The frontier of human longevity, on a Lexis plane

TidyTuesday 2023-05-30 · The 100 oldest verified women and men

Published

June 9, 2026

Session 1 · co-developed

This page came out of a live, turn-by-turn conversation between Jon Minton and Claude (Fable 5) — questions, plots and interpretation built together in real time. The Session 2 pages were instead produced by Claude working autonomously.

The 200 oldest verified people ever recorded — 100 women, 100 men, from the Gerontology Research Group’s validated lists. Demographers will want this on a Lexis diagram: calendar time across, age up, every life a 45° line. On that plane, this dataset is not a list of names but a frontier — the upper-right edge of documented human survival.

Code

library(tidyverse)

cent <- read_csv("data/centenarians.csv", show_col_types = FALSE) |>
  mutate(
    as_of = as.Date("2023-05-25"),          # dataset compilation date
    end_date = coalesce(death_date, as_of),
    birth_dec = decimal_date(birth_date),
    end_dec = decimal_date(end_date),
    end_age = end_dec - birth_dec,
    gender = str_to_title(gender),
    alive = still_alive == "alive"
  )

sex_pal <- c(Female = "#8e3b6b", Male = "#2a6f8e")
theme_set(theme_minimal(base_size = 13))

Two hundred lifelines

Each line traces one person from their 105th birthday to their death (or to May 2023 if still alive, marked with an open ring). The 45° slope is not a styling choice — on a Lexis plane, living a year moves you one year right and one year up.

Code

label_people <- cent |>
  filter(name %in% c("Jeanne Calment", "Kane Tanaka", "Sarah Knauss",
                     "Jiroemon Kimura", "Maria Branyas Morera"))

ggplot(cent) +
  # faint cohort guides every decade
  geom_abline(
    intercept = seq(-1810, -1960, by = -10), slope = 1,
    colour = "grey93"
  ) +
  geom_segment(
    aes(x = birth_dec + 105, y = 105, xend = end_dec, yend = end_age,
        colour = gender),
    linewidth = 0.45, alpha = 0.65
  ) +
  geom_point(
    data = filter(cent, !alive),
    aes(end_dec, end_age, colour = gender), size = 0.9
  ) +
  geom_point(
    data = filter(cent, alive),
    aes(end_dec, end_age, colour = gender),
    shape = 21, fill = "white", size = 2, stroke = 0.8
  ) +
  geom_text(
    data = label_people,
    aes(end_dec, end_age, label = name, colour = gender),
    hjust = -0.06, vjust = -0.55, size = 3.1, fontface = "italic",
    show.legend = FALSE
  ) +
  scale_colour_manual(values = sex_pal, name = NULL) +
  scale_x_continuous(breaks = seq(1960, 2020, 10)) +
  scale_y_continuous(breaks = seq(105, 122.5, 2.5)) +
  coord_fixed(xlim = c(1960, 2031), ylim = c(104.5, 123.5), expand = FALSE,
              clip = "off") +
  labs(
    title = "Every verified extreme life, 1968–2023",
    subtitle = "Deaths (dots) pile up along a remarkably flat ceiling — except one woman who sailed three years beyond anyone else",
    x = "Calendar year", y = "Age"
  ) +
  theme(legend.position = c(0.05, 0.88), panel.grid.minor = element_blank())

Lexis diagram of the 100 oldest verified women and men. Lifelines drawn from age 105; open rings are people still alive in May 2023. Axes share one scale, so every life runs at exactly 45°.

Three things the Lexis plane shows that a ranked table hides:

The thickening. Lifelines crowd rightward: most of the top-200 died after
1. That’s partly real (bigger, longer-lived birth cohorts) and partly verification — modern deaths are far easier to document.
The flat ceiling. Deaths accumulate against an edge around 117–119 for women, ~114–116 for men, and that edge has barely risen in thirty years.
The Calment anomaly. Jeanne Calment’s lifeline exits the swarm and keeps going — 122.45 years, dying in 1997. No one verified since has come within three years of her.

The record that refuses to fall

Code

frontier <- cent |>
  filter(!alive) |>
  arrange(end_dec) |>
  mutate(record = cummax(end_age), .by = gender)

ggplot(frontier, aes(end_dec, record, colour = gender)) +
  geom_step(linewidth = 1) +
  annotate("text", x = 1999, y = 121.7, hjust = 0, size = 3.2,
           colour = sex_pal["Female"], fontface = "italic",
           label = "Calment, 122.45 — unbroken since 1997") +
  annotate("text", x = 2014, y = 115.6, hjust = 0, size = 3.2,
           colour = sex_pal["Male"], fontface = "italic",
           label = "Kimura, 116.13 (2013)") +
  scale_colour_manual(values = sex_pal, name = NULL) +
  labs(
    title = "The longevity record has stagnated — especially for women",
    subtitle = "Maximum verified age at death recorded to date (within this top-100 sample)",
    x = NULL, y = "Record age at death"
  ) +
  theme(legend.position = "bottom")

Running record: the highest verified age at death achieved up to each point in time, by sex.

This stagnation is one of the livelier fights in contemporary demography. One camp reads the flat ceiling as evidence of a hard limit to human lifespan (Dong, Milholland & Vijg’s much-disputed 2016 Nature paper put it near 115). Another finds no detectable plateau in the hazard — Barbi et al. (2018) estimated mortality flattening around 50% per year after 105, which makes records a waiting game governed by how many people reach 105. And a third, more corrosive line of argument (Newman) holds that extreme-age records are heavily contaminated by clerical error and fraud — on that view, the cleaner the data get, the harder records become to break, and Calment’s outlier status is itself suspicious (a claim her validators vigorously contest).

This dataset can’t settle that fight — it’s the top of the order statistics, not the hazard. But it visualises the stakes: a 3-year, 26-year-old gap between one lifeline and every other human being ever verified.

Sample caveat

These are the top 100 per sex, so the early decades of the “running record” lines are right-censored: anyone who held the record in, say, 1975 but ranks below today’s top-100 isn’t in the data. The frontier shape is trustworthy from roughly the late 1980s onward.

--- title: "The frontier of human longevity, on a Lexis plane" subtitle: "TidyTuesday 2023-05-30 · The 100 oldest verified women and men" date: 2026-06-09 --- ::: {.callout-tip icon=false} ## Session 1 · co-developed This page came out of a live, turn-by-turn conversation between Jon Minton and Claude (Fable 5) — questions, plots and interpretation built together in real time. The [Session 2 pages](index.qmd) were instead produced by Claude working autonomously. ::: The 200 oldest *verified* people ever recorded — 100 women, 100 men, from the Gerontology Research Group's validated lists. Demographers will want this on a **Lexis diagram**: calendar time across, age up, every life a 45° line. On that plane, this dataset is not a list of names but a *frontier* — the upper-right edge of documented human survival. ```{r setup} library(tidyverse) cent <- read_csv("data/centenarians.csv", show_col_types = FALSE) |> mutate( as_of = as.Date("2023-05-25"), # dataset compilation date end_date = coalesce(death_date, as_of), birth_dec = decimal_date(birth_date), end_dec = decimal_date(end_date), end_age = end_dec - birth_dec, gender = str_to_title(gender), alive = still_alive == "alive" ) sex_pal <- c(Female = "#8e3b6b", Male = "#2a6f8e") theme_set(theme_minimal(base_size = 13)) ``` ## Two hundred lifelines Each line traces one person from their 105th birthday to their death (or to May 2023 if still alive, marked with an open ring). The 45° slope is not a styling choice — on a Lexis plane, living a year moves you one year right *and* one year up. ```{r lexis} #| fig-height: 4.9 #| fig-width: 11.5 #| fig-cap: "Lexis diagram of the 100 oldest verified women and men. Lifelines drawn from age 105; open rings are people still alive in May 2023. Axes share one scale, so every life runs at exactly 45°." label_people <- cent |> filter(name %in% c("Jeanne Calment", "Kane Tanaka", "Sarah Knauss", "Jiroemon Kimura", "Maria Branyas Morera")) ggplot(cent) + # faint cohort guides every decade geom_abline( intercept = seq(-1810, -1960, by = -10), slope = 1, colour = "grey93" ) + geom_segment( aes(x = birth_dec + 105, y = 105, xend = end_dec, yend = end_age, colour = gender), linewidth = 0.45, alpha = 0.65 ) + geom_point( data = filter(cent, !alive), aes(end_dec, end_age, colour = gender), size = 0.9 ) + geom_point( data = filter(cent, alive), aes(end_dec, end_age, colour = gender), shape = 21, fill = "white", size = 2, stroke = 0.8 ) + geom_text( data = label_people, aes(end_dec, end_age, label = name, colour = gender), hjust = -0.06, vjust = -0.55, size = 3.1, fontface = "italic", show.legend = FALSE ) + scale_colour_manual(values = sex_pal, name = NULL) + scale_x_continuous(breaks = seq(1960, 2020, 10)) + scale_y_continuous(breaks = seq(105, 122.5, 2.5)) + coord_fixed(xlim = c(1960, 2031), ylim = c(104.5, 123.5), expand = FALSE, clip = "off") + labs( title = "Every verified extreme life, 1968–2023", subtitle = "Deaths (dots) pile up along a remarkably flat ceiling — except one woman who sailed three years beyond anyone else", x = "Calendar year", y = "Age" ) + theme(legend.position = c(0.05, 0.88), panel.grid.minor = element_blank()) ``` Three things the Lexis plane shows that a ranked table hides: 1. **The thickening.** Lifelines crowd rightward: most of the top-200 died after 2000. That's partly real (bigger, longer-lived birth cohorts) and partly verification — modern deaths are far easier to document. 2. **The flat ceiling.** Deaths accumulate against an edge around 117–119 for women, ~114–116 for men, and that edge has barely risen in thirty years. 3. **The Calment anomaly.** Jeanne Calment's lifeline exits the swarm and keeps going — 122.45 years, dying in 1997. No one verified since has come within three years of her. ## The record that refuses to fall ```{r frontier} #| fig-height: 5.5 #| fig-cap: "Running record: the highest verified age at death achieved up to each point in time, by sex." frontier <- cent |> filter(!alive) |> arrange(end_dec) |> mutate(record = cummax(end_age), .by = gender) ggplot(frontier, aes(end_dec, record, colour = gender)) + geom_step(linewidth = 1) + annotate("text", x = 1999, y = 121.7, hjust = 0, size = 3.2, colour = sex_pal["Female"], fontface = "italic", label = "Calment, 122.45 — unbroken since 1997") + annotate("text", x = 2014, y = 115.6, hjust = 0, size = 3.2, colour = sex_pal["Male"], fontface = "italic", label = "Kimura, 116.13 (2013)") + scale_colour_manual(values = sex_pal, name = NULL) + labs( title = "The longevity record has stagnated — especially for women", subtitle = "Maximum verified age at death recorded to date (within this top-100 sample)", x = NULL, y = "Record age at death" ) + theme(legend.position = "bottom") ``` This stagnation is one of the livelier fights in contemporary demography. One camp reads the flat ceiling as evidence of a **hard limit to human lifespan** (Dong, Milholland & Vijg's much-disputed 2016 *Nature* paper put it near 115). Another finds **no detectable plateau in the hazard** — Barbi et al. (2018) estimated mortality flattening around 50% per year after 105, which makes records a waiting game governed by how many people *reach* 105. And a third, more corrosive line of argument (Newman) holds that extreme-age records are heavily contaminated by **clerical error and fraud** — on that view, the cleaner the data get, the harder records become to break, and Calment's outlier status is itself suspicious (a claim her validators vigorously contest). This dataset can't settle that fight — it's the top of the order statistics, not the hazard. But it visualises the stakes: a 3-year, 26-year-old gap between one lifeline and every other human being ever verified. ::: {.callout-note} ## Sample caveat These are the top 100 *per sex*, so the early decades of the "running record" lines are right-censored: anyone who held the record in, say, 1975 but ranks below today's top-100 isn't in the data. The frontier shape is trustworthy from roughly the late 1980s onward. :::