Dr Catrin Fflûr Huws, University of Aberystwyth, et al - Written evidence (ITS0023)
Dr Rhianedd Jewell, Dr Hanna Binks and Dr Róisín Costello
Interpreting and translation services in the courts House of Lords Public Services Committee Evidence
Summary
This evidence, based on research undertaken at Aberystwyth University and Trinity College, Dublin, focuses on court interpretation in official language settings. Welsh, Irish, and English are used as examples, but many of our findings are generalisable to other languages.
Introduction
This evidence is submitted by Dr Catrin Fflûr Huws, Senior Lecturer in Law at Aberystwyth University, Dr Rhianedd Jewell, Senior Lecturer in Professional Welsh at Aberystwyth University, Dr Hanna Binks, Lecturer in Psychology at Aberystwyth University, and Dr Róisín Costello, Assistant Professor in Law at Trinity College, Dublin. We are an international research team whose work, funded by the British Academy, the Arts and Humanities Research Council, Research Funding Wales, and the Socio-legal Studies Association, focuses on official language interpretation in Wales and the Republic of Ireland. Our findings have wider relevance to other languages used for interpretation.
Articles discussing our work in greater detail may be found in the following sources:
Huws, CF, Jewell, RM & Binks, H 2022, 'A legislative theatre study of simultaneous interpretation in legal proceedings', International Journal of Speech, Language and the Law, vol. 29, no. 1, pp. 37-59. 10.1558/ijsll.20610
Jewell, R, Huws, CF & Binks, H 2022, 'Cyfieithu Cyfiawn? Cyfieithu ar y pryd yn llysoedd Cymru'. in R Jewell & R Williams (eds.), Y Gymraeg a Gweithle'r Gymru Gyfoes . Gwasg Prifysgol Cymru | University of Wales Press.
Huws, C.F. Jewell, R.M, Binks, H, Shafto-Humphries, N, Schwede, L. 2024 Simultaneous interpretation in interpretermediated remote legal proceedings: some observations from a forum theatre study. 57 Comparative Legilinguistics https://doi.org/10.14746/cl.2024.57.2
Róisín Á Costello and Judge Úna Ní Raifeartaigh Irish Language Rights and the Courts of Justice Act 1924 in, editor(s)Niamh Howlin , The Courts of Justice Act 1924, Dublin, Four Courts Press, 2024,
Róisín Aine Costello Irish Language Rights under the 1922 Constitution: Symbolism and Substance in, editor(s)Laura Cahillane and Dónal Coffey , The 1922 Constitution, Dublin, Palgrave Macmillan, 2023,
Róisín Aine Costello Lost in Translation: Courtroom Interpretation and the Right to a Fair Trial in DPP v HM and BO [2021] IECA 315 68 Irish Jurist (2022) 130
Huws, C.F. Jewell, R.M, Binks, H, (2023) Remote Hearings Post-Covid Remote-Hearings-Post-Covid-(2)-1.pdf (aber.ac.uk)
1) To what extent do the current interpreting and translation services provided in courts meet the needs of those involved in proceedings, including defendants, witnesses, prosecutors and legal professionals?
The particular focus of our research has been on the use of interpretation services in Wales, and this very much represents the gold standard in terms of quality interpretation services. The training accreditation of interpreters by Cymdeithas Cyfieithwyr Cymru Homepage (cyfieithwyr.cymru) with specific additional competencies required for court interpretation means that the Welsh model is very much the model for other languages, in that a high standard of accuracy is ensured, as well as the professionalisation of the service. By comparison, there is no centralised training or accreditation of interpreters for court proceedings in Ireland with the result that it can be challenging to locate interpreters, and to confirm their ability to interpret legal proceedings. The fact that HMCTS’s Welsh Language Unit also have the responsibility of sourcing interpreters and ensuring their quality. Even within this context there are a number of aspects to which attention may be drawn:
- Our research indicates that interpreters are given no information in advance about the nature of the case, in order to enable them to prepare adequately in terms of the common vocabulary. Even within a criminal context, the vocabulary required for offences involving violence are likely to be different from offences involving theft.
- Our research also demonstrates that, commonly, only one interpreter is employed in a case. This means that the individual may have to interpret for significantly longer periods of time than is recommended, and may be required to interpret for multiple parties e.g. questioner and respondent, or for multiple witnesses. This imposes a significant cognitive burden on the interpreter, but also makes the process confusing for listeners who are unable to separate the speakers when all are interpreted for by the same individual. The result may be incorrect recollection of the testimony of particular speakers, or inaccurate recall of the impression of confidence, trustworthiness, or lack of same, imputed by the listener to a particular speaker.
- It is also not uncommon for interpreters to be asked to interpret in two different directions in the same case, a request which goes against the recommended practice of the profession due to the additional cognitive strain placed on the interpreter and the increased potential for error.
- We have found in our research that appropriate planning is needed in order to appropriately locate the interpreter within the court room. The interpreter needs to see those for whom he/she is interpreting, without obscuring the speaker from the listeners and/or distracting the listener by drawing their focus from the speaker to the interpreter’s body language or expressions thus interfering with the reception of the original speech. Ideally an interpreter should be positioned in a place that allows them to see the speaker, and thus ensure the tone, register and emotion of their interpretation mimics that of the speaker. However, the interpreter should, ideally, be out of view of the listeners whose attention should remain on the speaker themselves.
- We have also identified a need to explain the role of the interpreter to listeners, who may not be aware of the interpreter’s role (i.e. that they are not representing the witness but are literally relaying their testimony as it is heard), and the possible differences between the original speech and the interpreter-mediated evidence. In Welsh for example, the form of agreement must agree with the tense of the verb, thus there are different words for ‘yes I was’ ‘yes it is’ ‘yes they did’ and thus the interpretation into English as ‘yes’ may seem significantly shorter than the original. A similar issue arises in Irish where positive and negative responses to questions cannot be translated into ‘yes’ and ‘no’ but must be associated with an associated verb. The absence of “direct” and single word answers can be viewed by listeners as evasive. Proper direction should thus be given to explain the possibility of such divergences, and the weight they should be given by any listener.
- Relatedly, a number of linguistic features also need to be considered when interpreting including interpreting or eliding pauses and hesitations in speech, and interpreting cultural nuances. Accurate interpretation needs to incorporate these features in order to ensure that the fairness of the process is maximised and a true representation of the fluency, hesitation, perspective and character of the evidence and the speaker is portrayed.
- In providing explanations about interpretation and the role of the interpreter, the role of the interpreter must be explained both to the speaker as the listeners as being only to interpret the questions, answers and statements of the speaker and questioners. The interpreter is not a source of explanations of the technical terms used by parties in court. The judge, or barrister must retain responsibility for translation of the latter and this should be explained to the listeners and the speaker so that if a speaker asks what a legal term means, that question is relayed through the interpreter to the court, and the answer in turn relayed through the interpreter to the speaker. Our research suggests that listeners need to be better prepared for the experience of using simultaneous interpretation which may be entirely new to many. Interpreters are often only able to provide a brief explanation to listeners about the practicalities of using the headphones (e.g. adjusting volume, finding the correct channel) but listeners should be briefed about the need to switch back and forth between live testimony in one language and an interpretation provided via the headphones and about the delay between the original speech and the interpretation provided. Furthermore, the opportunity to practise use of the headphones would avoid confusion at the beginning of a case and missed details as listeners settle into the process.
- It is important to note that this preparation should be extended to include both Welsh and Irish speakers, for although they do not necessarily need to make use of the interpretation, their understanding of the process can impact the way in which they interact with the interpreter, for example, by adjusting their speed of delivery, allowing for the time delay between the speech and interpretation, and appreciating the significance of interpretation to a bilingual case.
- The court also needs to accommodate the interpretation. Even though the simultaneous model of interpretation used in Wales means that the interpretation is almost concurrent with the original language, questions and interruptions need to consider that the interpreter must incorporate this, and that listeners may also need to switch between listening via headphones and listening directly. The court should adopt an approach for asking questions from the bench and dealing with interjections from other legal representatives by noting how and when those may be made to both the parties and the interpreter so that a pause for such interjections can be anticipated.
- In terms of their own wellbeing, there is also a need to ensure that interpreters are adequately prepared for information that may be distressing. Unlike other participants in the process, the interpreter is required to present, or be presented with, harrowing information in the first person and to relay it without alteration in the first person. Because of this feature of their role, it is more difficult for the interpreter to distance themselves from the circumstances of the case, and more likely that they may experience distress as a result of the testimony they interpret.
I. How have interpreting and translation services changed in recent years?
Several more demands have been made on interpretation services over recent years. Firstly, the increased use of remote hearings during the Covid era, and continuing in several areas of court and tribunal work makes different and additional demands on interpreters, as reading visual and contextual cues is more difficult online. Secondly, there are increased demands on interpretation because there is greater linguistic diversity (i.e, more language communities who do not speak the State language as their primary language), and a greater likelihood that people may lack fluency in English or Welsh or Irish sufficient for the formal processes of a court. Another change is the increased support for BSL interpretation, and the need to provide BSL interpretation for D/deaf witnesses, jurors, and legal professionals.
Cost cutting and the backlog in cases is likely to have an impact as well, as courts may wish to encourage more participants to use English or Welsh to speed up the processes, and therefore, there may be a greater reluctance to ensure the availability of appropriate interpreters, especially, as in the Welsh context, the participants in our research have noticed an increased demand for interpretation from English to Welsh as well as from Welsh to English. This issue is exacerbated by there having been a reduction in the number of qualified interpreters, and the working conditions (adequacy of remuneration, unreasonable expectations in terms of the time spent translating, and the lack of support for dealing with distressing evidence) have been suggested as possible causes for this.
2) What are the key issues in the provision of interpreting and translation services and what impact do they have on the running of the courts, public trust, interpreters and translators.
Ensuring that interpreters are properly trained and accredited for the specific role of interpretation, and the specific task of interpretation within a court setting.
Ensuring that working conditions are appropriate in order to ensure that interpreters are able to prepare adequately, and that an undue cognitive and physical burdens are not imposed upon them by being asked to interpret for lengthy periods of time without a break, or without appropriate equipment, or to interpret in more than one linguistic direction.
Ensure that that interpreters are appropriately remunerated for their work, including the payment of wasted costs when they attend at court and the hearing is cancelled or delayed.
Ensuring that interpretation and interpreters are accommodated appropriately in the court process with an awareness of their requirements in terms of space, ability to see the person for whom they are interpreting and interjections from other parties in court, or the bench.
Ensuring that listeners understand the role of the interpreter, and how the original language is modified because of the requirements of translation such that an answer may be longer or shorter - and the interpretation is still correct.
Ensuring that listeners understand the nature of interpretation in practice, and are adequately prepared for the delay between speech and its interpretation and the potential overlap between the original testimony and the interpretation.
I. Is there data on the number of miscarriages of justice due to ITS error?
The difficulty here is that identifying that a miscarriage of justice has occurred is challenging, because the affected party does not understand the interpreted evidence, and other participants in the court process may not understand the original evidence. Nevertheless, the Irish case of DPP v HM and BO [2021] IECA 315 is an example of a criminal conviction which was subsequently rendered unsafe because of flawed interpretation. Further examples from the wider literature arise from the cultural differences between language – an example from the Welsh/English context of our research is that the word ‘trais’ in Welsh is used to denote both the tort of battery, and the crime of rape, and an interpreter might have to be mindful to use a phrase such as ‘trais rywiol’ in order to denote the latter, but not the former.
Our research indicates however, that wider miscarriages of justice may occur because of listeners’ attitudes to users of interpretation. The use of an interpreter may draw attention to immigrant status, and users may be perceived as being wilfully obstructive, particularly if they are perceived as being able to understand and communicate through English. The inconvenience of having to switch between interpretation headsets and the original speaker, or having to alternate between the evidence being given in the original language and the interpretation, may also cause the listeners to judge the witness negatively including by viewing them as self-indulgent or un-cooperative.
As a result of these factors, we would suggest that any available data on miscarriages of justice is likely to underestimate the true picture in respect of the impact of interpretation on the outcomes of proceedings.
3) Are the required qualifications and experience of interpreting and translation services in the courts consistent?
We do not believe that they are. The training and accreditation system for Welsh in Wales is standardised and a consistent process, and therefore a good standard of interpretation can be obtained. For other languages, there is less consistency, and also a lack of awareness of linguistic variations between, for example, the Spanish spoken in South America, and the Spanish spoken in Spain, and therefore a language-proficient interpreter may not be able to translate cultural nuances or slang accurately.
I. Are the recommended requirements standardised across all governing bodies, contractors, and institutions?
Our experience is predominantly the Welsh context, where there is an accreditation system for interpreters, and additional accreditation for court interpreters. This is the standard that we might expect from any interpreter used within a legal setting. However, this standard is higher than for other languages, where interpreters can self-accredit, and their linguistic competence for the role is not assessed.
A potential source of inconsistency arises from the situation where a judge evaluates a person’s linguistic capacity to be sufficient for them to be able to proceed without interpreter. Even though a person is not fluent in the State language, the trial is deemed fair based on the judge’s evaluation of their linguistic competence. Furthermore, even if an interpretation is provided, a trial may be deemed to be fair, even if there were flaws in the interpretation. The inconsistency here is that the monolingual judge is not able to evaluate whether a person’s ability to articulate themselves would be greater if they spoke a language fluently, and is not able to evaluate the extent to which the trial is unfair if the interpretation is inaccurate, because they do not understand how the interpreter interpreted the information for the party/witness, or how that person’s evidence was transposed for the benefit of the court.
However, with BSL interpretation, there is a different consideration, namely the equality of treatment (in addition to the fairness of the trial), and the additional expectation of ensuring that no discrimination occurs.
The situation with Welsh is again different, in that Welsh has official status within the court process, and may therefore be used alongside English. This has two consequences. Firstly, there is the possibility that listeners will be able to understand both the original language and the interpretation. However, it also means that the court user does not have to demonstrate that the trial would be unfair if it were conducted in English. The right to use Welsh is therefore a freedom of expression right, as much as it is a fair trial right.
The different needs of interpretation, and the different origins of the right mean that the approach to interpretation is conceptualised differently, and thus prioritised differently.
II. Are the current requirements fit for purpose?
The current requirements are not necessarily fit for purpose, as the expectation is that the interpretation should be sufficient, in the judge’s opinion, to ensure that the defendant understands the process sufficiently. Firstly, if the judge cannot understand the language of the witness or the nature and quality of the interpretation, he/she is not well-equipped to be able to evaluate whether the witness has a sufficient understanding of what is being said, and therefore they may be unaware of the fact that the witness is receiving information that is different from the information that was being communicated. In our study two interpreters were asked to interpret the phrase ‘Wnaethoch chi weld y diffynnydd yn cymryd yr eitem?’ One interpreter interpreted this as ‘Did you see the defendant take the item?’ while the other interpret it as ‘Did you see the defendant steal the item?’ Both are accurate translations, but the more specific accusation of ‘steal’ may lead the witness to reply ‘no’ even though they may have seen the witness ‘take’ the item. While variations are inevitable in communication, what is lacking is the awareness of the potential for these variations, and their implications.
The question of whether a person has anything further to say, may be interpreted as a request for any further information to be communicated, but may equally interpreted as an opportunity to present a plea in mitigation – something that may be detrimental if any judgment regarding guilt has yet to be made. It is also worth noting that legal language in itself is often difficult for parties to understand in contexts where no interpreter is required. This makes a subjective assessment by a judge with familiarity both with the technical language being used, as well as the language of the court, a challenging standard to apply consistently.
Regard must be had to ensuring that:
4) What quality assurance and complaints procedures are in place in relation to interpreting and translation services in the courts?
I. How easy is it for people to report or submit a complaint?
See above, the information regarding the potential for participants to be unaware of problems, errors, or inaccuracies.
II. What data exists on the number and types of complaints made?
Not known.
5) How easy is it to recruit and retain skilled interpreters and translators to work in the courts?
Our research indicates that there are a number of gaps, even within the Welsh context. There are fewer people training as interpreters, and particularly as interpreters for the courts. Our research participants indicated that the role does not pay enough, and that they are not paid if they attend at court, but the hearing has been cancelled or postponed. The burden of the role, combined with the inadequacy of the compensation (including wasted travel costs) means that the role neither attracts nor enables a sufficient number of interpreters to be retained, and as discussed above, there has been a reduction in the number of Welsh/English interpreters who are accredited to provide simultaneous interpretation in the courts between 2020 and 2024.
A further dimension of our research is the gender-matching of witnesses and interpreters. In Wales, and elsewhere, interpreters are predominantly female, while participants in the legal process are predominantly male. Where simultaneous interpretation is used, the witness’s voice is replaced by the interpreter. We have undertaken some work on exploring the situation where a female witness is translated by a male interpreter, but the statistics on crime indicate that a higher proportion of those accused of a criminal offence are male, and further work is needed on gender-matching male witnesses with female interpreters to determine what impacts, if any, the allocation of gender mixed interpretation and speaker pairs has on the reception of the evidence – and on potential distress for the interpreter, in particular in cases involving sex offences of gendered violence. The impact of unmatched gender may be significant, most obviously in these situations, but also in situations where strength or the proportionality of measures taken in self -defence may be in issue.
The question of recruitment therefore also raises questions concerning the recruitment of interpreters who are able to speak authentically for the witnesses.
The primary opportunity is the high standard already extant within the legal system in that the methods for training, accrediting, and engaging an interpreter ensure its consistent quality, and a clear process, and this is a model that could be adopted more widely.
Barriers and pitfalls exist in the form of a lack of awareness of variation in meaning across languages and cultures. There is a lack of awareness that information does not transpose across language and culture as neatly as might be assumed by those who are monolingual, although the differences between the English of the United Kingdom and the English of the United States gives some flavour of this – an American might walk down the street in their vest and pants with fewer eyebrows raised than the English person.
The potential for prejudice against users of interpretation is a further barrier, if the speaker is cast as the outsider, or as the person who is being wilfully obstructive. In such cases the listeners may already be prejudiced against them. This is compounded if the interpreter does not have the requisite command of the formal vocabulary of a court, or of the need to interpret the evidence accurately, and at speed, and where the court lacks awareness of the need to ensure that the interpretation provided is of a sufficiently high standard. Achieving this requires appropriate remuneration and recognition of the professional interpreter’s expertise.
6) What is the potential role of new technology (such as artificial intelligence, machine translation and the digitisation of court proceedings) in the future of interpreting or translation services in the courts?
This is something that must be approached with great caution. The research into interpretation indicates that the act of interpretation is not merely translation of the words but also of tone, meter, hesitation and emotion, and interpreters must also be conscious of the cultural connotations of words in different languages, the breadth and narrowness of meaning, politeness markers, and the indications of agreement or disagreement across languages. It is not clear to us that artificial intelligence will be capable of replicating the professional skills of an interpreter in a manner compatible with a fair trial.
It should also be noted that the training models used for artificial intelligence are ill-equipped to develop language libraries for minority languages which have reduced online presences. The impacts of this have been demonstrated in existing tools such as Google Translate which provide far less accurate renderings into minority languages such as Irish, for example. Such tools break down entirely when asked to translate dialects within minority languages whose spelling or word choice departs from standardised forms or when faced with a variety of linguistic registers that deviate from the standard written form. This is considered further below.
Our own research contains an example of a defendant accused of theft. We experimented with the same evidence being interpreted by several interpreters concurrently. In this context, interpreting the question put to the speaker (as some of the interpreters did) as ‘did you see the defendant steal the item?’ was correct, but the use of the term ‘steal’ is more accusatory than the construction adopted by other interpreters namely ‘did you see the defendant take the item?’ A human interpreter is able to identify the implicit bias inherent in the former question, where as AI translation may persist with a more accusatory term unless it is corrected appropriately.
No. Interpretation is a specialised skill, and there is a need to understand nuance and cultural connotations as well as the words used.
Not known.
Not known.
7) What is the current capability and accuracy of market leading artificial intelligence and machine translation tools in relation to ITS?
Use of translation tools (as opposed to spoken interpretation tools) reveals a number of problems across languages. Welsh/English translation tools often demonstrates the features of
-confused word order – ‘a dress blue’ rather than a blue dress
-a failure to match agreement with question tense.
-inaccurate transposition of meaning
-inaccurate transposition of words with multiple meanings e.g. set. Artificial intelligence tends to supply the commonest interpretation – ‘a group’, rather than alternative meanings such as ‘to coagulate.’
- inaccurate translation of figurative language
- inability to distinguish between different linguistic registers which could impact the perception of an individual speaker or of the relationship between speakers
Interpretation tools are likely to compound these problems as they may fail to transpose accent and dialect accurately and may make significant errors of interpretation through misunderstanding the original voice. Experiences of using subtitling software in our research indicate that such technologies routinely failed to transpose legal terminology and terms of art. Thus, phrases such as ‘a lease’ were transposed as ‘Elise.’
The software also fails to transpose terminology whose origins are not English reducing Welsh or Irish words or names to phonetic strings of letters that bore no relation to the spoken terms.
It must also be remembered that court users, particularly where they are bilingual, may use loan words and calques, and this again may lead to a failure of translation.
Caution must also be had to the accuracy of negative indicators – whether something ‘is’ or ‘is not’ is something that may be missed by AI ITS software, but it is also to be borne in mind that this may be rendered differently in different languages. – the negative clause may be at the end of the sentence in one language although it is at the beginning in another. In monolingual English speech for example is possible that a microphone may fail to pick up a slurred ‘It isn’’ and transpose it as ‘it is.’ However, when two languages locate the negative indicator in a different place, the AI translation may fail to transpose the word order correctly. For example, a phrase such as ‘She is not.’ places the negative indicator at the end of the sentence, but transposed into Welsh, the negative indicator would be at the beginning ‘Nid yw hi’ and the change of word order changes the meaning ‘Dyw hi ddim’ (‘She does not’ rather than ‘She was not.’)
A failure of translation may also arise if a sentence contains both positive and negative indicators particularly in a complex sentence e.g. ‘Is it true to say that you did not see Alice hit Bob?’ A failure of translation in verb agreement may mistranslate this sentence – either giving the indication that the statement is true, or contradicting the questioner’s assertion, as might be conveyed by ‘yes I did see Alice hit Bob.’
However, it is also possible that a negative sentence in one language may be parsed as a positive sentence in another. The sentence ‘Did you not see that incident?’ implies that the respondent did in fact see the incident, and has perhaps forgotten this information or that the respondent’s denial of seeing the incident is questionable. However, rendered into Welsh ‘wnaethoch chi ddim gweld y digwyddiad?’ communicates only that the respondent did not see the incident – ‘You did not see the incident?’ rather than ‘I am doubtful of your assertion that you did not see the incident.’
How does this vary between languages (e.g. low resource languages or languages with relatively few written language samples), interpreting (speech to text) and translation (text to text)?
Not known. However, with dominant languages within the relevant jurisdiction, or widely learned languages, there is a greater likelihood that discrepancies or inaccuracies may be detected, or suspected. There is a greater danger with languages where non-speakers have no knowledge of their structure (including community languages, minoritised languages) that errors may be undetected. For example, French, German, and Spanish are all widely-acquired languages within the United Kingdom, and therefore a non-speaker may be aware that in French for example a noun is gendered and that the adjective follows the noun. However, there may be less awareness among non-speakers of Welsh, Irish, Gaelic and Ulster Scots of the linguistic differences between these languages and English.
27 September 2024