An I for an AI: A Corpus-Based Study on the (In)direct Translation of Japanese First-Person Singular Pronouns into Finnish

(1)

UNIVERSITY OF EASTERN FINLAND PHILOSOPHICAL FACULTY

SCHOOL OF HUMANITIES English Language and Translation

Henry Laine

AN I FOR AN AI: A CORPUS-BASED STUDY ON THE (IN)DIRECT TRANSLATION OF JAPANESE FIRST-PERSON SINGULAR

PRONOUNS INTO FINNISH

MA Thesis

February 2021

(2)

ITÄ-SUOMEN YLIOPISTO – UNIVERSITY OF EASTERN FINLAND

Tiedekunta – Faculty Philosophical Faculty

Osasto – School School of Humanities Tekijät – Author

Henry Laine Työn nimi – Title

An I for an AI: A Corpus-Based Study on the (In)direct Translation of Japanese First-Person Singular Pronouns into Finnish

Pääaine – Main subject Työn laji – Level Päivämäärä – Date

Sivumäärä – Number of pages

English Language and Translation

Pro gradu -tutkielma x

February 25, 2021

89 pages Sivuainetutkielma

Kandidaatin tutkielma Aineopintojen

tutkielma Tiivistelmä – Abstract

The descriptive branch of translation studies has offered us new tools to categorize and predict translational behavior. On the one hand, we have translation norms, which describe and predict how a specific group of translators tend to act. On the other hand, we have translation universals, which are hypotheses on the nature of all translation. This thesis study examines two translation universals in particular, namely standardization, which predicts that translators tend to replace less common source- language elements with more common target-language ones, and the unique item hypothesis, which predicts that translators do not tend to employ those features of the target-language that lack counterparts in the source language.

The two universals are investigated by examining the translation of Japanese first-person singular pronouns into Finnish both directly from the original source language and indirectly via an intermediate language, namely English. Many of such pronouns lack counterparts in English but have (near) equivalents in Finnish. The Finnish pronouns are, therefore, unique when compared to English but not unique when compared to Japanese. By changing the source language (Japanese or English), it is possible to see whether there is a change in the frequencies of different Finnish personal pronouns as translation equivalents and whether direct and indirect translators follow different translation norms.

The material consists of eight Japanese novels and their twelve translations: four direct Finnish translations, four direct English translations, and four indirect Finnish translations. A sample of 286 Japanese pronouns and 445 target-language renditions has been extracted from the material both manually and with corpus tools. The translators’ solutions have been then categorized into three translation strategies based on the model presented by Mauranen and Tiittula (2005): these strategies are retention, omission, and paraphrase. Furthermore, the retention occurrences have been categorized based on the pronoun variants used by the Finnish translators.

The results show that both English and Finnish translators have preferred employing the strategy of retention, i.e., replacing a source-language pronoun with an equivalent target-language one. The Finnish translators have also used the strategy of omission extensively, i.e., omitted the source pronoun but indicated the first-person singular with a personal suffix. The strategy of paraphrase, in turn, is scarce in the material, i.e., the translators have avoided completely omitting references to the first-person singular.

The results also show that the Finnish translators have tended to choose the standardized target pronoun minä instead of any of the non-standardized ones (mä, mää, mie, etc.). Therefore, the results of this study do not concur with the unique item hypothesis, but they fall in line with the universal of standardization.

Avainsanat – Keywords

indirect translation, standardization, unique item hypothesis, translation norms, personal pronouns

(3)

ITÄ-SUOMEN YLIOPISTO – UNIVERSITY OF EASTERN FINLAND

Tiedekunta – Faculty Filosofinen tiedekunta

Osasto – School Humanistinen osasto Tekijät – Author

Henry Laine Työn nimi – Title

An I for an AI: A Corpus-Based Study on the (In)direct Translation of Japanese First-Person Singular Pronouns into Finnish

Pääaine – Main subject Työn laji – Level Päivämäärä – Date

Sivumäärä – Number of pages

Englannin kieli ja kääntäminen

Pro gradu -tutkielma x

25.2.2021 89 sivua Sivuainetutkielma

Kandidaatin tutkielma Aineopintojen

tutkielma Tiivistelmä – Abstract

Tämä tutkielma tarkastelee käännösuniversaaleja, käännösnormeja ja epäsuoraa kääntämistä.

Käännösuniversaaleista tutkimuksen kannalta oleellisimmat ovat standardisoituminen ja uniikkiaineshypoteesi. Standardisoitumisen mukaan käännökset ovat lähdetekstejään yleiskielisempiä, kun taas uniikkiaineshypoteesin mukaan käännöksissä aliedustuvat kohdekielelle ominaiset piirteet, jos lähdekielessä ei ole niille vastineita. Tarkempana tutkimuskohteena ovat japanin yksikön ensimmäisen persoonan pronominit ja niiden englannin- ja suomenkieliset käännösvastineet.

Tutkimus pyrkii selvittämään, miten japanin kielen pronomineja käännetään niin suoraan alkuperäiskielestä kuin epäsuorasti välikielenä toimivan englannin kautta eli mitkä ovat pronominien kääntämistä koskevat käännösnormit. Useimmilla japanin pronomineilla ei ole vastineita englannissa, mutta niille löytyy (osittaisia) vastineita suomesta. Englantiin verrattuna suomalaiset pronominit ovat näin ollen uniikkeja. Sen sijaan japaniin verrattuna ne eivät ole uniikkeja. Eri lähdekielistä käännettyjä tekstejä vertaamalla onkin mahdollista saada selville, vaikuttaako lähdekieli (japani tai englanti) kääntäjien pronominivalintoihin.

Tutkimusaineisto koostuu kahdeksasta japaninkielisestä romaanista ja niiden kahdestatoista käännöksestä: neljästä suorasta suomenkielisestä käännöksestä, neljästä suorasta englanninkielisestä käännöksestä ja neljästä epäsuorasta suomenkielisestä käännöksestä. Aineistosta on sekä manuaalisesti että korpustyökaluja hyödyntäen poimittu otos, joka kattaa 286 lähdekielistä pronominiesiintymää ja 445 kohdekielistä käännösvastinetta. Vastineet on jaoteltu Maurasen ja Tiittulan (2005) mallin mukaisesti kolmeen käännösstrategiakategoriaan: näitä kutsutaan tutkielmassa nimityksillä säilytys, poisto ja parafraasi. Lisäksi säilytysesiintymät on jaoteltu sen mukaan, mitä yksikön ensimmäisen persoonan pronomineja suomalaiset kääntäjät ovat käyttäneet lähdekielisten pronominien vastineina.

Tulokset osoittavat, että englannin- ja suomenkieliset kääntäjät ovat käyttäneet ensisijaisena strategianaan säilytystä eli he ovat korvanneet lähdekielen pronomineja kohdekielen yksikön ensimmäisen persoonan pronomineilla. Suomalaiset kääntäjät ovat lisäksi turvautuneet poistoon, eli he ovat häivyttäneet pronominin mutta ilmaisseet persoonasuffiksin avulla yksikön ensimmäisen persoonan.

Parafraasia on sen sijaan käytetty harvakseltaan kaikissa käännöksissä, eli kääntäjät ovat vältelleet häivyttämästä kokonaan viittauksia yksikön ensimmäiseen persoonaan. Tulokset osoittavat myös, että suomen pronomineista kääntäjät ovat suosineet yleiskielistä pronominia minä eivätkä ole valinneet käännösvastineiksi arkikielisiä pronomineja (esim. mä, mää ja mie). Tämän takia tutkimuksen tulokset eivät ole uniikkiaineshypoteesin mukaisia, mutta ne tukevat standardisoitumisen universaalia.

Avainsanat – Keywords

epäsuora kääntäminen, standardisoituminen, uniikkiaineshypoteesi, käännösnormit, persoonapronominit

(4)

1. Introduction

The rise of machine-readable text corpora and their eventual adoption in translation studies have made it possible to study the phenomenon of translation more systematically, since scholars can now automate some of the phases – such as data extraction – of a translation study. Furthermore, corpora can be used to investigate new translational phenomena. Two such objects of study have been suggested by Baker (1993), namely universal features of translation and (descriptive) translation norms. According to Baker (1993: 243), universal features of translation, or translation universals, are “features which typically occur in translated text rather than original utterances and which are not the result of interference from specific linguistic systems.” These universals include such phenomena as lengthening and simplification (see Chesterman 2011). However, as has been pointed out by Chesterman (2011: 176), these should be seen as generalizations about translation formulated as hypotheses, which will need to be validated or invalidated by further research.

Translation norms, in turn, are a central concept of the descriptive branch of translation studies, which has been promoted in particular by Toury (e.g., 1980; 1995). He defines translation norms as “the translation of general values or ideas shared by a community […]

into performance instructions appropriate for and applicable to particular situations” (Toury 1995: 55). He also points out that the breaking of norms can lead to sanctions in situations where an individual has the option of choosing from different types of behavior, such as in translation (Toury 1995: 55). Norms, therefore, encourage a certain kind of behavior while discouraging other kinds.

Since translation is a norm-governed activity (see Toury 1995), translation norms can offer us an explanation why certain translators act in certain ways in certain situations. If most Finnish translators, for instance, favor source-oriented translation strategies when they translate Japanese literature, it can be said that the norm is to employ said strategies and that the Finnish translators are conforming to the said norm. However, we do not know what many of the current (or past) translation norms are (cf. the above-mentioned example), and therefore there is a need to map such norms. These norms can then be used to explain and predict translational behavior.

This thesis centers around the two above-mentioned phenomena, but it also combines them with a less studied phenomenon in translation studies, namely indirect translation (other terms include pivot translation, relay translation, and second-hand translation; see Assis

(6)

2

Rosa, Pięta, & Bueno Maia 2017: 115). A major characteristic of indirect translation, as opposed to direct translation, is that the source text is a translation of an original work, i.e., an intermediate text. Therefore, we can ask the question of whether the use of an intermediate language, i.e., the language of the intermediate text, results in a translation that is somehow different from a comparable direct translation. In other words, it is unclear whether there are differences in translation norms between translators who translate comparable source- language texts directly or indirectly into one and the same target language. (In this thesis, these two types of translators are denoted by the terms direct translator, i.e., a person translating directly from the language of the original, and indirect translator, i.e., a person translating indirectly via an intermediate language.)

In order to study these three broad translational phenomena, there is a need to examine a more specific phenomenon that can be identified at the textual level. In the case of this study, the specific phenomenon is Japanese first-person singular pronouns. More specifically, the aim is to investigate how Finnish translators translate Japanese first-person singular pronouns into Finnish either directly from Japanese or indirectly via an intermediate language, which is in this case English. Consequently, the research material consists of English and Finnish translations and their Japanese source texts. These source and target texts are novels, and the former have been compiled into an electronic corpus. The research aim can be summarized with the three research questions presented below:

1. How are first-person singular pronouns rendered in the direct English material, in the direct Finnish material, and in the indirect Finnish material?

2. In what ways are the two Finnish materials similar or dissimilar in regard to the rendering of first-person singular pronouns?

3. Can the causes of the potential dissimilarities between the direct and indirect Finnish material be traced back to the direct English material?

Since this study also investigates translation universals, as mentioned above, we can use these as the basis for our hypotheses. In particular, two translation universals are central to this study. Firstly, it has been hypothesized by Tirkkonen-Condit (2002: 209) that those target- language linguistic items that are not present in the source language tend to be under- represented in the target text. This universal statement is known as the unique item hypothesis.

Since the intermediate language of this study, i.e., English, lacks the multiple first-person

(7)

3

singular pronouns that both Japanese (watashi, ore, boku, etc.) and Finnish (minä, mä, mie, etc.) have, it is justified to assume that direct Finnish translations are characterized by larger variation in first-person singular pronouns than indirect Finnish translations are. Secondly, this hypothesis coincides with another translation universal, namely the law of growing standardization: “in translation, textual relations obtaining in the original are often modified, sometimes to the point of being totally ignored, in favour of [more] habitual options offered by a target repertoire” (Toury 1995: 268; square brackets in the original). Therefore, it can be assumed that all three target-text types suffer from standardization, i.e., that many Japanese pronouns have been converted, or standardized, into the common target-language variants, namely I and minä, or have been replaced with other standardized textual elements, such as Finnish personal suffixes (e.g., ones attached to predicates: juon ‘I drink’).

There is one more universal that can be tested with the current research model, namely explicitation, which, simply put, hypothesizes that translations are more explicit than their source texts (see Klaudy 1998). If this is applied to the translation of first-person singular pronouns, the prediction is that the target texts contain more information specifying the speaker. However, due to the nature of first-person singular pronouns and the differences between the source (Japanese) and target-language (English and Finnish) pronominal systems, it is more likely that the universal of explicitation does not apply in the case of this study.

Firstly, it is difficult to render first-person singular pronouns more explicitly without breaking the rules of idiomaticity. For instance, it is unnatural for a person to refer to him- or herself in the third person (cf. I went shopping and Peter went shopping, i.e., the speaker is Peter).

Secondly, neither English nor Finnish has first-person singular pronouns that contain more semantic or pragmatic information than their closest Japanese counterparts. For instance, many Japanese first-person singular pronouns indicate the gender of the speaker (e.g., the feminine atashi), whereas similar English and Finnish pronouns are gender neutral (e.g., I and minä). Therefore, in this regard, the actual hypothesis of this study is that there is a decrease in the level of explicitness when first-person singular pronouns are translated from Japanese into English or Finnish.

This thesis has been divided into sections that examine the above-mentioned translational phenomena either from a theoretical or analytical perspective. The theoretical framework of this study is introduced in sections 2, 3, 4, and 5: section 2 centers around translation universals, section 3 discusses translation norms, section 4 concerns indirect translation, and

(8)

4

section 5 examines culture-bound translation problems, which include Japanese personal pronouns. In turn, sections 6 and 7 concentrate on the actual thesis study: section 6 presents the material and methods, and section 7 includes the analysis and results. Finally, section 8 brings together the theoretical and analytical parts in the form of a concluding discussion.

(9)

5

2. Translation universals

One of the concepts relevant to the descriptive branch of translation studies is translation universals. They are generalizations about the process of translation and its products, i.e., translations, and can therefore both explain and predict translational behavior. Such generalizations constitute the focal point of this section, which first discusses some general aspects of translation universals (section 2.1), such as terminological and typological issues, and then takes a closer look at four different universals due to their relevance to this thesis study. These are explicitation (section 2.2) and standardization (section 2.3) as well as interference and a special type of it, namely the unique item hypothesis (section 2.4).

2.1. Universal tendencies

The quest for the universals of translation had been going on even before the advent of computerized corpora (see Chesterman 2004b), but it was a paper by Mona Baker (1993) which really prompted (corpus-based) research on translation universals. In her paper, Baker (1993: 243) defines translation universals as “features which typically occur in translated text rather than original utterances and which are not the result of interference from specific linguistic systems.” In this definition, the keyword is typically, since it implies that universal features of translations do not need to apply to each and every translation. In fact, Malmkjær (2008: 53) has stressed that Baker’s (1993) notion of translation universals is compatible with the notion of linguistic universals.

Indeed, according to Song (2001: 6), linguistic universals can be categorized into absolute and non-absolute ones: absolute universals have no exceptions and must therefore be applicable to all languages; non-absolute universals (or universal tendencies), in turn, are applicable to most languages. An example of the former universal type is that “[a]ll languages have pronominal categories involving at least three persons and two numbers,” whereas an example of the latter type is that “[i]n declarative sentences with nominal subject and object, the dominant order is almost always one in which the subject precedes the object” (Greenberg 1963: 61, 75).

Although linguistic universals cover both absolute rules and general tendencies, there has been some debate in translation studies on whether universal features should be inherent to all

(10)

6

translations or just most of them. Tymoczko (1998: 655, 659), for instance, argues that in order to make truly universal claims, we would have to have access to translations produced in all languages, cultures, and periods, such as prehistoric translations. She further points out that even if we had access to these translations, the universal generalizations would be so general that they likely would be useless to translation studies. Mauranen (2008: 35), on the other hand, has pointed out that generalizations can be formulated based on limited available data, as has been done in other fields of study. Indeed, any branch of research that employs statistical methods can make generalizations based on a smaller sample representing a larger population of entities, such as all human beings or all translations.

Another problem with the stricter view on translation universals is that due to empirically confirmed exceptions, we would have to discard most (if not all) of the proposed universals.

Baker (1996: 176–177), for instance, lists four translation universals. A more detailed taxonomy has been proposed by Chesterman (2004a: 39–40), who categorizes universals based on differences between translations and their source texts, on the one hand, and between translations and comparable target-language originals (i.e., non-translations), on the other. Therefore, if a universal concerns differences between translations and source texts, Chesterman calls it an S-universal (S for source); if a universal concerns differences between translations and non-translations, he calls it a T-universal (T for target). The S-universals in his taxonomy include lengthening, interference, standardization, dialect normalization, explicitation, reduction of repetition, and the retranslation hypothesis, whereas the T- universals cover simplification, conventionalization, untypical lexical patterning, and the unique item hypothesis.

There are, however, some issues concerning the categorization of universals. On the one hand, some universals can be seen as both S- and T-universals, such as explicitation (i.e., translations tend to be more explicit than source texts or non-translations; see section 2.2). On the other hand, some proposed universals overlap with other proposed universals. For instance, dialect normalization, which hypothesizes that dialectal language is transformed into non- dialectal language in translation (see Englund Dimitrova 1997), is a subcategory of standardization, which, in turn, hypothesizes that translations contain more standardized language, i.e., more common language elements, than their source texts do (see Toury 1995;

section 2.3 in this thesis).

(11)

7

Besides the typological issues, there are also issues more closely related to translation universals per se. There is, for instance, the issue that some universals contradict each other.

Pym (2008: 319) has pointed out that, for example, the universals of simplification and explicitation are contradictory, since the former involves the shortening of sentences, whereas the latter involves adding additional information to translations and, therefore, results in the lengthening of sentences. Indeed, some universals, such as standardization and explicitation, may predict contradicting results when we examine single phenomena, such as extralinguistic culture-bound elements (see Laine 2020), but when we examine translations as a whole, it is difficult to determine which of the contradicting universals is the dominant one. It is also possible that some of the proposed universals are not as universal as others but may be conditioned. For instance, if the source culture is relatively unknown in the target culture, translators are more likely to standardize than to explicitate (this is just a hypothetical example based on the observations reported in Laine 2020).

Regardless of the problems discussed above, there is a need to do more research on translation universals, since, at this point, they are still commonly considered to be hypotheses on translation tendencies (see e.g. Chesterman 2011: 176) and, therefore, need to be validated or invalidated by further research. However, due to the sheer number of suggested universals, it is not possible to discuss them all here. Therefore, the next aim is to cover those universals that are most central to this thesis study, namely explicitation, standardization, interference, and the unique item hypothesis.

2.2. Explicitation

The universal of explicitation has its origins as a translation strategy introduced by Vinay and Darbelnet (1995 [1958]: 342): “A stylistic translation technique which consists of making explicit in the target language what remains implicit in the source language because it is apparent from either the context or the situation.” Explicitation was later proposed to be a universal feature of translation by Blum-Kulka 2000 [1986]), who examined explicitation on the level of discourse. More specifically, she investigated shifts of cohesion and coherence (i.e., covert and overt textual markers) in translation from English into French and Hebrew.

She then formulated the so-called explicitation hypothesis:

(12)

8

The process of interpretation performed by the translator on the source text might lead to a TL [target-language] text which is more redundant than the SL [source-language] text. This redundancy can be expressed by a rise in the level of cohesive explicitness in the TL text. This argument may be stated as “the explicitation hypothesis”, which postulates an observed cohesive explicitness from SL to TL texts regardless of the increase traceable to differences between the two linguistic and textual systems involved. It follows that explicitation is viewed here as inherent in the process of translation. (Blum-Kulka 2000: 300; italics in the original)

In its broadest sense, the explicitation hypothesis has been understood to mean that a target text is more explicit than its source text or a comparable text originally written in the target language (Pym 2005: 30). Therefore, in this regard, explicitation has been seen both as an S- universal (e.g., Øverås 1998; Marco 2012; Makkos & Robin 2014) and a T-universal (e.g., Vehmas-Lehto 1989; Puurtinen 2004).

There is still debate on what is meant by the term explicitation. It is apparent from Blum- Kulka’s (2000) definition that she considers explicitation to correlate with redundancy.

However, this has been criticized by Séguinot (1988: 106), who argues that Blum-Kulka’s definition is too narrow, since explicitation does not have to equal redundancy. Instead, Séguinot (1988: 106) suggests that explicitation can manifest itself in three ways in translation: something which is not present in the source text is added to the target text, something which is covert in the source text is expressed overtly in the target text, or a source text element is given a more central position in the target text. Séguinot (1988: 108) further argues that explicitation “should [...] be reserved in translation studies for additions in a translated text which cannot be explained by structural, stylistic, or rhetorical differences”

between the source and target language.

Séguinot’s (1988) strict view on the concept of explicitation has not been adopted by Klaudy (1993; 1998), who sees the translation shifts caused by structural differences as valid cases of explicitation. Her typology of explicitation shifts includes four categories, which are obligatory explicitation, optional explicitation, pragmatic explicitation, and translation- inherent explicitation. According to Klaudy (1998: 82–83), obligatory shifts are caused by the syntactic or semantic differences between the source and target language: if these shifts are not carried out by the translator, the target text will go against the grammatical rules of the target language. In Finnish, for instance, it is grammatically possible to omit the subject of a clause. This is not possible in English, and therefore a translator translating from Finnish into English has to make the implicit subject explicit in order to follow the grammatical rules of the target language (e.g., Menin kauppaan. → I went to the store.).

(13)

9

Optional explicitation shifts, in contrast, are not dictated by grammatical differences, since the translator can choose from multiple more or less explicit translation alternatives (Klaudy 1998: 83). Klaudy (1998: 83), however, points out that avoidance of optional shifts will result in translations that sound unidiomatic or unnatural. This might be correct for translations as a whole but not for certain linguistic items. For instance, a Finnish translator translating a Japanese subjectless clause can choose between keeping the subject implicit and making it explicit in the target text (e.g., Ie wo deta. → [Minä] lähdin kotoa. ‘I left the house.’). Both alternatives are idiomatic and grammatically correct, but the overuse of the latter will likely result in a translation which statistically differs from comparable texts originally written in Finnish (see e.g. Mauranen & Tiittula 2005).

The third explicitation category, i.e., pragmatic explicitation, has been pointed out to be a subcategory of optional explicitation by Englund Dimitrova (2005: 38). According to Klaudy (1998: 83), this type of explicitation shift is caused by the differences between the source and target culture: members of the target culture do not share the same cultural background and knowledge as members of the source culture. In order to make this implicit information more accessible to the target audience, translator can, for instance, employ the strategy of specification and add more linguistic material to a culture-bound element (e.g., Seinäjoki → the city of Seinäjoki; see Klaudy 1998: 83). There is, however, empirical evidence that translators do not tend to specify culture-bound elements (e.g., Pedersen 2011; Laine 2020), which does not support the explicitation hypothesis.

The fourth and final category of explicitation, i.e., translation-inherent explicitation, has been described by Klaudy (1998: 83) to be “attributed to the nature of the translation process itself.”

Englund Dimitrova (2005: 38) has suggested that this hypothetical category is a subcategory of obligatory explicitation. Becher (2011: 20), in turn, has pointed out that Klaudy’s (1998) fourth category mirrors the belief in translation studies that explicitation is inherent in translation, i.e., a translation universal. He also admits that it is difficult to find examples of translation-inherent explicitation because he does not know what they are supposed to look like (Becher 2011: 23). Indeed, Klaudy (1998) does not offer any concrete examples of this type of explicitation.

Regardless of the issues with the categorization of explicitation shifts, Klaudy’s (1998) categories of obligatory and optional explicitation allow us to categorize instances of explicitation into those that can be easily explained by grammatical differences and into those

(14)

10

that will have to be explained by other means (by idiosyncrasies, norms, universals, etc.). In fact, there is nothing stopping us from applying this method of categorization to other translation universals, such as standardization.

2.3. Standardization

The universal of standardization has been introduced into translation studies (or at least popularized) by Toury (1995: 268; square brackets in the original), who calls it the law of growing standardization and gives it the following definition: “in translation, textual relations obtaining in the original are often modified, sometimes to the point of being totally ignored, in favour of [more] habitual options offered by a target repertoire.” This has been interpreted by Pym (2008: 314) to mean that in translation, less common source-language features tend to be replaced by more common target-language features. A similar view is held by Leppihalme (2000b: 260), whose interpretation of standardization is that it entails replacing “unusual source elements [...] with more common target language elements.” Based on these interpretations, it seems that standardization is similar to what Kenny (2001: 65) calls normalization and defines as “the exaggerated use in translated texts of features that are typical of the target language.” Therefore, standardization (or normalization) can be seen not only as an S universal, but also as a T universal (see Kenny 2001: 65–66).

There is plenty of research on standardization, and many studies seem to examine standardization as an S universal by contrasting translations with their originals (e.g., Malmkjær 1998; Leppihalme 2000b; Kuusi 2016; Polcz 2016), although not all of these study standardization or translation universals explicitly. One case study where standardization is seen as an S universal has been carried out by Englund Dimitrova (1997), who has examined how dialect has been rendered in the English and Russians translations of Vilhelm Moberg’s Swedish novel Din stund på jorden (1963). In the case of both the English and Russian translation, she observed that dialectal language had been transformed into marked colloquial language and that phonological and orthographical markers had been changed into morphosyntactic and lexical markers, such as a syntactically complete sentence containing the dialectal contraction göra’t (göra det ‘to do it’) into a subjectless but otherwise standardized sentence (see Englund Dimitrova 1997: 59).

(15)

11

Based on these results, Englund Dimitrova (1997: 63) hypothesizes that translators tend to change informal or regionally specific language varieties into formal or regionally less specific language varieties. Therefore, a Finnish-language example of such a shift would be the replacement of the regionally specific pronouns mie (‘I’) and sie (‘you’; singular) with the regionally unspecific equivalents minä and sinä; these two pronoun types also exemplify the potential shift from informal and non-standardized language into formal and standardized one.

In addition to qualitative analysis, standardization has also been studied quantitatively with larger materials and corpus tools. Kenny (2001), for example, has examined creative word forms (e.g., Abwaschwassertee ‘dishwater tea’) and collocations (e.g., dreieckiges Auge

‘triangular eyes’) in the GEPCOLT corpus, which consists of German literary texts and their English translations. Her results show that of the 117 creative word forms and 59 creative collocations that she identified, 52 (44%) word forms and 13 (22%) collocations have been standardized when translated from German into English (see Kenny 2001: 117, 207). In other words, she found tentative evidence that translators have a tendency (to some degree) to replace less common language items with more common equivalents.

Another corpus-based study on standardization has been carried out by Nevalainen (2003).

Unlike Kenny’s (2001) study, however, Nevalainen’s (2003) approach has been to compare target-language originals and translations, i.e., to study standardization as a T universal. In fact, his research material consists of two monolingual subcorpora of the Corpus of Translated Finnish (see Mauranen 2000), and these in turn consist of 60 fictitious texts originally written in Finnish and 64 comparable texts translated into Finnish. Furthermore, in order to study standardization (he uses the term conservatism), he examines how often certain colloquial features, such as the first-person pronouns mä(ä) and sä(ä) as well as the colloquial equivalents of the word nainen (‘woman’), occur in the two Finnish corpora.

At a general level, the results of Nevalainen’s (2003) study show that colloquial expressions are used less in Finnish translations of fiction than in similar Finnish originals; there are also differences in what kind of colloquial expressions are preferred by authors of originals, on the one hand, and translators, on the other. In fact, similarly to Englund Dimitrova (1997; see above), Nevalainen (2003: 19) has observed that authors tend to prefer phonologically marked expressions, whereas translators tend to employ lexically marked ones. For instance, at a more specific level, Finnish authors of originals employ more often colloquial first-person singular

(16)

12

pronouns (e.g., mä) than Finnish translators do; they, in turn, use more interjections and expletives (e.g., ai ‘oh’ and jaa ‘well’) than Finnish authors (see Nevalainen 2003: 14, 17).

As we have seen, previous research on standardization suggests that translators have a tendency to avoid using less-common forms of expression. An explanation to such behavior has been suggested by Pym (2008: 325), who argues that standardization is one method of avoiding communicative risks. Indeed, if a translator is making a choice between retaining a source-language item (which may not be known by the target audience) and replacing it with a more common target-language item (which is known by the target audience), it is no wonder why he or she would pick the latter option, as it is the safer alternative from a communicative perspective. However, according to Pym (2008: 325), the other method of avoiding risks is to push some of the responsibility onto the author of the original by sticking closely the source text (see also Pym 2015). This, however, requires accepting some degree of interference from the source text or language.

2.4. Interference and the unique item hypothesis

The universal of interference has its origins in the fields of contact linguistics and second language acquisition as well as in contrastive analysis (see e.g. Mauranen 2004: 66). In contact linguistics, the concept of interference has been defined by Weinreich (1968 [1953]:

1) as “[t]hose instances of deviation from the norms of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language.” In second language acquisition, interference has been interpreted as learning difficulties and errors caused by the differences of an individual’s first and second language (see e.g. Lado 1957). In turn, the similarities between these two languages should facilitate language learning: “Those elements that are similar to his native language will be simple for him, and those elements that are different will be difficult” (Lado 1957: 2). This general influence of the first language on the second language is known as transfer, which, as explained by Lado (1964: 222), can be either positive or negative, and interference is synonymous with negative transfer. Therefore, for an English speaker learning Swedish, for example, the adjective–preposition pair bekant med (‘familiar with’) should be easier to learn (i.e., positive transfer) than the similar pair intresserad av (‘interested of’; cf. interested in; i.e., negative transfer or interference).

(17)

13

The notions of transfer and interference are not exclusive to second language acquisition but have also been adopted by translation studies. Although Baker’s early definition of a translation universal excludes interference (see Baker 1993: 243), it has still achieved the status of a translation universal, or a translation law as Toury (1995: 275) puts it: “[I]n translation, phenomena pertaining to the make-up of the source text tend to be transferred to the target text.” As opposed to second language acquisition, Toury’s concept of interference appears to be closer to transfer than to negative transfer. In fact, he argues that negative transfer is manifested in deviations from the typical use of a target language, whereas positive transfer is visible in situations where conventionalized and typical target-language features are selected because they are similar to corresponding source-language features (for further terminological discussion, see Mauranen 2004).

Regardless of the terminological issues, interference has captured the interest of other translation scholars. Mauranen (2004), for instance, has attempted to find signs of interference (or transfer) by examining with corpus tools how much variation there is between original Finnish texts and Finnish translations. Her material consists of four types of texts compiled into four different corpora: original Finnish texts, Finnish translations from ten source languages, Finnish translations from English, and Finnish translations from Russian.

Mauranen’s results show that the three translation corpora differ from the corpus of original Finnish, which she interprets as a sign of interference. However, her results also show that translations from Russian deviate more from original Finnish texts than translations from English do, which indicates that a target language might be more prone to interference from certain source languages.

Mauranen’s (2004) study has been guided by the belief that translations are somehow inherently different from original target-language texts and that at least some of these differences can be explained by the influence of one (source) language onto another (target) language (see e.g. Chesterman 2011). Another such scholar is Tirkkonen-Condit (2002; 2004;

2005), who, in order to explain (some of) the differences between target-language originals and translations, has formulated the so-called unique item hypothesis. She hypothesizes that translators use less those features of a given target language that lack “linguistic counterparts”

(Tirkkonen-Condit 2002: 209) or “straightforward linguistic counterparts” (Tirkkonen-Condit 2004: 177) in the source language. These kinds of features, or unique items as Tirkkonen- Condit calls them, should therefore be under-represented in translations. She exemplifies these items with a few Finnish verbs that indicate sufficiency but lack lexicalized equivalents

(18)

14

in English: jaksaa (‘be strong enough’), ehtiä (‘have enough time’), and viitsiä (‘have enough initiative’). Other unique items identified by Tirkkonen-Condit (2002: 215) include Finnish particles (e.g., -kin ‘also’) and impersonal references (e.g., on mentävä ‘[one] has to go’). In other words, the behavior described by the unique item hypothesis is a specific kind of interference, since the lack of counterparts to unique items in the source langue affects the decisions of translators, i.e., the source language or text interferes with the process of producing a target text.

The concept of unique items is not without its problems, some of which have been pointed out by Chesterman (2007). For instance, he criticizes Tirkkonen-Condit’s (2002; 2004) definition of being too loose and asks how we can differentiate between items or elements that are unique to the target language but not to the source language (Chesterman 2007: 6–7). In order to answer this question, Chesterman (2007: 7–8) refers to Catford’s (1965) concept known as unit shift. According to Catford (1965: 79), a unit shift is a change from one unit rank into another. The unit ranks from highest to lowest are sentence, clause, group, word, and morpheme (Catford 1965: 8). Therefore, Chesterman (2007: 7) argues that those items that are unique to the target language cannot be “readily” translated back into the source language without a unit shift. He later clarifies (or provides an alternative definition) that a unique item is a linguistic item whose “translation equivalent only maintains unit correspondence at some higher level or levels, not at given lower levels” (Chesterman 2007: 8). We can use Chesterman’s definition to test whether some of the potential unique items are in fact unique.

The Finnish particle -kin cannot be translated into English with a morpheme but has to be rendered using a target-language adverb: Hänelläkin oli hauskaa (‘He was also having fun’).

In this case, the Finnish morpheme -kin has been changed into the English adverb also, i.e., a unit shift has occurred upward. However, when translating into the opposite direction, the English adverb does not have be rendered using a Finnish morpheme. It can also be rendered using a target language adverb, namely myös (‘also’): Myös hänellä oli hauskaa. Therefore, the morpheme -kin can be classified as a unique item, whereas the adverbs myös and also are not unique items. Furthermore, previous corpus research on the particle -kin (e.g., Tirkkonen- Condit 2005; Tirkkonen-Condit & Mäkisalo 2007) concurs with the unique item hypothesis, as there is evidence of the particle being under-represented in Finnish texts translated from English, such as in fictitious and academic texts, although in Finnish subtitle translations, the particle occurs almost as frequently as in original Finnish texts. (In fact, certain proposed

(19)

15

unique items, such as the clitic particle -hAn, are over-represented in subtitled Finnish (see Tirkkonen-Condit & Mäkisalo 2007)).

Chesterman’s (2007) definition of a unique item seems to include many of the possible Finnish unique features studied by translation scholars, such as being able to omit subjective personal pronouns (e.g., [minä] elän ‘I live’; see Mauranen & Tiittula 2005) and being able to combine conjunctions with the negative verb ei ‘not’ (e.g., jottei ‘in order not to’; see e.g.

Puurtinen 2004). However, it also excludes some.

In translation from Finnish into English or Swedish, certain target-language verb constructions could be candidates as unique items, since they do not have close equivalents in Finnish. These verb constructions, also known as particle verbs in English, are formed by combining a verb and a particle (e.g., put out and tycka om ‘like’). However, many of these verbs can be replaced with independent verbs without losing the core meaning (e.g., extinguish and gilla ‘like’). Finnish, in contrast, relies on independent verbs to convey these core meanings (e.g., sammuttaa ‘put out/extinguish’ and pitää ‘like’). Therefore, if we translate these kinds of phrasal verbs into Finnish, we will have to perform a unit shift downward, i.e., from a phrase to a word. This should not be possible according to Chesterman’s (2007) stricter definition. However, there is empirical evidence of particle verbs being under-represented in the language of Finnish speakers learning Swedish (see Laurén 2005) and in English translations from Romance languages, most of which lack particle verb constructions, but not in English translations from other Germanic languages, such as Swedish (see Cappelle & Loock 2017). Regardless of whether or not particle verbs can be classified as unique, they are in line with the unique item hypothesis.

Neither Chesterman’s (2007) nor Tirkkonen-Condit’s (2002; 2004) definition of a unique item is completely satisfactory: Chesterman’s definition is too strict, as illustrated above, whereas Tirkkonen-Condit’s definition is too loose or subjective, since one can ask the question of what constitutes a straightforward linguistic counterpart to another linguistic item.

Therefore, there is the question of equivalence. For instance, all Finnish first-person singular pronouns (mä, mie, meitsi, etc.) can be readily or straightforwardly translated with the English first-person singular pronoun I, i.e., they share at least some degree of equivalence. However, one can argue that the English pronoun shares a closer resemblance to the standardized Finnish pronoun minä, since the semantic and pragmatic content of these two is more similar in comparison to the content shared by I and the non-standardized Finnish pronouns, such as

(20)

16

mie. For instance, I and minä in themselves do not necessary reveal where the speaker is from.

The dialectal mie, in contrast, indicates that the speaker’s place of origin is (likely) Eastern Finland. Therefore, if we agree that I is not a close enough equivalent to mie, the non- standardized Finnish pronouns fit better the definitions formulated by Tirkkonen-Condit (2004) and Chesterman (2007): they lack straightforward equivalents and cannot be translated without a unit shift (e.g., mie [word] → I, who is from Eastern Finland [group]).

As has been illustrated by the three subsections in this section, the four universals (i.e., explicitation, standardization, interference, and the unique item hypothesis) are predictions on what translators have done in the past and what they will do in the future. However, since universals are synonymous with tendencies, it is likely that we will find instances of translational behavior that cannot be explained by translation universals alone. If we want to explain such behavior, we can place translators into groups based on the features that differentiate some translators from all the other translators, such as the source or target language, the genre of the source text, or the gender of the translator. This will allow us to describe how a specific group of translators tend to act, i.e., to investigate translation norms.

These are the subject of the following section.

(21)

17

3. Translation norms

In addition to translation universals, one of the concepts relevant to the descriptive branch of translation studies is translation norms. However, whereas translation universals are assumed to manifest themselves in all or most translations and can be used to predict translation behavior on a larger scale, translation norms affect translations on a more well-defined level, since they can vary between languages and cultures, as well as across different time periods (see e.g. Eskola 2004: 85). This section first discusses the relationship between prescriptive and descriptive norms as well as between norms and other socio-cultural constrains (subsection 3.1) and then presents one possible way of placing descriptive norms into different categories (subsection 3.2).

3.1. Norms, rules, and idiosyncrasies

In translation studies, norm theory has been pioneered by Toury (e.g., 1980; 1995), who gives the concept of translation norm the following definition, which is based on sociology:

the translation of general values or ideas shared by a community – as to what is right and wrong, adequate and inadequate – into performance instructions appropriate for and applicable to particular situations, specifying what is prescribed and forbidden as well as what is tolerated and permitted in a certain behavioural dimension [...]. (Toury 1995: 55)

Toury (1995: 55) points out, however, that norms do not have to be explicitly formulated in order to exist: they exist if community members adhere to them. Therefore, we can distinguish between two types of norms, namely descriptive and prescriptive. According to Chesterman (1997: 53), descriptive statements of norms are hypotheses about human behavior based on observational evidence; prescriptive statements, in turn, are linguistically formulated instructions or orders that people are expected to follow. In other words, prescriptive norms tell us what translators are ordered to do, whereas descriptive norms tell us what translators actually do. For instance, the Finnish language authorities held the view that the construction alkaa tekemään (‘start doing’) is not correct standard Finnish and recommended instead the construction alkaa tehdä. However, this restriction was lifted in 2014 by the Finnish Language Board, and both construction types are now regarded as standard language (see e.g.

Kankaanpää 2014). Since many translators are bound by the rules of standard language, this raises the question of whether they still follow the old prescriptive norm (if they ever did), i.e.,

(22)

18

is the current descriptive norm that translators prefer the previously recommended construction to the newly allowed one?

In any case, translation is seen by Toury (1995: 54) as an activity that is subject to different kinds of socio-cultural constraints, of which norms are just one. He places them in the middle of a continuum whose opposite poles are rules, on the one hand, and idiosyncrasies, on the other. Furthermore, Toury’s idea of placing rules, norms, and idiosyncrasies on a continuum has been adopted by Pedersen (2011: 30), who has expanded the continuum with a few new concepts that govern translational behavior: his model includes idiosyncrasies, regularities, conventions, norms, rules, and laws.

According to Pedersen (2011: 31), idiosyncrasies explain translational behavior of individual translators, whereas regularities are “recurring patterns” which have not been socially noticed.

Conventions and norms, in turn, have been socially noticed, but the difference between these two is that conventions can be easily broken, whereas going against norms will result in sanctions (Pedersen 2011: 31). In this regard, norms are similar to rules, since violations of both norms and rules are sanctioned (Pedersen 2011: 32). Rules, however, steer translational behavior even more strongly because they are often “codified and authoritarian”, such as the rules of grammar (Pedersen 2011: 32). Therefore, it is less acceptable to break rules than it is to go against norms or conventions. Finally, laws are seen by Pedersen (2011: 33) as the same as translation universals (e.g., the law of growing standardization; see section 2).

Toury (1995: 54) has pointed out that the socio-cultural constraints can change their domain over time, i.e., idiosyncrasies can become norms, norms can become rules, rules can become norms, and so on. For instance, by going against the prevalent norm of slang not being used in Finnish literature, Pentti Saarikoski and his translation of The Catcher in the Rye (published in 1961) encouraged the use of colloquial language in Finnish-language novels (see Tiittula &

Nuolijärvi 2016). In other words, Saarikoski’s idiosyncrasies contributed to the formation of a new norm, which replaced the previous one.

3.2. Categorization of norms

There are multiple ways to place norms into different categories (e.g., Toury 1995;

Chesterman 1997). Toury (1995), for instance, distinguishes between three kinds of norms, namely preliminary norms, the initial norm, and operational norms. According to Toury

(23)

19

(1995: 58), preliminary norms govern what kind of texts or text types have the chance to be translated and whether the target culture tolerates indirect translations. In Finland, for instance, the source language of most literary translations is English nowadays (Statistics Finland 2019:

155). There are also a few Japanese novels being translated each year, and some of these have been translated from an intermediate language (of the thirteen Finnish translations of Japanese novels published between 2010 and 2019, two were indirect translations; Fennica 2020).

Therefore, the two Finnish preliminary norms are that English and Japanese literary texts are translated into Finnish and that indirect translations are tolerated (but perhaps not preferred) if the original source language is Japanese.

The initial norm, as described by Toury (1995: 56–57), governs whether the translator adheres to the norms of the source culture or the target culture. More specifically, Toury states that if the translator is following the norms of the source culture, he or she is aiming for an adequate translation but if the translator instead follows the norms of the target culture, he or she is aiming for an acceptable translation. For instance, early Finnish translators of literature tended to standardize colloquial language found in source texts (see Tiittula & Nuolijärvi 2016), i.e., they were adhering to the norms of the target culture. Later Finnish translators, in turn, tended to use more spoken language varieties (see Tiittula & Nuolijärvi 2016), i.e., they were adhering to the norms of the source culture. Therefore, it could be argued that later Finnish translations are more adequate than earlier translations, which, in turn, are more acceptable.

In any case, Toury’s (1995) notions of adequacy and acceptability are similar to Nida’s (1964:

159) formal equivalence (i.e., the aiming for a literal translation) and dynamic equivalence (i.e., the aiming for a translation that has a similar effect on its readers as the source text has on its readers). They are also similar to the concepts of foreignization and domestication (i.e., preserving source-culture items vs. replacing them with target-culture equivalents in translation) examined by Venuti (e.g., 1995; for further discussion, see Ruokonen 2004).

Toury’s (1995) notions of adequacy and acceptability have been criticized by Hermans (1999:

76), who points out both conceptual and terminological issues. On the one hand, he argues that there cannot be a completely adequate translation, since this would entail constructing the textual relationships of the source text in the target language. Therefore, as Hermans puts it, the only adequate translation is the source text itself. On the other hand, Hermans (1999: 77) argues that the terms adequate and acceptable are confusing (see also Chesterman 1997: 64)

(24)

20

and instead recommends the terms source-oriented and target-oriented (these terms have also been adopted by other scholars, such as Pedersen (e.g., 2011)).

Finally, Toury’s (1995: 58–59) third norm category consists of operational norms, which govern the translation process itself. These he categorizes further into matricial norms and textual-linguistic norms. According to Toury, the former determine how closely the target text follows its source text on a higher textual level, i.e., they determine whether the source text is translated completely or partially, how the textual material is arranged or how it is segmented.

Textual-linguistic norms, in turn, affect decision-making on a lower textual level, which concerns sentences, phrases, words, etc. (Toury 1995: 59). In accordance with Toury (1995), Hermans (1999: 76) sees matricial norms governing translational decisions on the macro level and textual-linguistic norms on the micro level. In other words, the former concern decisions that affect the whole text, whereas the latter affect individual parts of the text. For example, if the target culture does not tolerate references to religion, the translator may have to omit whole paragraphs or sections from his or her translation, i.e., make macro-level decisions, or to change the word God in a sentence to something less religious, i.e., make micro-level decisions (on censorship and translation, see e.g. Pokorn 2012).

This section has discussed some of the features that have been attributed to norms in translation studies. They are similar to translation universals in that both can be used to predict translational behavior. However, unlike universals, the predictive power of norms is fairly limited, since different groups of translators and translations can follow different norms.

Therefore, the group that the norm describes must be well defined, since a single variable, such as the source language, can distinguish one group from another. For instance, those Finnish translators who translate Japanese literature directly from the original source language might not adhere to the same norms as those translating it indirectly via English, i.e., via an intermediate language. This type of activity is examined closer in the following section.

(25)

21

4. Indirect translation

Unlike its direct counterpart, the act of translating text or speech indirectly, i.e., via an intermediate language, has occupied a peripheral position within translation studies, although recently, it has gained some popularity as a research object (see e.g. Assis Rosa, Pięta, &

Bueno Maia 2017). The product of such an act or process, i.e., an indirect translation, is here defined as a “translation based on a source (or sources) which is itself a translation into a language other than the language of the original, or the target language” (Kittel & Frank 1991:

3). Therefore, the aim of this section is to first (subsection 4.1) discuss how indirect translation, as a form of trilingual communication (in its simplest form), differs from mono- and bilingual communication, i.e., from non-translation and direct translation, and then (subsection 4.2) to discuss some of the reasons for and benefits and disadvantages of translating indirectly.

4.1. Indirect communication

We can point out the communicative peculiarities of indirect translation by contrasting it with direct translation with the help of two communication models formulated by Nida (1966 [1959]). Below is Nida’s (1966: 16) model for bilingual communication, i.e., a model that illustrates direct translation.

Although Nida’s (1966) model shown in Figure 1 describes bilingual communication, it also contains the model for monolingual communication, which functions as the starting point for

C1

S1

C2

R1

S2 R2

Figure 1. Nida’s (1966) model for bilingual communication M2

M1

(26)

22

multilingual communication. In Figure 1, the left-hand rectangle illustrates the different components of monolingual communication. The smaller rectangle, i.e., M1, is the original message uttered by S1 (source) and aimed at R1 (receptor); the larger rectangle, i.e., C1, in turn, is the cultural context where the communication is carried out (Nida 1966: 14). When a Japanese author, for example, writes a novel aimed at his or her countrymen, the author (S1) is working in the cultural context (C1) that is Japan. The novel, in turn, is the actual message (M1), which is read by its target audience, i.e., by Japanese people (R1).

Monolingual communication is similar to the first phase of bilingual communication: a source (S1) produces a message (M1) intended for receptors (R1) in a certain cultural context (C1).

The difference is, however, that in bilingual communication, the message is mediated to another culture by a translator, as shown by the right-hand triangle in Figure 1 (see Nida 1966: 16–17). Therefore, the translator assumes both the role of a receptor (R1) and a source (S2): taking the original message as a basis, the translator produces another message (M2), which has its own receptors (R2) in the target culture (C2;Nida 1966: 17). If we build on the previous example, the novel by the Japanese author (M1) is read and interpreted by an English translator (R1), for example. He or she then translates the novel into English (S2), and the translation (M2), in turn, is read by new receptors (R2) in the target culture (C2; e.g., in the United States or South Africa).

Moving from direct translation to indirect translation, then, entails adding yet another cultural context with its source, message, and receptors. This is illustrated by Figure 2 below, which shows Nida’s (1966: 18) model for trilingual communication.

Figure 2. Nida’s (1966) model for trilingual communication M1

S1 R1 S3 M3 R3

M2

S2 R2

C1 C2 C3

(27)

23

In Figure 2, the left-hand rectangle (C1) is the original source culture, whereas the right-hand circle (C3) is the final target culture; the triangle in the middle, in turn, is the mediating culture (C2; see Nida 1966: 18). Therefore, indirect translation is similar to direct translation in that in both types of communication, the translator simultaneously assumes the roles of a receptor (R1 and R2) and a source (S2 and S3). The difference is, however, that the indirect translator is conveying a message (M2) that is an approximation of the original message (M1; Nida 1966: 18). Based on this framework, we can yet again expand on the previous example.

The Japanese novel (M1), which has been translated into English, is now translated into Finnish. This time, however, the source text is not the Japanese original but the English translation (M2). Therefore, the Finnish translator (R2 and S3) is interpreting and translating another translation, and as a result, there is only an indirect relationship between the Finnish translation and the Japanese original. In other words, indirect translation increases the cultural distance between the original source culture and the final target culture, which can be seen as one of the downsides of indirect translation.

4.2. Reasons, benefits, and disadvantages

If indirect translation is stigmatized by being somehow inferior to direct translation (see e.g.

Ivaska & Paloposki 2018), then why are indirect translations preferred in certain situations? A number of possible reasons have been listed by Ringmar (2007: 6). Firstly, he suggests that an indirect translation is preferred if the intermediate language is more prestigious than the language of the original or if it simply is not possible to acquire the original source text. For instance, the modern translations of the Bible are based on intermediary texts, such as a Hebrew translation, since the (Aramaic) original has been lost (Dollerup 2000: 21–22).

Secondly, Ringmar (2007: 7) points out that some translations are based on intermediary texts because the latter have already been sanitized of controversial content. For instance, many translations of children’s books into the languages of Yugoslavia were based on Russian adaptations due to them already being adjusted to the communist ideology (e.g., references to religion being omitted; see Pokorn 2012).

However, the most obvious reason for indirect translation, according to Ringmar (2007: 6), is the lack of translators: either there are no translators available who can translate from a certain language into another, or there are not enough such translators to meet the demand. This latter reason is similar to what has been suggested by Leppänen (2013: 54–55). She hypothesizes

An I for an AI: A Corpus-Based Study on the (In)direct Translation of Japanese First-Person Singular Pronouns into Finnish

UNIVERSITY OF EASTERN FINLAND PHILOSOPHICAL FACULTY

SCHOOL OF HUMANITIES English Language and Translation

Henry Laine

AN I FOR AN AI: A CORPUS-BASED STUDY ON THE (IN)DIRECT TRANSLATION OF JAPANESE FIRST-PERSON SINGULAR

PRONOUNS INTO FINNISH

MA Thesis

February 2021

ITÄ-SUOMEN YLIOPISTO – UNIVERSITY OF EASTERN FINLAND

ITÄ-SUOMEN YLIOPISTO – UNIVERSITY OF EASTERN FINLAND

Contents

1. Introduction

2. Translation universals

2.1. Universal tendencies

2.2. Explicitation

2.3. Standardization

2.4. Interference and the unique item hypothesis

3. Translation norms

3.1. Norms, rules, and idiosyncrasies

3.2. Categorization of norms

4. Indirect translation

4.1. Indirect communication

4.2. Reasons, benefits, and disadvantages