<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">

Andrew Chesterman


1996d      Psst! Theory can be useful! Kääntäjä  10, 4-5.


Psst! Theory can be useful!

Kääntäjä - Översättaren 10.12.1996



Human or machine?

The Finnish general text of the recent Licensed Translator Exam was an article about how the municipalities in Finland had tried to cope with the economic crisis, how they had had to cut down services, etc. The translation was supposed to be published "in a local government publication" (true, rather a vague function, but better than nothing). One sentence began as follows:

Sekä tilastoihin ja kunnissa tehtyihin haastattelututkimuksiin perustuvat selvitykset että puhtaasti valtakunnallisiin tilastoihin perustuva palvelukatsaus osoittavat terveydenhuollon tehostuneen...

Not particularly elegant Finnish, OK; but what did the translator candidates do with it? Here are eight versions (some are slightly edited). Seven are by people taking the exam, and one is by a computer (the TranSmart system, by courtesy of Petteri Suoranta of Kielikone). Question: can you spot the computer one?

1.  Reports based on both statistics and interviews in municipalities, and service reviews based on national statistics show that health care has become more effective.

2.  Both the surveys based on statistics and interviews carried out in municipalities and the survey of services which is based on purely nationwide statistics show that Finnish health care has become more efficient.

3.  Studies based on statistics and interviews done in the municipalities and a service review based on national statistics both indicate that the health services have improved.

4.  Studies based on statistical and interview research conducted in municipalities, as well as a service survey based purely on national statistics, demonstrate that the efficiency of health care has improved.

5.  A study on services which is based on statistics as well as interviews made in the communes and on purely nationwide statistics shows that health care has become more efficient.

6.  The effectiveness of public nursing has improved. This is shown both by reports based on statistics and local interviews and by a survey based on strictly national statistics.

7.  The health service has become more efficient according both to statements based on statistics and interview researches done in the municipalities and on a service review based on purely national statistics.

8.  Both the reports which are based on the statistics and on the interview studies that have been done in the municipalities and the service review which is purely based on national statistics show that the public health service has improved.

Before I tell you the answer, note that all these versions are strikingly similar in some ways. This is partly because the TranSmart system is pretty good, but partly because trainee translators often seem to translate like machines, carefully processing every word of the original into a grammatical target-language form. The results are comprehensible, but not particularly good translations. Professional humans can do better. Why? Because professional humans have access to theoretical concepts that machines do not have, and these concepts affect the way they translate. The depressing thing about the human versions here is that they seem to reveal a total absence of such theoretical concepts. What kind of theoretical courses have these candidates attended, I wonder? What theoretical books have they read about translation? Or are they of the opinion that theory is a waste of time for a practical translator? If they have in fact studied any translation theory, it seems to be kept in a very separate compartment in their brains, not to be confused with actual practice. Have they never read anything about the relation between theory and professional development (e.g. Viaggio 1994)?

When assessing the scripts, I had the feeling that many of these would-be professional translators were working like amateur carpenters, trying to make a decent book-case but without using obvious things like a saw, hammer, screw-driver etc. Here are a few simple theoretical tools that would have been useful in deciding how to translate the example sentence.



This means changing the word class. Transposition is the term used by Vinay and Darbelnet (1958), but other theorists have used other names. This seems to be an unknown idea to the writers of the versions above (why?). Finnish nouns come out faithfully as English nouns, participles come out faithfully as participles, and so on. The result is rather like what a good machine translation program could produce... Look at version eight, for instance: this was the machine one. (I have cheated slightly here: the machine had three lexical problems, offering "cleanly" for puhtaasti, "look" for katsaus and "intensified" for tehostuneen. It would be a relatively simple matter to adjust the computer's dictionary accordingly.)

But is there something sacred about these word classes here? Suppose, for example, that we translate kunnissa as an adjective: municipal: this was a solution that occurred to very few candidates indeed, and yet it makes it much easier to produce an acceptable version, less cluttered with nouns and prepositions.



A well-known related tool is that of deverbalization. This is a key term in the training used at the ESIT school of interpretation and translation in Paris (see eg. Lederer 1994). It means simply that a translator or interpreter has to get away from the surface structure of the source text, to arrive at the intended meaning, and then express this intended meaning in the target language. (I will bypass here the deconstructionist arguments that there is no objective meaning there in the first place. Translators have to believe that there is something there, after all...) In other words, deverbalization is a technique used to avoid unwanted formal interference: professional translators need to process the intended meaning in their own words, rather than try to mechanically manipulate source-text structures.

What have our translators done? In most cases, they seem to have processed the source structure bit by bit, twisting it into some form of grammatical English, without stopping to ask what the hell this is actually supposed to mean. I'd love to ask them: if you had this idea to express in English, but no Finnish source text to distract you, would you really express it like this? Are these words in the translation "your own words"? Would you really clutter the sentence with repetitions of "based on"? Would you really choose to have such a long subject? Would you yourself use "both" in this way? Do you actually like using such complicated noun phrases? More importantly: do you think your readers find such a style pleasant to read, easy to understand?

Versions (6) and (7) have attempted some deverbalization. In (6) there is an additional sentence break which makes the idea clearer; and in (7) the sentence has been turned back to front, allowing a light subject that is easier to process.



To translate this sentence acceptably, it helps to have the concept of iconicity in your tool-box. Roughly speaking, iconicity is the matching of form and meaning, so that the form reflects the meaning or the experience that is being described. It is a well-known pragmatic maxim of clarity that iconic expressions are easier to process than non-iconic ones. (Switch on after plugging in is not iconic. Plug in before switching on is iconic, because the order of information expressed matches the order of the events described) The iconicity concept helps to solve the problem with both in our example sentence.

If you start <nobr>Both X and Y...</nobr> readers will naturally assume that the X and the Y are indeed the two things you are connecting with both. The trouble is, in several versions here we have the structure <nobr>Both X and Y... and Z</nobr>, so that the reader does not know whether the picture [is] supposed to be <nobr>(X+Y) & Z,</nobr> or <nobr>X & (Y+Z),</nobr> or even <nobr>X & Y & Z.</nobr> The structure is non-iconic: there is an awkward clash between form and meaning. We have both, which suggest two things, but a list of three things. In Finnish, the problem does not arise, as we have <nobr>sekä X ja Y että Z,</nobr> and we know that the basic pair is marked by sekä and että. We know that the intended picture is therefore <nobr>(X+Y) & Z.</nobr>

How can we show what goes with what in English, that the basic pair is presumably "municipal" & "national"? How can we produce a form that is iconic? Version (1) tries to do this by repeating based on and by adding a comma after municipalities, but this does not work well enough, especially as the necessary second comma, after national statistics, is missing. Repeating based on makes for a very cluttered sentence because there are still loose and's floating around. Version (3) postpones the both, after a repeated based on: not a bad idea, but there are still too many and's. Version (4) drops both and replaces it with the longer phrase as well as: a nice insight. Version (5) uses as well as to link the two smaller elements X & Y, not the main ones (X+Y) & Z. This is not iconic, for the version uses a longer phrase to separate the minor points and a shorter word (and) to separate the main points; an iconic one would match the size of the linking phrases with the hierarchical importance of the items linked: bigger links for bigger items. Version (6) repeats by to show the major elements: a good choice.

The difficulty is to find an iconic solution which can be combined with the need to deverbalize. So let's appeal to the concept of relevance.



Readers of the translation will of course be different from the readers of the original, and will have different cognitive backgrounds etc., they will have different ideas about what is relevant to them (Gutt 1991). The translator's job is to translate what is relevant: this may mean explaining or adding or omitting things occasionally. Gutt (1990: 157) puts the point in a way that links it nicely with the deverbalization concept:

What the translator has to do in order to communicate successfully is to arrive at the intended interpretation of the original, and then determine in what respects his translation should interpretively resemble the original in order to be consistent with the principle of relevance for his target audience with its particular cognitive environment.

Hönig and Kussmaul (1982; Kussmaul 1995) use a similar concept, which they call the maxim of sufficient degree of precision: professional translators go to the level of precision needed, not beyond it.

So: consider the original Finnish. Do we need to translate, as separate words, tehtyihin, perustuvat selvitykset, puhtaasti, perustuva, palvelukatsaus? After all, what is the point here? (a) That there are two sources of evidence referred to; (b) one of these is municipal and the other is national; (c) the municipal source is (presumably) based both on statistics and on interviews; and (d) the national source is a statistical one. Suppose we simply translate the passage as: Municipal statistics and interview studies, and also national statistics, show that... Have we lost anything that is relevant? Surely any intelligent reader will infer that there has been a national survey of some kind which has produced these statistics; indeed, that the statistics are "based on" this survey; that if I say "national" I normally do not mean "and municipal" (so "purely" is unnecessary).

Alternatively, how about Municipal statistics and interview research, and also a national service review, show that ...? Here we have dropped the statistics bit from the national research bit, and highlighted the service idea; but we can surely assume that any national survey will be based on statistics of some kind, maybe we don't need to say this explicitly. (By the way, both these suggested translations are iconic: and also plus the commas separate the two main items, whereas the minor items are linked by the shorter and. Neither of them bother to say "both".)

Another way of making this same point is to use the concept of implicitation. That is, one of a translator's tools is the technique of making information implicit: not everything needs to be explicit, not even everything that was explicit in the original. The opposite procedure is of course explicitation. Version (2) is the only one here that uses this tool: the writer has realized that if the translation is going to be published abroad, it might be helpful and relevant to add an explicit reference to Finland occasionally: Finnish health care. Good thinking!

But doesn't all this mean that the translation changes the style of the original? Well, in what respects should the style remain the same here, do you think? Same degree of garbage? If the translation of a text like this (we are not talking about literature, after all) ends up being clearer than the original, so what? Improving the original is another useful concept.



Gutt, E-A (1990) 'A theoretical account of translation - without a translation theory.' Target 2, 2, 135-164.

Gutt, E-A. (1991) Translation and relevance. Cognition and context. Oxford: Blackwell.

Hönig, H.G. and P. Kussmaul (1982) Strategie der Übersetzung. Ein Lehr- und Arbeitsbuch. Tübingen: Narr.

Kussmaul, P. (1995) Training the translator. Amsterdam: Benjamins.

Lederer, M. (1994) La traduction aujourd'hui. Paris: Hachette.

Viaggio, S. (1994) 'Theory and professional development: or admonishing translators to be good.' In Dollerup, C. and A. Lindegaard (eds) 1994, Teaching translation and interpreting 2. Insights, aims, visions. Amsterdam: Benjamins, 97-105.

Vinay, J.P. and J. Darbelnet (1958) Stylistique comparée du français et de l'anglais. Paris: Didier.





Back to [Publications list]                            Back to [Homepage]