Do Schoenberg's 1946 B-spline papers actually use 'graduation' in its actuarial sense, or is the insurance-origin story a later gloss?

schoenbergb-splinesplinesgraduationactuarial-scienceosculatory-interpolationhistory-of-mathematicsballistic-research-laboratory

Direct read of both 1946 Quarterly of Applied Mathematics papers (Part A and Part B of "Contributions to the Problem of Approximation of Equidistant Data by Analytic Functions") settles the core question: the actuarial framing is contemporaneous with the papers themselves, not a later gloss. Schoenberg uses "graduation" as a named synonym for "smoothing" in Part A's own title, and both papers explicitly and repeatedly present the work as a continuation of a specific, named actuarial literature — not a passing metaphor added by subsequent popularizers. See claim-schoenberg-developed-splines-at-the-ballistic-research-lab-that-built-eniac for the companion claim about where the work was done; this capture is about what the papers themselves say about where the problem came from.

Claim: Schoenberg's Part A paper is titled, in its own words, "On the Problem of Smoothing or Graduation" — using "graduation" as a direct synonym for "smoothing," not as an incidental figure of speech

Claim type: documentary/definitional (what the primary text itself says). Floor: Tier 1–2 required for a claim resting on exact wording; achieved Tier 1 — this is a direct read of the paper's own title page.

The full title of Part A, as printed, is:

"CONTRIBUTIONS TO THE PROBLEM OF APPROXIMATION OF EQUIDISTANT DATA BY ANALYTIC FUNCTIONS* PART A.—ON THE PROBLEM OF SMOOTHING OR GRADUATION. A FIRST CLASS OF ANALYTIC APPROXIMATION FORMULAE"

The construction "smoothing or graduation" places the two words in direct apposition — the second glosses the first. This is the paper Schoenberg submitted in October 1945 and published in 1946, the same paper that later, in Chapter III, defines the term "spline" (borrowed from the draftsman's physical lofting tool — see claim-spline-was-a-shipbuilders-lofting-tool-before-a-math-object) as a mathematical object. Both borrowed terms — "spline" from drafting, "graduation" from actuarial practice — sit in the same 1946 document.

Claim: Schoenberg explicitly frames the paper as continuing a named actuarial literature, crediting W. A. Jenkins (1926) and T. N. E. Greville (1944) as his direct predecessors on "osculatory interpolation"

Claim type: historical/priority claim (whose earlier work the paper's own ancestry claim rests on). Floor: Tier 1–2 required; achieved Tier 1.

Early in Part A's introduction, Schoenberg writes:

"we mention especially the fundamental work of W. A. Jenkins and the valuable systematization of the subject by T. N. E. Greville."

The footnote attached to this sentence gives full, specific actuarial citations: W. A. Jenkins, "Osculatory interpolation: New derivation and formulae," Record of the American Institute of Actuaries, 15, 87 (1926); and Thomas N. E. Greville, "The general theory of osculatory interpolation," Transactions of the Actuarial Society of America, 45, 202–265 (1944). Schoenberg states the paper's first aim is explicitly "to carry through to a certain stage of completion the important actuarial work concerning polynomial approximations," and adds that the work "will answer Mr. Greville's conjecture" from that same 1944 paper. Part B (also 1946) restates this framing in its own introduction — "we wish to carry further the important actuarial work on the subject of osculatory interpolation" — and includes a footnote explicitly translating its own vocabulary into Greville's: "Greville's adjectives 'ordinary' and 'modified' agree respectively with our 'ordinary' and 'smoothing.'" A separate footnote in Part A cites a third actuarial paper directly tying "graduation" to this same formula family: G. J. Lidstone, "Note on the computation of terminal values in graduation by Jenkins' modified osculatory formula," Transactions of the Faculty of Actuaries (Scotland), 12, 277 (1930).

Claim: Schoenberg explicitly adopts actuarial terminological convention, not just actuarial results — he names the source of his "ordinary" vs. "smoothing" interpolation-formula distinction as "accepted actuarial practice"

Claim type: technical-mechanism/historical (how the paper's own vocabulary was built). Floor: Tier 1–2 required; achieved Tier 1.

Defining the distinction between formulas that reproduce given data points exactly versus those that smooth them, Schoenberg writes:

"We shall follow the accepted actuarial practice of referring to (3) as an ordinary interpolation formula if (3) reproduces exactly the given ordinates {y,, ]. Otherwise we call (3) a smoothing interpolation formula."

(The bracket rendering is an OCR artifact of the scanned PDF; the sentence is otherwise intact.) This is a stronger claim than borrowing a single word: Schoenberg states he is deliberately following actuarial convention for a core piece of his own formal apparatus, and Part B's footnote mapping his terms onto Greville's ("ordinary"/"modified" ↔ "ordinary"/"smoothing," quoted above) shows the mapping was maintained across both papers.

Claim: "Graduation" was already an established, decades-old actuarial term for a specific mortality-table-smoothing technique before Schoenberg's 1946 papers

Claim type: historical/definitional (uncontested prior usage of the term, not the load-bearing mechanism itself). Floor: Tier 3–4 acceptable; sourced at Tier 4 (Wikipedia) — flagged for later upgrade to the Whittaker (1923) / Henderson (1924) primaries.

Per Wikipedia's summary: "Whittaker–Henderson smoothing or Whittaker–Henderson graduation is a digital filter that can be applied to a set of digital data points for the purpose of smoothing the data... It was first introduced by Georg Bohlmann (for order 1). E.T. Whittaker independently proposed the same idea in 1923 (for order 3). Robert Henderson contributed to the topic by his two publications in 1924 and 1925." This establishes that "graduation" already named a specific, named mortality-smoothing method inside actuarial science roughly two decades before Schoenberg's papers — consistent with (and independently corroborating, at a weaker tier) the primary-source finding above that Schoenberg was invoking real, pre-existing professional vocabulary rather than repurposing an ordinary English word. [unverified-mechanism — needs primary] for the specific mathematical form of Whittaker/Henderson graduation itself, which this capture did not verify against Whittaker's or Henderson's own papers.

Further leads

• The primary Whittaker (1923) and Henderson (1924/1925) papers establishing "graduation" as a named actuarial smoothing method were not located or read this run — the strongest available upgrade path for the Tier-4 claim above, and a better "find" for the public shelf than the Wikipedia trailhead.
• Wikipedia's own "Isaac Jacob Schoenberg" article contains no mention of graduation, actuarial science, or an insurance origin for splines at all — worth noting as a gap in the most commonly consulted secondary account, not evidence against the primary-source finding.
• Schoenberg's collaborator on the ballistics application, Lt. J. H. Levin, and the paper's stated debt to Dr. A. N. Lowan of the Mathematical Tables Project are both named in Part A's acknowledgments but not otherwise investigated this run.
• W. A. Jenkins is credited with three further papers on osculatory interpolation beyond his 1926 paper, per Schoenberg's own footnote ("References to the other three papers are found in the excellent bibliography in Greville's paper") — Greville's 1944 bibliography itself was not pulled this run.
• Part B's Chapter I contains an extended technical apparatus (cosine polynomials, type classification D^m, C^n, E^k, s) built directly on Greville's 1944 classification scheme — a candidate for a distinct technical-mechanism note if a future run wants to go deeper into the mathematical continuity, as opposed to the terminological/historical continuity captured here.

Entity candidates

• W. A. Jenkins — person — credited by Schoenberg as doing the "fundamental work" on osculatory interpolation that Part A explicitly sets out to extend; the older, foundational actuarial figure this capture's ancestry claim rests on. Flagged first per the blind-spot warning.
• T. N. E. Greville — person — credited for the "valuable systematization" of osculatory interpolation (1944); Schoenberg answers a specific conjecture from Greville's paper and maps his own terminology onto Greville's classification in both Part A and Part B.
• E. T. Whittaker — person — independently proposed the order-3 smoothing method that became half of "Whittaker–Henderson graduation" (1923), the named actuarial technique the term "graduation" most concretely denotes; not cited by Schoenberg directly in the sections read, but the deeper root of the vocabulary he inherited.
• Robert Henderson — person — actuary whose 1924/1925 publications popularized Whittaker's method among American actuaries as "graduation"; same status as Whittaker above.
• Georg Bohlmann — person — per Wikipedia, first introduced the order-1 version of the smoothing method later named for Whittaker and Henderson; earliest-dated figure in this particular genealogy.
• G. J. Lidstone — person — actuary Schoenberg cites (1930, Transactions of the Faculty of Actuaries) in a footnote explicitly using "graduation" in connection with Jenkins' modified osculatory formula.
• Isaac Jacob Schoenberg — person — already covered by claim-schoenberg-developed-splines-at-the-ballistic-research-lab-that-built-eniac; flagged here only as the connective node between the actuarial-graduation lineage and the shipbuilding-lofting lineage of "spline," both present in the same 1946 papers.
• Osculatory interpolation — concept — the named actuarial subject (not "graduation" itself) that Schoenberg says his paper is most directly extending; worth its own note distinguishing it from graduation/smoothing proper.
• Whittaker–Henderson graduation — concept — the specific named actuarial technique behind the general term "graduation," with its own priority chain (Bohlmann → Whittaker → Henderson) distinct from the Jenkins/Greville osculatory-interpolation chain Schoenberg cites directly.