I’ve been on Latin reddit discussions for years and one single educational theory comes up again and again as if it were the only way to learn a language: Krashen and his comprehensible input hypothesis.

Put simply, a learner should be introduced to the each feature of the language incrementally, by receiving input that contains their previous level of competence plus the next feature (i + 1).

This is extremely similar to Vygotsky’s Zone of Proximal Development which can be summarised as “for optimal learning, give a student enough challenge but not too much”. And “if a challenge is too great, give them scaffolding (i.e. help) until they can do it with the scaffold, then gradually take the scaffold away.”

There is sound wisdom in advice as basic as “give them just enough challenge” – that it is what we aim to do in teaching every subject, not just languages.

The error comes in believing that if a student successfully translates a sentence which happens to contain the target grammar feature, they necessarily understood the target grammar feature.

I see this all the time when teaching the accusative case to a beginner level class of mixed ability.

A lot of textbooks, when introducing the accusative, just keep on piling on sentence after sentence in the format of Subject-Object-Verb (SOV):

puer puellam videt.

The boy sees the girl.

As teachers we try to verbally explain that the ending of “puella” changed here, and this is the accusative case and you should pay atttention to it. That’s all well and good, but the experiential learning is more powerful for the student, and it often overpowers the verbal, abstract, theoretical explanation. What the student learns experientially from seeing SOV sentence after SOV sentence is that “the first noun I see is the subject, the second noun I see is the object”. By contrast, a single-letter difference on the end of the second word seems of minor importance. “Oh, so the second word gets a spelling change. Cool. Well I already guessed it was the object – correctly! – without noticing the one letter difference. I don’t need that letter.” And the more SOV sentences they see, the more their reliance on assuming the SOV word order is reinforced.

Step back and think how little a change an -m is on the end of puellam. When you are reading, you don’t read every single letter in every word. You glance at each word and it triggers recognition. Ifm there’sm one differentm letter sometimesm you can still recognise what word it’s supposed to be, and ignore the extra letter, just like you ignored the extra -m’s in this sentence. Seeing puella and puellam as variant spellings of the same word does not seem like that big of a deal to a lot of beginner Latin students, if they even registered there was an -m at all.

So now, I’m going to use Lingua Latina Per Se Illustrata (LLPSI) as an example of how ‘comprehensible input’ can reinforce an unintended habit. I chose this because the Krashen crowd likes this book. However, it is far from the only textbook that does this. I could have chosen almost any.

Chapter 3 introduces the accusative case. Therefore, I tallied up the word order in all the 41 sentences containing an accusative case noun in Ch. 3 (but I excluded sentences near the end which contained the relative pronoun, which I would treat as a seperate and more difficult topic).

These are the stats: a student can correctly assume that the first noun is the subject in 36 out of the 41 sentences. That’s 88% for just assuming English-like word order.

In the remaining five sentences, a student can try translating the first noun in each phrase as the subject and not necessarily believe they made a mistake from the context. If the teacher does not immediately correct them, many students will work on the assumption that these answers could be correct:

Quīntus īrātus est et Mārcum pulsat.

*Quintus is angry and Marcus punches. (in the next sentence, Marcus indeed punches Quintus, and Julia then complains that Marcus punches Quintus.)

Iūlia plōrat et Aemiliam vocat: “Mamma!”

*Julia cries and Aemilia calls. “Mamma!” (It makes sense that the mum might call to her daughter to figure out why she’s crying. Then again, it would be odd for the next bit of speech to be Julia’s, but sometimes the word for “she says” is left out. And sometimes students just forget who is meant to be the mum or daughter.)

pater dormit neque tē audit

*Father is sleeping and you don’t hear [him]. (That’s why you didn’t know he was asleep, you just didn’t hear whether he was awake or asleep)

Iūlius Quīntum videt eumque interrogat

*Julius sees Quintus and he asks (the father really does do the asking in the next sentence.)

Mārcum verberat, quia puer improbus est.

*Marcus is hitting [people], because he is a rude boy. (Rude boys do beat up people.)

Students can use an incorrect method to translate sentences that happen to contain the accusative case, and apparently get it right almost all the time, reinforcing their erroneous assumptions. The input is comprehensible and it ‘contains’ that one extra grammar feature (i + 1), but it does so in such a way that it gives the same reward and reinforcement to ignoring the accusative case ending as it does to actually paying attention to the endings.

There are a variety of solutions to this particular problem – for example, a teacher can train students on three word sentences with totally random word order (SOV, OSV, SVO, OVS), showing that they must rely on word endings alone for meaning. I don’t prefer this approach personally, because it gives the impression that Latin word order is random, as if it were concocted to trip up students like a puzzle.

Another approach is to give a “fill in the blanks” exercise to students where they have to put the nominative and accusative endings on the Latin words to show they know they’re supposed to be there. Unless you’re training students to compose in Latin (in which case this is a good scaffolding step), I think from a reader’s perspective, this type of exercise just reinforces to students that they can still read and understand the meaning of the Latin, even with the endings removed, so endings are really unnecessary for comprehension. No wonder Latin lost its endings in the Romance languages, if you can read “Marc__ Iuli__ pulsat” and already guess what it’s supposed to say! Noun endings are clearly superfluous to getting the meaning.

My favoured approach to teaching students to really heed the accusative case is to give them many examples of a sentence type I call “SOV-OV” and work on them in an environment that gives instant feedback:

In this sentence, the format of the question shows that the answer must either be “and the daughter greets” or “and greets the daughter”.

The words have to be carefully chosen. The subject and object must be people (who, unlike inanimate objects, are capable of both acting and receiving action), and the verb must make sense as an action either person could do to the other. The student makes a choice not based on what they could assume from context, but only on the grammatical information in the word they are focusing on, and instantly, the computer tells them if their choice was correct.

Also, the word order in this sentence is something they are likely to encounter in real stories, as the implied subject is carried over without it needing to be restated. This is more natural than getting sentences in totally jumbled word orders, or having to insert an unknown “he/she” at the start of a sentence without a stated subject.

Just so that students don’t learn “the word in this position must be an object”, I sprinkle its opposite version amongst the example questions too. This is the SOV-SV sentence that would accompany the above example:

The instant feedback is key, or else students will believe they were right when they were actually wrong. I’ve been making these as online exercises via the platform Education Perfect, so that as soon as students select their answer, they are given instant feedback (and an explanation for why they were wrong, if they were wrong). You could also convert these into multiple choice questions on a free platform like Quizlet. If I have to teach in a tech-free classroom (a rarity in this pandemic now), I give students an answer key to their practice sentences and instruct them to correct their sentences as soon as they finish each question.

Habits (like assuming the first noun is a subject) are not broken by verbally explaining this to young students and hoping they do it on their own when they are caught up in the flow of reading. Students learn more powerfully by “doing”, and a poorly designed experience can teach them things we don’t intend. When designing comprehensible input for a target grammar feature, the reading activity has to reward only the right reasoning, or it will unintentionally reinforce a lazy and erroneous assumption.