Linguistics and translation

Cliquez ici pour voir ce version en Français

I'm not a translator, I'm a linguist, so my question is, can linguistics tell human translators anything useful? I think it can.

Most linguistic theories involve several levels of analysis of text (I use text here to include transcriptions of speech). For example texts can be analysed from the point of view of phonology – the organised system of sounds in a language. They can be analysed from the point of view of morphology – the way that words in a language can be analysed into meaningful units (or not, as the case may be). Then there's syntax: the analysis of words organised into sentences; semantics – the analysis of the meaning of words and sentences; pragmatics – what people achieve by the use of sentences; and there's discourse – the analysis of sentences organised into larger texts. One popular conception of the task of translation is the transfer of a structure in a source language to a structure in a target language. What are these structures that are transferred?

One answer – one that might seem obviously correct to many – is that it's a meaning structure that gets transferred. So, the appropriate level of textual analysis required for translation would be semantics. To demonstrate what this means we can look at a favourite example of mine – one I've used often, even commenting here on nakedtranslations. Consider the Dutch sentence

Ik zwem graag

If we are to translate this to English we are unlikely to worry much about transfering anything concerning the phonology, the morphology or even, in this case, the syntax. Look at the syntax. The sentence in Dutch has a syntactic structure something like this:

[Pronoun [Verb(finite) Adverb]]

Now the English translation equivalent is

I like to swim

Which has the structure

[Pronoun [Verb(finite) [Inf Verb(non-finite]]]

(A finite verb is one that is marked for tense, a non-finite verb one that carries no tense information. Don't worry about the labels just now, or the missing detail: just notice that the analyses are different). Transfering the Dutch structure to English wouldn't do the job. What is required is to make some representation of the meaning of the source sentence for transfer to the target language. It's not obvious by the way how to do this at first sight but it's perfectly possible. In fact you would probably find that most linguists would give this a representation something like

(Pred(Argument_1, Argument_2))

Where Pred represents swim/zwem and Arguments 1 and 2 are the subject Ik/I and some relatively abstract analog of graag/like to. My point here is simply that this example demonstrates that syntactic transfer is not always appropriate.

Of course, it might be in other cases. There might even be occasions when phonology is an appropriate level of tranfer. The only examples I can think of are poetry and texts to be sung. A particular problem concerns haiku. The most usual haiku form is a three line text with a syllable count 5 – 7 – 5. Here's a famous haiku by the zen master Basho:

Furu ike ya
kawazu tobikomu
mizu no oto

Now, it's part of what a haiku is that it has this metrical structure and so there is a case to be made that it should be preserved when translating. It's a very difficult job. As it happens, some translators have tried it and you can see examples both where they do and they don't on this site. Here's one of them that does preserve the syllable count in the English:

The old pond is still
a frog leaps right into it
splashing the water

(Translated by Earl Miner & Hiroko Odagiri and cited from The Bureau of Public Secrets).
Here in order to preserve the phonological structure the translators have compromised on preserving the syntax and one might easily argue the meaning. There is no equivalent of right in the second line of the translation in the Japanese original. So here's a case where the approriate level of transfer between source and target was felt to be phonological as well as semantic.
I can also think of examples where the appropriate level of transfer is neither phonology, syntax or even semantics but pragmatics. There's an English idiom not in this lifetime. You can use it in situations like the following:

A: Do you think he'll get married?
B: Not in this lifetime!

Let's imagine translating this into Swedish (just for a change of language). I think the following would do it

A: Tro du at han ska gifta sig?
B: Absolut inte!

The Swedish phrase absolut inte would be literally rendered absolutely not. No transfer from the English at any level other than the pragmatic would make for an appropriate translation.

What about discourse?
There are occasions when the appropriate level of transfer is discourse. I said above that discourse analysis includes relations between sentences or units bigger than sentences but strange as it sounds a sentence can be bigger than a sentence - when it's considered from situated point of view. OK, that's a bit opaque but look at some examples, this time from Chinese.

haizi mai le shu
child buy PERF book
the child bought a book
(le/PERF is a particle that indicates the action is complete).

Notice that Chinese doesn't have any analogs for the indefinite and definite articles here. Notice, and move right on. Consider

haizi mai le yi-ben shu
child buy PERF one-item book
the child bought a book

Here yi-ben is the numeral one and a classifier which I've glossed item. Its use in this case is very much like an indefinite article in English. Now things get interesting. Look at these examples.

shu, haizi mai le
book child buy PERF
The child bought the book

Chinese encodes discourse level information about definiteness partly by the position of a phrase within the sentence. Now for translation of the Chinese into English this information has to represented in the structure to be transfered from source to target. In building the translation we have to see just how information is encoded at different linguistic levels in the two languages.

But it's not just the nature of the source or target language that can dictate the appropriate level of transfer, it can also be the genre or type of text.
I said above that there are cases where the translator of a literary text might wish to transfer at the phonological level. But at the other extreme there are cases where the transfer is entirely at the discourse level. If I am translating the instruction manual for some piece of equipment, then I won't much concern myself with any transfer below the level of the global, information structure of the text, I won't care about preserving any syntactic features.

So linguistics can do this much for translators (even if linguists haven't yet built a robust model of how translation is carried out): it can provide translators with categories and vocabulary for discussing and thinking about translation problems. In the examples I've used, thinking about translation as transfer at different linguistic levels helps to make explicity why some or other strategy is appropriate or succesful.

So, linguistics can be the translator's friend. I don't think you all have to read Chomsky, but I think that just as linguists can benefit from understanding what translators do, translators can plausibly find linguistic analysis a useful part of their armory.

Reprinted by permission from
Naked Translations: English to French translation