How comparable are personal reading-speed measurements between paper, e-reader and phone for the same reader and text?

Open question

The reading-log protocol on this wiki records pages and minutes per session, derives pages per hour per book only, and states that it is not a reading-speed test because pages differ between formats. A reading-speed measurement would use passages whose word count is known (for a digital text, wc -w counts words as non-white-space sequences delimited by white space, which is one definition and not the only one) and time the reader on each. The question is what such a measurement means across formats: if the same reader reads passages from the same book on paper, on an e-ink reader and on a phone, the three figures differ by some amount, and it is unclear how much of that is the format and how much is the passage, the font size, the line length, the ambient light, the time of day, the order in which the formats were tried, or the reader's knowledge that they are being timed. It is also unclear whether a comprehension check changes the pace and therefore whether a figure taken without one is comparable with a figure taken with one.

What a useful answer contains

• A protocol that states the word-count rule, the passage length, how passages are assigned to formats (rotated so each format gets each passage position), and what is held constant in typography.
• A statement of what "reading" meant: whether a comprehension question followed, and whether a passage read too fast to answer it was discarded.
• A log of at least several weeks from one reader with all three formats, showing per-session figures and their spread, not just means, and whether the ordering of formats stayed the same from week to week.
• Which conditions, when changed on purpose (font size, line length, light), moved the figure by more than the week-to-week spread within one format.
• What happened when a second reader followed the same protocol: whether the between-format differences were similar in sign, in size, or neither.
• An honest account of any difference that vanished once passage order or time of day was controlled.