Speech-language assessment in a linguistically diverse setting : Preliminary exploration of the possible impact of informal ‘ solutions ’ within the South African context

Speech-language therapists (SLTs) working in the context of cultural and linguistic diversity face considerable challenges in providing equitable services to all clients. This is complicated by the fact that the majority ofSLTs in South Africa are English or Afrikaans speakers, while the majority of the population have a home language other than English/Afrikaans. Consequently, SLTs are often forced to call on untrained personnel to act as interpreters or translators, and to utilise informally translated materials in the assessment and management of clients with communication impairments. However, variations in translation have the potential to considerably alter intervention plans. This study explored whether the linguistic complexity conveyed in translation of the Western Aphasia Battery (WAB) test changed when translated from English to isiZulu by five different first-language IsiZulu speakers. A qualitative comparative research design was adopted and results were analysed using comparative data analysis. Results revealed notable differences in the translations, with most differences relating to vocabulary and semantics. This finding holds clinical implications for the use of informal translators as well as for the utilisation of translated material in the provision of speech-language therapy services in multilingual contexts. This study highlights the need for cautious use of translators and/or translated materials that are not appropriately and systematically adapted for local usage. Further recommendations include a call for intensified efforts in the transformation of the profession within the country, specifically by attracting greater numbers of students who are fluent in African languages.

settings, nurses, cleaners, general assistants, and family members are frequently called upon to translate or interpret for patients who do not speak and/or understand English.This observation is confirmed by Elderkin-Thompson, Silver and Waitzkin (2001), who report that bilingual individuals with no formal training as translators frequently translate for patients who are non-English speaking.Much meaningful information may be lost since the person translating may not have an understanding of terms used, may use the notion of direct translation or translate the message differently, thus creating confusion and changing the meaning of the item to be translated (Elderkin-Thompson et al., 2001).Previous research has also shown that ad hoc interpreters can commit many errors of interpretation such as omitting important information as a result of limited vocabulary during translation (Flores, Laws, Mayo, Zuckerman, Abreu, Medina & Hardt, 2003).It can thus be anticipated that results obtained from mediated SLT assessments may be compromised, leading to incorrect diagnoses, particularly where untrained translators or interpreters are used.In speech-language therapy this challenge is more complex because how communication occurs is a crucial aspect, not just what is communicated as in most other healthcare setting scenarios.
Owing to the limited focused research on this challenge within the South African context, the ad hoc arrangements made during assessment and management of clients from linguistically and culturally diverse populations creates translation and adaptation difficulties which arguably influence the normative interpretation of assessment measures, and consequently management plans.Despite the availability of frameworks and guidelines for adapting or translating test measures (e.g.Geisinger, 1994), as well as published guidelines for working with interpreters, anecdotal evidence suggests that these are infrequently followed by SLTs in the South African context; this is believed to have a negative impact on speech-language therapy service provision as a whole.
One of the populations with whom SLTs commonly work is the neurologically impaired adult population.Van Schoor, van Niekerk and Grobbelaar (2001) reported that South Africa is rated among the highest in the world for casualty admission rates secondary to motor vehicle accidents.Global statistics reveal that South Africa has among the highest HIV/AIDS infection rates in the world, with an estimated 410 000 new infections in 2010 (Statistics South Africa, 2010).Mochan, Modi and Modi (2003) show a relationship between HIV infection and cerebrovascular accidents as a result of vascular abnormalities, coagulation disorders and cardio-embolic disease.This high incidence of stroke concurs with the findings of the Southern Africa Stroke Prevalence Initiative (SASPI) Project Team (2004), who report an incidence of cerebrovascular accidents of 243 per 100 000 people.As a result of the above-mentioned factors, it can be assumed that in South Africa the number of patients with speech, language and/or cognitivelinguistic impairments as a consequence of neurological impairment is high, resulting in an increased need for speech-language therapy assessment and intervention.
One commonly used formal English assessment tool for speechlanguage and cognitive-linguistic assessment of the neurologically impaired adult population is the Western Aphasia Battery (WAB) (Kertesz, 1982).This tool includes subtests relating to content, fluency, auditory comprehension, repetition, naming, reading, writing and calculation (Kertesz, 1982), and thus assesses areas of functioning in which patients with traumatic brain injuries also experience difficulty (Kennedy & Yorkston, 2000).
Formal assessment measures provide a baseline for intervention, which may be compromised when a patient's first language is not English.This becomes abundantly clear in a multilingual context like South Africa where the majority of formal assessment measures available are written in English and standardised in different settings.As a result, ad hoc translators are often used in clinical encounters, and this may influence the assessment findings in a number of ways.The untrained individual's lack of familiarity with the speech materials presented and compromised competence with the language that the clinician speaks may influence the accuracy of translation.Translators who are neither competent in English nor the language being translated may not perform as well as native speakers of that language.Hence, SLTs need to take cognisance of these factors and how their lack of knowledge of the native languages of their clients can negatively affect the accuracy of assessment results, preventing them from fully meeting their clients' needs.This heightened awareness should range from case history taking to the actual assessment to the test material used in the assessment, and therefore more emphasis should to be placed on development of assessment and intervention materials in all languages.
In response to this need within the South African SLT profession, the current study explored whether differences existed in the translation of the WAB by five first-language isiZulu speakers.In a multilingual context, it is well documented that many of the formal speech-language assessments are inappropriate for first-language English speakers, and indeed speakers of other languages.It is therefore crucial for all avenues to be explored in order to appropriately adapt or develop the most accurate and efficient assessment tools for the African continent.

Methodology
Prior to the study being conducted, permission was obtained from the University of the Witwatersrand Non-Medical Ethics Committee (Protocol number: H090404).

Aim
To determine whether the linguistic complexity of the test items of the WAB changed when translated from English to isiZulu.

Objective
To determine whether differences existed in the translation of the WAB from English to isiZulu when translated by five first-language isiZulu speakers.

Research design
A comparative qualitative research design was employed in order to determine whether similarities and differences existed within the data rather than starting with predefined concepts (Ovretveit, 1998).

Data collection
The initial part of the study involved translating the WAB from English to isiZulu in order to investigate whether the syntax and semantics as well as item complexity and familiarity changed during translation.This translation was performed by the third author (KM -translator 1).Another translation was then performed by a senior lecturer from the Department of African Languages (ALL -translator 2).The two sets of translations were compared and one set was agreed upon for use as a baseline comparative set.For the purpose of this study, the translation made by KM was chosen to be the baseline for comparison.This choice was based on the fact that she had detailed knowledge on the test material being used for translation, as well as the purpose for which the assessment tool was established.Furthermore, while the translations made by both translator 1 and translator 2 were considered to be accurate, translator 1's choice of words was thought to be less academic/formal, and therefore more familiar to the participants.Once the WAB had been translated, the three subtests deemed by the researchers to comprise the most complex linguistic features were identified: the subtests of auditory verbal comprehension, sequential commands and spontaneous speech.These were then utilised in the second phase of the study where verbal translations were conducted with the five participants (Tables 2 -4).
Data collection took place with each participant individually in a quiet private consulting room at the University Speech and Hearing Clinic.The researcher dictated each item of each of the subtests individually, and the participants were instructed to verbally translate these from English to isiZulu.These translations were digitally audio-recorded and later transcribed.

Participants
Five first-language isiZulu speakers were recruited using convenience sampling.The choice of isiZulu speakers was based on the fact that this is the most widely spoken African language in South Africa (Stats SA, 2003), and represents the most commonly spoken African language at the University Speech and Hearing Clinic.Furthermore, both the second and third authors are first-language isiZulu speakers.Inclusion criteria stipulated that all participants needed to be over the age of 18 years, be first-language isiZulu speakers, and should consider themselves to be proficient in English as per self-report.Table 1 provides a description of the participants.
All participants were provided with information letters and consent forms in both English and isiZulu prior to the commencement of data collection.Where the participants were unable to read, the researcher verbally explained the study to them in isiZulu and verbal assent was accepted in lieu of written consent.Ethical considerations were guided by the principles of the South African Medical Research Council (South African Medical Research Council, 2003).

Data analysis and trustworthiness of findings
Data were analysed qualitatively using comparative analysis.Each translation was transcribed and back-translated from isiZulu to English.Thereafter, each translation was compared against the baseline translation.Furthermore the translations were compared with each other so as to identify differences in translation across the five participants.Trustworthiness of findings was accounted for by means of back-translation of the baseline WAB, as well as independent validity checks of two of the five participants' translations to eliminate bias.

Results
Results are presented in accordance with the aims of the study.Results of the translations of the most linguistically complex subtests of the WAB are depicted in Table 2 (Tables 2 -4 are placed at the end of the article to facilitate the flow of the article).

Comparison of translations made by the researcher and African Languages Lecturer (ALL)
Initially, the translations made by the researcher and the ALL were compared.While some differences relating to sentence structure and vocabulary were noted, these appeared to be mainly dialectal in nature and did not affect the overall linguistic complexity of the test items.The reasons behind these differences appeared to pertain to the ALL's more formal and academic use of isiZulu, while the researcher's choice of words was more colloquial.
Within the subtest of auditory verbal comprehension, a few changes to structure and complexity of the questions were noted.An example of this can be seen in the translation of the question 'Is the door closed?'In English this phrase consists of the structure Copula Determiner Noun Adjective, whereas in isiZulu the structure changes to Adjective Noun which is translated to 'Uvaliwe umnyango?'Although the structure of the question changes, the meaning remains the same, and when translated back into English the question stays the same.Similarly, by translating the question 'Are you wearing red pyjamas?' to isiZulu ('Ugqoke izimpahla zokulala ezibomvu?') the sentence structure changes to 'Are you wearing pyjamas that are red?'This indicates that the meaning does not change, although the length of the question increases.This was also noted in the phrase, 'Do you eat a banana before you peel it?'('Uhlubaubhanana ngaphampi kokuba uwudle?').In this phrase the Noun Phrase (NP) is followed by the Verb Phrase (VP) in English.However, in isiZulu the VP is followed by the NP, further indicating that the structure of the question changes, yet does not compromise the meaning of the question.When translated back to English, the phrase reads as 'Do you peel a banana before you eat it?' In the subtest of spontaneous speech, differences were also noted.The researcher translated the phrase 'What is your occupation?' to 'Usebenza laphi?' which, when translated back into English means 'Where do you work?'This implies that the response obtained could refer to place of occupation but not the type of occupation.Some differences were also noted in the translation of the auditory verbal comprehension subtest.For example, the researcher translated the command 'Shut your eyes' as 'Vala amehlo' whereas the ALL translated it as 'Chimeza'.Both of these translations are accurate representations of the English command as a patient would perform the command correctly if translated either way.The only important difference in the two translations is that the translation made by the researcher can be back-translated into English as 'Close your eyes' whereas the translation made by the ALL would be translated to English as 'Shut your eyes'.Another difference was noted in the command 'Point to the chair'.People from the more rural parts of KwaZulu-Natal and those who have formally studied isiZulu tend to use the word 'isihlalo' as used by the ALL, for 'chair'.However, the researcher translated the word 'chair' as 'isitulo'.The word 'isitulo' is a 'borrowed word' taken from the Afrikaans word 'stoel' and is used more commonly in more urban areas, whereas 'isihlalo' is the more authentic version of the word 'chair' in isiZulu.
In some instances, the word order of the sequential commands changed.An example of this is seen in the command 'Point to the comb with the pen' , which was translated to 'Ngepeni, khomba ikama', meaning 'With the pen, point to the comb'.Although this change did not compromise the overall meaning of the command, it has the potential to affect a patient's understanding of the command and may also change the sequential order in which they carry out more complex commands.
Most of the changes that were noted from the initial translation of the WAB were concerned with structure and vocabulary.Since the ALL had studied isiZulu at both an undergraduate and postgraduate level, her translations were more formal as opposed to the more colloquial translations by the researcher.This implies that level of education may play an influential role on translation.These effects of education may also be true for the patient population.This is of significance to SLTs since the risk of misdiagnosis of communication impairments may be increased by failure to ensure effective translation of assessment materials or by assessing the patient in a language that he/she does not understand (Stolk, Ziguras, Saunders, Garlick, Stuart & Coffey, 1998).This suggests that SLTs should be cautious in their interpretation of assessment findings elicited through informally mediated consultations.

Comparison of verbal translations made by the five participants
The results of this aspect of the study have been tabulated in Table 2.
Based on these results, it was noted that many more discrepancies were noted across the translations, specifically those concerning vocabulary used, as well as sentence structure and semantics.

Variations in translation relating to vocabulary
As evident in Table 2, much of the vocabulary used in the subtest may be considered inappropriate for assessment in South Africa.This concurs with the findings of Mosdell et al. (2010), who also found that test items in the Boston Naming Test and the Cookie Theft Test were both linguistically and culturally biased, affecting the reliability of the tests when used in the South African context.Over and above the effect that culturally inappropriate vocabulary may have on the responses obtained from a patient, the nature of the vocabulary may be altered by the translator if he/she has limited vocabulary in his/her own language and/or in English.For example, when looking at nouns such as Smith, Brown, Windsor and Toronto, it may not be easy for a second-language speaker of English to relate to such words, and may thus affect the results of the testing procedure.Anderson (1992) recommends that materials used during assessments and therapy should be culturally and linguistically sensitive so as to appropriately guide patient management.Vocabulary such as 'hotel' and 'snow' in the WAB may be inappropriate to use in a country such as South Africa, especially in the government healthcare context, since many people in this context may not be familiar with the concept of a hotel.Similarly, many people in South Africa have never seen snow before or heard what it is called in languages other than their own.For this reason it may be difficult to comment on when it snows because of the fact that they do not have any experience on which to base this.The WAB was standardised in New York and therefore the expected response for the absurdity question 'Does it snow in July?' is 'no'.However, in South Africa it is quite possible to have snow in certain parts of the country at this time of year.This is therefore clinically relevant as a patient may then be thought to be presenting with impaired comprehension because of the fact that he answered the question incorrectly.
The word 'axe' is also considered inappropriate for use in a South African assessment battery.This is due to the fact that some patients may be unfamiliar with the name of this object and therefore may have difficulty in answering the question.For example, participant 4 translated 'Do you cut grass with an axe?' as 'Unganqamula utshani ngombese?' meaning 'Can you cut grass with a knife?'This may be due to the fact that she does not have the vocabulary to make a connection between the English word 'axe' and the isiZulu word 'imbazo'.Implications of this mistranslation which leads to changed meaning entirely are considerable and have direct impact on test scoring as well as eventual diagnosis.

Errors in translation of syntax and/or semantics
Word order is an aspect that may be altered during translation and may have negative effects on SLT assessments.Berndt (2001) states that verbal short-term memory is frequently affected in patients with neurological impairments.Therefore a change in sentence structure may have a negative effect on the neurologically impaired patient's ability to comprehend an instruction.
From the subtest of auditory verbal comprehension ( a dog or a horse?'In all of these instances, patients with neurological impairments may have difficulty in responding accurately since the nature of the question has been changed from one which requires a yes/ no response to a forced alternative, and the processing requirements being assessed have also been simplified.In the questions ' Are you a man or a woman?', ' Are you a doctor?' and 'Is this an assessment room?', participant 5 consistently asked the question by saying 'Could you/this be a woman/doctor/therapy room?'This increases the level of complexity of the question and requires a degree of reasoning in formulating an answer, which may be difficult for the neurologically impaired patient, or the patient's reasoning may be difficult for the SLT to follow.In the question ' Am I a man or a woman?' participants 1, 2, 3 and 4 all changed the nature of the semantics, thus posing the questions 'Is she the female here?',' Am I a woman?', 'Is she a male or a female?' and ' Are you a man or a woman?' respectively.Another change in semantics can be seen in the question 'Will paper burn in a fire?' where participant 4 translated it to 'Is paper burning?', thereby altering the meaning conveyed in the question and the expected response.These examples not only reveal the impact on complexity of the stimuli, but also highlight the influence of informal translation on validity of the instrument. The only semantic differences in the spontaneous speech subtest (Table 3) were in the question 'What is your occupation?'This may be because few people use the term 'occupation' when they speak about work, but tend to use the word 'job' or 'work', and may explain participant 2's translation to 'Where do you work?' which resulted in a change of semantics.In the same question, participant 4 mistook the meaning of the word 'occupation' for the meaning of 'address' in the question 'What is your occupation?', thereby completely altering the meaning of the question.
Analysis of the subtest of sequential commands (Table 4) revealed that in the command 'Point with the pen to the book' participants 2 and 3 changed the word order to 'Khomba ibuku ngepeni' and 'Khomba incwadi ngepeni' respectively, meaning 'Point to the book with the pen', thus changing the sequential nature of the command.Furthermore, participant 4 translated the same command to 'Khomba ikamu neball peni' meaning 'Point to the comb with the pen' thus changing the semantics of the command.Similarly, in the command 'Point to the window then to the door', participants 1 and 4 both translated the command as 'Point to the window and the door', and in the command 'Point to the pen and the book', participants 3 and 5 both translated the command to 'Point to the pen and then point to the book.' These errors in translation affect the expected sequential order of the response and consequently patients may be misdiagnosed as having auditory memory impairments and/or sequencing problems incorrectly.In the command 'Point to the comb with the pen', participants 1 and 4 translated this as 'Point with the pen to the comb' which, although it maintains the semantics of the command, changes the complexity.
Similarly, the command 'With the book, point to the comb' yielded errors in translation, where participant 2 translated it as 'Point with the comb to the book', thus changing the semantics, and participant 4 translated it as 'On the book point to the book' thus providing a nonsensical command.The command 'Put the comb on the other side of the pen and turn over the book' was problematic for almost all of the participants because of the increasing complexity of the command, with only participant 5 giving an accurate translation of the command.An important factor highlighted by Berndt (2001) is that most neurologically impaired patients have impairments related to sentence comprehension tasks, especially with semantically reversible sentences that are syntactically complex.This implies that a patient who already presents with a neurological impairment may have marked difficulties in answering a question with complex semantics and syntax, especially where translation is necessary; where inaccurate translations are performed, the assessment process is therefore severely hampered.

Discussion
As illustrated in this preliminary study, numerous differences and errors in translation may occur when untrained translators are used, and these have the potential to alter the responses obtained from the patient.This may result in misdiagnosis and consequently inappropriate intervention.While it is also not possible to fully identify the characteristics that make one translator more accurate than another, there are a number of factors that can be identified from the current study.The following factors need consideration when translations are conducted from English to isiZulu: The impact of isiZulu as a tonal language on translation Stress and intonation patterns impact substantially on the semantics of spoken language in isiZulu.This implies that lexical tone can be used to attach different meanings to words which share the same phonemic content (Kuun, Zimu, Barnard & Davel, 2005).The effects of this were noted, for example, during the transcription of participant 4's question 'Am I a man or a woman?' which was translated to 'Ungumuntu wesifazane noma wesilisa?'As a result of the intonation patterns of the language, stress can be placed in different places on the word 'ungumuntu' thereby changing the meaning of the sentence.The subtle difference between 'ungumuntu' (is she/he) and 'ungumuntu' (are you) will affect the patient's response.This is an important consideration since the ability to accurately detect intonation is a cognitive function involving the right hemisphere (Kuun et al., 2005); therefore it has significance in the assessment of a neurologically impaired individual.

The impact of gender on translation
Although the sample size was small, it was observed that generally the male participants translated the test items with the closest level of accuracy to the original stimuli.It is unclear why the males in this study consistently gave more accurate translations than the females; however, this finding is of interest given that in most clinical encounters, females are more likely to be requested to assist with translation when a language barrier exists.Within the South African context, females dominate the nursing profession, and they are also the ones who most often take the role of caregivers.This reality is also confirmed by Friedland and Penn (2003), who state that it is typical within the South African healthcare sector for nurses and cleaning staff, who are predominantly female, to act as translators in cross-linguistic encounters.
Interestingly, on a few occasions in the current study, female participants used the carrier phrase 'please' during translation of test items.For example, in the spontaneous speech subtest, both participants 2 and 3 translated 'Tell me a little about why you are here' to 'Please tell me a little about why you are here'.Similarly in the sequential command 'Raise your hands', participant 2 translated this to 'Please raise your hands'.While the sample is too small to draw definitive conclusions as to why this occurred, it is possible that the female participants, because of their 'nurturing' nature attempt to establish a rapport with the 'patient' during the process of informal translation.This use of the carrier phrase may also be related to cultural norms in isiZulu which may dictate subservience on the part of females in general societal functions and engagements.

The impact of education on translation
Based on the current findings, it was evident that level of education also appears to play a role in the type of translation.Participant 4 has the lowest level of education.A number of variations were noted in the translations made by this participant compared with the other translators, with the most significant variation being in the translation of 'What is your occupation?' Owing to the fact that participant 4 may not have had extensive exposure to formal English because of her level of education, she was unable to understand the term 'occupation' , but instead interpreted it as 'address' .Ercikan (1998) states that a translation made by an interpreter must reflect not only the meaning of the original item, but should also maintain the same relevance, intrinsic interest and familiarity of the item content.Furthermore, Ercikan (1998) argues that if this is not done, what the item measures may be altered.The current authors agree with this statement as evidence from the current study provides some support for it.
A further example of the impact of education on translation can be seen in the subtest of auditory verbal comprehension, where participant 4 was unable to provide the correct term for the object 'axe', indicating that she did not have sufficient vocabulary to make the translation.However, having said this, the participant with the highest level of education did not always translate the test items with the closest degree of accuracy.When the impact of level of education on translation was discussed with ALL, she revealed that 'reasons' for translation also play a role in how translation is done, and this may not necessarily be related to level of education.Furthermore, she reported that translating items for a 'language assessment' may pose significant challenges for the translator and requires a higher level of precision than 'routine' translation.Therefore, despite the fact that in the South African context English is learned at school by the majority of the population and it can be argued that level of education may give a translator access to a greater vocabulary or to a greater degree of bilingualism, it may not necessarily improve their ability to make translations without changing the inherent meaning, especially in cases where the purpose of the translation has not been explicitly described.

The role of multilingualism and language exposure on translation
The number of languages that a translator is able to speak appears to play a role in the type of vocabulary that is used during translation.This was observed, for example, in the subtest of auditory verbal comprehension when the question, ' Are the lights on in this room?' was posed.Participants 1, 2 and 3 translated this sentence correctly.However, participants 4 and 5 substituted the isiZulu word 'ugesi' with the English word 'light'.This can possibly be explained by looking into where the participants are from.Participant 4 is from Newcastle in KwaZulu-Natal, where isiZulu is the dominant language, but many people from different backgrounds live there, and a number of languages are spoken.The same applies to participant 5 who is from Soweto.Therefore, it is understandable that they would not use the authentic isiZulu term, 'ugesi' for 'lights'.Nevertheless, it is unlikely that this translation would negatively influence the response elicited.The use of 'borrowed' English words during translation may also be related to age, in that the younger participants are more likely to have had more exposure to English as a result of being born during a time when more people were educated with English as a medium of instruction, and also having had more exposure to English on television.

Conclusion
Although this study employed a small sample size, it has effectively highlighted some of the potential dangers associated with non-formal and non-systematic use of informal translators in the translation of formal speech-language assessment tools.One may argue that the WAB is not the most frequently used assessment tool, yet the authors believe that similar results would have been obtained had another formal test been translated, given that the nature of the differences in translation were mostly related to vocabulary and semantics.
The findings suggested that test items may be affected by the language, culture, gender and level of education of the translator.This concurs with the findings of Mosdell et al. (2010); however, at this point these factors cannot be considered as definitive because of the preliminary nature of the study.These factors therefore warrant more in-depth attention in future studies where bigger sample sizes can be used.
While the effects that translation may have on the process of assessment of language impairments are undeniable, the solution to this is less clear.As SLTs we can continue to motivate for the employment of formal trained translators/interpreters, yet issues associated with the diversity of cultures and languages in the South African context may not be notably reduced in the presence of a translator.While the goal of employing a translator in cross-cultural consultations would be to reduce the effects of language barriers on the accurate assessment of speech and language impairments, this study produces evidence that mediated consultations may in fact be more complex in speech-language therapy than in general health communication, and warrant a different type of solution.Efforts to develop and standardise assessment tools in African languages, coupled with increasing the numbers of SLTs who speak African languages, need to be intensified as it appears these would be the most reliable and logical ways of ensuring that appropriate and effective SLT service delivery is achieved within this context.Tell me a little about why you came here today.
Tell me a little why you are here.
on the other side and put it on top of the book 4 Beka i-ball peni phezu kwekamu bese ngiyayiphendula bese unikeza mina Put the pen on top of the comb and then I will turn it over and give it to me 5 Beka ikamu kuloluhlangothi lwencwadi bese uphendula incwadi Put the comb on the other side of the book and then turn over the book

Table 2
), a number of semantic differences in translation were apparent.Participant 3 translated the question 'Are the lights on in this room?' as 'Are the lights on in this room or not?' Similarly, participant 3 changed the nature of the question 'Does March come before June?' to 'What comes first, March or June?' and translated 'Is a horse larger than a dog?' to 'What is larger,