I was Twarguing recently with my District colleague Shauna Néro. “Grammar feedback and direct instruction in the language classroom are pointless,” I said.
“I don’t believe it,” she said.
This is a pretty classic argument between a C.I. practitioner and a traditionalist skill-builder. Fortunately, science doesn’t care whether or not we “believe” it, and our argument, as we shall see, has been mostly settled for us by VanPatten, Keating and Leeser.
Let’s prep our brains for this. Here we go:
Sentence A: “Chris is brewing beer in Vancouver on Saturday.”
Sentence B: “Chris in Squamish rock-climbing is on Sunday.”
Which took you longer to understand? Which made you momentaraily go “huh?” I bet it was B.
Because, as egg-heads and linguists say, “you have acquired a well-developed mental representation of English,” sentence B takes you a half-second more to read than A, and it takes you a couple of microseconds longer to answer Question B than Question A. You probably noticed your brain going “no, wait, sentence B should be __________” i.e. “re-thinking” the sentence properly. This is what the brain that has fully acquired English does.
Now here’s another example. In Hindi, khanna means “food” and he means “there is.”
Sentence C: Khanna he?
Sentence D: He khanna?
You probably figured out that both questions mean “is there food?” Did one take you longer to read than the other, or to figure out? Probably not. Now, here’s the thing. Sentence D is wrong. If you are a native speaker of Hindi, you read D and briefly went “huh?” You understood it both ways, but your processing speed differed depending on whether or not your Hindi was later-learned or acquired as your/a first language.
This, it turns out, is standard experience with L1. If the vocabulary and grammar have been acquired, and the sentence makes sense, grammatical errors will make us measurably slow down in our reading and question-answering. In other words, if we really have the language acquired– wired in– we will slow down slightly when processing errors in the language. If we have acquired it less thoroughly, we will get the meaning but our processing will not slow when there are word order or grammatical issues.
This is why you can read Noam Chomsky’s famous sentence “colorless green ideas sleep furiously” and it won’t intuitively “feel” or “sound” weird, despite obviously being nonsensical (like Lewis Carrol’s “’twas brillig and the slithy toves did gyre and gimble in the wabe”).
It turns out that measuring processing speed is a pretty good way to figure out how well someone has acquired mental representation of a language. If we process errors slowly, it shows that we have the language “wired in,” and if we process them quickly (or don’t process them at all), we have not yet acquired mental representation of the part of the language being looked at.
Bill VanPatten and crew asked two interesting questions in an experiment detailed Vanpatten Keating & Leeser 2012:
a) Do native (L1) speakers of Spanish “slow down” processing different types of errors?
b) Do non-native (L2) Spanish speakers slow down equally for different types of errors?
They exposed Spanish L1s and L2s to Spanish sentences with three types of errors: wh- question, adverb placement, and verb conjugation errors.
With verb conjugation, where one should say la chica juega fútbol (the girl plays soccer), subjects read errors such as *la chica juego fútbol (the girl I play soccer).
With adverb placement, where one should say Juan no viaja más a Francia porque no tiene dinero (Juan no longer goes to France because he has no money), subjects read *Juan no más viaja a Francia porque no tiene dinero. Note here that the errors follow English word order.
The question-word errors involved sentences like ¿dónde comen tus padres cuando hacen visita a Chicago? (Where do your parents eat when they visit Chicago?) and *¿dónde tus padres comen cuando hacen visita a Chicago? Again, the error sentences follow English word order.
Note that both correct and incorrect sentences where possible use the same number of words and the same words.
VanPatten, Keating and Leeser showed the subjects sentences (some with errors, some normal) on the computer screen, followed by a question. The computer timed from when they saw the sentence to when they entered an answer, and then showed them the next question. The focus here was not getting the right answer (it was basically impossible to get a wrong answer) but rather on seeing how long it took people to process grammatical and ungrammatical sentences.
It had been hypothesised that the L1s would process Spanish errors of any type at the same speed, because their mental representation of Spanish is so solid that any error would stick out. This turned out to be the case. L1s– native Spanish speakers– processed conjugation, question word and adverb placement errors at the same speed. For the L2s, things were quite different.
Verb conjugation had been (repeatedly) addressed in the college students’ Spanish classes. There had been explicit instruction and explanations, practice, listening, reading etc, all of which used verb conjugation as focus– both direct and implied (implicit)– of instruction. Using adverbial phrases– e.g. saying “no longer” (no…más)– as well as question formation had been regularly present in reading & listening, but had not been explicitly dealt with.
What VanPatten, Keating and Leeser found was that there were significant differences in L2s’ processing speed depending on the error type in the sentence. For the verb conjugation error sentences, the L2 group– despite having been taught, and having read and ‘practiced’ it– did not slow down in their processing. With the adverb placement and question-word error sentences, the L2 group– despite not having been ‘taught’ or having ‘practiced’ them— slowed down processing markedly. If they slowed down processing, that meant they had developed a mental representation thereof, and if they did not slow down processing it, they had not (yet) acquired mental representation of the item in question.
In other words, the L2s had acquired a solid mental representation of adverb phrase placement and question word usage despite not having been explicitly taught them, and they had failed to develop a solid mental representation ofverb conjugation despite having been taught it and having practiced it.
You probably want to re-read that last paragraph. Let’s rephrase: people acquired things from the Spanish input which they weren’t explicitly taught and had not practiced, and they did not acquire other things from the Spanish input which they were taught and had practiced. As VanPatten wrote me– echoing Krashen of 35 years prior– “the research on effects of instruction consistently fails to show any effect for [on] the implicit system,” and this study shows exactly that.
Let’s rephrase again: you can teach– and “practice”– all the ______ you want, but people are not going to pick up on ____ if their brains are not ready for it.
The implications are staggering, and here they are in no particular order.
First, textbooks are at best irrelevant and at worst an active and expensive impediment to learning. Sequenced grammar, as I have argued, not only makes language acquisition boring, but is ineffective. Why?
- We do not learn grammar as “units” or “skill sets.” There is no mental “unit” called “the -ar verb” or “indirect object pronouns.”
- students do not acquire grammar on a teacher or book-dictated schedule
- As Vanpatten, Long, Krashen etc note, we do not acquire what textbooks present as “rules.” What we call “rules” are surface descriptions of very complex subconscious processes.
- as Susan Gross has pointed out, people may well be ready for input that the text does not present, and not ready for what it is showing us.
Second, this study suggests that teaching with fully unsheltered grammar from Day 1 is the way to go. In other words, all grammatical structures– verb tenses, pronouns, moods, what have you, no matter how “complex” or “advanced” or “un-L1-like” they may appear from the traditionalist’s perspective– should be present in input as soon as learners begin acquiring the language. As Susan Gross has argued, and as Long (1997) says, “[t]he idea that what you teach is what they learn, and when you teach it is when they learn it, is not just simplistic, it is wrong.” In addition, if there as an un-L1-like feature in the L2 (e.g. the Spanish subjunctive for English L1 speakers), it may require more processing practice– input– than supposedly “L1-like” features…and should therefore be present sooner.
This being the case, using sheltered grammar– what all textbooks do when they leave “advanced stuff” like the subjunctive for later– is wrong, because it
- denies the brain at least some of the input it needs
- overemphasises at least some what the brain is not ready for
Eric Herman points out that this is a logical, but not an empirical, argument for unsheltered grammar, noting that we would need to study under what conditions– unsheltered or sheltered input– better mental representation appears. My gut feeling (and classroom experience) is that unsheltered instruction will develop better mental representation, but more slowly than sheltered, as there is a greater variety of syntax etc to soak up.
Third, this puts a massive hole into any kind of proficiency level system which is tied to specific grammar or vocabulary. Most assessment systems in traditional textbooks– the typical end-of-chapter reading, listening, writing etc tests– are invalid. Sure, kids can consciously wrestle with verb conjugations or pronouns or whatever by “studying hard” the night before the test, and on test day they say “hurry let’s do the exam before we forget,” and they are right– they immediately forget whatever they’ve memorised. However, vocab-and-grammar-specific tests often
- do not assess proficiency of what kids actually know
- assess proficiency of things kids may not be ready for
If your State, District, school etc says, for example, “French Level 2s must know how to conjugate verbs using être in the passé composé,” they are flat-out wrong-headed. Kids should be exposed to verbs using être in the passé composé (if the input is comprehensible and interesting), and it is certainly appropriate to expect them to understand, but forcing them to use these (especially in idiotic discrete-item grammar tests) is wrong.
It is also wrong to ask students to output specific features of language in a test situation. While the brain can recognise the meaning of things it has not acquired, asking it to output these things properly is asking the impossible. So a writing or speaking exercise which says “you must use the verb ___, the following nouns, etc, in your dialogue, paragraph etc,” is bad practice. This is when you get kids writing “yo tener divertido” or “je suis quinze ans.”
Fourth, the most obvious lesson here is that teachers are wasting their time with grammar lessons, explanations and practice. The brain is going to pick up what it wants only when it wants. While obviously teachers should answer grammar questions, and explicitly deal with grammar to support meaning (e.g. by saying “in French, the –ons on aimer means ‘we,’ and goes with nous“), anything else is a waste of time. Ditch those stupid fill-in-the-blank workbooks.
Fifth, if grammar lessons– and teaching that organises curriculum around grammar– do not deliver the wanted results, the question of “what should we be teaching?” arises. The answer, I think, is threefold:
- multidimensional language (a.k.a. unsheltered grammar) with “everything” in the mix
- compelling comprehensible input which will keep learners focused (and therefore acquiring)
- vocabulary beginning with more frequently-used and moving toward lower frequency
I’d urge people to look at Wiktionary’s frequency lists and note that, in terms of the most commonly-used words, all grammar is present. It’s not like the subjunctive is seldom (or later)- used in Spanish or French. Indeed, the typical order of presentation of grammar in textbooks– numbers=> present tense=> past tense=> subjunctive– is backwards to how often these features are used in languages.
Sixth, activities which focus students on input of language– reading and listening– will be more effective than those involving grammar manipulation, output, etc. If a teacher notes that students have not acquired ____, reading/listening to a ton of input that contains ____ would be the best bet…but, as VanPatten et al show, it’s no guarantee they’ll pick it up when you want them to. We do know that if they understand, they are acquiring…just not necesarily on the teacher’s schedule.
Seventh, there is news for “communicative” teachers, who like having students use the target language to exchange information: it will be almost impossible to have students generate accurate language either to make questions or to respond to them. Why? Because– by definition– what a ‘communicative’ teacher is asking a student to do is the thing they do not know how to do. If the experimental subjects had not developed mental representation of verb conjugation after competent native speaker and textual exposure to it, how can we expect students to acquire (say) verb conjugation by both doing it and getting poor input from other learners? As Bill VanPatten puts it on his podcast, “asking students to produce what they are trying to acquire is, in a sense, putting the cart before the horse.”
OK, there we go. People’s brains, not their teachers, set the acquisitional agenda. Let’s respect that by providing good comprehensible input, not insisting that students master the passé composé by November 27th at 2:43 PM, and not harping about grammar.