Linguistic variation in mediated diplomatic communication: a full multi-dimensional analysis of interpreted language in Chinese Regular Press Conferences

The study examines how interpreted diplomatic language differs from non-interpreted diplomatic speech by analysing co-occurring linguistic features through a full multi-dimensional analysis (MDA). It responds to limitations in prior corpus-based interpreting research that often relied on translation universals, manually selected features, and limited statistical approaches, and that underexplored contextual influences on interpreted language use. The context is Chinese Regular Press Conferences (Ministry of Foreign Affairs) compared with non-interpreted U.S. Department of State press briefings—both high-stakes institutional settings in which spokespersons address journalists’ questions. The purpose is to uncover holistic, co-occurrence-based patterns that characterise interpreted diplomatic language and to relate these to contextual factors. Research questions: (1) What similarities and differences exist between interpreted diplomatic language and non-interpreted counterparts in co-occurring linguistic patterns? (2) Which contextual factors contribute to these patterns in interpreted diplomatic language? The study is important for advancing corpus-based interpreting studies and for informing pedagogy in diplomatic interpreter training by revealing multi-dimensional variation beyond isolated features.

Research on diplomatic language highlights precision, formality, ambiguity, and strategic persuasion, emphasizing interpreting’s crucial role. Prior work on interpreted diplomatic discourse has examined norms, agency, and selected features (e.g., lexical density, modality, hedging), showing complex mediation but often overlooking broader co-occurrence patterns and contextual effects. Limitations include: insufficient attention to situational parameters (field, mode, tenor), narrow feature sets, and reliance on difference tests that may miss underlying functional dimensions. Multi-dimensional analysis (MDA) offers a robust alternative, clustering features into functional dimensions via factor analysis. Standard MDA (Biber, 1988) uses predefined features, while full MDA inductively selects features tailored to the discourse domain. Although MDA has grown in translation and interpreting studies, applications to interpreted diplomatic language remain limited. Existing studies (e.g., Zou & Wang, 2021; Sheng & Li, 2024) indicate interpreted diplomatic language can be more literate, information-dense, and less involved, but they are constrained by predefined feature sets, statistical assumptions, or limited control of speaker effects. This study addresses these gaps by applying a tailored full MDA to interpreted diplomatic press conference discourse, aiming to reveal nuanced, context-specific co-occurrence patterns.

Design: Full multi-dimensional analysis (MDA) comparing simultaneously interpreted (SI) English outputs from Chinese Regular Press Conferences with non-interpreted (NS) English from U.S. Department of State press briefings. Corpus: Transcripts sourced from official MFA (English) and U.S. State Department websites. Only spokespersons’ replies retained; paralinguistic elements, abbreviations, and bracketed notes removed. - SI: 515 texts; 1,140,302 tokens; mean text length 797; period 2020-09 to 2022-11. - NS: 242 texts; 1,062,135 tokens; mean text length 4389; period 2021-02 to 2022-11. Feature selection and extraction: Initial 113 linguistic features (67 grammatical, 40 n-grams, 6 mediated-language textual features) identified from prior literature. After grouping/removing certain grammatical and punctuation-related features and excluding very rare features (<0.003 per 100 tokens), 88 features remained. Tools: MAT 1.3.3 for 54 grammatical features; PatConc for 40 n-grams; WordSmith 7.0 for word-length measures; MATLAB R2020b for Top 10 vocabulary coverage; Textalyser for lexical density. Frequencies normalized per 100 words; z-scores computed per text. MDA procedure: Principal factor analysis (principal axis factoring), Varimax rotation; absolute loadings ≥0.30 retained. Data adequacy: KMO=0.94; Bartlett’s test χ2=55,569.586, p<0.001. Scree plot supported a five-factor solution explaining 42.80% shared variance. Dimension scores per text computed as sum of positive-loading feature z-scores minus sum of negative-loading feature z-scores; dimensions were interpreted functionally. Statistical testing: Mann-Whitney U tests compared SI vs NS at dimension level, feature level, and aggregated overall style scores (non-normal distributions). Overall formal-informal style scores computed by weighting dimension scores by explained variance; D3 scores reversed to align polarity of formality/orality.

Five dimensions identified: (1) Involved vs. Informational Production, (2) Objective vs. Addressee-focused Narration, (3) Literate-Oral Continuum, (4) Information Elaboration, (5) Narrative vs. Non-narrative Concerns. Significant SI–NS differences on D1, D2, D4, D5; no significant difference on D3. - D1 (Informational vs Involved): SI loaded negative (more informational): mean −25.74 (SD 11.82); NS positive (more involved): mean 54.77 (SD 11.55); z=−22.209, p<0.001. Negative features include lexical density, longer words, attributive adjectives, nominalizations, noun-centric colligations; positive features include first-person pronouns, demonstratives, contractions, present tense, analytic negation. - D2 (Objective vs Addressee-focused Narration): SI positive (more objective): mean 0.93 (SD 2.65); NS negative (more addressee-focused): mean −1.99 (SD 5.26); z=6.651, p<0.001. Objective side includes perfect aspect, adverbial subordinators, downtoners, pronoun it; addressee-focused side includes second-person pronouns, public verbs, predictive modals. - D3 (Literate–Oral Continuum): No significant difference (SI mean 0.07, SD 3.06; NS mean −0.16, SD 2.67); z=0.675, p=0.500. Positive end features: prepositional phrases and noun-centered colligations (art+N+of; prep+art+N+of), indicating literate, information-dense style; Top 10 vocabulary coverage indicates spoken constraints. - D4 (Information Elaboration): NS higher elaboration (mean 1.19, SD 2.98) vs SI lower (mean −0.56, SD 2.68); z=−7.888, p<0.001. Positive features include adjective-related colligations (V+adv+adj; be+adv+adj; adv+adj+N) and amplifiers. - D5 (Narrative vs Non-narrative): NS more narrative (mean 0.47, SD 3.46) vs SI more non-narrative (mean −0.22, SD 2.58); z=−4.489, p<0.001. Narrative side: past tense, third-person pronouns, seem/appear, longer sentences; non-narrative side: split auxiliaries, pp+prep+V-ing. Overall style: SI significantly more formal than NS by the weighted overall score (SI mean −16.71, SD 7.77; NS mean 35.55, SD 7.82); Mann-Whitney U: z=−22.209, p<0.05. Both SI and NS occupy similar positions on the literate-oral continuum but SI aligns more with formal registers.

Findings indicate that interpreted diplomatic language in Chinese press conferences tends to be more informational, objective, non-narrative, and overall more formal than non-interpreted U.S. press briefings. These patterns reflect contextual influences: communicative aims to efficiently convey authoritative policy content, strict interpreting norms emphasizing adequacy, clarity, and explicit logical relations, and the structured Q&A format that limits interactive dynamics in Chinese settings. By contrast, U.S. briefings encourage interaction, follow-ups, and conversational tone, yielding greater involvement, addressee focus, elaboration, and narrative elements. Similarity on the literate-oral continuum is attributed to the inherently scripted and high-stakes nature of diplomatic discourse and to interpreters’ tendency toward literate constructions, balancing spoken delivery with careful composition. Reduced information elaboration in SI may stem from shared background knowledge assumptions and interpreting constraints (time pressure, cognitive load), encouraging focus on novel information while relying on implicature for shared content. The non-narrative emphasis in SI aligns with expository and argumentative goals of diplomatic messaging—clarifying policies and positions rather than recounting events. The overall higher formality of SI is consistent with the seriousness and sensitivity of diplomatic topics, multilingual audiences, and Chinese transcription practices that record edited exchanges, all of which constrain interpreters’ stylistic flexibility and promote a formal, precise register.

This study applies full MDA to interpreted diplomatic language, identifying five functional dimensions and demonstrating that, relative to non-interpreted diplomatic speech, interpreted output is more informative, objective, less elaborated, more non-narrative, and overall more formal, while both varieties similarly traverse the literate-oral continuum. The work advances corpus-based interpreting studies by uncovering co-occurrence patterns tied to communicative context and offers pedagogical implications for diplomatic interpreter training—providing empirically grounded targets for stylistic and strategic practice. Future research should refine factor solutions and validation procedures and expand to other institutional contexts to test generalizability and explore speaker- and event-level influences.

Methodological constraints include use of a fixed loading threshold rather than sample-size-adjusted thresholds, reliance on scree plot instead of parallel analysis for factor number determination, and total explained variance below 50%, suggesting room for improved modeling. Cross-loadings indicate overlap among dimensions; future work should consider oblique rotations (e.g., Promax, Direct Oblimin) and sensitivity analyses across rotation methods. Results may be sample-specific due to corpus composition and transcription practices.