References
Reaction-time testing is full of numbers that get repeated until nobody remembers where they came from. This page lists every external reference this site uses, and — more usefully — what each one is actually being used to support. If a claim on this site is not traceable to one of these, we have tried to rewrite it as an observation about how this site measures rather than present it as a research finding.
Nothing here has reviewed or endorsed this site. These are references we read, not authorities speaking on our behalf. Our own measurement method is documented separately on themethodology page.
Robert J. Kosinski (2013). Clemson University.
A Literature Review on Reaction Time
Used for: The reference summary this site uses for population reaction-time figures. Reports the long-standing laboratory means of roughly 190 ms for light stimuli and 160 ms for sound in college-age adults, and the standard division of reaction time into simple, recognition, and choice tasks. Also the source of the Hick (1952) proportionality finding described on the choice reaction time test.
James T. Eckner, Jeffrey S. Kutcher, and James K. Richardson (2010). Journal of Athletic Training, 45(4), 327–332.
Used for: The clearest published demonstration that apparatus choice changes the score for an identical task. The same football players averaged about 203 ms on a falling meter stick but about 268 ms on a validated computerized battery — a difference of roughly 65 ms, with the computerized measure also far more variable (SD 44 ms versus 20 ms). This is why this site publishes its own method instead of implying a browser reading is equivalent to a laboratory one.
David J. C. MacKay (2018). Cambridge University Press.
Information Theory, Inference, and Learning Algorithms
Used for: Source of the normal distribution complementary error function this site uses to estimate a percentile rank from a reaction time. The error function approximation and its maximum absolute error of roughly 1.5×10⁻⁷ are given in the appendix tables.
Milton Abramowitz and Irene A. Stegun (eds.) (1964). National Bureau of Standards, Applied Mathematics Series 55.
Handbook of Mathematical Functions
Used for: The reference table for the error-function approximation implemented in the percentile code. Formula 7.1.26 is the one used, chosen for accuracy rather than speed of evaluation.
George A. Miller (1956). Psychological Review, 63(2), 81–97.
The Magical Number Seven, Plus or Minus Two
Used for: The origin of the seven-plus-or-minus-two figure quoted on the number memory test page. Miller argued about span measured in bits rather than a fixed count of digits, which is why this site presents the figure as a starting point rather than a limit.
Nelson Cowan (2001). Behavioral and Brain Sciences, 24(1), 87–114.
The Magical Number 4 in Short-Term Memory
Used for: The re-analysis arguing short-term memory capacity is closer to four chunks than seven, distinguishing items that survive a distractor from those that do not. Cited on the number memory page alongside Miller so readers see both the original figure and the revision.
W. E. Hick (1952). Psychological Review, 59(3), 89–95.
On the Theory of Rate of Information
Used for: The finding that choice reaction time grows with the logarithm of the number of possible stimuli. This is why the choice reaction time test on this site fixes a four-colour mapping rather than adding more options.
Gianluca Del Rossi, Alfonso Malaguti, and Samanta Del Rossi (2014). Journal of Athletic Training, 49(3), 356–359.
Practice Effects Associated With Repeated Assessment of a Clinical Test of Reaction Time
Used for: Ten sessions of a ruler-drop task over five weeks, in 43 participants. Mean reaction time fell from 264.9 ms in session 1 to 252.0 ms in session 10 — about 13 ms — but almost the entire gain arrived between sessions 1 and 2 (a 7.2 ms drop that did not itself reach significance), and no two consecutive sessions after that differed significantly. The authors conclude a practice session should be discarded before recording a baseline. This is the evidence behind the site treating a first attempt as a warm-up rather than a measurement.
Gebremariam Der and Ian J. Deary (2006). Psychological Bulletin, 132(2), 201–231.
A meta-analysis of the effects of ageing on human speed and reaction time
Used for: A large-scale synthesis of the age-graded reaction-time figures this site uses as reference bands. Reaction time is fastest in the early twenties and slows from middle age onward, with the effect larger for complex than for simple tasks. Variation between individuals inside one age band is wider than the gap between adjacent bands, which is how this site presents the numbers.
W3C / W3C TAG (2015). W3C.
High Resolution Time — W3C Recommendation
Used for: The specification defining `performance.now()` and `Event.timeStamp`, the two clocks this site uses for measurement and for the latency probe.
Mozilla Developer Network contributors (2024). MDN Web Docs.
Used for: Reference for the policy this site serves, including the split between `style-src` and `style-src-attr` described on the methodology page.
Being clear about the limits is more useful than a longer list of references. A few things worth stating outright:
If you find a figure on this site that contradicts the sources above, we want to know. Thecontact page is the fastest route. Corrections already made are noted where the relevant claim appears.