Science projects by their nature are interdisciplinary, and this alone makes them a wonderful tool for learning. The application of previously unconnected skill sets in one project is valuable enough, but a greater power of experimental design in learning is that it can be as simple or advanced as the student is ready for. The experimental design for an advanced science project can expand the use of math, reading and writing, scientific concepts, and computer literacy in several predictable and modular ways, beginning with the research review.
The Research Review
With science projects for a younger audience, often the research review is limited to a few assigned readings on the topic. The science project becomes an elucidation of the major concepts in the topic. For an advanced project, the review is largely left to the students, with teachers acting as guides to appropriate sources of information. Whether the student is performing a case study, an ethnography, or an experiment, a thorough research review will inform the process and lead to better results.
A research review will help in most every area of the advanced science project. Reading the work of researchers on the topic chosen will inform the creation of a more relevant and insightful hypothesis. Also it is a good idea to be familiar with problems in the field of study. It may be hard to collect a proper control group, as is the case with much behavioral research. It may be that certain extraneous variables cause a lot of problems in some types of research, or measurement of certain variables may be prone to bias. Scientists working with these topics will have found ways to get around these problems.
A research review should be systematic. Catalog the databases you search or journals you look through. If internet searches are a way you locate articles, then include that. A list of keywords that are used will come in handy later in case there is difficulty finding information about the specific topic or an instructor has further keyword and database suggestions.
The goal for a research review should ultimately be about finding similar experiments and the existence of any theory for the topic. The experiment should be about testing current knowledge or expanding it. The only way to find out which is by reading what has been done to date. Getting this context for an hypothesis goes a long way toward making the argument for its relevance.
Advanced experimental design requires close attention to extraneous variables, especially in controlling their influence over the dependent variable(s). The hypothesis should be formed as a prediction. If X is done, then Y will happen. X is the independent variable, the action being taken by the researcher. Y is the dependent variable(s) being measured by the researcher. Extraneous variables are defined as alternative explanations for changes in Y. In other words, to determine the effects of X, the influence of extraneous variables must be removed or minimized.
Some types of experiments, especially lab-based, will allow the use of random sampling and a control group. For example, we could homogenize and sterilize soil, randomly choose 100 marigold seeds, and treat 50 seeds with a chemical abrasive to weaken the seed coat. After planting and observing the germination rates of each group, we could safely say whether the chemical abrasive increases or decreases the germination rate of marigold seeds.
Imagine the same experiment performed outdoors in the garden. The seeds are now exposed to a world of extraneous variables. If the treated seeds failed to sprout, we could blame it on the abrasive, but there is a chance that those seeds were washed away in rain, were eaten by animals, were kept from germinating through the presence of inhibitor chemicals (such as found in the black walnut), and any number of other possible explanations. Performing experiments in the real world requires close attention to extraneous variables, and this requires creativity. We could in this instance insure the control group does its job by attempting to sprout both groups in a damp paper towel.
A control group has to control for extraneous variables. The beauty of using a control group properly is that it will control all extraneous variables, whether you are aware of them or not. Control groups are not always possible. Imagine a new situation, two teachers predict that oral recitation of the multiplication table alongside multiplication games will help students learn better than recitation alone. Why will this be difficult to test?
It will be impossible to set up a proper control. Students cannot be randomly assigned to the two classrooms. Students already will have been assigned to classes, probably based on their skill levels. Let’s assume for the sake of argument that both classes are remedial, similar skill levels. What if both teachers are interested in finding evidence to support taking extra time to play the multiplication games? The one assigned to use the games (the teacher of the experimental group) will be more positive and enthusiastic about this than the teacher of the control class, who will likely feel that her students are not getting the best instruction. Teacher enthusiasm becomes an important extraneous variable that remains uncontrolled. Can you think of ways to control for teacher (and student) enthusiasm?
This experiment is further limited in a related way. Take another look at the hypothesis, specifically “will help students learn better.” An hypothesis has to be precise. This one is saying that, should the experiment turn out favorably for recitation plus games, ALL students would benefit from the use of both teaching tactics. How might we reword the hypothesis to better state the group results will generalize to?
Limits on Generalizing
All experiments have limitations and advanced experimental design requires we know and clearly state those limitations. For the marigold seed experiment, we have to consider the source of the seed. If it all came from one seed provider, then the results can only be generalized to marigold seeds from that provider. Using seeds from three of the top providers would allow us to generalize more freely to all marigold seeds.
In the case of the education experiment, results could generalize to remedial 7th grade math students at Forest Hills Middle School. We could change the experiment so that it generalized to a larger group, but this will require the participation of more math teachers and likely the principal. It will also require a matching, instead of a control, group. A matching group is one chosen for its similarity along one or more variable(s). If we wanted to find a matching group for our math class, we would look for a math class in the same grade, with a similarly experienced teacher, and a set of students that typically got the same grades and had the same socioeconomic status, racial, and gender makeup.
Now let’s consider a student-led experiment. All your math teachers have said, if you do all the assigned homework, then you will get better grades on tests. You think this teacher gives way too much homework, and you want to prove the hypothesis false. Enlisting the help of seven other students in the class who feel the same, you set out to do just that. Here are the important questions:
What is the hypothesis?
Can a control group be established? If not, can we create a matching group to rule out important extraneous variables?
What limitations exist in this experiment? How far do we expect results to generalize?
If you have further questions about the experimental process, write to me!