Avoiding translation

This topic came up briefly in another thread but I wanted to make it the main focus here. Do you think that translation is bad or not very useful? Personally, I find translation immensely helpful in several ways: it allows me to compare structures of different languages, can quickly give me an accurate idea of the meaning of a word or check my understanding of it, and can help detect holes in my knowledge of TL vocabulary and locutions (L1 to L2 translation). On the other hand, I know that many people avoid translation as though it is a plague that will contaminate everything about their learning of a language, or they think that using translation means you are forever doomed to have your L1 running through your head, and so on.

What are your thoughts about using translation in language learning?

I'm fully in favour of translation.

The common criticisms of translation are unfair.

Some claim translation teaches errors arising from idiom -- eg people saying "I have 15 years" is commonly stated as being "because of translation", but as far as I'm concerned, it's because they're not being taught to translate properly.

Then there's the claim that translation doesn't teach language, but only teaches you to translate. I don't believe this is always true. Yes, large passage translations tend to be unwieldy and require too much thought, but that's not true of short translations. After all, understanding your native language requires almost no effort whatsoever. If I give you a simple sentence in your native language, you almost don't notice the words, and instead understand the meaning in a purely mental form. I have given you a thought. Now you can use language to express that same thought, only this time, you'll use a different language. Simple.

Furthermore, translation exercises allow a much wider variation in the material practised. Target-language-only exercises, particularly at the early stages of learning, tend to be very repetitive, because there's no way to reliably indicate to the student what you require of them. So you mechanically churn through "the cat is on the table, the box is on the floor, the boy is on the chair" etc, and eventually you aren't even thinking about the meaning of the structure, instead just saying a couple of nouns with a bunch of "glue words".

Producing language is a matter of making a series of choices, and monolingual prompts do not force the learner to make linguistic choices, instead presenting most of the choices half-made and having the learner simply "juggle" the last few pieces into place.

Translation forces the learner to recall the language that they're supposed to be learning, and that's what real language practise is, isn't it?

I like translation for understand something, but I don't like to translate a paragraph or longer.

It depends on how much you lean on translation. Even now sometimes in Cebuano I'll know all the words I just read but still be like, what does it mean? This is especially useful at the beginning when I'm in a big cloud of uncertainty. If I go through it several times with a translation, I will eventually be able to throw away the English for that particular work. As time goes on though, if I kept using translations (other than a word here and there) then I don't think I would ever get good at using context clues in my TL.

I suggested to someone trying to learn Greek to use an English translation to check their understanding, but what ended up happening is that they never figured out the clues for the correct English word order, they just used the translation to tell them.

I'm in favour of translations, but it is a tool that has been used in problematic ways, and there are situations where it should be avoided.

As I wrote in another thread it is idiotic to give explanations in a language which you haven't learnt yet - and under explanations I also count translations in dictionaries. So for beginners the use of monolingual dictionaries and grammars is thoroughly harmful and should be avoided like the plague. And when language related texts in the target language (without a translation) are introduced it would be better to start out with semi-narrative articles (for instance about etymology or specific idioms) rather than tools like grammars and dictionaries, where you expect to get quick and precise answers to specific problems rather than entertaining background information.

Translations back and forth are a staple in language teaching, but I have always been sceptical about 'free' translations. If you want to study a text which really is above your level (in accordance with the principles behind the use of comprehensible texts) then it is practical to have a translation within reach for assistance and confirmation. But a free translation will often be misleading, not least in questions of sentence construction. Let's take an eample: in many language, including the Slavic ones, the genitive case is used after numbers (with some exceptions at the lower end of the scale). So if you think "five tickets" in English you will find it strange and irrelevant to remember to use the genitive in Russian - it seems to be an added complication invented to scare of potential learners of the language. The solution is to think "five of tickets" instead, because this reflects the logic behind the Russian construction. It is true that idiomatic expressions may have meanings that go beyond the sum of the elements, but even then there is mostly some kind of logic behind them - and if you can see that logic then it will make it easier to remember the complete expression.

'Beautiful' and idiomatic translations in your own language are something eager amateurs or paid professionals make to help those lazy bums who can't be bothered to learn a foreign language. Literal and hyperliteral translations are there to help you to learn that foreign language, and once you have learnt it you can (and should) drop them.

The hyperliteral translation beginning, but I also look at neuropsychology to justify why plain translation is actually OK. (Oh God... I'm starting to sound pretentious again...)

There are two major mechanisms in the brain: activations and inhibitions. At the risk of oversimplifying, the brain "activates" lots of neural pathways simultaneously to try to find a solution. When it finds a solution, the activated solution "inhibits" all the other paths, effectively switching them off.

Activations can be associative -- consider Pavlov's dogs. They were trained to associate the bell with food, so the perception of the bell triggered (activated) the anticipation of feeding.

Now, when you have the urge to tell someone (for example) that you love them, you will probably activate that expression in all your languages, because every version is tied to the same concept. Your native language will normally be most strongly activated -- this is only natural. What allows us to speak in other languages is that the outside environment boost the activation of the other language, or actively inhibits your native language. If the person you love doesn't speak your language, and you speak all the time in his/hers, then his/her language will win the activation/inhibition race in your brain.

However, if you associate a phrase with a hyperliteral translation only, and never the idiomatic translation, you might actually be training your brain not to think of the target phrase as the same as the native language equivalent, and in the long run, when you want to say "five cars" your brain will translate it literally because it's "five cars" and not "five of cars".

If you've ever done Michel Thomas's French, Italian or Spanish courses, he teaches the equivalent of "to wait" by association with the English "wait", but uses the hyperliteral "to await" to demonstrate the structure. As the course progresses, he uses "await" less and less, eventually relying solely on "wait" as the prompt for translation, associating the most common term in the target language with the most common equivalent in the native language.

Actually I seem to remember reading a study saying it was better to use L1 translations on cards both because it gave something well-known to associate with and because of the converse, i.e. that the words of the L2 translation would not be as well known, thus leading to a worse mental concept of a word. I can't remember the study now though or where to find it. And as you say, translation works and works well. Perhaps with a 2nd language in a language family the use of L2 only cards would merit more consideration.

A picture in addition to the L1 word /phrase is fine, but not only a picture in case there is more than one nuanced meaning where they are not 100% synonyms or there is a higher topical class. Like with "insect", the picture of any insect along with the word itself gets the job done. But with the picture alone, how does one know whether the general term or the specific one is meant? Or "run" vs "race/dash"?

I've tried the picture flashcard method, but my biggest issue is the time needed to create the cards. A dictionary look-up L2 -> L1 takes seconds, particularly if I'm cut/pasting electronically. When I added the task of finding a picture (via Bing/Google images), carefully trying to select the image that really show the meaning, etc. ... it takes minutes. It can take even longer the further you get from concrete nouns.

In the end, even if it was 'X% more efficient' from a learning standing point, if the cards take 5x longer to create, it isn't a win for me. Particularly for those of us who only have 30-60 minutes a day to devote to languages. (I'd rather spend that time working with native content than building cards.)

As I have mentioned previously in other threads, card creation is already a big time suck, so I too am averse to anything that takes a lot of additional time to add to the card, like rummaging around for an appropriate image. I'd rather just spend a few extra seconds cutting/pasting example phrases and sentences onto the card instead. People who add images and perhaps audio as well to all their cards either don't have nearly as big a deck as I do, or don't use it for as long. Or they simply have many more hours a day to devote to the task.

My answer to this question is complex and personal. Your mileage may vary.

Making translations manually to learn the language (A levels). I've never been interested in making my own written translations, except as a fairly advanced exercise. Whether I go from L1->L2 or L2->L1, the process suffers from several drawbacks:

1. It requires somebody else to correct the translations, or they're of very limited utility.
2. If large amounts of translation are being used for rote memorization, they're less effective than Anki.
3. Making my own translations forces too much clarity, too soon. Almost any real, interesting text will contain quite a few nuances that really aren't important below C1, and which can be efficiently ignored. Producing a good translation requires paying careful attention to the most difficult parts of the text, parts which an A2 or B1 student should probably just skim.

I imagine this could be of use in a classroom setting, for checking student comprehension. But assuming there's no need to give students grades, it seems relatively inefficient.

Making translations manually to discover weak spots (C levels). As sctroyenne has pointed out, translating into the L2 can be quite useful at ~C1 levels. A C1 student should be able to compose directly in their L2 with no particular difficulties, but they may still be subconsciously restructure their argument to work around their weaknesses. Translating from the L1 to the L2 can make some of the weakness visible.

Using translations to create comprehensible text (A levels). I'm very much in favor of this, especially in the beginning. It's one of the cheapest and easiest ways to make input comprehensible. Assimil, the older Linguaphone courses, Listening/Reading, Iversen's use of Google translate—all are potentially useful.

Translating while trying to listen (B levels). Personally, I find this to be poisonous. To boost my listening comprehension, I really needed to kill off my translation reflexes. L2->L1 translation while listening to native speech was actively harmful.

Translation + SRS: Translating L1<->L2 to pass cards. I really dislike any style of card which encourages me to see an L2 word and automatically produce an L1 word, mostly because it creates the exact same reflexes I needed to kill off to improve my listening comprehension. The other direction is marginally more useful, because it makes speaking slightly earlier in the beginning.

Translation + SRS: Using L1 to help understand L2. This is a slightly different approach, where the L1 text may be present on an SRS card, but isn't necessarily used unless a hint is needed. Early-stage bilingual MCD cards work like this, and it's especially helpful with subs2srs cards (which often have L2 audio on the front, and L2 + L1 subtitles on the back).

So translations are a tool, and even sometimes a powerful one. But they're never my main goal. My main goal is always to directly understand and produce L2 content, even if I sometimes use translations as a tool to get there.

I don't really believe in "damaging" a language, except maybe by doing massive amounts of speaking with a very bad accent and seriously incorrect grammar. Adults can suffer from fossilization, even in the presence of sustained massive input. But I don't think translation can "damage" a language, unless it somehow means that the student always thinks in terms of their L1 while speaking their L2, and keeps that up for so long that they build deeply ingrained habits. But aside from that, I think that most alleged "damage" is simply stuff that hasn't been learned yet.

As for translating while listening, it was definitely a problem for me, personally. I had spent so long reviewing L1<->L2 vocabulary cards in Anki that I'd trained myself to reflexively think "of which" every time I saw dont, for example. This worked fine in one-on-one conversation, because people would just slow down and simplify their vocabulary until I had time to translate in my head. But it was a hard habit to break when I started seriously listening to native media, because I'd spent so much time drilling dont → "of which" into my brain, and similarly ingraining another 1,000+ translations like that.

To fix this, I needed to learn to listen to French, and to understand it as French, without translating. (I could already do this for household stuff, thanks to massive exposure, but not for news radio or TV, which used a much larger and more advanced vocabulary.) This helped a lot, and it undid all my bad habits. And I redesigned all my Anki card formats to avoid making the same mistake again—none of my cards encourage me to drill a specific translation, not even in Egyptian. My goal is either to understand the text on the card directly, or to correctly fill in a blank.

For me, this seems like a personal problem, and so I hesitate to offer "don't translate" as advice for others. The most I will say is, "If you're having trouble with listening comprehension because you've trained yourself to translate reflexively, and if you can't get your speed up, maybe it's time to try a different approach." I can easily imagine that other people will never see this problem at all.

This is one of the great challenges of giving language learning advice to novices: untangling things which were important to me from things which were important to everybody. I think a lot of polyglots overgeneralize from personal experience: witness the way that Benny insists on speaking very early, and Khatzumoto resists speaking until very late. I think that "don't translate" falls into the category of advice that was important to me, because my language-learning process went wrong in a very specific way, and not translating was the easiest way to fix it.

The two things (L1->L2 and L2->L1) are very different. Look back at Peregrinus's comment about picture flashcards. How do we know that the picture represents "insect' rather than "green aphid"?
Either:
A) we don't
or:
B) we don't process it as a "picture" at all; it's a pictograph, and a pictograph represents a word.

After all, what is a word but an arbitrary label that hangs off a concept.

Now here's a question: who uses flashcards going pic->L2? Now how about L2->pic?

I don't remember ever seeing any discussion of this, and it's generally taken for granted that you start with the pic and recall the word, whereas with L1<->L2, the direction is always under discussion.

But the point of the pic in pic->L2 is to evoke the concept, and that's what L1->L2 is about - evoke concept to prompt recall of the target word. L2->(anything) is about testing recall, not prompting it (if that makes any sense). Despite the superficial similarity, they are very different inside - don't they feel different to you?

You've picked a brilliant example, as it finally clarifies that eternal argu... uhm.. debate about single-word vs in-context flashcards.

Basically, "of which" isn't a natural representation of a familiar concept in English. It doesn't evoke a concept, which means the only thing you can associate the L2 with is the literal L1 words. If you have something like "mother", it invokes a clear and unambiguous concept, and you don't have to consciously think about the L1 word.

So maybe the question of direction 1<->2 has distracted us from a more important point: flashcards are not a means of teaching a truly knew concept - we use mostly to either train the labelling of a known concept, or to train subcategorisation or specialisation of known concepts.

If a concept is truly new and unknown, like "dont" is to an English speaker, flashcards are never going to do the job. But for "banana", they're pretty effective.