Language testing of asylum claimants: a flawed approach
August 7, 2014 — Comment
Written by Aisha Maniar
Following a critical Supreme Court judgment on the Home Office’s use of controversial language analysis tests to determine the nationality of asylum seekers, Aisha Maniar asks: why does the government insist on using these tests?
Language is a crucial element of the identity of each and every one of us, and a marker of social and cultural inclusion. Over the past twenty years, it has increasingly been used by western states as a means of determining political and bureaucratic identity – nationality – and consequently to reject the claims of undocumented asylum seekers on the basis that the language they speak is not that of their claimed country of origin. And where a language analysis places the claimant’s linguistic origin elsewhere than the country from which they are seeking asylum, not only is the asylum claim rejected, but removal is, in many cases, to the wrong country, which the language analysis deems them to come from.
Decisions made by the immigration authorities in asylum cases can be a matter of life and death. Can language, which has only a tenuous link to nationality – a stateless person has the former but not the latter – provide a reliable tool in determining the origin of asylum seekers?
Language analysis testing in determination of asylum claims is a relatively new area of linguistics, first used in the mid-1990s in Scandinavian countries, and now used elsewhere in Europe as well as in Australia, New Zealand and Canada. Some countries have their own government department providing the service, but most, including the UK, outsource it to small commercial companies. There is no set method for such testing, but in the UK, a typical language analysis involves a taped telephone interview of the asylum seeker by a language analyst, lasting around twenty to thirty minutes, during which the analyst works through a standard list of questions and carries out a linguistic analysis and ‘knowledge assessment’ of the interviewee’s knowledge and experience of their stated country/region of origin. With a linguist, the analyst then assesses, to one of five degrees of certainty, whether the interviewee originates in the claimed country. The report can be produced within fifteen minutes of the interview, and is often treated as conclusive.
Following a pilot on several nationalities – including Iraqis, Afghans and Tamils who at the time made up a large number of applicants – the method has been used in the UK for the past decade, but has been subject to much criticism.
The UK Home Office uses the linguistic analysis services of Swedish company Skandinavisk Språkanalys AB, or Sprakab, a private commercial company which works almost exclusively for governments and in the public sector, providing similar services to the governments of Canada, Australia and the Netherlands, among others. Sprakab says that it has carried out over 40,000 tests in its fourteen-year history and currently provides 4,000 tests annually worldwide. Given that the ultimate purpose of such testing is to help determine entitlement to protection under international law, the use of a private profit-based corporation is a key criticism. The company’s insistence on the anonymity of its linguists and analysts ‘for their own safety’ has also been criticised for hampering transparency and understanding of the decision-making process.
According to Sprakab’s website: ‘The results are very reliable and […] provide a clear picture of an individual’s language background’, yet before a court, the company manager ‘agreed that linguistic analysis could not determine a nationality’; ‘an individual’s language background’ is not an indicator of their nationality.
Unqualified analysts, unreliable results
There is confusion over the role and qualifications of analysts and linguists employed by Sprakab. Home Office guidance states that ‘Sprakab analysts have linguistics backgrounds and experience in dialectology’ and linguists ‘have the equivalent of a master’s degree in either linguistics or phonetics’. In many cases, this has proved not to be true. Very often the linguist is qualified in a related field (such as having a language degree) and the analyst’s main qualification is the ability to speak a language. Many analysts are employed to analyse the mother-tongue competency of asylum seekers in languages which they themselves do not have as a first language.
Additionally, Sprakab’s own standards lack the scientific accuracy one should expect of a forensic linguistic procedure. In over two decades of language analysis testing by Sprakab and other similar companies, there is no empirical evidence that language analysis bears any relevance to the determination of the citizenship or nationality of any individual. The general consensus among professional linguists and experts who provide testimony in court is that language analysis testing is not fit for purpose. Dr Derek Nurse, an expert on the minority Somali Bajuni community, describes the Sprakab analyses and conclusions he has examined as ‘brief, careless, lacking in supporting evidence, and unconvincing’. Other experts have called it ‘unwise‘ to rely on Sprakab reports to make legal decisions, with one submission to parliament stating that the process is‘deeply flawed, unprofessional, flies in the face of expert opinion, and infringes the Human Rights of large numbers of vulnerable individuals’.
A large number of basic linguistic errors that undermine Sprakab’s purported experience and expertise have been pointed out. For example, contrary to the assumptions behind the tests, monolingualism is not the norm in many societies. In addition, the method does not take code-switching into account, a phenomenon in which a bilingual or multilingual speaker alternates between two or more languages, or language varieties, in a single conversation; it is often done unconsciously. It is also common for a speaker of a minority language to speak automatically in the dominant/majority language of a country/region when speaking to people outside of their own community. Nonetheless, a taped recording of less than thirty minutes is expected to provide certainty in assessing an individual’s nationality.
Other language issues that are ignored are more specific to refugee communities. For example, years and decades of war and upheaval can have a massive impact on the language patterns of a whole region – whole groups of people are dislocated, within and outside borders – and new social interactions are born. In addition, the reality of refugee diasporas is that generations now grow up and live permanently displaced in refugee camps in bordering countries. This dislocation inevitably affects the language of whole communities. Trauma can also have an impact on linguistic ability.
In the case of the Somali Bajunis, who are almost always assessed as being Kenyan Swahili speakers, experts observed other issues such as, for example, the fact that the average educational level of members of this island community may not equip them to answer ‘knowledge assessment’ questions about Somali politics and history. Many Bajunis are not familiar with using telephones; in at least one case, it was the first time the interviewee had ever used a telephone. Several experts have also reported that applicants seem ‘afraid and confused‘ by the impersonal telephone interview approach.
Language analysis cannot determine nationality
Leaving aside the methodology applied, the main criticism, particularly among professional linguists, is that such tests are of limited value, and in particular ‘cannot be used reliably to determine national origin, nationality or citizenship’, as these are ‘political or bureaucratic characteristics, which have no necessary connection to language’. In recognition of the growing use (and misuse) of language tests for determination of origin, in 2004, a group of international linguists produced Guidelines for the Use of Language Analysis in Relation to Questions of National Origin in Refugee Cases, mainly for the use of governments and decision-makers, to emphasise the limitations of language testing and to deal with the difficulties. They point out that language tests can disclose the region of socialisation, but conclude that ‘language analysis should be used with considerable caution in addressing questions of national origin, nationality or citizenship’. The Guidelines, which offer clear and practical instructions for the use and limits of language testing, have since been endorsed by many professional linguistic organisations, including the British Association for Applied Linguistics (BAAL) and the International Association of Forensic Linguists (IAFL). But neither the Home Office nor Sprakab follow the Guidelines (they both have their own), and a decade on from their introduction, persisting problems show that they have not had quite the intended effect on the government and non-linguist decision-makers.
‘Anti-asylum seeker agenda’
Refugee advocates argue that language analysis testing is a ‘way of getting around the legal barriers‘ to send people back to war-torn countries. The UK has an obligation under European human rights and international refugee law to consider asylum claims and to provide protection to those fleeing persecution and violence. According to Dr Diana Eades, the use of language analysis testing falls within a ‘wider anti-asylum seeker agenda’.
One needs to look no further than the politicised language of the Home Office guidance: ‘The purpose of language analysis … is to assist in identifying an individual’s true place of origin where it is in doubt [and] deter fraudulent claims based on false claims of origin for actual or perceived benefit’. Testing is conducted ‘either because particular doubts are held’ or ‘because an inadequately documented individual claims to be a nationality/national origin that may be targeted under an exemption to the Equality Act 2010’. The scales are weighted against the applicant before the interview has even taken place.
As of February 2013, an exemption has been applied to all asylum applicants claiming Syrian, Kuwaiti or Palestinian nationality, so if inadequately documented, they are automatically tested, and not just where there is ‘doubt’. The rationale for this is circular: language analysis testing, in particular between October 2011 and May 2012 (done, of course, by Sprakab) showed that ‘abuse was particularly apparent for [these] three claimed nationalities or national origins’.
Taken in the context of the UK’s response to Syria, the world’s largest refugee crisis in recent years, the purpose may be simply to deter applicants from that country. The UN estimates that approximately 2.7 million people have fled the war in Syria in the past few years, with the vast majority seeking refuge in neighbouring countries. In Britain, only fifty people have been accepted as refugees this year under the government’s Vulnerable Persons Relocation package, which was adopted by the government only after a big public campaign. In the case of Somalis, widely targeted by such testing, it has been claimed that ‘applicants may have undergone language analysis simply because they claimed to be Somali, rather than because of doubts about them as individuals’. Whole groups of people are targeted rather than considering individual applications on their own merits, as required under the Refugee Convention.
Over the past decade, legal challenges, using expert criticisms frequently based on the Guidelines, have resulted in some improvements to language analysis testing in the UK, and a large number of Home Office decisions made on the basis of such analyses are overturned at appeal, at great cost to the Home Office. But legal decisions have left the basic structure of language testing intact. In a test case brought by a Somali Bajuni woman, the panel of judges at the Upper Tribunal gave guidance on how Sprakab reports are to be used, endorsing them subject to certain safeguards. The case, RB (Somalia), was subsequently upheld in all main respects by the Court of Appeal in 2012. It controversially ruled that where the analyst expressed ‘a high degree of certainty’, little further evidence would be needed to establish the interviewee’s nationality – giving the green light to the Home Office to present language analysis as ‘conclusive’ evidence that asylum seekers were not from their claimed country. The approach was compounded by analysts expressing their opinions as to the general credibility of the interviewee, based on answers to ‘knowledge’ questions. Both these abuses were nailed by the Supreme Court in Secretary of State for Home Department (Appellant) v MN and KY (Respondents) (Scotland). MN and KY both claimed to be from the minority Benadiri clan, which in KY’s case at least was sufficient to make her a refugee, and in both cases, Sprakab testing was relied on to reject their asylum applications. The analysts claimed they were Kenyans and not Somalis. The anonymous Sprakab analyst assessing KY had not visited Somalia since 1990 (eighteen years before the test took place), was not an analyst for her language and, while university-educated, had no background in linguistics or languages. Adding his own value judgement to her ‘knowledge assessment’, he stated ‘Her knowledge sounds rehearsed’.
The Supreme Court endorsed the Scottish Court of Session’s ruling that the Sprakab analysis could not be relied on, adding that Sprakab analysts should not go beyond their scope as experts by offering their opinion on an asylum seeker’s credibility – ‘Expert witnesses should never act or appear to act as advocates’ – and that if anonymity is to be granted to analysts, there must be safeguards for the context within which this applies.
In many ways, the Supreme Court ruling simply endorsed the rules set out in the Guidelines as well as the Home Office’s own guidance – such as not relying on language analysis testing alone – and those of Sprakab. But unlike the experts, the court did not question the purpose of language analysis or call for its use to be abandoned.
Two plus two equals five
There are many reasons why asylum seekers arrive without documentation or adequate proof of their identity. It does not make the work of those tasked with processing their claim any easier. Nonetheless, Britain has a history of policies of deterrence of asylum seekers, and its record as one of the biggest arms manufacturers and exporters in the world, and its support for warmongers and war criminals which make it (albeit indirectly) a major producer of refugees, do not allay concerns that language analysis, along with other quasi-scientific approaches, are designed for deterrence rather than for scientific accuracy. Untested approaches such as language analysis testing will inevitably be the norm, in spite of the costs involved. A redacted 2012 study on the impacts and economic costs and benefits of language analysis testing offers no clear conclusions on the benefits of this approach, but concedes that asylum applications have been steadily falling in recent years.
There may be some benefit in language testing, but ultimately the decision made is political. The reasons for which individuals seek asylum and for which it is granted or denied are also political. The very premise for this approach is couched in political terms. As Diana Eades states, ‘the ultimate problem here is not a linguistic one. Linguists are not responsible for, nor qualified to, provide a solution to this problem, namely the validation of nationality claims’. Language testing may be an answer, but not if validating the nationality of asylum seekers is the question.
 See Home Office asylum process guidance: ‘Language analysis’.  See discussion in Secretary of State for the Home Department v MN and KY  UKSC 30.  For this and many other criticisms see the decision by the Inner House of the Court of Session in MN and KY’s case.  See Secretary of State v MN and KY above.  See note 1 above.  See in particular the report by D Nurse, Overview of Sprakab linguistic analyses of Bajuni refugee claims 2004-2010 (2010, modified 2013).  Cited by the Inner House of the Court of Session in MN and KY’s case, note 3 above.  From the Guidelines for the use of language analysis in relation to questions of national origin in refugee cases. See an earlier IRR News article, ‘The use and abuse of language analysis in asylum cases’ (21 July 2005).  An exemption to the Equality Act 2010, granted by ministerial authorisation, allows immigration officers to discriminate against groups defined by nationality, national or ethnic origin for the purposes of immigration control, by, for example, subjecting members of such groups to more intensive examination or carrying out additional tests.  See Sarah Craig, ‘The use of language analysis in asylum decision-making in the UK – a discussion’ in Journal of Immigration, Asylum and Nationality Law (2012) 255.  For other quasi-scientific approaches to assessment of origin see ‘Bad science?’, IRR News (24 September 2009). See also Craig, note 10 above.
The Institute of Race Relations is precluded from expressing a corporate view: any opinions expressed are therefore those of the authors.