I would like to test ~50 randomly-selected Americans by asking them to put 1.7
ounces of water in a glass, then measure how close they got. Then test the
same number Europeans who grew up with the metric system, this time asking
for 50 ml.
But you are stacking the deck (biasing the test) in favor of metric by using a rather common, rounded metric size (50 mL) vs. an awkward, non-rounded Olde English size (1.7 fl. oz.). Metric DOESN'T NEED to have an unfair advantage! If we give metric that kind of advantage and then metric wins people will disregard the results by correctly stating that metric had an unfair advantage.
A better test would be to use 2 fl. oz. for the Olde English quantity and the equivalent 59 mL for the metric value. That gives the Olde English the advantage of having the bias in its favor and yet I'd be willing to bet the metric users could still do better.
Then there is the possibility of doing a double test, using both sets of values above. Have the Olde English user try to estimate (a) 1.7 fl. oz. and (b) 2.0 fl. oz., while having the metric user estimate 50 mL and 59 mL.
Another possibility (using just a single test) is to make both samples simple numbers. There is really no reason why they have to be equal, although it may be useful if they are similar. So, use 2 fl. oz. and 50 mL.
To prove the worth of metric vs. English, insist upon tests that are neutral or balanced regarding what is to be tested. Devising good, scientific experiments requires a lot of careful thought.
(PS You also have overlooked the not-so-insignificant problem of the accuracy with which the MEASUREMENTS have to be made to check the subjects' estimated amounts. Who would measure the amounts to check them, an Olde English user or a metric user?)
Regards, Bill Hooper Fernandina Beach, Florida, USA ======================== SIMPLIFICATION begins with SI ========================
