Contribution
The paper shows that Parfit’s premises do not force his conclusion. If persons are aggregates of stages, the formal mismatch between identity and the psychological relation is a mismatch between relations on different things, and the relation that matters among stages can be identified with the relation of belonging to one person. The cost is paid in counting: persons may overlap, and we must count them by a relation weaker than identity. It also gives the first worked treatment of degrees of personal identity by supervaluation.
Key insight
Do not compare identity with the R-relation; compare the I-relation with it. “Stage S is I-related to stage S′” means “some one continuant person has both as stages”, and that relation can be intransitive and graded if persons can overlap and personhood can come in degrees. Define persons as the maximal R-interrelated aggregates, and the two relations cannot come apart.
Assumptions
- Persons have stages. Continuant persons are aggregates of person-stages (n. 4 prefers mereological sums). Postscript B (1983) defends this.
- R is fixed in advance. “Pretend that the open questions have been settled” about which version of continuity and connectedness matters (p. 20). The argument works for any version.
- R is made symmetrical and reflexive by stipulation, merging its forward and backward forms (pp. 23–24).
- Counting need not be by identity. It must be by an equivalence and at least partial indiscernibility relation (p. 28).
- Bizarre cases are set aside. Stages with no “personal time” ordering are “so indeterminate” that they “cannot serve as counterexamples” (n. 5).
Key results
- The two answers (pp. 17–19). “What matters in survival is mental continuity and connectedness” (p. 17), versus “what matters in survival is survival”, identity “between the I who exists now and the surviving I” (p. 18). “If the two answers disagreed … we would have to prefer the platitude of common sense” (p. 18).
- Parfit’s argument (p. 19). (a) identity is one-one and not a matter of degree; (b) mental continuity and connectedness need not be; so (c) what matters is not identity. Lewis grants (a) and (b) (p. 19) and refuses a verbal redefinition of “identity” (p. 20).
- Different relata (p. 20). “Identity among stages has nothing to do with it, since stages are momentary. Even if you survive, your present stage is not identical to any future stage.”
- The I-relation is the R-relation (pp. 21–22). A person is a maximal I-interrelated aggregate; if I = R, “something is a continuant person if and only if it is a maximal R-interrelated aggregate of person-stages” (p. 22).
- Why I need not inherit identity’s form (p. 23). The argument that it must “wrongly takes it for granted that every person-stage is a stage of one and only one continuant person”. The definition must use the indefinite article: some continuant person has both stages.
- Fission and fusion (pp. 24–25, Figure 1.1). The I-relation “will fail to be transitive if and only if there is partial overlap among continuant persons”. Fission is two continuants C1 and C2 sharing pre-fission stages; fusion, sharing post-fusion stages.
- Overpopulation and tensed identity (pp. 26–29).
- With overlap there are more continuants than stages at a time, and we count stages (p. 26).
- “Continuants C1 and C2 are identical-at-time-t if and only if they both exist at t and their stages at t are identical” (p. 26). “Tensed identity is not a kind of identity” (p. 27).
- The roads: the Chester A. Arthur Parkway and Route 137 merge, and one crosses both by crossing one (p. 27).
- Counting fusion products as two but fission candidates as one uses identity-at-all-times-up-to-t (p. 28).
- “Ned”, the name of a person who will fission, is ambiguous, harmlessly until the fission (pp. 28–29).
- Longevity (pp. 29–31). If any two stages of Methuselah within 137 years are R-related and no others are, the persons are exactly the 137-year segments; continuum many share his 300th birthday, and tensed identity counts them as one.
- Degree (pp. 32–36).
- Personal identity can come in degrees because personhood can (p. 32), as in fusion with someone very different.
- Naive degrees fail: an aggregate R-interrelated to degree 0.9 but inside one R-interrelated to 0.88 “passes … to degree 0.9” but “flunks the maximality test to degree 0.88. Therefore it is a person only to degree 0.02!” (p. 34). On Methuselah, no segment is a person to any significant degree.
- Supervaluation: each cut-off for R is a delineation; at each, R and I coincide; so the measure of delineations on which two stages are I-related equals that on which they are R-related (pp. 35–36), after van Fraassen (n. 9).
- Perry (pp. 36–39). Perry’s continuant persons are “lifetimes”, the aggregates of stages R-related to a given stage at a given time; Lewis’s are “branches”. Perry identifies only time-indexed subrelations of R and I, and must exclude persons “not determinable” at a time. Counting from no time, “How many persons were involved in an episode of fission long ago? I say: two. Perry says: three. Or else he says: none now determinable” (p. 39).
- Postscript A (1983, pp. 73–75), read but not the paper’s text. A shared pre-fission stage cannot desire survival for one continuant without desiring it for the other, so the most commonsensical desire available is plural, “let at least one of us survive” (weak) or “let all of us survive” (strong). The weak desire is equivalent to wanting I-relatedness to future stages. Leiber’s survival insurance, a mind recorded and played back into a new body, satisfies the weak desire.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | The formal mismatch between identity and R does not show that what matters is not identity, since they relate different things | strong | p. 20; the stage/continuant distinction |
| C2 | The I-relation and the R-relation are the same relation in all possible cases | conditional: holds by definition once persons are maximal R-interrelated aggregates | pp. 22–23; the definition is “at least, I cannot think of any that clearly is not” (p. 22) |
| C3 | Fission and fusion are cases of partial overlap of persons | moderate, given C2 | pp. 24–25 |
| C4 | Counting persons by tensed identity is acceptable | moderate | the roads analogy (p. 27); indiscernibility at a time (p. 28); the author calls it “somewhat counterintuitive” (p. 38) |
| C5 | Degrees of I-relatedness equal degrees of R-relatedness under supervaluation | strong as a construction | pp. 35–36 |
| C6 | Lewis’s branches are better persons than Perry’s lifetimes | moderate | the overpopulation objection when counting from no time (p. 39) |
Concepts
- person-stage — a “more or less momentary” time-slice of a continuant person (p. 20); a physical object that does many things a person does, but briefly (Postscript B).
- R-relation — the relation of mental continuity and connectedness among stages that matters in survival, made symmetrical and reflexive (pp. 20, 24).
- I-relation — holding between two stages iff some one continuant person has both as stages (pp. 21, 23).
- tensed identity — identity-at-t: having identical stages at t; a relation among continuants that is weaker than identity when they share stages (pp. 26–27).
- delineation — a cut-off on the degree of R-relatedness (p. 35).
- branch / lifetime — Lewis’s and Perry’s continuant persons respectively (p. 37).
Connections
- Parfit’s “Personal Identity” (filed beside this reading) is note 2 and the argument of pp. 19–20. Lewis reports Parfit’s Methuselah-like immortals (p. 30) and his fusion case (p. 32) from it. Lewis’s C1 accepts that paper’s C1 and C5–C6 and denies its C2.
- Locke (filed beside it). Lewis’s note 1 replaces memory with “quasi-memory”, Shoemaker’s term that Parfit uses, which answers the circularity charge against Locke.
- Goldstein and Lederman (LIT-206, NOTE-138) and Birch (LIT-111, NOTE-183). See the LIT entry: stage-sharing models instances of one model, and Leiber’s recording is the case Birch would deny.
- Meincke (LIT-142). Both count individuals that overlap. Lewis keeps identity sharp and lets persons share parts; Meincke gives up substance counting for a process view and accepts counts “between one and two”.
Bearing on the record
No THEORY document cites it. If the record files a THEORY on whether copies or parallel instances of an AI system are one subject or many, this paper is the strongest case that the answer can be “many, counted as one until they diverge” without giving up identity. It carries no instruction for machine-learning practice.
Note of 2026-10-16: THEORY-476 now cites LIT-446 as a source.
Limitations
What is shown and what is asserted:
- Shown: that Parfit’s (a) and (b) do not entail (c) once stages and continuants are distinguished (C1); and that supervaluation gives equal degrees (C5).
- Stipulated: that persons are the maximal R-interrelated aggregates (C2). Lewis offers it as the claim that makes the reconciliation work, and its only support is that he cannot think of a counterexample.
- Conceded as costly: counting by tensed identity (C4).
- Not addressed: whether common sense really wants identity of a continuant, or only something a stage can want. Parfit pressed this in the same volume, and Lewis’s answer, the plural desire of Postscript A, came in 1983.
- No first-person argument. The paper does not consider Williams’s case that expectation follows the body; it takes the psychological answer as given.
Open questions
- What does the shared pre-fission stage rationally fear or hope? Postscript A’s answer, a weak plural desire, says what it can want; the paper does not say how it should feel about the branch that will soon die.
- Is the I = R identification stable under a different choice of R? The paper says any version will do (p. 20), but its maximality condition makes which aggregates are persons sensitive to that choice.
Corrections
None to the brief’s citation. One addition: the pages (17–40) the brief did not give.