the construction is not.
| level | what happens to the transitive construction |
|---|---|
| A1 to A2 | Present but infrequent. Learners often use a transitive verb with no object at all |
| A2 to B1 | The sharpest rise in frequency in the whole scale. Accuracy stays where it was |
| B1 | Object-case accuracy around 70 per cent. Roughly one use in three is wrong |
| B1 to B2 | The accuracy leap. 86 per cent target-like, which the project counts as mastery |
There is a stretch, and it is exactly the stretch people describe as a plateau, where the amount of Finnish you attempt goes up steeply and the proportion you get right does not move. Your error count rises. Your level rises too. Both things are the same event.
The researchers state the pattern plainly: the growth in quantity happens between A2 and B1, while accuracy does not increase until B2.1 Negation follows a different schedule again, with frequency flat across the levels and the accuracy growth landing on the A2 to B1 step.

What you are leaving behind is avoidance
The most useful thing in the data is not about B1 at all. It is about what A-level Finnish looks like from outside.
At level A, the study found, learners tend to leave the object out of sentences that need one: they use transitive verbs intransitively more often than more advanced learners do. The example the authors print, reproduced with the learner’s own spelling, is Minä sain uuden valokuvian ja näytän sinulle, where the second half has a verb, a recipient and no object at all.1
An A-level sentence stays safe by saying less. Leave out the object and there is no case to choose. Stick to the present tense and there is no past-tense stem to form. Use on and you never need the conditional.
Crossing into B1 means giving up every one of those exits at roughly the same time. The sentences get longer, the choices multiply, and the error rate goes up because you are finally in a position to make those errors at all.
This is the part that is worth saying to yourself in the bad month. A learner who has stopped making object-case errors has not necessarily improved. They may just have stopped writing sentences with objects in them.
The choice that stays wrong longest
Finnish has two ways of marking a total object and one way of marking a partial one, and the choice is not free: it turns on whether the action is complete, whether the verb is negated, and whether the thing is countable. A negated clause takes the partitive regardless of everything else, and the standard grammar devotes a section to the conditions under which the object loses its -n.3
| Written | What changed |
|---|---|
| Luin kirjan. | The book got finished |
| Luin kirjaa. | The reading was going on |
| En lukenut kirjaa. | Negated, so partitive, and the completion question does not arise |
| Ostan auton. | One whole car |
| Ostan autoa. | Marked, and it means you are in the process of buying |
Nothing on that list is exotic. All of it is the first hundred verbs. And the measured result is that two CEFR levels after meeting it, a learner is still getting roughly one in three wrong.
That is the honest expectation to carry. Not that the object case is a thing to fix before moving on, but that it is a thing that stays partly broken while everything else improves, and then resolves at B2.
Why the leap sits where it does
The authors put their own explanation in the title of the chapter. The CEFR calls the B levels the Independent User band, and their suggestion is that the jump reflects something real about independence rather than an accumulation of forms: to stop depending on support you have to start attempting things you cannot yet do reliably.1
There is a second, duller explanation and they raise it too: the CEFR levels may simply not be equally spaced, so more happens between A2 and B1 than between any other pair. The data cannot separate the two.
Either way, the practical reading is the same. The A2 to B1 stretch is where the number of things you are trying to do outruns your control of them, by design, and it stays that way for a level.


The input problem underneath it
One more thing changes at the same point, and it is not linguistic.
Finland has a genuinely good supply of simplified Finnish. Selkokieli is a defined register with its own published guidelines, its own certification mark and its own news service, and Selkokeskus estimates around 800,000 people in Finland need it.4 What it is not is a ladder. There is one simplified register, aimed at everyone who finds reading hard, and no graded step above it, which leaves ordinary Finnish as the next thing to read.
That is the part that feels like losing ground: not a drop in ability, a step change in what is available to practise on.
The two halves of the problem also want different treatment, and only one of them solves itself. Comprehension responds to volume, so more listening at whatever difficulty is to hand, for longer. Accuracy does not, because it needs somebody to tell you which of two forms you produced, and Finns correct in conversation far less often than learners expect. A teacher of nearly 30 years told Yle that teaching materials barely address the gap between the standard language and the spoken one.5 The correction has to be arranged.
The cliff is also not the same height for everybody. In the integration-training data, prior education and the level people started at both predicted where they finished, which is a more uncomfortable finding than the plateau itself and a better guide to how much time to budget.6
So for the year after A2: volume for the listening, and a deliberate arrangement for the correction, because only one of those two arrives on its own. For the volume half there is a ranking of Finnish television by how much real speech is in it, which is a more useful filter than picking by genre.
Footnotes
-
On Becoming an Independent User - Maisa Martin, Sanna Mustonen, Nina Reiman and Marja Seilonen, EuroSLA Monographs 1, pages 57-80, reporting the Cefling project at the University of Jyväskylä. 669 texts by adult learners of Finnish as a second language, rated on a purely communicative CEFR scale from A1 to C2. The project’s DEMfad model counts a structure as mastered at 80 per cent target-like use ↩
-
L2 learner corpora - Kielipankki, the Language Bank of Finland, which holds the Cefling and Topling learner data ↩
-
Iso suomen kielioppi § 934 - Institute for the Languages of Finland, on the conditions under which the object appears without its -n ↩
-
Selkokieli - Selkokeskus, the Finnish Centre for Easy Language. Selkokieli is a simplified register with published guidelines and a certification mark; Selkokeskus puts the number of people in Finland who need it at about 800,000 ↩
-
D-äänteen hankala käytös on vain yksi veikeä erityispiirre, kun vieraskielinen opettelee suomea - Yle, interviewing Auli Suonio, head of training at Vantaan aikuisopisto ↩
-
Koulutustaustan ja lähtötaitotason merkitys suomen kielen oppimiselle kotoutumiskoulutuksessa - Tanja Seppälä, Puhe ja kieli 42:1 (2022), pages 17-44, University of Jyväskylä ↩











