S2. Source editor’s annotation There is nothing for January or February. This [gap] does not result from the loss of a notebook. In fact, the year 1821 follows [December] 27, 1820, in the same notebook. And it begins thus: “1821. March. Quite simply, Joubert noted nothing in these notebooks during those [two] months. After “March,” three dates are indicated, the 18th, 19th and 20th, each followed by three exclamation marks.”
T1. The two relative clauses, comparison and opposed preserves/destroys remain without qualifying the objects of either action. The complete March 21 date belongs to this main unit; its raised source call 2 follows mars before the date’s separate period and points to the source editor’s missing-month account. That annotation supplies no new main entries.
T7. The source-editor account is preserved as reported notebook evidence, not fresh manuscript access. [Gap], [December] and [two] are contextual whole-word counterparts of lacu[…] , d[…]cembre and de[…] , without recovered French endings. The whole lost article after et before 20 has no separate English lexical counterpart in the coordinated enumeration, so the intact number remains unbracketed. Both missing months, the explicit 27 and 1820, the same notebook and all three March day numbers remain. The incomplete December month is qualified and does not repair the prior main December 7 reading. The quotation opens before 1821 and retains the visible final close; no unsupported early closing quotation after March is supplied. The shorter quoted March retains both of its source quotation boundaries, with its following comma placed inside the English closing mark. This is one roman annotation owned by the March 21 date call, not a series of synthetic main entries.
T100. The complete 21, two relative clauses, comparative and opposed verbs remain. Raised 2 follows mars before its separate period; the annotation does not supply new main entries for missing months.
T101. One roman source-editor paragraph reports missing months and notebook continuity. The right edges retain only lacu , d before the next-line cembre , and de before mois . After et no whole article is certified. These four losses remain explicit, with no manuscript access or reconstructed word claimed. The quotation visibly opens before 1821 and has a visible final close after exclamation; any further close after Mars lies in the damaged edge and is not certified. The short quoted mars later has both guillemets. The explicit 27 and 1820 are retained, but the month word is incomplete; this narrative does not repair the preceding main December 7. Reported March 18/19/20 do not become synthetic dated main entries.
S3. Source editor’s annotation Nothing for April or May. — Nothing for August.
T2. The nominal opening, beautiful/bad contrast, equal truth/beauty and alone remain. Which one finds renders qu’on trouve ; and which preserves the explicit et qui . [Letters], [alone], second-quotation [the], [Providence] and only the second [Patin] are contextual whole-word counterparts of lettr[…] , seu[…] , l[…] , Pr[…]vidence and Pati[…] . Their missing French endings are not recovered, and the first Patin is intact. The Latin is freshly translated in its wisdom/folly order; the separate roman French quotation reverses that order. Only the first English quotation’s words correspond to the italic Latin words through mundum ; its quotation marks remain roman. No source closing guillemet or stop is certified after mundum ; the English closing quotation is supplied. The comma after the second [Patin] is also supplied English punctuation, not recovered at the damaged edge. The external bibliography remains unverified. Its small raised round mark is retained diplomatically as ° after 195, without identifying an ordinal letter, degree value or source-note call; volume I remains Roman. The complete June 10 date owns source call 3 and its separate missing-month annotation.
T8. Both Nothing statements and the intervening dash remain. The months are source-editor observations, not invitations to insert empty dated entries into the main text. This complete roman annotation belongs to source call 3 after juin in the June 10 date, before its separate period. The source spelling août retains its qualified compact circumflex.
T102. The full 10 and raised 3 after juin before its period remain. Original color/native right edges certify only lettr , seu , l , Pr before the next-line vidence , and Pati in the second Gui Patin reference. Their interrupted remnants do not establish the missing endings or punctuation. The first Gui Patin name is complete. In the French source, the Latin words through mundum are italic, with the opening guillemet and following parenthesis roman. No French closing guillemet or punctuation is certified at the damaged right edge after mundum ; none is silently supplied. The later French quotation has both its visible opening and closing guillemets and retains the reversed folly/wisdom order. The ending of the external locator has a small raised round mark, copied diplomatically as °, then period and closing parenthesis. Its contour does not justify an identified ordinal letter or a new source call. Tome I is Roman, Paris/Jean Petit/1692 remain exact, and no external edition or quotation source is authenticated.
T3. The opening subordinate That, full identifying parenthesis, present/perfect distinction and course/mechanisms pair remain. Reserved for himself retains the reflexive force; no independent theological explanation is supplied. [Left] contextually interprets liv[…] as livré without restoring its missing letters or participial ending in French; the intact auxiliary has and object it remain outside that lexical bracket. The repeated French subject after mais is naturally shared in English. This unit is undated.
T104. The right edge retains liv ; the following r -like part lacks its full contour, and no ending is certified. The dependent Qu’il , roman identifying parenthesis, present dédaigne and perfect a/s’est constructions remain. The unit is undated.
T4. The nominal opening, repeated birth and parenthetic image remain. The complete source word miroir , including its final r , supports unbracketed mirror. The separate following […] marks uncertified trailing material; no additional word, closing parenthesis or period is recovered from it. The English closing parenthesis and terminal period are supplied punctuation, not certified French readings. The image is not replaced by a doctrinal formula drawn from elsewhere. This is the first of two separate paragraphs under the complete June 21 date.
T105. The complete 21 owns two separately indented paragraphs. Both naît forms show compact circumflex contours in color/native. The complete word miroir retains its final r ; the following right-edge ending has no certified closing parenthesis or punctuation, so the separate gap denotes unknown trailing material without declaring a lost word. After the second naît , only an incomplete left bowl remains; no whole d or de is certified. A whole-word gap remains. No subject is supplied for the second nominal paragraph, and the two paragraphs are not merged.
T5. This remains a separate paragraph, with no supplied subject. Both alternatives, their and the repeated born relation remain; whose contemplation is not further specified. [Of] contextually supplies the wholly lost linking word after the second naît . No whole d or de is certified or restored in French. Both paragraphs share June 21 without being merged.
T6. The unusual English ternity/duity mirror ternité/duité instead of silently substituting familiar theological terms. Both parenthetic explanations and the repeated essential relation remain. [Divine], [as] and [the] contextually render divi[…] , […]omme and […]a without recovering their missing French letters. The English comma after divine nature is supplied; no source comma is recovered at the page join. The opening has no inserted finite verb. One paragraph continues from 902 onto 903 under the complete Sunday, July 1 date; its raised er belongs to the printed day ordinal and is not a source call.
T106. The full weekday, Arabic 1 and raised er remain. At the 902 ending only divi is complete; the next n -like contour is cut before its full right stem. On 903 only omme and a are complete at the two left openings; the missing c/l are not restored. No comma is recovered at the page join. This is one continued paragraph, with both parentheses and repeated essential relation. The unusual ternité/duité are printed literally rather than normalized to a familiar doctrine. Full color 902/903 and bounded original-size/native 434 targets were read; independent contour confirmation is not claimed.
T9. The place cue and printed attribution remain in the body. The short quotation is freshly translated from the French without consulting an English Montaigne; no external locus is authenticated. French scavons is retained literally. The September day preserves only the complete terminal 0 after an unresolved loss; no preceding digit is supplied. The place, quotation and attribution are roman and do not become part of the date.
T107. The final 0 is complete. The preceding narrow upper stroke is compatible with 1 but its lower contour is interrupted in both the original-size color and native 435 views; no complete preceding digit is certified. The earlier navigation reading 10 is not treated as textual authority. The body place cue, full quotation and roman attribution remain separate from the date. Scavons is literal. The opening parenthesis, quotation and complete terminal punctuation remain identifiable despite the worn left margin; no additional loss marker is inserted.