Languages, the census and the National Health Service

‘Welcome to the Head and Neck Outpatient Department. If you want to change the language touch one of the buttons below’. This invitation to choose your preferred language for checking in seems like a perfect use of technology to broaden access to a service and engage with a diverse public.

NHS check-in

However, there are a number of odd things about this check-in screen.

To begin with, the choice of languages. Apart from English, there are seven: Gujarati, Gaelic, Italian, Latvian, Polish and two versions of Chinese. However, these do not seem to be chosen on any obvious basis. According to the 2011 census, the 10 non-English languages with the highest numbers of speakers in the district (those who declared their ‘main language’ as one other than English) were Polish, Greek, German, Gujarati, French, Spanish, Arabic, Chinese, Russian, and Bulgarian. The last of these had just over 100 speakers (out of 134,000 people in the district overall). Italian was a few places behind; Latvian, with 14 speakers, a long way down the list. There were no speakers of Gaelic recorded at all.

There is a reasonable amount of overlap between the languages offered on the screen and the list derived from the census, but even so it raises questions: why Gaelic (spoken in distant parts of Scotland) rather than Welsh, spoken relatively close by? Why Italian but not Greek?

An alternative way of deciding on the languages to offer would be by proficiency in English. For example, in the national census, of the people who claimed French as their main language, only 5% said they knew little or no English. Greek and Spanish are around 10%. On those grounds, one might decide there was little need for a French, Greek or Spanish option. On the other hand, 17% of people whose main language was Arabic or Bulgarian said they had poor or no knowledge of English, so those languages still seem to have a stronger claim to be included.

A second peculiarity of the check-in screen is the way the languages are labelled. ‘Gaelach’ for Gaelic is simply wrong: this word means ‘Irish’. The expected label would be ‘Gàidhlig’ for Scottish Gaelic, or ‘Gaeilge’ for Irish Gaelic. ‘Latvisk’ is the Danish (or Norwegian or Swedish) for ‘Latvian’ but the Latvian for ‘Latvian’ is  ‘latviski’ or ‘latviešu valoda’. Furthermore, the Chinese labels say ‘Cantonese’ and ‘Mandarin’ but the actual distinction should be (and probably is) between the traditional script (used in Hong Kong and Taiwan) and the simplified script used in mainland China. That is the distinction made in the language options in the Hospital Trust’s own website, which shows a different range of options – Polish, Urdu, Gujarati, Turkish and the two versions of Chinese – ‘Simplified Chinese’ and ‘Traditional Chinese’ (see screenshot below). Although in the British context, most of the ‘simplified’ readers would be Mandarin speakers and most of the ‘traditional’ readers would be Cantonese speakers, the real difference is in the form of the script, not the language.

University Hospitals of Morecambe Bay NHS Foundation Trust - languages

This does not inspire confidence in the quality of the translations in the automated check-in, nor does it seem very respectful to the speakers of the languages concerned. Imagine you are a speaker of English, and are invited to press a button labelled ‘Inglish’.

All this raises issues of how the National Health Service – and other public services – engage with multilingualism (and, of course, multilingual literacies). Although in the past, language information has been very scarce in England, this is arguably not so any more. The 2011 census in England broke new ground. Language questions had previously been asked in Ireland, Scotland and Wales, but never in England, and the census authority, the Office for National Statistics (ONS) was not convinced of the case for one. However, the ONS eventually became convinced that there was a need for language information, and the 2011 census included two questions on language: ‘what is your main language?’ and (if the answer to that was a language other than English), ‘how well do you speak English’.

In the consultation period before the 2011 national census in England, many people within the health services were arguing for a language question to be included on the census questionnaire. For example, Peter Aspinall, a senior research fellow at the Centre for Health Services Studies at the University of Kent, urged the National Health Service to press for the inclusion of language questions in the 2011 census, citing ‘pronounced ethnic disparities in health and health care’ Aspinall (2005: 364).

So how is the information gathered being used? It is not necessarily the case that the census is the best way to collect language information, and it can be argued that the quality of the information collected (or of the questions asked) is not of the desired standard. Even so, it is not clear that the information is being used at all. Perhaps the NHS lacks the resources to exploit the language information which we now have. Which leads to the question: who makes decisions about languages other than English within the NHS, and on what basis?

References and sources

Aspinall, Peter J. 2005. Why the next census needs to ask about language: Delivery of culturally competent health care and other services depends on such data. BMJ 331, 363–4

Office for National Statistics, 2011 Census. Table DC2210EWr – Main language by proficiency in English (regional).

Office for National Statistics, 2011 Census. Table QS204EW – Main language (detailed), local authorities.



Muslim Women and English: Cameron gets his figures wrong

David Cameron announced an initiative yesterday to provide £20m for  English lessons, apparently to be targeted at Muslim women. In an interview on BBC Radio 4, the  Prime Minister said:

The statistics are clear that there are 38,000 Muslim women who really don’t speak hardly any English at all and perhaps as many as 19,0000 who speak it very badly, and this is about building a more integrated, cohesive, one nation country where there’s genuine opportunity for people.

 (There’s a link to the audio file of Cameron’s interview in this BBC news item).


Unfortunately in this case, the statistics are not clear. Most people listening to Cameron would get the impression that there are 228,000 Muslim women who speak little or no English, but that’s a misinterpretation of the census figures.

Census data from the 2011 census, available from the Office for National Statistics, shows that of 846,000 Muslim females aged 16 and over, 38,000 reported that they spoke no English, and another 152,000 stated that they did not speak English well. That’s 190000 altogether, not 228,000. While it’s likely that this was a genuine mistake, it does not help when figures like these are exaggerated on the media.

Another figure mentioned in this news story – that 22% of Muslim women have poor or no English – seems to be based on the correct statistic, i.e. 190,000 of the 846,000.

As with any statistics, though, it’s important to look both at where they came from, and at the finer detail.

The relevant statistical table, CT0557 – Religion by proficiency in English by sex by age appeared on the Office for National Statistics (ONS) website on 7th January, in a section for ‘ad hoc’ data. This appears to mean that it was produced as the result of a request, and was not part of the normal programme of census data analysis carried out by the ONS. Most likely it was requested by the government itself. The data in the table was the result of census questions on gender, age, religion, main language and English proficiency; of these, the religion question was optional – the only optional question in the census – so we cannot be sure that every Muslim reported themselves as such (there could be good reasons why Muslims might prefer not to answer this question). Equally relevant, proficiency in English was calculated from responses to the question ‘How well can you speak English’ with the possible answers ‘Very well’, ‘Well’, ‘Not well’ and ‘Not at all’. Respondents had to judge their own abilities in English. The ONS believe the results produced by self-assessment are sufficiently accurate, though there is recent research suggesting that it is problematic.

So, when we evaluate the statistics, we have to bear in mind that not all Muslims may be included, and some people may have over- or underestimated their knowledge of English.

Next, the detail. A fact not mentioned in the news stories is that 51% of Muslim females over age 3 reported English as their main language. Although overall, the census data shows that 22% of Muslim women have poor or no English, the breakdown by age shows very substantial differences. Among Muslim women aged 45 to 64, almost 40% do not speak English well or at all; aged 25 to 44, 18%;  aged 16 to 24, 6%; and aged 3 to 15, 3%.

There is thus a very clear trend that even where English is not their main language, younger Muslim women are able to speak it well. In the cohort who are the mothers of the near future, only 6% may be unable to introduce English to their children at home. Among the under-16s, currently targets of the Goverment’s Prevent strategy against radicalisation,  only 3% may have problems speaking English.

Cameron’s announcement is almost a re-run of the controversy created by the then Home Secretary, David Blunkett in 2002. Blunkett published an essay in a national newspaper in which he said, referring to the disturbances of the previous year,

I have never said, or implied, that lack of fluency in English was in any way directly responsible […] However, speaking English enables parents to converse with their children in English, as well as in their historic mother tongue, at home and to participate in wider modern culture. It helps overcome the schizophrenia which bedevils generational relationships. In as many as 30% of Asian British households, according to the recent citizenship survey, English is not spoken at home.

Blunkett’s claim that ‘in as many as 30% of Asian British households English is not spoken at home’ seems not to be derivable from statistics found in his department’s 2001 Citizenship Survey. Data from that study shows that of 3263 respondents who gave their ethnic background as South Asian, approximately 2200 or 67.4% regularly used English. Another 1056 respondents (32.4%) said they spoke no English. Thus the total proportion of homes where English was reported to be spoken was 67%.

However, the survey questions appear to be about the respondent’s personal use of languages: ‘What languages do you regularly speak at home?’ and ‘Which language do you speak most often at home?’ and assuming they were understood in that way, they do not tell us about the use of any other members of the household. The sample was designed to be representative of the population of adults aged 16 and over, so it overlooked the children – the group most likely to have good English proficiency in these minority communities.

Iconicity, branding and translation in orthography


Cover of the ‘Complete Russian Songbook’ published in 1917, showing the ‘old’letters <і>, <ҍ>, <ъ> which were reformed later that year

I have recently written a paper on ‘Iconicity, attribution and branding in orthography‘ which will appear in a special issue of Written Language and Literacy later in 2015. You can read the preprint here. Here’s an interesting historic example of iconicity and branding in Russian, which I didn’t use in the paper. It is also an example of the difficulties of literary translation, and how matters of spelling have social meaning which can be difficult to grasp from the outside.

Iconicity is a concept familiar in linguistics largely due to the work of Irving and Gal (e.g. 2000). When a feature of language becomes iconic, its relationship with its users is transformed so that ‘linguistic features that index social groups or activities appear to be iconic representations of them, as if a linguistic feature somehow depicted or displayed a social group’s inherent nature or essence.’(Irvine and Gal 2000: 37). Androutsopoulos (2010) has an excellent example, from the literary writing of Feridun Zaimoglu whose characters are second generation Turkish migrants in Germany. Their nonstandard language, ‘Kanak Sprak’, with its spellings which deviate from standard German, represents iconically the distance between its users and mainstream German society.

Attribution is the process whereby a linguistic feature becomes associated with a particular group of users. Attribution is necessarily a discursive process, and the user (and non-user) groups are likewise discursively constructed.

Branding is the name I give to a process where a specific visual element of the script, for example a letter or a conventional element like a diacritic mark, becomes an emblem of a group of people who use it in their writing practices. There are many examples of such branding, for example in the Spanish counterculture where <k> replaces the standard <c> or <qu> representing the sound /k/ (see Sebba 2007, Screti 2015), or the ‘little hook’ or  haček  whichis seen as emblematic for Slavic languages’ in Austria (Busch 2013: 206).

Branding does not always involve iconicity, although it may do so, if the branding element becomes associated with the non-linguistic behaviour of its users – in other words if it becomes an emblem of something more than the language itself. The haček is emblematic of certain languages, but appears not to be an icon as it is not associated with anything specific outside of language. On the other hand, the use of <k> by Spanish anarchists could be seen as doing the kind of symbolic violence to the conventions of the Spanish language which its users would like to do to the society as a whole. In this case, the <k> ‘brand’ functions as an icon in the sense of Irving and Gal.

So to the example from Russian. The Russian orthography was modernised by Peter the Great in the early 18th century but retained a number of anomalous or redundant letters. One of these was the so-called ‘hard sign’, used to mark an unpalatalised (‘hard’) consonant at the end of a word (this is the last character in the last word in the illustration above, which is the cover of a songbook published in 1917). The ‘soft sign’ was used to mark palatalised (‘soft’) consonants at the end of words. But since all final consonants were either hard or soft, only one of these signs was actually needed. Either the hard or the soft sign was redundant, and in practice it was the hard sign that could be dropped most economically.

IMG_20150407_185815 The ‘hard sign’ (tvyordy znak)

By the late 19th Century it seems that the practice of not writing the hard sign at the end of words had become an emblem of a kind of social enlightenment. In his famous short story Lady with Lapdog, Anton Chekhov uses it as part of a very short description to give a portrait of the character of the wife of the male protagonist (Spoiler alert: she is not the eponymous ‘lady with lapdog’). In one translation,

She was a tall woman with dark eyebrows, erect, dignified, imposing, and, as she said of herself, a “thinker.” She was a great reader, omitted the “hard sign” at the end of words in her letters, and called her husband “Dimitry” instead of Dmitry (translated by Ivy Litvinov)

Here the ‘hard sign’ has become iconic. Rejection of the old, redundant letter can be seen to be linked to the rejection of an old, redundant way of  thinking and a society in need of change. Not writing this letter marks out the writer as socially progressive and an intellectual. So here is a case of iconicity in orthography with <ъ> (hard sign) as the brand.

While the mere  mention of the practice of not writing the hard sign seems to have been enough to trigger an understanding of the social attitudes of Chekhov’s character in his original readers, the same is not true for non-Russian readers. This particular sentence has given translators a fair bit of trouble. Apart from editors of scholarly editions, translators are unlikely to want to explain the subtleties of Russian spelling and its social meaning to readers. So how have they dealt with this? Here’s a selection of other translations:

She was a great reader, preferred the new ‘advanced’ spelling, called her husband by the more formal ‘Dimitry’ and not the familiar ‘Dmitry’ (Penguin books 1964, translation by David Magarshack)

She read a great deal, used phonetic spelling, called her husband, not Dmitri, but Dimitri ( ; translator possibly Constance Garnett)

She read a great deal, used the new orthography, called her husband not Dmitri but Dimitri (translation by Richard Pevear and Larissa Volokhonsky)

A fifth verson, translated by S.S. Koteliansky and published by Penguin in  1941, omits all mention:

She read a great deal, called her husband not Dimitri but Demitri

Of course there is no ‘correct translation’ although that of Ivy Litvinov is most faithful to the Russian original. I do find it a bit sad that S.S. Koteliansky gave up on it altogether!

It turns out that Kate Sutherland has already written a blog about translating Lady with Lapdog, so thanks to her for some of the versions which I didn’t know about.

Historical note: the ‘hard sign’ at the end of words was officially dropped after the Russian Revolution of February 1917.


Androutsopoulos, Jannis (2010). Ideologizing ethnolectal German. In: S. Johnson and T. Milani (eds.), Language Ideologies and Media Discourse. New York: Continuum, 182-202.

Busch, Brigitta (2013). The Career of a Diacritical Sign: Language in Spatial Representations and Representational Spaces. In Pietikainen, Sari and Helen Kelly-Holmes (eds), Multilingualism and the Periphery, pp. 199-221. Oxford University Press.

Irvine, J. T. and Gal S. (2000). ‘Language Ideology and Linguistic Differentiation’, in Kroskrity, P. V. (ed.) (2000). Regimes of Language: Ideologies, Polities and Identities, 35-83. Oxford: James Currey.

Screti, Francesco  (2015) The ideological appropriation of the letter <k> in the Spanish linguistic landscape, Social Semiotics, 25:2, 200-208.

Sebba, Mark (2007). Spelling and Society: The Culture and Politics of orthography around the world. Cambridge University Press.

Post Literacies in the linguistic landscape


What could be more iconic of the British Empire than the red pillar box emblazoned with the monogram of the monarch? Both South Africa and the Irish Republic kept their pillar boxes after independence but the Irish ones were promptly repainted in green, while the remaining South African ones are still red. Meanwhile in both countries they provide a space for displaying official language policy. The South African one (photographed in 2012, but probably unchanged for at least 20 years) reflects the Apartheid-era policy of treating English and Afrikaans as equal official languages in accordance with the 1910 constitution, but ignoring the rest (there are now 11 official languages under the 1996 constitution). The Irish pillar box (photographed in 2006) shows English (the second official language according to the constitution of the Republic) and Irish (the ‘national language and first official language’), in that order (i.e. the second official language first). In both the South African and cases, exactly the same content is given in both official languages and the fonts and letter sizes are identical for both, metaphorically establishing the languages as equally important.

Image Image

Despite the South African constitution declaring 11 languages now to be official, the most recent incarnation of the posting box is just in one language – English. the same, incidentally, is true of the stamps.


Image Image