Royce 1970 and the Waterfall Misreading

Narrative

The primary text is short and easy to check. Royce's paper appeared in the August 1970 WESCON proceedings (pages 1 to 9), and a bibliographic record notes a reprint in the Ninth International Conference on Software Engineering proceedings of March 1987 (pp. 328 to 338). The 1987 reprint explains why some citations carry a 1987 date. Royce opens by saying his views come from nine years of spacecraft mission planning, commanding and post-flight analysis software. He then admits uneven success on time and cost. A MIT Press anthology reprint (2021) shows the paper is still treated as canonical. Verdict on "Royce coined waterfall": myth. Every secondary source retrieved agrees the word is absent from the paper.

The naming chain is supported by documents, though the sources are secondary. Bell and Thayer (TRW, 2nd ICSE, 1976) wrote that Royce "introduced the concept of the 'waterfall' of development activities". Blog analyses of that passage say they reproduced the page-2 sequential diagram and not Royce's later figures. Boehm's 1987 remark, quoted on Wikipedia, is the best contemporary scholarly hedge. He called Royce's paper the one generally thought to define the stagewise waterfall model, yet noted that Benington and Hosier had earlier approximations and that Royce already included prototyping. Verdict on "Royce's Figure 2 was his recommendation": contested at best. Secondary sources quote him calling the implementation "risky and invites failure" and endorsing the later Figure 4 approach as "fundamentally sound". Royce's five additions (design first, documentation, do it twice, test planning, customer involvement) are consistent across accounts, but this lane did not retrieve the paper's full text for page-level checking.

Procurement is where the evidence is thinnest and the sources least reliable. DOD-STD-2167 (4 June 1985) listed sequential phases, and 2167A (29 February 1988) drew vendor criticism for bias towards waterfall through its formal reviews and audits. Yet 2167 section 4.1.2, quoted on Wikipedia, already allows overlapping iterations and "incremental build". MIL-STD-498 (November 1994) replaced 2167A and offered grand design, incremental and evolutionary strategies, according to a practitioner summary. Several retrieved sources contain dating errors: Smartsheet places 2167A in 1985, and Mosaic Projects gives 1996 for MIL-STD-498. Treat these as unreliable. An arXiv thesis cites Larman and Basili for a Brooks-chaired Defense Science Board report of October 1987 prompting the DoD to permit iterative development. I could not verify that claim against the report itself.

Larman and Basili (IEEE Computer, June 2003, about 2,500 Google Scholar citations) is the key counter-narrative. It traces iterative and incremental development to the mid-1950s, argues that each decade's thought leaders supported it, and that many large projects used it successfully. Gerald Weinberg later recalled that colleagues thought waterfalling a huge project was "rather stupid", which suggests documentation overstated waterfall's real dominance. Walker Royce's testimony, relayed via Larman and Basili and reproduced in a book excerpt, says his father "was always a proponent of iterative, incremental, evolutionary development". Walker's own textbook also calls the paper the most quoted source of the waterfall model. That is a family witness reported second-hand, not archival evidence. Verdict on "nobody read past page 2": plausible but unproven, since it rests on inference and blog-level reading.