A source, a model, and a test
What is the Handbook?
The Enchiridion, or Handbook, is a short collection of Stoic practical guidance compiled by Arrian from Epictetus’ teaching. It asks how we should judge events, choose our responses, and live with outcomes we cannot command. You do not need to know Stoicism to read this study: each model now sits beside the passage it interprets. [1]
How to read the comparisons
Read the source on the left, then the translation choices and runnable model on the right. On a narrow screen they appear in that order vertically. The alignment is by claim or argument, not by physical code line: declarations, case facts and tests often have no direct sentence in the source. The result underneath says what the checks establish and what remains open.
The source column reproduces Higginson’s English translation, not the original Greek. This translation is identified as public domain in the United States by Project Gutenberg. Longer chapters have expandable context; theological and historical assumptions are retained there even when the model selects only one claim.
The Handbook was compiled by Arrian from Epictetus’ teaching. We use Higginson’s I–LI numbering, cross-checking selected passages against Oldfather and the Greek. [1–4] This revision reviews the whole short work for coverage and strengthens selected executable models. It does not claim a complete formalization of every sentence.
Each example distinguishes the author’s claim, our reconstruction, and the conclusion computed from it. Examples use invented people and situations. Assertions test both warranted results and results that must not follow. Successful execution is evidence about the model and interpreter; it does not establish the truth of Stoicism or a treatment effect.
Compared with the earlier study, this version corrects unknown-name classification, the direction of IX’s conditional, the identification of proof tiers with psychological states, and the scope of the two handles. It also repairs an interpreter defect: aspect-derived conclusions now track the underlying fact for cascading retraction.
A partition is not a dictionary of every possible phrase
The doctrine divides what is ours from what is not. A finite classifier still needs evidence about a particular object. An unrecognized phrase, a typo, or an unclassified instance establishes neither side. Axioma checks the currently registered extent; it does not prove that every possible object has been classified. The named faculties and externals are our encoding of Epictetus’ examples, not a natural-language understanding engine. [1, I]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- The two categories become
UpToUsandNotUpToUs; opinion and body are named samples. - Additional assumptions and scope
- A finite registered domain is a modeling restriction. A missing classification is not evidence for the opposite class.
What running this establishes. The tests detect a classification gap. They do not prove Epictetus’ division for every possible thing.
Derivation, revision, and psychological claims
The following rules reconstruct selected claims; choosing strict or defeasible implication is a modeling decision. A theorem is a strict consequence of supplied premises, not proof of their truth or of someone’s permanent tranquility. Both kinds of consequence can disappear when their last premise is withdrawn. cancel suppresses a conclusion without changing its judgment premise. An empty answer means that this model no longer derives the claim; it does not establish its negation or a clinical result. Independent support matters. [1, V, LI]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
confines_concern → free → tranquilreconstructs part of I; a judgment-to-disturbance bridge reconstructs V.- Additional assumptions and scope
- The intermediate predicates and strict/defeasible rule choices are ours. Calling a conclusion a theorem describes its derivation, not an observed mental state.
What running this establishes. Removing a premise withdraws dependent conclusions. It does not make a person’s distress disappear.
Explicit indifference is different from no valuation
Epictetus locates good and evil in the use of choice. Here “indifferent” means an attributed judgment about moral value, not that health or property has no practical use, and not that every modern reader must accept that judgment. An unvalued fact is a different case. This encodes only the axiological part of XXXI; its providence and ritual claims remain outside the model. [1, XXXI–XXXII]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- Restricting good and evil to what is ours motivates the attributed valuation
value_indifferentfor wealth. - Additional assumptions and scope
- Wealth is our example. The code does not adopt the chapters’ theological or divinatory premises; their full context remains available.
What running this establishes. An explicitly indifferent valuation differs from the absence of a valuation. Neither is a universal judgment that the object has no practical importance.
Descriptions must remain attached to an existing fact
A qua handle labels a fact; it does not identify who holds the description or when. This example deliberately uses descriptive predicates rather than naming an event “at peace.” Two tags may coexist. Both generic and relation-specific rules depend on the underlying event, so cascading withdrawal removes their unsupported consequences and tags. Plain forget is non-cascading. cancel retains metadata for uncancel; it is not a substitute for premise revision. [1, XI]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- “Lost” and “restored” motivate two descriptions of
departed. The code’s labelgiven_backnames the second description. - Additional assumptions and scope
- The estate is invented; the divine giver is not modeled. Withdrawal of the fact is an interpreter lifecycle test, not an event in Epictetus’ narrative.
What running this establishes. Descriptions can coexist, but their derived consequences require the underlying event fact.
A person’s appraisal, an event, and a stage
The loss/return example belongs to XI. We add explicit person and stage arguments and a revisable bridge from appraisal to a prediction. These are modeling additions, not quotations or established psychological laws. An imagined loss alone must not satisfy a rule about an actual loss. A later appraisal must not overwrite another person or an earlier stage. An observed distress report remains even when the model predicts bearability. XI also appeals to a giver and return; that theological premise is not proved by this reconstruction. [1, XI]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- XI’s redescription and XLIII’s handles motivate
takes_handle(person,item,stage,handle). - Additional assumptions and scope
- Combining these chapters for an estate loss is an extension. The predicted response requires an explicitly adopted hypothetical bridge; observations remain separate.
What running this establishes. Changing an appraisal changes a conditional prediction. It does not erase a recorded experience or establish a clinical effect.
The brother’s injustice does not disappear
XLIII uses a brother’s injustice and the alternative focus on kinship and shared upbringing. The model retains the wrong act while changing the chosen basis for a response. “Suggested response” is our operational interpretation of a bearable handle; it is neither forgiveness already achieved nor denial that injustice occurred. [1, XLIII]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- The brother’s injustice, kinship, shared upbringing and chosen focus become separate relations.
- Additional assumptions and scope
- A response follows only when the learner adopts the kinship bridge and directs attention accordingly. The norm is not derived from kinship alone.
What running this establishes. The suggested response changes while the injustice remains a fact. No test establishes an obligation to tolerate abuse.
A valid comparison and an invalid inference
Greater wealth supports a comparison of possessions, not personal moral superiority. A missing derivation alone would only show that our program lacks a rule. The finite countermodel additionally demonstrates that the proposed implication is invalid without another premise. The Socrates example makes the same discipline visible: disagreement alone does not refute intrinsic badness. Epictetus supplies a further conditional linking what is terrible to Socrates’ judgment; omitting that bridge changes his argument. This test exposes the need for the bridge, not a refutation of Epictetus. [1, V, XLIV]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- The wealth comparison is restricted to possessions. V’s Socrates argument is checked as a separate Boolean entailment.
- Additional assumptions and scope
- For the Socrates argument, the bridge from objective terribleness to Socrates’ judgment is an additional premise, not a truth of propositional logic.
What running this establishes. A countermodel exposes an invalid inference without claiming either the moral superiority or terribleness proposition is false.
A situation, a faculty, and a stated norm
X links kinds of trial to faculties, rather than merely listing the faculties. XXX and XXXII connect relationships and duties. Our simplified friend example makes the normative bridge explicit: descriptive facts alone do not establish an obligation, and an obligation does not establish compliance. This is not a complete account of conflicting duties, divination, or the historically specific family obligations in the text. [1, X, XXX, XXXII]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- Trials select faculties; relationships and an adopted norm select a recommendation to help.
- Additional assumptions and scope
- Our friend-in-need case adapts XXX/XXXII. A norm is represented explicitly; recommended conduct and performed conduct are different predicates.
What running this establishes. The recommended faculty and duty follow under the given rules. Compliance and good consequences do not follow automatically.
Observation, assent, and deliberate action
XLV distinguishes observed haste from judging motives or moral quality. Not deriving badness does not prove goodness. IV preserves an undertaking alongside the aim of acting well; read with VIII, acceptance need not mean abandoning action. The final predicate below names their conjunction. It demonstrates a possible compatible state, not a causal theorem that acceptance makes action successful. [1, IV, VIII, XLV]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- Observed conduct is separated from evaluation. A second model combines acceptance of an outcome with retaining an aim.
- Additional assumptions and scope
- The bath case is stipulated. Compatibility is defined by co-presence of the two attitudes, not by a tested psychological mechanism.
What running this establishes. Observation alone yields no moral verdict. Accepting an outcome neither prevents action nor proves successful completion.
“Only if” must not become “if”
“The will is not impeded unless it chooses” restricts impediment: in this propositional abstraction, impeded implies consents. It does not by itself license the converse, consents implies impeded. We enumerate all Boolean valuations of those two predicates. The bodily impediment in the chapter is a different predicate; the formalization must not conclude that physical illness is unreal. [1, IX]
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- Let
imean impediment to the will andcmean consent: the proposed necessary condition isi → c. - Additional assumptions and scope
- This Boolean abstraction leaves bodily illness and lameness outside its two variables. Reading “unless” as a necessary condition does not license the converse.
What running this establishes. The enumeration finds consent without impediment consistent with the stated condition; impediment without consent is excluded by that premise.
It seemed right to him
Epictetus asks the reader to distinguish an offender’s impression from the reader’s own judgment. We can represent that distinction without endorsing the offender’s belief. The chapter then makes stronger claims about harm and a gentle response; those require their own premises. This example tests the attribution boundary first.
AXIOMA MODEL · OUR RECONSTRUCTION
- Translation choices
- The offender’s judgment, the observer’s attribution of that judgment, and the stipulated wrongness of an act are represented separately.
- Additional assumptions and scope
- The case and the bridge for identifying mistaken judgment are ours. Nested contents are explicit stored reports, not an automatic calculus of iterated knowledge.
What running this establishes. A report of someone’s belief does not endorse its content. Removing the contrary case fact removes the basis for identifying the mistake; no psychological harm or calm is inferred.
The whole Handbook, honestly mapped
The table records the earlier corpus’ coverage and the current revision, including remaining modeling tasks. “Partial” is not “proved”; an inventory or a belief string is not a reconstruction of an argument. The revised examples above add IV/VIII/XLV, X/XXX/XXXII, the actual XLIII case, XLIV’s countermodel, and the corrected IX abstraction. XI gains person, stage, event and observation boundaries. Most remaining passages still need separate models.
Higginson groups the final material into L and LI. Oldfather separates it into 50–53: Higginson L corresponds to Oldfather 50–51, and LI to 52–53. The numbering is edition-dependent. In XVI, the MIT Carter transcription differs on outward sympathy from Higginson, Oldfather and the Greek; it is not our sole textual authority.
Open the 51-section audit
| Section | Subject | Earlier coverage | Revision and remaining task |
|---|---|---|---|
| I | Control, ownership and examination of impressions | Partial | Corrected: unknown phrases remain unknown and finite partition gaps are tested. Still partial: agent-relative control and influence versus guaranteed outcomes. |
| II | Desire and aversion; temporary restraint during training | Stub | Existing file enumerates faculties only. Represent agent, desired object, attainment, training stage and reservation; risk is not identical to inevitable immediate failure. |
| III | Mortality and fragility of what is loved | Absent | Represent object nature, anticipated loss and the proposed appraisal bridge; do not equate remembering mortality with an observed absence of grief. |
| IV | Bathing and the second aim of preserving one's moral purpose | Absent | Added a compatibility model with VIII: accepting an outcome can coexist with deliberate action. No causal or guaranteed-serenity conclusion is established. |
| V | Judgments, disturbance and the Socrates argument | Partial | Corrected proof-tier and revision claims; strict and defeasible support are tested. The Socrates countermodel shows why the additional epistemic bridge is needed. The psychological causal claim remains an adopted premise. |
| VI | The horse's excellence and one's own use of impressions | Absent | Prevent transfer of an owned object's excellence to its owner; distinguish possession from attribution. |
| VII | The ship and the summons | Absent | Represent the analogy and priority of a summons; do not turn metaphor into literal universal transport facts. |
| VIII | Wish events as they occur | Attribution only | Added a compatibility model with IV; acceptance does not infer passivity or successful outcomes. The chapter's promised tranquility is not empirically established. |
| IX | Bodily impediment versus impediment to moral purpose | Partial | Corrected the necessary condition and tested the converse countermodel. Bodily and moral impediment must remain distinct in further models. |
| X | Choose the faculty appropriate to the trial | Inventory only | Added trial-to-faculty mapping. The selected faculty is a recommendation, not evidence that it was exercised. |
| XI | Restored rather than lost | Partial | Corrected event support and added person, stage, hypothetical bridge and independent observations. The divine-giver premise remains a separate interpretive commitment. |
| XII | The cost of equanimity and modest training cases | Absent | Represent tradeoffs and practice progression; preserve the historical servant setting as the author's setting, not a recommendation by the modeler. |
| XIII | Progress versus concern for reputation | Absent | Represent competing aims and attention; the claimed incompatibility needs an explicit bridge rather than a syntactic fact. |
| XIV | Others' lives, desire and dependency | Absent | Distinguish one's choice from controlling another person or a lifespan; model dependence on another's allocation of valued externals. |
| XV | The banquet analogy | Absent | Represent waiting, moderate taking and foregoing as distinct policies; do not treat divine status as a measured consequence. |
| XVI | Outward sympathy and inward appraisal | Absent | Separate expressions of sympathy from the internal judgment; cross-check translations, since the MIT transcription conflicts with Higginson and Oldfather here. |
| XVII | Assigned role and quality of performance | Partial | The companion now connects a particular assigned role, an adopted norm and actual performance. Assignment alone does not establish compliance or justify a social order. |
| XVIII | Omens and the use of events | Absent | Separate an omen report, predicted external outcome and proposed moral use; recording an omen must not entail its truth. |
| XIX | Contests, freedom and envy | Absent | Distinguish a conditional strategy from guaranteed success in external contests; restrict the meaning of unconquerable. |
| XX | Insult, judgment and delaying assent | Absent | Represent sequence and delay, plus bodily assault separately from moral affront; no inference that physical injury is unreal. |
| XXI | Daily reflection on death and exile | Absent | Represent a prescribed practice and a claimed effect independently; repeated advice is not a causal proof. |
| XXII | Ridicule and persistence in philosophy | Absent | Represent present ridicule and the forecast of later admiration as distinct claims with temporal scope. |
| XXIII | Pleasing others versus one's philosophical aim | Absent | Specify what abandoning an aim means; do not infer moral failure from every socially considerate act. |
| XXIV | Civic usefulness without sacrificing integrity | Duty label only | The current label is not a derivation. Model means, integrity conditions and civic benefit; it limits a fatalistic reading of I and VIII. |
| XXV | The price of invitations and external goods | Absent | A useful programming/KR case: alternatives, costs, retained goods and chosen tradeoffs, with no invented utility numbers attributed to Epictetus. |
| XXVI | Consistency between one's own and another's loss | Absent | Compare two appraisals of analogous events; invariance is a normative demand, not an assumption that persons are identical. |
| XXVII | The target analogy and the status of evil | Absent | Analyze the teleological bridge explicitly; an analogy alone does not prove that evil has no nature. Keep translator's explanatory footnote separate. |
| XXVIII | Entrusting one's mind to an abuser | Incidental prohibition | The synthesis records a prohibition but does not formalize the bodily/moral analogy or its premises. |
| XXIX | Consider requirements and consequences before undertaking | Absent | Represent prerequisites, capacity, costs, uncertainty and commitment; undertaking the training does not entail victory. |
| XXX | Duties arising from relations | Duty inventory | Added a duty from a relationship plus an explicit norm; no compliance is inferred. Conflicts among duties and the full family context remain unmodeled. |
| XXXI | Piety, providence and relocating good and evil | Partial | Corrected explicit indifference versus no valuation. Providence, blame and ritual conditions remain outside the executable fragment. |
| XXXII | Divination, uncertainty and prior duty | Duty label only | The shared friend-duty model adds an explicit normative bridge. The standalone file remains a duty inventory; the divination and uncertainty argument still needs its own model. |
| XXXIII | Many rules of conduct with qualifications | Absent | Requires clause-level work, not one Boolean: silence, speech, oaths, company, provision, sexuality, spectacles, visits and self-display have different conditions. |
| XXXIV | Delay before pleasure; compare later responses | Absent | Represent alternative trajectories and later appraisal. A hypothetical consequence is not an observed fact. |
| XXXV | Doing what is right despite public censure | Absent | Keep the correctness condition explicit; believing oneself right is not interchangeable with being right. |
| XXXVI | Disjunction/conjunction and courtesy at dinner | Absent | Separate the propositional analogy from the social principle; suitability for appetite is not suitability for the shared occasion. |
| XXXVII | Roles beyond one's capacity | Absent | Represent capability and the opportunity cost of abandoning a feasible role; no duty-compliance inference from the role name alone. |
| XXXVIII | Guard the ruling faculty in each action | Absent | Differentiate bodily and moral precautions, rather than one undifferentiated harm predicate. |
| XXXIX | Measure of possessions and escalation | Absent | Model need, fit and excess; the slippery-slope claim needs a stated causal or normative status. |
| XL | Gendered honor and social incentives | Absent | Preserve attribution and historical context; do not silently universalize the gendered standard as an axiom endorsed by Axioma. |
| XLI | Bodily maintenance and attention to reason | Absent | Distinguish necessities from excess and resource allocation; it is not a claim that every bodily activity is bad. |
| XLII | Another person’s impression and mistaken judgment | Absent | Added an explicit two-level belief attribution and a conditional mistake check. Belief, endorsement, case truth and psychological effect remain separate. A general nested-knowledge calculus is not claimed. |
| XLIII | The brother's injustice and the bearable handle | Partial adaptation | Added the actual brother case, shared upbringing and selected focus. Injustice remains represented when the response changes; no psychological outcome is guaranteed. |
| XLIV | Invalid inferences from wealth or eloquence to personal superiority | Absent | Added the appropriate possessions comparison and a finite countermodel to moral-superiority inference. This is a small entailment check, not a general moral theory. |
| XLV | Observation versus judging motives | Mismatched demonstration | Replaced the mismatched cancellation example with observation versus moral verdict; neither badness nor goodness follows from haste alone. |
| XLVI | Practice rather than proclaiming principles | Absent | Represent the difference between reported knowledge and conduct; saying the doctrine does not prove practicing it. |
| XLVII | Private training rather than display | Absent | Represent intention, audience and practice; a public report of austerity must not entail virtue. |
| XLVIII | Signs and stages of progress | Mismatched simplification | The companion now tests two recorded signs and the failure to infer perfect wisdom. The remaining behavioral signs and training dynamics still need models. |
| XLIX | Interpreting Chrysippus versus following nature | Absent | Separate text comprehension, attribution and behavior. Interpretation can be correct while compliance is absent. |
| L | Adopted rules, urgency and present effort | Absent | Represent ongoing commitment and missed opportunities; avoid treating one lapse as a proven irreversible loss of all progress. |
| LI | Practice, justification, logic and closing maxims | Prose only | The companion now separates explaining a norm from following it. The closing maxims and their fate/moral-harm premises remain unformalized. |
Where the language still needs pressure
Concept inheritance, finite partitions, explicit attribution, strict and defeasible rules, fact revision, and ordinary enumeration are sufficient for these models. Person and stage are ordinary relation arguments; they do not provide a temporal or psychological semantics automatically. Global qua tags do not themselves represent nested belief, a source’s reliability, or a person changing their mind.
The harder remaining tests include a compositional semantics for nested knowledge beyond XLII’s explicit stored attributions, conflicting duties, XXIX’s uncertain plans, XXXIII’s qualified prescriptions, and XXVII/XXXI’s teleological and theological premises. Historical social and gender assumptions need explicit attribution. The Boolean countermodel proves only a small stated entailment claim; no general theorem about the entire Handbook or the completeness of Axioma’s inference follows.
This provides a reusable method for the later ACT study: preserve source context; label definitions, assumptions and empirical claims; test consequences and counterexamples; revise premises; keep observations separate from predictions. ACT itself is not revised by this study.
References
- Epictetus. The Enchiridion. Compiled by Arrian. Translated by Thomas Wentworth Higginson. Liberal Arts Press, 2nd ed., 1955. Project Gutenberg 45109. Primary text for the complete I–LI coverage review.
- Epictetus. The Discourses as Reported by Arrian, the Manual, and Fragments, vol. II. Translated by W. A. Oldfather. Loeb Classical Library, 1928. Manual. Selected passage and numbering checks.
- Epictetus. The Enchiridion. Translated by Elizabeth Carter. MIT Internet Classics Archive transcription. Consulted comparatively; see the XVI transcription caveat above.
- Epictetus. Encheiridion. Edited by Albert Watanabe. Dickinson College Commentaries, 2020. ISBN 978-1-947822-13-9. Greek text and commentary, chapter 16. Check of outward sympathy and inward appraisal.
What follows — and from what?
The useful result is an inspectable argument, including its assumptions and its limits.