The question turns up everywhere: how long from B1 to B2 in Spanish? Published answers range from 200 to 600 hours, which is wide enough to be no answer at all, and almost nobody explains why the range is that wide.
There is a real number underneath it, and there is a better question sitting behind that. The number is worth having, because “it depends” is the answer people are given and it is not actionable. But the reason the published figures scatter across a factor of three turns out to be more useful than any single figure, and it changes what you would do with your next month. This is the same problem in a different costume from understanding far more than you can produce.
The short answer
Cambridge puts the step at roughly 150 to 200 hours, and the published disagreement is mostly about what counts as an hour. Their guided learning hours are cumulative: about 350 to 400 to reach B1 and 500 to 600 to reach B2.
The load-bearing word is guided. Those are supervised hours with structured output, of the kind an intensive programme delivers: enrolments at the US Foreign Service Institute’s School of Language Studies run from eight to forty-four weeks depending on the language, full-time. An hour there and an hour of watching a series with subtitles are both called an hour, and they are not the same unit.
So the honest answer to “how long” is a question back: how many of your hours contain an attempt at producing Spanish you cannot yet produce? For most learners the answer is a small fraction, and that fraction, not the total, is what sets the timeline.
Why the number people quote is never the number they mean
Before the mechanism, it is worth clearing up why searching this produces such a mess.
A course provider quoting 200 hours means 200 hours in their classroom. An intensive programme quoting 600 means six hundred hours of full-time supervised instruction plus homework. A forum post quoting 1,000 means every minute the writer spent near the language, including films they half-watched. None of them is lying and none of them is measuring the same thing, and the ratio between the loosest and the strictest definition is easily five to one.
The practical effect is that a learner benchmarks their own loose hours against somebody else’s strict hours, concludes they are slow, and adds more loose hours. That is the wrong correction, and it is the one the wide range invites.
Why this step costs more than the last one
The early stages reward any effort. New vocabulary sticks because it connects to nothing and every word is a first encounter. Grammar is finite and testable. Progress is weekly, because the distance between nothing and something is short.
B1 to B2 differs in kind. Most of the vocabulary you need you have already met, so the work is making it producible rather than recognisable. Grammar stops being new rules and becomes the full range of rules you hold, under pressure, without simplifying. And the errors that matter turn subtle: wrong register, flat phrasing, structures quietly avoided. Not wrong, just limited.
The consequence is that progress becomes invisible to the instruments around you. Your multiple-choice score does not move. Your streak does not reflect it. You feel stuck because nothing external confirms movement, and that is a measurement failure rather than a learning failure.
This is why the timeline question is emotionally loaded in a way that “how long from A1 to A2” never is. At the earlier step you can see yourself moving, so the number is trivia. At this one you cannot, so the number becomes the only evidence available that anything is happening at all, and a bad number does real damage. Someone told six months who takes two years concludes they are the problem, and someone told two years who could have done it in one simply does it in two.
What the hour estimates are actually counting
The published figures assume guided instruction with structured output practice. Not hours of series, not app exercises, not reading.
Those things build comprehension, which was never the constraint at B1. Batia Laufer’s longitudinal work on passive and active vocabulary is the uncomfortable evidence here: across a year of instruction the ratio of active to passive vocabulary fell from 89% to 73%. The learners improved throughout, and their comprehension outran their production. More input aimed at a system that already recognises far more than it can say widens the very gap you are trying to close.
The learners who cross faster are not usually working more hours. They are spending a higher share of the same hours in the bottom row.
What targeted production means specifically
Not “speaking practice”, which is where this advice normally stops and where it stops being useful. Four things, and the fourth is the one people skip.
Tasks that force the structure you avoid. If you never use the subjunctive after cuando for future events, you need a task with no indicative exit. The moment the easy route exists, you will take it without noticing, because that is what avoidance is. Elaine Schachter documented the pattern in 1974: learners produced almost no relative clauses rather than producing them wrongly, which made an error count rate them as strong.
Feedback on the dodge, not only on the error. Quiero que vienes is wrong and gets corrected. Creo que no es buena idea is right, and it is the reason you are still at B1. Something has to notice the second one, and almost nothing does, because there is nothing to flag.
Repetition until it is automatic, spread across days. One correct subjunctive is not ownership. Roediger and Karpicke’s test-enhanced learning work found restudying winning on an immediate test and losing badly on a delayed one, and Cepeda and colleagues’ synthesis of 839 assessments of distributed practice across 317 experiments found spacing reliably beating massed repetition. Three short sessions across three days beat one long one, and will feel less productive throughout.
Speed as a target, not a by-product. Robert DeKeyser’s skill acquisition account has knowledge move from declarative through procedural to automatic, and Norman Segalowitz’s cognitive fluency is the variable that decides whether a form arrives inside a conversational turn. A structure you can produce in four seconds is one you will route around, so it does not yet count as crossed.
Our own arithmetic, offered as an observation
Here is a rough model rather than a finding, and it is worth labelling as such.
If you take the Cambridge step at 150 to 200 guided hours, and if a typical self-directed week contains perhaps a fifth of its hours in the bottom row of that table, then ten hours a week yields around two effective hours. At that density the step takes a long time, which matches what most people report.
The useful implication does not depend on the exact fraction. Doubling the density halves the timeline, and you do that by removing the practice that only maintains what you already have rather than by adding hours to a week that does not have them.
It is also worth being clear about what this model is not. We have not measured the fifth, and anyone who gives you a precise conversion between app hours and guided hours is making it up. What we can say is that the two are different units, that published estimates count only one of them, and that the gap between the two explains almost all of the disagreement you will find when you search this question. A learner comparing their thousand app hours against a published four hundred and concluding they must be unusually slow is comparing the wrong things.
The strongest case for just putting in the hours
The volume argument deserves its best version, because it is not wrong.
Nobody has crossed B1 to B2 without a great deal of contact with the language, and the people who get there are, overwhelmingly, the people who stuck at it. Hours are also the only variable most learners actually control: you can decide to do ten a week, and you cannot easily decide to have better feedback. Advice that dismisses volume in favour of a quality you cannot buy is not advice.
There is a fair argument against over-engineering, too. A learner who spends their study time optimising their study time is a familiar and unproductive figure, and plenty of people have reached B2 on nothing but stubbornness, conversation and television.
Why composition still beats volume
The problem is what unstructured volume selects for.
Your conversation partner understands you and does not push back on a correct sentence that could have been more complex, so the range you bring is the range you rehearse. Paul Nation’s four strands put meaning-focused input, meaning-focused output, language-focused study and fluency development at roughly a quarter each, and the typical plateaued learner is running one and a half of those four. Adding hours to an unbalanced diet does not balance it.
Laufer and Goldstein’s four degrees of word knowledge, validated across 435 learners, is the same point at the level of a single item: recognition and active recall are separate rungs, and hours that only exercise the bottom rung produce a larger bottom rung.
Where the measurement comes in
You cannot raise density without visibility. Which structures am I still avoiding, where does production fall short of comprehension, what did I nearly say before simplifying it. That is diagnostic work, and it is exactly the signal ordinary practice does not generate.
If you are aiming at a formal milestone, DELE B2 speaking exposes precisely this gap, because it does not test what you know. Preparing properly for it is the work described above, which is why the exam is a reasonable forcing function even if you never sit it. Its grouped mark scheme is the point: comprehension cannot carry production, so the exam refuses the trade your daily life keeps offering you.
What we do with this at Suelto
The unit we track is not hours, which is the whole argument of this note.
Territory shows which structures are fog, partial or clear, so the week can be aimed rather than filled. The conversation step records avoidance as a signal rather than scoring the substitute as correct, because the dodge is where the remaining distance lives. Latency sits alongside accuracy everywhere, since a form you produce slowly has not crossed. And review is spaced rather than massed, following the evidence rather than the feeling of productivity.
The gap test gives you a concrete snapshot of the specific structures holding you at B1, in about ten minutes with no signup, rather than another level label. The method page sets out what the rest rests on.
About 150 to 200 guided hours, of which the ones in your week are probably a small fraction. The way to go faster is not a longer week. It is a week with considerably less maintenance in it.