How to Impress a Science Fair Judge

I always entered my school science fairs as a child. In elementary school I'll admit that my parents did most of the work, but as I progressed through middle and high school I did more and more of the work myself. My first round of science fair participation culminated in a grand prize award at our local county science fair.

These days I am well into my second round of science fair participation: I now go to science fairs as a judge. The change in perspective is great, but the ingredients of a winning science fair project are as clear to me now as they were in high school. Having just finished judging a competition at a local science and technology magnet school I thought I'd share some thoughts on several easy ways to impress judges and improve your chances. The great thing about these suggestions are their very simplicity. Follow these six simple tips and I guarantee improved results at your next science fair.

• Tip 1. Pick a good project.

This is the probably the hardest part for beginners, but a few simple guidelines are in order. A good project does not have to be complicated, involve exotic equipment, or require highly advanced skills. A good project identifies some hypotheses or goals which are clearly explained. A good project has not one, but a sequence of tests or experiments which are clearly described in a written set of procedures. A good project has a clear set of results which are supported by the data. A good project includes an acknowledgement of the limitations of experimental design and data analysis, including alternative explanations or suggestions on how to improve the experimental procedures.

The important point is that a good project allows one to follow the scientific method. In my mind this is more important than the novelty of your project or even the correctness of the results.

There are lots of books and web sites that list ideas for science fair projects. I have two overall suggestions: 1) choose a project which interests you, and 2) keep it simple enough where you can understand the whole project. As a judge I'd rather see a simple project that has been well executed than a complicated project that has been poorly done.

One of the best projects I ever saw was a botany project done by a seventh grader. It was a Saturday night and the room was packed with parents and presenters. While most of the botany projects were overflowing with green plants and admiring parents, there was a gap in the audience at one project that appeared to offer up only a few pots containing dead grass. Botany is not my area, but I felt sorry for the young man standing there, so I asked him to explain his project to me. His face lit up as he explained that he had heard a story about the bluegrass of Kentucky, and had decided to perform an experiment to create grass in colors other than green.

In his first experiment he had 3 pots of grasses, one he watered with diluted red food coloring and one with diluted blue food coloring. The last pot of grass was watered with tap water as a control. A month went by and the color of the grass did not change. In his second experiment he increased the amount of dye by using straight food coloring, but again nothing happened. He concluded that whatever this green stuff was, it must be pretty strong. He decided he needed to get rid of the green dye in the plants before proceeding, and so he began his next experiment by "watering" his plants with Clorox bleach. Sure enough the grass turned white. Unfortunately it also died. It was at this point he ran out of time, hence the sorry appearance of his project. His final conclusion was that the green in the plants must be necessary for the life of the plant.

The beauty of this science fair project was not complexity or even the validity of the science. It was that the student had obviously done the project himself, that he had been persistent in the face of failures performing multiple progressive experiments, each of which was reasonably well designed and thought out. Finally the student had clearly learned something from his experiments. I walked away thinking that given enough time this was the kind of young man that would eventually discover chlorophyll.

• Tip 2. Understand your project.

This may seem obvious but you should understand what you have done in your project and why you did it. A number of projects at our local magnet school were performed by students who had internships at a prominent local research institute. The data analyses that these students had performed were far more sophisticated than the analyses of the students who had not been working in a research environment. The problem was that few if any of these students could explain in simple terms why they performed the analyses in the way that they did.

Let me give a specific example. One young lady had performed a statistical analysis technique called a least squares fit in order to fit a mathematical model to some data. This technique is typically taught in first or second year college courses. This student was quite bright and had reproduced the mathematical development of least squares fitting in a page of complicated looking equations in her research write up. It was quite impressive, but as a judge I wanted be certain she understood what she had done. So I asked her why she had chosen to perform a least squares fit. I wanted to know in non-mathematical terms what a least squares fit was all about. (The answer I was looking for was that a least squares fit is a method to adjust a mathematical model to minimize the differences between the model and a set of data. As a judge I wasn't looking for math, just words describing a simple concept.) Unfortunately she froze and couldn't answer the question. If I had been the only judge to ask her the question, I would have thought that her freezing up had more to do with nerves than anything else. But it turned out that another judge independently asked her the same question, with the same result. In the end we were forced to conclude that this very able student had performed a rather sophisticated statistical analysis that she didn't quite understand. In my opinion that was the difference between first place and third place.

More recently I reviewed a project analyzing the Gulf Stream, a topic I know quite a bit about. In my interview I asked the student three questions, all of which were a bit unfair: why is there a Gulf Stream, are there other "Gulf Stream-like" currents in the world, and why does the Gulf Stream separate from the US coast at North Carolina. I say these were unfair questions because it took me several years of graduate school to understand the answers to these questions. The last question in particular was one of the outstanding questions in physical oceanography during the 20th century. But I wasn't probing for PhD-level knowledge, I just wanted to see how much the student understood about the subject of his study. In fact I was looking for the following three simple answers: varying wind fields over the Atlantic cause a general clockwise circulation in the Northern hemisphere, and this circulation is intensified along the Western boundary of the basin by the spinning of the Earth; other western boundary currents exist, for example the Kurishio in the northwestern Pacific; and I don't know. The final answer shouldn't surprise you - if you don't know the answer to a judge's question, just admit that you don't know.

The student stumbled a bit in attempting to answer my questions, but they still did well in the competition. I did suggest to the student though that they should be able to answer some general questions about the topic they are studying. It is very important in science to understand enough about your project to be able to be able to place it in a larger context for the general public. It should not be hard to anticipate the easy questions about your topic and prepare to answer them.

• Tip 3. Be able to concisely explain your project.

Concise means brief. Judges do not have a lot of time to spend on each project. So you need to be able to completely describe your project, making all the neat points you need to make, in a short period of time.

As you may be able to tell from my writing, I have a tendency to run on. I was even worse years ago, when I worked on my winning high school project. So I actually wrote out two speeches describing my project, one was 2 minutes long, and one was 5 minutes long. I practiced them to the point I was comfortable with points I needed to cover, but not so much as to make it seem rehearsed. This is a bit of work, but I suggest you do the same. At the very least write out an outline of the major points you want to make. It turns out that good communication is a big part of good science, so a solid concise story will impress the judges. And remember, concentrate on the big picture, not the tiniest details

• Tip 4. Carefully analyze and refine your experimental design.

In each experiment you perform, the outcome or result of the experiment should depend on some input variable which you control or measure. The outcome of a good experimental design depends on only a single input variable. In a poor design the outcome depends on a number of input variables or factors that act together in ways that are poorly measured or understood.

In a recent science fair a student examined the question of whether sunlight affects the rate of melting ice. Their procedure was to measure the time it took several ice cubes to melt, with some set on the ground in direct sunlight and some set on the ground in the shade. The input variable they controlled was the absence or presence of sunlight. The outcome was the time to melt. Unfortunately, the time to melt also critically depended on the temperature of the ground on which the ice cube was placed, which I can imagine depends on the amount of sunlight falling on the ground. So there was an uncontrolled and unmeasured variable in the experiment. The fact that the student did not recognize the clear flaw in their experimental design pushed them out of the running for an award. Recognizing and acknowledging the limitation might have raised the project to honorable mention status. To get a top prize the student would have had to recognize the flaw, and repeated the experiment with the ice cubes placed on an insulating material, like styrofoam.

• Tip 5. Be persistent.

A good experimentalist is persistent. One of the projects that did well at a recent fair involved the effects of different colored lights on heating water. The student built a styrofoam and glass apparatus with black pipes containing the water to make the measurements. Unfortunately the first set of measurements showed no result. He then made modifications to the apparatus, changing its dimensions in an attempt to make it more sensitive. Again he got a null result. Finally he completely redesigned the apparatus and this time got some positive results. In discussing the project with the student it was clear that he didn't quite understand the detailed physics involved with the measurements he was making. In fact some of his explanations were exactly backwards from scientific truth. And yet he won an award because he impressed me and the other judges with his persistence.

• Tip 6. Make multiple measurements to beat down the noise.

I complained in tip 2 about a student who had performed a sophisticated statistical analysis that she didn't understand. I didn't mean to suggest that statistical analyses are unimportant. Just the opposite - you should use statistics to analyze your data. But they don't have to be complicated. In fact, you should be able to analyze your data in a manner simple enough for anyone to understand. In the next few paragraphs I am going to try to teach you enough statistics to impress most judges. And I am going to do it with only one math equation and one graph.

Everytime we make a measurement of anything, whether it be length, color, or anything else we can measure, the measurement contains some error. That error can be broken up into two basic types: random and systematic. Random errors are those that fluctuate randomly from measurement to measurement. The typical example is if you were to measure the number of times a flipped coin landed heads up in 10 attempts, the result is random. Although on average it will land heads up 5 times out of 10, the actual result fluctuates each time you make the measurement. Systematic errors don't fluctuate and are due to some systematic problem with the measurement. For example, if I gave you a ruler that was marked in inches, but told you the ruler was marked in centimeters, then I will have introduced a big systematic error into your measurements. Unfortunately the source of most systematic errors are not that obvious. It is a goal of good experiment design to minimize systematic errors.

Statistical analyses has to do with the random fluctuations we see in measurements. Most of the time we use the arithmetic mean or average to represent the results of a series of measurements. To get the mean you sum all of the measurements together and then divide by the total number of measurements. (That is my one equation you need to know to do science fair projects.) We average results together because we want to reduce the effect of the random errors in our measurements. The idea is that some of the measurements will be lower than the true value, and some will be higher. Averaging tends to cancel these errors out, providing a result closer to the true value.

The net result is that you won't do well at the science fair with just one measurement. Taking a bunch of measurements, and averaging the results together tells the judges that you are beginning to understand the way things work. Experimental repetition and averaging may be enough to get you to honorable mention, but there is one more concept you need to understand to get to the top prizes. You'll need to know the answers to the questions of how much do I have to average, and how accurate is my result.

Now I promised only one equation, which I've already used up. But I'll mention that the answers to the above questions involve another calculation of a quantity called the standard deviation of the data. For most science fair projects it is probably sufficient to have a simple mental picture of what the standard deviation measures, and how it is used. If you can compute it, all the better, but in most cases the judges will be suitably impressed that you even understand the concept.

The idea is simple. If we did our multiple experiments of counting the number of heads in ten coin flips, we'd find a certain percentage of the time we'd end up with 5 heads. Likewise we could figure that some other percentage of the time we'd end up with 4 heads or 6 heads. There is even some small percentage of the time that we flip the coin 10 times and get all 10 landing heads up. This would be rare, but it will sometimes happen. Each of these percentages of time represents the probability that a particular outcome (such as 4 heads out of 10) will occur. If we plot these probabilities we will see that the most probable outcome is 5 heads, and the probability falls off from there. This is the way probabilities often work, with some peak in probability around the mean and less likely outcomes occuring far from the mean. The probability curve for our 10 coin experiment is shown below.



The standard deviation measures the width of this kind of probability curve. If the standard deviation is large, the probability peak is wide and any one measurement might fall a long way from the true mean value. If the standard deviation is small, the probability peak is narrow and it is likely that most of the measurements you make will be very close to the correct value - at least assuming there are no systematic errors.

Even if you don't compute the standard deviation of your measurements using a standard formula, you can easily estimate the standard deviation directly from your data. On estimate of the standard deviation is one half of the spread of the middle two-thirds of your data points. (This is not exactly correct, but it is close enough for our purposes.) For example, given the six readings 1.2,1.3,1.4,1.5,1.6,1.7 the mean would be 1.45 and the spread in the middle four readings (also known as the standard deviation) is plus or minus 0.2. This will tell you how much on average any one measurement might be in error.

Finally, I'll give you a rule of thumb which generally works. To reduce the standard deviation by a factor of N you generally need to average N*N results together. Thus if you have standard deviation of 4 for a single measurement, and you wanted to reduce it to 1, you'd have to average 16 measurements together. This may seem discouraging (100 experiments are required to reduce the standard deviation by a factor of 10!), but this is why I say that more repetitions are better.

If you didn't exactly follow all of this, just remember that more repetitions are better because they provide more certainty that your measurements are not just a fluke. That for the most part is what statistics is all about.

In fact, one more example that my colleague Frank Monaldo uses, may make this whole concept easier. Imagine I asked you to estimate the batting average of a major league baseball player. How many at bats would you need? You could estimate the average from just one at bat, but the answer would either be .000 or 1.000, both ridiculous values. Two at bats would provide a slightly better estimate, but you'd need 100 or more at bats to start to get a really good idea. There is a great deal of randomness in any one at bat, but the batting skill of a player can be estimated by averaging many at bats. And the more at bats you average, the better the estimate - at least up to point.

Major league batters are often ranked by their "lifetime" batting average. But no one really computes a lifetime batting average. To do that you'd have to include all the player's little league, high school, college, and minor league games, along with any old-timers games they played after they retired. Ofcourse averaging little league and major league batting averages would be like comparing "apples" and "oranges", which is exactly my point. You should only average together results from identical experiments.

For example, there were several projects at a recent fair studying soil erosion. Each student performed 10 experiments, and averaged the results together. The individual data showed a trend though because they reused the same soil each time - the soil was dry at the start of the first experiment, but increasingly wet at the start of each subsequent experiment. The student who received honorable mention recognized the problem and allowed 24 hours between the experiments, although even this was not enough time for the soil to completely dry. Any of these projects might have garnered a top 3 finish if they had simply recognized the trend in their data, admitted the problem, and decided to not include the first 2 or 3 experiments in the average, thus limiting the experiment to the study of erosion of already wet soil.

This article describes six simple ideas about how you can better prepare yourself for competing in a science fair. None of these suggestions should take you much time, but I guarantee the judges will be impressed. And best of all, all of these points work just as well in the real life world of science. Good luck.