Erik_Mag
My point in using "consistently" was in respect that variations in pushing the start/stop button can be substantially less than the reaction time (in my line of work reaction time = propagation delay and variations in timing = jitter).
As it happens, I had the chance to play around with this in Professor Galanter's lab. The specific issue is the uncertainty in latency between two haptically-similar events (here, using the thumb to press the stopwatch button while observing a 'zero crossing' timing event, whether it is to start or stop the device). Most RT studies seem to concentrate on more complex actions, notably the combination of perception and reflex arcs involved in foot-braking an automobile. The specific data for stopwatch manipulation involves far shorter nerve paths, and with training (as my European correspondent pointed out) quite a bit of the latency variation corresponding to jitter could be avoided. But it is still present.
There is better portable instrumentation available today, which is capable of reading neuromuscular actuation directly (to the necessary precision) and associate that with absolute event timing from a more precise source (e.g., optical sensing using edge detection, for this example of timing mileposts). For an investment of about $300 plus a suitable smartphone (which can be obsolescent; you only need 'communications' if you have no other way to download software) you could conduct proper data gathering and perform reasonable analysis on the results for this very specific application. The results would be useful research in a number of significant respects involving accuracy of handheld timing -- if it wouldn't already tell you what Big Jim already knows: the variance is already on the order of the precision needed for measurement.
Using some of his examples, if the variance is on the order of the 0.1sec commonly quoted in what there is of the literature, even his longest 'mile' (which is out 54 feet) produces a difference at 100mph that is only about three times the variance. Something pointedly not mentioned is whether the distribution is symmetric - does a given individual routinely tend more toward anticipatory variance, or lag? Again you could produce a correction table, if you are of nitpicking persuasion ... and this might actually be 'important' in critical cases, for example if you absolutely need to know as a railfan that your train went 100mph instead of 97.94.
Another way to put it is that the 1/100th second resolution of the digital stopwatch should be fine for 99+% of the users and finer resolution would be needed by only a few very skilled users.
I might point out that for 100mph timing, a 1/100 minute stopwatch gives you nominal precision to a mile per hour. The 1/100sec resolution is fine ... until you start invoking that number of decimal places in the precision of the measurement. Lest you think this is trivial, the "126.1mph" claim for Mallard relies upon a very short trace at the indicated speed, which brings up the timing precision of the motor driving the chart recorder, probably demonstrably well short of 1/100sec accuracy and perhaps whatever the corresponding measure to wow/flutter would be for charts. Again, agreeing with Big Jim, it's hopelessly nitpicky to derive meaning from this kind of record ... but don't try telling that to a wide cross-section of English railfans!
One advantage of the digital stopwatch is that it is easier to read than an mechanical stopwatch.
That is true even if you have a true 3sec/10ms stopwatch (which is a joy to behold if you like fine machinery) but a much more important difference is the longer muscular 'throw' and effort, and internal mechanical factors, involved in starting or stopping a mechanical watch. Somewhere I have data about the difference in haptic touch for various kinds of typewriter-keyboard feedback ... and the much larger difference, including key-to-key variance, an important potential input for biometric security, for mechanical typing.
The big problem I have with digital stopwatches is the deceptive precision they give. I happily wear a watch that, among its other features, provides a 1/1000 timing function (and the quartz chip in the watch supposedly has an oven to permit that precision, although to date I haven't checked this scientifically!). The slop in the mechanical pusher switches alone is a substantial multiple of .001sec, and there is of course no breakover or other haptic confirmation when the button is 'making' for the timing command. (My somewhat unscientific testing indicated somewhere around .044sec variance when quickly starting and stopping by pure reflex) While there are contacts for more direct electrical actuation, I have to conclude that at least that last digit will vary so much that it can't be relied on for something meaningful. But tell that to the average guy who bought one of these watches for three-digit-precision bragging rights!