Conclusion

Is structural difference in a language pair associated with difficulty in translating or interpreting complex sentences in that pair? The tentative answer given here to that question is: Yes. This study has established preliminary associations between structural difference in a language pair and recorded rates for indicators of difficulty in translating or interpreting complex sentences in that pair, with some differences according to the mode of language transfer.

Can anything be done to reduce that difficulty? In my opinion, the short answer is: Not much. Various tactics can be practiced and applied for coping with the working memory constraint in simultaneous interpretation, as described in the previous section. But the main take-away from this study is that translating or interpreting complex sentences can be much more difficult between languages with very different structure than between structurally similar languages – so much so, particularly in simultaneous interpretation, that the challenge can seem insurmountable. Even if not much can be done about it, it can still be useful to be aware of that difficulty – its nature, its causes, how intractable it can be, as well as its potential effects.

This study is now complete. It’s proposed a semantic parsing method for segmenting complex sentences in a way that allows for better cross-linguistic comparison than a syntactic approach. It’s used that method to count and record rates for indicators of difficulty in translation or interpretation into five languages from typologically different families. And it’s associated those rates with structural difference in the language pair, sentence complexity and mode of language transfer.

When we say “languages with very different structure,” what languages are we talking about besides the ones considered here, especially among languages likely to be used in professional translation or interpretation? Based on typological classifications by Dryer (2013a) and Schmidtke-Bode and Diessel (2017), languages with the same complex sentence structure as English (relative and complement clauses which both typically branch to the right) include most Indo-European languages, as well as languages like Arabic, Hebrew, Indonesian, Swahili, Thai and Vietnamese. Languages with somewhat different structure from English (relative clauses which typically branch either way and complement clauses which typically branch to the right) include Armenian, Finnish and Hungarian. Languages with moderately different structure from English (relative clauses which typically branch to the left and complement clauses which typically branch to the right) include Sino-Tibetan languages, Basque and Georgian. Languages with opposite structure from English (relative and complement clauses which both typically branch to the left) include Turkish, Japanese and Korean.

The findings of this study may help confirm the impressions of many professionals, who feel from experience that translating or interpreting complex sentences can be much more difficult between languages with very different structure than between structurally similar languages. The findings also confirm that structural difficulties manifest themselves differently in different modes of language transfer. Because of space and time constraints in subtitle translation and the working memory constraint in simultaneous interpretation, those two modes appear to be associated with more parallel transfer and easier-to-process output structure than legal translation. But that greater ease in production and processing seems to come at the price of more changes in hierarchical structure and therefore potentially in meaning.

Impressions of the comparative difficulty of transferring complex sentences between languages with major differences in structure can seem obvious to professional translators or interpreters with relevant experience. Still, stating such impressions explicitly can be somewhat taboo. This is true for a number of reasons, including good ones like professional solidarity. This study hopes to go some way towards dispelling that taboo, by helping to highlight structural difference in a language pair as a major potential factor of difficulty – in addition to other linguistic and cultural factors which can complicate the task of translating or interpreting between languages from different families and different parts of the world.

Particularly for simultaneous interpretation, the natural constraint on working memory can make it nearly impossible to interpret complex sentences between structurally very different languages with anything approaching the completeness, accuracy, emphasis and style of a good written translation.

To draw an analogy, a translator can be likened to a swimmer in a pool. They first survey a sentence, then dive in and start swimming forward. The greater their expertise, the more skillful they are in technique, elegance and speed. And they’re in control, as the water of the text is still, allowing them to proceed as quickly or as slowly as they like.

Continuing the analogy, a simultaneous interpreter can be likened to a swimmer in a river with a current. They have less control as they swim than the translator does in the pool. Where the current isn’t too fast, they have time to maneuver around obstacles. They can speed up or slow down relative to the flow. But ultimately it’s the speed of the current which determines their pace. And if they’re swimming downstream, the direction of flow helps immensely by propelling them in the right direction.

This propelling effect is so essential and so constant that interpreters take it for granted when swimming downstream – working between languages where propositions follow each other in similar order. It’s only if an interpreter has to swim upstream – working between languages where propositions come at them in reverse order – that they realize how tough it can be to fight the flow. If the current isn’t too strong, they can manage, though with considerably more effort than swimming downstream, and less gracefully. But if the speaker’s propositions come rushing at the interpreter in a very different order from an order in which they can process and reformulate them coherently, the task can become overwhelming.

When professional simultaneous interpretation was first developing, some experts claimed that interpretation between languages with very different structure was simply not possible: “Some languages, such as Japanese, do not permit simultaneous interpreting, due to the complexity of their grammatical structure” (Bower 1959, cited in Davidson 1992: 1). “That’s impossible,” said Sen Nishiyama, one of the pioneer consecutive interpreters in Japan, in 1945. “The word order of English and Japanese is exactly opposite. It just can’t be done” (Torikai 2009: 92).

This study suggests that there may be some truth to those early impressions. Major structural difference in some language pairs (like between a European language and Turkish, Mandarin or Japanese) may well mean that it simply isn’t possible to interpret complex sentences in such a pair with the same degree of accuracy, detail and coherence as between two European languages. For different reasons and perhaps to a different extent, the same may also apply to subtitle translation in a structurally very different pair. As for legal translation between languages with very different structure, the difficulty may be felt mostly by the translator and reflected less in the output than for the other two modes – although that output may also be harder to read than the output of translation in a structurally similar pair.

The semantic parsing method proposed in this study is fit for its purpose, as confirmed by the high degree of consistency found in the reliability check. But there’s room for improving it. In particular, an attempt could be made to define more complete and objective criteria for deciding if a given proposition is an assertion and is therefore functionally independent. An attempt could also be made to define criteria for confirming whether syntactically nested or grammatically deranked propositions, which are characteristic of some languages, have less assertive force and are therefore more functionally subordinate than corresponding structures in other languages.

An approach like the one taken in this study could be applied to different genres. A more rigorous version, or a partially automated version, of the proposed semantic parsing method could be developed. Some issues raised here could be investigated in more depth, such as comparing the cognitive difficulty of processing different numbers and levels of nested structures, measuring the cognitive difficulty of reordering propositions, or making a typology of changes in semantic relations or other changes characteristic of translating or interpreting between languages with very different structure. Other related issues could be explored further, such as the importance of preserving logical order or rhetorical effect in translation or interpretation between languages with very different structure. Two issues of output quality in simultaneous interpretation of complex sentences highlighted in the reliability check are the tendency to restate and reformulate information and the tendency to use non-standard syntax. It could be interesting to compare simultaneous interpretation of speeches with complex sentences into typologically different languages, in terms of how natural and clear the interpretation is perceived to be, and how well it’s understood.

I hope the findings of this study will prove interesting and useful to linguists, students, teachers, and professional translators and interpreters. It may help confirm impressions from experience, inform individual or policy decisions, or provide a basis for targeted training or future research. If so, it will have achieved its aim.