A critical aspect to the scientific method is the hypothesis. A hypothesis must be falsifiable. This means there must be some way to demonstrate the hypothesis was incorrect. As a general rule, it is not possible to prove a hypothesis correct. A hypothesis can be demonstrated to be consistent with results, but further or more detailed testing may invalidate it. But if a hypothesis cannot be proven incorrect, then any test is inherently inconclusive. For the same reason that a hypothesis cannot be proven correct, a positive result from a test is not necessarily meaningful. However, if many test of many different aspects of a hypothesis all corroborate the hypothesis, then it may be considered true with high confidence.

Let us say that we learned that plants use sunlight for photosynthesis. We can assume that we've already made an observation and it was our research that turned up this claim. Let us take this claim, and form a hypothesis. Assume we will be performing the test in some home environment. Here are some examples:

• Photosynthesis converts CO₂ and sunlight into sugar.

• Photosynthesis doesn't require sunlight
• Photosynthesis cannot take place in the absence of sunlight

These are similar claims, but it is important that a hypothesis be within the scope of an experiment. We absolutely could perform an experiment to test the first one, however, that requires highly specialized equipment. It is also necessarily for a hypothesis to be falsifiable; compare the second and the third. The second one makes a claim that cannot be falsified without more specialized equipment; perhaps the photosynthetic process requires some chemical provided by other cells which require sunlight.

We will select the third hypothesis. However, it is important that we note our background assumptions (and preferably what research supports those assumptions). For the third claim to be falsifiable, we require the assumption that a plant which cannot perform photosynthesis can be differentiated from a plant that can. In this case, our research has indicated that for the plant species selected, it presents with green full foliage when it is healthy, and that it wilts and turns brown when it is unhealthy. We have assumed that the health of the plant is positively correlated to photosynthesis.

Now, for our test to be conclusive, we must have a control subject and an experimental subject. Ideal we will have groups, to reduce the probability that outside influences have caused the observed results. The control group will kept in conditions which our research has indicated will result in a healthy plant. To keep the test specific to the hypothesis, we must keep the conditions for the experimental group identical in all ways except exposure to sunlight. We also want to restrict human interaction with all aspects of the experiment.

Let us use an automated watering system perhaps a soaker hose. Fertilizer can be provided before the test, it will just limit our timeline a tad. We'll keep both groups in identical enclosures with identical ventilation systems. We could make a Plexiglas box with a hole cut for a computer case fan. The only difference will be that the control group's enclosure is transparent, and the experimental group's enclosure is opaque. We can paint the experimental group's enclosure with several coats of reflective paint (white should be good enough). We should place the experiment somewhere where the control group will get adequate sunlight, but the opacity of the experimental group will not result in a temperature difference. We must also ensure that the ventilation system for both groups is constructed in such a way that it minimizes light introduction. The fans can have a filter put on them, and maybe a pipe with a downward bend. It is important that the plants be left in the experiment for some time, one would not expect to see a difference overnight. Research will help determine the appropriate timeline for the species of plant involved.

We also want to reduce the possibility of human bias in the results. We can't really get a true double blind with this setup, but we can get close. We'll have to have at least two people involved in the process. Person A sets up the enclosures. Person B places the subjects in the enclosures, and at the end of the experiment removes them; Person A must be absent for both. The subjects must be identified in some code which has no meaning to their group, only Person B knows which plant was in which group. Person A will judge the health of the subjects; the location of the judging must be removed from the enclosures and Person B must be absent. Ideally different people will be used for each step, and the people judging the health will have no idea what the purpose of the experiment is. After judging, A's results can be compared with B's identifiers to see if there is a difference between health and the group the plant was in.