Sonia Colina reports on a linguistic phenomenon we don’t often think of:
Despite efforts to increase research participation, racial and ethnic minorities remain underrepresented in results. A review of 5,008 papers in three pediatric journals from 2012 to 2021 revealed that only 9% of these studies included non-English speaking volunteers.
Language is a key barrier to participation, as even those with some English proficiency are less likely to participate in studies when recruitment materials aren’t in their native language. Language barriers also hinder a person’s ability to provide informed consent to participate.
This problem is not likely to fade away. The number of people with limited English proficiency in the U.S. grew by 80% between 1990 and 2013, going from nearly 14 million to 25.1 million people. As of 2022, this number rose to 26.5 million people. Excluding people with limited English proficiency is not only unethical, as these groups deserve the same access to experimental and evolving therapies as the English-speaking population, but also limits how applicable research findings are to the general population.
One way to eliminate language barriers is by translating research documents. As a translation scholar, I strive to discover ways to improve translation quality to benefit the research community and broader society. Translation in research, however, is not straightforward. Not only must the translated materials be accurate, they also have to serve their intended purpose.
More at the link, of course, and I hope attention is paid. Thanks, Bathrobe!
Just a part of a much broader problem: provision of proper healthcare of all kinds, not just enrolment in research projects, is frequently seriously impaired by lack of adequate resources for people who don’t speak English well.
In the NHS, the problem in face-to-face consultations is “dealt” with by the pretence that telephone services are just as good as actual translators. I don’t believe that the people who crafted these pretend “solutions” are genuinely so stupid as to think that they are adequate: it’s just a fig-leaf. To be fair, when the lower-than-vermin Tories have been deliberately starving the NHS of adequate funding for over a decade, you can see why such decisions are made.
The problem is exacerbated by diktats to the effect that we are forbidden from relying on accompanying relatives to interpret, in case said relatives make mistakes in translation, thereby exposing the organisation to legal liability. At least, it would be exacerbated, if one were at all inclined to pay any attention at all to such diktats. Much money could be saved by sacking all those whose jobs are to sit on their arses in offices far from any actual patients and dream such things up. Might be able to afford more translators then …
This piece simply assumes that being a guinea pig is good for the guinea pig as distinct from the availability of guinea pigs being useful for the Progress of Science. But why would you believe that? Socially marginal groups (in any given society) often have entirely plausible reasons, given the oft-poor ethical track record of medical research, to not include insisting on being represented pro rata among guinea pigs as a particularly important desideratum.
Overcoming language barriers (not to mention cultural barriers that may be statistically correlated …) in providing adequate medical care (with procedures or meds that have passed from the “experimental” stage to the “standard recommendation” stage) to non-guinea-pig patients seems an entirely different issue.
I was intrigued by her excursion into translation theory in this article, that is, her advocacy of the “functional approach” to translation, with a couple of apt examples.
One of the traditional, rather tired, oppositions in translation theory is that between “literal” and “free” translation. A literal translation is one that closely tracks the words of the original; a free translation licenses the translator to depart from the literal meaning and capture the intent of the original. However, this opposition raises more questions than it solves. For one, it assumes that dictionaries are an infallible guide to words and their accurate translation equivalents — which is nonsense.
One old concept from translation theory that attempts to capture this “freedom” more accurately is Nida’s “dynamic equivalence”, the idea that a translation can be equivalent to the original in a “dynamic” rather than word-for-word sense.
But I find the functional approach a more useful concept. It recognises that translations can depart from strict literalism in a more clearly defined way: the translation must perform the same function as the original. Needless to say, pinning down the function of a translation is also difficult, but it at least suggests a specific way to test for aptness. The example given, “Consigo pensar claramente” rather than “O meu pensamento é claro”, is a nice one. (One does wonder, however, what sort of native speaker of Portuguese would think that “O meu pensamento é claro” is a good translation if it is that unclear to ordinary speakers…)
an entirely different issue
Depends a bit: there are areas where the available treatment options are pretty much all experimental to some degree: for example, many lymphomas and leukaemias (as a close relative of mine can testify.)
Experimental in the strict formal sense, too, not just the customary medical “nobody really knows what the hell is best in this condition, so let’s try this” sense. You end up enrolled in a clinical trial as part and parcel of your treatment, more or less routinely. Come to think of it, in a better world, that would probably be the best way to conduct all treatment schedules.
I have three or four points of contact with this topic. The first is the most immediate and personal, since I am just now emerging from a scary illness in a healthcare culture quite remote from the comfortable and familiar Australian system.
Persistent fever, headache, complete loss of appetite … several diagnoses would be available from qualified and accessible professionals here in Bali. So why have I not consulted them? Well, I would have if things got worse. But I knew already what the interventions would be: hydration, rest, and paracetamol. I could see to all that by myself. It’s not COVID, and I was able to watch for specific developments that might signal something truly ominous. Heading home in a few days, in travelworthy condition.
Pharmacies here are interesting and generally useful, though there always doubts about the qualifications of personnel. And still these days, doubts about the authenticity of the medicines on offer. I called into one to buy a thermometer (essential tool), and was offered all kinds of spurious stock concoctions to speed my recovery: some known to me, some inscrutable. I ignored the recommendations from caring but unversed shopgirls (no pharmacist was in sight) and left with just the thermometer and some handy electrolytes. Language was not an issue, but culturally founded expectations were. In Australia no pharmacy is permitted to do business without a professional pharmacist on duty; they’re always present, knowledgeable, resourceful, and of high integrity.
Second, in multicultural Australia I’ve seen little hesitation to accept informal help – linguistic translation, cultural translation – in healthcare transactions. My wife and I regularly smooth things for a cranky 84-year-old man whose L1 Serbian and Hungarian and L2 (L3?) English make things arduous for his vascular surgeon and a host of other caring professionals. Reluctance to wear his hearing aid is another difficulty, along with a general linguistic inflexibility in the professionals themselves: but they’re very ready to receive our active interventions. Legal considerations seem remote.
@Bathrobe, the example illusrates that literal translation means better quality of back translation (which they say is a common quality-assesment tool). But it clearly can’t illustrate dynamic equivalence (WP says that Nida himself replaced it with “functional equivalence”)
PS. Or well, sorry. You did not meant that it illusrates “dynamic equivalece”.
The problem is exacerbated by diktats to the effect that we are forbidden from relying on accompanying relatives to interpret, in case said relatives make mistakes in translation, …
There is a genuine difficulty here: with recent-immigrant families (especially refugees) often it’s the kids who are the competent English speakers. a) They might not know the English terms for obscure parts of the body, or obscure sensations; b) their elders might not be comfortable telling their kids such intimate details, and use euphemisms which’ll only get lost in translation.
And of course all sorts of cultural gradients if the patient is female, the relative male. (Also if the medic is female: demands to see a ‘proper doctor’, not a nurse. My niece (in Manchester) was still getting this when already a Senior House Surgeon.)
So how much do medics rely on testament from the patient? Can they actually diagnose from examination and tests alone?
(I’m not saying it goes any better with telephone services alone. Clearly properly trained translators at the bedside is what it needs.)
There are just too many aspects of translation.
E.g. semantic units of different level matter to speakers, so you have to choose between (semnatic/functional) equivalents of words and (semantic/functional) equivalents of other units.
It is functional both times
Yes, of course there are some expressions who exact components are of little importance to speaker but which as a whole used in very exactly same situations and for the same reason and to convey the same idea as some expression in another languages. But there are texts where everything matters.
___
In this case the situation is unusual in that these materials have very little value. They were composed with a specific purpose, now you can compose a new text in a new langauge or you can translate this text.
The only thing that matters is the purpose, authencity does not matter at all.
There are just too many aspects of translation.
E.g. semantic units of different level matter to speakers, so you have to choose between (semnatic/functional) equivalents of words and (semantic/functional) equivalents of other units.
It is functional both times.
Yes, of course there are some expressions whose exact components are of little importance to the speaker but which as a whole are used in exactly same situations and for the same reason and to convey the same idea as some expression in another language. But there are texts where details from different levels matter.
___
In this case the situation is unusual in that these materials have very little value. They were composed with a specific purpose, now you can compose a new text in a new langauge or you can translate this text.
The only thing that matters is the purpose, authencity does not matter at all.
@ drasvi
Thanks for updating me on Nida.
The point is that back translation doesn’t necessarily provide an accurate guide to the quality of a translation (because “a perfect match in backtranslation often reflects a translation that is too similar to the original”). I’d heard of back translation as a way of ensuring that the translation is a good match to the original, but that was a long time ago. I must admit that it made sense at the time, although a little thought shows that it doesn’t.
I have little experience with medical matters, except for one embarrassing example when I was working for a Japanese market research company. The company was keen to break into the medical survey market. My highly self-confident and “commonsense” but ignorant Japanese boss took on a job that I was put in charge of. We duly translated the questionnaire into Japanese and had our interviewers administer it to nurses. There was one question the Japanese nurses were nonplussed by, which involved certain units of measurement in taking patients’ symptoms or administering medication. I’m afraid I don’t remember what the purpose or the unit was, but the upshot was that the British client used European units in the questionnaire whereas Japanese nurses use American units.
The client justifiably complained about our work. If we were professionals in medical surveys we should have known that the units had to be changed. Our head office’s response was that “our company is listed in the market research handbook” so we must be ok (cringe).
Third, as an editor assisting research teams to secure funding I have often dealt with grant schemes administered by Australia’s NHMRC (National Health and Medical Research Council). I am impressed with the thorough attention applications are expected to give, and do give, to ethnic and linguistic diversity in the populations they study. This dimension seems as salient as gender, socioeconomic status, and age, though of course the details will depend on the specific nature of the proposed investigation. I don’t know how this fits with health-related research elsewhere in the world, but it’s an unmissable characteristic here.
@bathrobe
You are blaming Japanese management for face-saving. If the complaint had been worded in a different way and not with what I assume was a more legalistic or hectoring approach on the part of the complainant, I suspect they would have got a different reply. I think many European companies would also react very cautiously to complaints in this situation, i.e., would want to be paid or face-saved before repairing the error, especially if there were vague requirements or pressure to deliver without completing a full in-house quality assurance or testing campaign.
So how much do medics rely on testament from the patient? Can they actually diagnose from examination and tests alone?
Depends on the specialty, and depends on the disease. There are certainly plenty of diseases that it is completely impossible to diagnose without the patient’s testimony.
There’s also the fact that it’s common for there to be several reasonable treatment options, all with their own pros and cons. If you can’t communicate with the patient, you end up deciding for them, which is Bad. And you won’t be able to warn them properly about side-effects/downsides of treatment, so they may stop the treatment, whereas if they’d been better informed, they might have decided that the side-effects were a price worth paying.
It’s also an objective fact that patients are more likely to get better if they trust their doctor (including in domains where it’s a mystery how this could work causally.) If you can’t communicate, you can’t build trust.
I’m afraid I don’t remember what the purpose or the unit was
This sounds like it might be blood glucose, which is mg/dl in the U.S., Germany, and Japan, but the proper SI unit of concentration mmol/l elsewhere.
An interesting and slightly related article I’ve just read:
Translation as discrimination: Sociolinguistics and inequality in multilingual institutional contexts
Thanks, that’s definitely of related interest. The abstract:
I have to say (having only skimmed it) it seems to me to be conflating issues of translation with the general problem of bureaucratic inhumanity, which may not be helpful.
The stuff about blind faith in Google Translate is also interesting. I don’t think this only happens because of racist disinclination to consult native speakers, though: I’ve seen many a ludicrous example in English versions of signs in France, for example.
It happens partly because native speakers are not conveniently to hand, and might even – horrors – be expected to be paid for their input, but also because most people have no conception of how crap Google Translate really is, not least because of the ever-popular language-as-bag-of-words model they are working with.
(This intersects with the whole Artificial “Intelligence” thing, of course. GT has no understanding whatsoever of any text in any language. No wonder it’s crap, especlally whenever people try to use it to translate anything that isn’t simply a string of disconnected propositions.)
The stuff about “referential transparency” is on the nose. Very often, the effect of translation is to produce something that indeed, means “the same” as in the source text: the trouble is, that it is not at all what any native speaker would actually say in that situation: the translation has failed.
(I used to encounter this in English in Ghana.)
Socially marginal groups (in any given society) often have entirely plausible reasons, given the oft-poor ethical track record of medical research, to not include insisting on being represented pro rata among guinea pigs as a particularly important desideratum.
There are probably some cases (Australian Aborigines being an a priori likely example) where you might legitimately want to include some members of a locally socially marginal group in your medical research trial, because they are also (relatively) genetically divergent from the local majority population, to exclude the inconvenient possibility that the treatment you are researching in fact works differently for them (because it depends on a mutation that they lack, or because it is affected by a mutation that they have and the majority population doesn’t).
On further consideration there are probably also some (perhaps more frequent) cases where the treatment you are researching might have a bad interaction with a different disease that happens to be almost entirely limited to the lower classes… those are probably harder to catch even if you do try to include lower-class patients in your test groups, though.
I don’t disagree with JanFoM about optimal research design, just saying that for a variety of historical, cultural, and political factors, interest in participation may not be evenly spread across all relevant demographic dimensions, not least because how generally benignly versus suspiciously people view the sort of institutions and researchers involved may vary in ways correlated with some of those demographic dimensions.
And of course getting a better set of research subjects (from a demographic cross-section perspective) will generally require more time, effort, and funding. Language barriers further exacerbate that.
Two things about Prof. Colina. First, her focus and expertise is on translation between Spanish and English – since Spanish is by far the most commonly-spoken language by current U.S. residents, there are both greater incentives to get translation into Spanish right and certain economies of scale in doing so. So if that’s not being done very well, imagine what that says about U.S. institutions that would ideally be able to communicate with relevant immigrant populations in Albanian or Tamil or what have you. Second, some of the titles of her publications (such as “Fonología generativa contemporánea de la lengua española” and “Optimality-Theoretic Studies in Spanish Phonology” suggest the possibility of some Chomskyan affiliations …
Actually, maybe I was being overhasty in buying the notion that this was actually a problem from the perspective of the research yielding high-quality results. Census Bureau data estimates approx. 3.4 million people in the U.S. who speak “Chinese” at home, of whom approx. 1.8M are LEP and 1.6M are fluent in English. No doubt there are some sociological differences between those two subsets but if you have an adequate proportion of research subjects of Han-Chinese ancestry in your study, does it matter if they’re all recruited from the fluent-in-English subset of that population? (And of course there are also Americans of 100% Han ancestry who speak only English at home, even if they have some heritage-language ability to talk to grandparents on the phone etc.) Certainly if the concern is how genes might affect the efficacy or dangerousness of an experimental medication, the language spoken or not spoken by the bearer of particular genes would be irrelevant.
The LEP percentage varies considerably by non-English language spoken at home, but it’s hard to find at a quick skim (this is ACS table B16001 for you Census-data buffs …) a potentially relevant population where the LEP percentage is so overwhelmingly high that it would obviously not be practical to oversample the English-fluent subset of the relevant group. And of course you need to figure out the right level of generality for “relevant population.” Are Haitian-Americans and Jamaican-Americans distinct from each other in ways that are relevant for typical medical research? If not, you don’t have to worry about LEP Haitian immigrants if you have enough Jamaicans in your sample.
I would guess things like stress are a factor here? Cortisone suppressing the immune system or something like that?
My anecdote: I was at the big city hospital a few months ago, and the nurse/doctor in the room next door was communicating with an elderly patient using a Cantonese translator, over the phone. Cantonese is probably the third most spoken language here (after English and Spanish), and probably more so among older, hospital-going people. It seems that even in this near-optimal case, it is not worth it for the hospital to employ a staff Cantonese interpreter, preferring a contractor over a mediocre phone line, who probably works as part of a pool which serves institutions all over. Perhaps the Chinatown hospital is different.
On the other hand, there always seem to be enough Spanish-speaking staff on hand to translate in person.
areas where the available treatment options are pretty much all experimental to some degree
and there are many areas where it’s very easy to be defined into only having access to standard treatments through clinical trials, for potentially scientifically-defensible but fundamentally ass-covering reasons dictated by hospital lawyers (and probably insurance companies).
for example, as someone i know recently found out, if you have any kind of autoimmune diagnosis – including things like psoriasis as well as, say, crohn’s disease – in the u.s., at the moment, that’ll be your only path to immunotherapy for cancer (at least certain fairly common kinds; i can speak with certainty about adenocarcinoma of the lungs). never mind that the ‘side effects’ at issue may be things you already experience continuously, or/and things that have no chance of going unnoticed.