Somewhere in the last few years the question stopped being academic. A whole generation of Spanish learners now starts with comprehensible input, watches hundreds of hours of video, tracks the hours in a spreadsheet, and is told, explicitly, not to study grammar. Not “study less grammar.” Not “study it later.” Do not study it.
The case for that is stronger than most grammar defenders admit. Anyone who has met a learner with six years of conjugation tables and no ability to order lunch knows the failure mode being reacted against is real.
But there is an unusual amount of evidence on this specific question, including one experiment that ran for decades across an entire country. This note works through it: what “without grammar” actually means, the strongest version of the case for it, what the immersion evidence shows, which Spanish forms never arrive on their own and why, and where that leaves you.
The short answer
Can you learn Spanish without grammar? You can get further than most teachers admit, and you will stall in a place you can predict in advance. Input alone builds vocabulary, word order and comprehension reliably. What it does not reliably build is accuracy in the forms whose meaning is already carried somewhere else in the sentence.
In Spanish those forms are a short, specific list: gender agreement, preterite against imperfect, the subjunctive, ser against estar, and the clitic pronouns lo, le and se. They share one property. Getting them wrong almost never stops you being understood, so nothing in the input ever punishes the error, and it hardens into a permanent feature of your Spanish.
The useful question is not grammar or no grammar. It is how much, when, and in what form. The evidence-backed answer is: not much, at the moment you have a production problem, and always followed by production rather than more explanation.
What “without grammar” actually means
Three different claims travel under the same banner, and they are worth separating because they are not equally strong.
The theoretical claim is Stephen Krashen’s. In Principles and Practice in Second Language Acquisition he separates acquisition, the unconscious process that produces fluency, from learning, the conscious study of rules. Learned rules never become acquired competence. They can only act as a Monitor: an editor that checks output after the acquired system has produced it.
The strong practical claim is ALG, Automatic Language Growth, developed by J. Marvin Brown at the AUA language centre in Bangkok. Students listened for up to a year without speaking. Brown’s position went further than Krashen’s: conscious study does not merely fail to help, it actively damages your ceiling.
The mainstream practical claim is what most learners actually meet today. Dreaming Spanish, the largest Spanish comprehensible-input platform, states it plainly: “We acquire language by understanding messages in context, not by studying grammar.” No drills, no memorisation, progress measured in hours of understood input. Steve Kaufmann and the Refold community hold a softer version again: read a grammar guide once as a reference, then stop.
These are different bets. The first is a theory of mind, the second is a claim about damage, the third is a claim about efficiency. Most of the argument online conflates them.
The strongest case for skipping grammar
Take Krashen’s argument seriously, because on its own terms it is good.
The Monitor only runs when three conditions hold at once: you have enough time, you are focused on form rather than meaning, and you know the relevant rule. Krashen’s point is that ordinary conversation supplies none of the three. Someone asks you a question, you have about half a second, your attention is on what you want to say, and the rule about the subjunctive after negated belief verbs is not available at that speed even if you can recite it perfectly at a desk.
That observation is correct and it is the single most useful thing the input movement has put back into mainstream practice. A rule you can state is not a rule you can use. Anyone who has spent a year on conjugation tables and still cannot conjugate in conversation has lived the proof.
The mistake is the step that comes next: because conscious rules cannot run in real time, therefore they contribute nothing. That step does not follow, and there is a large experiment showing it does not.
The experiment nobody designed as one
Canadian French immersion has been running since the 1960s. Children start at five. Subject matter is taught in French. Over a school career they accumulate thousands of hours of comprehensible input in an environment far richer than any app can build.
If input alone were sufficient, this is where it would show.
Merrill Swain and colleagues measured what actually happened. Reception was the success story: listening and reading comprehension approached native levels. Production did not follow. After six or more years, students were still producing systematic errors that native-speaking children of the same age never made, clustered in verb morphology, grammatical gender and the formal versus informal distinction. Their French was fluent, confident, and audibly non-native in ways that had stopped changing.
This is why Swain proposed the Output Hypothesis. Her argument was not that input is unimportant. It was that comprehension lets you succeed semantically without ever processing the grammar, because you can get the meaning from context, word order and content words alone. Production is what forces you to notice the parts you skipped.
The immersion case matters because it is the ceiling. It is more input, better input, and earlier input than any adult learner will ever assemble. If input alone were going to close production, it would have closed it there.
Why some Spanish forms never arrive on their own
Not every part of Spanish behaves this way, and the split is predictable.
Forms arrive from input when noticing them is unavoidable. You cannot misunderstand no or basic word order for long, because getting them wrong breaks the sentence and the breakage comes back at you immediately.
Forms stall when three things line up: the form is hard to hear, its meaning is already carried elsewhere in the sentence, and getting it wrong costs you nothing socially. Richard Schmidt’s noticing hypothesis is the mechanism: input does not become intake until it is consciously registered. Recent work goes further and finds that redundancy actively hinders adult grammar learning, because a cue that adds no information is a cue the learner has no reason to process.
Spanish has an unusually large number of forms that fit that profile.
The subjunctive is the clearest case, and it has been measured. Fernández Cuenca and Jegerski’s eye-tracking study (2022) notes that subjunctive forms run at roughly 10% of finite verbs against 90% indicative, and that the morphology “is usually redundant and thus not critical to meaningful communication,” because the trigger verb in the main clause has already told you the speaker’s attitude. On regular verbs the entire distinction is one vowel, in the middle of a sentence, unstressed.
So: a form you rarely hear, that you cannot easily hear, that you do not need in order to understand the sentence, and that nobody will correct you for omitting. There is no mechanism by which raw exposure fixes that. It is not a motivation problem, and it is not a question of more hours. We put actual numbers on how often it comes up in how much subjunctive you really need, and on what dodging it looks like in subjunctive avoidance.
The clitics are the same story from another angle. In our audit of the 1,000 most common Spanish words we pulled the RAE frequency list apart and found se sitting at rank nine, lo at twenty-one and le at twenty-seven, with all ten of the top ten being grammatical words. The most frequent items in the language are the ones a no-grammar approach has no way to teach and a word-list approach has no way to define.
What happens when you do teach it
The counter-evidence is not a matter of opinion. It has been meta-analysed twice.
Norris and Ortega’s synthesis of instruction studies (2000) pooled 49 samples and found that focused instruction produces large and durable gains, with explicit approaches outperforming implicit ones.
Spada and Tomita’s follow-up meta-analysis (2010) tested the obvious objection: fine for easy forms, but surely complex grammar has to be acquired rather than taught. Across 41 studies they found larger effects for explicit instruction over implicit for both simple and complex features, and, importantly, that the benefit showed up in spontaneous use, not only in controlled tests. The prediction that complex forms resist teaching did not hold.
Then there is Bill VanPatten’s input processing work, which shows the failure happening in real time. His First Noun Principle says learners assign the agent role to the first noun or pronoun they meet. Give an intermediate Lo saluda María and they reliably read it as “he greets María” rather than “María greets him.” Every word in that sentence is high-frequency and known. More input of the same kind does not fix it, because the learner comprehends it, incorrectly, and moves on satisfied.
That is the whole problem in one sentence. Input teaches you what you notice. If your parser is quietly wrong, additional input confirms the wrong parse.
The other way to fail
None of which rescues the traditional approach, and it would be dishonest to pretend otherwise.
The learner who has done grammar and nothing else is the reason the input movement exists. They can name six subjunctive triggers, complete a preterite table under exam conditions, explain por against para better than most natives, and go silent when the person behind the counter asks a follow-up question. Their knowledge is real and it is in the wrong format: declarative, slow, and never converted into anything retrievable at conversational speed.
Both ends fail, and they fail differently. Input-only produces fluent, confident, permanently inaccurate Spanish. Grammar-only produces accurate, careful, unusable Spanish. Neither camp is wrong about the other; they are both right, which is a good sign that the binary is the problem.
So how much grammar do you actually need?
Less than a textbook, more than zero, and in a different shape than either camp usually delivers.
The distinction that does the work is Michael Long’s, between focus on form and focus on forms. Focus on forms is the traditional syllabus: grammar points marched through in order, in isolation, because chapter seven comes after chapter six. Focus on form is brief attention to a structure inside meaningful communication, triggered when the structure is what stands between you and the thing you were trying to say. Same content, opposite delivery, and only one of them has the evidence.
Nation’s four strands put a number on it: a balanced course is roughly a quarter meaning-focused input, a quarter meaning-focused output, a quarter deliberate language-focused study, and a quarter fluency development. Deliberate grammar work is one of four, not zero of four and not four of four.
In practice that means three things. Grammar in minutes, not hours. Triggered by a production problem you actually have, not by a chapter number. And followed immediately by producing the form under mild time pressure, because the explanation is not the learning event.
What we do with this at Suelto
Suelto is not an input app and it is not a grammar course, and the reason is everything above.
Territory is the diagnostic layer. It maps your Spanish on two axes, Grammar (can you produce the structures?) and Function (can you do things with the language?), and every category sits in one of three states: fog, partial, or clear. Fog means we have never seen you attempt it. That is the whole point: the forms that fossilise under input alone are precisely the ones you never attempt and never get corrected on, so they are invisible to any system that only scores what you did. Territory is built to show absence, not just error.
The conversation step tracks avoidance for the same reason. When you write creo que no es buena idea instead of no creo que sea buena idea, nothing is wrong. It is a correct Spanish sentence. But if you take that exit three sessions running, that is a fossilising subjunctive, caught before it hardens. Grammar lessons get scheduled off that signal rather than off a syllabus position.
Lessons themselves are short and immediately drilled, and verbs get a dedicated timed step, because a form you can only produce in four seconds is a form you will avoid in conversation.
And input is in there too. Daily reading, listening, vocal shadowing and a Discover feed of real Spanish media, because the input people are right about volume and we are not interested in relitigating that. The method page lists the research each part rests on. If you want to see where your own fog sits, the gap test takes about ten minutes and needs no signup.
The honest summary is not glamorous. You cannot learn Spanish without grammar, you should not learn Spanish as grammar, and the gap between those two sentences is where almost all the useful work happens.