Translation Recognition in Mandarin Speakers

Background

Translation is about finding the equivalent meaning to a word in a different language, but there very rarely exists complete term-to-term equivalence across languages. This partial conceptual equivalence explains why some second language (L2) learners appear to have a good command of L2 vocabulary but still find themselves unable to convey a certain message from one language to another if an exact translation equivalent does not exist. This is especially the case when the two languages of the speaker are not closely related (e.g. Chinese and English). Despite the ‘messiness of bilingualism’ (Pavlenko, 2011), people are still able to understand and use a second language by ‘re-naming the world’ during L2 acquisition (Pavlenko,2011). This raises an interesting question as to how and to what extent cross-linguistic semantic differences can have effects on translation processes. To date, there has been no research that explores the degree of semantic similarity between words from two highly differentiated languages to examine the role of cross-linguistic semantic similarity in bilingual processing. The current study is designed to fill in the gap by investigating how cross-linguistic semantic differences between Mandarin Chinese and English words can impact translation performance of Mandarin-English bilinguals.

Research Questions / Hypotheses

The main aim of the present study was to investigate how cross-linguistic semantic similarity can have effects on translation performance of different-script bilinguals as a function of lexico-semantic characteristics, and to determine whether these semantic effects are modulated by translation direction and bilinguals’ level of L2 proficiency.    Four specific aims and hypotheses were proposed:    1. To investigate the impact of cross-linguistic semantic similarity on translation recognition response times (RTs)    2. To investigate whether cross-linguistic semantic similarity affects translation RTs as a function of lexico-semantic word properties namely concreteness, valence, and word frequency.    3. To examine whether the effect of cross-linguistic semantic similarity is influenced by translation direction (L1-L2/L2-L1).    4. To investigate whether the effect of cross-linguistic semantic similarity is influenced by bilinguals’ level of L2 proficiency.

Participants

One hundred thirty-five participants (109 female) were recruited via online advertising and through the University of Melbourne Research Experimental Program. Nine participants who listed English and Cantonese as their first-acquired language in the LEAP-Q were removed from the dataset, leaving a final sample of N=126.

Methods

Procedure:    Each participant completed three online tasks in the following order (1) Translation Recognition Task, (2) the LexTALE test, and 3) the LEAP-Q.    Materials:    For translation ‘target’ materials that require a ‘yes’ response to the Translation Recognition Task (TRT), stimuli consisted of 216 pairs of translation equivalents taken from the Mandarin Chinese translation norms currently being collected by De Deyne (2020). Each translation pair consisted of a cue word, either in Chinese in the forward translation condition or in English in the backward translation condition, and a target word that is the dominant translation produced by at least 50% of the native Mandarin participants in De Deyne’s translation norm study.    The stimuli also consisted of 216 non-translation pairs in which the translation target was a distractor requiring a ‘no’ response to the TRT. Half of the non-translation pairs were selected from the same 12 semantic categories as the target materials, and each of the cue word was paired with a distractor that belonged to the same semantic category (e.g. income paired with 财富cáifù [wealth], both from the category work). For the other half of the non-translation materials each cue word was paired with a distractor from a different semantic category (e.g. wife paired with 秋天qiūtīan [autumn]). The two types of distractors provided a mixture of semantic similar and dissimilar non-translation pairs and thus covered a wide range of semantic similarity.

Results

The findings indicated that cross-linguistic semantic similarity was a significant predictor of translation times, and it had distinct effects on translation ‘target’ and ‘distractor’ materials, such that greater similarity speeded recognition of translation pairs but slowed down recognition of non-translation pairs.    Lexico-semantic variables including concreteness, valence, and word frequency were also found to be significant predictors of translation RTs. Yet for translation pairs, there was no interaction of these factors with the similarity effect. For non-translation pairs, the similarity effect was dependent on concreteness and valence, but not on word frequency.    The present study also aimed to examine whether the two critical assumptions of the MHM can be generalised to Mandarin-English bilinguals whose languages differ in script. Inconsistent with the translation asymmetry assumption, we found that translation RTs did not differ significantly between the two directions. A significant interaction between the similarity effect and translation direction was observed only among non-translation pairs, such that backward translation was more interfered by semantic similarity than forward translation. Regarding the developmental hypothesis, we found that bilinguals’ level of L2 proficiency did not modulate the interaction between the similarity effect and translation condition.

Implications

The present study’s results possess important educational implications for second language learning and teaching. Our findings suggest that different word properties can have distinct impacts on the translation process. This finding highlights the need to utilise distinct teaching and learning strategies for different types of words. For instance, abstract words are more difficult for L2 speakers to translate than concrete words because abstract words tend to share less semantic overlap with their translation equivalents. Therefore, when it comes to learning abstract L2 vocabulary items, L2 understanding may benefit from strengthening connections between the L2 lexicon and the conceptual system, so that direct links can be established between the lexical and conceptual representation of the L2 word with decreased reliance on its L1 equivalent (Tytus, 2014). When it comes to learning concrete vocabulary items that tend to have a high degree of semantic overlap with their translation equivalents across languages, learning might be facilitated by strengthening direct links between L2 words and their translation equivalents in the L1 lexicon.