The single biggest mistake in all the science fair projects I evaluated yesterday was no replication or incorrect replication, so today’s post is geared toward covering this specific topic.
All science experiments MUST be replicated. That means you have to repeat everything you did, exactly the same way (to the best of your ability), a minimum of 3 times. Why 3? The first time, you will have a number, but you can’t have much confidence in how good that number is. So you repeat the experiment to see how repeatable the response is.
The second time you do the same experiment, you might get the same answer, but chances are you will get a slightly different number. How different, will depend on several things including the precision of your equipment, the degree to which the experimental procedure was standardized, and the natural variability of the relationship between the independent and dependent variables. The problem now, is that you have 2 numbers and you don’t know which one is “better”, so you repeat the experiment again.
With 3 numbers in hand, you have an indication of how repeatable the response is. If your numbers are all similar, you can have more confidence in them. You should report all three along with an average, if that is appropriate for your grade level. If, however, 2 numbers are similar and 1 is really different, then you might think back to that trial (and refer to your notes if you are keeping a scientific notebook) to speculate on what might have resulted in the difference. You can always repeat the experiment again. In general, the greater the level of replication, the more data you will have, and the more you can say about the data and the experiment. In science there is usually a tradeoff between the level of replication and the use of resources, including time, supplies, equipment, etc.
KEY: Many students think/claim they have replication because they have “tested” 3 juices, or metals, or parachute designs, or balls, or whatever, but that is not the correct type of replication (i.e. incorrect or improper replication). Projects need to include 3 trials for EACH item (not just 3 items).