Corps de l’article

1. Introduction

Among the proposed recurrent tendencies of translated language (Baker 1993), a strong preference for conformity to conventions or norms in the target language (TL), sometimes even to the extent of exaggeration (Baker 1996: 177), has been highlighted. This preference is referred to as normalisation (also known as “standardisation,” “conservatism,” or “conventionalisation”). It is related to Toury’s (2012) law of growing standardisation, which posits that “in translation, source-text textemes tend to be converted into target-language (or target-culture) repertoremes” (Toury 1995/2012: 304). Chesterman (1997: 72) elaborates this law as “translators tend to replace text-specific items with institutionalised items: translations tend to be less idiosyncratic, more conventionalised, than their originals.” The tendency implied here is accompanied by an avoidance of creative or unusual linguistic forms (Kenny 2000; Bernardini and Ferraresi 2011), so that, overall, translations exhibit a greater degree of conventionality than comparable original texts. “Conventionality,” in this sense, refers to the idea that much language use is routine (Stewart 2000). Stewart (2000: 80) hypothesises not only that translated texts are more conventional than non-translated texts in the same language, but also that translations done into the L2 are even more conventional than translations done into the L1.

Normalisation, or a tendency towards increased conventionality in translation, is most often regarded as a consequence of socio-cultural or economic constraints that influence translation (Kenny 2001). Since translators translate with readers’ expectations about acceptability in mind, this biases translators to orient their translation towards existing norms and conventions. Translations that deviate from these expectations for acceptability might cause the translation to be criticised, ignored and rejected by the target audience (Kenny 2001: 67). In contrast to this social explanation of normalisation, cognitive-linguistic explanations of normalisation assume that the category prototype and highest-level schema of the TL might exert “magnetism” (Halverson 2017: 15), which encourages the use of “specific TL lexical and grammatical structures that correspond to those salient nodes and configurations in the schematic network” (Halverson 2003: 218). As a consequence, an overrepresentation or exaggeration of specific target-language lexical and grammatical features occurs in translation, which accounts for the normalisation tendency (Halverson 2003: 218-221).

As pointed out by Van Oost, Willems, et al. (2016), Baker’s (1993; 1996) definition of normalisation empirically implies two possible frequency patterns in the TL. First, there could be no significant frequency differences of a particular feature in translated and non-translated TL texts (therefore, “normalisation”; Van Oost, Willems, et al. 2016: 9). Second, there could be a significantly higher frequency of the feature in question in translations compared to non-translations in the same language (therefore “over-normalisation”; Van Oost, Willems, et al. 2016: 3).

However, in addition to these two patterns, there is also logically a third scenario, in which the linguistic feature in question demonstrates a significantly lower frequency in translated texts than in original texts in the TL. This scenario provides evidence against normalisation, in that translations diverge from non-translations in ways that are not typical for the TL. This tendency has been investigated under various guises in translation studies, and is usually accounted for by source-text or source-language influence. Toury proposes a counterbalance to the law of increasing standardisation, namely the law of interference: “phenomena pertaining to the make-up of the source text tend to force themselves on the translators and be transferred to the target text” (1995/2012: 310). Interference is also termed transfer, “source-language shining through” (Teich 2003), or cross-linguistic influence (CLI) and refers to the idea that translations show influence from the source language (SL) or source text (ST), due to the basic cognitive condition of the prior cognitive activation of the SL, leading to what has been described as cross-linguistic priming effects (Kruger and Van Rooy 2016). This cognitive mechanism of interference has been established in Tercedor Sánchez, López Rodríguez, et al. (2013).

Another form of CLI is evident in the so-called “Unique Items Hypothesis” (Eskola 2004; Tirkkonen-Condit 2004; Cappelle 2012), according to which linguistic items or features which are unique to the TL and not present in the SL may tend to be under-represented in translations compared to non-translations in the same language because their selection is inhibited by their absence in the SL, and they therefore occur less frequently in translations (Eskola 2004; Tirkkonen-Condit 2004).

Clearly, in respect of the conventionality of translations compared to non-translations, there is therefore a tension between CLI and (over-)normalisation. They should be considered two constantly competing tendencies, affecting a translation’s degree of conventionality. The relationship between these two forces is complex and influenced by numerous other variables (for example, sociocultural norms, text type conventions, translator style; see Lefer and Vogeleer 2013). As a consequence, normalisation may be counterbalanced (Volansky, Ordan, et al. 2015) and sometimes even overruled by CLI (Capelle and Loock 2013; Lefer and Vogeleer 2013). Their interaction could be described as points along a cline, where normalisation (or over-normalisation) and CLI are positioned at each extremity.

To investigate the tension between normalisation and CLI, this study makes use of Chinese children’s literature translated from English as a test case. Translated children’s literature is a genre in which incentives towards normalisation are particularly strongly felt, since the acceptability of the translated text to target readers (including both children and adults, for example, the parents who buy books) is highly emphasised in this text type (Puurtinen 1995; 2006; O’Connell 1999; Shavit 2006). This emphasis might be further strengthened in the genetically distinct language pair of English-Chinese (Xiao 2011). Translators translate with readers’ expectations about acceptability in mind, which biases them to translate in ways that conform to existing norms and conventions in Chinese. However, it is also likely that the ever-present effects of CLI might be evident, particularly in the form of the underrepresentation of Unique Items – cases where lexicogrammatical items that are unique to Chinese have no equivalent in English. To our knowledge, the tension between normalisation and CLI has not been widely studied in this genre, nor in this language pair (see Section 2). This study therefore aims to investigate normalisation in a self-built comparable corpus of children’s books translated from English into Chinese, using modal particles (a unique item in Chinese; see Section 2.2 for more detailed discussion) as a linguistic indicator.

The following section (Section 2.1) reviews corpus-based studies of normalisation in relation to CLI in translated language with a focus on Chinese translated from English. Section 2.2 discusses the use of modal particles in Chinese children’s books. This section concludes by tying together the threads of the discussion in formulating the research questions informing this study. The methodology section (Section 3) starts with a discussion of the corpus in terms of text collection and processing, followed by a description of the selected operationalisation, and then a discussion of the data extraction and analysis procedures. Section 4 presents and discusses the findings of the study. Conclusions, implications and limitations are summarised in Section 5.

2. Literature review

2.1. Normalisation and CLI in tension

Translations could be more conventional, more conservative and less creative compared to their ST (normalisation as a S-universal; Chesterman 2004; see Mauranen 2000; Olohan 2004; Baker 2007) as well as to non-translated texts in the TL (normalisation as a T-universal; Chesterman 2004; see Kenny 2001; Marco 2009; Delaere, De Sutter, et al. 2012; Delaere and De Sutter 2013). On the one hand, there is evidence that translators do overuse conventional patterns or typical features of the TL (Mauranen 2000; Kenny 2001; Olohan 2004; Baker 2007; Marco 2009; Delaere, De Sutter, et al. 2012; Delaere and De Sutter 2013; Szymor 2018), supporting the tendency towards (over-)normalisation. On the other hand, studies also show that translations tend to reflect more “unusual word combinations” than originals (Mauranen 2000: 120, 137); structures more typically used in the SL than the TL (for example, passive constructions; Hansen and Teich 2001; Dai and Xiao 2011; Kenny and Satthachai 2018) and lower frequency of some prototypical TL features, such as manner-of-motion verbs and there-BE constructions in English translations from French (Cappelle 2012; Cappelle and Loock 2013), providing evidence in favour of CLI in the form of the Unique Items Hypothesis.

Research on normalisation in translated Chinese often yields mixed results (Hu 2006; Wang and Qin 2010; Xiao and Dai 2010; Xiao, He, et al. 2010; Xia 2014; Xiao and Hu 2015). In one of the earlier studies on the topic, Hu (2006) considers normalisation along with other features of translated language and sets out to operationalise them at lexical and syntactic levels in contemporary Chinese translated fiction. The study yields contradictory findings. At a syntactic level, the lower frequency of passive constructions and a stronger tendency of negative and afflictive prosodies associated with the passive converge with conventional use in Chinese, producing persuasive evidence of normalisation. However, the higher frequency of grammatical words realising hypotaxis in the translated texts shows a deviation from the norms of the TL, as original Chinese texts prefer paratactic structures. A tendency towards foreignness is also evident in longer and more complicated attributives (unusual in Chinese). These findings suggest a complex interplay between normalisation and CLI.

Expanding this approach, Xia (2014) conducted a systematic and comprehensive study of normalisation in a diachronic parallel and comparable corpus of fiction texts translated from English to Chinese, their English STs and Chinese non-translated fiction. She measures normalisation at both the macro and micro linguistic levels to explore the manifestations of normalisation over two historical periods (1930s-1949; 1988-2007). She also attempts to seek explanations for these changes from a social-cultural perspective. The findings demonstrate that translations from both periods show a normalisation tendency, although this varies both in nature and in degree. Earlier translations tend to show normalisation effects reflected in the significantly more common use of high-frequency words and monosyllabic words in translations than non-translations, while contemporary translations exhibit a stronger tendency towards conservativeness, realised by a weaker compositionality of affixes, fewer creative collocations and other linguistic features. In general, the tendency towards normalisation appears to be stronger in the more recent translations. However, there is also evidence of a tendency towards non-typical linguistic features in the translations, indicating CLI effects from English, for instance, in the more frequent use of pronouns, connectives and prepositions, in line with Hu (2006). According to Xia (2014), the differences over time may be ascribed to changes in power relations between the source and target language and cultures as well as the position translation occupies in the target literary system. The study concludes that translated texts are a mixture of normalisation and denormalisation (Xia 2014).

The tension between normalisation and CLI has been an explicit focus of several corpus-based studies (Hansen-Schirra 2011; Xiao and Hu 2015; Van Oost, Willems, et al. 2016). Hansen-Schirra (2011) investigates typical and atypical features associated with fictional writing in translated and non-translated English texts, using the Translational English Corpus (TEC), the British National Corpus (BNC) and the CroCo Corpus (Hansen-Schirra, Neumann, et al. 2012),[1] a bidirectional parallel corpus consisting of English originals and their German translations as well as German originals and their English translations. The findings show that both typical and atypical features occur more frequently in the TEC compared to the BNC, meaning that translations show contradictory tendencies in relation to normalisation (Hansen-Schirra 2011). Based on an analysis of the bidirectional parallel CroCo Corpus, she demonstrates that the overuse of atypical fiction features can be ascribed to CLI effects, due to the literal translation of the corresponding source-text structures (Hansen-Schirra 2011: 147). The co-occurrence of normalisation and interference results in the hybridisation of target texts, which are dissimilar to both the SL and the TL (Hansen-Schirra 2011). By investigating the rendition of prepositional phrase placement, a syntactic variation existing in both Dutch and German, in translations of contemporary literary fiction between the two languages, Van Oost, Willems, et al. (2016) not only confirm the co-occurrence of normalisation and SL shining through, but also find an asymmetry, in that CLI is strongly present in German-Dutch translation whereas normalisation is more dominant in the reverse direction. They interpret this as providing evidence for Toury’s (1995/2012) hypothesis that the tolerance for interference tends to increase when translation happens from a highly prestigious language (for example, German) to a less prestigious language (for example, Dutch; see Toury 1995/2012; Van Oost, Willems, et al. 2016).

Xiao and Hu (2015) also investigate normalisation and SL shining through by measuring the frequency of idioms, modal particles and passives in translational Chinese. The findings show that the use of these items is quite often affected by the ST. They conclude that SL “shining through” may have more significant effects than TL normalisation in translation from English to Chinese, particularly in terms of modal particles (Xiao and Hu 2015). Further support for CLI effects come from Wang and Qin (2010), Xiao and Dai (2010) and Xiao, He, et al. (2010). In particular, Wang and Qin (2010) find that the suffix – 性 (xìng) shows a stronger word formational capacity in translation, which contradicts the findings of Xia (2014). They argue that this is the consequence of affix-by-affix translation from the English affixes -ity, -ness and -dom.

The degree of conformity to linguistic and sociocultural norms in translation varies by SL and also text type (Delaere, De Sutter, et al. 2012; Delaere and De Sutter 2013). According to Delaere, De Sutter, et al. (2012), the more heavily edited text types, including fiction and journalistic texts, tend to use more standard language than less heavily edited text types (for example, administrative texts). Delaere and De Sutter (2013) furthermore verify that even within the same sociocultural context, the tendency to conform to linguistic norms depends on text type and the readers targeted. In this sense, the operationalisation of normalisation should include more specific features conventional of a particular TL and a particular text type (Lefer and Vogeleer 2013). The next section deals with a typical feature of Chinese children’s literature, which is a unique item that exists in Chinese, but has no direct equivalent in English: modal particles.

2.2. Modal particles in Chinese children’s literature

Mandarin Chinese makes use of a rich system of particles, which occur in sentence-final position to mark the speaker’s mood or attitude towards the proposition. They are broadly captured under the concept of “sentence-final particles” (Li and Thompson 1981) or more specifically referred to as “modal particles” (Xiao and Hu 2015). Frequently used modal particles include 了 (le), 呢 (ne), 吧 (ba) and 吗 (ma).

Modal particles are typically used in speech, or in writing that reflects or recounts conversations (Li and Thompson 1981). They do not have meaning by themselves, but are context dependent, and thus each particle can be used in different contexts to express different emotions (Chappell 1991; Bross 2012). Modal particles are of interest because of their non-equivalent semantic and pragmatic functions in Chinese in contrast with English. Chinese is one of the few languages (including also German and Japanese) that make use of modal particles. In this sense, they may be viewed as a unique feature of Chinese as English does not have a direct equivalent structure to Chinese modal particles. In English, the same functions can be realised by the use of auxiliaries, modal verbs, special word order or intonation (Xiao and Hu 2015).

A greater use of modal particles is considered a particular feature of Chinese spoken language and children’s literature. Children’s literature tends to use modal particles more frequently than other types of writing to slow down reading speed and soften the tone (Zang 2010). The use of modal particles increases vividness in literature, which would be attractive to children and arouse resonance in them as they help child readers visualise scenarios using their imagination (Mei 2015). Modal particles can also increase the readability of children’s books. The typical association of modal particles with Chinese children’s literature makes this operationalisation particularly suited to the investigation of the feature of normalisation in this study.

Previous studies (mostly master’s theses) have investigated modal particles in relation to the tension between normalisation and CLI in English-Chinese children’s literature translation (Zang 2010; Mei 2015; Jiang 2016). For instance, Jiang (2016) investigates normalisation at lexical, syntactic and discourse levels in Chinese children’s literature translated from English. She finds tendencies of both normalisation and deviation. Normalisation is most remarkable at the lexical level (including identical part-of-speech distribution patterns, overuse of typical modal particles, greater occurrence of high-frequency words in translations vs. non-translations) and deviation is most detectable at the syntactic (for example, longer sentence segments) and discourse levels (for example, more frequent use of conjunctive markers). Jiang (2016) discusses the causes and effects of normalisation and deviation: while normalisation could be ascribed to the construal of target child readers on the part of translators, deviation is considered to be driven by SL interference (Jiang 2016). She also expresses concerns that inappropriate normalisation (for example, over-normalisation) and deviation in translated books for children aged 3-6 could cause reading difficulties and burden their comprehension. This is an important consideration, and ties in with the emphasis on target-audience acceptability in the translation of children’s literature.

Despite the fact that a handful of studies have touched upon the use of modal particles in investigating the features of translated language in children’s literature translated from English to Chinese, modal particles have not been comprehensively and systematically investigated in relation to the interplay between normalisation and CLI. Moreover, the existing corpus-based studies often suffer from a small-scale and unbalanced corpus design. Against this background, this study aims to shed more light on the occurrence of and the reasons for these two tendencies in a relatively larger comparable corpus of Chinese translated and original children’s books by answering the following two research questions:

  1. Are there significant differences in the frequency and use of modal particles in Chinese children’s literature translated from English, compared to non-translated Chinese children’s literature, suggesting that translations diverge from non-translations in respect of the degree of conventionality?

  2. If there are differences, do these reflect tendencies of over-normalisation or CLI, or an interplay between these two forces?

3. Methodology

3.1. Corpus composition and compilation

This study is based on a self-built comparable corpus of translated and non-translated Chinese children’s literature. The Translated Chinese Children’s Literature Corpus (TCCLC) and Non-translated Chinese Children’s Literature Corpus (NCCLC) are constructed to be as comparable as possible in terms of size, sampling period and content domain. A total of 22 texts of Chinese children’s books translated from English are included in the TCCLC while 20 texts of original Chinese children’s books are included in the NCCLC. The token count of the TCCLC is 1,168,137 and 1,215,259 for the NCCLC. The corpora were compiled according to the following criteria:

  1. Children’s books published in mainland China in the period 2000-2017. The timeframe is slightly extended backwards to 1998 for translated books, as a consequence of the fact that some famous translations by well-known translators do not, to our knowledge, have more recent editions.

  2. Books suggested by publishers/editors or online booksellers as suitable reading for older children, aged 7-11. This information was obtained from either the back-cover blurb of the book or classification by age groups in online bookstores.

  3. Classic fiction books for children. This decision was made to reflect the reality that, in the translation industry in China, classic children’s books dominate the market of translated Chinese children’s literature.

The majority of texts were sourced online. Only a small number of texts were sourced by purchasing e-books from online bookstores in epub format, which were then converted into text files by the OCR module of CamScanner.[2] All electronic text files were subsequently proofread and corrected manually in order to ensure accurate renderings of the original texts. The corpora include running text with characters only, with metadata stored in a separate text file for easy retrieval. In books for children aged seven and older, visual material significantly decreases, and it can usually be removed without loss of meaning. The corpus avoided the over-representation of any individual author, translator or publisher.[3] Detailed information about the books included in each subcorpus can be found in Appendices 1 and 2.

As Chinese is written without spaces separating words, a process of segmenting text strings into word tokens (referred to as “segmentation” or “tokenisation”) was conducted using a freely accessible segmentation tool, SegmentAnt 1.1.2 with the NLPIR/ICTCLAS engine.[4] Segmented texts were then loaded into the corpus-analysis software WordSmith Tools 7.0.[5]

3.2. Frequency of modal particles as operationalisation of conventionality: Data extraction

There are more than twenty types of modal particles in Chinese (Chao 1968; Li and Thompson 1981) and the focus of this study is on the most commonly used five. A bottom-up method for identifying the most commonly used modal particles was used. The most frequent modal particles were selected from the word list generated by the Wordlist function in WordSmith Tools of the combined corpus of translated and non-translated texts (see Table 1). These particles are呢 (ne), 吗 (ma), 吧 (ba), 呀 (ya) and 啊 (a).[6]

The concordances of these modal particles extracted from both corpora were manually cleaned by deleting irrelevant cases, where these tokens were not used as modal particles.[7] These concordances were used to calculate the normalised frequency (per 1,000 words) of each modal particle, per file. The overall normalised frequency of all five modal particles combined per file was also calculated. These normalised frequencies were used as the basis for statistical analysis.

Table 1

Modal particles selected for investigation[8]

Modal particles selected for investigation8

-> Voir la liste des tableaux

As discussed above, modal particles do not have a lexical meaning of their own; instead their meaning is context-dependent. Applying the mood type classification of Halliday and Matthiessen (2004) to Chinese, a configuration of modal particles with regard to mood type can be produced (see Table 2; also see Huang and Liao 2002).[9] The column “Optionality” in Table 2 indicates whether the use of a particle is compulsory in realising the mood type. If a particle is optional, it means that the speaker does not necessarily need to use the particle to realise the mood type.

Table 2

Classification of modal particles by mood type in modern Chinese[10]

Classification of modal particles by mood type in modern Chinese10

-> Voir la liste des tableaux

In the presentation of the findings and discussion (see Section 4), the specific classification of the modal particles investigated in the study will be further refined building on this classification.

As far as normalisation is concerned, there are three possibilities. If modal particles occur at similar frequency in translated Chinese children’s books and in non-translated Chinese children’s books, this may be taken as evidence of normalisation or conventionalisation. If modal particles occur significantly more frequently in translated than in non-translated Chinese children’s books, it may be inferred that translators are over-adjusting their translations to the typical norms of the TL, thus hyper-conventionalising or over-normalising their translations. Alternatively, modal particles may occur significantly less frequently in translated Chinese children’s books than in non-translated Chinese children’s books. In this case, this may be ascribed to the effects of CLI, specifically in the form of the Unique Items Hypothesis. As English lacks the counterpart of Chinese modal particles, the activation of English suppresses the activation of this unique feature of Chinese, leading to an under-representation of this feature in translated Chinese children’s books. A further possibility is that the lower frequency of these particles in translated Chinese might be ascribed to a different kind of conservatism on the part of translators: given their association with informal, colloquial language, translators may avoid them in favour of a more standardised, written style.

3.3. Data analysis

The normalised frequency of modal particles for the two subcorpora (the TCCLC and NCCLC) was compared to answer Research Question 1. To evaluate whether the differences between the two groups were not just due to chance, the independent samples t-test was used in the first instance. This test has three assumptions that need to be satisfied: independence of observations, normal distribution of data and homogeneity of variance. The first assumption of independence of observations was met in all cases. To assess the second and third assumptions, a quantile-quantile (Q-Q) plot was first created to visually assess whether the data were sufficiently normally distributed and then Levene’s test was carried out to test the assumption of homogeneity of variance. If the test statistic (p-value) was larger than 0.05, then the equal variances assumption could not be rejected.

Where the data were sufficiently normally distributed and the assumption of homogeneity of variance was met, subsequent descriptive statistics are presented using the mean as a measure of central tendency, and standard deviation as a measure of dispersion, and the t-test was used to determine whether the difference in means in the two subcorpora is statistically significant, with p < 0.05 set as the level of significance. If these assumptions were not met, the median is used as a measure of central tendency in reporting, and the interquartile range as a measure of dispersion. In these cases, the non-parametric two-samples Mann-Whitney U-test was used to assess the differences between the two subcorpora, with p < 0.05 regarded as a significant result.

In addition to the quantitative analysis, a small-scale qualitative analysis of particular modal particles was also conducted. This allows us to better understand the quantitative findings and to further explore possible explanations for the observed findings, answering Research Question 2.

4. Findings and discussion

4.1. Statistical results

Figure 1 shows that modal particles are more frequent overall in the NCCLC (a mean of 7.74 per 1,000 words) than in the TCCLC (6.25 per 1,000 words). However, the t-test shows that this difference is not statistically significant (t = -1.51, p = 0.14).

Figure 1

Normalised frequency of all modal particles (per 1,000 words) by translated status

Normalised frequency of all modal particles (per 1,000 words) by translated status

-> Voir la liste des figures

It is important to point out that, as shown in Table 2, in most of the mood types, the use of these modal particles in the TL is optional. Meanwhile, as unique items in the TL, they have no counterparts in the SL, which means that the use of these particles in the TL cannot be subconsciously activated by any lexical representation in the SL. Therefore, the use of these modal particles in the translation may be seen as a result of translators’ conscious decision-making. Keeping in mind both the absence of modal particles in the SL and the optionality of the use of modal particles in the TL, the fact that there is no significant difference in the frequency of modal particles in the two subcorpora suggests that translators clearly aim to normalise their translations to the conventions of the TL genre, by adding these particles to the target texts – and are largely successful in doing so.

Meanwhile, it is equally important to point out that, although the difference is not significant, these modal particles are more frequently used in the non-translated Chinese texts in the NCCLC than the translated Chinese texts in the TCCLC. The relative under-use of the particles in the translated Chinese texts may therefore reflect the opposing pull of CLI in the translation process.

A subsequent question is whether there are any differences in the use of individual modal particles. The results show that three out of the five modal particles investigated do not demonstrate significant differences in the two subcorpora (see Table 3).

Table 3

Results of statistical tests for individual modal particles

Results of statistical tests for individual modal particles

-> Voir la liste des tableaux

As shown in Table 3, of the five particles under investigation, only 吗 (ma) is more frequently used in the TCCLC than the NCCLC, although the difference is not significant. It is interesting to point out that, as shown in Table 2, 吗 (ma) is the only one among the five particles that can be used in one particular mood type only (that is, in the biased polar interrogative type), and 吗 (ma) is not optional in achieving the mood type when in use. An alternative in achieving a similar semantic meaning would be to use an unbiased polar-interrogative instead (see further Li 2007), which is common in Chinese, but a less common type of interrogative in English (a typical example structure is “Do…or not?”). The more frequent use of 吗 (ma) in the translated Chinese texts might be due to the intention to achieve a syntactic equivalence, where a yes-no polar interrogative in English (a typical example structure is “Do…?”) has triggered the choice of a direct equivalent mood type in Chinese, typically realised by the use of 吗 (ma) at the end of the sentence.

Only two individual particles, 呢 (ne) and 呀 (ya), do demonstrate significant differences in the two subcorpora, with both particles showing higher frequencies in the NCCLC (see Figures 2 and 3), in line with Xiao and Hu’s (2015) finding that modal particles are significantly more frequently used in non-translated texts than in translated texts. This provides evidence for a significant divergence in conventionality between translated and non-translated Chinese children’s books in respect of the use of these modal particles, with the translated texts less conventional than the non-translated texts.

Figure 2

Normalised frequency of 呢 (ne) (per 1,000 words) by translated status

Normalised frequency of 呢 (ne) (per 1,000 words) by translated status

-> Voir la liste des figures

The boxplot in Figure 2 shows that the median value for the (relatively frequent) modal particle 呢 (ne) in the TCCLC is 1.29 per 1,000 words, whereas the NCCLC has a higher median value of 1.89. The modal particle 呢 (ne) is significantly more frequent in originals than in translations (U = 139, p < 0.05), in line with the overall trend.

Figure 3

Normalised frequency of 呀 (ya) (per 1,000 words) by translated status

Normalised frequency of 呀 (ya) (per 1,000 words) by translated status

-> Voir la liste des figures

The boxplot in Figure 3 shows that the median values for the modal particle 呀 (ya) in the TCCLC and NCCLC are 0.23 and 0.79 per 1,000 words, respectively. Clearly, the non-translated subcorpus has a higher median value. The modal particle 呀 (ya) is significantly more frequent in originals than in translations (U = 122, p < 0.05), in line with the overall trend. In addition, there is a higher degree of variability in the non-translated corpus, with a higher IQR of 1.08, compared to 0.53 for the translation.

In an attempt to further interpret the causes of the significant differences in the use of the two modal particles呢 (ne) and 呀 (ya) between the two subcorpora, that is, whether the differences in frequency are biased by the mood types in which they have been used to express various emotions, analysis has been conducted to examine their diverse pragmatic functions in the context. Tables 4 and 5 present the results of statistical tests for modal particles 呢 (ne) and 呀 (ya) by mood type, respectively.

Table 4

Results of statistical tests for modal particle 呢 (ne) by mood type

Results of statistical tests for modal particle 呢 (ne) by mood type

-> Voir la liste des tableaux

Table 5

Results of statistical tests for modal particle 呀 (ya) by mood type

Results of statistical tests for modal particle 呀 (ya) by mood type

-> Voir la liste des tableaux

As can be seen from Table 4, it is only in the declarative mood that the use of 呢 (ne) occurs at a significantly different rate between the TCCLC and NCCLC (U = 322, p < 0.05), with the NCCLC demonstrating a higher frequency (0.56 per 1,000 words in the NCCLC compared to 0.28 per 1,000 words in the TCCLC). When used in interrogatives, although there is no significant difference found between the two subcorpora, 呢 (ne) is more frequently used in the NCCLC than in the TCCLC (1.21 per 1,000 words vs. 0.86 per 1,000 words). This means the significant difference in the overall frequency of 呢 (ne) between the two subcorpora is mainly driven by its use in the declarative mood. In Table 5, it is evident that there are significantly more uses of 呀 (ya) in the NCCLC than in the TCCLC across all the mood types: declarative, interrogative and imperative (in all three cases, p < 0.05). The tendency of under-use of the modal particle 呀 (ya) in translated Chinese children’s books, in comparsion to non-translated Chinese children’s books, cuts across all the mood types more generally.

4.2. Further analysis of 呢 (ne) and 呀 (ya)

As explained in Section 2.2, modal particles do not have a lexical meaning of their own; instead, their meaning is context-dependent. In this section, a more detailed discussion of the two modal particles that are significantly more frequently used in the TCCLC than the NCCLC, 呢 (ne) and 呀 (ya), is presented, based on an analysis of their association with mood type and other particles. Specific attention is given to the question of why these two modal particles demonstrate a significantly different frequency while others do not.

4.2.1 呢 (ne)

As shown in Table 2, 呢 (ne) may be used in both declaratives and interrogatives, and in both cases its use is optional in realising the mood type. When occurring to express the declarative mood, it functions to signal that a proposition is contrary to what has been expected (Chappell 1991) and shows the speaker’s (or narrator’s) wish to convince the interlocutor (Cao 2005[11]). For instance, in Example (1a), the speaker states that he definitely does not want to be a monk. 呢 (ne) is added to further stress the statement of not wanting. In this case, 呢 (ne) is optional, as seen in Example (1b).[12]

In cases where 呢 (ne) is used to express the interrogative mood, it may be used in either an unbiased polar question (A-or-not-A type of question in English) or a content question (WH question in English). In the former case (that is, an unbiased polar question), the interrogative mood type is achieved by repeating the first element of a verbal group (either an auxiliary or a lexical verb) with a negative particle in between (Li 2007: 121), and therefore the use of 呢 (ne) is optional and an omission of the particle will not impact the interrogative meaning. In the latter case (that is, a content question), a further distinction should be made between two different situations. In instances where there is a co-occurrence with an interrogative word (WH-elements in English), 呢 (ne) may be elliptical. This is because the interrogative mood is actually realised by an interrogative word the use of which is compulsory, whereas the optional use of the modal particle 呢 (ne) functions to soften the tone so as to make an enquiry readily acceptable (Cao 2005 in Note 11. Therefore, the omission of 呢 (ne) would not cause significant loss of meaning. For instance, in Example (2a), the interrogative mood is basically conveyed by the interrogative word 哪(nǎ)儿(er) [where], whereas the modal particle 呢 (ne) is merely used to soften a direct interrogation. As shown in Example (2b), without 呢 (ne) the sentence is still acceptable.

In a different scenario, where an interrogative word which indicates the interrogative meanings of when, where, why, how and so on is omitted, the optionality of the modal particle 呢 (ne) changes and it then becomes compulsory to use the particle to realise the interrogative mood. However, this use is highly context-based where the omitted interrogative meaning must be recoverable in context. For instance, in Example (3a), from the answer given by the addressee, what 你(nǐ)家(jiā)里(lǐ)人(rén)呢(ne) [your families ne] means is actually 你(nǐ)家(jiā)里(lǐ)人(rén)哪(nǎ)里(li)去(qù)了(le) [where have your families been] and here the interrogative word 哪(nǎ)里(li) [where] has been omitted. In this case, the modal particle 呢 (ne) bears the function of realising the interrogative tone, without which, the interrogative mood disappears and the sentence is no longer grammatically correct (see Example [3b]). As the recovery of the interrogative meaning in such cases relies heavily on context, it may present a challenge to young readers whose ability in recovering the elliptical meaning may not yet be well developed. Translators may be aware of this and it is therefore reasonable to argue that a safe strategy in the translation of children’s literature is to avoid the omission of an interrogative word in translating a WH-interrogative, which in turn makes the use of 呢 (ne) optional. To sum up, except in interrogatives where an interrogative word is omitted, the use of 呢 (ne) is optional (Wang 2006).

Based on the discussion above, we believe there are mainly two potential situations where CLI has led to the significant under-use of 呢 (ne) in the translated Chinese texts in the TCCLC. Firstly, the modal particle 呢 (ne) is generally optional in realising the mood type in Chinese, either in a declarative or an interrogative. The optionality of the use of the modal particle gives translators choices: they may or may not use the particle in the translation. In this situation, the use of the particle could be viewed as a conscious choice of normalising the translated text to conventions in the TL genre. However, the under-use of the particle in the translated texts, compared with the non-translated texts, also indicates that CLI plays an important role in the translator’s decision-making: if there is no modal particle in the SL triggering the use of 呢 (ne) in the TL, the translator’s spontaneous cognitive response might be to translate without using it. Such a tendency is particularly evident in the case of a declarative. This point can be illustrated by a comparison of two similar situations found in the NCCLC and TCCLC, where the non-translation (in Example [4]) does make use of 呢 (ne) while the translation (in Example [5]) does not, even though adding 呢 (ne) sounds more idiomatic and would be typical for a native writer. It appears likely that it is the lack of an equivalent in the English ST that suppresses the use of 呢 (ne) in translation and that CLI inhibits translators’ use of the modal particle呢 (ne) despite their attempts to conventionalise usage to TL norms, resulting in a relatively lower frequency in the TCCLC.

Secondly, in the case of translating WH-interrogatives from English into Chinese, the effect of CLI could be further enhanced, as there are direct Chinese equivalent interrogative words that the translator can use in realising the interrogative mood, and translators may consider that the omission of the interrogative words in the translation may present a challenge to young readers. In this case, translators are more likely than not to maintain the use of an interrogative word in the translation, which could further discourage the use of 呢 (ne) since the use of the latter is optional in this context.

4.2.2 呀 (ya)

The modal particle 呀 (ya) can be used in an exclamative, an interrogative or an imperative, and similar to the case of 呢 (ne), it is optional to use 呀 (ya) in these mood types. Therefore, as an unique item in Chinese, the optional use of 呀 (ya) may help explain the existence of both trends in the translation choice: it indicates the normalisation trend when 呀 (ya), which has no equivalent in English, is used in the translation; and the trend of CLI as 呀 (ya) is found to be much less used in the translated texts in the TCCLC compared to native Chinese texts in the NCCLC.

However, apart from the interplay between normalisation and CLI, as observed so far, another factor also seems to play a role in the significantly more frequent use of 呀 (ya) in the non-translated Chinese books in the NCCLC. In Chinese, both modal particles 呀 (ya) and 啊 (a) can be used in an exclamative, an interrogative or an imperative, and in each case they are interchangeable (Chappell 1991). When used in an interrogative, similar to the case of 呢 (ne), 呀 (ya) or 啊 (a) expresses doubtful questioning. By using 呀 (ya) or 啊 (a) in an exclamative, the speaker (or narrator) indicates that what he or she is experiencing is astonishing or surprising. As illustrated in Example (6a), by using the modal particle 呀 (ya), the speaker indicates that what she is seeing is unexpected for her and this 呀 (ya) could be changed to 啊 (a), as illustrated in Example (6b).

When used in imperatives, 呀 (ya) has a hortatory use in prompting or urging the interlocutor to do something (Chappell 1991). In Example (7a), the modal particle 呀 (ya) is used by the speaker to give a command to urge the listener to take action. Again, 呀 (ya) and啊 (a) are interchangeable, as in Example (7b).

Comparing the use of 呀 (ya) and 啊 (a) in native, non-translated Chinese children’s books, there is no significant difference in frequency, as the two particles have nearly identical medians in the NCCLC (呀 [ya]: 0.79; 啊 [a]: 0.78). When compared with translations, however, a significant difference in frequency is evident for 呀 (ya), which is more frequent in the NCCLC (0.79 per 1,000 words) than in the TCCLC (0.23 per 1,000 words; U = 122, p < 0.05).

A likely explanation for this finding is that there may be some influence of writers’ and translators’ personal preferences. The quantile-quantile plot of 呀 (ya) shows non-normal distribution of the data (see Figure 1 in Appendix 3), with outliers in both the TCCLC and NCCLC. The most frequent use of 呀 (ya) in the TCCLC occurs in the book 红(hóng)头(tóu)发(fà)安(ān)妮(ni) [Anne of Green Gables][14], with a frequency of 3.98 times per 1,000 words, while the most frequent use of 呀(ya) in the NCCLC occurs in the books 下(xià)次(cì)开(kāi)船(chuán)港(gǎng) [“Next time depart” Bay][15] (4.92 times per 1,000 words) and 大(dà)林(lín)小(xiǎo)林(lín) [Big Lin and Small Lin][16] (4.97 times per 1,000 words). The frequency of 呀 (ya) in the rest of the texts in the two subcorpora varies from 0 to 1.81 times per 1,000 words. While the CLI effect driven by the Unique Items Hypothesis as well as the optionality of the use of 呀 (ya), as discussed above, in all likelihood account for the significantly higher frequency of 呀(ya) in the original subcorpus than in the translation subcorpus, there also appears to be some effect of individual preference by authors (and translators) in selecting 呀 (ya) rather than the inter-changeable alternative.

5. Conclusion

In sum, we have found that the overall use of modal particles demonstrates no significant differences between the two subcorpora, namely, translated and non-translated Chinese children’s literature. Furthermore, the investigation of each individual modal particle shows that only two of the five commonly used particles demonstrate significant differences in their frequencies, with the translation subcorpus having fewer occurrences. There is therefore evidence of both normalisation and CLI playing a role in conditioning the frequency and usage patterns of modal particles in translated Chinese children’s literature, in comparison with non-translated Chinese children’s literature. Translators’ attempts to normalise their translations to target-culture genre expectations are evident in the overall insignificant difference in the frequency of modal particles between the translated Chinese books and the non-translated original Chinese books. As the particles are unique to the TL, have no counterpart in the SL, are commonly associated with children’s literature in Chinese and are also, in most cases, optional, using them appears to be motivated by translators’ attempts to meet the expectations or conventions of the genre in Chinese. However, we have also found evidence suggesting that, in tandem with the impulse to normalise, there is a CLI effect: the relatively more frequent occurrence of modal particles in the original Chinese children’s books than in the translations suggests that the absence of equivalent constructions in the SL, a negative form of CLI (as captured in the Unique Items Hypothesis), pushes the translation towards the other direction on the cline of conventionality. Since English does not have a structure that is directly equivalent in form and function to Chinese modal particles, modal particles can be regarded as a unique feature of the TL. The activation of the SL (English) suppresses the activation of this unique feature of Chinese, which has no counterpart in English. Furthermore, the optional use of some particles in realising mood types in Chinese may have further suppressed their use in translation. Consequently, modal particles tend to be under-represented in the translation subcorpus compared to the non-translated subcorpus.

The results of this study do not support the claim that translations are relatively more conventional than comparable non-translated texts – at least not in respect to the feature investigated here, modal particles. That this is the case, even in a text type so strongly shaped by expectations of target-culture acceptability as children’s books, raises important questions about the nature of normalisation and its manifestation in different text types and language pairs. Clearly, normalisation and CLI, as competing tendencies in translation, are both in play, and their interplay is conditioned by a complex set of factors. On the one hand, there is normalisation demonstrated through the use of largely optional modal particles; on the other hand, there is the under-representation of the particles in the translated Chinese texts compared to the original Chinese, indicating the effect of CLI. In the “the tug-of-war” between the two competing tendencies, CLI seems to prevail, which is especially obvious in the case of two particular modal particles (呢 [ne] and呀 [ya]), which are found to be significantly less used in translation. The qualitative analysis further suggests that their occurrence could also be influenced by translators’ personal style or preference.

The findings therefore suggest, as other researchers (Dai and Xiao 2011; Cappelle and Loock 2013; Lefer and Vogeleer 2013) have found, that under some conditions CLI wins out over conventionalisation, potentially highlighting the fundamentally cognitive constraints of translation: interference is cognitively almost inescapable. What leads to tolerance for and the realisation of interference is certain prevailing socio-cultural conditions (“the relative prestige of cultures and languages”), translators’ professionality and text type (Toury 1995/2012: 311-315). Therefore, the law of standardisation could be counterbalanced, to different extents, by the law of interference, depending on the relative strength and interplay of these (and possibly other) factors.

This study makes a contribution to the growing body of research that considers the occurrence of and the reasons for normalisation, in relation to CLI. However, the corpus design of this study is comparable in nature, which limits the investigation of CLI and thus the tug-of-war between normalisation and CLI, since translations cannot be comprehensively and systematically compared with their source texts. A comparison of translations and non-translations, both in the SL and TL, will yield a more complete picture – an important avenue for future studies that aim to better understand the relationship between normalisation and CLI.