Royce 1970 and the Waterfall Misreading

Narrative

The financial press has almost nothing on the Royce and waterfall origin story itself. Searches of the Financial Times, WSJ, The Economist and Bloomberg returned no feature on the 1970 paper, Figure 2 or the 1976 Bell and Thayer naming. What exists is adjacent coverage of the downstream cost: a Bloomberg Law report (c. 2023) on a GAO finding that the Pentagon still relied on an outdated "waterfall" system and had not moved to Agile in some weapons programmes, and a Reuters piece (March 2025) on a Hegseth memo making the Software Acquisition Pathway the preferred route for DoD software. Treat both as evidence of procurement legacy, not of Royce scholarship. The primary-source history sits in academic and practitioner outlets, and this lane should be read as a gap flag rather than a verdict.

On what Royce wrote, the sources agree on the core points. The word "waterfall" does not appear in his paper (Martin Fowler, Wheeler). Wheeler quotes Royce calling the simple sequential method "risky and invites failure", and Fowler adds a counterweight: Royce called the illustrated approach "fundamentally sound", and he expected iteration between adjacent steps while treating longer backtracks as a serious problem. That makes "Royce opposed waterfall" a contested simplification rather than a clean fact. The five additions (design first, documentation, do it twice, test planning, customer involvement) are summarised in secondary sources. Boehm's 1988 spiral paper, which is a primary source, already credits the 1970 model with feedback loops confined to successive stages and a "build it twice" step, and notes it became the basis for most acquisition standards.

On naming and adoption, an arXiv study (2023) traces the earliest use of "waterfall" for Royce's model to Bell and Thayer (1976), and a 2025 arXiv history by Saravanos repeats the point. Dating of the paper varies: many citations give 1987, which is the ICSE reprint, while the original is the 1970 WESCON paper. Claims about DoD-STD-2167 conflict. ScienceDirect's overview calls it a "grievous misreading" of Royce, but Mosaic Projects argues 2167 (1985) was an iterative standard and that the sequential emphasis came with 2167A (1988), while Wikipedia cites section 4.1.2 of 2167 allowing an incremental build approach. Larman and Basili (2003) report that the DoD changed its waterfall-based standards at the end of 1987 after the Brooks-chaired Defense Science Board report, per a secondary arXiv citation. The 1994 date for abandonment is asserted by Mosaic and ScienceDirect and needs checking against the standards themselves.

On myth versus support, Larman and Basili's abstract supports the claim that iterative and incremental development dates from the mid-1950s, and Walker Royce's remark that his father was always a proponent of iterative, incremental, evolutionary development is quoted via Larman's book, so it is second-hand. Bossavit (InfoQ) challenges the popular "nobody read past page 2" story, arguing Royce did not originate the sequential-phases model; this tells against the neat misreading narrative, though Bossavit's full chapter was not retrieved. Benington's 1956 paper is the usual alternative origin. The "Royce was Agile" retelling is oversimplified: Fowler notes practitioners in enterprises still describe waterfall as viable, and a July 2026 ProjectManagement.com commenter warns that the misreading explanation is too tidy.