published in each syllabus or guide; the five disagree on assignment symbol and index base
Textbook pseudocode with its own typesetting package
Cormen 2003
package on CTAN 2026
Executable specification for distributed designs
Newcombe 2015
10 systems, 7 teams (the company's own count)
Pseudocode as a design step before code
McConnell 2004
a chapter of a widely read book (not read here)
Notebooks as literate programs
Rule 2018
over one million public notebooks
Upper bounds (achieved)
Flowcharts were read faster, with fewer errors, at all three complexity levels, with few exceptions, in 114 participants (Andrzejewska 2022; abstract).
Professionals coded 31% faster from PDL than from prose in one setting (29.7 against 20.5 minutes; Boehm-Davis 1983).
A pseudocode line corresponds to one or two code lines in a human-written corpus (Kulal 2019).
Over 320 thousand algorithm blocks have been extracted from arXiv sources (Toksoz 2024).
Lower bounds (not yet shown)
No study found that compares a published pseudocode against real code for reimplementation by developers, with a fixed task and reader group.
No study found that tests the advice to add pseudocode "to decrease ambiguity" (Gundersen 2018).
No controlled comprehension experiment of literate programs was found in this search (a thesis, Bertholf 2000, exists and was not read).
No study found that measures line numbers in pseudocode, or the typed-assignment arrow against an equals sign, although the boards use arrow, equals sign and SET ... TO.
No study found that compares the pseudocode of Cormen, Knuth and Sedgewick side by side; Sedgewick's 4th edition ships Java, so the contrast is also pseudocode against code.