Beyond Accuracy versus Style

---

What fidelity asks of an English Works of Love

Series: A Corrigible Kierkegaard, Essay 2
Status: public scholarly working paper; open for correction
Version: 0.1, 19 August 2026

Abstract

Discussion of English Kierkegaard often contrasts accurate but awkward translation with stylish but less exact translation. This essay argues that the opposition is inadequate, although the underlying tensions are real. Accuracy is not a single property: it includes lexical distinction, grammatical relation, conceptual recurrence, rhetorical movement, tone, biblical resonance, and completeness. Style is not a decorative addition to meaning: in Kierkegaard, repetition, direct address, humour, feigned dialogue, and long argumentative pressure often perform the thought. Yet readability also matters, because an English calque may obscure a relation that the Danish makes clear. Drawing on criticism of the Swenson, Hong, and Hannay traditions and on examples from the present working translation, the essay proposes a multidimensional review rather than a ranking of translators by one virtue. A close/readable architecture can help, but only if both layers remain accountable to the same complete Danish source. The aim is modest: not to resolve the long-standing debate, but to state the trade-offs in a form that readers can test passage by passage.

Keywords: Kierkegaard; translation; accuracy; style; readability; repetition; rhetoric; Works of Love

1. The appeal of a false choice

Readers of Kierkegaard translations often recognise a familiar contrast. One version stays close to Danish words and syntax but sounds laboured. Another reads as compelling English but appears to have made interpretive decisions on the reader’s behalf. The first is called accurate; the second stylish.

The contrast captures real experiences, but it becomes misleading when either term is treated as simple. A grammatically close sentence may mislead an English reader because the English construction does not function as the Danish one does. A beautifully recast sentence may preserve Kierkegaard’s argumentative force while changing a recurring word upon which a later distinction depends. Neither “literal” nor “readable” settles the matter.

M. G. Piety closes her comparison of Swenson and Hong by expressing the hope that translators will move beyond the style–accuracy dichotomy.¹ Her examples also show why the dichotomy persists. A translation may be preferred for voice and still contain an error. Another may be less attractive in English and nevertheless preserve a distinction the first erased. Intellectual fairness requires both judgements.

Alastair Hannay describes the difficulty in complementary terms. Kierkegaard writes in multiple genres and in a Danish now culturally and linguistically distant even from many Danish readers. A translation that allows his presence to be felt must somehow hold together the seriousness underlying the text and the liveliness of its surface.² “Somehow” is where most of the work lies.

2. Accuracy has several objects

To call a translation accurate is incomplete unless one asks: accurate to what?

2.1 Lexical distinction

When Kierkegaard places Christenhed and Christendom in one passage, their difference carries the contrast between Christendom as a social-historical condition and Christianity as the teaching or faith. Rendering both with one English word loses an object that accuracy should preserve.

2.2 Grammatical relation

Small words and endings can determine the argument. The scope of er det in the passage on the certainty of being loved distinguishes a claim that makes sense from one that does not. The enclitic Danish definite article in Stavningen does not necessarily identify a special kind of spelling. Accuracy here lies in construing function, not reproducing visible form.

2.3 Conceptual recurrence

Kierkegaard develops concepts through repetition. Kjerlighed, Elskov, Forkjerlighed, Næsten, Pligt, Gjerning, and their compounds establish relations across chapters. Unmotivated variation in English can dissolve those relations. Mechanical consistency can also distort them where one Danish word changes force by context. Accuracy must therefore attend to both recurrence and range.

2.4 Rhetorical movement

A long Kierkegaardian sentence may make a reader undergo qualification, reversal, suspension, or acceleration. Dividing it into short English assertions may state the propositions while changing the experience of reaching them. Conversely, preserving every clause boundary may turn intelligible Danish motion into English fog. The translator must decide what the syntax is doing before deciding how much of it English can carry.

2.5 Voice and address

Kierkegaard’s “you,” questions, exclamations, apostrophes, ironies, and staged interlocutors are not detachable ornament. They reposition the reader. “You shall love” is both quoted command and present address. A version that converts it into impersonal exposition may be lexically sound while rhetorically inaccurate.

2.6 Completeness

A translation cannot be accurate to a paragraph it omits. The project’s own early omission and compression findings are a reminder that completeness precedes finer evaluation.³

These objects may conflict in a particular sentence. The difficulty does not make judgement arbitrary. It makes the grounds of judgement plural.

3. Style is not polish

“Style” can sound optional, as though meaning were first transferred and then improved by a writer. That picture is ill suited to Works of Love.

Kierkegaard’s prose repeatedly enacts its claims. The hidden spring is approached through accumulating comparisons that prevent the reader from mastering its depth. The command to love presses upon a second-person “you.” The discourse on perseverance forces incompatible movements—lying at the point of death and walking upright, hurrying with time and lingering with eternity—into the same imaginative field. The argument about reading and spelling moves from the fluent grandeur of a resolution into the dismembered hours of daily action.

These effects belong to meaning because they train attention. To preserve only the conclusion would produce an abstract Kierkegaard who tells readers what he thinks without making them feel the resistance his thinking addresses.

Kristiina Taivalkoski-Shilov’s account of voice in machine-assisted literary translation is relevant here. Literary voice arises through recurrent textual features and layered speakers; technological systems may homogenise those features even while producing acceptable sentence-level prose.⁴ Kierkegaard’s voice is especially vulnerable because it contains other voices: “the poet,” the worldly objection, the anxious learner, the scriptural command, and the author who speaks “without authority.”

Style, then, is not whatever makes the English elegant. It is the patterned manner in which the text thinks and addresses.

4. Three passages from the working translation

4.1 The hidden lake

The close layer of Chapter I begins:

Love’s hidden life is in the innermost, unfathomable, and then again in an unfathomable connection with the whole of existence.

The readable layer begins:

Love’s hidden life is in the innermost, beyond fathoming, and in turn has an unfathomable connection with existence as a whole.⁵

The close version makes the recurrence of uudgrundeligt conspicuous and retains “the whole of existence.” It also sounds translated. The readable version varies the first occurrence to “beyond fathoming” and moves towards ordinary English rhythm while retaining “unfathomable” in the second position. It sacrifices exact verbal repetition to avoid a heavy opening.

Which is more accurate depends on the use. A reader tracing recurrence needs the close version. A reader entering the paragraph’s lake-and-spring imagery may find that the readable cadence better carries the movement. The important fact is not that both versions exist, but that the variation is visible and defensible. If the repeated adjective is judged conceptually essential, the readable choice should change.

4.2 Spelling the resolution

Kierkegaard writes that daily particulars resemble Stavningen, der river Ordene fra hinanden og i Stykker. The working translation gives:

They resemble spelling, which tears words apart into pieces.⁶

This is stylistically plain and grammatically sound. It avoids making the definite form mean “the spelling,” as though Kierkegaard had identified an eccentric spelling method. Yet “into pieces” partly duplicates “apart,” and the Danish fra hinanden og i Stykker carries both separation from one another and breaking into parts. A tighter readable candidate—“They resemble spelling, which pulls words apart and breaks them into pieces”—may carry the two actions more distinctly.

No appeal to literalness decides between them. The grammar excludes one error; rhythm and semantic distribution decide among the remaining English options.

4.3 Standing certain of love

In the passage on Christ and Peter, the current text reads:

Humanly speaking, to stand unconditionally certain of being loved is not to love, for it places one above the equal relation of friend to friend.⁷

The version correctly attaches “it is” to “not to love,” avoiding the Swenson construction criticised by Piety. Its English is nevertheless open to improvement. “Stand unconditionally certain” is stiff, and “equal” interprets a relation that the Danish names simply as Forholdet mellem Vennen og Vennen. A closer version would be:

Humanly understood, to be unconditionally certain of being loved is not to love, since that is to stand above the relation between friend and friend.

The current readable wording tries to make the implied reciprocity explicit. The proposed close wording makes the Danish steps visible. This is a case where passing the main grammatical test does not end stylistic and interpretive review.

5. The translators should not become caricatures

English-language Kierkegaard reception often organises preferences around translator names. Such orientation is useful: translators do exhibit recurring practices. It becomes unhelpful when “Hong,” “Swenson,” or “Hannay” becomes shorthand for a single virtue or defect.

Piety’s discussion is valuable because it resists that simplification. She can prefer Swenson and Swenson–Lowrie stylistically while documenting places where the Hongs are more accurate. A Notre Dame Philosophical Reviews discussion of Hannay’s Concluding Unscientific Postscript similarly presents the Hongs as comparatively consistent and literal and Hannay as accurate and readable, willing to recast syntax; it also notes that scholarly usefulness includes such matters as source-edition page references.⁸ The review concerns another work and should not be transferred mechanically to Works of Love. It nevertheless illustrates the multiple functions an edition may serve.

The present project should be treated in the same way. Its apparatus may be useful where its prose is not. Its readable text may succeed where a glossary policy fails. Its method may be soundly stated and incompletely implemented. A fair review need not deliver one total verdict.

6. Readability can reveal or conceal

Readable English is sometimes described as a concession to readers. It may instead be required for accuracy. If an English calque causes readers to parse a relation incorrectly, awkwardness does not preserve the source; it places a new obstacle before it.

Readability becomes distorting in at least four situations:

1. Explanation replaces ambiguity. The English resolves an uncertainty Kierkegaard leaves active.
2. Variation removes recurrence. A repeated Danish word becomes a series of elegant English synonyms.
3. Paragraphing changes pressure. A sustained address becomes a list of detachable assertions.
4. Normalisation removes strangeness. Reversal, paradox, or unusual imagery is treated as a problem in Kierkegaard’s prose rather than part of its work.

It becomes revealing when it:

1. makes Danish grammatical relations intelligible in English;
2. removes accidental calques while preserving conceptual difficulty;
3. gives rhetorical emphasis an English means rather than a Danish-shaped shell;
4. lets humour, urgency, or tenderness register as such.

The relevant distinction is therefore not difficult versus easy English. It is source difficulty versus translation difficulty. The former may need to remain; the latter should ordinarily be removed or explained.

7. A multidimensional review grid

The following grid is proposed for each consequential passage:

Dimension	Question	Evidence
Completeness	Is every source clause represented?	Aligned Danish unit
Lexicon	Are distinctions and recurrences preserved or responsibly varied?	Concordance, glossary, historical dictionary
Grammar	Are scope, reference, tense, modality, and syntax correctly construed?	Danish construction and context
Concept	Does the English preserve the relation to the surrounding argument?	Chapter and cross-chapter reading
Rhetoric	Does the English carry repetition, address, irony, image, and movement?	Source pattern and oral reading
English	Is the target text intelligible without introducing a false relation?	Read-aloud and independent reading
Apparatus	Are material losses or alternatives disclosed?	Note, decision record, close layer

This grid does not produce a score. Numerical weighting would conceal the fact that one lexical error can be decisive while several minor rhythmic losses may be tolerable. Its purpose is to prevent “accurate” and “stylish” from ending inquiry.

8. Implications for the close and readable layers

The two-layer method is justified only if it distributes functions without distributing responsibility.

The close layer should:

• be complete;
• preserve important recurrence and order;
• expose lexical relations;
• remain intelligible English;
• identify where Danish has been retained diagnostically.

The readable layer should:

• be capable of continuous reading;
• preserve rhetorical action rather than merely propositions;
• avoid interpretive additions that the close layer cannot support;
• record consequential departures;
• remain revisable when cross-chapter evidence changes.

Neither layer is “the accurate one.” The close text makes some forms of accuracy easier to inspect. The readable text tests whether the interpretation can live as English prose.

9. What an AI-assisted workflow changes

AI assistance heightens rather than resolves the tension. A model can reproduce a glossary instruction with impressive consistency, smooth a long sentence within seconds, and imitate an elevated register. Each capacity creates a corresponding risk: the wrong default becomes systematic; the sentence’s pressure disappears; a generic solemn voice replaces Kierkegaard’s changing voices.

Computational tools can help locate recurrence, compare layers, count omissions, and retrieve candidate passages. They cannot decide which stylistic feature is doing conceptual work. Venuti’s recent argument is useful here: computational analysis and language models may serve a human interpretive act, but translation remains an inscription of interpretation rather than recovery of an invariant message.⁹ The project should not invoke that position as an endorsement. It should accept the responsibility it assigns.

Limitations

This essay relies mainly on examples from the First Series working files and on published discussions of a limited number of English translators. It does not offer a comprehensive stylistic comparison of complete editions. Some source page references in the local working files still require final verification. Judgements about English cadence remain contestable even where the Danish grammar is clear.

Conclusion

Accuracy and style cannot simply be merged into a promise that the translator will achieve both. They name several obligations that may pull against one another. The responsible response is to identify those obligations at the passage level, state the loss attached to a choice, and keep conceptual and rhetorical evidence in view together.

This approach produces no single ranking of Swenson, Hong, Hannay, or the present project. It permits a more modest and useful conclusion: one rendering may be better in one respect and worse in another, and sometimes one respect is decisive. The work of criticism is to show why.

Contribution and AI-use statement

The project editor and series instigator proposed that the essays address criticism of both established translators and AI translation without assuming that the present project deserves a favourable verdict. OpenAI Codex assisted with source location, comparative analysis, and drafting. The human project editor must approve any published version and remains responsible for the evaluative claims. Published translations are paraphrased or quoted only briefly; the argument rests on the Danish source and cited criticism.

Bibliography

Hannay, Alastair. “Translating Kierkegaard.” In The Oxford Handbook of Kierkegaard, edited by John Lippitt and George Pattison, 385–401. Oxford University Press, 2013. https://doi.org/10.1093/oxfordhb/9780199601301.013.0021.

Piety, M. G. “Some Praise for the Hongs as Translators.” Piety on Kierkegaard, 12 May 2026. https://pietyonkierkegaard.com/2026/05/12/some-praise-for-the-hongs-as-translators/.

Taivalkoski-Shilov, Kristiina. “Ethical Issues Regarding Machine(-Assisted) Translation of Literary Texts.” Perspectives 27, no. 5 (2019): 689–703. https://doi.org/10.1080/0907676X.2018.1520907.

Venuti, Lawrence. “The Human Translator versus the Machine: Algorithms of Interpretation.” Critical Inquiry 52, no. 3 (2026). https://doi.org/10.1086/739756.

Notes

Correction history

• 0.1 — 19 August 2026: first complete public-working-paper draft.

1. M. G. Piety, “Some Praise for the Hongs as Translators,” Piety on Kierkegaard, 12 May 2026. ↩
2. Alastair Hannay, “Translating Kierkegaard,” in The Oxford Handbook of Kierkegaard, 385–401, abstract. ↩
3. Works of Love — Methodological Review after Chapters 1–3 (project record). ↩
4. Kristiina Taivalkoski-Shilov, “Ethical Issues Regarding Machine(-Assisted) Translation of Literary Texts,” Perspectives 27, no. 5 (2019): 689–703. ↩
5. Chapter I working translation (project record), close and readable layers for the hidden-lake paragraph. ↩
6. Chapter III.A assembled working text (project record), source unit 35. ↩
7. Chapter IV assembled working text (project record), source unit 2. ↩
8. Notre Dame Philosophical Reviews, review of Hannay’s Concluding Unscientific Postscript. ↩
9. Lawrence Venuti, “The Human Translator versus the Machine: Algorithms of Interpretation,” Critical Inquiry 52, no. 3 (2026). ↩

---

Editorial responsibility. The project editor and series instigator initiated the translation and this essay series, established its methodological and ethical direction, commissioned the research and drafting, and retains authority over revisions and publication. OpenAI Codex assisted with the work described in each essay but is not an author. No Danish specialist or Kierkegaard scholar is represented as a reviewer, participant, or endorser.

Invitation to engagement. Interested readers are welcome to discuss, criticise, or correct these papers publicly elsewhere. Such engagement does not imply participation in or endorsement of the project.