I’d like you to picture a young student sitting in a classroom. The subject is Introductory Spanish. On the blackboard is written the name of the teacher and the agenda for today’s class: “Pop quiz”.
Our teacher stands at the head of the class and calls on each student in turn.
“Michael, answer question 14.”
After considering how many desks stand between Michael and herself, with some modest arithmetic our student calculates that she will be answering question 21. Hastily, she pages through the textbook in the hope of coming up with something that won’t embarrass her. She lights upon an answer in the nick of time, and the anxiety subsides.
Four years of pop quizzes, cramming for tests, and staring at verb charts will pass by, and our student will have completed high school Spanish. Unfortunately, our student’s Spanish ability will extend no further than being able to order a beer.
As useful as “Una cerveza, por favor” may be, we should expect more from four years of (often painful) effort towards learning Spanish. But failure is all too often the norm.
And if this is true of modern languages, it’s doubly true of ancient languages.
What is “grammar-translation”?
If you’ve studied Latin in most formal settings, you’ve likely been exposed to a method called “grammar-translation” which works something like this:
You’re shown a variety of grammatical rules, including what seems like endless charts listing word forms appropriate to be used in various situations (e.g. -us, -a, -um), and told to memorize them.
You are then given practice applying these forms by translating sentences to and from the language you’re learning.
Four years of that kind of Latin often results in the ability to describe how many parts Gaul is divided into, but little else.
Although the grammar-translation method may seem old-fashioned, it is in fact a relatively modern approach: it has its origins in the way Latin began to be taught in the 19th century, and spread from there to be used for the modern languages as well. Interestingly, the grammar-translation method displaced methods that were much more like the ones backed up by current second-language acquisition research. We’ll see what those look like soon.
For now, though, what’s clear is that the grammar-translation method and its descendants don’t work. But why?
Grammar-translation and related modern language teaching methods don’t work because they’re predicated on a mistake about what is going on when you learn a language. Research into linguistics and language acquisition have now caught up with the older forms of language teaching that predated the 19th century grammar-translation method. There was never a compelling reason to teach that way, but especially now that the linguistic and pedagogical theory reflects the experience of past centuries of successful language teaching – rudely interrupted by the Prussians and Victorians – we do not have to continue their mistake.
So what is the mistake? The grammar-translation method assumes that a language is something that can be described using rules that reflect the language’s surface-level behaviour.
These are the familiar rules we learn in traditional language classes: there’s a rule for how adjectives agree with nouns, there’s a rule for how the past tense is formed, a rule for where the verb goes in the sentence, etc. Once you’ve learned all those rules, and could apply them on tests using the vocabulary you’d memorized along the way, you could say that you knew the language.
But this isn’t true: knowing a language is something other than knowing all of its grammatical rules.
Against grammatical rules
Why is it that learning all the rules of a language does not mean you’ve learned the language?
This is an especially mysterious thing: after all, grammatical rules aren’t false. To use Latin as an example, rules such as “Use -īs for the ablative plural of 1st and 2nd declension nouns” or “cum takes the indicative when it means ‘when’” are accurate generalizations about the Latin language.
Given that the rules accurately describe the language, you could be forgiven for thinking that a person who has learned all of a language’s grammatical rules will have learned that language. But, as we’ve seen anecdotally, it doesn’t seem to be the case.
After all, what does it mean to successfully “learn” a language? If you say you have actually learned Spanish, it means something like: “I have in my mind a certain kind of knowledge about how Spanish works and this knowledge allows me to understand native speakers of Spanish and to be understood by them in return.” There is obviously a broad spectrum here, since at a beginner level you may have enough knowledge of how Spanish works to blunder your way through a conversation with a native speaker. But the closer your knowledge gets to what a native speaker has, the more ease you’ll have using the language, and the more true it is to say that you “have learned Spanish.”
If learning a language means coming to know what a native speaker of that language knows (even if you can’t necessarily use all that knowledge with the same degree of speed or accuracy as a native speaker), we need to ask what a native speaker knows. And, as it turns out, the knowledge of language possessed by a native speaker of a language is a different kind of knowledge than the knowledge that learning grammatical rules gives you. In other words, the problem with our imagined Spanish class was not that it didn’t teach you enough; it taught you the wrong stuff.
We can see this illustrated by examining our knowledge of our native language. I’ll use English here because that’s my native language: if it’s not yours, see if you can come up with equivalent examples in your native language.
Before you blink or take another breath, give the 3rd person singular pluperfect passive indicative form of to eat. Can you manage it without fainting? Don’t try.
Now for another, different experiment: Imagine coming home from college. Your mom has made your favorite dinner for you, but traffic took so long that your younger brother ate it all for a midnight snack. Poor you. The next day, you are telling your girlfriend about this poor welcome home. You say (fill in the blank): “I got home and the entire roast ______!” Here, it’s not in any way hard to come up with the form had been eaten.
The curious thing is this: had been eaten is the 3rd person singular pluperfect passive indicative form of to eat. For some reason, however, it’s not easy for native speakers to retrieve that form using its grammatical description.
In other words, you have implicit knowledge – the knowledge which shows up when you produce sentences and allows you to come up with forms like had been eaten when the situation demands it – that outstrips your explicit knowledge of the exact same facts – namely, that to eat has the 3nd person singular pluperfect passive indicative form of had been eaten.
You may object: perhaps an English speaker untrained in grammar would have difficulty, but someone with a proper linguistic education would surely succeed. To this, I present myself as a counterexample. My professional education is as a linguist, and specifically as a morphologist: someone who studies how words change to express meaning.
And I don’t perform any better at this than does the average person: to figure out the 3rd person singular pluperfect passive indicative form of to eat, I had to go through a whole series of steps in my mind: I had to retrieve the fact that pluperfect corresponds to the had VERB-ed construction, I had to work out the fact that the 3rd person singular form of have is had, I had to put the whole thing into the passive, which (I had to remember) is expressed by the be VERB-ed construction in English, and finally I had to remember that I didn’t have to worry about indicative vs subjunctive in a past tense form like the pluperfect.
The case of had been eaten shows us that the explicit knowledge of language that grammatical rules gives us and the implicit knowledge of language that native speakers possess are distinct things.
Although explicit and implicit knowledge may relate to the same facts about the language, for example that to eat has a particular form used in sentences such as I got home and the entire roast ______!, only implicit knowledge makes you able to do the things a native speaker does: use language in real time.
What the grammar-translation method and its various descendants teach is almost exclusively explicit knowledge. Is it any wonder, then, that few people learn languages in these classes?
The complexity of linguistic knowledge
The linguistic knowledge of native speakers is not only implicit, but also far more complex and abstract than the explicit knowledge encoded in grammatical rules.
Here’s an example of the kind of complexity and abstractness involved in our implicit knowledge of language. Consider the following pair of sentences, of which the first is a statement and the second a question which may have given rise to that statement.
Sam has been reading the Aeneid.
What has Sam been reading?
Note that the question word is way out at the start of the sentence, and its answer is all the way at the end of the sentence. You can also make questions like Sam has been reading WHAT?!, where the question word is in the same place as its answer in the corresponding statement. For this reason, it’s common in linguistics to consider the “real home” of the English question word what to be in the same spot as the corresponding answer word.
Let’s look at our question-answer pair again, this time keeping track of the “real home” of what by including a little “ghost” in that position – you’ll understand why a little bit later.
Sam has been reading the Aeneid.
What has Sam been reading 👻?
Both of these are fine sentences of English, and we could come up with a grammatical rule that describes the relationship between the two. Indeed, when English is taught to speakers of other languages, they learn just such a rule, which looks something like this:
To make a statement into a question, replace the thing you are questioning with a wh-word, move that word to the front of the sentence, move the main verb before the subject (inserting a form of do if the main verb is not a modal or auxiliary), and voilà: you have a question. In other words, replace the Aeneid with the question word what, then put what as the first word of the sentence, and finally move has before Sam.
Sam has been reading the Aeneid → What has Sam been reading?
This is a lovely rule, but it has the major disadvantage of being incorrect. Consider the following pair:
Sam left because he wanted to read the Aeneid.
*What did Sam leave because he wanted to read 👻?
Here, when we follow the rule, we’re left with an ungrammatical sentence of English (these non-sentences are marked by linguists with an asterisk before them, like so: *). And this is not an isolated exception: there are many sentences that work this way. Here are some more examples:
Sam met someone who translated the Aeneid
*What did Sam meet someone who translated 👻?
Sam likes Helen’s translation of the Aeneid.
*Whose does Sam like 👻 translation of the Aeneid?
Sam likes the Aeneid and the Odyssey.
*What does Sam like 👻 and the Odyssey?
Each of these sentences presents a counterexample to the (already rather complicated) question-formation rule that a learner would be presented with.
We could, of course, amend the rule – but how? What do these exceptions have in common? Is it that they’re longer sentences? That can’t be it, because the following pair works without any issue:
Sam thinks that I said that Helen translated the Aeneid.
What does Sam think that I said that Helen translated 👻?
It turns out that these restrictions on question formation in English depend on the underlying structures of the sentences, structures which native speakers have no conscious knowledge of.
But the way these speakers use English betrays a great deal of unconscious knowledge of these structures – if English speakers did not have such knowledge, they would not find the ungrammatical sentences above so unacceptable. If you are a native English speaker, you may have even chuckled reading these strange, ungrammatical sentences. Do you know why they are so funny and wrong? Probably not – you just do.
Perhaps even more shocking is the fact that all native speakers of English converge on more or less the same abstract, implicit rules, despite growing up in very different circumstances. And crucially, they do so without having been taught: no one ever told an English-speaking child that *What does Sam like 👻 and the Odyssey? is not a valid English sentence. No one ever needed to.
Cases like these are everywhere in English, and in every language. Consider also the case of adjectives in English: if you want to describe a house which is old, big, red, there’s only one way to do it naturally: a big old red house. *A red big old house doesn’t quite have the same ring to it, does it?
Native English speakers were never taught this, but they all obey the rule without expending seemingly any mental energy; learners of English as a second language are often taught this, but for them it’s often a struggle to achieve the same results as the native speakers.
What’s made the native speakers act like native speakers is a miraculous process that occurs in childhood. The implicit knowledge of language – in all its complexity and abstraction – grows within children’s minds of its own accord, needing nothing but access to the necessary fuel, which is provided by the language the child is exposed to in the context of communication.
This fuel is called the input to the language acquisition process and it is the crucial ingredient in children’s acquisition of their first language.
First- and second-language acquisition
But what about those of us who are no longer children? It doesn’t seem that we can achieve as adults what we once achieved as children, at least not with the same ease.
It is a fact that, in general, the ultimate attainment of adult language learners falls short of – or differs in some way from – what is achieved by child language learners. Most second-language learners speak their second language with an accent, for example. The differences between first- and even advanced second-language speakers of a language extend also to subtle aspects of the grammar, as has been found in many studies over the years.
This seems to paint a bleak picture for second-language learners. What became of the prodigious powers we enjoyed as infants? Are they forever lost to us as adults?
Not necessarily. It is true that the course of second-language acquisition is different than the course of first-language acquisition: second-language acquisition takes longer, seems more effortful, and doesn’t tend to reach the same level as first-language acquisition.
But it is also true that the circumstances of second-language acquisition are usually very different from those of first-language acquisition: second-language acquisition has less urgency, has less time devoted to it, and usually occurs alongside learning to read in the second language, which usually happens later on in first-language acquisition.
For this reason, there is controversy in the research about whether native-likeness is actually impossible, or rather just unlikely, given the circumstances of second-language acquisition.
Although it does seem clear that there are almost always some differences in the eventual outcome of first- and second-language acquisition, it’s just as clear that second-language speakers can and do attain a level of knowledge that enables them to live, work, read, and even write in their target language.
To do this, these second-language speakers must have very high achievement in their target language. And this, I think, is worth aiming for, even if you may not be able to pass for someone born and raised in a community that speaks your target language. (This is, perhaps, less of a temptation for learners of ancient languages.)
Just as there are some differences between first- and second-language acquisition, there are also some powerful similarities. To explain these, I need to tell you about how language acquisition works as a developmental process.
By “developmental process”, I mean that the growth of the mental representation of the language being acquired goes through predictable stages.
For example, a researcher named Roger Brown found in the 1970s that children acquiring English as a first language come to know the various bits of grammatical structure in English (such as a, the, the plural -s, the past tense -ed, etc.) in a predictable order. Brown and his colleagues found that children acquire irregular past tense forms such as went before the regular past tense suffix -ed.
They acquire the verb be first as a copula in sentences like Dog is good, then as an auxiliary verb, as in The dog is eating. They also acquire contracted forms of be only after the full forms, and in the same order: first the copula The dog‘s good and then the auxiliary The dog‘s eating.
In the years following Brown’s work on this developmental sequence in first-language acquisition, other researchers found that the same sequences apply to second-language acquisition of English as well.
Their results showed also that the order of acquisition in the second-language is not affected by other variables: not by whether it happens in a classroom and, perhaps most surprisingly, not by what the learner’s first language is. Unfortunately, not a lot of work has yet been done to extend these findings to the acquisition of languages other than English.
The existence of similar developmental sequences have been found in the acquisition of language structures like negation. For negation, the sequence for people learning English as a second language looks like this (for children learning English as their first language, the stages are very similar):
- The negator no is placed before a phrase: No he can eat dinner.
- The negator no/not (or the fixed form don’t) is placed within the phrase: He don’t can eat dinner.
- Modal verbs (such as can, should) are negated correctly: He can’t eat dinner.
- The auxiliary do is negated correctly in its various forms: He didn’t eat dinner.
Similar stages have been reported for a variety of other structures in English and in a few other languages.
What is most surprising about these developmental stages is that learners do not seem to be able to skip any. Learners do vary in how long they spend at each stage, but each stage invariably appears.
But this leads us to a question: The intermediate stage of He don’t can eat dinner is not part of the adult, native grammar. It would not have been taught to the learner, so where did it come from? And why does it always seem to show up? It seems that it is the byproduct of the stages of the development of the learner’s knowledge of language.
These striking findings suggest two things: (i) there are important similarities between first- and second-language acquisition, and (ii) language acquisition is an organic process and one that proceeds according to its own logic.
The role of input
Now that we’ve surveyed the similarities and differences between first- and second-language acquisition (and found the similarities deeper than the differences), we can move on to what makes the process of acquisition happen. With very few exceptions, every mainstream theory of second language acquisition believes this acquisition occurs through the processing of large quantities of input, that is, exposure to language within the context of communication.
This is in contrast to what many language teachers and learners believe about language acquisition: that it occurs as a result of learning grammatical rules and practising them – in other words, via something like the grammar-translation method or its descendants.
If learning and practising grammatical rules worked, the “Una cerveza, por favor” story I told you at the start of this article – which is a story of missed opportunities in language teaching – would sound outlandish. Unfortunately, my guess is that it is all too familiar.
But that’s an anecdote. We can do better than that. There’s also solid experimental evidence for this position. One strong piece of evidence is that the developmental stages of language acquisition are the same inside the classroom and out.
The researcher Rod Ellis (1989) studied the developmental sequence of the acquisition of German as a second language within a classroom context and compared it to what had previously been found in non-classroom contexts. The results are telling: the developmental sequence is the same, whether the language is learned inside or outside of the classroom. (It’s worth noting that Ellis found that classroom instruction may speed up the process, but the process itself remains the same.)
In other words, non-classroom learners, who aren’t exposed to grammatical rules or explicit practice, nevertheless undergo the same process of development as learners who are exposed to grammatical rules and explicit practice. Given that the same process occurs outside and inside the classroom, formal classroom instruction must not be necessary.
Whatever is fuelling this process must be common to both learning environments. That thing is input: language the learner is exposed to within the context of communication. This last word, communication, is important because it excludes things like isolated example sentences showing the operation of some grammatical rule. Communication means that someone (whether it’s someone you’re speaking with, someone you’re listening to, someone you’re reading…) is trying to get a message to you. It’s in the act of trying to comprehend the message that the magic happens.
This has huge implications for how we learn languages. If you do seek out a class (and though not necessary in the strict sense, it has been shown to help the process), you should look for one that has the potential to provide you with a lot of input. There are many ways teachers can accomplish this, and the specifics depend on things like the language being taught, the size of the class, and the level of the students.
For example, you might get lots of input in an ancient language by reading progressively more challenging texts together with a teacher, who can explain anything unclear by paraphrasing in the target language, as the Ancient Language Institute does. Teachers can also provide lots of input by making up stories to tell the class. In a modern language, a teacher has even more options, such as showing TV shows or movies to students.
Note that, in any case, grammatical rules do not need to be explicitly taught: we now know that doing so is unnecessary for acquisition and, besides, these rules aren’t what learners acquire anyway.
You can apply these principles to self-study as well: get yourself as much input as possible and don’t stress over not being able to come up with things like the dative plural of fifth declension nouns on the spot.
If you’re interested in this kind of thing, I have some recommendations for you. First, go consume everything Bill VanPatten has produced for non-specialists. You can start with his podcast Tea with BVP or this talk (in a series of YouTube videos) where he explains the outlines of second language acquisition theory to a group of language teachers.
If you’d prefer a book, check out VanPatten et al. (2019). Key Questions in Second Language Acquisition, which goes through key questions in the field and presents an overview of what the research had to say about it as of 2019. Everything is written in such a way as to be easy to understand for non-specialists, so it’s a good way into the literature.
Another great source is Stephen Krashen, the originator of lots of the thinking about the role of input in second-language acquisition. He’s made a great deal of his writing available for free on his website. He also gives great interviews, which go a long way towards clarifying his ideas: here is a good example.
And if you’re looking to learn a new language with this method in a classroom setting, check out the Ancient Language Institute for Latin, Ancient Greek, and Biblical Hebrew. Comprehensible input for other ancient languages is relatively rare, although I’ve made some beginner reading resources for Old English. YouTube is also a gold mine for comprehensible input material for many languages, including Russian and Korean. If you’re looking for teachers or resources for other languages, try adding phrases like “natural method”, “direct method”, “comprehensible input” to your searches.
Whether you choose to teach yourself a new language or learn from a teacher, make sure that you absorb lots of comprehensible input. Follow this path and Una cerveza, por favor or Gallia est omnis divisain tres partes will be the start of your abilities, not the end of them.
Colin Gorrie is a linguist on a mission to bring linguistics out of the ivory tower. He is also a serial language learner. In trying to apply what he knew about linguistics to his language learning practice, he came across comprehensible input-based approaches. After seeing how well they worked, he became an evangelist for them too. Colin received his PhD in linguistics from the University of Arizona in 2014, where his studies focused on linguistic typology, the methodological foundations of linguistic theory, and Celtic linguistics. He is also an avid conlanger (maker of constructed languages). You can follow Colin on Twitter, YouTube, or through his website.