In principle, one can conduct naturalistic observation nearly anywhere. There are some exceptions, however, for example, the naturalistic observation of human sexual behavior, or the observation of juries. You cannot observe a jury unless you happen to be one of the jurors. So, to study sex and juries, we need other methods.
Animals make natural subjects for naturalistic observation. A good example is Schaller's work with the mountain gorilla. When Schaller first started wondering about the behavior of gorillas, very little was known about them outside of their behavior in zoos. So, Schaller arranged to go to Africa to study gorillas. He soon discovered that gorillas were hard to find. Slowly, and with much effort, he began to see signs of their presence. He began to see where they had been eating and sleeping a few nights before. Then he was closer; their beds were hours old. One day, while sitting in a clearing, a group emerged all around him. From then on, he was able to watch them closely. He gradually discovered a great deal about gorilla behavior, including the fact they did not like cameras pointed at them. Other investigators followed Schaller after his pioneering work, including Diane Fossey. She was later killed by local gorilla poachers; her story is told in the 1988 movie, "Gorillas in the Mist."
Humans also make good subjects for naturalistic observation. The behavior of second graders makes a good example. The answer to questions about what second graders are like can be found by naturalistic observation. To observe them, a scientist would arrange to go to a second-grade classroom. Some schools are set up with one-way mirrors or cameras, so in those cases, observation would be fairly easy. Most schools are not set up that way, so the second graders would know that they were being observed. That would, quite naturally, change their behavior, at least for awhile. It might take a week or two of observation before the second graders settled down and reverted to their usual behavior. Then, naturalistic observation could start.
Naturalistic observation is probably the best method around for the process of formulating new hypotheses. By becoming intimately familiar with subjects and their behavior, one can start to make hypotheses about them and their behavior. Those hypotheses can then be tested by other methods. We will cover those other methods soon. So, the notion that naturalistic observation is a good first step in science derives from its utility in formulating hypotheses.
There are problems with the use of this method. The main one is the issue of bias. Researchers are always going to bring preconceived ideas to a naturalistic-observation episode. Then, unwittingly, those preconceptions will shape the observations themselves. For example, you may think that second-grade boys are more vocal than girls. So, you may pay more attention when you hear boys talking and shouting, but less when you hear girls doing the same. How can you deal with this problem?
One way to deal with the problem of bias is to have multiple observers. If you are the only one consistently to observe some behavior, then, maybe you are adding it to the situation, or, maybe the others just cannot see it. However, now you and the other observers could look for it more closely. Another way to deal with bias is to record the situation. Then, it can be analyzed repeatedly. Still another way is to have naive observers. You train observers, but you do not tell them why they are observing. Then, those observers will be less likely to see the situation in the light of the hypothesis.
Another problem is reactivity. Imagine what would happen right after you sat down in that second grade classroom. The students would not behave as they did before you entered. But, after a few days or weeks you would become part of the 'woodwork' of the classroom and the students would revert to their usual behavior. Your naturalistic observations could begin then. Note, that classrooms equipped with one-way mirrors or cameras would not have the problem of reactivity.
So, naturalistic observation is a good first step in research; it is good for formulating hypotheses, but care needs to be used to control for bias and reactivity.