ISSN: 1579-9794
Hikma 24(1) (2025), 1 - 31
Learner Translations in Contrast: An English-Basque-
Spanish Case Study
Traducciones de aprendices en contraste: un estudio de
caso inglés-euskera-castellano
MARLÉN IZQUIERDO
marlen.izquierdo@ehu.eus
Universidad del País Vasco (UPV/EHU)
NAROA ZUBILLAGA
naroa.zubillaga@ehu.eus
Universidad del País Vasco (UPV/EHU)
Fecha de recepción: 27/03/2024
Fecha de aceptación: 27/02/2025
Abstract: This article reports on a contrastive descriptive study of learner
translations into Basque and Spanish of the same source text originally written
in English. It is a preliminary case study with a twofold aim: first, to bring to
contrast translations into different target languages (TLs) that have been
produced under the same circumstances; and two, to identify and test the
usefulness of corpus metadata in the interpretation of the translation product.
The study was conducted within the framework of the MUST (MUltilingual
Student Translation) project. Accordingly, all learner translations were first
aligned, and then annotated with regard to three different error categories:
transfer, language, and meta-text. The juxtaposition and contrast of the
translational phenomena identified in each target language showed that the
Basque data set reveals more language-related problems, while the Spanish
data set features slightly more transfer-related errors. In interpreting the
results, we benefited from the learners’ metadata our MUST corpora are
enriched with. In particular, we observed that those learners who considered
Basque to be their one and only mother tongue performed the best in terms
of language.
Keywords: Contrastive descriptive translation study, Learner corpora,
English-Spanish-Basque, Diglossia
Resumen: El presente artículo da cuenta de los resultados derivados de un
estudio descriptivo-contrastivo de traducciones de estudiantes al euskera y al
castellano de un mismo texto originariamente escrito en inglés. El objetivo de
dicho estudio preliminar era doble: primero, contrastar traducciones a
2 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
distintas Lenguas Meta (LMs) producidas en circunstancias iguales; y
segundo, identificar y testar la utilidad de un corpus de metadatos en la
interpretación de la traducción en cuanto producto. El estudio se realizó en el
marco del proyecto MUST (MUltilingual Student Translation). Para ello, en
primer lugar, se alinearon todas las traducciones de los estudiantes y, a
continuación, se realizó la anotación de tres categorías de error diferentes:
transferencia, lengua y metatexto. La yuxtaposición y el contraste de los
fenómenos traductológicos identificados en cada lengua meta mostraron que
el grupo con lengua meta euskera revela más problemas relacionados con la
lengua, mientras que el grupo con lengua meta español presenta ligeramente
más errores asociados con la transferencia de contenido. En la interpretación
de los datos, nos beneficiamos de los metadatos de los estudiantes con los
que están enriquecidos nuestros corpus MUST. En concreto, observamos
que los alumnos que consideraban el euskera como su única lengua materna
eran los que obtenían mejores resultados lingüísticos.
Palabras clave: Estudio descriptivo-contrastivo de traducción, Corpus de
aprendices, Inglés-español-euskera, Diglosia
INTRODUCTION
Empirical approaches to translation research have traditionally focused
on product-oriented studies that observe and describe real world translation
phenomena from a corpus approach. TRACE, TRALIMA-ITZULIK or ACTRES
are some examples of research groups working on corpus-based translation
studies in Spain. Echoing Olohan’s (2003) call for “contextualising translation
by combining corpus-based investigations with other kinds of methodologies
and analyses” (p. 419), this study tries to broaden the scope of empirical
research by combining methodologies proper of two neighbouring disciplines,
namely, corpus-based contrastive linguistics (CBCL) and corpus-based
translation studies (CBTS). The former explores differences and similarities
between at least two languages on the basis of a tertium comparationis that
enables the contrast. The latter examines strategies, resources, and norms in
the rendering of content from a source language (SL) into, at least, a target
one (TL). The notion of equivalence is key to the quality evaluation of
translated texts.
Many scholars have underlined the need to integrate in such empirical
observations more social, contextual, and cognitive data (De Sutter and Lefer,
2019). In this regard, for a few years now the convergence between corpus-
based and process-oriented translation studies is shaping current empirical
translation studies (TS) (Kotze, 2019), thus connecting the three branches of
TS, namely, product-, process-, and function-oriented research (Holmes,
1972). This effort requires new-generation corpora that are “more carefully
Marlén Izquierdo and Naroa Zubillaga 3
Hikma 24(1) (2025), 1 - 31
designed to take consideration of translators’ backgrounds and the
circumstances of text production” (Kotze, 2019, p. 356). A prime example of
such an endeavour is the MUltilingual Student Translation (MUST) project, a
learner translation corpus enriched with standardised metadata related to the
source text, the translation, and the students (Granger and Lefer, 2020). It has
been within the framework of this project that members of the TRALIMA-
ITZULIK research group have started to investigate translations made by
students from various languages, such as English, German, or Spanish, into
Basque (Sanz Villar, 2024; Izquierdo Fernández and Zubillaga Gómez, 2022).
Accordingly, this study aims to contribute to empirical translation research into
Basque with a contrastive descriptive translation case study from a learner
corpus approach, a novel type of study in the field. In particular, we will
compare the translations of the same source text into Basque and Spanish as
rendered by translator trainees to answer the following questions:
i) Does a given source text (ST) pose the same problems to learners
translating it into two contact languages?
ii) Do the same chunks trigger the same or different translational
phenomena in each target language (TL)?
Considering the diglossia situation in which the languages under
contrast are used for translation and/or other communicative purposes, and
benefiting from the students’ metadata, the current study will interpret the
findings in the light of the learners’ linguistic background to answer our third
research question:
iii) To what extent do translation products differ depending on whether
the learner’s mother tongue is Basque (EU) or Spanish (ES)?
1. CORPUS-BASED CONTRASTIVE LINGUISTICS AND TRANSLATION STUDIES
The disciplines of contrastive linguistics (CL) and translation studies
(TS) have long been related as two branches of “interlingual linguistics” that
enabled empirical approaches to language through contrast and/or contact
(James, 1980). Ever since the scientific development −and settlement− of
corpus linguistics in the 1990s, CL and TS have evolved over the course of
time in a parallel and complementary way, shaping the field of cross-linguistic
research (Izquierdo Fernández, 2007). The envisioned potential of corpus
linguistics “in refining understanding of how languages relate to one another”
(Granger et al., 2003) has taken shape in various ways: in a large number of
tailored, periodic conferences (e.g. UCCTS); in widely-cited, top-edited
publications (Baker, 1995; Granger et al., 2003; Doval Reixaand Sánchez-
Nieto, 2019; Lavid-López et al., 2021), and journals (Languages in Contrast,
Target, IJCL, Corpora, or Learner Corpus Research, to name few); in a wealth
4 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
of open-ended research (Granger and Lefer, 2020); and, most importantly,
through a steady increase of applications developed within both fields (see
Doval Reixa and Sánchez-Nieto, 2019). As a result, a great deal of present-
day cross-linguistic research is by definition corpus-based, relying on various
types of corpora to fine-tune theories with descriptive adequacy.
Both comparable and parallel corpora have played a key role in the
cross-fertilization of CL and TS, in various ways: i) unveiling language-related
aspects that might escape the human eye but are observable through
comparable/parallel concordance data; ii) conferring robustness and maturity
to empirical research thanks to statistics, corpus-driven techniques, and
systematised annotation schemes at various levels; iii) opening up research
avenues by combining different types of data; and iv) enabling data
triangulation (Marco Borillo, 2019), among other possibilities.
A productive and extended procedure in joint corpus-based contrastive
linguistics (CBCL) and corpus-based translation studies (CBTS), to which we
will refer for short as CBCL-and-CBTS, entails the use of a comparable corpus
to analyze a given language pair, and the use of a parallel corpus to examine
the translation from one of the languages into the other (Rabadán Álvarez,
2007). Insights from the comparable contrastive analysis are then considered
control data in the interpretation of parallel-based findings. Less common,
however, has been the use of translated data as contrastive data. CBCL-and-
CBTS research, on the other hand, is no longer restricted to a pair of
languages; multilingual corpora are more and more frequently used in cross-
linguistic research (Hunston, 2002; Johansson, 2007). Moreover, various
directions of analysis are considered (e.g. PAGES project), as well as various
modes of communication (e.g. EPTIC project). Most importantly, the nature of
the language data under analysis has also diversified, with learner language
being a central object of study from the point of view of CBCL-and-CBTS. The
use of learner corpora has contributed positively to a reapproachment
between the abovementioned disciplines, not only between one another, but
also with the field of language pedagogy. As Granger and Lefer (2021) state,
learner corpora are believed to strengthen synergies between contrastive
linguistics and translation studies, as the present study seeks to prove.
To add one more trend that draws from all the interdisciplinarity above,
this case study reports on a contrastive descriptive translation study, where
the same source text in English (EN) is translated into two target languages,
namely, Basque (EU) and Spanish (ES) by translator trainees. In other words,
we will contrast students’ translations arising from the same source text in two
contact languages, a type of analysis never attempted, to the best of our
knowledge.
Marlén Izquierdo and Naroa Zubillaga 5
Hikma 24(1) (2025), 1 - 31
2. LEARNER CORPORA IN EMPIRICAL TRANSLATION RESEARCH
As repositories of authentic, acted-out cross-linguistic
correspondences, parallel corpora, as well as comparable corpora, are of
direct benefit not only to translators but also to learners of a foreign language,
be it for general purposes or for specific purposes, such as translation
(Bowker, 1999). Notwithstanding this advantage, to truly progress towards a
meaningful teaching-learning experience it is necessary to observe first-hand
what learner language use is like. In short, the more learner data we examine,
the more insights we gain into their linguistic/translator competence and,
therefore, the easier it might be to raise student awareness of their
learning/training needs.
Accordingly, building corpora of students' translations is motivated by
the need to extend the fruitful combination of learner corpus research (LCR)
and CBTS. Among the first initiatives to combine translator training research
and CBTS, Bowker’s corpora created by translators (CCBT) stands out as “a
type of learner corpora that can be used to investigate difficulties encountered
by trainee translators” (Bowker, 2003, p. 169). Similarly, the Translation
Teaching and Learning Corpus (TTLC) project was carried out to “give priority
to actual learner needs and integrate both language-based and process-
based translating skills” (Tiayon, 2004, p. 119). Other ensuing successful
projects worth highlighting have been the UPF learner translation corpus,
featuring the language pair English-Catalan (Espunya Prat, 2014); special
mention deserves the undergraduate learner translator corpus (ULTC) that
features translations from English or French into Arabic (Alfuraih, 2020). This
kind of proposals unquestionably meant a breakthrough in empirical
translation research, even though there was still a lot of room for improvement.
The language combinations lacked variety, translations were analysed in only
one direction, and inconsistencies or lack of systematicity in error-annotation
prevented the comparability of findings across research projects, let alone a
generalization of results with real pedagogical application (Granger and Lefer,
2020). In addition, learner corpus research focused most of its attention on L2
teaching-learning, to the detriment of translation training. This apparent
neglect, however, would nevertheless benefit student translation corpora in
the years to come, making available for researchers long-tested and widely
attested protocols in corpus design, methodology, analysis, and application.
In order to overcome the abovementioned shortcomings, the
Multilingual Student Translation project (MUST) was launched in 2016 as an
ambitious, yet well-thought, initiative that has grown over the past years and
at the time of writing brings together researchers in the fields of translation,
contrastive linguistics, and language/translation pedagogy from 38
universities (Granger and Lefer, 2020). The MUST project provides
6 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
researchers with relevant metadata about students that might help understand
the process and outcome of translation, as well as to help understand
translation as a social process. Students’ metadata are related to their mother
tongues, the foreign languages they know, or the languages used in their
education, among others. We are particularly interested in the students’
linguistic background, given that the two target languages of our contrastive
descriptive translation study, namely, Basque (EU) and Spanish (ES) are
languages in contact in a diglossic environment. The relevance of this fact is
twofold. First, the study of “languages in contact” has always been framed
within cross-linguistic research (James, 1980; Izquierdo Fernández, 2007).
Second, we would hypothesise that the social reality that shapes our students’
communicative competence, as characterised by diglossia, might have an
effect on the translation process, arguably projected or reflected on the
translation outcome that we will juxtapose and contrast (see section 6).
3. DIGLOSSIA IN THE BASQUE COUNTRY
The territory where Basque is spoken is divided in two nations (Spain
and France) and three regions or communities. The sociolinguistic situation of
the Basque language differs from one territory to the other, since the official
status of the language is different, and the measures oriented to its promotion
vary as well. In the Basque Autonomous Community (a.k.a. Euskadi), Basque
is the official language together with Spanish and all children learn Basque at
school. Nevertheless, since the approval of the 138/1983 Basque Law ruling
the use of official languages at non-university education, there are three
different educational models, namely A, B and D, according to which the role
of Basque as the language medium of instruction differs. Out of these models,
it is only Model D that provides integral teaching in Basque except for the
subjects of Spanish and a foreign language, most usually English. This model
has become today the most popular one (Urdalleta Lete, 2023). Nevertheless,
while this formal instruction enables language acquisition, it does not
guarantee a settled habitual use of the language. In fact, according to the
latest sociolinguistic survey, 53% of the population over 16 can speak or
understand Basque (Basque Government, 2016, p. 4) but just 20.5% use it on
a daily basis and as their main language (Basque Government, 2016, p. 23).
The case of the Autonomous Region of Navarre is particularly complex.
The region is divided into three linguistic zones, where the status of Basque
varies from fully to partially or no official. Consequently, the presence of
Basque at schools is enhanced to different degrees. Nowadays, 23.2% of
people over 16 can speak or understand Basque in Navarre (Basque
Government, 2016,p. 4), but just 6.6% use it as their main language (Basque
Government, 2016, p. 23).
Marlén Izquierdo and Naroa Zubillaga 7
Hikma 24(1) (2025), 1 - 31
This diglossic situation has its impact on the actual use of, even attitude
towards, the Basque language by learners. For example, Barnes Mason and
García Fernández (2011) found out that most of the children at school age
use Spanish regardless of the educational model. In this regard, extensive
research on the issue of minority language acquisition has been done on
Basque (Austin, 2009; Almgren and Manterola Garate, 2016), also in a
trilingual context with English (Leonet et al., 2019). Nevertheless, to the best
of our knowledge, not much attention has been paid to the role of Basque (in
contrast with Spanish) in the development of translators’ competence or
students of translation training. The few studies analysing Basque as a
language of translation that have been done so far focus on literary
translations published by professional translators (Zubillaga Gómez et al.,
2015; Zubillaga Gómez, 2016; Sanz Villar, 2018).
4. METHODOLOGY
Framed within the MUST project (see section 2), we carried out a
description of translational options benefiting from source text (ST) to target
text (TT) alignments on two parallel corpora. The study is mainly qualitative in
that translations are annotated for errors, as well as for particularly good
translational options. In addition, some quantitative data is provided for the
sake of contrast. The last stage of the study considers the learners’ linguistic
background, which might explain the contrastive results of the translation
annotation process.
In the following sub-sections, we will detail which data was used for the
study, and the actual procedure followed to analyse it.
4.1. The Data
Data was retrieved from two MUST sub-corpora, i.e., English-Basque
(EN-EU) and English-Spanish (EN-ES). Each of these corpora is a multiple
translation corpus (Espunya Prat 2014), as there is more than one translation
for the same ST. In turn, each corpus contains several translations as output
of different learning activities.
For this particular study, we selected data from one specific ad hoc
translation task, designed in exactly the same way for students translating into
EU and those doing it into ES. The ST, common to both language pairs, was
an opinion article written by Eli Davies and published by The Guardian, on its
online version, in January 2021 (Davies, 2021). The topic dealt with in the text
was the loneliness pandemic brought about by COVID19. The article was
originally 994 words long, but it was shortened to a bit less than 600 words to
make the translation task suitable for a continuous assessment activity.
Students were given 105 minutes to accomplish the task during class time, in
8 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
a computer-equipped classroom, and were allowed to use tools and resources
to assist their translation, such as online dictionaries, machine translation
engines, or corpora, among others. This task would be marked, i.e., it would
add on to the students’ final mark in the subject, a fact all the translator
trainees were aware of. The task represented 15% of the total grade and was
corrected by the teachers, who later provided the learners with feedback.
Therefore, the task was realised in a parallel but independent way by learners
enrolled in translation courses for each language pair. In this way, we
managed to select parallel data with the greatest comparability degree
possible, thus resembling a comparable parallel corpus (Hareide, 2019).
Table 1 shows the corpora composition:
EN-EU
EN-ES
No. translations
15
17
No. words/translated
sample
7016
10547
No. words/translation
467.7
620.4
Degree year
3rd
3rd
Degree course
Translation Practice IV:
English-Basque I
Translation Practice IV:
English-Spanish IV
Task timing (minutes)
105
105
CAT Tools allowed?
Yes
Yes
For marking?
Yes
Yes
Table 1. Comparable parallel data sets
Source. Elaborated by the authors
To compile two parallel corpora that were in turn comparable, a tertium
comparationis was established to enable the contrast of the two translational
data sets. This was done on the basis of i) the students’ profile described in
demographic (gender, age, educational level, and training), as well as in
linguistic terms (e.g., years of L2 learning and bilingual mindset); ii) task
instructions and conditions for completion (e.g., time, setting, software used
and grading); iii) size of data sets, notwithstanding the minor differences in the
number of translations collected per language pair, i.e., number of
translations, and consequently number of types;1 and a final aspect that
warrants comparability of our data would be iv) data annotation, as described
in the following section.
The EN-ES group included 17 students (3 males, 18%, and 14 females,
82%) enrolled in the third grade of Translation and Interpreting at the
University of the Basque Country. 13 of them (76%) considered ES their L1,
1The number of translations does not reflect the number of students taking the course, but the
number of the students who consented that their translations be included in MUST.
Marlén Izquierdo and Naroa Zubillaga 9
Hikma 24(1) (2025), 1 - 31
while the remaining 4 (24%) acknowledged both ES and EU as their mother
tongue. The ES translated sample is over 10 thousand words large. The
subject for which they conducted the translation task was Prácticas de
traducción IV: inglés-español IV (Translation practice IV: English-Spanish IV).
All 17 students in this group chose ES as their A language and EN as their B
language.2 Most of them came from the Basque Autonomous Community,
which means that the majority knew some EU, as both ES and EU are
compulsory in the Basque educational system. In fact, even the students who
reported having only ES as their mother tongue (76%) have mostly been
schooled in Basque, as indicated in the metadata.
Regarding the EN-EU group, it consisted of 15 students (3 males, 20%,
and 12 females, 80%), also in the third grade of Translation and Interpreting
at the University of the Basque Country. The EU translated sample amounts
to slightly more than seven thousand words. The subject for which they
conducted the translation task was Itzulpen Praktikak IV: ingelesa-euskara I
(Translation practice IV: English-Basque I). While the number of translation
practical courses into their, respective, A language is the same for both
student sets, the opportunity to work on EN as a source language is thrice as
big for trainees whose A language is Spanish when compared to those with
Basque as their A language (Ministerio de Educación, Cultura y Deporte,
2011). On the basis of such a difference, we could hypothesise that the
learners translating into Basque may commit more transfer-related errors, on
the assumption that they have a poorer command of the source language (see
section 6).
4.2. The procedure
We approached our analysis with the classical procedure followed in
contrastive linguistics, namely, selection, description, juxtaposition, and
contrast (Kreszwoski, 1990). The selection of our data has been detailed in
4.1. We will focus on the description, juxtaposition, and contrast of
translational options in the following lines.
4.2.1. Description
At the second stage of our analyses, we annotated our data using the
translation-oriented annotation system (TAS) tailored to the MUST project,
and accessible through the Hypal4Must interface (Fictumova et al., 2017;
Granger and Lefer, 2020). Although there is currently an upgraded second
2As is usual in Translation Studies at universities, all students have to choose their A, B and C
languages when they enroll. “A” language is their ‘first language’, “B” the second one and “C” their
third one. All subjects are organised in such a way that students translate from “B” and “C” into
their “A” language, except for some inverse translation subjects, where students also practice
translating from their “A” language into their “B” language.
10 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
version of the annotation system, namely TAS 2.0, this was released after we
had already annotated our data using the very first version. TAS 1.0., as well
as TAS 2.0., does not focus only on translational errors; it gives users the
possibility of tagging translational options that are considered particularly
good.
By employing TAS 1.0., we could classify translational options into four
categories, namely, ST-TT transfer (TR), language (LA), translation
procedures (TP), and meta-text (MT). At the time of annotation, the third
category was still under construction, so we described the translational
phenomena observed in the TTs using the categories of TR, LA, and MT.
These categories represent a first level of annotation of our data.
Going further down to greater detail, every category breaks down into
a second level of analysis where a number of sub-categories label a given
type of TR, LA or MT error or phenomenon. As such, the TR category is further
divided into 5 sub-categories as illustrated in Figure 1.
Figure 1. ST-TT transfer sub-categories
Source. Hypal4Must (Granger and Lefer, 2020)
The first sub-category specifies whether the ST-TT Transfer error is one
affecting “content” [TR-CT]. The second one (Lexis TR-LE) would be useful to
annotate that a wrong lexical translational option has been chosen. In case
that the transfer triggered deviations at the level of discourse and/or
pragmatics, the third sub-category (TR-DP) would mark it so. When deviations
affect register and/or cultural aspects from ST to TT, there is also a possibility
of annotation (TR-RC). Finally, the fifth sub-category within level-one transfer
phenomena would relate to changes in the translation brief (TR-TB). However,
our learners did not have to use one for their translation task.
Likewise, there are five LA sub-categories as shown in Figure 2. Every
sub-category tags errors that take place at different levels of linguistic
Marlén Izquierdo and Naroa Zubillaga 11
Hikma 24(1) (2025), 1 - 31
analysis: grammar (LA-GR), lexis and terminology (LA-LT), cohesion (LA-
CO), spelling and punctuation (LA-ME), and finally, style (LA-ST).
Figure 2. LA sub-categories
Source. Hypal4Must (Granger and Lefer, 2020)
The third category, MT, breaks down into two types of annotations
(Figure 3):
Figure 3. MT sub -categories
Source. Hypal4Must (Granger and Lefer, 2020)
The first annotation addresses positive translational solutions. The
other, by contrast, addresses translational solutions that are considered
negative as suspected to stem from SL intrusion.
Each of the TR and LA sub-categories may be further detailed at a third
level of delicacy; our data was annotated at all three levels, which was not free
from difficulty, given the high granularity of TAS 1.0.
In fact, the description stage was rather challenging to carry out for two
main reasons. First, as a three-tier translation-oriented annotation system,
TAS 1.0. was rather complex to implement, reason why TAS 2.0. has been
refined to diminish granularity and it features fewer sub-categories (Granger
and Lefer, 2021). Second, the annotation of each language data set was to
be done independently from each other, which could double the opportunity
for discrepancy between language annotators. Accordingly, to conduct the
annotation in an individual but parallel and comparable way, a basis for some
degree of inter-rater consistency was laid. Aiming for the highest degree of
inter-rater reliability, two coders first considered a few translations into each
language together, to have the possibility of interpreting TAS 1.0. jointly,
discussing their understanding of each tag, and deciding what TAS annotation
12 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
seemed to be appropriate for certain errors that were either common to both
language pairs, or recurrent, or ambiguity-ridden. In short, we aimed to ensure
coherence and consistency in our TAS management.
Once the description stage was completed, the number of annotations
per language data set was 351 annotations in the EU translations and 399
annotations in the ES data set. On average, there are 21 annotations per text
in the EN-EU translations and 22.1 in the EN-ES translations. There are,
therefore, fewer annotations in the first combination than in the second. What
this might be due to will be discussed further in the article (see section 5).
To ease readability, in this study, we will juxtapose annotations at the
first and second level (see section 4.2.2.). In the proper contrastive stage (see
section 5), we will be commenting on third-level annotations to discuss
similarities and, most importantly, differences between Basque and Spanish
translations of the same original English text.
4.2.2. Juxtaposition and contrast
In the bosom of the MUST project, a powerful interface and browsing
software was developed to serve both teaching-learning and research
purposes, namely, Hypal4Must. This is an extension of the Hybrid Parallel
Text Aligner (Hypal), a web-based interface originally developed to align
parallel texts (Obrusník, 2014). Although initially tested on the Czech-English
pair, Hypal was conceived of as a language-independent tool, which has
enabled its customisation for MUST purposes, now homing multiple corpora
and a teaching tool section (Granger and Lefer, 2021). Hypal4Must is
programmed to enable a number of corpus functionalities such as data
upload, lemmatisation, and part-of-speech tagging; text alignment, at the
paragraph and sentence level; text annotation; and corpus browsing, yielding
both statistical as well as qualitative information.
Figure 4 juxtaposes information related to the number of annotations at
the first level in each target language:
Marlén Izquierdo and Naroa Zubillaga 13
Hikma 24(1) (2025), 1 - 31
Figure 4. Juxtaposition of first-level annotations in EU and ES translations
Source. Elaborated by the authors
As illustrated, the LA-related annotations outnumber TR-related ones
in EN-EU translations (left column). In fact, students having EU as their target
language show 32.4% more language annotations (LA, 63.6%) than transfer
annotations (TR, 31.2%). LA-related annotations rank second. On the other
hand, far from there being a clear distinction between LA and TR errors in the
EN-ES data set, these categories share a similar percentage; TR-related
annotations represent 45.51%, while 44.08% of the annotations involve LA. In
other words, there are very few more TR annotations (181) than LA ones
(176). Finally, MT is clearly the least frequent category in both languages. Yet,
a striking cross-linguistic difference is observed, for there are double the
number of annotations in EN-ES than in EN-EU. A further difference lies in the
phenomena tagged under such a category; while all MT instances in the EN-
EU data set feature source-language intrusion cases into the TL, only in the
EN-ES translations were remarkably good translational options also
annotated.
Similar to Figure 4, Figure 5 juxtaposes the second-level annotations in
each language-pair translation set:
5.2
31.2
63.6
10.4
45.1
44.1
MT
TR
LA
First-level error categories
EN-EU EN-ES
14 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
Figure 5. Juxtaposition of second-level annotations in EU and ES translations
Source. Elaborated by the authors
We observe that within the LA category the most recurrent annotation
hints at grammar issues (LA-GR). Even though this is common to both data
sets, grammatical errors stand out among EU student translations. Other
grammatical errors were also identified in terms of the cohesion and
coherence of the TT (LA-CO); lexis and terminology choices (LA-LT);
mechanics of the text, i.e., spelling and punctuation, among others (LA-ME);
and issues having to do with the style of the translation/translator (LA-ST), or
whether the TT feels redundant or heavy, among other possibilities. The
occurrence of such errors differs across the two language pairs, with EU
student translations displaying first LA-GR, followed by LA-ME, LA-CO, LA-
LT, and LA-ST. In the EN-ES data set, however, LA-CO annotations represent
the smallest group. LA-LT ranks second, followed by LA-ME and LA-ST
categories. Table 2 illustrates such a diverging order of frequency of the same
LA-bound error subcategories, together with examples:
EN-EU
EN-ES
Examples3
LA-GR
1st
1st
(1)
ST. I’ve been thinking a lot about
loneliness over the last few years as I’ve drifted
in and out of various forms of it myself.
3 For the sake of contrast, we have chosen ST chunks that trigger errors in each TL whenever
available.
Marlén Izquierdo and Naroa Zubillaga 15
Hikma 24(1) (2025), 1 - 31
TT (EU). Asko pentsatu dut bakardadeaz azken
urteotan, ni neu ere
bakardadearen hainbat
aldaera pairatu baititut
[Much thought I have
loneliness (about) last years (in the), me myself
also loneliness (of) various forms suffered
because - the subject “me myself” has to be in
ergative form, nik neuk, as it is a
subject of
transitive verb].
TT (ES). He estado reflexionando acerca de la
soledad estos últimos años y he experimentado
con ella y sus distintas formas
hasta que
experimentarla en su mayor esplendor, como
es lógico, en 2020. […and its forms different until
that to experience it…]
LA-ME
2nd
3rd
(2) ST. The historian Fay Bound Alberti,
who has written a “biography” of the condition,
argues that…
TT (EU)
. Fay Bound Alberti historialariak
baldintzaren inguruko «biografia» idatzi du Ø eta
pentsatzeko modu horrek... [Fay Bound Alberti
the historian the condition about “biography”
written has Ø and to think the way this… - in
Basque you would put a comma before ‘and’]
TT (ES). El historiador Fay Bound Alberti Ø que
ha escrito una “biografía” de la situación,
sostiene... [missing comma]
LA-CO
3rd
5th
(3) ST. …and perhaps that’s why we find it
so difficult to talk about or admit to.
TT (EU). …eta agian horrexegatik egiten zaigu
hain zaila horren inguruan hitz egitea edota Ø
onartzea […and perhaps that is why make it us
so difficult about that to speak and to accept - the
second infinitive verb onartzea or “accept” needs
a pronoun in absolutive form: hura].
TT (ES). …y quizás es por eso nos resulta tan
difícil hablar de ello o admitirlo [linking “that”
[QUE-conjunction] is missing between eso and
nos].
LA-LT
4th
2nd
(4)
ST. …by the global health crisis and its
accompanying lockdown.
TT (EU). -----
TT (ES).
…la crisis sanitaria mundial y el
bloqueo que la acompañaba. […the crisis health
worldwide and the block that escorted it]
LA-ST
5th
4th
(5)
ST. …to work on my PhD
16 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
TT (EU). ------
TT (ES). …para sacarme un doctorado. […to
take out a PhD]
Table 2. LA-error types in EU and ES: order of frequency and examples4
Source. Elaborated by the authors
On the other hand, the various types of annotations tagging errors
related to the actual transfer from the SL to the TL are equally distributed in
EN-EU and EN-ES student translations. Among the various TR errors
identified, content (TR-CT) annotations stand first, followed by annotations
referring to problems in the building of the TT at the level of discourse and
pragmatics (TR-DP). Finally, the smallest percentage of transfer-related
annotations relates to the lexical choice made to render a given meaning from
the ST into the TT (TR-LE). Even though TAS 1.0. broke down this category
into more error types referring to register and cultural issues (TR-RC), or the
translation brief (TR-TB), none of them were observed in our translations.
EN-EU
EN-ES
Examples
TR-CT
1st
1st
(6)
ST. I’ve been thinking a lot about loneliness over
the last few years as I’ve drifted in and out of
various forms of it myself.
TT (EU). Azken urteotan
bakardadean asko
pentsatu dut
, ni neu bakardadeak dituen
hainbat formetatik sartu eta atera bainaiz [Last
years alone much thought I have,.... – “thinking
alone” is something different as “thinking about”]
TT (ES). He estado reflexionando acerca de la
soledad estos últimos años y he experimentado
con ella y sus distintas formas hasta que
experimentarla en su mayor esplendor, como es
lógico, en 2020. […about loneliness these last
years and have experienced with it…]
TR-DP
2nd
2nd
(7)
ST. Like so many living alone during lockdown,
I’ve felt incredibly isolated.
TT (EU). ------
TT (ES). Yo también me he sentido apartado
durante el confinamiento, al igual que muchas
personas. [I too have
felt casted during the
lockdown, same as many people - inverse clause
order!]
4 For the sake of contrast, we have chosen ST chunks that trigger errors in each TL whenever
available.
Marlén Izquierdo and Naroa Zubillaga 17
Hikma 24(1) (2025), 1 - 31
TR-LE
3rd
3rd
(8)
ST. Pages appeared on the NHS and Red Cross
websites….
TT (EU). NHS
eta Gurutze Gorriko
webguneetan...
[NHS and Red Cross(of)
websites… - NHS has not been translated]
TT (ES). Aparecieron secciones páginas web del
NHS y en la Cruz Roja que…. [Appear sections
pages web of the NHS and…]
TR-RC
-
-
TR-TB
-
-
Table 3. Order of frequency of TR-error types in EU and ES
Source. Elaborated by the authors
Example (6) shows a change of content in the transfer from original
English into target Basque and Spanish. In the TT (EU) example, the
translation means that it is “while being alone” (bakardadean) that a lot of
thinking has been done (asko pentsatu dut). Although affecting another chunk
of the original sentence, a distortion of content is also observed in TT (ES)
whereby a subordinate causal clause, “…as I’ve drifted…” is rendered into
Spanish as a complex clause introduced by coordinating conjunct y (“and”).
Not only is the content changed, but also the actual discourse, for the
underlying causal relation between main and subordinate clauses of the ST
disappears altogether in the TT (ES), with a coordination of two clauses at the
same propositional level. This phenomenon is also observed in (7) TT (ES),
where the order of the two clauses is inverted. Finally, Example (8) illustrates
the least frequent translational option in either language, which consists in
leaving an original item untranslated. The instances found among our data
affect mainly the abbreviation “NHS”, i.e., National Health System, which is
left as such in both translation data sets.
5. DISCUSSION
As an answer to our first research question, our data reveals that the
same ST does not necessarily pose the same problems for students
translating it into two −contact − target languages. Cross-linguistically, the
distribution of annotations per error categories would suggest that trainees
working into Basque have had more difficulties dealing with the target
language, for Basque translations seem to be linguistically worse done than
the Spanish translations. Such results would not refute a hypothesis posited
earlier in the article (see section 4.1.) as to whether the EN-EU learners could
have greater difficulties in decoding the original text, given that they have not
worked on English as a source language as much as their EN-ES
counterparts.
18 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
By contrast, Spanish translations appear to deviate from the ST in terms
of source content to a greater extent, which would hint at the learners’
difficulties in the craft of “transferring” meanings. Let us discuss the overall
difference in the categorization of the annotations per language pair.
First, the translator trainees’ competence in the target language may
account for the nature of the errors tagged, mainly those with a high
recurrence rate and observed to be common to all translations in the data set.
This is believed to be so in the case of grammar problems among the Basque
translations, where they stand out. In particular, the most frequently annotated
language error in the EN-EU group concerned grammar inflection (15.5 % of
all the annotations). Basque is a highly inflected language, which turns out to
be a challenge when it comes to acquiring the language, whether as an L1 or
L2 (Ezeizabarrena Segurola, 2012). The noun-phrase is inflected in 17
different ways for case and it is, furthermore, an ergative-absolutive language,
where the case used for the agent of a transitive verb is ergative, formally
marked by -k. This is exactly what many students have forgotten in their
translations, as illustrated in Example (1) in Table 2, where the subject should
have been nik neuk. The result is, therefore, an ungrammatical construction.
EN-ES grammar problems, which rank second in frequency after lexical
and terminological issues (LA-LT), affect mainly the choice of verb tense,
rendering not an ungrammatical construction but mostly an inaccurate
translation as (9) illustrates.
(9) ST. I've felt incredibly isolated.
TT (ES). Me sentí increíblemente aislada.
On the other hand, most LA-related errors in this data set have to do
with the translation of multiword and idiomatic expressions. Where the ST
features an idiomatic expression, trainees are expected to use an equivalent
expression in the TL whenever possible. This has not always been the case,
as exemplified in the translation of “struck a chord” in (10).
(10) ST. When I read these words, they struck a chord.
TT (ES) a. Cuando escuché estas palabras, me tocaron la fibra
sensible.
TT (ES) b. Estas palabras me conmovieron.
While some students produced an equivalent term, in meaning and
form, such as tocar la fibra sensible (TT [ES] a), other options were single-
word verbs conveying only the meaning, as in (TT [ES] b).
Marlén Izquierdo and Naroa Zubillaga 19
Hikma 24(1) (2025), 1 - 31
Flawed interpretation of the source text may also give shape to the
actual annotations. Distortion of the original content is clearly a pervasive
problem for the learners translating into Basque as well as for those working
into Spanish. Most often, this distortion yields an inaccurate transfer as
observed in Example (6) in Table 3. The chunk “I’ve been thinking over
loneliness” has been mistranslated into Basque by more than half the
students. We suspect that the students have actually understood the meaning
of the ST but not managed to render it in Basque. In fact, we would argue that
the actual translational option stems from cognitive interference of Spanish;
the English phrase “think about loneliness” is equivalent to Spanish pensar en
la soledad, literally, ‘think in the loneliness’. Arguably, the Spanish preposition
is negatively transferred into the Basque translations. In a bilingual scenario,
this error could be interpreted in the light of “L2 intrusion,” a tag that would be
parallel to TAS “source language intrusion” (MT-SLI). Considering the
diglossic status of Basque and Spanish, we would claim that the dominant
language intrudes, affecting both translation process and product (Sanz Villar,
2018). As a matter of fact, 17 instances of MT-SLI were found in the EN-EU
data.
Arguably, the fact that the “well translated” perspective (MT-PLUS) was
taken into account only when tagging EN-ES translations, as opposed to the
EN-EU translations where only the MT-SLI cases were acknowledged, would
be due to a rater’s “idiosyncratic vision of good practice” (Graham et al., 2012,
p. 4). In other words, the researchers did not show equal awareness of the
need or relevance of marking good translational options, which hints at a
slightly different interpretation of the corpus data.
Aware that many variables may shape the translational output of two
different language data sets, in our second research question we aimed to
observe whether one given chunk would trigger the same or different
translational phenomena in each TL. To this end, we narrowed our search to
the fifth paragraph of the text, as this seemed to be rather problematic in both
translation tasks. It triggered 84 annotations in the EN-ES group and 78 in the
EN-EU group. Our corpus is aligned at the sentence level, so Table 4 shows
the sentences composing the fifth paragraph with the overall number of
corresponding annotations:
5th paragraph sentences
EN-EU
EN-ES
A few months into the Covid-19 outbreak, people
started to talk about a corresponding
loneliness pandemic.
LA-ME (4)
LA-GR (3)
MT-SLI (2)
LA-CO (1)
TR-CT (1)
LA-LT (4)
LA-GR (3)
LA-ME (1)
LA-ST (1)
TR-CT (4)
20 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
LA-LT (1)
12
TR-DP (1)
14
Pages appeared on the NHS and Red Cross
websites advising how to cope with the isolation
thrust upon us by the global health crisis and its
accompanying lockdown.
MT-SLI (8)
TR-LE (7)
LA-ME (5)
LA-GR (3)
LA-LT(1)
24
LA-GR (6)
LA-ME (2)
LA-ST (1)
LA-LT (1)
TR-CT (5)
TR-LE (4)
TR-DP (1)
MT-SLI (5)
25
But before this year, various reports claimed that
loneliness had reached dangerous and even
life-threatening epidemic levels, and in 2018
Theresa May launched a UK government
loneliness strategy.
LA-GR (9)
LA-ME (9)
TR-CT (3)
21
LA-GR (14)
LA-LT (2)
LA-ST (1)
LA-ME (3)
TR-CT (7)
TR-DP (2)
29
Such concerns have always been particularly
heightened during winter and around Christmas,
a time when charities and politicians frequently
urge festive revellers to think about and reach out
to the lonely and vulnerable.
LA-GR (7)
LA-ME (6)
LA-LT (4)
LA-CO (2)
MT-SLI (2)
21
LA-GR (5)
LA-LT (3)
LA-ME (2)
LA-ST (1)
MT-PLUS (3)
TR-CT (1)
TR-DP (1)
16
TOTAL ANNOTATIONS
78
84
Table 4. Annotations of most problematic paragraph and common problem-
triggers (in bold) into EU and ES
Source. Elaborated by the authors
The order of the first-level annotation categories differs between the
groups; there are 55 LA (70.5%), 12 MT (15.4%), and 11 TR (14.1%)
annotations among the Basque translations. In the Spanish data set, LA
stands first with 50 (59.5%) annotations, followed by 26 TR (31%) cases and
finally 8 MT (9.5%) annotations. Likewise, the most recurrent LA error
committed by Basque learners is one either of spelling or punctuation (LA-
ME), very closely followed by grammatical mistakes (LA-GR). By contrast, this
paragraph triggered mainly LA-GR problems into Spanish, followed by errors
at the lexical and terminological level (LA-LT). Greater differences were
observed within the TR category, with Basque translations featuring mainly
lexical errors (TR-LE), as opposed to a majority of content distortions (TR-CT)
in the Spanish translation of the paragraph. Finally, for once the MT
annotations in the Basque translations outnumber those in the Spanish data,
with the qualitative difference that the former report only source language
Marlén Izquierdo and Naroa Zubillaga 21
Hikma 24(1) (2025), 1 - 31
intrusion while the latter required tagging occurrences of “good translations”
(MT-PLUS).
On the other hand, we observe discrepancies but also similarities if we
focus exclusively on the chunks that posed problems into both TLs, namely,
“corresponding loneliness pandemic” (sentence 1), “pages appeared” and
“NHS” (sentence 2), and “loneliness had reached dangerous and even life-
threatening epidemic levels” and “loneliness strategy” (sentence 3).
Cross-linguistically, “corresponding loneliness pandemic” has been
dealtwith differently. While “corresponding” seems to be the problem for
Basque trainees, many Spanish translations reveal problems at the compound
level.
TT (EU). COVID19aren agerralditik hilabete batzuetara, birusarekin
zetorren “bakardadearen pandemiaz” hitz egiten hasi zen [COVID19 (of)
outbreak months some (after), virus (with) coming ‘loneliness (of) pandemic’
(about) to speak started – the phrase “the loneliness pandemic coming with
the virus” seems not natural in Basque].
TT (ES). …hablar de una correspondiente pandemia solitaria.
“Pages appeared” and “NHS”, either together or independently, have
triggered the same kind of problems into both EU and ES. Many translations
of the first chunk are mostly tagged as MT-SLI, in view that they sound “literal”
irrespective of the TL. By contrast, the second item has remained untranslated
several times (Example 12):
TT (EU). NHS eta Gurutze Gorriko webguneetan orrialdeak agertu
ziren [NHS and Red Cross (of) websites (on) pages appeared – “NHS” has
not been translated and “pages appeared” does not sound natural].
TT (ES). Aparecieron páginas en los sitios web del NHS y…
The translation of the two items in the third sentence has brought to
light the learners’ difficulties with phraseological complexity, as already seen
in Example (11). As a result, ill-built phrases abound among both translation
data sets, mostly featuring non-grammatical solutions that would call into
question the learners’ competence in the TL.
TT (EU). hainbat txostenek zioten bakardadea epidemia mailara iritsi
zela eta arriskutsuak zein hilgarria zela [various reports said loneliness (the)
epidemic level (to) arrived and dangerous (plural) as well as lethal was – both
adjectives “dangerous” and “lethal” do not agree in the same way: plural and
singular].
22 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
TT (ES). la soledad había alcanzado niveles peligrosos e incluso,
amenazantes epidemiológicos…
In Example (14), the equivalent compound in either language would
require a preposition that is either an ill choice or missing in the target text.
TT (EU). Theresa May momentu horretan lehen ministroa zenak
«bakardadearen estrategi» izeneko kanpaina jarri zuen abian [Theresa May
moment that (in) prime minister was (who) “loneliness (of) strategy” called
campaign put forward – when reading “strategy of loneliness”, one could think
that the aim is reaching loneliness].
TT (ES). ...en 2018 Theresa May lanzó una "estrategia de soledad"
del gobierno británico.
In the light of the data so far discussed, it would not seem safe to state
that one given chunk gives way to the same kind of errors by learners working
into two contact languages. Notwithstanding this difference, it has been
observed that certain chunks operating at the lexical level are more likely to
lead to similar errors.
The high rate of LA errors among the EU translations has been an
interesting finding. Related to this, our third and last research question goes
beyond the translation product and aims to delve into the translation process,
on the assumption that the linguistic background of the trainees might explain
their decisions and the translational options so far identified. Indeed, the issue
of what a language user’s mother tongue is in a diglossic context is fraught
with controversy as much as it is filled with provoking thoughts. As briefly
explained in 3 above, the sociolinguistics of the Basque Country is rather
complex. For our research purposes, it pertains to acknowledge that the
diglossic language situation in which Basque is spoken is one where Spanish
is felt as dominant or “high” language, whereas Basque would be the
dominated or “low” language (Ferguson, 1959). This diglossic situation has its
impact on the actual use of, even attitude towards, the Basque language by
learners. In fact, it may trigger different feelings. On the one hand, Basque
being the original language of the territory, it may be felt as the only mother
tongue, even though all Basques are unquestionably competent in Spanish.
On the other hand, there are also people who feel comfortable in both
languages, and others who barely use Basque although they have most likely
been formally instructed in the language, especially among the age range our
learners would belong to. Taking this complex sociolinguistic situation into
account, we decided to look closer at those metadata about the mother tongue
in the EN-EU group and relate them to the annotations for language. Figure 6
illustrates the type and percentage of first-level TAS categories per EN-EU
trainee. Out of the 15 students, four of them, namely students 4654, 4655,
Marlén Izquierdo and Naroa Zubillaga 23
Hikma 24(1) (2025), 1 - 31
4656, 4659, reported that Basque was their one and only mother tongue. Such
students “consume” Basque daily, in various facets of life; Basque is their
formal medium of instruction, their home, peer, and leisure language.
Figure 6. TAS first-level annotations per EN-EU trainee
Source. Elaborated by the authors
As can be seen, it is two of the students who claimed Basque to be their
only mother tongue that have received the smallest number of language
annotations (students 4654 and 4656). Accordingly, we would argue that, in a
diglossic environment, knowing the “low” language does not guarantee a
proficient command by the speakers. Rather a native-like command stemming
from an all-round use of the language appears to be essential. Previous
research on language acquisition corroborates this fact (Ezeizabarrena
Segurola et al., 2009; Ezeizabarrena Segurola, 2012; Zawiszewski et al,
2011). It seems, therefore, that exposure to the language, its varieties and
different registers is a key factor to master the language. Given the exploratory
nature of this case study, these results cannot be generalised. However, they
hint at a possibility worthy of further exploration; our contrastive descriptive
study might contribute to the field of empirical translation studies by identifying
a niche of research where the actual amount and type of use of contact
languages in bilingual settings and the language quality of translation students
may be correlated.
CONCLUSION
The aim of this article was to report the preliminary results of a
contrastive descriptive translation study from a learner corpus approach.
Within the framework of the MUST project, two sets of students were asked
24 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
to translate the same source text from English into Basque and/or Spanish,
the learners’ assumed mother tongue or the A language they chose in their
university degree. Generally speaking, translation trainees working into
Spanish seem to have more problems with content transfer, while Basque
students show difficulties at using the target language. Nevertheless, it is also
observed that translation at the lexical level is prone to less divergence, across
target languages, than at the discourse level.
Another conclusion we reach is that TAS 1.0 cannot escape the
subjectivity of the human eye that necessarily examines each and every
translation in search of errors and/or options worth annotating. As a matter of
fact, TAS 2.0. has deleted some levels of annotation as well as ambiguous
tags, displaying less granularity. In the same way that human skills and
limitations give shape to a translation, it is a human mind with an
idiosyncratically evaluative attitude that judges the degree of correctness,
acceptability and/or appropriateness of the translation under correction.
In addition to describing the actual translational options suggested by
the learners, we also aimed to understand the outstanding difficulties in
producing grammatically correct Basque translations. At this stage, the
students’ metadata our corpora are enriched with were essential to figure out
that a poorer exposure to the Basque language in the diglossic context our
learners live and learn might account for the many LA-related problems. Being
the dominated language, Basque is not experienced as much or as diversely
as is Spanish. On the other hand, being the dominant language, Spanish may
intrude in the trainee’s Basque output as a consequence of some cognitive
interference in bilingual settings.
Our study has been an exploratory one, yet it contributes to existing
research in various ways. First, it adds a new trend in contrastive studies
whereby the translations into two target languages of the same source text
are contrasted. In addition, it integrates learner data into contrastive and
translation studies. Most importantly, it has taken a step further by trying to
understand translation products from a wider perspective, using the trainees’
metadata as indicative of the process of translation. Where the trainees claim
Basque to be their one and only L1, they perform better. Finally, as expected,
this preliminary study has shed light on some limitations that need addressing
before embarking on larger contrastive descriptions of learner translations.
For example, our metadata being common to the MUST project, it has been
insufficient to fully reflect the sociolinguistic or socio-political status of minority
languages like Basque. Even a clear explanation regarding what L1, L2 or
foreign language should be understood as to trainees seems appropriate prior
to collecting their metadata. On the light of the results, we also believe that
understanding TAS tags is one part of the annotation process only; foreseeing
Marlén Izquierdo and Naroa Zubillaga 25
Hikma 24(1) (2025), 1 - 31
the co-text where certain tags may be relevant is also a factor for the coders
to agree on.
Finally, further research may be taken on the results obtained. For
example, a similar analysis of larger phraseological data would attest whether
it is really the case that cross-linguistic differences at the lexical level are
fewer. Likewise, a comparison between TAS-annotated human and machine
translations could help us understand how determinant the human variable
may be. On a pedagogical note, the same Hypal4Must software we have used
to process and annotate the translations remains to be used in the translation
class. The benefit of this would be for the trainees to learn from their own
translations as well as previously done learner translations. Browsing a
students’ corpus, no matter the language combinations, trainees would
become aware of potential equivalents, common errors, and a range of
translational strategies or solutions.
ACKNOWLEDGMENTS
Research for this paper has been done in the framework of the
TRALIMA-ITZULIK Research Group, GIU21/060, sponsored by the University
of the Basque Country (UPV/EHU).
We would like to thank the two anonymous reviewers for their
comments, which have helped us improve a first draft of this paper.
We would also like to thank the translation and interpreting students
who kindly consented to our using their translations as part of the MUST
corpus project.
REFERENCES
Alfuraih, R. F. (2020). The undergraduate learner translator corpus: a new
resource for translation studies and computational linguistics. Lang
Resources & Evaluation, 54, 801-830. https://doi.org/10.1007/s10579-
019-09472-6
Almgren, M., & Manterola Garate, I. (2016). The development of narrative
skills in learners of Basque as a second language. Education Inquiry,
7(1), 27-46. https://doi.org/10.3402/edui.v7.27627
Austin, J. (2009). Delay, interference and bilingual development: The
acquisition of verbal morphology in children learning Basque and
Spanish. International Journal of Bilingualism, 13(4), 447-479.
10.1177/1367006909353234
26 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
Baker, M. (1995). Corpora in translation studies: An overview and some
suggestions for future research. Target, 7(2), 223-243.
https://doi.org/10.1075/target.7.2.03bak
Barnes Mason, J., & García Fernández, I. (2011). Vocabulary growth and
composition in monolingual and bilingual Basque infants and toddlers.
International Journal of Bilingualism, 17(3), 357-374.
10.1177/1367006912438992
Basque Government. (1983). Decreto 138/1983, por el que se regula el uso
de las lenguas oficiales en la enseñanza no universitaria en el País
Vasco. https://www.legegunea.euskadi.eus/eli/es-
pv/d/1983/07/11/138/dof/spa/html/webleg00-contfich/es/
Basque Government. (2016). VI. inkesta soziolinguistikoa.
https://www.irekia.euskadi.eus/uploads/attachments/9954/VI_INK_SO
ZLG-EH_eus.pdf?1499236557
Bernal, J. M. (2001). Diglosia y funciones sociales de las lenguas en Grecia
(1830-1941). Minerva: Revista de Filología Clásica, 15, 115-116.
Bowker, L. (1999). Exploring the potential of corpora for raising language
awareness in student translators. Language Awareness, 8(3-4), 160-
173. https://doi.org/10.1080/09658419908667126
Bowker, L. (2003). Corpus-based applications for translator training: exploring
the possibilities. In S. Granger, J. Lerot, & S. Petch-Tyson (Eds.)
Corpus-based approaches to contrastive linguistics and translation
studies (pp. 169-183). Rodopi.
https://doi.org/10.1163/9789004486638_014
Davies, E. (4th of January, 2021). To solve the problem of loneliness, society
needs to look beyond the nuclear family. The Guardian.
https://www.theguardian.com/commentisfree/2021/jan/04/loneliness-
society-nuclear-family-lockdown-communities
De Sutter, G., & Lefer, M.A. (2019). On the need for a new research agenda
for corpus-based translation studies: A multi-methodological,
multifactorial and interdisciplinary approach. Perspectives, 28(1), 1-23.
https://doi.org/10.1080/0907676X.2019.1611891
Doval Reixa, I., & Sánchez-Nieto, M. (Eds.). (2019). Parallel corpora for
contrastive and translation studies. New resources and applications.
John Benjamins.
Marlén Izquierdo and Naroa Zubillaga 27
Hikma 24(1) (2025), 1 - 31
Espunya Prat, A. (2014). The UPF learner translation corpus as a resource
for translator training. Lang Resources & Evaluation, 48, 33-43.
https://doi.org/10.1007/s10579-013-9260-1
Ezeizabarrena Segurola, M. J. (2012). The acquisition of the (in)consistent
ergative marking in Basque: L1 and early L2. Lingua, 122(3), 303-317.
https://doi.org/10.1016/j.lingua.2011.11.009
Ezeizabarrena Segurola, M. J., Manterola Garate, I., & Beloki Lizarralde, L.
(2009). Euskara H2 goiztiarraren ezaugarrien bila: adizkiak eta
gramatika-kasuak haurren ipuin-kontaketetan. Euskera, 54(2-1), 639-
681.
Ferguson, C. (1959). Diglossia. Word, 15, 325-340.
https://doi.org/10.1080/00437956.1959.11659702
Fictumova, J., Obrusnik, A., & Stepankova, K. (2017). Teaching specialised
translation. Error-tagged translation learner corpora. Sendebar, 28,
209-241.
Graham, M. J., Milanowski, A. T., & Miller, J. B. (2012). Measuring and
promoting inter-rater agreement of teacher and principal performance
ratings. Center for Educator Compensation Reform.
https://files.eric.ed.gov/fulltext/ED532068.pdf
Granger, S., & Lefer, M. A. (Eds.). (2020). The complementary contribution of
comparable and parallel corpora to crosslinguistic studies. Languages
in Contrast, 20(2).
Granger, S., & Lefer, M. A. (October, 2021). Translation-oriented annotation
system manual version 2.0. Centre for English Corpus Linguistics –
UCLouvain.
Granger, S., Lerto, J., & Petch-Tyson, S. (2003). Corpus-based approaches
to contrastive linguistics and translation studies. Rodopi.
Hareide, L. (2019). Comparable parallel corpora: A critical review of current
practices in corpus-based translation studies. In I. Doval Reixa, & M.
Sánchez-Nieto (Eds.), Parallel Corpora for contrastive and translation
studies (pp. 19-38). John Benjamins.
Holmes, J. (1972). The name and nature of translation studies. In J. Holmes
(Ed.), Translated! Papers on literary translation and translation studies
(pp. 67-80). Rodopi.
Hunston, S. (2002). Corpora in applied linguistics. CUP.
28 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
Izquierdo Fernández, M. (2007). Corpus-based cross-linguistic research:
directions and applications. Interlingüística 17, 520-527.
Izquierdo Fernández, M, & Zubillaga Gómez (2022). Empirical translation
studies: contrasting learner translations in a diglossic environment.
Paper presented at the 6th Learner Corpus Research Conference, 22-
24 September 2022, Padova (Italy).
James, C. (1980). Contrastive analysis. Longman.
Johansson, S. (2007). Seeing through multilingual corpora: on the use of
corpora in contrastive studies. John Benjamins.
Kotze, H. (2019). Converging what and how to find out why: An outlook on
empirical translation studies. In H. Kotze (Ed.), New empirical
perspectives on translation and interpreting (pp. 333-371). Routledge.
Kreszwoski, T. P. (1990). Contrasting languages. The scope of contrastive
linguistics. Mouton de Gruyter.
Lavid-López, J., Maíz-Arévalo, C., & Zamorano-Mansilla, J. R. (Eds.). (2021).
Corpora in translation and contrastive research in the digital age. John
Benjamins.
Leonet, O., Cenoz Iragui, J., & Gorter, D. (2017). Challenging minority
language isolation: translanguaging in a trilingual school in the Basque
Country. Journal of Language, Identity and Education, 16(4), 216-227.
https://doi.org/10.1080/15348458.2017.1328281
Marco Borillo, J. (2019). Living with parallel corpora. The potentials and
limitations of their use in translation research. In I. Doval, & M. Sánchez-
Nieto (Eds.), Parallel corpora for contrastive and translation studies.
New resources and applications (pp. 39-56). John Benjamins.
Ministerio de Educación, Cultura y Deporte. (2011). Memoria de grado de
Traducción e Interpretación en la UPV/EHU. https://gestion-
alumnos.ehu.es/tmp/Memoria%20Verificada%2006-02-2014.pdf
Obrusník, A. (2014). Hypal: A user friendly tool for automatic parallel text
alignment and error-tagging. Paper for the 11th Teaching and Language
Corpora Conference (TALC 2014).
Olohan, M. (2003). How frequent are the contractions?: A study of contracted
forms in the translational English corpus. Target, 15(1), 59-89.
10.1075/target.15.1.04olo
Rabadán Álvarez, R. (2007). Divisions, description and applications: The
interface between DTS, corpus-based research and contrastive
Marlén Izquierdo and Naroa Zubillaga 29
Hikma 24(1) (2025), 1 - 31
analysis. In Y. Gambier, M. Shlesinger, & R. Stolze (Eds.), Doubts and
directions in translation studies: Selected contributions from the EST
Congress, Lisbon 2004 (pp. 237-252). John Benjamins
SanzVillar, Z. (2018). Interference and the translation of phraseological units
in a parallel and multilingual corpus. META, 63(1), 72-93.
https://doi.org/10.7202/1050515ar
Sanz Villar, Z. (2024). German-to-Basque translation analysis of multiword
expressions in a learner translation corpus. Íkala, Revista De Lenguaje
y Cultura, 29(1), 1-21. https://doi.org/10.17533/udea.ikala.354417
Tiayon, C. (2004). Corpora in translation teaching and learning. Language
Matters, 35(1), 119-132. https://doi.org/10.1080/10228190408566207
Urdalleta Lete, I. (8 September, 2023). Hegoaldeko hamar ikasletik seik
ikasiko dute D ereduan. Berria. https://www.berria.eus/euskal-
herria/hegoaldeko-hamar-ikasletik-seik-ikasiko-dute-d-
ereduan_1341536_102.html
Zawiszewski, A., Gutiérrez Sigut, E., Fernández Fernández, B., & Laka
Mugarza, I. (2011). Language distance and non-native syntactic
processing: Evidence from event-related potentials. Bilingualism:
Language and Cognition, 14(3), 400-411.
10.1017/S1366728910000350
Zubillaga Gómez, N. (2016). (In)direct offense. A comparison of direct and
indirect translations of German offensive language into Basque.
Perspectives: Studies in Translatology, 24(3), 486-497.
10.1080/0907676X.2015.1069858
Zubillaga Gómez, N, Sanz Villar, Z. & Uribarri Zenekorta, I. (2015). Building
a trilingual parallel corpus to analyse literary translations from German
into Basque. In C. Fantinuoli & F. Zanettin (Eds.) New directions in
corpus-based translation studies (pp. 71–93). Language Science
Press.
APPENDIX 1
Source text (ST) provided to the learners for their translation into
Basque and Spanish:
To solve the problem of loneliness, society needs to look beyond the
nuclear family
Eli Davies
30 Learner Translations in Contrast: An English-Basque-Spanish Case Study
Hikma 24(1) (2025), 1 - 31
Like so many living alone during lockdown, I’ve felt incredibly isolated.
It’s time to rethink the way communities work.
“I have been trying, for some time now, to find dignity in my loneliness,”
wrote the poet and critic Maggie Nelson in her 2009 book Bluets. When I read
these words, they struck a chord.
I’ve been thinking a lot about loneliness over the last few years as I’ve
drifted in and out of various forms of it myself, the most extreme form coming,
unsurprisingly, in 2020. Is there dignity to be had in it? Perhaps not, and
perhaps that’s why we find it so difficult to talk about or admit to.
A few months into the Covid-19 outbreak, people started to talk about
a corresponding “loneliness pandemic”. Pages appeared on the NHS and Red
Cross websites advising how to cope with the isolation thrust upon us by the
global health crisis and its accompanying lockdown. But before this year,
various reports claimed that loneliness had reached dangerous and even life-
threatening epidemic levels, and in 2018 Theresa May launched a UK
government “loneliness strategy”. Such concerns have always been
particularly heightened during winter and around Christmas, a time when
charities and politicians frequently urge festive revellers to think about and
reach out to the lonely and vulnerable.
There isn’t much talk in all this, though, of what loneliness actually is,
what it feels like or where it comes from. In these scenarios it is an affliction:
distant, othered and slightly frightening, coming to us in the form of elderly
people at Christmas, the recently widowed, those who are unloved or
forgotten about. But thinking about it as some kind of disease is wrongheaded.
The historian Fay Bound Alberti, who has written a “biography” of the
condition, argues that this way of thinking suggests “that it’s coming from the
outside, rather than being something that is a social problem.”
Loneliness, then, is partly produced by the way we organise the world
– and to address it we need to seriously rethink how we approach our public
spaces, housing arrangements and relationships. This includes questioning
our dependency on certain forms of relationship – the couple and the nuclear
family – as units of social organisation.
I first started properly thinking about all this around the summer of 2017.
I had recently come out of a 12-year relationship and, not unrelatedly, had
moved to Ireland to work on my PhD, that most solitary of endeavours. After
years of coupled domesticity, I was living alone. Solitude is not the same as
loneliness, of course – as Nelson puts it, “loneliness is solitude with a problem”
– and some of this was OK: I read, I walked, I wrote, I went out and made new
Marlén Izquierdo and Naroa Zubillaga 31
Hikma 24(1) (2025), 1 - 31
friends. But my isolation, coupled with the rawness of a recent heartbreak,
frequently was a problem.
At such times I would message friends and family back home, filling my
phone screen with three, four, five WhatsApp chats, and in these exchanges
they described their own struggles: too much to do, not enough time or space
for themselves, an excess of people and stuff to look after. The contrast in our
predicaments seemed, at times, completely absurd and, above all, wasteful. I
often wondered whether it wouldn’t make sense to bolt my household on to
one of theirs, to redistribute some of my caring resources and they, in turn, to
share some of the human company that I often craved.
As indicated, this is a shortened version of the original text, which can
be found at To solve the problem of loneliness, society needs to look beyond
the nuclear family | Eli Davies | The Guardian.
