Cognitive load in remote simultaneous interpreting: place name translation in two Mandarin variants

The study investigates whether interpreting into a non-habitual language variant affects accuracy in remote simultaneous interpreting (RSI), focusing on Mandarin Chinese variants (mainland vs. Taiwanese). While these variants are largely mutually intelligible, they differ in phonology, syntax, and lexicon, which can influence listener perceptions and comprehension. Market realities and remote settings may require interpreters to work into variants they do not routinely use, potentially elevating cognitive load. The research question centers on how using a non-habitual Mandarin variant impacts interpreters’ ability to accurately render foreign place names over the course of an RSI task, with attention to potential fatigue effects. The study aims to inform training, recruitment, and practice by clarifying performance implications of variant switching.

• Lexical access and working memory in SI: SI demands concurrent listening and speaking under limited working memory resources. Automatic lexical access to high-frequency words and formulaic sequences in the interpreter’s habitual variant reduces cognitive load and errors. Bilingual lexical selection requires managing competition between two active languages; variant switching may add an inhibitory step, slowing access and increasing effort (Gile, 2009; Bialystok et al., 2008; Kirk et al., 2018).
• Language variants in SI and user expectations: Interpreters adapt output to context, including register and politeness strategies. Listeners often rely on formal cues (e.g., accent, variant) to judge quality and may react negatively to deviations from their norms. Awareness of language variants is recommended in training and practice (Cheung, 2022; Setton & Dawrant, 2016).
• Place names as problem triggers: Names and numbers are specific, objective units for assessing SI accuracy and are recognized problem triggers due to low redundancy and predictability. While many place names can be transcoded, difficulties arise when target forms are phonologically dissimilar to the source or differ by translation approach (transliteration vs. semantic translation). Lists of place names exacerbate difficulty by increasing cognitive load and reducing predictability (Atkinson, 1984; Gile, 2009; Chmiel & Spinolo, 2022).
• Mandarin variants and foreign name translation: Mainland and Taiwanese Mandarin often share transliteration practices but diverge for certain foreign person and place names due to sociopolitical and convention differences. Some items differ lexically while sounding similar; others differ both lexically and phonologically, or by approach (e.g., semantic translation in mainland vs. transliteration in Taiwan for “Montenegro”), potentially increasing difficulty when producing a non-habitual variant.

Design: Between-groups experimental study comparing accuracy of place name translation in RSI when interpreting into a habitual (mainland Mandarin; control) versus non-habitual (Taiwanese Mandarin; Taiwan group) variant. Three time-anchored lists of place names were embedded in a 30-minute English presentation; the third item in each list served as the test item because its Chinese translation differed between variants.

Participants: 67 professional English/Chinese conference interpreters based in mainland China were recruited; 52 recordings were analyzable (technical issues/withdrawals reduced the sample). Inclusion criteria: native mainland Mandarin speakers; born, and educated (primary through undergraduate) in mainland China in Mandarin; RSI experience; physically present in mainland China during the experiment. Groups: Taiwan group n=25 (instructed to use Taiwanese Mandarin translations), Control group n=27 (no special instruction, i.e., mainland Mandarin). Gender: Taiwan 6 male/19 female; Control 4 male/23 female. Professional experience: majority <10 years (Taiwan 22; Control 25), with a minority >10 years (Taiwan 3; Control 2).

Materials: Experimental speech based on an authentic online presentation (UN SDGs private sector). An English-speaking Singaporean male recorded a 30-minute talk (108 wpm) in a studio. Three lists of foreign place names were inserted at specific times: List 1 at 05:11 (…South Africa, Sudan, Côte d’Ivoire…), List 2 at 13:48 (…Kosovo, Serbia, Montenegro…), List 3 at 23:33 (…Cambodia, Myanmar, Laos…). In each list, the third item was the test item. Test items and variant-specific Chinese translations: (1) Côte d’Ivoire—Taiwan: 象牙海岸 (xiang-ya hai-an); Mainland: 科特迪瓦 (ke-te-di-wa). (2) Montenegro—Taiwan: 蒙特內哥羅 (meng-te-nei-ge-luo) [transliteration]; Mainland: 黑山 (hei-shan) [semantic translation]. (3) Laos—Taiwan: 寮國 (liao-guo); Mainland: 老撾 (lao-wo).

Preparation: One day prior, participants received group assignment and a bilingual glossary of 20 place names (including the 9 list items) with target translations in their designated variant, presented alphabetically; simplified Chinese characters used for both groups. No other prep materials.

Procedure: July 2022; online via a professional RSI platform. Interpreters logged in 1 hour early; two technicians conducted system/sound checks and collected RSI recordings. Participants had both audio and visual access to the presentation. Post-task, Mandarin interpretations were saved as MP3.

Transcription and coding: Automatic transcription via iFlyTek; two research assistants manually verified, focusing on the three test items. Scoring: For each test item, the interpreter’s rendition was coded correct if it matched the designated variant; otherwise incorrect. Incorrect included: use of the other variant’s form, mispronunciations, source-language forms, omissions, or substitutions (e.g., “and a neighboring country”). Self-corrections within the rendition to the correct variant were counted as correct. Order of items was ignored.

Analysis: Group comparisons for each test item using independent-samples t-tests on correctness (binary aggregated as mean correct per group).

• Tabled outcomes (counts correct/incorrect): • Côte d’Ivoire (05:11): Taiwan 22/3 (correct/incorrect), Control 26/1 — no significant group difference. • Montenegro (13:48): Taiwan 5/20, Control 20/7 — significant group difference. • Laos (23:33): Taiwan 2/23, Control 15/12 — significant group difference.
• Inferential statistics (means reflect proportion correct per participant): • Côte d’Ivoire: Taiwan M=0.88, SD=0.33; Control M=0.96, SD=0.19; t(50)=1.11, p=0.14 (ns). • Montenegro: Taiwan M=0.20, SD=0.41; Control M=0.74, SD=0.45; t(50)=-4.55, p<0.001. • Laos: Taiwan M=0.08, SD=0.28; Control M=0.56, SD=0.51; t(50)=-4.15, p<0.001.
• Error trends increased over time for both groups, consistent with fatigue or reduced attentional resources.
• Interpreting into a non-habitual variant (Taiwan group) was associated with significantly lower accuracy on later test items (Montenegro, Laos), suggesting growing cognitive load/fatigue effects when inhibiting the habitual variant and retrieving less-automatic forms.

The findings indicate that variant switching does not uniformly degrade performance: early in the task (Côte d’Ivoire) accuracy did not differ between groups, likely aided by forewarning via the glossary and higher attentional reserves. However, by mid and late stages (Montenegro and Laos), the Taiwan group, tasked with a non-habitual variant, showed significantly fewer correct renderings than the control group. This pattern supports the interpretation that sustained RSI with variant inhibition increases cognitive load and fatigability, diminishing accuracy over time. The Montenegro item exemplifies differences in translation approach (semantic translation vs. transliteration), which may demand additional processing. The Laos item further illustrates how phonological and lexical differences between variants can strain automaticity. Professional strategy trade-offs (e.g., omission, approximation) might have been used to preserve fluency under load but at the expense of accuracy. These results extend SI performance models by highlighting variant-specific lexical access and inhibitory control as additional constraints, reinforcing the need to consider language variants in recruitment, preparation, and training, especially in RSI contexts where fatigue effects can be pronounced.

This experiment shows that interpreting into a non-habitual Mandarin variant can impair accuracy for challenging items, especially as RSI progresses and fatigue accumulates. The study contributes by integrating non-habitual variant use into discussions of RSI accuracy and cognitive load, demonstrating time-dependent effects. Practically, it supports specifying language variants in recruitment, encouraging interpreter specialization or multi-variant preparedness, and incorporating variant-specific materials (e.g., divergent place names) into training. Future research should incorporate real-time cognitive load measures, examine item frequency effects with corpus checks, gather post-task strategy reports, expand participant samples, and test other languages with multiple variants (e.g., English, Portuguese) to generalize findings and refine training and professional standards.

• No physiological measures of cognitive load or fatigue due to remote setup; in-person data (e.g., pupillometry, heart rate variability) could clarify mechanisms.
• Test item frequency was not controlled via reference corpora; familiarity differences could affect accuracy.
• No post-task survey/interviews to capture decision-making strategies and motivations.
• Sample size, while moderate, could be larger for greater statistical power and robustness.
• Generalizability may be constrained by Chinese-specific variant differences in foreign name translation; cross-linguistic replication is needed.
• Remote design precluded tighter environmental control but improved ecological validity for RSI contexts.