Many people seem to misunderstand what an experiment is, which is unfortunate as experimentation is the foundation of the scientific method. The sterotypical experiment involves a lab and test tubes, but experiments certainly aren't limited to the lab.
An experiment involves controlling one 'input' variable, holding all others constant (to the best of your ability) and measuring the effect on an output variable of a change in the control variable. The 'quality' of an experiment depends in part on how well you can control the other variables. By this standard, the lowest quality experiments are the 'natural experiments' frequently used in biology due to restrictions imposed by budget or ethics. For example, you cannot perform an experiment that involves killing people, but you can look for naturally arising situations that are similar except for the 'control' variable and see how it effects survival.
An experiment also has to test an hypothesis, theory or law, otherwise it is pointless. Or to put it more politely, it is merely an observation. You need to make a prediction based on your hypothesis. For the experiment to be valid, your observations need to be likely to be different from your prediction if your hypothesis is wrong. However, if your hypothesis predicts one thing, but can be 'reinterpreted' to accomodate any likely observation, your experiment is worthless.