Where empathy goes untranslated: the Chinese reflexive Ziji and the recoding of perspective in two complete English translations of Hong Lou Meng

The translation of literary narrative across typologically divergent languages often exposes deep chasms in how human consciousness and narrative perspective are grammatically encoded. A new empirical study published in Frontiers in Psychology sheds fresh light on these mechanics by examining the Mandarin polyfunctional anaphor ziji (自己) and its English renderings across two complete, independent translations of the classic 18th-century Chinese novel Hong Lou Meng, widely known in English as The Dream of the Red Chamber or The Story of the Stone. Authored by researcher Hongyu Zhu, the investigation utilizes a purpose-built parallel corpus to analyze how self-reference and empathetic viewpoint are systematically transformed, retained, or relocated when moving from a "speaker-anchored" language to a "structure-anchored" one.

The Research Framework and Corpus Architecture

To capture the complex behavior of ziji, the study constructed a meticulous parallel corpus comprising 1,296 unique source sentences containing the reflexive token, drawn from all 120 chapters of Hong Lou Meng. These sentences were aligned with two monumental English translation projects: the version by Yang Xianyi and Gladys Yang, published between 1978 and 1980 by the Foreign Languages Press, and the Penguin translation initiated by David Hawkes (chapters 1–80, published 1973–1980) and completed by John Minford (chapters 81–120, published 1982–1986).

Because the novel’s profound literary artistry depends heavily on rapid, often unmarked shifts into and out of the interior psychological landscapes of its characters, ziji functions frequently as a logophoric marker. In these contexts, it signals that the narrative voice has momentarily merged with a character’s inner vantage point, serving as what linguists term an empathy expression. However, English grammar imposes strict structural boundaries on its own reflexive pronouns, such as himself or herself, which are typically clause-bound and lack the sweeping perspectival affordances of the Mandarin term.

To evaluate how translators navigate this mismatch, every source token was annotated for its syntactic function across four distinct categories: subject position (SUBJ), object position (OBJ), adnominal possessive (POSS), and appositive intensifier (INT). Concurrently, every target rendering was classified under a five-way strategic schema: reflexive pronouns (REFL), appositive intensifiers (INTENS), possessive own (OWN), lexicalization (LEX), and zero marking (ZERO), where no overt reflexive or intensifying form appears at the corresponding site.

Quantitative Findings: The Dominance of Zero Marking

The empirical analysis yielded three principal findings that challenge conventional assumptions about literary translation and cross-linguistic equivalence. First, the study revealed that the modal formal outcome for ziji in both English translations is ZERO—meaning the complete absence of an overt reflexive or intensifying form at the target site. Specifically, the Yangs translation left 63.5% of ziji sites without overt marking, while the Hawkes–Minford translation recorded a zero rate of 49.1%.

Rather than indicating random omission, this formal absence was found to track structural affordances within English grammar precisely. Zero rates peaked dramatically where English lacks a categorial equivalent—namely, subject-position ziji, where the Mandarin term acts as the nominative subject of a finite clause. Zero rates dropped to their lowest levels in object position, where English supplies a direct equivalent in pronouns like himself. This identical gradient (SUBJ > POSS > INT > OBJ) across both translation teams underscores that zero rendering acts as the visible footprint of a fundamental typological gap.

Furthermore, when testing for broader contextual cues—specifically whether ziji was preceded by a predicate of cognition, perception, or speech—the data revealed no statistically significant association with zero marking in either translation. Consequently, the researchers emphasize that high zero rates measure local formal absence rather than an automatic, corpus-wide loss of viewpoint.

The Translator Paradox: Reversing Conventional Characterizations

Perhaps the most startling quantitative discovery concerns the systemic differences between the two translation teams. Within translation studies scholarship, the Yangs’ version has long been characterized as source-oriented, literal, and culturally faithful due to its institutional mandate under the Foreign Languages Press, whereas the Hawkes–Minford Penguin edition has been viewed as reader-oriented and comparatively free.

However, when evaluated on a strictly parallel sub-corpus of 318 fully paired sentences spanning chapters 81 to 120—where both teams translated identical source inputs—Hawkes–Minford retained overt reflexive and intensifying marking significantly more often than the Yangs. McNemar’s test confirmed this divergence with high statistical significance ($p < 0.001$), revealing that the "free" translation stayed closer to the surface morphology of the source text than the supposedly literal translation.

Qualitative analysis helps explain this paradox through the lens of explicitation. The Yangs’ prose is notably economical, pruning redundant self-referential markers to achieve stylistic concision. In contrast, Hawkes and Minford frequently expand upon the source text, deploying devices like himself, he himself, or his own to keep the character’s self-relation visible on the surface of the English sentence, even if those devices do not fully replicate Mandarin logophoric dynamics.

Qualitative Insights: Relocation of Perspective and Discourse Architecture

While quantitative counts establish patterns of formal presence and absence, the study’s qualitative case studies demonstrate that viewpoint preservation cannot be judged solely by surface-level morphology. Through detailed examinations of representative passages, the research outlines three distinct fates for empathetic ziji: perspective reduction, containment in weaker English forms, and relocation into broader discourse architecture.

For instance, when subject-position ziji is rendered via syntactic restructuring—such as converting a perceptual sequence into an absolute participial phrase in English—the internal focalization can be externalized, reducing the reader’s immediate access to a character’s subjective despair. Conversely, adnominal uses of ziji frequently find a reliable receptacle in the English possessive own, which successfully re-anchors possessions to a character’s sphere of consciousness when licensed by an attitude-holder matrix verb.

Most intriguingly, the study uncovers instances where the loss of a localized reflexive pronoun is compensated for by a redistribution of perspective across the surrounding discourse. Translators occasionally rebuild self-address through dramatized second-person dialogue paired with vivid stage directions, or capture profound solitude through lexical accumulation, combining reflexive pronouns with descriptive adjectives and adverbial modifiers. These stylistic maneuvers suggest that viewpoint equivalence between a speaker-anchored language and a structure-anchored language must be audited at the level of the discourse unit rather than through isolated, form-by-form comparisons.

Broader Implications for Translation Pedagogy and Typology

The implications of this research extend far beyond Chinese-English literary scholarship. By demonstrating that translational zero-marking is governed by structural affordances rather than arbitrary preference, the study provides a robust empirical model for investigating anaphora and logophoricity across languages.

For translation practitioners and educators, the findings offer three practical maxims. First, translators must audit the discourse function of terms like ziji before selecting a target equivalent, asking whose consciousness a clause inhabits rather than defaulting to mechanical word-matching. Second, translators should actively leverage possessive own as a primary vehicle for maintaining perspectival texture. Third, reverse-direction translators working from English into Chinese must remain vigilant against missing implicit interiority, ensuring that empathy markers are appropriately planted to prevent externally narrated readings of internal focalization.

Ultimately, the study refines our understanding of source-orientation and target-orientation in translation studies, portraying them not as fixed positions on a single scale, but as complex bundles of level-specific habits. As natural language processing and parallel corpus methodologies continue to evolve, investigations of this caliber provide vital natural experiments in how human languages construct, negotiate, and translate the inner lives of narrative characters.