Measuring Cup Accuracy
The recent thread on measuring spoon accuracy got me wondering about the accuracy of my measuring cups. I have five glass measuring cups: a 4 cup Pyrex, a 2 cup Bennington Flameware, a 1 cup Bennington Flameware, a 1 cup Anchor Hocking, and an ancient 1 cup Pyrex.
Using my digital scale, I tared each cup and filled them with water to various marks. One cup should weigh 8 oz. The results:
Pyrex 4 cup: 1c = 7.3 oz, 2c = 15.4 oz, 3c = 24.5 oz, 4c = 33.2 oz
Flameware 2 cup: 1c = 8.2 oz, 2 c = 17.1 oz
Flameware 1 cup: 1c = 7.4 oz
Anchor Hocking 1 cup: 1c = 8 oz
Pyrex 1 cup: 1c = 7.9 oz
Though none are enough off to throw off a recipe, it is interesting to know the difference between them.
Your measuring technique may introduce a bias as well. The surface of a liquid in a confined space is curved, because of surface tension. In laboratory work, one uses a narrow container and reads at the miniscus. I don't know how an ordinary measuring cup is calibrated, but in any case I'm sure you don't have tenth of an ounce precesion in reading it. Just because your scale reports it doesn't mean that it is significant in the context of your experiment.
Did you do multiple trials to test the consistency of your measurement? How many? What is the variance?
That's why chemists use volumetric flasks for publishable experiments. If you are really interested in that level of accuracy (pastry is really the only area where quite that level seems necessary), you can buy them online and through scientific supply catalogs. You can also learn more about creating the closest possible execution of a theoretical model under online headings referring to Quantitative Measurement in relation to chemistry.
I can't resist. Correct as usual, the best way to make the comparisons is to repeat the experiment a number of times, typically 10 is an acceptable number. Then you run the statistics, average and standard deviation for each of the measuring cups. You can then look at stastical varriation and determine if there is a stastical difference and not jsut random error. Yes, way more work than it's probably worth, but necessary if one is to make a specific claim as to the accuracy of ones experiment.
Much better put than I did. Now looking back at what I wrote, I think some people understand my statements, but others may not.
What I meant is that if I measured a "1-cup" once and got a -10% error, then we don't know if this -10% is due to the cup or due to me or both -- probably both.
Now, if I measured the "1-cup" three times and got -10%, -8% and -12% deviations, these errors have a different meaning than if I got -10%, +8%, +2% errors.
Of course, to really investigate this, one need the cup to be measured by at least another people to validate. The reason is that some people are naturally biased. As such, one cannot assign all the "systemtic error" to the cup. For example, one may not know how to measure read the meniscus. In that case, the investigtor will constantly introduce a systemtic error.
Neverthless, having another person to validate the results would be "too much" for most situations.